Está en la página 1de 329

The Philosophy ofPhilip Kitcher

The Philosophy
ofPhilip Kitcher

Editedby MarkCouch
Jessica Pfeifer

Oxford University Press is a department of the University of Oxford. It furthers
the Universitys objective of excellence in research, scholarship, and education
by publishing worldwide. Oxford is a registered trade mark of Oxford University
Press in the UK and certain other countries.

Published in the United States of America by Oxford UniversityPress

198 Madison Avenue, NewYork, NY 10016, United States of America.

Oxford University Press2016

All rights reserved. No part of this publication may be reproduced, storedin

a retrieval system, or transmitted, in any form or by any means, withoutthe
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by license, or under terms agreed with the appropriate reproduction
rights organization. Inquiries concerning reproduction outside the scopeofthe
above should be sent to the Rights Department, Oxford University Press,atthe

You must not circulate this work in any otherform

and you must impose this same condition on any acquirer.

Cataloging-in-Publication data is on file at the Library of Congress

ISBN 9780199381357

Printed by Sheridan, USA


Mark Couch and Jessica Pfeifer
1. Kitcher against the Platonists 14
Reply toRosen
2. Kitchers Two Design Stances 45
Karen Neander
Reply to Neander
3. Proximate and Ultimate Information in Biology 74
Paul E.Griffiths
Reply to Griffiths
4. Bringing Real Realism Back Home:APerspectival Slant 98
Michela Massimi
Reply to Massimi
5. Unificationism, Explanatory Internalism, and Autonomy 121
James Woodward
Reply to Woodward
6. Special-Science Autonomy and the Division of Labor 153
Michael Strevens
Reply to Strevens
7. Toward a Political Philosophy of Science 182
Reply toDupr
8. Kitcher on Science, Democracy, and Human Flourishing 206
Reply toDaston
9. Deliberating Policy:Where Morals and Methods Mix 229
Nancy Cartwright and Alexandre Marcellesi
Reply to Cartwright and Marcellesi
10. Function and Truth in Ethics 253
Reply toSmith
11. What to Do While Religions Evolve before Our Very Eyes 273
Daniel Dennett
Reply to Dennett



Nancy Cartwright is Professor of Philosophy at the University of Durham

and the University of California, SanDiego.
Mark Couch is Associate Professor of Philosophy at Seton Hall University.
Lorraine Daston is Director at the Max Planck Institute for the History of
Science, Berlin.
Daniel Dennett is University Professor of Philosophy at Tufts University.
John Dupr is Professor of Philosophy of Science at the University of
Paul E. Griffiths is Professor of Philosophy at the University of Sydney.
Philip Kitcher is John Dewey Professor of Philosophy at Columbia
Alexandre Marcellesi is a graduate student in philosophy at the University
of California, SanDiego.
Michela Massimi is Professor of Philosophy of Science at the University of
Karen Neander is Professor of Philosophy at Duke University.
Jessica Pfeifer is Associate Professor of Philosophy at the University of
Maryland, Baltimore County.
Gideon Rosen is Stuart Professor of Philosophy at Princeton University.
Michael Smith is McCosh Professor of Philosophy at Princeton University.
Michael Strevens is Professor of Philosophy at NewYork University.
James Woodward is Distinguished Professor of History and Philosophy of
Science at the University of Pittsburgh.


Philip Kitcher is one of the most influential philosophers in the contempo-

rary period. He is known for a number of important works he has written
during the course of his career. He has written over fourteen books and
160 papers and other publications. The scope of these works spans many
of the traditional areas of philosophy, including, among other things, phi-
losophy of science, philosophy of biology, philosophy of religion, ethics,
philosophy of mathematics, and epistemology. He is commonly read by
students and scholars who work in these areas and who need to be familiar
with the important positions he defends on these subjects. But it is also
the case that Kitcher has been widely read by the general public. His book
Abusing Science:The Case against Creationism (1982), for example, has had
many printings, and he has written articles for such venues as the New
Republic and the NewYork Times, among others. It is fair to say that Kitcher
has been a very influential figure both within and outside of the academy.

* For help with this volume we are grateful to a number of people, including Kyle
Stanford, Peter Godfrey-Smith, Laura Franklin-Hall, Jim Thomas, Silvena Milenkova,
Philip Kitcher, the contributors, Seton Hall University, University of Maryland,
Baltimore County, and discussions with several others. Aspecial thanks to our editors
at Oxford University Press, Peter Ohlin and Emily Sacharin, for their guidance dur-
ing this project, and Nisha Dayalan and Judith Hoover for their help in preparing the
manuscript. Philip Kitcher would also like to say thank you to the contributors for each
of their chapters.
This volume provides an examination of various areas of Kitchers phi-
losophy. In this introduction we provide some background for the reader by
describing a number of his major works and how his interests have devel-
oped over the years. Given the many works that have been written, there
wont be an attempt to cover everything.
A good place to begin is with the first book Kitcher published, which
was on the issue of scientific creationism. In the 1970s some of the cre-
ationists in the United States were becoming more open with their view
that the scientific evidence that existed did not undermine the creation
story in the Bible. This culminated in a number of books defending scien-
tific creationism. In 1982 Kitcher wrote Abusing Science:The Case against
Creationism in reply to the emergence of this view, arguing that the cre-
ationist authors were bending the science to support their positions. In
this work Kitcher carefully explains how to think about scientific evidence
and the content of evolutionary theory and other issues that were rel-
evant to the creationists approaches. He patiently describes their mis-
takes in trying to make the science appear to support their positions. He
explains that the problems with their approaches become apparent once
one is clear about what the science actually says and how to understand
it properly. Several of the ideas Kitcher touches on in this book he would
develop further at a latertime.
In other work during this period Kitcher turned his attention to issues in
the history and philosophy of mathematics (which was the area of his doc-
toral work). While the focus was different, this raised issues related to what
came before. In The Nature of Mathematical Knowledge (1983), Kitcher aimed
at understanding the historical development of mathematical knowledge.
His interest here is with understanding how mathematics developed as a
practice and how an appreciation of this informs our account of mathemat-
ics itself. He takes the view that mathematical knowledge is empirical in the
tradition of John Stuart Mill. We should understand mathematical practice
as depending on perceptual experiences in the origins of the subject and
developing from there through a sequence of rational transitions to more
complex parts of mathematics. Such an approach does not involve making
reference to abstract objects to explain mathematics. Characterized in this
way, mathematics takes its place alongside other empirical disciplines that
serve to improve our knowledge.
Around this time Kitcher also devoted several years to gaining a better
understanding of the sciences and their features. One area on which he
focused was the notion of scientific explanation. The proposal he offered
was an account of explanation that drew on earlier work in the area. The
distinctive feature of the account was to appeal to the notion of unification

[2] The Philosophy of Philip Kitcher

as a criterion for what counts as a good explanation. Ascientific explana-
tion, he writes in Explanatory Unification (Kitcher 1981), aims to unify
the phenomena of nature as a way of improving our understanding. The
account is intended to move beyond the traditional deductive-nomological
model of explanation that had been accepted earlier and provide an account
that better reflected the practice of scientists. This approach remains one of
the central models of explanation in the literature to the present day and
continues to be widely discussed.
Another interest he developed during this period was improving our
understanding of the biological sciences, which had been neglected by
previous generations of philosophers of science. Kitcher wrote a number
of papers focusing on issues in biology that have become well-known in
the field. These include 1953 and All That: A Tale of Two Sciences (1984),
concerning the attempted reduction of classical genetics to molecular biol-
ogy. He also did work on the notion of function (Function and Design
[1993c]), altruism (The Evolution of Human Altruism [1993b]), and how
to make sense of the levels at which natural selection operates (The Return
of the Gene [Sterelny and Kitcher 1988]). A focus on the different aspects
of biological theory and its implications would remain an area of interest
for many years.
This period of work raised issues related to our understanding of some
particular features of the sciences. It was during this time that Kitcher
also described his perspective on the sciences in broader terms. In 1989
he was instrumental in the founding of the Science Studies Program at the
University of California, San Diego that focused on philosophy, history,
and sociology of science. His major work that developed from his interest
in this area, The Advancement of Science (1993a), provides an account of sci-
entific knowledge in general. The immediate concern was to deal with the
problems that existed after the publication of influential work in the his-
tory and sociology of science by Thomas Kuhn and other critics that called
into question the traditional conception of scientific knowledge. The critics
claimed that people are often held captive by a simplistic image of science
as developing in an ideally rational and objective way. Kitcher agreed
that the traditional image is misleading and that the historical record
reveals the development of science is in fact a messy affair. However, he
suggested that the view that scientific development is merely the product
of sociological and historical forces resulted in its own form of simplifica-
tion. In contrast he argued that we can accept that science develops in a
broader, social context, but this should not lead us to think that we should
view the growth of science as irrational in the end. What is needed is a
more nuanced understanding of the practice of science and how it is able

Introduction [3]
to provide us with objective knowledge about the world. To explain this
Kitcher offers his readers improved accounts of such notions as explana-
tion and rationality as they occur in the practice of science. Once these are
understood, we are in a better position to understand the development of
scientific knowledge. We can recognize the critics concerns without having
to give up the view that science develops in a largely rational manner.
After this Kitcher continued to develop his work along these lines. While
he worked on other issues, he never strayed far from his interest in under-
standing the sciences in general. Several years had passed since the publica-
tion of The Advancement of Science before Kitchers next book on the subject
appeared, and in the intervening years he had come to think that some of
his earlier views needed to be revised. In a sequence of works he explains
what should be retained from this previous work and how it should be
modified to better reflect the practice of the sciences. In Science, Truth, and
Democracy (2001b), he makes a number of suggestions. Whereas previously
he had thought there was a single, overriding goal of the sciences, he came
to think this was misleading. The sciences develop in their individual ways,
and we should recognize a plurality of practical and epistemic aims that
exist since this view is more consistent with how the sciences have arisen
historically. In addition, Kitcher argues that more attention should be paid
to the issue of how the sciences fit into society, particularly its democratic
aspects. It is evident that science develops in relation to the broader society
of which it is a part. For this reason part of understanding science involves
understanding the role of science in society and how to think of the relation
between scientists interests in their research and the needs of the larger
society (e.g., think of how some scientists are interested in pure research
and how this may differ from citizens interests in solving particular medi-
cal conditions). The suggestions Kitcher made tried to be sensitive to the
interests of both and describe the proper role of science in society. The view
he presented he called well-ordered science.
The details of this view are further developed in Science in a Democratic
Society (2011b). Here Kitcher provides more specific accounts of what
a well-ordered science would look like under his conception, making his
suggestions more concrete and working out how they apply in the circum-
stances. There is a balance that needs to be struck between the expertise of
the scientist on factual questions relevant to public policy and the recogni-
tion that value-laden decisions about which policies to pursue should be
informed by input from the public. How this balance should be struck, and
the difficulties involved in it, are addressed with a number of proposals.
These works focus on understanding the role of science in society and
the value of scientific research. But the specific issue of value and its

[4] The Philosophy of Philip Kitcher

relation to ethics was something that Kitcher had yet to examine in detail.
This is the subject of his other work during this period where he turned to
the nature of ethics. In The Ethical Project (2011a), he offers an account of
moral evaluation. The account he develops views moral evaluation as an
ordinary form of human behavior like others. He argues that we should
conceive of ethics as arising out of our ancestors altruistic tendencies and
that these later developed into agreement about ethical rules that apply to
members of our group. Ethics is a type of social technology that enables the
members of society to navigate certain problems that arise in living with
others. Human societies work out ethics as they develop over time and
reflect upon the character of their previous practices and how these may be
changed. The result is that ethics is a normal part of human behavior and
shouldnt be seen as requiring religious backing or appeals to unusual fac-
ulties of ethical perception. Kitcher not only offered a distinctive approach
to explaining ethics as it has been understood by philosophers through the
years, but he provided the materials to address issues that he had yet to
examine in his other works, and so was able to round out his perspective.
In other work Kitcher continued to develop his previous views and took
on other areas of interest. One area he returned to was the subject of reli-
gion in his book Living with Darwin: Evolution, Design, and the Future of
Faith (2007b). He again discusses the question of religion in the context
of evolutionary theory and shows that intelligent design theorists have
misunderstood the scientific evidence. He considers the implications of
evolutionary theory for belief, as well as the role of religion in society.
In addition, he ventured into issues related to literature, analyzing the
work of James Joyce in Joyces Kaleidoscope: An Invitation to Finnegans
Wake (2007a) and Thomas Mann in Deaths in Venice:The Cases of Gustav
von Aschenbach (2013). The focus of these concerned a number of themes
regarding what gives meaning to our lives and the tensions that exist
between certain social and ethical values. These represent merely some
of the ways that Kitchers interests have developed as he has continued to
refine hisviews.
This review has tried to describe a number of areas that Kitcher has worked
on over the years. There are two points that become clear to anyone familiar
with reading his works we would like to emphasize. The first concerns the
interesting analysis Kitcher provides of the issues he discusses in his writ-
ings. Kitcher is known for writing with clarity and for developing his views
with careful attention to detail in working out the positions he considers. It
is evident to his readers that in doing this he frequently raises fundamen-
tal and deep questions about whatever philosophical issues he is consider-
ing. These are the kinds of works that open up new ways of thinking about

Introduction [5]
important problems and repay careful reading. The second point relates to
the breadth of Kitchers knowledge, which becomes apparent as soon as one
considers the several areas on which hes worked. The range of his contri-
butions to traditional areas of philosophy, as well as other areas of broader
interest, is seldom found today among scholars in any field. We would sug-
gest that anyone who believes philosophy has become narrow in scope should
see in Kitcher an example of someone who has avoided this sort of parochial


The above review has described some of the central areas of Kitchers
large body of work. While it would be impossible to cover all the areas in
which Kitcher has made a significant contribution, we have tried to cre-
ate a volume that represents the breadth of his research. The contributors
have been asked to raise critical issues about different aspects of this work,
and Kitcher has been given the opportunity to reply. The remainder of this
introduction provides a summary of the chapters.
Gideon Rosens essay, Kitcher against the Platonists, is a critique of
Kitchers anti-platonism in mathematics that focuses on Kitchers (2012b)
more recent work. In both that work and his earlier work (1983), Kitcher
attempts to avoid platonism by arguing that the truth of mathemati-
cal claims does not require the existence of mathematical objects of any
sort. Kitcher (1983) defends a non-face-value semantics for mathematical
claims, arguing that the subject matter of mathematics is actually a hypo-
thetical collecting activity of idealized agents. Later (2012b) he defends
the view that mathematics is a collection of games and that mathematical
claims are not descriptions and dont have a subject matter of any sort. He
provides a novel defense of this formalism by arguing against any view that
attributes a subject matter to mathematics, including platonism. In par-
ticular he argues that platonists have no good explanation of how symbol
manipulation could lead to the discovery of new abstract objects. There is a
gap between the basis for mathematical claims and the ultimate standard
of correctness in mathematics on the platonist view. In response, Rosen
argues that a version of platonismmoderate platonismcan answer
Kitchers charge. Unlike Benacerraf (1973), Kitchers critique of platonism
does not rely on a general constraint on knowledge. Rosen argues that
without such a general constraint Kitchers conclusion does not follow for
a moderate platonist. The moderate platonist can accept that there is a gap

[6] The Philosophy of Philip Kitcher

but argue that the gap is unproblematic. Rosen also discusses objections to
Kitchers formalism, most significantly that it fails to make sense of vari-
ous aspects of mathematical practice and especially applied mathematics.
Hence even if Kitchers argument against platonism were successful, his
own formalism would fare even worse in making sense of the practices
of mathematicians, which is precisely the sort of concern Kitcher raises
against the platonist.
Karen Neanders Kitchers Two Design Stances is a response to
Kitchers (1993c) account of functions and his criticisms of etiological
accounts. Kitcher is critical of etiological accounts on two grounds:first, on
such accounts the role that selection must play in explaining the presence
or maintenance of traits is too demanding; second, etiological accounts fail
to make sense of the practices of many biologists who appear unconcerned
with the selective history of a trait when ascribing functions. In response
to the latter, Neander argues that the etiological account can in fact make
sense of biologists focus on the current selective utility of a trait, so long
as we are careful to distinguish what individuates a traits function (its eti-
ology) from how biologists figure out what that function is, which often
involves looking at a traits current utility. In response to Kitchers first con-
cern, Neander argues that his criticisms rightly apply to an ultra-strong
etiological notion but that there is a middling-strong notion that is defen-
sible against his objections. Where the middling-strong account requires
only that selection played some role (relatively recently) in explaining the
presence or maintenance of the trait over some of the available alternatives
that actually existed, the ultra-strong etiological notion requires consider-
ation of whether the trait would have offered a selective advantage over
all possible alternatives. Kitchers view, in contrast, is even less demand-
ing. It attempts to combine the insights of both the etiological account and
Cumminss (1975) causal role account. Traits of an organism have a func-
tion because of the causal contribution they make to the whole organism in
light of the general constraint provided by selectionthat organisms need
to survive and reproduce. Neander is critical of Kitchers view for failing to
make sense of the very practices he is interested in capturing. In particular
his view leaves underspecified how to account for the possibility of mal-
function of token traits.
Paul Griffithss Proximate and Ultimate Information in Biology
focuses on Kitchers principle of causal democracy, which Kitcher (2003a)
argues is integral to an appropriate response to genetic determinism. In
contrast to the views Griffiths defended a decade ago, he here attempts to
use the concept of information to buttress Kitchers principle, motivated
by recent advances in our understanding of information that Griffiths and

Introduction [7]
others have been engaged in developing. He distinguishes between proxi-
mate and ultimate information and describes new accounts of each that can
be used to help characterize gene-environment interaction in a way that
respects Kitchers causal democracy principle. He argues that we can com-
bine the insights of interventionist accounts of causation with Shannons
information theory to develop an account of proximate information in
terms of causal specificity. Moreover by revising Sheas (2013) account of
ultimate, teleological information, Griffiths defends a notion of biologi-
cal teleology that can figure in proximate explanations of development.
This also allows him to show how the teleological notion of information
might be more closely aligned with the proximate account he discusses.
Both notions of information are consistent with Kitchers principle since
both leave open which causal factorsgenetic or environmentalmight
be carriers of information. Hence rather than being a barrier to under-
standing gene-environment interaction, Griffiths argues, biological infor-
mation might prove useful in vindicating Kitchers argument that a correct
response to genetic determinism requires patient, empirical study of the
relative importance of various causal factors in development in a way that
respects the principle of causal democracy.
Michela Massimis Bringing Real Realism Back Home: A Perspectival
Slant is an attempt to rescue Kitchers (2001a) Real Realism from an inad-
equacy she believes it faces by bringing it back to Kitchers earlier Kantian
roots. Massimi focuses on Kitchers response to Laudans (1981) histori-
cal argument against realism, and specifically his use of the distinction
between working posits and idle wheels of theories. While she considers
Kitchers argument one of the most persuasive replies to such challenges,
there are historical cases wherein Kitchers approach seems inadequate. She
diagnoses the problem as resulting from his stringent notion of success.
She distinguishes success from above (which might be Nagels view from
nowhere or the real realists view from now) and her own preferred success
from within. She argues that our current vantage point is not privileged; it
is just one perspective among many. Rather than assess past theories from
our current perspective, we should assess them using their own standards
of success, but from other subsequent or rival perspectives, which include
the richer information such perspectives have at hand. She maintains that
false claims could not satisfy such a criterion of success. Her perspectival-
ism thereby provides the Real Realist with an alternative route to defend-
ing realism without privileging our own perspective. Where Kitcher relies
on our own perspective to pick out those parts of theories that are deemed
true from our own perspective, her perspectival realism identifies claims
that we have reason to believe are true, since they are justifiably retained

[8] The Philosophy of Philip Kitcher

in the shift from the original perspective to the perspective(s) from which
they are assessed.
Unificationism, Explanatory Internalism, and Autonomy by Jim
Woodward is directed at Kitchers unificationist account of explanation, as
well as how unification relates to arguments about autonomy. Following
Morrison (2000), Woodward maintains that not all unification is associated
with explanation; moreover something more than unification is required.
In particular explanations must capture the external or ontic relation of
difference-making that Woodward (2003) defends elsewhere. Nevertheless
one significant lesson we can learn from Kitcher is that internalist or epis-
temic concerns are crucial to providing an adequate account of explanation.
It matters what we can calculate and measure, and our interests clarify why
we focus on the explananda we do. These internalist and externalist con-
cerns are not in fact in conflict with one another. Rather it is because of
the stable patterns in nature that we are able to construct tractable models
of phenomena that interest us. Woodward also distinguishes two types of
unificationist projects: EU1 involves explaining a large number of phenom-
ena in terms of a few factors and is often tied to successful reduction; EU2
involves showing that certain factors are irrelevant to some phenomena
and is often tied to showing that some upper-level phenomena are auton-
omous from lower-level micro details. EU1 notes a common cause of mul-
tiple phenomena, whereas with EU2 it is the commonality or universality
that is itself being explained. Moreover different levels of explanation are
typically not competing with one another since they often cite difference
makers that are relevant to different explananda. Hence Woodward argues
that the defense of the autonomy of special sciences does not depend
on one explanation being more unifying than another. The autonomy of
the special sciences can be established by showing that they successfully
describe stable difference-making relations for their intended explananda.
Michael Strevenss Special- Science Autonomy and the Division of
Labor takes aim at Kitchers arguments against reductionism. Strevens
argues that explanatory autonomy (and the division of cognitive labor that
goes along with it) is in fact compatible with explanatory reductionism. He
distinguishes two senses of explanatory relevance: objective and contex-
tual. Based on his kairetic account of explanation and his view on transitiv-
ity, Strevens claims that fundamental physics is always and everywhere
objectively explanatorily relevant. However, this does not entail that phys-
ics is always contextually relevant for explanations. Contextual relevance
is decided in part by us, based on how cognitive labor is divided. Strevens
discusses two ways that scientists divide labor among complementary
fields:functional compartmentalization and functional stratification. The

Introduction [9]
approach of compartmentalization involves plugging black boxes into
a systems inputs and outputs, while stratification involves black-boxing
lower-level phenomena and building a model of the system out of the black
boxes. Stratification makes clear how explanatory autonomy is compat-
ible with reductionism. While objective irrelevance can lead to functional
stratification, scientists often decide to black-box lower-level phenomena
that are objectively relevant as a way of efficiently dividing cognitive labor;
in such cases the lower-level phenomena are objectively relevant, but prac-
tical considerations about how to efficiently divide labor entail that they
are contextually irrelevant. Such practical considerations do not entail that
contextual irrelevance is merely pragmatic or observer relative but depends
on what Strevens calls functional difference-making. Hence the world
allows for functional stratification, which enables scientists to efficiently
divide cognitive labor. Explanatory autonomy is thereby preserved in a way
that is consistent with explanatory reductionism.
In Toward a Political Philosophy of Science, John Dupr directs his
attention to Kitchers notion of a well-ordered science. While he thinks
Kitchers goals are laudable, he is less sanguine about whether or to what
extent well-ordered science is achievable and skeptical that Kitchers pro-
posed methods for realizing it are the most fruitful. Dupr focuses on two
main issues:how we ought to decide which research to fund (or even allow
to be pursued) and how democratic decisions should be made about the
application of science to public policy. He argues that implicit in Kitchers
work is the idea that science, democracy, and ethics are all social technolo-
gies. What he considers especially enlightening in Kitchers work are the
ways that science and democracy can come into conflict, which Dupr
sees as especially problematic in the information age. However, Dupr is
skeptical that Kitchers proposed solutions to ill-ordered science are either
workable or helpful. He argues that it is unclear how Kitchers idealized
conversations can be harmonized with actual conversations. He also ques-
tions the relevance of such idealized conversations for addressing the dis-
cord between democracy and science, given that such discord is a problem
of social technology. In addition, while he thinks the citizen juries that
Kitcher recommends are perhaps successful in some cases, such juries are
often ill-suited to the task. Given the current social system we inhabit, more
systematic political changes are needed. Where Kitchers focus is primarily
on equality of voice, Dupr argues that well-ordered science is hampered by
the inequality of resources that our current social system promotes.
Lorraine Daston, in Kitcher on Science, Democracy, and Human
Flourishing, focuses on Kitchers attempt to reconcile science and democ-
racy and his use of history in defending his views. First, she questions

[10] The Philosophy of Philip Kitcher

whether his account of the history of modern science is correct and whether
its being wrong might affect his argument. She argues that the insistence
on the autonomy of science emerged in the twentieth century primarily
as a way to defend basic research during a period of increased public fund-
ing of science. This correction to the history, she contends, might require
rethinking his arguments. Her second focus is on whether his use of his-
tory is legitimate. Kitcher uses history to ground his claim that mutual
engagement and well-ordered science might be approximated and that sci-
ence and democracy would be better off if they were. The argument for the
former relies in part on what she calls speculative history about early
human social groups. While she is skeptical of such a history, her primary
concern is whether it can ground the kinds of claims on which Kitchers
arguments depend. In particular she questions whether it gives us reason
to believe that we are capable of enlarging our domain of mutual engage-
ment to all humans in the way Kitcher envisages and, more significantly,
whether we ought to do so. She distinguishes two ways we might make use
of history:to broaden our perspective on what is possible and to ground
normative claims. The former is legitimate, and in that regard Kitchers
work is highly successful. However, she questions the second, arguing that
history alone is not sufficient but requires a compelling vision of a better
way of life, which she thinks might be supplied by his notion of human
flourishing. She argues, though, that Kitchers requirements for mutual
engagement on a global scale might themselves undercut the possibility of
human flourishing.
In Deliberating Policy: Where Morals and Methods Mix, Nancy
Cartwright and Alexandre Marcellesi honor Kitcher by discussing a topic
about which he would no doubt care deeply. Kitcher has had a long-stand-
ing concern with the use of science in public policy. As they note, Kitcher
takes moral and social values to be intrinsic to the practice of the sci-
ences (Kitcher 2001b, 65), believing that great care should be taken in
determining how such values ought to enter scientific practice. Cartwright
and Marcellesis essay focuses on how morals and methods can mix in del-
eterious ways, especially when science is used to guide policy decisions in
the age of evidence-based policy. Policy decisions ought to be based on (1)
whether the policy will be effective and (2) whether it is morally, politi-
cally, socially, and culturally acceptable. Greater weight is often given to
(1) because it is believed that we have better methods for answering (1)
than for answering (2). However, we are overconfident in our judgments
about (1) based on mistaken ideas about objectivity, certainty, and cau-
sality. We bank on certainty, believe that objective methodssuch as
randomized controlled trials (RCTs)are the best path to such certainty,

Introduction [11]
and think that causality is linear and God-given. Causal relations are far
more complex, while the objective relations we discover through RCTs are
local, surface-level, and expressible only in language specific to the RCTs.
Instead of using other types of investigation that would be a better guide
to causal structure and hence a better guide for policy decisions, we over-
generalize from a few objective RCTs without adequately addressing the
moral ramifications of doing so.
In Function and Truth in Ethics, Michael Smith raises concerns about
Kitchers (2011a) account of ethical truth as developed in The Ethical Project.
Kitcher builds ethical truth out of ethical progress. Ethical rules count as
true if they are retained as ethical codes progress. Smith argues that this
account of moral truth leads to problems once we realize that progress is
to be understood in terms of promoting ongoing cooperation. On Kitchers
account there is a gap between the ethical rules we need to adopt in order
for ethical practice to serve its functionwhich Smith argues Kitcher must
understand as promoting ongoing cooperationand the moral beliefs
many of us (including Kitcher) hold. Hence Kitchers views about the func-
tion of ethical practice, together with his pragmatic naturalist account of
the truth of ethical claims, entail that many of our ethical beliefs are false.
Moreover ongoing cooperation is sometimes aided in crucial ways by the
fact that such false beliefs (beliefs that are false by Kitchers lights) are
widely shared. Fortunately we can accept Kitchers account of the function
of ethical practice without adopting his account of ethical truth. Smith
considers two alternatives he maintains are preferable:noncognitivist and
Kantian accounts of ethical truth. He defends both of these possibilities
against Kitchers objections. Either would also allow us to disambiguate the
causal question of why we have adopted the rules we have and the justifi-
catory question of what rules we ought to adopt. We can thereby accept
Kitchers account of the function of ethical practice, while leaving open
what function ethical practice ought to serve and what moral beliefs we
can legitimately assert aretrue.
Daniel Dennetts essay, What to Do While Religions Evolve before Our
Very Eyes, focuses on Kitchers (2011c) essay Militant Modern Atheism,
in which he argues that the New Atheists fail to account for the positive
role religion can play in peoples lives. Consequently their militant athe-
ism is likely to be counterproductive in the end. Kitcher argues that it
is possible to maintain a religious life even in the face of criticisms the
modern atheists have effectively wielded, and for at least some people it
is beneficial to do so. He distinguishes between the belief model of reli-
gion and the orientation model, arguing that the orientation model opens
up such possibilities and more adequately accounts for the aspects of

[12] The Philosophy of Philip Kitcher

religion that provide fulfillment in peoples lives. While noting that he and
Kitcher agree in most respects, especially in the ultimate goal they seek,
Dennett takes issue with Kitcher on the best strategy for achieving this
goal. This difference results from two fundamental disagreements: about
the benefits and costs of maintaining religion and about whether the ori-
entation model is ultimately sustainable. Dennett agrees that religion can
provide peoples lives with meaning, but he argues that the potential costs
of maintaining religion are too great: xenophobia, violence, and so on.
Moreover he sees the maintenance of religion in any form as a distraction
from and a means of delaying the work needed to replace religion with pref-
erable secular institutions. Dennett also argues that the nonsecular variet-
ies of the orientation model are unsustainable. Once the orientation model
is made explicit, those with religious commitments either lapse into the
belief model or engage in faith fibbing. He sees these problems as particu-
larly acute in the modern age of informational transparency.1

1. Abibliography of Kitchers works can be found in the useful volume by Wenceslao

J.Gonzalez (2011).

Introduction [13]

Kitcher againstthe Platonists



Mathematics is replete with results that affirm (or seem to affirm) the exis-
tence of mathematical objects. For example:

(1) There are at least three prime numbers greater than15

(2) The equation x3 + 1=0 has three complex solutions
(3) There exists a finite simple group of order 246 320 59 76 112 133 17
19 23 29 31 41 47 5971

These results furnish the basis for what appears to be a straightforward

argument for platonism: the view that mathematics is concerned, inter
alia, with a domain of abstract entities. Take (1):Unless you are willing to
reject grade school arithmetic, you must agree that there are prime numbers
greater than 15, from which it follows that there are numbers. But it is obvi-
ous on reflection that numbers are not physical objects. It is just silly to ask
where the number 17 is located, or how much it weighs, or how fast its mov-
ing. And it is likewise silly to think that the number 17 might be a nonphysi-
cal mental entity, like a Cartesian mind or an afterimage. So the only thing
to think about the number 17and the complex solutions to x3 + 1=0, and
the monster groupis, first, that these things exist and, second, that they
are neither physical nor mental and that they are therefore abstract objects.1

1. There are several senses in which an object may be said to be abstract. This
usage follows a tradition deriving from Frege ([1918] 1984), but the objects of pure
There are exactly three ways to resist this argument. You can take the
eccentric view that numbers and the rest are (despite appearances) con-
crete entities (Forrest and Armstrong 1987). You can step back from ordi-
nary mathematics and hold that while the existence theorems may be good
mathematics, they are not true and so cannot serve as premises in a sound
argument (Field 1980). Or you can holdand this is trickierthat while
the existence theorems are true and so fit to serve as premises, their truth
does not require the existence of mathematical objects of anysort.
This last position is tricky for obvious reasons. It is a plain contradiction

There are prime numbers greater than 15, but there are no numbers.2

And it sounds almost as bad tosay:

There are numbers, but there are no abstract objects.

The anti-platonist who takes this route is therefore in a tight cornera

corner so tight that it is unclear whether his view makes any sense at all.
Does he agree that that there are prime numbers greater than 15? He must,
if he accepts basic mathematics. Does he believe that these numbers are
concrete entities? Surely not, if he is sensible. So how can he deny that
numbers are abstract entities without simultaneously denying their exis-
tence and so contradicting the ordinary mathematical assertion with which
he began?3


Philip Kitchers views in the philosophy of mathematics have evolved,

but early and late he has always been an anti-platonist of this third sort.

mathematics are presumably abstract in every sense if they exist at all. See Rosen
(2014) for discussion of the terminologicalpoint.
2. Compare the closing sentence of Benacerraf (1965, 73):If truth be known, there
are no such things as numbers; which is not to say that there are not at least two prime
numbers between 15 and20.
3. One important feature of this argument for present purposes is that it is not a
semantic argument. It does not assume a Tarskian account of mathematical truth,
or any other such determinate account. The argument uses, but does not mention,
mathematical vocabulary. It thus puts pressure on any theorist who is happy to use
mathematical vocabulary in the usual ways, regardless of his or her semantic views. For
a more complete statement of the argument, see Rosen and Burgess (2005).

K i t c h e r a g a i n s t t h e P l at o n i s t s [15]
Kitcher has never been tempted to identify the objects of mathematics
with concrete things or to dismiss ordinary mathematics as a false but use-
ful fiction. His view has always been that settled mathematics is just fine as
it is, but that its claims, properly understood, do not concern a domain of
mathematical objects.
Kitchers 1983 book, The Nature of Mathematical Knowledge, defends a
version of anti-platonism according to which the subject matter of math-
ematics is not a domain of entities but rather the hypothetical collecting
activity of an idealized human agent. Like many anti-platonist strategies
from the period, this one works by constructing a nonface value seman-
tics for (part of) the language of mathematics. Such semantic theories yield
a mapping S S* from ordinary mathematical claims like There is a prime
number greater than 15 to claims in a modal language that do not seem to
require the actual existence of abstract entities:roughly, claims of the form
If the concrete world had been thus and so, then such and such would have been
the case. The mapping is designed to associate truths (falsehoods) in the
mathematical language with truths (falsehoods) in the modal language in
such a way as to preserve intuitive entailment relations among claims. But
more than this:the mapping is supposed to give the meaning of the original
mathematical claim and so to show it to be the sort of claim that does not
require the existence of mathematical objects for itstruth.
Since Kitcher has abandoned this approach Iwill not dwell on his par-
ticular version of it. But it is worth asking how views of this sort respond
to the quick argument for platonism sketched in the previous section. The
reductive nominalist cannot deny that there are prime numbers greater
than 15, since his view will map this mathematical claim onto a modal
claim he accepts by means of a semantic mapping that is designed to pre-
serve truth value. Nor can he deny that there are numbers, since the lat-
ter claim is a logical consequence of the first and his mapping is designed
to preserve logical relations.4 Instead he must sayand this is the tricky
bitthat while there are indeed infinitely many numbers of various sorts,
it is a kind of nonsense to ask whether these numbers are abstract or con-
crete, whether they exist in space, and so on. The paraphrase procedure that
gives meaning to statements in the language of mathematics associates
each ordinary mathematical statement with a definite (modal) content; but

4. These points are emphasized in Alston (1958). The neglect of this paper in the
literature on mathematical platonism is striking, especially in view of the fact that
Alstons paper was reprinted in the field-defining collection Benacerraf and Putnam
(1964). Astriking exception is Wright (1983), the first important work to emphasize
the significance of Alstons point for the metaphysics of mathematics.

[16] The Philosophy of Philip Kitcher

that procedure does not extend to mixed claims like The number 15 exists
(does not exist) in space or The monster group weighs more than a kilo-
gram. Statements that mix the languages of mathematics and physics or
metaphysics in this way are not false, as the quick argument for platonism
sketched above assumes, but rather meaningless on the model of Its five
oclock on the sun (Wittgenstein 1953, 350). The plausible reasoning that
leads us from ordinary existence claims in mathematics to the metaphysi-
cal claims characteristic of platonism thus involves a subtle lurch from
sense to nonsense. We can all agree that the existence theorems are true
and that certain mixed statements like The number 17 exists in space are
to be rejected. The mistake is to assume that because this statement is to
be rejected, its negation must be true. The right response, licensed by the
semantics, is rather to insist that the mixed statement and its negation are
both to be rejected. And so the case for platonism is blocked.5
This subtle position is worth exploring, but the main point against it
should be clear. It is always awkward to rest a philosophical argument on
claims to the effect that certain apparently meaningful statements are
really meaningless, especially when clear-headed speakers of English rou-
tinely affirm the claims in question. If we had a credible theory of meaning-
fulness, that would be one thing. But we dont. And in the absence of such a
theory, the fact that many competent speakers find a claim fully meaning-
ful and indeed true is powerful evidence that it is ultimately meaningful
and apt for truth. But let us put this issue to one side for now, since, as
Isay, Kitcher no longer endorses this sort of reductive nominalism.
Kitchers (2012b) latest work in the philosophy of mathematics defends
a more radical anti-platonist position: the formalist view that mathemat-
ical statements do not have a subject matter of any sort. Platonists and
reductive nominalists agree that mathematics is a descriptive science, in
the sense that it normally makes sense to ask whether things are as any
given mathematical claim says they are. The platonist takes mathematics
to be in the business of describing abstract entities; the modal nominal-
ist assigns it a modal subject matternot a domain of objects, but still
a putative domain of facts. For the formalist, by contrast, the so-called
statements of mathematics are not representations or descriptions of
any sort. Despite their superficial resemblance to meaningful declara-
tive statements in other areas, they are more like configurations of chess
pieces. It makes sense to ask whether a configuration of chess pieces is

5. To my knowledge the reductive nominalists have not made this point explicitly,
but Ibelieve it is the only way for them to evade Alstonspoint.

K i t c h e r a g a i n s t t h e P l at o n i s t s [17]
derivable in chess, that is, whether it can arise through play accord-
ing to the rules. But it makes no sense to ask whether a configuration of
chess pieces correctly represents its subject matter. According to the view
Kitcher now holds, mathematics is a collection of games for transform-
ing strings of nonrepresentational squiggles. These games are governed
by rules implicit in our practices and sometimes known explicitly to
mathematicians. We can make combinatorial, metamathematical state-
ments about these strings and the rules that govern them. These state-
ments constitute what Frege ([1903] 2013, 93) calls the theory of the
game, and for all Kitcher says, they may be genuinely representational.
But There are prime numbers greater than 15 is not a metamathemati-
cal statement about the game. It is a configuration of pieces within the
game. The rules governing the manipulation of these pieces may give the
string a kind of meaning. But the string and its parts do not stand for
anything, so it makes no sense to ask what the objects it describes are
like, or whether it describes them correctly.
How does this position block the quick argument for platonism
sketched at the outset? Kitcher does not say, but the answer must be
this: Whereas a string like 17 is a prime number may be derivable in
the game of arithmetic and hence assertible in a sense, metaphysical
statements like The number 17 is not in space or The number 17 is an
abstract object arelike

is green.

This is neither a configuration of chess pieces nor an English sentence,

but a monster. And the same must go for statements that mix the lan-
guage of mathematics and the language of metaphysics. We could try to
extend that mathematics game so as to include rules governing the trans-
formation of such strings, but that would invite confusion and serve no
purpose. So, like the reductive nominalist, the formalist blocks the case

[18] The Philosophy of Philip Kitcher

for platonism by insisting that certain mixed statements involving math-
ematical vocabulary and vocabulary drawn from other areas amount to
Kitchers innovation in this recent work is not the formalist view itself,
versions of which have been known for many years (Detlefsen 2005), but
rather the argument for it. Kitchers paper develops a new critique of pla-
tonism that emphasizes the role of free postulation in the history of
mathematics. As Kitcher notes, if this argument is successful, it threatens
any view according to which mathematics has a subject matter, including
reductive nominalism of the sort Kitcher himself once endorsed, leaving
formalism or something like it as the only view left standing. My aim in
this paper is to restate this argument and to say how the platonist might


Before we turn to the argument, Ishould say a word about its official tar-
get. As is well known, the view we have called platonism comes in two
flavors (Chihara 1973). Both hold that mathematics is concerned with a
domain of immaterial abstract objects. The hardcore platonists distinctive
claim is that these objects play something like a causal role in mathemati-
cal practice:that mathematicians are somehow aware of them or sensitive
to them, and hence that our mathematical beliefs are sometimes shaped
by the objects they represent. The moderate platonist denies this, insist-
ing that abstract objects do not impinge on us in any way. To put the con-
trast dramatically, the hardcore platonist holds that if the numbers had not
existed (per impossible, but so what?), the history of mathematics would
have been quite different, whereas the moderate holds that it might have
unfolded just as itdid.6
I mention this familiar contrast because Kitcher often writes as if hard-
core platonism were the only form of platonism on offer. Thus after an ele-
gant review of the history that led to the acceptance of imaginary numbers,

6. In the recent history of the philosophy of mathematics, hardcore platonism is

standardly imputed to Gdel (probably incorrectly) on the strength of a famous sen-
tence:We do have something like a perception also of the objects of set theory, as
is seen from the fact that the axioms force themselves on us as being true (Gdel
1964, 271). Moderate platonism has been defended by many authors, most notably
Quine (1961).

K i t c h e r a g a i n s t t h e P l at o n i s t s [19]
a history in which certain early figures (Cardano and Bombelli) stumbled
on the complex roots of cubic equations only to dismiss the new numbers
as subtile and useless, Kitcher (2012b, 182)writes:

We have, I claim, a satisfactory historical explanation of what occurred, even

though it never reveals a change in the relations between any mathematician and
any abstract objects. Moreover, dragging in the world of abstract objects would
be quite mystifying. It would generate the puzzle of why Cardano and Bombelli
hesitate because the new numbers are subtile and uselessisnt the impor-
tant issue whether they are part of the abstract realm, and if so, why dont they
sneak a Gdelianpeek?

More generally Kitcher supposes that any platonist must hold that when
the domain of mathematics is extended, new objects are discovered
or worse, detectedin roughly the sense in which Mendel discovered/
detected genes. Thus after entertaining the view that the sort of symbol
manipulation that led to complex arithmetic simply counts as a way of
detecting new abstract objects, Kitcher responds:

In none of these instances do we have any serious account of how the symbolic
manipulations serve as a way of detecting the alleged abstract entities. In the
Mendelian case, its possible to provide a positive causal explanation for why
the detection via pea plants works. Mendel himself saw part of this, and thats
why he could take his observations of the pea plants to be ways of detecting
underlying factors. He could justifiably use his instrument because he had
an account of how the phenomena he was trying to detect were related to the
properties he was able to observe.... Imagine properly educated counterparts
of Bombelli, Euler, Hamilton and Lagrange who fully subscribe to the Platonic
wisdom. Like Mendel, they would surely reflect on how their instruments, in
this case their symbolic practices, enable them to detect the underlying entities,
and platonic wisdom would supply them with no answer. Thus if they had what
is supposed to be the correct philosophical view of the matter, they would not
have been able to proceed as they did. (18485)

I will return to this passage later. The point to emphasize for now is that this
talk of detection makes sense only if the view under discussion is hard-
core platonism. Moderates deny that mathematical objects are detected, on
the ground that detection is a causal process, and so reject the demand to
say how the complex numbers were detected.
Of course the moderate platonist does believe in mathematical discovery
in a bland sense. To discover a fact in the bland sense is simply to come to

[20] The Philosophy of Philip Kitcher

know it for the first time. To discover an object is to come to know for the
first time that it exists (or that it has the features that make it notewor-
thy). Anyone who believes in the growth of mathematical knowledgeas
the moderate does, if he is not a skepticbelieves that mathematicians
make discoveries in this sense all the time. Euler discovered that ei + 1=0;
Cantor discovered the alephs. The moderate will insist, however, that such
discovery does not involve detection and is therefore quite unlike Mendels
discovery of genes or any discovery that involves interaction with the
newly discovered objects.
Kitchers deployment of the rhetoric of detection can give the impres-
sion that he is focused on hardcore platonism, and that would be unfor-
tunate, since the hardcore platonist is a straw man in a contemporary
context. With the possible exception of Roger Penrose, there are no promi-
nent living representatives of the species, and despite one oft quoted line
from Gdel, the view has not figured prominently in the recent history of
the philosophy of mathematics, where the important platonists after Frege
have all been moderates.7 And this is not surprising. Hardcore platonism is
a wild view, flatly inconsistent with a naturalistic view of inquiry, whereas
moderate platonism is (as we have seen) little more than a restatement of
common sense about mathematics. Any serious critique of platonism must
therefore engage the moderate version of theview.
A careful reading of Kitchers paper yields an argument that can be
stated without saddling the platonist with talk of detection and which
would therefore indict any form of platonism if it were successful. In the
next section Ipresent a formulation of that argument designed to high-
light its ecumenism.


The background for the argument is Kitchers (standard textbook) account

of the development of complex arithmetic beginning in the sixteenth cen-
tury. In Kitchers version of this history, Bombelli and other early innova-
tors began to calculate with symbols involving 1 and to achieve results
(e.g., about the roots of cubic equations) that were independently verifi-
able. Mathematicians were initially queasy about these techniques, a quea-
siness Kitcher interprets as uncertainty about the rules governing the use

7. Notably Quine 1961; Putnam 1971; Wright 1983; Burgess 1983. Burgess and Rosen
(1997) is a full dress defense of moderate platonism (there called antinominalism).

K i t c h e r a g a i n s t t h e P l at o n i s t s [21]
of these new symbols and about the ultimate fruitfulness of their introduc-
tion. These doubts were fully allayed by the beginning of the nineteenth
century as the fruitfulness of the new apparatus was placed beyond doubt
and the now standard interpretation of complex arithmetical operations
as operations on points in the Argand plane guaranteed the consistency
(relative to established analytic geometry) of the game involving them.
Kitchers key claim is that this standard history is satisfactory, not just in
the sense that it tells us who did what and why but in the further sense that
it vindicates the introduction of complex analysis, showing it to have been
warranted by every pertinent epistemological or methodological standard.
With this paradigm of a satisfactory historical account in place, Kitcher
(2012b, 17273) presents four claims about mathematical discovery that
any platonist should be tempted to accept:

(A) Innovations in mathematics often consist in the discovery of new

abstract objects.
(B) Satisfactory explanations of the discovery of a new entity X must offer
an account of X and the ways in which it became accessible.
(C) Historical explanations of cases of mathematical discovery never offer
any account of abstract objects and of ways in which such objects
become accessible.
(D) Some historical explanations of mathematical discovery are eminently

These claims are obviously inconsistent, so the platonist must reject one
of them. Kitchers argument is that unlike the formalist, who can happily
reject (A), the platonist has no good options.
Moderate platonists should accept (A), provided discovery is under-
stood in the bland sense given earlier. Mathematical innovation may not
involve the detection of new objects. But if to discover a thing is just to
come to know for the first time that it exists (or that it has the features
that make it interesting), then of course mathematicians discover new
abstract objects from time to time. Needless to say, the discovery will often
be spread out over time and over many people. (Asking Who discovered
zero? is like asking Who discovered the stone ax?) But if there was a time
when no one knew anything about X and a later time at which important
facts about X are clearly known, then we can infer that X was discovered
along theway.
(B) is likewise unexceptionable, properly understood. The metaphor of
access is of course misleading, since access to a thing normally involves
spatial or causal contact, and the moderate platonist will deny that we have

[22] The Philosophy of Philip Kitcher

(or need) access to abstract objects in that sense. But there is a legitimate
demand in the vicinity of (B), namely(B*):

(B*) Satisfactory explanations of the discovery of a new entity X must

offer an account of how people first came to know that X exists.

When the claim is put like this, the platonist must accept it. But he will
instantly add that it is somewhat unclear what it means to offer an account
of how people first came to know that X exists. Any complete account will
have two aspects:there will be an account of how people first came to believe
that X existsan historical account of the various psychological and social
processes that led to the formation of this beliefand then there will be
an epistemological gloss on this account designed to show that the belief in
question, formed in this way, was not a mere opinion but a case of knowl-
edge (or reasonable belief). Now in practice any account of this sort will
leave a great deal unsaid. The textbook account of Le Verriers discovery of
Neptune is in one sense satisfactory as it stands. It lays out Le Verriers evi-
dence in such a way as to make it clear that one might come to know about
the existence of a planet on such a basis. But it does not include a worked-
out theory of evidence, or of inference to the best explanation, or what have
you. Ahistorical account of this sort will make it clear that Le Verriers evi-
dence was in fact good evidence. But it need not answer the philosophical
question Why is the evidence that led Le Verrier to posit Neptune good
evidence for the existence of a thing like Neptune? We could insist that a
satisfactory account of an episode of discovery always include a philosophi-
cal account of why the discovery counts as a discovery. But if we do, then
Kitchers (D)will be clearly false. We should therefore understand a satis-
factory account of a discovery as one that presents the grounds on which
the new existential belief was formed in such a way as to make it clear to
the sensitive reader that these grounds were in fact good groundsideally
grounds on the basis of which the proposition in question might have been
known. And if we understand the word in this way, (B*) is unassailable.
Passing over (C)for a moment, we can agree that given this last point,
(D)is likewise unassailable. Standard historical accounts of mathematical
discovery tell us in more or less detail how mathematicians came to know
the existence theorems, which is to say that they tells us how mathemati-
cians came to believe these claims in such a way as to make it clear that these
beliefs were reasonable. Suppose a mathematician discovers a new solution
to the field equations of General Relativity by writing down an expression
for the solution and then proving that it is in fact a solution. The historian
can tell us in more or less detail how this happened, reconstructing both

K i t c h e r a g a i n s t t h e P l at o n i s t s [23]
the context of discovery and the proof itself, filling in the gaps, and so on.
But qua historian she will not broach the question Why is mathematical
proof of this sort a source of knowledge? That doesnt mean that the ques-
tion does not arise. The point is simply that ordinary historical accounts
presuppose that proof is a source of knowledge, leaving it for philosophers
and foundationally minded mathematicians to explore this presupposi-
tion. Asatisfactory historical account of discovery need not include a com-
plete philosophy of mathematics, though of course it must be consistent
with some such account if its explanations are to be correct.
Now Kitchers examplethe discovery of the complex numbersis
much more interesting, since this discovery was not a matter of proving a
theorem in an existing framework but rather of devising a new framework
(or extending an old one) with new rules of proof and calculation. The pla-
tonist must indeed regard this as a discovery in the undemanding sense. In
1500 no one knew the key existence theorems of complex arithmetic. By
1850 they were common knowledge. So over the course of this period the
relevant objects must have been discovered. Kitchers historical account
lays out the key mathematical developments in such a way as to make it
clear to the sensitive reader that mathematicians were reasonable in pro-
ceeding as they did. The account is therefore satisfactory in the sense
outlined earlier. Of course it is also incomplete. Not only does it leave out
details in the historical sequence, as is inevitable; it does not even begin to
address the epistemological question Why was it reasonable for mathema-
ticians to proceed as they did? This is a philosophical question, and Kitcher
himself clearly believes that it needs an answer that the historical account
itself cannot supply. After all, Kitchers formalist philosophy of mathemat-
ics is designed to supplement the historical account at just this point. So
Kitcher cannot object if the platonist must also provide a supplement:an
account of why the historical processes that led mathematicians to believe,
for example, that every polynomial with complex coefficients has a com-
plex solution counts as a way of coming to know this proposition.
The platonist must therefore reject (C), as Kitcher anticipates. Of course
if we rewrite (B)as (B*), we should rewrite (C)as(C*):

(C*) Historical explanations of cases of mathematical discovery never

offer any account of how people first came to know existential
truths about abstract objects.

But so rewritten, (C*) is easy enough to reject. When a mathematician

proves an existence theorem, she proves the existence of what isaccording
to the platonistan abstract object. Ahistorical account may explain how

[24] The Philosophy of Philip Kitcher

she came up with the proof, and it may set out the proof in such a way as
to make it clear that it is a proof. In normal contexts this will amount to an
account of how the mathematician came to know the existence theorem in
question. Of course the historical account will not normally include a dis-
cussion of why the proof is a proof or of why proof is a source of knowledge.
But if silence on this point disqualifies the historical account as an account
of how the existential truth is known, then historical accounts of discovery
in the sciences are hardly ever satisfactory, since they are almost always
similarly silent on epistemological questions.
What about the case Kitcher focuses on? Does the standard historical
account of the development of complex arithmetic explain how mathema-
ticians came to know that certain abstract entities exist? Well, it certainly
explains how mathematicians came to accept existence claims that, accord-
ing to the platonist, affirm the existence of abstract objects. Moreover it
explains how they came to accept these claims in such a way as to show the
mathematically sensitive reader that it was reasonable for mathematicians
to accept them as they did. So why cant the platonist say, as seems true,
that this historical narrative offers a perfectly good account of how people
came to know about these abstract entities, or, in other words, a perfectly
good account of how the complex numbers were discovered?
Kitchers (2012b, 18385) discussion of this gambit challenges the pla-
tonist to say when in this process the complex numbers were first detected.
But as Ihave emphasized, detection is a red herring. The question for the
platonist is whether at some point in this process mathematicians came to
know, for example, that every polynomial has a complex root and whether
the historical account shows how they came to know this. In my view
the platonist is thoroughly within his rights to answer yes to both ques-
tions:yes to the first question because the fact was unknown in 1500 and
well known by 1850, and yes to the second question because the mathe-
matically sensitive reader can learn a great deal from the historical account
about how this fact came to be known over this period.


Of course there is still a perfectly good philosophical question for the pla-
tonist:Why is that sort of history a way of coming to know (or reasonably
believe) a proposition about abstract objects? This is analogous to the ques-
tion one might ask after absorbing the standard historical account of the
discovery of Neptune:Why does that evidence count as evidence for the
existence of a distant planet? It is no knock against the historical accounts

K i t c h e r a g a i n s t t h e P l at o n i s t s [25]
that they do not answer this question. But of course it must have an answer
if the platonists rejection of (C*) is to be tenable.
The dialectic here is subtle, so let me clarify the point. Suppose the pla-
tonist concedes that at this point he does not have a satisfactory philosoph-
ical account of why the considerations that led mathematicians to accept
the existence of complex numbers count as a route to knowledge of abstract
objects. To concede this is simply to concede that he has not solved every
philosophical problem that his view raises, and that by itself is no objection
to his view. The platonist may be within his rights to say, in a Moorean vein,
Mathematicians certainly do know that x3 + 1=0 has three complex solu-
tions. And since the only thing to think about these solutions is that they
are abstract entities, their grounds for accepting this claim must amount
to a way of knowing about abstract objects. Since Ilack a general theory of
knowledge, Icannot explain why these grounds suffice for knowledge. But
they clearly do, and that is enough to answer Kitchers challenge. On the
other hand, if it could somehow be shown that there can be no adequate
platonist answer to this question, then the platonists Moorean response
would fail and Kitchers challenge wouldstand.
This is of course the standard gambit on the anti-platonist horn of Paul
Benacerrafs (1973) famous dilemma for theories of mathematical truth.
But to sustain this argument it is not enough to ask the platonist for his
epistemology. One must wield an epistemology of ones own to show
that if the subject matter of mathematics is as the platonist takes it to
be, then the usual ways of coming to believe a mathematical proposition
could never amount to ways of knowing. This is how Benacerraf himself
proceeds, invoking the causal theory of knowledge, and this is how Hartry
Field (1989) proceeds, invoking an explanatory constraint on knowledge.
Kitchers argument resonates with Benacerrafs, as Kitcher notes, but it is
distinct on precisely this point. Kitcher does not wield a general constraint
on knowledge against the platonist. If my analysis of the argument is cor-
rect, however, some such constraint is needed at just thispoint.
Moderate platonists will differ in their positive accounts of why the
grounds cited by the historian (and the mathematician) constitute adequate
grounds for believing the existence theorems, understood as claims about
abstract entities, but let me sketch the response Iprefer. The key thought is
that the ordinary internal justification for a mathematical claimwhether
it amounts to proof in an existing system or the sort of informal argu-
ment that justifies the acceptance of a novel axiomalways amounts to
adequate justification for the claim in question, the sort of justification
that suffices for knowledge when all goes well. To ask why the mathemati-
cal case for p is a good case for p is like asking why the inductive case for

[26] The Philosophy of Philip Kitcher

a generalization in the natural sciences is good. At any given stage in the
history of science, our best guide as epistemologists to the normative prin-
ciples governing sound reasoning on scientific subjects is the practice of
actual scientists at their best. Scientists accept standards for determining
when a claim is supported by the evidence, and while any given standard
of this sort can be questioned while holding others fixed, there is no higher
standpoint from which to criticize the most fundamental principles of this
sort. This is a version of Quines repudiation of first philosophy as modi-
fied and applied to the mathematical case by Penelope Maddy (1997). The
only difference is that Iwould be emphatic where Maddy is neutral (Rosen
1999). If a statement S is fully acceptable by the most stringent norms that
operate in mathematical or scientific practice, then unless those norms
are criticizable on scientific or mathematical grounds, we are justified not
just in accepting S for certain purposes but in believing that S is true.
Anyone who doubts this commits the error of the skeptic. She assumes
that a claim may be credible by the best, most refined scientific standards
and yet still not credible, as if there were some more authoritative standard
to which the claims of science are answerable. Since Im no skeptic, Ithink
that the mathematical case for complex numbers must at some point have
been good enough. But to say this is just to say that when mathematicians
accepted complex arithmetic on the strength of what mathematics itself
deems good grounds, they thereby came to accept these claims on what
are in fact good grounds, and hence to know them (if they are true and
there is nothing funny about thecase).
Kitcher (2012b, 184)is aware of this gambit and rejects it on the ground
that in none of these instances do we have any serious account of how
the symbolic manipulations serve as ways of detecting the alleged abstract
entities. Iclaim that we do have such an account when we conjoin the story
about the mathematical basis for accepting complex arithmetic with the
philosophical thesis that to accept an existence claim for the reasons that
support it within mathematics is, when all goes well, to acquire the knowl-
edge that that claim is true, and so to discoverif not detectthe abstract
objects with which that claim is concerned. If this conflicts with some nave
thought, sourced from elsewhere, that knowledge always involves detec-
tion or quasi-causal access to the subject matter of ones knowledge, then
so much the worse for that nave thought.
A passage already cited gets to the nub of Kitchers (2012b, 185)beef
with this sort of insouciant platonism:

Imagine properly educated counterparts of Bombelli, Euler, Hamilton and

Lagrange who fully subscribed to the platonistic wisdom. Like Mendel, they

K i t c h e r a g a i n s t t h e P l at o n i s t s [27]
would surely reflect on how their instruments, in this case, their symbolic
practices, enable them to detect [better:discover] the underlying entities, and
platonistic wisdom would supply them no answer. Thus if they had what is sup-
posed to be the correct philosophical view of the matter, they would not have
been able to proceed as they did. They would have had to hesitate, as Bombelli
did, although for considerably different reasons.

Looking past the rhetoric of detection, we may put the point as follows.
According to the platonistmoderate or otherwisethe truth of a math-
ematical claim is one thing; its acceptability by mathematical standards is
another. Mathematicians who took this view would therefore presumably
want some sort of proofor at least reason to believethat acceptability
by mathematical standards is a guide to truth. They would want a sound-
ness proof, in effect, not just for the system they presently accept but also
for the informal procedures that lead them to modify that system, as in the
case at hand. But platonism provides no such proof. This shows, first, that
these mathematicians were not (consistent, thoughtful) platonists, since
they saw no reason to demand proof of the soundness of their procedures.
But it also shows, more importantly, that platonism cant be the right phi-
losophy of mathematics since it would force us to represent this episode
in the history of mathematics as a matter of mathematicians blundering
around as sleepwalkers without adequate justification (Kitcher 2012b,
169), when in fact it is one of the great success stories in the history of the
My moderate platonist answers that since intramathematical justifica-
tion is justification enough, the heroes in Kitchers story were fully justified
in proceeding as they did (insofar as their procedures were mathematically
unexceptionable), and would have proceeded just as they did even if they
had been apprised of the platonistic wisdom. Just as astronomers can pro-
ceed by means of epistemic rules that they cannot justify independently, so
mathematicians can proceed by means of whatever principles guide their
practice without any independent proof of soundness. This does not mean
that astronomy does not aim at (transcendent) truth; the same goes for
mathematics, mutatis mutandis.8

8. Amore complete development of this point might proceed as follows:Although

there have occasionally been dissenters even in the modern periodfor example,
Kronecker and constructivist opponents of the axiom of choicemodern mathe-
matical practice for the most part presupposes a principle of plenitude, according to
which, very roughly, there are as many mathematical objects/structures as there could
possibly be. Given this presupposition, the introduction of complex numbers, ideal
elements in geometry and algebra, Cantorian sets, infinitesimals, higher categories,

[28] The Philosophy of Philip Kitcher


This account represents mathematicians as sticking their necks out:as aim-

ing at one thing (a true characterization of certain abstract objects) on the
basis of criteria (conformity to the rules governing the manipulation of
mathematical symbols) that do not immediately guarantee that this aim
has been achieved. As Kitcher notes, any account of mathematics on which
mathematics has a subject matter will have this feature.9 Even if mathemat-
ics is not concerned with abstract entities but rather with the hypothetical
activities of an ideal collector, there will still be a gap between the ultimate
aim of mathematical assertiontruth with respect to this modal subject
matterand the criteria on the basis of which mathematicians assess can-
didate assertions. Kitchers (2012b) brief against platonism is best under-
stood as a brief against any theory that posits such a gap. The argument
is not, as with Benacerraf, that any such theory leads to skepticism about
mathematics. It is rather that any such account will represent the heroes of
Kitchers historical narrative as proceeding either blindly or recklessly, fail-
ing to notice, or at least to worry about, the gap between the grounds for
mathematical assertion and the ultimate standard of correctness in mathe-
matics, namely, conformity to an invisible subject matter. Ihave responded
by conceding the gap while insisting that it is neither reckless nor irre-
sponsible for mathematicians to ignore it. More could be said in defense of
this response, but let me turn instead to Kitchers proposed alternative:an
account of mathematics on which the gap simply disappears.

and the rest thus cannot fail:if the posited structures are consistent, they exist. This
principle is notoriously hard to formulate; this is an important unsolved problem in
the philosophy of mathematics. (See Maddy 1997 and Balaguer 1998 for discussion.)
A fully adequate platonist response to Kitchers challenge would include a formula-
tion of this principle and then the claim that this principle stands to mathematics
as the metaphysical presuppositions of perceptual knowledge, inference to the best
explanation, and the rest stand to natural science. If settled science simply takes it
for granted, without independent justification, that the external world exists, that the
laws of nature are as simple as they can be given the phenomena, and so on, then it is
rationally permissible for natural scientists to take these things for granted. The mod-
erate platonists view is that the same goes for the principle of plenitude presupposed
by the method of free postulation in mathematics.
9. We might conceive mathematics as modal logic, or the study of structures, or
as the idealized science of human operations. Some of these conceptions dont make
much headway with respect to Benacerrafs original dilemma but none of them has
any advantages over Platonism with respect to the challenge of fitting the ontology
to the evolution of mathematical practice. For all of them say too much, introduc-
ing shadow entities behind the languages with which the mathematicians play their
increasingly sophisticated games (Kitcher 2012b,185).

K i t c h e r a g a i n s t t h e P l at o n i s t s [29]
Kitchers (2012b, 186)positive view is that mathematics is a game for
the manipulation of meaningless symbolsor better, a family of evolving
and overlappinggames:

Mathematicians and commentators on mathematics use true to mark out the

statements at which mathematics aims. On many occasions, what the mathema-
tician is seeking is a licensed transition within a well-established system:he or
she wants to produce a certain kind of statement using the transitions that are
allowed. There are other times... at which mathematicians look for modifica-
tions of those systems that will accord with broader methodological rules....
As Iput it more colloquially earlier, they are trying to find new games that are
worth playing. If they are successful, then new language will be adopted and
their successors will hail some sentences in those languages as being worth
inscribing in the books, as legitimate starting points for further transitions, in
short, as true. So thanks to Bombelli, we take it to be true that (2 + i)3=2 + 11i,
and thanks to Euler we suppose its true that i-i=e/2. On the account Iwant to
defend, this simply means that we can reach these equations by acceptable rules
in language-games that are worth playing.

This is a version of what Michael Resnik (1980, ch. 2)calls game formal-
ism. Kitcher cites Wittgenstein as a predecessor, though of course the view

For the formal conception, arithmetic is a game with signs, which one may well
call empty, thereby conveying that (in the calculating game) they do not have
any content except that which is attributed to them by their behavior with
respect to certain combinatorial rules (game rules). Achess player makes use
of his pieces in a similar fashion; he attributes certain properties to them that
constrain their behavior in the game, and the pieces are only external signs of
this behavior. (Thomae 1898, quoted in Frege [1903] 2013, v.2,9798)

Kitchers innovation is to historicize the formalist view of mathematics,

emphasizing that the symbolic games that constitute mathematics can
change over time and that these changes may be governed by methodologi-
cal desiderata that license modifications and innovations so as to yield new
games that are worth playing, either for their intrinsic interest or for
their usefulness as instruments.
Stated somewhat more explicitly, the view is this:At any given stage in
its development, mathematical practice will involve an alphabet of unin-
terpreted symbols, a grammar that brands certain strings as well formed,
an archive of accepted strings that have been inscribed in the books as

[30] The Philosophy of Philip Kitcher

licensed starting points for derivations, a set of rules for deriving new
strings from accepted strings, and a set of norms for modifying existing
rules, extending the alphabet, and introducing new rules. These rules and
norms may be implicit, in the sense that mathematicians may grasp them
without being able to state them. Some may be formal in the sense that a
computer could implement them; others may be fuzzier and harder to for-
malize. Astring is derivable at t iff there is a derivation of it from strings
that are accepted at t in accordance with the rules in place at t. The proxi-
mate aim of ordinary mathematics is the production of derivable strings,
in the following sense: When mathematicians assess the correctness or
acceptability of a mathematical claim, they check for derivability. When a
string has been derived and the derivation confirmed, it is licensed as an
accepted string, which may then serve as a starting point in subsequent
As is well known, two rather different versions of formalism can be
developed from this starting point. The less radical variety holds that ordi-
nary mathematical statements are meaningful, truth-evaluable claims with
a metalinguistic subject matter (Curry 1951). On this sort of view, when
a mathematician assertorically utters an otherwise meaningless string
Ssay, ei + 1=0his statement is really shorthand for something like
S is derivable in such and such a game. This sort of view gives math-
ematics a linguistic/combinatorial subject matter. A mathematical state-
ment so understood is true in the ordinary sense just in case the linguistic/
combinatorial facts are as it says they are. Indeed it gives mathematics a
platonistic subject matter, since the strings and derivations in which this
sort of mathematics trades are presumably abstract types rather than con-
crete inscriptions made of chalk or ink. And yet the view does not posit
a yawning gap of the sort that worries Kitcher. For this sort of formalist,
a mathematician who affirms a mathematical statement on the basis of a
correct derivation does not stick his neck out very far, since the existence
of the concrete derivation he produces guarantees that the (platonistic
combinatorial) truth condition of his assertion is satisfied.
For better or worse, however, this is not Kitchers view. On the version
of formalism Kitcher prefers, the mathematical statement itself says noth-
ing. It is not the description of a fact. It is like a configuration of chess
pieces. The mathematician who puts it forward is not making a combina-
torial statement about the game that he is playing; he is simply playing
the game. We can say, if we like, that a mathematical statement is true
iff it is derivable in the relevant game, and Kitcher does sometimes speak
in these terms. But truth in this sense is very different from truth in
other areas (hence the question mark in Kitchers title). It is a platitude

K i t c h e r a g a i n s t t h e P l at o n i s t s [31]
governing truth as we ordinarily conceive it that a statement is true if and
only if things are as it says they are. But according to Kitchers formalist
approach, a mathematical statement does not say anything. We do not call
a configuration of chess pieces true when the rules permit its derivation
from the starting position. More generally we do not call a move in a rule-
governed practice true simply because it is permitted by the rules. Better
to say, as Kitcher is clearly tempted to say, that like configurations in chess,
mathematical strings cannot be true or false. They can be derivable or not, and
that isthat.
The main argument for this sort of formalism is that it provides the basis
for a fully vindicatory reading of the history of mathematics in which epi-
sodes of radical innovation, like the introduction of complex numbers (or,
as Kitcher might have said, Cantorian sets) do not involve a bold leap into
the beyond but rather simply the introduction of new games constrained
by norms of consistency and fruitfulness. Even those of us who resist for-
malism should feel the force of this consideration. Occasional dissenting
voices to the contrary notwithstanding, it is a defining feature of modern
mathematical practice that mathematicians are free to innovate as they
likeintroducing new assumptions, objects, and structures, constrained
only by (hard-to-articulate) norms of consistency and fruitfulness. Any
philosophy of mathematics that cannot ratify this Cantorian idea is false
to the best modern self-understanding of the subject. Kitchers formalism
is expressly designed to satisfy this desideratum.
I have sketched a version of platonism that is also (I think) well placed to
satisfy it, so Ideny that this argument favors formalism. But even if formal-
ism were better placed to accommodate the fact of Cantorian freedom, this
would not settle the matter. It remains to consider the objections to this
sort of formalism, since they may tip the scales in the opposite direction.
One familiar objection comes from Gdels theorems. Let G be a Gdel
sentence for (say) ZFC of the sort that figures in the proof of the first
incompleteness theorem. Afamiliar sort of reflection persuades us that G
must be true (if ZFC is consistent, as we think it is). But G is not deriv-
able in ZFC or in any other accepted theory, so truth and derivability must
Kitchers (2012b, 191)response is that the sequence of worthwhile sys-
tems proceeds indefinitely:

One of the directions in which it can extend consists in the addition to any for-
mal system adequate to arithmetic of the pertinent Gdel sentence, to yield a
new formal system for which the same extension can be carried out . . .. We
learn from Gdel that there will be no first order system adequate for the whole

[32] The Philosophy of Philip Kitcher

of mathematics. That lesson is perfectly compatible with the thesis Iespouse, to
wit, that for any mathematical truth there is a worthwhile system within which
that truth can be reached by licensed transitions.

The suggestion is that mathematical practice involves a metarule that per-

mits the extension of any acceptable formal system by the addition of its
Gdel sentence to yield a new, equally acceptable system in which that sen-
tence may (trivially) be derived. Since the new system is worthwhile, the
derived sentence is true in the only sense in which mathematical state-
ments can be true. The formalist can therefore echo the widely held verdict
that even though G cannot be derived in ZFC, it is nonetheless true (if ZFC
is consistent).
Of course the formalist cannot quite echo the informal reasoning that
typically persuades us that G is true. That reasoning, after all, involves
reflection on what G says. We reason as follows:G says that there is no
number that codes a proof of G in ZFC. But as Gdels construction shows,
if ZFC is consistent then there is indeed no such number. So if ZFC is con-
sistent, what G says is true. This sort of reasoning is not available to the
formalist, since according to him, G does not say anything.
Even if we waive this point, there is a more interesting difficulty in
the vicinity. Take G to be Con(ZFC):an arithmetical sentence that codes
the assertion that there is no proof of 0=1 in ZFC. The correct view of
this statement is that it is almost certainly true (hence almost certainly
unprovable in ZFC) but that it is nonetheless, at present, unacceptable as
an axiom (or theorem) of mathematics. Someone who derives a statement S
in ZFC + Con(ZFC) has not thereby proved S. Her proof does not license
the use of S itself as a premise in further derivations. At best she has
proved the conditional If Con(ZFC) then S. The evidence that persuades
us that Con(ZFC) is almost certainly true thus resembles the evidence for
Goldbachs Conjecture. Whatever force this evidence may have, it is not the
sort of evidence that licenses inscribing the sentence in question in the
books, at least not according to the methodological norms of mathematics
as we havethem.
What can the formalist say about Con(ZFC)? Is it derivable in a game
worth playing and hence true in the only available sense? Well, if the fore-
going is correct, the norms governing actual mathematical practice do not
license a transition from ZFC to ZFC + Con(ZFC). The latter theory is obvi-
ously worth studying, but it is not at present acceptable or authorized. If
it were, we would be happy to say that the consistency of ZFC is like the
consistency of Euclidean geometry: a theorem of mathematics. But that
is not what we say. Like any statement beyond ZFC, Con(ZFC) comes with

K i t c h e r a g a i n s t t h e P l at o n i s t s [33]
a question mark (if only a very faint one in this case). Unlike the axiom of
infinity or the axiom of choice, Con(ZFC) is not an acceptable, fully detach-
able resource for proving theorems. So if mathematical truth simply con-
sists in derivability in a system that is fully acceptable for this purpose,
Kitchers formalist must say that Con(ZFC) is not true. Of course this is not
to say that Con(ZFC) is false. The formalist can say instead that Con(ZFC)
is like the continuum hypothesis as he presumably understands it:neither
provable nor refutable in any fully acceptable mathematical system, and
hence, by the formalists lights, neither true norfalse.
It would be disastrous if the formalist were forced by this admission
to the metamathematical conclusion that there is no fact of the matter
whether ZFC is consistent. But Kitcher is not quite forced to this conclu-
sion. The metamathematical, combinatorial claim that there is no proof of
0=1 in ZFC is, for all Kitcher says, a fully contentful claim whose truth
does not consist in its derivability. That claim might therefore be true, in
the ordinary sense, even if the number-theoretic statement Con(ZFC) is
neither derivable nor refutable, hence neither true nor false in the only
sense appropriate to it. The difficulty is that this severs the link, essential
to foundational research in mathematics, between metamathematical
claims of consistency and derivability, on the one hand, and the ground-
level mathematical claims that we normally take to formalize or code
them. This part of mathematics is predicated on the assumption that
we can convert modal or combinatorial claims about the consistency of
formal systems into mathematical claimsclaims about the existence of
models or about the existence of (numbers coding) formal derivations.
The objection is that Kitcher-style formalism would call this aspect of
mathematical practice into question. Understood as a modal/combinato-
rial question, the consistency of ZFC appears to be a factual question with
an answeralbeit a question we cannot answer within established math-
ematics. By contrast the mathematical question whether there exists a
number that codes a proof of 0 = 1 in ZFC, or the question whether
there exists a model of the ZFC axioms, must be understood as a question
that has no answer at all, since the candidate answers are underivable in
every authorized game and hence untrue in the only sense pertinent to
such claims.
Be all this as it may, the main source of resistance to formalism does
not come from these somewhat recherch considerations but from a
more basic source. Mathematics presents itself as a domain of inquiry.
As we normally think, mathematicians raise questions, entertain con-
jectures, make assumptions, engage in reasoning, and often come to

[34] The Philosophy of Philip Kitcher

believe mathematical propositions for good reasons. The philosopher
for whom mathematics has a subject matter can take these appearances
at face value. The formalist, by contrast, must reinterpret the practice
from top to bottom. Consider the number theorist who (as we would
normally say) first wonders whether every even number is the sum of
two primes, then checks the numbers up to 17 billion and comes to sus-
pect that every even number is the sum of two primes, then finds what
he takes to be a proof and comes to believe this proposition, but then
discovers a mistake in his proof and comes to doubt it. Aphilosophy of
mathematics that cannot represent this sequence of events is plainly
inadequate. And the trouble is that for the formalist, there is no such
thing as wondering whether or suspecting/ believing/
doubting that
every even number is the sum of two primes. This makes as much sense
as wondering whether

Of course there is such a thing as wondering whether a certain string is

derivable in an authorized game. As I have stressed, this metamathemati-
cal proposition may express an ordinary truth-evaluable proposition. So a
formalist may say, Insofar as mathematics looks like a first-order inquiry
in which one first wonders whether S and then comes to believe that S for
what one takes to be good reasons, the appearances are misleading. All of our
wondering and reasoning is really metamathematical. The ground-level lan-
guage of mathematics is not a language in which questions can be asked and
If the formalist is forced to respond in this way, his view is false to the
self-understanding of mathematicians and the phenomenology of math-
ematical practice. Mathematicians use the language of mathematics to
formulate their questions and conjectures, to express their doubts, and
to give their reasons. For the formalist all of this is at best elliptical. The

K i t c h e r a g a i n s t t h e P l at o n i s t s [35]
platonist by contrast can take these aspects of mathematical practice at
face value, and this must certainly be reckoned a clear advantage for the
This brings us to the last and most important objection to formalism,
first pressed by Frege ([1903] 2013, 91) and widely regarded as deci-
sive.10 Suppose Iwant to know how many socks to buy and reason as fol-
lows:Each kid needs six pairs of socks, and Ive got 12 kids. 6 12=72.
So Ineed 72 pairs of socks. This line of reasoning is clearly cogent, and we
could easily represent it as a valid argument if we wished. The standard
(Fregean) reconstruction would involve both mathematical and non-
mathematical premises, together with bridge principles like The num-
ber of socks=n iff there are n socks, where the left-hand side involves
a numerical singular term and the statement on the right involves the
(readily analyzable) adjectival use of a number word. According to the
platonist, this argument supports its conclusion in the familiar way, that
is, by showing that it is impossible for the premises to be true and the
conclusion false. For the formalist, by contrast, the mathematical prem-
ises arelike

strictly incapable of ordinary truth, and the bridge principles are hybrids
involving meaningful English expressions on the right and meaningless
squiggles on the left. An argument of that sort cannot support its conclu-
sion in the familiarway.11
Of course the formalist can try to recast the reasoning as involving a
truth-evaluable metamathematical premise at the crucial point:

10. For example, by Dummett (1991, ch.20).

11. It is an important bit of unfinished business for the formalist to give an account
of the norms governing these bridge principles. They are not statements of pure

[36] The Philosophy of Philip Kitcher

I have 12kids.
Each kid needs 6 pairs ofsocks.
The string 6 12=72 is derivable in an authorizedgame.
So Ineed 72 pairs ofsocks.

But this looks like terrible reasoning.

It is clear what the formalist needs: a metatheorem to the effect that
when we reason from a mixture of nonmathematical and mathematical
premises, via licensed rules for mathematical symbol manipulation, to a
nonmathematical conclusion, that conclusion is always a modal conse-
quence of the original nonmathematical assumptions by themselves. Such
results may or may not be attainable; one would need to take the question
case by case. The crucial point for present purposes is that ordinary applica-
tions of mathematics proceed without any sense that some such result is
We can put this point as a tu quoque. Kitcher objects to platonism on
the ground that it represents mathematicians as sticking their necks
out at crucial moments, making bold ontological claims, when in fact, as
mathematicians understand their own practice, there is no felt risk and
no demand for special justification. The objection is that platonism fails to
vindicate the practice of mathematicians as we find it. In response we may
say that whatever the merits of this charge against the platonist, it applies
in spades against the formalist. Everyday applications of mathematics
not to mention sophisticated applications in the sciencesall proceed as
if mathematical reasoning were genuine reasoning and as if mathematical
premises figure in applied mathematics as genuine premises that might fur-
nish reasons for accepting the nonmathematical conclusions they support.
This makes good sense if mathematical statements express truth-evaluable
thoughts. But it makes no sense if mathematical statements are like con-
figurations of chess pieces. Or better:this makes no sense in the absence of
a metatheorem justifying the use of this uninterpreted formalism, a theorem of
a sort that no mathematician has ever felt the need to supply. Imagine ordinary
applied mathematicians going about their business fully apprised of the

mathematics, and so their truth (or acceptability) cannot consist simply in their deriv-
ability in some formal game. Rather the norms governing their use must derive in
part from these formal rules and in part from the independently given meanings of
the nonmathematical vocabulary they contain. To my knowledge, formalists have had
nothing to say about how these two constraints might interact.

K i t c h e r a g a i n s t t h e P l at o n i s t s [37]
formalist wisdom, treating meaningless strings of squiggles as if they were
premises in arguments for factual conclusions:

Like Mendel, they would surely reflect on how their instruments, in this case,
their symbolic practices, enable them to arrive at knowledge of these ordinary
matters of fact, and formalist wisdom would supply them no answer. Thus if
they had what is supposed to be the correct philosophical view of the matter,
they would not have been able to proceed as they did. They would have had to
hesitate, as Bombelli did, although for considerably different reasons.

So even if Kitcher is right to say that platonism represents mathematical

practice as involving bold assumptions that mathematics itself does not
pause to justify, the same must be said of formalism. In fact the problem is
more acute for formalism. The moments in the history of mathematics that
generate Kitchers problem for platonism are relatively rare:the introduc-
tion of brand-new existence assumptions that cannot be derived within
a previously accepted theory. By contrast the moments that generate our
problem for formalism are the opposite of rare:they arise whenever math-
ematics is applied, which is to say that they arise millions of times eachday.
It is telling that Kitchers (2012b, 18891) discussion of applications
focuses on the most elementary sort of case: the use of numbers in counting
collections. It is universally agreed that if this were the only use for number
words, there would be no ground for regarding mathematical statements as
descriptive. The impulse to do so comes when we begin to use mathematical
statements in reasoning. It is at this point that the realist has a clear advan-
tage: she can say that this reasoning is just what it seems to be, and she can
elaborate the details in Freges way. Kitchers formalist must say, No, the
appearances here are profoundly misleading: we treat these sentences as if
they were fit to serve as premises in the usual way, but they arent. Any use we
make of them depends on an assumptionthe conservativeness of the for-
mal apparatusthat no mathematicians (and few philosophers) have felt the
need to justify. This suggests that even if Kitchers formalism is the correct
philosophy of mathematics, it is not a quietist philosophy of mathematics: the
sort of philosophy that represents our practice as unproblematic as it stands.

Reply toRosen

Gideon Rosen offers astute challenges to my recent discussion of platonism

in mathematics. He questions both my critique and the positive account

[38] The Philosophy of Philip Kitcher

of mathematical truth I defend. I shall start with my suspicions about
Thinking of mathematics as concerned with a domain of abstract objects
allows a straightforward understanding of mathematical truth but raises
worries about epistemological access to that domain. Paul Benacerrafs
(1973) seminal paper provided a way of making those worries precise.
Benacerrafs commitment to a sharply formulated causal constraint on
knowledge allows the platonist to circumvent objections by denying that
mathematical knowledge is subject to that constraint. I hoped to use the
history of mathematics to close the loophole.
Heres my challenge in its baldest version. Discoveries about a particular
sort of object occur only when there is a change in the cognitive relations
between some people (the discoverers) and objects of that sort. The history
of mathematics reveals cases of mathematical discovery. We can under-
stand those episodes without describing any changes in cognitive relations
between mathematicians and the abstract realm mathematics is supposed
to describe.
According to Rosen, theres a version of platonism, moderate platonism,
that escapes my argument. I have supposedly assimilated discovery to the
detection of objects, and, because detection is a causal process, there is
apparently no advance beyond Benacerraf (although, as Rosen goes on to
point out, its possible to formulate a version of my argument that doesnt
turn on the notion of detection).
There are, I believe, two notions of detection. The stronger notion
requires a causal connection. Yet, as my admittedly imprecise talk of
changed cognitive relations (Kitcher 2012b, 170) was intended to sug-
gest, Ihad something weaker in mind:to detect a kind of object is to come
to be in a state in which you are able to know some proposition making
essential reference to that kind of object (a state nobody has previously
enjoyed). Iwanted to issue a challenge:Here are the processes mathemati-
cians undergo on occasions of detection in this weaker sense; tell us how
undergoing processes like that yields knowledge of new abstract objects.
(To be fair, Rosens imputation of the stronger notion is quite understand-
able, given some of my less guarded remarks about Gdelian peeks.)
In passages Rosen doesnt directly address I tried to imagine ways in
which themoderate!platonist might answer this challenge. The best I
could think of was to say that producing expressions that denote complex
numbers and manipulating them in ways that are mathematically fruitful
is sufficient for detecting complex numbers (2012b, 184)a formu-
lation that seems not so distant from Rosens own preferred account. As
I went on to argue, if that were correct, it would be hard to understand

K i t c h e r a g a i n s t t h e P l at o n i s t s [39]
why the pragmatic considerations about fruitfulness, so prominent in the
history of the acceptance of complex numbers, are relevant. Moreover,
how exactly do the symbol-manipulating processes of the mathematicians
involved provide information about an abstract realm?
Lets start with a concrete example of the sort that plainly moved
Benacerraf. You buy an expensive present for someone dear to you. You go
to great efforts to ensure that it is well hidden. Yet before you spring the
great surprise the dear one offers a compelling demonstration of advance
knowledge, a description of the present both richly detailed and accurate.
When you ask how this knowledge has been obtained, you receive an aston-
ishing response:I sat down and thought very hard about what you might
give me, exploring the possibilities on paper. They led me to the description
Ive just given you. Isuspect that, were this to happen, you would not be
convinced by the explanation. Youd suppose that your best-laid schemes of
concealment had gone (as Burns puts it) agley.
We dont believe that people can obtain knowledge about physical
objects unless (to use Benacerrafs phrase) the space-time worms of
knower and object intersect (or, more exactly, come close to intersecting).
Thats what motivates the causal constraint. Moderate platonists will not
be moved by this, seeing it as a crude and misleading analogy. Yet a puzzle
remains: there seems to be a gap between the mathematician (Bombelli,
Euler, Hamilton) at the writing desk and the realm of abstract objects. How
do the scribblings performed at the desk, or the thoughts that generate
them, yield knowledge of the abstract realm? How do the cognitive rela-
tions change during the process of discovery?
Really modest platonists might try to close the gap by force. The really
modest platonist declares that all there is to being an abstract object is
to be susceptible to being known through various intellectual activi-
ties (expressed, perhaps, in experimental scribblings). I strongly suspect
that any attempt to provide a semantics for talk of abstract objects that
would articulate the declaration would lead to difficulties parallel to those
my critics discerned in my own earlier efforts (Kitcher 1983; see Burgess
and Rosen 1997, 2023). But in any case, the only residual issue between
really modest platonism and my position would be a dispute over preferred
modes of speech.
Rosens preferred moderate platonism attempts to do more than this.
He sees the task as one of explaining why the grounds cited by the his-
torian (and the mathematician) constitute adequate grounds for believ-
ing the existence theorems, understood as claims about abstract entities
and takes this to be part of the practice of mathematics in just the way
that using inductive generalization is part of the practice of the natural

[40] The Philosophy of Philip Kitcher

sciences. Rosen supposes that raising any further question about how the
gap between what goes on at the mathematicians desk and the abstract
realm is to be bridged rests on the nave thought that detection, in the
strong, causal sense, is requireda thought we should discard.
It seems to me that this is simply to dodge the crucial question. Amore
appropriate analogy with the natural sciences would focus on those
moments at which new instruments or techniques are introduced to dis-
close previously unsuspected denizens of the universe. Questions about
the ability of the tool or technique to do what its proponents claim for it
are often highly significant. Can the telescope really disclose features of the
heavens? Are the alleged structures within the cell artifacts of the micro-
scope or the method of preparation? What exactly does an fMRI reveal? On
all these occasions scientists (not just philosophers) try to understand how
the processes that generate judgments can reveal properties of the objects
they are attempting to characterize.
If the same question were to occur with respect to the use of an estab-
lished technique or instrument, scientists would not have to evade it by
declaring that offering a fully worked-out theory of evidence was none
of their business. In advance of the telescopic identification of Neptune,
Adams and Le Verrier could have explained how the observed deviations in
the orbit of Uranus were related to the positions and masses of other bod-
ies (in the rough vicinity), and thus how their observations served to detect
such bodies. After the telescopic observation they could have noted that
telescopes work by focusing light radiated from distant objects, so that the
magnified image represents the distant planet. The conversation doesnt
stop at the point where Rosen supposes that nothing more is to besaid.
The episodes in my historical narrative reveal mathematicians intro-
ducing new symbols and rules for their manipulation, with the resulting
system becoming accepted when its shown how fruitful it is in answering
questions hailed as significant. I ask, How does doing that sort of thing
give knowledge of propositions, construed as the platonist understands
them, as referring to elements of a realm of abstract objects? Rosens
mathematician (and his moderate platonist) answer, Thats the way math-
ematics works! The extensions of mathematics produced in this way yield
knowledge of a wider and wider part of the abstract realm. Induction on
the history of mathematics tells us that. Surely induction on the history of
mathematics tells us something. It reveals that certain types of pragmatic
criteria are good for expanding mathematics. Yet it doesnt answer my orig-
inal question, since there are no instances in which the practice of extension
is explained as a practice that provides knowledge of the abstract realm. If
you want to defend the general ascription of a property to a population,

K i t c h e r a g a i n s t t h e P l at o n i s t s [41]
you had better have a sample of instances that clearly have that property,
from which you can generalize.
As I noted, a part of my view on which Rosen doesnt comment (2012b,
18385) explores various possibilities for the moderate platonist. I dont
see how Rosens preferred epistemology improves on these.
But what of my rival account, seen by Rosen as game formalism? He
is entirely correct to suppose that game formalism dates back to the nine-
teenth century and that it provides one way to elaborate the approach I
favor. Yet there was reason to invoke Wittgenstein (rather than Thomae).
For Wittgensteins later writings contain three themes from which I draw
inspiration. One, explicit in my 2012b paper, is the emphasis on language
games directed toward different functions (not in the description busi-
ness). A second, related concern rejects the idea of some problematic depth
behind the surface phenomena: Wittgenstein excoriates the idea of some
hard reality behind mathematical statements, seeing that reality as idle
(as, I think, it becomes in Rosens moderate platonism). The third, of which
I made little use in my paper, is to separate the idea of meaningfulness
from correspondence-truth conditions in favor of a connection with use.
Taking that third theme seriously, I reject the inference from the premise
that mathematical terms lack reference to the conclusion that mathemati-
cal statements lack meaning.
An alternative way to articulate my approachmathematical pragmatism
is to start from the idea that the meanings of mathematical terms are con-
stituted by their conditions of use, both within mathematics and in the
uses to which mathematical vocabulary is put in scientific inquiry. (More
on Rosens penetrating questions about this practice later.) The history
of mathematics is a process in which mathematicians not only prove new
theorems using established language but also periodically extend their
languages. Mathematical progress consists in finding answers justifiable
by the settled rules to questions recognized as significant (and we should
allow for progressive revisions of the rules and for progressive reworking
of the notion of significance). The mathematical truths are the stable ele-
ments that emerge as we continue, indefinitely, to make progress. (I obvi-
ously echo thoughts of Peirce and James.)
Mathematics obtains its initial content from the ways arithmetic and
geometric vocabulary is used in operations of collecting, combining, and
comparing. Here my old Millian account (1983, ch. 6) can be refined to
explain the content of elementary mathematics. The history of math-
ematics reveals how further expressions become connected with those
introduced earlier, with new patterns of usage fixing the content of mathe-
matical language. As some people who have responded to my proposals have

[42] The Philosophy of Philip Kitcher

seen, it might be thought of as Hilbert historicizedalthough whereas
for Hilbert elementary (contentful) mathematics is understood from a
Kantian perspective, Mill plays the analogous role in mystory.
Offering this alternative to game formalism doesnt excuse me from
addressing Rosens two most important objections. As he argues, no
account of mathematical truth that identifies it with what can be derived in
a fruitful formal game can be adequateand matters are no better if truth
is viewed as what is achieved and preserved under mathematical progress.
Moreover my explicit claim that mathematical practice entangles mathe-
matical statements with other areas of inquiry only reinforces his challenge
to explain how the applications of mathematics are to be understood.
Mathematicians would claim that the Gdel sentence for ZFC is true,
and Rosen denies that I can endorse this judgment. Since Con(ZFC) is
presently unacceptable as an axiom (or theorem) of mathematics, Icant
appeal to some larger system in which it is derivable, and thus it fails to be
true. Apparently, then, the formalist (or the pragmatist) must abandon the
common mathematical judgment.
But why exactly is Con(ZFC) unacceptable as a new axiom? Not, Ithink,
because it comes with a very faint question mark. Sometimes, historically,
extensions of mathematics have come with question marks that were not,
initially, so faint (as my example of complex numbers reveals). The prob-
lem with simply tacking Con(ZFC) onto the other axioms is that it vio-
lates a methodological norm of mathematical practice, one that opposes
tacking, in favor of seeking extensions of mathematical systems that
systematize previously obtained results and answer unresolved questions.
Mathematicians would welcome systems that encompass ZFC, offering
rich new languages for further developing old parts of mathematics, for
opening up new questions of purely mathematical significance, and per-
haps providing new mathematical tools for empirical inquiry. Because of
the metamathematical considerations to which Rosen points, they would
maintain that any such future system deciding between Con(ZFC) and its
negation should contain Con(ZFC) as a theorem. Thus my initial formula-
tion was not quite adequate as an account of mathematical truth. Ishould
have said that, in addition to the statements derivable in accepted sys-
tems, a statement can properly be counted as true if there are grounds for
expecting that any extensions of those systems able to decide between the
statement and its negation would contain the statement in question. Gdel
sentences are the most prominent examples.
Rosens subtle discussion of the implications of the incompleteness the-
orems raises further issues that are best approached after considering his
most fundamental objection to formalism (or to pragmatism). As Rosen

K i t c h e r a g a i n s t t h e P l at o n i s t s [43]
points out, mathematical statements are often deployed in conjunction
with other statements in reasoning about many aspects of nature. Any
account that either denies truth (formalism) or understands truth differ-
ently (pragmatism) must explain how reasoning goes when the two types
of statementsmix.
For elementary arithmetic and geometry this is relatively easily done,
and my preferred way of doing it is along the lines of my reconstruction
of Mill (Kitcher 1980, 1983, ch. 6). For the higher parts of mathematics
I would adapt a classic logical empiricist idea: the symbols of the math-
ematical systems are given concrete applications in context, by linking
them to physical objects, properties, and operations on those objects and
properties. Newton connects the language of the calculus to velocities
and distances, conceived as measured through performing various opera-
tions using clocks and rulers. The mathematical systems are interpreted
anew in the contexts of various types of inquiry, so that applied (or
mixed, to use an old term) mathematical statements are candidates for
To leave matters there is only to point in a direction along which an
answer to Rosens criticism might be found. Actually to give that answer
would require another historical enterprise. What is required is to show
how mathematics can be applied in measuring the world in a succession of
reinterpretations of the available formalisms, where schemes of measure-
ment introduced at earlier stages are taken for granted as further steps are
taken, so that we move from the simple (Millian) ways in which elemen-
tary mathematics is used to the most sophisticated applications of com-
plex analysis, tensor calculus and linear algebra in contemporary physical
theory. Some of my own early work in the history of mathematics (Kitcher
1973)inspires me to think this is not impossible (at least not for the intro-
duction of the calculus). But, to the best of my knowledge, nobody has
yet offered a detailed history of measurement practices that would settle
Metamathematics should be viewed from the same perspective, as the
construction of a system for applying mathematical language to processes
that occur in the practice of mathematics itself (notably, the production of
proofs). Iwould hope that the historical approach to measurement would
enable us to see clearly just how a formalist/pragmatist can reconstruct the
reasoning mathematicians engage in when they reflect on the incomplete-
ness theorems, and thus address the questions Rosen poses. But for the
present Ican offer only this outline of a way of responding to some of his
doubts rather than providing fully developed answers.

[44] The Philosophy of Philip Kitcher


Kitchers Two Design Stances


P erhaps everyone who has thought much about the issue is now a func-
tion pluralist in some form or other. However, the interesting and sub-
stantial controversy concerns the nature and theoretical role of each notion
of function. Philip Kitchers (1993c) paper Function and Design develops
a pluralist view that in some ways is close to my own, but in other ways it
remains an influential challenge to the view that Idefend. Iwelcome this
opportunity to clarify how Kitchers theory differs from the theory Ifavor
and to explain why his argument for the form of function pluralism that he
supports is at best inconclusive, even though it raises important challenges
that need to bemet.
Kitcher (1993c, 379) begins his paper on functions by telling us that
there is some unity of conception that spans attributions of function
across diverse contexts. These contexts include those in which functions
are attributed to artifacts in everyday life, as well as to parts or processes
in organisms in creationist as well as contemporary biology. This unity, he
writes, is founded on the notion that the function of an entity S is what S
is designed to do (379). This might suggest that Kitchers analysis simply is
an etiological analysis.
Most etiological analyses tell us that an entitys function depends on
its history of selection. As Kitcher (1993c, 379) says, when we attribute
the function of releasing the metal bar to the mousetraps lever, we do
so because we believe that it was put there to do just that. And when
Harvey attributed the function of circulating blood to the heart, he prob-
ably believed that the wise and beneficent designer foresaw the need for
a circulation of blood and assigned to the heart the job of pumping (380).
As Kitcher further comments, it is also generally recognized that both
intentional selection and natural selection can ground functions.
However, Kitcher contrasts his analysis of functionsneedless to say
favorablywith etiological analyses. The main difference is that, on his
analysis, the links to intentions and to selection can be more or less direct
(1993c, 380). He is, it emerges, supporting a pluralist proposal in which
more than one notion of function is employed in attributions of the form
The (or a) function of S is to do Z. He calls one a strong etiological notion;
the other is less demanding, with close ties to Robert Cumminss (1975)
According to Cummins, the (or a) function of a part of a system is to do
Z if it contributes Z-ing to a complexly achieved capacity Z* of a containing
system that a researcher is trying to explain by means of a functional analy-
sis, also known as a mechanistic or operational explanation. The crucial
difference is that, on Kitchers analysis, the relevant complexly achieved
Z* capacities are not pragmatically determined but are capacities for which
the system is designed.
Unlike Cummins, Kitcher thinks that the etiological notion of function
also has an explanatory role in biology. However, Kitcher argues that the
etiological notion is too onerous, ontologically and epistemically, to be of
much use in explaining how organisms operate or function. In my view
Kitchers argument is more worrying than the one originally offered by
Cummins, which was a critique of Larry Wrights (1973) formulation of the
etiological notion.2
To assess Kitchers argument for function pluralism, I first need to
explore an interpretive question concerning his more and his less demand-
ing notions of function. Precisely how demanding is each of them? Ishall
argue that his more demanding notion is more demanding than it needs to
be and that his less demanding notion is open to two interpretations:on
one it is not demanding enough for the explanatory role that Kitcher
assigns it; on the other it is a middling-strong etiological notion that might
play both of the explanatory roles that Kitcher identifies. Ishall also argue
that Kitchers argument for function pluralism in any case leaves it open
whether there is a middling-strong etiological notion of function that
might play bothroles.

1. Kitcher uses the phrase strong etiological conception several times (e.g., 1993c,
389). He refers to a less demanding account of functions once (1993c,388).
2. For other arguments for function pluralism, see Amundson and Lauder 1994;
Godfrey-Smith 1993; Brandon2013.

[46] The Philosophy of Philip Kitcher


A preliminary question concerns the relation between design and selec-

tion. Kitcher speaks of both intentional selection and natural selection as
sources of design.3 It is unclear whether he wants to allow that there might
be other sources of design, and I return to this in a moment. However,
Kitcher seems to speak as if being selected to do Z suffices for being
designed todoZ.
If Iuse a teapot to water my plants, Iselect it for watering plants, but
Ido not design it for watering plants. And in selecting the right instrument
for a procedure, a surgeon does not design it for the procedure. Someone
might reply that if the surgeon selects an instrument from a tray by pick-
ing it up and holding it in her hand, then she holds it in her hand by design
(for the procedure). Still this is not the same as saying that the surgeon
designed the instrument for the procedure. The claim that something is
designed to do Z is usually taken to suggest that it was shaped, structured,
or organized to do Z (or that it was, so to speak, fashioned to doit).
Along the same lines, if a biological entity is exapted, it has a new adap-
tive effect for which it was not originally fashioned. Although the entity
might be refashioned for its new role, this does not always happen. An exa-
ptation might be maintained in a population by natural selection without
modification for its newrole.
However, Kitcher seems to use design such that, if S is (was) selected to
do Z, S is (was) designed to do Z.4 Were Iwrong on this, and his notion of
design is stronger, then his strong etiological notion of function is even
stronger than I am here suggesting. And this would strengthen and not
weaken the argument of this paper. However, Ido not think that this is a
misreading. So Ipoint out Kitchers slightly unusual use of design to put us
on guard against its usual connotations.
There are a couple of reasons someone might object to Kitchers claim
that natural selection is a source of design that are worth noting, though
Idisagree with both objections.
One is that a person might think that natural selection does not answer
Paleys question about how complex adaptations are explained.5 On the

3. See Kitcher 1993c,380.

4. See especially Kitchers 1993c,9n.
5. See Sober (1984, ch. 5; 1995)for support for the negative view on Paleys ques-
tion. As I here describe the negative and positive views, they are noncommittal on
the orthogonal question of whether natural selection explains why a particular indi-
vidual has a given trait (e.g., why Ihave an opposable thumb). The latter question raises
an additional (metaphysical) issue about the extent to which we essentially have the
ancestors that we do, and it is best treated separately.

K i t c h e r s T w o De s i g n S ta n c e s [47]
negative view of natural selection, good designs arise as a result of random
processes (such as mutations, genetic recombinations, and environmen-
tal changes that affect development). Then they are passed on through the
mechanisms of inheritance and development. Selection is only responsible,
on this view, for preserving and distributing good designs in the popula-
tion and for eliminating bad ones. In my view this is correct, except for the
word only. The negative view ignores the role of cumulative selection in the
evolution of complex adaptations, which do not arise in a single saltation.
So it fails to recognize how Darwin answers Paleys question.6 On the posi-
tive view, simple adaptations can arise without natural selection. Complex
adaptations could possibly arise without natural selection. But complex
adaptations are vastly more likely to arise if selection is part of the evolu-
tionary process, with its repeated rounds of fresh variation, selection, fresh
variation, selection, and so on.7 The chance of a random adaptive alteration
to an existing mechanism M (an earlier version of a wing, for example)
increases as a result of selection of M, since the more M is replicated, the
greater the chance that a lucky alteration to M (an improvement on the
wing) will arise. Of course selection of M will also increase the probability
that maladaptive alterations to M will arise too. But selection can eliminate
these, as well as pick up and run with the improvements. Complex adapta-
tions arise through this mix of selection and randomly arising alterations.
By eliminating bad designs and multiplying the instances of good ones,
selection channels evolution.
Kitcher anyway does not seem to use the term design to denote a process
that necessarily involves fashioning. So anyone who (in my view wrongly)
rejects the positive view of natural selection in favor of the negative view
need not reject his account of functions on that ground. Kitcher sometimes
speak of natural selection as fashioning traits, but he also distances this
manner of speaking from the positive view of selection.8
The other reason someone might object to Kitchers assumption that nat-
ural selection is a source of design is that it might seem anthropomorphic,
like talk of Mother Natures intentions. This might be thought acceptable
for casual talk but not for serious analysis. Of course Kitcher fully under-
stands that natural selection is a purposeless process that lacks foresight or
any other kind of sight. So this objection is a mere verbal quibble. Perhaps

6. See Neander (1995) for support for the positiveview.

7. If increasing the probability of an events occurrence is sufficient for counting
among the causes of an events occurrence, selection is generally among the causes of
complex adaptations.
8. This is how Iread Kitchers 1993c,9n.

[48] The Philosophy of Philip Kitcher

the scent of metaphor still lingers over talk of design in a way that it no lon-
ger lingers over talk of selection in contemporary biology. But it is perfectly
clear that neither Kitcher nor the orthodox contemporary biologists who
speak of a species design are suggesting that intention is involved.
In sum, Kitcher recognizes intentional selection and natural selection as
sources of design. And, in his idiolect, to say that something is selected to do
Z entails that it is designed to do Z, though to say that something is designed
to do Z does not entail that it is fashioned to do Z. There is a risk that, in
speaking of design in this way, Kitcher makes the etiological notion of func-
tion seem more demanding than it is. It could be hard to shake the usual
stronger connotations of design. But, having said this, Ishall adopt Kitchers
way of speaking about selection and design in the sections that follow.


In Function and Design, Kitcher describes the etiological analysis pro-

posed by Wright (1973, 1976), who spoke of neither design nor selection
in his formula (as he calls it) for function ascriptions, although he speaks
of these in the surrounding text. Kitcher (1993c, 384)remarks that more
recent etiological analyses solve certain problems for Wrights earlier
analysis only by sacrificing one of its virtues, its recognition of a common
feature in attributions of functions to artifacts and to organic entities.
I disagree with this assessment and with the claim that anything worth
having was sacrificed.
Before Iexplain why, it is worth noting that even Wright is a function
pluralist of a sort. That is, he recognizes many senses of function.9 Most
obviously there are social functions, such as weddings and funerals, as well
as logical and mathematical functions, such as conjunction and multipli-
cation. The values of one determinable can be a function of the values of
another in the way that pressure is a function of temperature in a confined
space. And people might complain that they do not function well under
duress, meaning roughly that they do not perform well. As Wright stressed,
an item might also be said to serve the function of Z-ing, or to function
as a Z-er, even if it does not have the function to Z. Some of these senses
of function are closer to the target notion than others, but it is function
ascriptions that take the form S has the function to Z in which Wright is

9. See Wright 1973,139.

K i t c h e r s T w o De s i g n S ta n c e s [49]
One of Wrights examples is of a belt buckle stopping a bullet and sav-
ing the life of the soldier wearing it. Wright says that the belt buckle serves
or performs the function of stopping the bullet and functions as a bullet
stopper, but it does not have the function to stop the bullet and save the
soldiers life, or not in the sense in which he is interested. Wright need
not suppose that we would never speak of the belt buckle as having the
function to stop the bullet, if we were (say) analyzing a hit by friendly fire
for the purpose of a military trial. Rather he is drawing our attention to a
certain notion of function by means of a locution that marks it out when
the contrast between the two locutions is explicitlydrawn.
Wrights main interest is in a teleological (or, if nonintentional, teleo-
nomic) notion of function. He tells us that the claim that the function of
the belt buckle is to hold up pants (and not to stop bullets, even if it did
both) is equivalent to saying that the belt buckle is there in order to hold
up pants (and not to stop bullets). In Wrights view all teleological explana-
tions invoke consequence etiologies. When an entity has a consequence
etiology, a consequence of the entity explains the entity. More specifically,
on Wrights analysis, an entity, S, has the function to do Z if and only if
(a)S does Z and (b)S is there (i.e., where it is and/or in the form that it is
in) because it does Z. Wrights hope is that this will work equally well for the
functions attributed to coffee mugs, calculators, and candelabra in every-
day life, as well as the functions attributed to biological entities, such as
hearts and hemoglobin, in pre-Darwinian biology and contemporary biol-
ogy. In this way its scope is ambitious.
However, Wright leaves certain details to be completed in context-
sensitive ways. For instance, he tells us that the formula is tenseless. In
other words, we may use whichever verb tense is appropriate in a given
case. Intentional design involves foresight, but natural selection does not.
So while a creationist might think that God gave mammals hearts with the
intention that they pump blood in the future, just as a potter gives a mug a
handle with the intention that it save drinkers from burning their fingers
in the future, a contemporary biologist will appreciate that natural selec-
tion cannot select traits on the basis of future advantage. Thus the second
requirement of Wrights formula is very schematic. In effect it says that if
S has the function to Z, S must be there because it does Z, or because it did
Z, or because it will do Z (or it is hoped that it will), depending on thecase.
Wrights first requirement says that S does Z, if doing Z is Ss function. It
is unclear if Wright intended this to be tenseless too, but he intended it to
deal with vestigial loss of function. It will, for instance, preclude the emus
vestigial wings from having the function to fly if it takes the present tense.
An emus vestigial wings are there (where they are and to some extent in

[50] The Philosophy of Philip Kitcher

the form that they are in) because the wings of the emus ancestors assisted
in flight (and drift has not yet fully dismantled them). But no emus wings
now assist in flight. It is less clear if all vestigial traits will count as vestigial
if the present tense is optional in the first requirement. In that case we
would need to know how traits are typed for the purpose of ascribing func-
tions to them (something that Wright does not discuss). We would need
an account of why the emus forelimbs, for instance, do not count as the
same type of trait as the forelimbs of flighted ancestors, for the purpose of
ascribing functions to them.
Actually, Wright does not specify whether S stands for tokens or traits
of a type. One might think that S should stand for traits of a type (i.e., for
Ss). It is hard to see how the formula can work for functions that derive
from natural selection otherwise, and Wright intends it to. Natural selec-
tion does not select tokens, and you do not have eyes in your head because
they allow you to see but because eyes in the heads of your ancestors
allowed them to see. But there are one-of-a-kind artifacts. Plus users might
bestow idiosyncratic functions on a generic artifact. If Ibuy a teapot to use
to water my plants and use it to water my plants, it acquires a function to
water plants because it is where it is (in my house) for plant watering.
Wrights formula is also problematic with respect to malfunction,
because it says that traits do not have functions if they do not and cannot
perform them. This issue is ameliorated if S stands for traits of a type. Then
there is space for some tokens to have functions that they cannot perform,
consistent with the first requirement. But this leaves us with the question
of how many instances of S must do Z:most, many, some, one?10 Wright
does not answer this question. In his very brief comment on this issue,
Wright does not appeal to the fact that S might stand for Ss (plural). He
instead suggests that we use a different tone of voice to attribute a function
to a trait that it cannot perform.
Kitcher reminds us that Wrights analysis suffers from certain counter-
examples that were introduced by Christopher Boorse (1976). The counter-
example that Kitcher describes involves a solitary scientist, working with
some equipment that passes gaseous chloroform through a pipe. A hole
accidentally forms in the pipe and lets the gas escape, which knocks the sci-
entist unconscious. Wrights formula entails that the hole in the pipe has
the function to let the gas escape because (a)the hole lets the gas escape

10. Ifirst defended a pure etiological analysis, which dropped the first requirement
and explicitly appealed to selection, in a widely circulated paper first presented to the
New Zealand Division of the AAP in 1980 as well as in my PhD dissertation, submitted
in 1983. See Neander (1991) for a more easily accessible early version.

K i t c h e r s T w o De s i g n S ta n c e s [51]
and (b)the hole remains in the pipe (where it is and in the form that it is in)
because the gas knocks out the scientist, who then cannot repair thehole.
A second counterexample from Boorse is of a man who kicks a dog,
intending to break its leg and cause it pain. If the man succeeds in his aim,
Wrights analysis entails that the break has the function to cause the dog
pain, for (a)it causes the dog pain and (b)it is there because it causes the
dog pain (that is, it was put there in order to dothat).
More recent etiological analyses usually drop Wrights first problematic
requirement (and seek other ways to handle vestiges). Plus they usually
require that the relevant history involve selection. The function of some-
thing is what an entity was (or entities of the type were) selected for doing.
This handles Boorses first counterexample, because the hole in the pipe was
not selected for leaking gas. More recent etiological analyses also usually
recognize that the details of the analysis need to be elucidated differently
for different kinds of selection. Most basically, while intentional selection
grounds artifact functions, it does not ground nonartifact functions. The
break in the dogs leg was intentionally selected to cause the dog pain, and
so it can have an artifact function to do so, on this type of analysis. But the
break in the dogs leg does not have a nonartifact function to cause the dog
pain, for there was no natural process of selecting leg breaks in dogs for
causing thempain.
Wrights aim to provide a univocal analysis for artifact functions and bio-
logical functions was abandoned. But even setting aside its other problems,
Wrights formula only anyway described a soft-focus unity of conception.
That is, it is vague or underspecified in certain ways that leave us with wriggle
room to fill out the details in different ways, as required for different kinds of
functions. The tenselessness of the formula is one instance of this. Its failure
to specify whether S stands for a token or for traits of a type is another. The
appeal to different tones of voice is yet another. In other words, recent etio-
logical analyses are explicitly acknowledging the disunity that was already
implicit in Wrights treatment. Recent etiological analyses can also equally
well accommodate a soft-focused unity of conception. All we need to do is
look less closely at the details. Whether artifact or nonartifact, an entitys
function is what it was (or entities of the type were) selected fordoing.
Further, there is no need for more unity of conception than this.
Scientists often refine everyday notions when they put them to use in spe-
cialized contexts. Scientists also often refine their technical terms in the
light of revisions to background theories. They can, moreover, be expected
to do so when the revisions are radical and the background theories are
central, in the way that the Darwinian Revolution was to the biologists
understanding of the functions of organismic traits.

[52] The Philosophy of Philip Kitcher

Note too that, while Wright was writing in a tradition of ordinary lan-
guage philosophy, the philosophical debate moved beyond this, to an inter-
est in the actual nature of functions and the theoretical role that function
ascriptions play in biology and elsewhere. Ireturn to the theoretical role
of function attributions in biology later. On the actual nature of functions
what we need to note is that, if functions are grounded in etiology, and if
the relevant etiology is subject to scientific investigation, the ontological
grounding of the functions is also subject to scientific investigation. Given
that Gods intentions are not in fact responsible for biological adaptations,
their functions are not in fact grounded in Gods intentions, no matter if
they were once thought to be. So if one wants to know what biological func-
tions are, it is not directly relevant that the creationists believed them to
be the product of Divine design. William Harvey, who died in 1657, did
not have natural selection in mind when he discovered that the function
of the heart is to circulate blood. However, this is consistent with todays
biologists having natural selection rather than Divine Design in mind when
they ascribe functions. The conception of biological functions can change
over time, even though the actual nature of biological functions has not
In closing this section it is worth repeating that the friends of etiologi-
cal analyses of functions have long recognized function pluralism of some
sort. Wright allowed that there are different senses of function, and Ihave
always taken this as a given in my own work on the topic. Focusing on the
etiological notion of function, my view is that different precisifications are
needed in different contexts, though there is a soft-focus unity of concep-
tion. People have employed somewhat different conceptions of functions
in different historical periods, but this is to be expected because they have
had different understandings of the sources of design in nature. Plus the
selection processes on which artifact and nonartifact functions depend are
different in significant ways.12


While Kitcher thinks that etiological theorists have lost some valuable
unity of conception, he also seems to be pressing for more pluralism than

11. See Neander (1991) for a more developed version of this argument.
12. Not only do we tend to allow that functions can be grounded in intentional selec-
tion or natural selection, but we also generally allow that natural processes of selection
might include antibody selection, cultural selection, and some learning processes.

K i t c h e r s T w o De s i g n S ta n c e s [53]
was conceded in the previous section. He claims that two notions of func-
tion are used in ascriptions of the form S has the function to do Z, one
of which is a strong etiological notion, which he takes to be the one that
proponents of etiological theories support (I question this later), and a
less demanding notion. In relation to the less demanding notion, Kitcher
claims that an indirect connection to design suffices. But just how indirect
does he think it can be? Kitcher first introduces the less demanding notion
in relation to intentional selection.
Here is his leading example:You are building a machine for a purpose.
Lets say that you are building it to ferry people across a river. You combine
certain parts:a couple of cables, a barge that is to be pulled across on the
cables, various parts of a motor to do the pulling, and so on. These parts
all have jobs that you explicitly intend them to do. In addition you acciden-
tally drop a screw into the machines inner workings, where it luckily lodges
between two parts and makes a connection without which the machine
would not work. You never learn about the dropped screw. You do not know
that it is in the machine, that it makes the needed connection, or that such
a connection is needed. The screw was not put in place originally, nor left in
place later, with the intention that it make such a connection. Kitcher says
that the screw has the function to make the connection in the machine.
Unfortunately there are two ways to read his claim, because the example
can be further elaborated in two different ways. For the first reading, sup-
pose that the screw accidentally falls into a screw thread and works its way
in, screwing the two parts together. On an etiological analysis the screw
has the function to screw two parts together, since it was designed to do
this and it retains its original design function while lodged in the machine.
Moreover it performs this function in the machine, even though no one
intended it to perform it in this particular machine.13 If this is what Kitcher
has in mind by an indirect connection to design, the connection to design
is not much loosened.
For the second reading, suppose that the screw gets stuck between two
parts without working its way into a screw thread or screwing two parts
together. Instead it gets wedged between two parts, and it makes an elec-
trical connection. In this case, that it is a screw is irrelevant. On this read-
ing Kitchers claim is that the screw has the function to make the electrical

13. Had the screw not made the connection, the machine would not have worked. So
if the screw had not made the connection, the machine might have been dismantled
and rebuilt or been further worked upon. So maybe the screw is there because it makes
the connection. It is therefore unclear, on Wrights analysis (which does not require
explicit intention), whether the screw has a function in the machine.

[54] The Philosophy of Philip Kitcher

connection in the machine, even though it was in no way intended to make
an electrical connection, either in this machine or in any other. On this
reading Kitchers claim is that a function of a part of a system can be an
entirely accidental contribution to a system doing what the system was
designed (if poorly)todo.
That Kitcher does not specify whether or not the screw acts as a screw
in the machine suggests the first reading, since it seems an obvious way to
understand the example. But the second reading seems more in tune with
Kitchers general comment on the case. For instance, when he says, The
function here is grounded in the contribution that is made towards the per-
formance of the whole machine and in the link between the performance
and the explicit intentions of the designer (1993c, 381), he does not men-
tion the screw having been designed.
Even on the second reading of Kitchers less demanding notion, one
might wonder if there is a significant disagreement between him and the
proponents of apparently less pluralist views, who allow that the screw
serves the function of making the electrical connection and that it functions
as an electrical conductor, even though it does not have the function to do
so. On the second reading Kitcher says that the screw has the function to
make the electrical connection, although it was not designed to. But is this
just a slight terminological variant? It would be if it were not for the sub-
stantial claim that Kitcher makes concerning the theoretical role played by
the less demanding notion. Iconsider this later; for now Iam only trying to
make sure Iunderstand his less demanding notion.14


Turning to pre-Darwinian creationist biology, Kitcher (1993c, 381)tells us

that some biologists might have thought that the Creator had a grand design
into which all of the details were explicitly intended to fit, while others
might have thought that the Creators plans were in part achieved through
secondary causes that endowed organisms with abilities to respond to
their needs. These biologists might have believed, for example,that:

the Creator intended that jackrabbits should have the ability to thrive in desert
environments... [but] that there was no explicit intention about the length of

14. The issue of how direct the connection between function and design is, in our
everyday conception of the functions of artifacts, should be settled with the help of
psychological investigations. See, for example, Matan and Carey2001.

K i t c h e r s T w o De s i g n S ta n c e s [55]
jackrabbits ears. Yet, because the length of the ears contributes to the mainte-
nance of roughly constant body temperature, and because this is a necessary
condition of the organisms flourishing (which is an explicitly intended effect)
the length of the ears has the function of helping in thermoregulation.(381)

The length of the jackrabbits ears is meant to be analogous to the dropped

screw. Unfortunately there are also two ways to read this example. Kitcher
describes these biologists as thinking that God put in place secondary causes
that give creatures the ability to respond to their needs. But what are these
secondary causes? Apparently they are thought to further adapt the details
of the creature to its environment. Perhaps these biologists think that God
arranged for the (Lamarckian) inheritance of adaptive acquired character-
istics, or that God arranged for natural selection to make limited modifica-
tions to a species design, or that God arranged for further adaptation to
occur through a process that is as yet unknown. But these biologists do not
seem to think that the length of the jackrabbits ears is entirely accidental.
So, again, there are two possible readings. On one, the less demand-
ing notion of function only slightly loosens the link between function
and design. Here it is only the link between the function of a part and
the explicit and advance intentions of God that is indirect. Some other
source of design is still involved. (Does Lamarckian inheritance of adap-
tive acquired characteristics count as a design process? Iam not sure how
to answer this, absent a general characterization of a design process. But it
seems to me that it should count, especially if God put it in place to further
adapt the features of organisms.)15 On a second way of reading Kitcher he
did not intend this. On the second reading, the possible secondary causes
need not be a design (or a selection) process.


Kitcher contrasts his less demanding notion of function with a strong etio-
logical one. Doesnt this exclude an etiological reading of his less demand-
ing notion? It would, except that Kitchers strong etiological notion is

15. Here is another question of interpretation. On Kitchers proposal all creationists

think that God had the explicit and advance intention that organisms thrive in their
habitats, but some think that certain detailed features (such as the length of the jack-
rabbits ears) could derive their functions indirectly from Gods intentions. But what
of middle-level features (such as the jackrabbits ears)? Do they require direct design,

[56] The Philosophy of Philip Kitcher

ultra-strong, as Iexplain in this section. This is also important for assess-
ing his claim that the strong etiological notion is too onerous for certain
uses in biology. Kitchers strong etiological notion is much more demand-
ing than any notion of function that Ihave had in mind in developing and
defending an etiological theory. However, he raises some important issues
that have been sorely neglected (and were even more so at the time). Icom-
pletely agree with Kitcher that his ultra-strong etiological notion is too
onerous, but Ishall argue that this leaves us with a middling-strong notion
that needs to be considered.
Kitcher tells us that the proponent of the etiological theory needs to
provide some answers, and that, once the answers have been supplied, the
full burden of the etiological notion of function becomes evident. In rela-
tion to this he identifies two main issues:(1)the time at which the envis-
aged selection must take place and (2) the relevant alternatives to that
which is selected. It is in discussing the second issue that the ultra-strong
nature of Kitchers etiological notion is revealed, but we need to take them
one at atime.
The first issue has to do with vestiges and exaptations. Feathers are
thought to have originated in flightless dinosaurs due to a role in ther-
moregulation. In modern birds feathers often have other functions. Some
have a function in flight, in camouflage, and/or in attracting mates. There
are also modern flightless birds, such as emus and penguins, which had
ancestors that flew. The feathers of these birds no longer have a function
in flight. How does an etiological theory of functions account for such
changes in function over time? (I here assume that Wrights first require-
ment has been dropped.)
Kitcher poses two questions in relation to this. The first is whether
an etiological analysis should require that the relevant selection explain
(a)the origination of an entity with a function, (b)its subsequent mainte-
nance, or (c)both. Is the relevant selection responsible for first fixing the
trait in a population, or for later maintaining it there (against deleterious
alterations), or both? In Kitchers view the best answer is(b).
I prefer a different option:(d)both directed selection and maintenance
selection might be relevant, as long as they are relatively recent.16 In any

16. There is another reason someone might want to privilege maintenance selec-
tion. One might raise Cumminss (2002) objection that selection is too fine-grained to
ground (appropriate) etiological functions. Cummins argues that, in order for wings
to have a function in flight, on an etiological analysis, there must have been a popula-
tion in which individuals had wings that allowed flight while other individuals lacked
wings that allowed flight. Cummins says that there was, rather, a gradual evolution of
wings, during which variants were selected for more effective flight or a different kind

K i t c h e r s T w o De s i g n S ta n c e s [57]
event it is not necessary to privilege maintenance selection in order to
accommodate vestiges and exaptations. True, a modern traits function
need not be the same as that for which ancient homologues were selected.
But think of the feathers of the flightless dinosaurs. Selection spread muta-
tions in the population that enhanced the ability of proto-feathers and
early feathers to contribute to thermoregulation. Their function in thermo-
regulation was grounded in this directed selection (as well as in any main-
tenance selection for them that was occurring at thetime).
We can handle changes in functions over time with a careful answer
to Kitchers second question, without restricting the relevant selection to
maintenance selection. The second question asks if the function-conferring
selection for a trait S of an organism O is (a)in the recent past, (b)in the
present, or (c)in the recent past and the present (relative to when O lives
and S comes into existence). Kitcher suggests that (b)is the popular choice
for biologists. The friends of the etiological theory generally choose (a)for
the reasons lucidly explained by Peter Godfrey-Smith (1994). I choose a
similar but longer answer that avoids the use of the vague term recent and
further obviates the need to privilege maintenance selection.17
Suppose we want to know if a feather has a function to assist in flight. In
principle the relevant history concerns the relevant lineage of historical homo-
logues. Clearly it matters if the feather belonged to a preflight dinosaur, an
emu, or an eagle, and it also matters what kind of feather it is (for example,
if it is a long tail feather as opposed to a down feather in either an adult or
a chick, since down is for insulation). Next the question to ask about this
lineage is if selection for assisting in flight operated on this lineage and, if so, if it
began prior to and continued up until the feather in question arrives on thescene.

of flight over other variants that flew less effectively or flew differently. One response
to this is that it ignores maintenance selection. Even when flight-enabling wings have
gone to fixation, highly deleterious mutations would undermine a capacity for flight
if not weeded out. This is a good reason not to ignore the importance of maintenance
selection, but it is not a reason to exclude directed selection from grounding functions.
Cumminss challenge deserves a longer discussion, but it seems to me to trade on shift-
ing inappropriately between coarse-grained and more fine-grained ways of speaking.
When we speak of wings having the function to enable flight, we speak in a very coarse-
grained way. Coarsely speaking, selection for flying more efficiently or selection for
flying in a different way is still selection for flying.
17. This follows Neander and Rosenberg (2012). Their proposal does not eliminate all
vagueness, but Ithink the remaining vagueness is unproblematic since it corresponds
to a plausible vagueness in functions. It is vague what counts as a cessation in selec-
tion. For how long must it cease? Aday or a season is too short. Are several genera-
tions enough? There is no sharp cut-off, but nor need there be. There is also vagueness
with respect to gradual transitions in selection for (say) leaping or gliding to flying.
However, this vagueness seems appropriate.

[58] The Philosophy of Philip Kitcher

If the feather is from a preflight dinosaur, it has no function in flight
because there was no selection for flight operating on the relevant lineage
before the feather came into existence. If the feather is from an emu, it
has no function in flight because, while selection for flight operated on
the lineage before the feather in question, it ceased before it and its emu
came into existence. If the feather is a long tail feather in an adult eagle (a
rectrix), it has a function in flight because selection for assisting in flight
(more specifically for generating thrust and lift) began and did not cease
in the relevant lineage (of rectrix feathers in eagles) before it came into
Let me stress that, in my view, this is how etiological functions are indi-
viduated in principle. In practice biologists spend most of their time exam-
ining structures and observing use and so on, as Kitcher comments. But
there is a big difference between what makes a thing a certain kind of thing
(ontologically speaking) and how we figure out if it is that kind of thing in
None of us gets out our chemistry set each morning to check for H2O
before drinking a glass of water. Even in an experiment, when care is para-
mount, a scientist might rely on the look of the liquid and on its source (the
faucet or the labeled bottle). This is consistent with water being necessarily
composed of H2O. People, including scientists, often recognize things on
the basis of their more easily accessible properties, even when less easily
accessible properties are considered essential. Afield biologist can be sure
that a platypus is a platypus on the basis of its superficial appearance. But
this does not begin to settle the question of whether species membership is
historical. Psychological studies show that people prefer to categorize and
label an artifact on the basis of their beliefs about its original intended use.
But we do not observe the intentions of the artifacts creator (unless we
make it ourselves). Iam sure that the cups in my cupboard were designed
for drinking hot beverages, but Idont even know who designedthem.
In short there is a difference between individuation and the criteria
employed, in practice, in identification. The etiological theory of functions
is a theory about individuation. So the fact that the biologists are usually
not discussing evolutionary questions when, in practice, they figure out
somethings function is far from decisive in favor of function pluralism.
This is not to deny that the etiological notion of function is onerous. It
is onerous. But biologists use a lot of onerous notions (of species, clades,
homologues, and so on). The question is if it is too onerous for the theoreti-
cal work that it is supposed to do or if it is onerous enough for it. This is a
large question that cannot be quickly settled. But Kitchers main argument
for the etiological notion being too onerous is not that biologists do not

K i t c h e r s T w o De s i g n S ta n c e s [59]
seem bothered about its burden in discussing functions (although his argu-
ment is partly that) but that it is too onerous for some purposes, for reasons
that Idiscuss in the next section.
Before that we need to consider the second issue that Kitcher raises for
the proponent of the etiological theory:the issue of the alternatives. The
full burden of the etiological notion of function is revealed, Kitcher says,
when we consider the alternatives with respect to which an entity with a
function must have been selected. But this is where it becomes apparent
that Kitcher (1993c, 383)has an ultra-strong notion in mind. Here is a key

Recognizing a trait, structure, or behavior of an organism as responding to a

selection pressure imposed by the environment (in the context of other features
of the organism that are viewed as inaccessible to modification without severe
loss of fitness) we do not necessarily commit ourselves to claiming that the
entity in question originated by selection or that it is maintained by selection.
For it may be that genetic variation in the population allows for alternatives that
would be selectively advantageous but are fortuitously absent. Thus the entity
is a response to a genuine demand imposed on the organism by the environ-
ment even though selection cannot be invoked to explain why it, rather than
the alternative, is present. In effect, it is the analogue of the luckily placed screw,
answering to a real need, but not itself the product of design.

Kitcher is saying that selection cannot necessarily be invoked to explain

why an entity is present, even if its presence is a response to selection
pressure. Iwould have thought that, if one state of affairs is a response to
another, the second is causally relevant to the first. But Itake Kitcher to
mean that a trait can be adaptive, and in that sense a response to selec-
tion pressure, without its presence being due to selection.
Under what circumstances can selection be invoked to explain why an
entity is present, in Kitchers view? His answer is that, in order for S to count
as having been selected for Z-ing, no more advantageous alternatives could
have arisen (given relevant constraints) that failed to arise or that did arise
but were eliminated by drift. Lets highlight this by calling it the Ultra-Strong
Selection Requirement for the ultra-strong etiological notion of function.
My own answer is that something much less demanding suffices. The
entity must have been selected against available alternatives, understood
to be the ones that arose and were not eliminated by nonselective forces
such as drift. Suppose that the trait in question is S3, and S3 does Z, which
meets some demand that the environment imposes on the organism. Also
suppose that there are more advantageous alternatives (S1 and S2) that

[60] The Philosophy of Philip Kitcher

might arise given relevant constraints. Further suppose that S1 never arises
(a mutation needed for it does not occur). Suppose too that S2 arises, but it
is eliminated by drift while its representation in the population is small. S3
is not selected over S1 or S2. We can all agree onthat.
But now suppose that, before S3 arose, S4 had been selected and gone to
fixation. Suppose too that there are some other less fit or deleterious alter-
natives (S5S10) that were not eliminated by drift. Then S3 can be selected
over the alternatives S4S10. Given these realistic suppositions, selection
cannot be invoked to explain why S3 rather than S1 or S2 is present or preva-
lent. Yet selection can be invoked to explain why S3 rather than S4S10 is
present or prevalent. If S3 was, in the latter sense, selected over other avail-
able alternatives because it did Z, then it was selected for doingZ.
I am in complete agreement with Kitcher that his ultra-strong etiologi-
cal notion of function is extremely onerous. But there is a middling-strong
notion that does not require satisfaction of the Ultra-Strong Selection
Requirement. For the middling-strong notion, S has the function to do Z
if its doing Z was (at least in part, since S might have other functions too)
responsible for Ss selection over available alternatives, these being ones
that arose and were not eliminated by nonselective forces such asdrift.


For most of Kitchers paper, his comments on his less demanding notion
of function seem to cast it as an etiological one, or as ambiguous in this
respect. At least on one reading, the dropped screw was doing what it was
designed to do (screwing two parts together). And, at least on one read-
ing the length of the jackrabbits ears was either thought by creationists
to result from a secondary source of design, or from Gods explicit and
advance intention regarding their length. This reading is also supported
by Kitchers Ultra-Strong Selection Requirement for his more demanding
notion of function. If this requirement is invoked in the case of the more
demanding notion, there is room for the less demanding notion to still
be etiological. The main difference between Kitcher and me, in that case,
would be with respect to whether the ultra-strong etiological notion is at
all useful. Idoubt thatitis.
However, when Kitcher gives a more general characterization of his
less demanding notion, the etiology seems to disappear, except insofar as
Kitcher clearly requires that the system as a whole must be designed (to
ferry people across the river, to thrive in a desert environment, and so on).
This is clearest in the comparison that Kitcher draws toward the end of

K i t c h e r s T w o De s i g n S ta n c e s [61]
Function and Design between his less demanding notion of function and
the one defined by Cummins (1975).
As mentioned earlier, Cummins claims that the (or a) function of a part
of a system is what it contributes to a complexly achieved capacity Z* of a
containing system, where Z* is a capacity that a researcher is interested in
explaining by means of a functional analysis, otherwise known as a mecha-
nistic or an operational explanation. In such an explanation, diverse func-
tions (Z1 Zn) are ascribed to the diverse parts (S1 Sn) of a system,
and it is the performance of these functions (in the appropriate spatial and
temporal order) that explains the systems abilitytoZ*.
Kitcher suggests that the relevant complexly achieved Z* capacity is
one for which the system is designed. When contemporary biologists are
explaining how organisms operate or function, the relevant Z* capacity is
one for which natural selection designed the organism. In general they are
designed to survive and reproduce in the environment in which they are
found, within the constraints imposed by hard-to-change features of their
Bauplan. This is the sole modification that Kitcher mentions making to
Cumminss account of the weaker notion of function. Thus he here seems
to come down decisively in favor of there being no need for the parts of
the organisms to have been selected for their functions. They need only
be responses to selection pressures insofar as they happen to be adaptive.
The modification to Cumminss account is a major improvement. For
one thing, Cumminss proposal is too liberal because it implies that some
entities that lack functions have functions. For example, it implies that the
various entities that contributed to the formation of the solar system had
the function to contribute to its formation. It implies that stars have the
function to send heavy elements into the interstellar medium to help form
vast clouds of molecular matter, that preplanetary clumps of matter have
the function to collide and accrete into larger clumps, and that various dif-
ferent elements and compounds have various functions to behave in the
ways they needed to behave if solar systems like ours are to form. Those
who defend Cumminss proposal might contend that such a system is insuf-
ficiently complex, but it is quite complex.18 Kitchers revised notion does
not suffer from this problem.

18.One might also wonder if the simplicity of some artifacts is a problem for
Cumminss proposal: cups, spoons, and doorstops have functions. Cummins can
respond that they are parts of larger systems that include their users. This puts a lot
of weight on how systems are demarcated. On Kitchers analysis functions belong to
parts of systems only if the systems were designed. So this helps. It takes the weight
off the need for complexity and the delineation of systems in determining which parts
have functions and whichdont.

[62] The Philosophy of Philip Kitcher

Cumminss analysis is also problematic because it implies that contribu-
tions to complex pathological processes are functions. A contribution to
the growth of a tumor, to the thickening of the heart muscle in cardiac
hypertrophy, or to the dementia that results from untreated tertiary-stage
syphilis is a function, on Cumminss analysis, just in case a researcher is
interested in explaining the process. Some contributions to pathological
processes are instances of normal or proper functioning. Many normal
functions are involved in cell replication, even in a cancerous growth, for
instance. But a part of an organism does not have a function in virtue of con-
tributing to a pathological process that a biologist is interested in explain-
ing. As Kitcher points out, this problem is neatly avoided by his proposal.
Organisms are designed to survive and reproduce, not to die of cancer or
suffer from cardiac hypertrophy or syphilitic dementia.
Is this a fair criticism of Cumminss analysis in the context of function
pluralism? There is a notion of function that respects the function versus
nonfunction distinction and the function versus dysfunction distinction.
And Cumminss analysis does not respect these distinctions. But if we are
now to assess his analysis as an analysis of one notion of function among
others, this changes the dialectic. They cannot all respect all of the same
distinctions, or there would be just one notion of function. We need an
independent principle for determining which distinctions each notion of
function must respect. In advance it will be controversial which notion of
function performs which theoretical role. So the best strategy here is to
turn to the theoretical role and work back from it, trying to decide which
distinctions are needed. Kitcher does this when he argues that his ultra-
strong notion of function is too onerous for the operational explanations
that the physiologists attempt to provide. Fair enough. But which distinc-
tions must the less demanding notion respect for the purposes of these
There are generally thought to be two main types of functional expla-
nation. One is the type that Wright emphasizes. In such an explanation
a function is ascribed to an entity, and the form or presence of the entity
is explained by its function. For example, if the function of the rectrix
feather in the eagle is to provide thrust and lift, its providing thrust and
lift explains why eagles have rectrix feathers. In my view the ultra-strong
etiological notion is too onerous even for this explanatory task, but the
middling-strong etiological notion of function can play thisrole.
The second kind of functional explanation is the one that Cummins
emphasizes:operational explanations of complex systems in terms of the
diverse functions of the parts. Cummins (1975) is considered the classic
statement of the importance of this explanatory role for functions. But

K i t c h e r s T w o De s i g n S ta n c e s [63]
we need to separate his claim concerning how the functions are analyzed
from his claim concerning the importance of this explanatory role. It is
apparently tempting to bundle them together. For example, in introducing
the etiological account and a Cummins-style causal-role account, Philippe
Huneman (2013, 2)says:

Both acknowledge that function is a concept used in some explanations, but

they diverge from the first step because the etiological account thinks that the
function of X being Y explains the presence of X whereas, for the causal role
theorist, the function of X being Y explains or contributes to an explanation of
the general proper activity of a system which includesX.

This is near enough right as a comment on Wrights and Cumminss work

(although Cummins recognized the legitimacy of the first kind of expla-
nation in the artifact case). But it is important, as I say, to separate the
analyses of the notions of function from the claims concerning the theoreti-
cal roles that these notions play. Ithink that a middling-strong etiological
notion is employed in explaining the general proper activity of a system.
So Iagree with Cummins and Kitcher that function ascriptions play an
important role in operational explanations in physiology. But the key ques-
tion is:Which notion(s) of function is (are) used in them? Is it an entirely
ahistorical one, as Cummins claims? Does history enter into the picture only
insofar as the organism as a whole is designed by natural selection to survive
and reproduce, as Kitcher claims (on the second reading of his view)? Or does
history enter the picture for each function of each part, as in myview?
Kitcher agrees with Cummins that the strong etiological notion of
function is not well suited for this type of explanation. He argues that biol-
ogists are not, and do not need to be, concerned with the burden of the
ultra-strong etiological notion when they give this kind of explanation,
except when they have evolutionary questions specifically in mind. Again
Iagree. But this leaves open the possibility that a middling-strong etiologi-
cal notion is employed in explaining how living systems operate. In Kitchers
view this type of explanation makes use of the function versus dysfunction
distinction.19 And, again Iagree. It does. Physiologists give a central place
to describing properly functioning systems. So one question we need to ask
is:Which analysis of function attributions best captures this distinction?
That physiologists ascribe functions that respect the function versus
dysfunction distinction might seem puzzling at first. An entitys having

19. For a more extended treatment of the points made here, see Neander (2015).

[64] The Philosophy of Philip Kitcher

a normal or proper function is not a causally potent property of it. Bobs
heart has the function to pump blood, but this will not save him if he is
suffering cardiac arrest. If his heart cannot pump blood or cannot pump it
well enough, then he will die, no matter what his heart would do if it were
functioning properly. Only actual blood pumping will save his life. If physi-
ologists want to explain physiological processes and outcomes, why dont
they restrict their attention to causally efficacious properties of parts?20
The puzzle dissolves when we appreciate that physiologists face a gen-
eralization problem. One of their tasks is to give useful general descrip-
tions of kinds or types of living systems. And abstracting from pathology,
by describing systems that are functioning normally or properly, contrib-
utes to doing so. This type of idealization abstracts away from some of the
variation between individuals rather than from the details of systems of
the kind or type. That is, a description of a normally or properly function-
ing system (such as the normal human circulatory system or the normal
human visual system) can be as detailed as we want, up to the point where
physics takesover.
Many scientists face generalization problems. But the problem for the
physiologists is especially profound given the stupendous complexity
of living systems as well as the mechanisms involved in inheritance and
development. The latter leads to sui generis genomes and fluid interac-
tions with the environment during maturation and thus to a great deal
of the potential variation being realized. The complexity compounds this.
The more complex a system of a given type, the more variables there are,
and so the more potential there is for variation between individuals (or in
the same individual over time). Some of this variation can be described.
Physiologists often describe individual cases, as well as normal polymor-
phism and pathological syndromes and so on. However, physiologists must
also find ways to usefully generalize. And they do so by idealizing, among
other things by abstracting away from pathological processes, describing
systems that are functioning normally or properly.
Kitcher is right that the mechanistic explanations given by physiolo-
gists make significant use of the function versus dysfunction distinction.
So his point against Cumminss analysis holds good, even in the context of
function pluralism. Nevertheless this leaves open the question of whether
the middling-strong etiological notion of function or a relatively ahistori-
cal notion of function best captures this distinction, as well as any other
distinctions used in such explanations.

20. See, for example, Kitcher 2003b, 302; Craver2007.

K i t c h e r s T w o De s i g n S ta n c e s [65]
Does (a)or (b)best capture the physiologists notion of the system that
is functioning normally or properly?

(a) Asystem that functions as designed, with each part able to do what it
was selected to do (along the lines of the middling-strong etiological
(b) A system that thrives anyhow, in its environment, within the con-
straints imposed by general hard-to-change features of the organisms

My answer is (a) and not (b). But neither Kitchers arguments nor mine
here settle this question.
If malfunction is possible, S has the function to do Z is compatible with
S lacks the capacity to do Z, if S stands for a token. If we use an etiological
notion, a token trait can malfunction because its function does not depend
on its own capacities but on what its historical homologues were selected
for. Kitchers analysis of his less demanding notion (on the less demand-
ing reading of it) seems to preclude the possibility of malfunction. Like
Cumminss analysis, it seems to entail that the functions of token traits are
among the traits actual causal capacities. (Readers will recall that a similar
problem arose in relation to Wrights analysis.)
However, if Kitcher allows S to stand for traits of a type (i.e., for Ss),
then there is room for individual instances to malfunction.21 But even so
there is a problem. Now the problem is that the analysis is underspecified.
We need to know (i)how many current tokens of a given type of trait must
have the adaptive capacity in question, (ii) how traits are typed for this
purpose, and (iii) in what environment(s) the relevant traits must be adap-
tive. (It cannot simply be whichever environment an organism is currently
found in; at any rate that does not seem to be the physiologists notion
of normal or proper function.) If the analysis is also supposed to apply to
artifacts, we will need to know how one-of-a-kind artifacts can still have
idiosyncratic functions with respect to which they might malfunction, and
how generic artifacts (such as accidentally dropped screws) are now to be
treated. Until the analysis is further specified, it cannot be fully assessed.
I have long believed that serious attempts to answer questions along
these lines push us toward an etiological notion of function and to dif-
ferent precisification for artifacts and the natural functions of parts and
processes in organisms. In my view, in relation to (i), the normal or proper

21. Godfrey-Smith (1993) makes thispoint.

[66] The Philosophy of Philip Kitcher

function of a token trait does not depend on its present capacities, or
on the present capacities of any traits of the type, but on the capacities
for which historical homologues were selected. In relation to (ii), Imain-
tain that traits are initially typed in lineages of homologues, and then
changes in the selection pressures operating on the lineage (for example, a
change in selection for flight to selection for paddling) permits more fine-
grained classification. In relation to (iii), Ibelieve that the performance of
a traits function need only have been adaptive in the past environment
in which the trait was recently selected for performing it. Each answer
needs discussion, and there is no fast way forward to a convincing conclu-
sion. But Igive more detailed development and defense of the etiological
theory of functions in other places, and interested readers might turn to


Kitchers pluralist theory of functions describes a so-called strong etio-

logical notion of function that is ultra-strong. He is right that it is too
onerous to be of use in the operational explanations that physiologists
provide. In my view it is too onerous to have a significant explanatory role
of any kind in contemporary biology. However, there is a middling-strong
etiological notion. Using this notion, the function of an entity is what
it was (or what traits of the type were) selected for, but selection need
be only a partial explanation of the traits presence or distribution in a
population. For much of Kitchers discussion it is unclear whether this
is the less demanding notion of function that he has in mind. However,
his more general characterization of his less demanding notion sug-
gests a different interpretation. It suggests that he is arguing for a role
for functions that need only contribute to a complexly achieved capacity
for which the system as a whole was designed. Iagree with Kitcher that
the operational explanations of the physiologists do not use his ultra-
strong etiological notion. Ialso agree with his important point that they
use a notion that respects the function versus dysfunction distinction.
However, Kitchers argument leaves open whether this is a middling-
strong etiological notion. For one thing, it is unclear whether his less
demanding notion of function adequately respects this distinction, since
it is underspecified with respect to the possibility of malfunction. In the
absence of a more fully specified alternative, the burden of the middling-
strong etiological notion might be justified by the need to capture that

K i t c h e r s T w o De s i g n S ta n c e s [67]
Reply toNeander

No contemporary philosopher has thought harder (or with more sophis-

tication) than Karen Neander about the concept of biological function. In
her essay Karen further articulates the etiological approach to functions
that has been a signal contribution to the philosophy of biology and poses
some important challenges to my own rival proposal. Ill begin by outlin-
ing the general motivation for my efforts at synthesizing the merits of
the etiological conception and its main rival (the decompositional notion
defended by Robert Cummins). Ill then address the ambiguities Karen
discerns in my account and explain why Icontinue to resist the call of pure
My essay Function and Design (1993c) was written in the wake of a
decade of controversies sparked by the publication of Gould and Lewontins
(1979) seminal paper. At the core of their critique of adaptationism was a
clear recognition of the extensive work needed if claims about the opera-
tion of natural selection are to be properly grounded. My own work on evo-
lutionary studies of animal behavior and of the differences between the
careful investigation of some aspects of nonhuman behavior and the casual
speculations about supposedly hard- wired human propensities had
echoed the same theme. In my 1985 book Icontrasted Geoffrey Parkers
meticulous studies of dung flies with headline-grabbing claims about xeno-
phobia and female coyness. Work like Parkers demonstrated, Ibelieved
(and still believe), the burdens assumed in supposing that a particular
organ, system, structure, or behavior was selected because it conferred a
specific benefit.
Whether or not Dobzhansky was correct to declare that nothing in biol-
ogy makes sense except in the light of evolution, it was already true in
the late 1980s, and is even more apparent today, that the vast majority
of the worlds biologists are not actively engaged in evolutionary studies.
Yet throughout molecular genetics, molecular cell biology, and molecu-
lar developmental biology, references to the functions of genes, proteins,
and organelles, as well as larger complexes and structures, abound. Were
the authors of these references really taking on the burden of committing
themselves to detailed ideas about the history of natural selection? Or were
they behaving as casually as the figures Ihad excoriatedwith what Inow
see as too much smart-alecky enthusiasmin my 1985 book? I found
it hard to answer either question affirmatively and, seeing etiological

[68] The Philosophy of Philip Kitcher

conceptions of function as committed to saying yes to one of them, sought
an alternative that would render the everyday practice of functional attri-
bution respectable.
Two further aspects of the work on animal behavior buttressed this
line of thought. Behavioral biologists routinely refer to Tinbergens classic
division of four important why-questions, in which the question of func-
tion is separated from the question of evolution. Even if the champions of
etiological approaches to function emphasize the recent history of selec-
tion, they effectively suppose that Tinbergens list contains two different
evolutionary questionsand if that was really part of his conception,
Iwondered (and wonder) why he didnt explicitly mention the important
kinship. Second, the most rigorous evolutionary studies of animal behavior
are typically circumspect in deploying optimality analyses (and thus in line
with some of the warnings issued by Gould and Lewontin):my 1993c paper
cites the work of Clutton-Brock and Albon on display behavior in red deer.
Iinterpreted these investigators as expressing caution about using attribu-
tions of function to infer that the focal trait emerged because, among the
available alternatives, it maximized reproductive success.
Inspired by these diagnostic thoughts, Iproposed to invert the standard
etiological picture. Instead of looking from the perspective of the organ-
ism and asking which past variants might have been selected, Itook the
view of the environment, which issues a fundamental Darwinian chal-
lenge: Reproduce as much as possible! To meet that challenge particular
organisms need to survive to the stage at which they reproduce and they
need sources of energy to keep going, and that starts a cascade of further,
specific requirements. The biologist looks at the organism in the here and
now, holding particular traits and structures fixed. She asks how the organ-
ism meets some need at some level of specificity. Mammals need food. The
environments of some mammals provide tough fibrous materials that they
can ingest. If they are to digest the plants, they require some ability to
break down cellulose, perhaps broad molars, good for grinding, or a gutful
of appropriate bacteria. In attributing a function, either to the dentition
or to the intestinal flora, the biologist traces the cascade of selective pres-
sures, as they descend from the fundamental challenge to reproduce, so
that the characteristic feature can be seen as a response by the mammals
under study to the problems theyface.
Although theres an evolutionary backdrop to the whole approach, no
specific hypothesis about a history of selection is presupposed. Asked
whether the teeth were selected for their ability to grind, the biologist may
simply shrug her shoulders, or even explain that, having pondered Gould

K i t c h e r s T w o De s i g n S ta n c e s [69]
and Lewontin, she has no interest in conjectures about the course of evo-
lutionary history. I look at the animals as they are now, she declares,
and Ifigure out the ways in which, given their major traits and environ-
ments, they face problems in surviving and reproducing. I then identify
the ways in which particular molecules, intracellular structures, organs,
and systems respond to those challenges. If she has read the philosophi-
cal discussions of the concept of biological function, she may add an ecu-
menical remark: So I agree with Cummins that my attributions rest on
causal analyses, based on information Ihave about the animals in the here-
and-now, but Cummins fails to see that the problem background arises
from Darwins fundamental ideas about lifefrom his hostile forces; the
Darwinian connection adds the idea that etiological theorists like Neander
highlight, but it does so in a less specific way, freeing me from any burden
of commitment to a particular style of selective history. With respect to
adaptationism, my conscience is clear.
Now to the details of Neanders critique. Consider first the dropped
screw. Neander rightly distinguishes two scenarios, one in which the screw
drops into a threaded holefunctioning as a screw!and one in which it
does something different (making an electrical connection). Ihad the sec-
ond of these in mind and should have been more explicit. On this second
scenario the placement of the screw was not selected, or even intentional;
its presence in the machine is a lucky accident. It has the function to make
the connection because the machine is supposed to do a certain job, and,
given major features of the structure, that requires an electrical connec-
tion between two separated bits of metal; without the fortunately placed
screw the connection would not be made. Ihope the parallel with the cas-
cade of more specific demands generated by Darwins fundamental chal-
lenge (Reproduce!) is evidentand Ialso claim that the screws function
has nothing to do with the history of its placement.
The second major ambiguity Neander rightly discerns in my account
concerns my remarks about the demands of selective history and the
extent to which natural selection has to pervade the history underlying
the emergence or maintenance of a function. This is a deep and important
critique, and the presentation in my 1993c paper is inadequate to meet it.
It now seems to me that the inadequacy stems from my having taken for
granted some morals Idrew from Gould and Lewontin that Ifailed to make
explicit. Inow hope to remedythat.
Let me begin with the line of thought most obvious in my 1993c paper.
If you contend that the function of X in Os is F because recent selection for
F explains the presence of X in current Os, you need to decide just how pow-
erful the role of selection must be:Under what styles of history involving

[70] The Philosophy of Philip Kitcher

selection can selection do enough to ground the attribution of function?
Neander recognizes two demanding conditions that might be placed on
etiological theories:

1. F confers a fitness advantage on bearers of X with respect to all the rivals

that could have been generated by heritable sources of variation.
2. F confers a fitness advantage on bearers of X with respect to all the actu-
ally arising rivals generated by heritable sources of variation.

I entirely agree with Neander that (1)is too strong. But what about(2)?
Theres an obvious tendency to think that (2)can fail, even though selec-
tion does a significant amount of work in explaining the contemporary
prevalence of X among the Os. Imagine that a few rivals to X with supe-
rior expected reproductive success actually emerged, but that these were
fortuitously eliminated from the population. With respect to most of the
alternatives, however, their inferior contributions to reproduction doomed
them. Here we have a scenario in which selection is largely responsible for
Xs success, even though chance helps out from time to time. (2) would
then be false, even though the role of selection in the history seems sub-
stantial enough to ground the attribution of the function. Recognition
of that motivates Neanders proposal of a middling-strong etiological
notion that doesnt live up to the demand imposed by(2).
But there are complications. Suppose chance figures differently in the
story. The genome of the Os has hotspots at which mutations would allow
for all sorts of superior alternatives to X. There are lots of them, and the
probability that a significant number of them would not appear is tiny. Yet
by a bizarre quirk of fate, none of the beneficial mutations ever occurs. This
is one of those occasions on which the vastly improbable happens. Instead
X beats out the (relatively few) inferior competitors that doarise.
This latter story deploys the notion of potential helpful mutations,
introduced in too simple a form in the overstrong (1). Yet although the
scenario satisfies (2), it still seems inadequate to explain the presence of
X as the result of selection. For a vital part of the story is the exceptional
luck X experienced:there were all these close possibilities that would have
dethroned X. We should no more explain Xs presence by citing a history of
selection than we should explain a childs passing an exam on the basis of
her deep mathematical knowledge, when the questions posed were drawn
from a list that includes large numbers of problems on topics about which
she knows nothing, but the actual draw generated only the simplest types
of arithmetical calculations.

K i t c h e r s T w o De s i g n S ta n c e s [71]
Thus (2) can sometimes be too strong and sometimes too weak. The
forms of the evolutionary histories prove crucial. Notice also that chance
can enter in both the modes considered so far. Amend the scenario about
the favorable mutations at the hotspots to allow some of them to emerge.
By coincidence, however, when those mutations arise they happen to
cluster in particular locales, and freak storms in these locales generate
catastropheshere an avalanche, there a number of falling trees, elsewhere
an incursion of predatorsand in every case these chance events elimi-
nate all the superior rivals. Once again the presence of X isnt explained by
recent selection but rather by the infrequent arising of beneficial mutants
and by their unfortunate bad luck on the few occasions on which such
mutants do appear.
I anticipate the response that the scenarios just envisaged are too con-
trived to dislodge a middling-strong etiological conception of the style
Neander favors. But they are only the entering wedge to appreciating the
complex combinations of evolutionary forces that can easily occur in the
history of a lineage. Central to the Gould-Lewontin critique of adaptation
is the thought that traits are bound together in packages. The genetic basis
for a focal trait gives rise in development to the acquisition of other char-
acteristics. Instead of selection for F we should think of selection for
F & G1 & G2 & Gn, where the Gi are the ancillary traits with which F is
Consider another type of example. X delivers F, but a rival X* does F
better than X; X* also manifests a deleterious trait G, and the selective dis-
advantage of G outweighs X*s superiority with respect to F. X wins out in
history because it is selected over X*, but this isnt selection for F. Rather
X is selected because it doesnt show G. Nevertheless F may still be the
Imagine an extreme case. Long ago mammals had blood- pumping
devices of greater efficiency. Unfortunately the genetic basis for developing
the organs also generated a physiological condition under which there was
a relatively high chance of a fatal response to some modestly uncommon
environmental factor. Agenomic reshuffling produced a new developmen-
tal program, yielding the ancestral mammalian heartsignificantly less
durable and less efficient than the older device but happily free of the phys-
iological side-condition. Ever since, in all descendant mammalian lineages,
the significant competition has been with genomic reversals that produce
the old developmental program. Recent competition and recent selec-
tion have pitted the older pumping device against the familiar mamma-
lian heart. The latter has wonhas been selectednot because it pumps

[72] The Philosophy of Philip Kitcher

the blood well but because it doesnt bring the risk of the fatal response.
Nevertheless the function of the heart is to pump theblood.
Ill close by considering two ways of defending middling-strong etiol-
ogy. The first is to take on board the Gould-Lewontin concerns about devel-
opmental programs and constraints. We should appreciate that selection is
always for F & G1 & G2 &... Gn and revise our function ascriptions accord-
ingly. If the scenario of the previous paragraph were true, the function of
the heart would not be to pump the blood but to pump the blood well
enough and simultaneously avoid the physiological side-condition. This
idea is akin to that envisaged by Michael Smith in his critique of my claims
about the function of the ethical project. In his honor we might call the
functions etiology assigns Smith functions. Replacing our normal function
ascriptions with attributions of Smith functions would involve a major
reform of biological practice. For revision would be needed in instances
where the package of traits, developmentally bound together by the genetic
basis for the focal entity (organ, system, and so on), contained constituents
with opposite effects on reproductive success. If Gould and Lewontin are
correct, these instances will be very common.
The second objection concedes that there are a few contrived exam-
ples in which proper etiology would expose a more complex selective story,
insisting that, in the main, we can suppose selection to be for some par-
ticular trait or effect, thus retaining our standard ascriptions of functions
(with occasional invocations of Smith functions). From my perspective this
response reflects ungrounded faith, and does so in an area in which faith
is entirely unnecessary. Return to my typical biologist, specifying the func-
tion of an enzyme or an organelle or a larger part of an organism. Igive her
the last word: Evolutionary history is complicated. I can only speculate
on what might have occurred. But Ican see, here and now, how Darwinian
competition bears on the organisms that interest me. Ican recognize the
causes of their success in meeting their challenges. In that light Iascribe
functions as Ido. No need for a detour through the past. What Isee in front
of me is evidence enough.

K i t c h e r s T w o De s i g n S ta n c e s [73]

Proximate and Ultimate Information



Significant progress has been made in philosophy of biology concerning the

nature of biological information since an exchange between Philip Kitcher
and myself a decade ago. At that time Iargued that the idea of biological
information was a barrier to understanding gene-environment interaction.
Here, however, Idescribe two new accounts of biological information that
provide powerful tools for characterizing gene-environment interaction
and implementing Kitchers principle of causal democracy.
All phenotypes depend on both genes and environment for their devel-
opment. This truism constitutes the interactionist consensus in biology
and philosophy of biology. The interesting question is how genes and envi-
ronment interact in the development of particular phenotypes. Kitcher
(2003a, 290) has argued that this can be settled only by patient empiri-
cal research that obeys his principle of causal democracy:Interactionists
ought to support a principle of causal democracy:if the effect E is the prod-
uct of factors in set S, then, for any C S, it is legitimate to investigate the
dependence of E on C when the other factors in S are allowed to vary.1

1. This paper appeared in Singh et al. (2001) before being reprinted in Kitchers
(2003a) collected papers. Kitcher himself cites the paper as 2000 (Kitcher 2003a, 13),
which was the physical publication, as opposed to the imprint, of the original volume.
Kitchers principle has been widely misrepresented. As the quotation
makes clear, causal democracy requires equality of opportunity, not equality
of outcomes. Nevertheless, like conservative political commentators, con-
servative philosophical commentators have represented causal democracy as
the demand that all causes be dragged down to the same level and the refusal
to acknowledge that some causes are more significant than others.2 But like
any good liberal, Kitcher is merely asking that all causes be given a chance to
reveal whether they play a significant role in development. He believes that
current empirical evidence suggests that genes and environment interact in
many different ways, depending on the phenotype being studied, so that the
relative significance of genes and environment must be assessed on a case-
by-case basis. He recognizes that the existence of powerful, standardized
techniques for investigating genetic factors provides a practical justification
for focusing on genes. But he thinks, or at least thought at that time, that the
focus on genetic factors to the exclusion of environmental factors is greater
than can be justified by these practical considerations.
Kitchers (2003a) key message in Battling the Undead is that patient
reiteration of interactionism and causal democracy are all that is needed
for a balanced and accurate assessment of the role of the genes in devel-
opment. He denies that there are any deeper conceptual reasons for the
persistent neglect of the role of the environment in development, contra
Richard Lewontin (1983), Susan Oyama (2000b), and Russell Gray and
myself (Griffiths and Gray 1994). Kitcher agrees that simplistic genetic
determinism can seem like a vampire, rising from the grave each time
it seems to have been dispatched, but he counsels patience rather than
searching for a conceptual stake in the heart (Kitcher 2003a,283).
The decade since Kitcher proposed the causal democracy principle has
seen greatly increased interest in the role of environmental factors in
development (Griffiths and Stotz 2013, ch. 5). Amajor reason for this has
been the rise of the developmental origins of health and disease paradigm
in medicine, which has redirected some of the vast resources available for
biomedical research. Research into obesity, for example, now targets not

2.See, for example, Franklin-Hall 2015; French 2012, 197; Okasha 2009, 724;
Rosenberg and McShea 2008, 174; Thornhill 2007, 206; Weber 2006, 607; Woodward
2011, 249. Most of these authors attribute the idea of causal democracy to Susan
Oyama (2000a, S333), citing her response to Kitcher where she says that she will not
adopt his phrase causal democracy because it introduces into already-complicated
discussions rather more additional baggage than is likely to be helpful. Perhaps these
authors have been misled by the fact that the phrase occurs in the title of Oyamas
paper. Kitchers democracy principle is not dissimilar to Oyamas demand for parity of
reasoning when comparing genetic and nongenetic factors in development, a demand
that is usually parodied in the same way as Kitchers democracy principle.

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [75]
just obesity genes like LEPR but the epigenetic effects of maternal nutri-
tional state on offspring physiology and broader exogenetic pathways from
parent to offspring, such prenatal and neonatal influences on food prefer-
ences. In evolutionary biology there has been a parallel rise of interest in
adaptive phenotypic plasticity, both within and between generations. In
many species a significant component of fitness differences results from
transgenerational environmental influences, or parental effects. Even
behavioral geneticists, the target of some of Kitchers strongest criticisms,
have recognized the need to broaden their research to embrace an interac-
tive picture of behavioral development (Hamer 2002). Some of the power-
ful techniques for studying genetic factors that seemed to provide practical
reasons to focus on genes have simply been repurposed for studying non-
genetic factors. To take just one example, second-generation sequencing
can be used for high-throughput screening of epigenetic marks. The spirit
of democracy seems to have been handsomely vindicated.


Kitcher acknowledges that there is something puzzling about how readily

people default to monocausal genetic explanations. If we are all interaction-
ists, why, then, do we always end up discussing whether genotypes are all-
powerful in development? (Kitcher 2003a, 290300). Susan Oyama has
argued that apparently commonsensical recommendations like Kitchers
causal democracy principle are stymied by the notion that some influences
are more equal than others, that form, or its modern agent, information,
exists before the interactions in which it appears and must be transmitted
to the organism either through the genes or by the environment (Oyama
2000b, 13). Oyama has carefully dissected the role of this idea in distorting, as
she sees it, the interpretation of research and the kind of research that takes
place. Ideas like causal democracy must fight against a deep-rooted conviction
that while phenotypes have many causes, only some of those causes contain
the information that specifies the phenotype, the others playing a merely
supportive role for the expression of that information. Her work builds on
that of behavioral development researchers such as Daniel Lehrman, Robert
Hinde, and Gilbert Gottlieb (see Griffiths and Tabery2013).
Kitcher is skeptical that there is any problem deeper than the usual
preference for overly simple, often monocausal explanations. Adecade ago
Idefended Oyamas diagnosis (Griffiths 2006a), arguing that genetic infor-
mation is frequently read as intentional information that is about some-
thing in the same way that sentences and thoughts are about something.

[76] The Philosophy of Philip Kitcher

As a result the relationship between genes and the outcomes about which
they contain information is assumed to be more deterministic than is sup-
ported by the actual data. For example,

if we describe the same gene as a genetically encoded instruction to be a homo-

sexual, then, intuitively, the presence of different genes at other loci, or prenatal
environments that do not support the cascade of gene expression, or postna-
tal environments that lead the brain to mature differently, all merely cause the
organism to misinterpret or disobey the instruction contained in the gene.
Furthermore, the gene retains its identity as a gay gene even in an individual...
who is, phenotypically, a heterosexual. (Griffiths 2006a,187)

This picture of how the gay gene causes same-sex preference does not
reflect the actual scientific content of behavioral genetic research (Hamer
etal. 1993), but it was clearly the picture operating in public discussion of
that work. The idea that genes are units of information meant the gay gene
was understood as an intentional causethe brain is constructed using
a homosexual blueprint, or it is instructed to be homosexual. But inten-
tional causation is utterly different from the interactive, context-sensitive,
difference-making role of genes envisaged by the interactionist consen-
sus (Sterelny and Kitcher1988).
In more recent work Iand my collaborators have tried to bolster such
anecdotal arguments with a program of experimental research on the folk-
biology of behavioral developmenthow development is understood by
people without formal education in biology (Griffiths etal. 2009; Linquist
etal. 2011). This work provides some empirical support for Oyamas con-
tentions that people hold a dichotomous view of development in which
some phenotypic traits express the organisms inner essence while oth-
ers are imposed on it by the environment; that traits that come from the
inside are thought to be resistant to modification by the environment;
and that this inner essence is nowadays thought to be in the genes.
In this essay, however, Iwant to consider how the concept of informa-
tion can play the opposite role, helping to vindicate the principle of causal


In 2006 I worked on the assumption that biological information must

be either causal information or intentional information (Sterelny and
Griffiths 1999). Causal information is the systematic dependence of one

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [77]
variable on another, the kind of dependence that is measured in informa-
tion theory. One variable carries information about another whenever the
values of the two variables are systematically related. Intentional informa-
tion is the context-insensitive aboutness described earlier. Intentionality
was introduced into philosophy over a century ago as the distinctive fea-
ture of human thought and language. In recent decades, however, phi-
losophers and biologists have argued that intentionality can be created by
natural selection. On this teleosemantic view a variable carries inten-
tional information if it evolved for the purpose of representing another
variable (Maynard Smith 2000; Millikan 1984). Hence intentionality can
be ascribed to bacteria and to genes, not just to humans.
Causal and intentional biological information can be thought of as
proximate and ultimate information, respectively (Griffiths 2013; Mayr
1961). Causal biological information is a way to describe the causal struc-
ture of a living system: How do the parts of the system depend on one
another? It can be characterized mathematically using information theory.
Hence the study of causal information is part of proximal biology. However,
if intentional biological information is teleosemantic information, then it
is a way to describe the purpose for which some aspects of a living system
evolved:What is the evolutionary function of these parts of the system?
This is an ultimate or evolutionary biological question. Knowing how a sys-
tem works will not tell us anything about the teleosemantic information
it contains, unless we also know the selection pressures that created and
maintain the system.
In sections 4 and 5 I describe a new theory of causal/proximate bio-
logical information. This is in the spirit of Sterelny and Kitchers (1988)
difference-making analysis of the sense in which a gene may be a gene for
a phenotype despite the interactionist view that every phenotype depends
on many variables. The new theory uses the recent interventionist view
of causal explanation (Woodward 2003). Some of my earlier claims about
causal/proximate biological information need to be revised in light of this
recent work. Ipreviously argued that causal/proximate information could
not distinguish genetic from nongenetic causes because it is ubiquitous.
Any variable that has an effect on the development of a phenotype will con-
tain information about that phenotype in the sense of information theory;
knowing the state of the causal variable reduces our uncertainty about the
state of the phenotype (Griffiths and Gray 1994; Maynard Smith 2000).
In this more recent work, however, my collaborators and Ihave used the
concept of causal specificity to discriminate between causes that provide
information for their effects and those that do not (Griffiths etal. 2015;
Griffiths and Stotz 2013). This reintroduces the possibility that genetic

[78] The Philosophy of Philip Kitcher

causes are the sole or main locus of developmental information. However,
Iwill show that our new approach provides a powerful, quantitative way to
state the principle of causal democracy and that at least some environmen-
tal causes contain enough information to deserve citizenship.
In section 6 Idescribe and build on a debate between Nicholas Shea and
myself over whether ultimate information can play a role in developmen-
tal explanations (Griffiths 2013; Shea 2013). Shea (2007) has developed a
sophisticated teleosemantic approach to information and claims that the
best way to defend the significance of nongenetic causes is to show that
it is not only genes that carry what he calls inherited information (Shea
2011). Iaccept that inherited information is a useful concept in evolution-
ary theory. However, Ihave argued that inherited information cannot cause
development. Irevisit this criticism in the light of significant developments
in philosophical accounts of biological teleology in the past decade. These
open the way to construct an ahistorical teleosemantics. This defines teleose-
mantic intentional information in terms of the current causal structure of
organisms and makes the presence or absence of this information a poten-
tial causal difference-maker in development.


It has long been argued that because the effect of an allele substitution
depends on many other factors, both other genes and the environment,
it is misleading to identify a single allele as the gene for a phenotype.
However, the fact that alleles produce phenotypes interactively does not
prevent their being salient causes of those phenotypes in the interactionist
picture favored by Kitcher. Alleles cause phenotypes by making a difference
to those phenotypes against a background of other factors. This idea was
spelled out in detail by Kitcher in an article with Kim Sterelny:

An allele Aat a locus L in a species S is for the trait P* (assumed to be a determi-

nate form of the determinable characteristic P) relative to a local allele B [at the
same locus] and an environment E just in case (a)L affects the form of P in S,
(b)E is a standard environment, and (c)in E organisms that are AB [genotype]
have phenotype P*. (Sterelny and Kitcher 1988,350)

Gray and I offered two criticisms of this definition (Griffiths and Gray
1994). The first concerned the definition of standard environment, to
which Kitcher (2003a, 29192) responded with an amended definition. The
second criticism was that the definition could equally license epigenetic

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [79]
marks for or incubation temperatures for phenotypes. Kitcher does not
regard this as a criticism and believes that we can and should treat genes
and environment symmetrically in this respect:Far from being a reductio
of the interactionist view, this point simply testifies to the democracy prin-
ciple introduced above(293).
The idea that there are genes (and other factors) for phenotypes is
closely related to the idea that genes contain information about pheno-
types. Sterelny and Kitchers difference-making approach can be readily
translated into information-theoretic terms. Their analysis identifies a
covariance between gene and phenotype when other factors are held con-
stant. We can regard the gene as a signal source, the phenotype as a signal
receiver, and the other factors as channel conditions. When the channel
conditions are stable, we can reduce our uncertainty about the state of the
phenotype by observing the state of the gene, so the gene carries informa-
tion about the trait (Griffiths and Gray 1994). Iobserved earlier that infor-
mation in this sense is ubiquitous. All developmental factors carry such
information. For many authors, the fact that causal information is found
in all factors affecting development is a reason to look for another kind of
information that is found only in genes, or in genes and some special selec-
tion of environmental factors. This has typically been teleosemantic inten-
tional information (Maynard Smith 2000; Shea 2007; Sterelny etal. 1996).
However, an alternative strategy is to develop a more discriminating causal
account of information using resources from the philosophy of causation
and information theory. It is this strategy that my collaborators and Ihave
pursued in our recentwork.
The influential interventionist theory of causal explanation provides
new resources for the study of causal difference- making (Woodward
2003). It provides formal criteria that distinguish causal from noncausal
relationships, based on the insight that causal relationships are relation-
ships that are potentially exploitable for purposes of manipulation and
control (Woodward 2010, 314). The theory treats causation as a relation-
ship between variables in a scientific model, using causal graph theory as
a canonical format in which to express these models. There is a causal rela-
tionship between variables X and Y if it is possible to manipulate the value
of Y by intervening to change the value of X. Intervention here is a techni-
cal notion with various restrictions. For example, changing a third variable
Z that simultaneously changes X and Y does not count as intervening on
X. Causal relationships between variables differ in how invariant they are.
Invariance is a measure of the range of values of X and Y across which the
relationship between X and Y holds. But even relationships with very small
ranges of invariance are causal relationships.

[80] The Philosophy of Philip Kitcher

The basic interventionist criterion of causation is deliberately weak,
admitting even causal relationships that hold only under very narrow con-
ditions. In more recent work Woodward and others have examined why
some causes are singled out as more salient or significant than others. They
have focused on how such causal selection occurs in biology (Stotz 2006;
Waters 2007; Weber 2006; Woodward 2010). One of the most prominent
proposals is that causes differ in the degree to which they are specific to
their effects.
Causal specificity can be illustrated by contrasting the tuning dial and
the on-off switch of a radio. Hearing the news is equally dependent on the
dial taking the value 576 and on the switch taking the value on. But the
dial seems to have a different kind of causal relationship with the news
broadcast than the switch does. The switch is a nonspecific cause, whereas
the dial (or digital tuner) is a specific cause. Interventions on a specific
causal variable can be used to produce a large number of different values
of an effect variable, providing what Woodward (2010, 302)terms fine-
grained influence over the effect variable.
The existing literature on causal specificity is mostly qualitative, rely-
ing on examples and intuition, and the authors recognize that greater
precision is needed (Weber 2006, 606). Woodward has suggested that the
limit of fine-grained influence is a bijective mapping between the values
of the cause and effect variables:every value of E is produced by one and
only one value of C and vice versa. The idea of a bijective mapping does
not admit of degrees, but we have developed an information-theoretic
framework with which to measure the specificity of causal relationships
within the interventionist account, with a bijective mapping as a limiting
case (Griffiths etal. 2015). Our measure formalizes the idea that, other
things being equal, the more a cause specifies a given effect, the more
knowing how we have intervened on the cause variable will inform us
about the value of the effect variable. This led us to propose a simple mea-
sure of specificity:

Spec:The specificity of a causal variable is obtained by measuring how much mutual

information interventions on that variable carry about the effect variable.

The mutual information of two variables is simply the redundant infor-

mation present in both variables. Where H(X) is the Shannon entropy
of X, the mutual information of X with another variable Y, or I(X; Y), is
I ( X ; Y ) = H ( X ) H ( X |Y )

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [81]
Mutual information is symmetrical:I(X; Y) = I(Y; X). So variables can have
mutual information without being related in the manner required by the
interventionist criterion of causation. However, our measure of specificity
does not simply measure the mutual information between C and E. Instead
it measures the mutual information between interventions on C and the
variable E. This is not a symmetrical measure because the fact that inter-
ventions on C change E does not imply that interventions on E will change
C:in general, I(C ; E) I(E; C), where C is read do C and means that the
value of C results from an intervention on C (Pearl2009).
Any two variables that satisfy the interventionist criterion of causation
will manifest some degree of mutual information between interventions
and effects. If the relationship C E is minimally invariant, that is, invari-
ant under at least one intervention on C, then C has some specificity for E,
that is, I(C ; E)>0. Conversely, if C E has some degree of specificity, then
the relationship is invariant under at least one intervention on C.
Elsewhere we have argued that a causal relationship in biology should
be regarded as an informational relationship when it is highly specific
(Griffiths and Stotz 2013). We are not the first to draw a link between
information and specificity. Woodward (2010, 312n21) has written, The
ideas of causal specificity and information are obviously closely linked.
Biologists tend to think of structures as carrying information when they
are involved in causally specific relationships. Iregret that Ilack the space
to explore this connection in more detail. Sahotra Sarkar (2004) and Ulrich
Stegmann (2014) have also argued that the salient causes in development
are the most biologically specific causes. Sarkars set-theoretic analysis of
biological specificity is very similar to Woodwards idea of a bijective map-
ping, and Isuggest that biological specificity is simply causal specificity in
a biological system.
Using our measure, a causal relationship will be highly specific whenever
C and E can take many values and there is a high degree of mutual informa-
tion between them. In informal terms the cause can make the difference
between many different states of the effect and can be used to exercise fine-
grained control over that effect. This is actually what Francis Crick (1958,
153)meant by information when he introduced the sequence hypothesis
and the central dogma of molecular biology: Information means here
the precise determination of sequence. The distinction between instruc-
tive and merely permissive causal interaction in developmental biology
is also a distinction between more and less specific causes in our sense.
The proposal to identify biological information with causal specificity in
biological systems is thus a classic explication (Carnap 1950). We construct

[82] The Philosophy of Philip Kitcher

a more precise substitute for an intuitive idea: Informational causes are
causes for which I(C ; E) is substantial. We then demonstrate that there
is significant intellectual continuity with the original, intuitive notion. In
the next section Ishow that our analysis of biological information meets
another requirement for a good explication:it provides a useful tool for the
relevant sciences.


Some authors have suggested that the very idea of causal specificity refutes
the principle of causal democracy (see note 2). This relies on the misrep-
resentation of causal democracy as the view that all causes are equally
significant, so it can be refuted merely by showing that it is possible to
discriminate. But Kitchers principle is that all causes should be given a
chance to show their significance. This is evidently compatible with the idea
of causal specificity and with other theories of causal selection.
When Crick advanced the sequence hypothesis and central dogma he
assumed that the sequence of the gene not only precisely determined the
sequence of the product but also completely determined it. The discovery
of alternative splicing in the 1970s showed that the sequence of the gene
can underdetermine the sequence of the product. Since then, alternative
splicing has turned out to be ubiquitous in eukaryotes and has been joined
by other mechanisms of pre-and post-transcriptional processing: mRNA
editing, co-transcription, programmed frame shift, trans-splicing, transla-
tional recoding, and protein trans-splicing. The transcriptomethe total
population of RNAs found in the cells of an organismis at least an order
of magnitude greater than the number of genes.
These mechanisms are employed because there are many different types
of cell, each of which uses the same genetic resources to make a different
set of products; even a single cell uses those resources differently at differ-
ent stages in its life cycle. This requires additional specificity of a kind not
captured by the original sequence hypothesis. Cricks biographer Robert
Olby (2009, 251, italics added)notes:

Clearly, in concentrating on this aspect of informational transfer he [Crick]

was setting aside two questions about the control of gene expressionwhen
in the life of a cell the gene is expressed and where in the organism. But these
are also questions of an informational nature, although not falling within Cricks

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [83]
This point was immediately obvious to Cricks contemporaries and led the
ciliate biologist David L.Nanney (1958, 712)to introduce the idea of epi-
genetic control systems:

This view of the nature of the genetic material... permits, moreover, a clearer
conceptual distinction than has previously been possible between two types
of cellular control systems. On the one hand, the maintenance of a library of
specificities, both expressed and unexpressed, is accomplished by a template
replicating mechanism. On the other hand, auxiliary mechanisms with differ-
ent principles of operation are involved in determining which specificities are to
be expressed in any particular cell.... To simplify the discussion of these two
types of systems, they will be referred to as genetic systems and epigenetic
systems. The term epigenetic is chosen to emphasize the reliance of these
systems on the genetic systems and to underscore their significance in develop-
mental processes.

The philosophical literature on causal specificity has not been blind to this
aspect of specificity, and Woodward (2010, 3045, italics added) has noted
that specificity includes systematic dependencies between a range of dif-
ferent possible states of the cause and different possible states of the effect,
as well as dependencies of the time and place of occurrence of E on the time
and placeofC.
It is evident that the additional research questions to which Olby refers
concern additional sources of specificity. Our account of biological infor-
mation as causal specificity chimes neatly with the way many biologists
use information in these contexts. Regulatory mechanisms that affect how
coding sequences are used in a particular cell at a particular time have
been described as amplifying the information in those coding sequences.
Biologists in this field search for the target sequence specificity of
forms of editing (Davidson 2002)or search for the missing information
needed to supplement the information in the coding sequence (Wang and
Nanney hypothesized that the utility of the epigenetic control systems
lies precisely in their ability to respond specifically to altered environmen-
tal conditions (1958, 713, italics added). He suggested that the influence
of these systems should be understood in terms of their specificity of
induction of developmental effects (715). We see the same language of
specificity employed in developmental biology when biologists distinguish
between the more specific instructive and the less specific permissive induc-
tive interactions.

[84] The Philosophy of Philip Kitcher

Identifying biological information with causal specificity provides a
rigorous analysis of what is meant by information in these contexts and
explains why these very disparate physical processes influencing the phe-
notype share a common informational currency. Causal specificity can pro-
vide that currency because it obeys Kitchers causal democracy principle.
It does not build into the definition of biological information any features
that prejudge the issue of which causal factors contain information.
My concern in this and the previous two sections has been with proxi-
mate information, information that does causal work in living systems.
In the following section Iturn to ultimate information, to definitions of
information based on the process of evolution. I raise the prospect that
these two kinds of information, currently serving very different kinds of
biological explanation, can be brought closer together.


The most developed account of ultimate information in the current lit-

erature is Sheas (2007, 2011)version of teleosemantic intentional infor-
mation, which he calls inherited information. Shea accepts the causal
democracy principle. It is an empirical question whether environmental
factors contain inherited information:

To make the case against gene centrism, DST [developmental systems theory]
should be pointing to the undoubted specialness of genes and saying, You know
that property, the one that makes genes so special? Well that property is found
not just in genes but in several other factors in development. That special role is
to transmit information, generated through a process of natural selection, down
the generations to inform development. (Shea 2011,61)

Shea differs from earlier teleosemantic theorists by requiring that rep-

resentations actually correlate with what they represent, so he describes
his account as an infotel theory, combining information theory and
A cause contains inherited informationif

(a) there is a consumer system which is caused by a range of tokens,

including tokens of type R, to produce a range of outputs, with a spe-
cific evolutionary function for each type of output, and where every
token satisfies (b)to (d)with respect to some content;
(b) Rs carry the correlational information that condition C obtains;

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [85]
(c) an evolutionary explanation of the current existence of the represent-
ing system adverts to Rs having carried the correlational information
that condition C obtains;and
(d) C is the evolutionary success condition, specific to Rs, of the output of
the consumer system prompted by Rs. (Shea 2013,5)

Shea defines an evolutionary success condition thus:The proximal evo-

lutionary explanation of the survival and reproduction of the representing
system adverts to Cs obtaining when Rs were tokened(5).
I agree with Shea that both genetic and environmental causes in devel-
opment can carry inherited information and that this fact can be used to
show that they sometimes play the same role in evolution. My concern is
that the presence of inherited information, whether in genes or in environ-
ment, cannot contribute to proximate explanations of biological develop-
ment. Conditions (c)and (d)imply that which information a gene carries
depends on the selection pressures that acted on ancestral copies of the
gene. It follows that physically identical genesan inherited beneficial
allele and a de novo mutation that produces the same allele, for example
need not share any information in Sheas sense. They will both share what
he calls correlational information, but the new mutation will not meet
conditions (c) and (d) and so will contain no inherited information. But
this difference in information can make no difference to how these genes
affect the developmental process, since the two are physically identical. The
presence or absence of information in Sheas sense is a difference that
makes no difference in development (Griffiths2013).
Shea (2013) argues that, despite this point, inherited information can
explain how organisms develop. He describes models that treat both genes
and environmental factors as sources of information about future environ-
ments and describes how these models are used to study the conditions
under which a phenotype will develop with or without an environmental
cue. Similarly biologists can study whether an organism should respond to
its own developmental environment or to its parents environment, signaled
by a cytoplasmic cue, by asking which is the more reliable source of informa-
tion about the environment in which the organism evolved. Sheas examples
are compelling, and his teleosemantic definition of information is appropri-
ate for such explanations. But these are evolutionary, ultimate explanations
and not proximate explanations. By modeling the adaptive advantage of
that design in past environments they explain why development is designed
in a particular way. They dont answer the proximate How? question any
more than an evolutionary model predicting the optimal design of a mito-
chondrion could have been used to solve the mystery of how oxidative

[86] The Philosophy of Philip Kitcher

phosphorylation is possible in biochemistry (see the excellent account in
Weber 2005). If a plant adds something to its seeds that causes those seeds
to flower early, it must add something more than inherited information.
In the remainder of this section Iwill explain how a concept of informa-
tion that can figure in proximate developmental explanations can be con-
structed by taking Sheas definition and censoring the claims it makes about
history. To do this Iwill need an ahistorical account of biological teleology.
Fortunately there has been an upsurge of interest in such accounts in recent
years (Christensen 1996; Griffiths 2009; McLaughlin 2001; Schlosser 1998;
Weber 2005; Wouters 2007). The common theme in these accounts is that
the function of a biological trait is the contribution it currently makes to
survival. But to make such an account work, survival must be understood
as survival and reproduction (Griffiths 2009). Focusing on survival with-
out reproduction makes it impossible to understand many functions of
physiology and behavior, since many life-history strategies sacrifice health
and physical integrity for increased reproductive fitness. In fact the whole
apparatus of evolutionary theory is needed to identify what an organism
is doing because some physiological mechanisms have functions that can-
not be characterized except by referring to the evolutionary process, such as
controlling the variance of the distribution of numbers of offspring.
The functions of a trait in this ahistorical sense are those features in vir-
tue of which the trait has survival value (Tinbergen 1963). Questions of
survival value are questions about the causes of current fitness. This aspect
of Niko Tinbergens four questions framework is often misunderstood,
but in fact, these questions ask whether any effect of the observed process
contributes to survival, if so how survival is promoted and whether it is
promoted better by the observed process than by slightly different pro-
cesses (418). The answer to these questions provides vital data with which
to ask further, evolutionary questions about the selection pressures that
produced the trait. Survival value itself, however, is not a historical mat-
ter:Even if the present-day animals were created the way they are now,
the fact that they manage to survive would pose the problem of how they
do this(423).
This ahistorical yet entirely evolutionary approach identifies the func-
tions of a trait with the features that are adaptive, whereas the historical
approach identifies functions with the features for which the trait is an
adaptation. Although adaptiveness and adaptation are both essential to the
theory of natural selection, many philosophers are skeptical about whether
an ahistorical definition of function is possible. This skepticism may stem
from the deficiencies of one well-known ahistorical theory (Bigelow and
Pargetter 1987). But the view that any ahistorical evolutionary view of

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [87]
function will be incoherent seems to me to be a non sequitur. Here are four
versions of the argument:3

1. What a trait is adapted for is fully determined by facts about the past,
although admittedly these may be hard to discover. But there is no
determinate fact about what a trait is currently adaptive for. It may do
something for one organism but not another, or in one locality but not
another. Reply:What a trait is adapted for is defined as what it was
adaptive for in the past. So the first cannot be determinate if the sec-
ond is not. To identify the evolutionary forces now acting on a popula-
tion we have to look for general patterns that may not hold for every
organism, but we have to do that to identify the evolutionary forces that
acted in the pasttoo.
2. Fitness depends on the environment. When studying adaptation we
know exactly which environments are relevant:those in the actual past.
But when studying why a trait is currently adaptive we have no objec-
tive basis for rejecting abnormal environments. Some animals live in
zoosshould we include them? Reply:This is a more concrete version
of the first argument. There were animals in zoos in the past too. We
need to make decisions about what constitutes a single selective process
whether we are looking at the past or the present.
3. The evolutionary trajectory leading to the current population is deter-
minate, a matter of past facts. But its future trajectory is indeterminate
because future events may interfere with our best prediction. Reply:The
ahistorical account does not need to predict the future. It only needs to
establish the causes of current fitness. Just as a moving object has an
instantaneous trajectory in space, a population has an instantaneous
evolutionary trajectory. In experimental studies of evolution, it is this
instantaneous trajectory that we actually study. Even what might seem
to be essentially predictive traits, such as variance of offspring num-
ber, can be defined at an instantthat is why we can design games
of chance like lotteries rather than having to construct them by trial
4. If functions are defined ahistorically, they will not explain why the traits
that have these functions exist. Reply:Ahistorical function is not meant
to replace historical function, any more than adaptive can replace
adaptation. The two are complementary. To explain why a trait exists

3. It has also been argued that we cannot describe organisms ahistorically because
their parts are defined by their adaptive function (Neander 2002; Rosenberg and
Neander 2009). For a refutation, see Griffiths (2006b).

[88] The Philosophy of Philip Kitcher

you need the concepts of adaptation and historical function. To explain
what a trait currently does for the organism (and to define adaptation)
you need the concepts of adaptiveness and ahistorical function.

With this ahistorical account of biological teleology, I can construct

a definition of teleosemantic intentional information free of history.
References to past evolution in Sheas conditions are simply replaced by
references to present evolution. Rather than inherited information, we get
adaptive information:
A cause contains adaptive informationif

(a) there is a consumer system which is caused by a range of tokens, includ-

ing tokens of type R, to produce a range of outputs, with a specific sur-
vival function for each type of output, and where every token satisfies
(b)to (d)with respect to some content;
(b) Rs carry the correlational information that condition C obtains;
(c) an explanation of the current fitness of the representing system adverts
to Rs carrying correlational information that condition C obtains;and
(d) C is the success condition, specific to Rs, of the output of the consumer
system prompted by Rs (that is, R increases fitness because C obtains).

So if we remove the historical content from Sheas theory of inherited

information, we are left with the claim that some state R (a)has an effect
on the organism, (b) carries information about the environment, (c) the
resulting pattern of Rs has survival value, and (d)each output increases fit-
ness because it fits a specific C. For example, the North American seed bee-
tle Stator limbatus follows alternative developmental pathways on different
hosts. To survive on Blue Palo Verde seeds, offspring must grow faster and
attain a larger final size than those developing on seeds of Catclaw Acacia.
Mothers bring this about by laying fewer, larger eggs on the Palo Verde
seeds than they do on the Acacia seeds (Fox etal. 1997). Having detected
which kind of seed it is depositing eggs upon, the mother signals to the
offspring to adopt one growth strategy rather than another. The egg mass
contains adaptive information because growth rate (output of the con-
sumer system) is caused by seed mass (R), which is correlated with the spe-
cies of tree on which the egg is laid (C), and fitness is enhanced because
different growth rates suit different tree species (in this case the egg mass
also contains Sheas inherited information, since this system is an adapta-
tion, but this will not always beso).

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [89]
This proposal can be made more comparable to the causal/proximate
notion of information described in sections 4 and 5 by stating it is as a rela-
tionship among three variables. C continues to denote Sheas environmen-
tal condition; R continues to stand for the state that signals this condition;
and Iintroduce E (effect) to denote the outputs of Sheas consumer system.
A causal variable contains adaptive informationif

(a) there is a variable E whose value depends on a variable R of which con-

ditions (b)through (d)hold:
(b) R correlates with some third variableC;
(c) E contributes to fitness by responding to R because R is correlated
(d) for each r j there is a success condition ck such that if the ei caused by r j
contributes to survival then it does so because ck is the expected value
of C givenr j.

A connection to the idea of causal specificity from section 4 can now be

seen. R is specifically caused by the environmental variable and specifically
causes a state of the organism. It is straightforward to express conditions
(a)through (d)formally using the specificity measure introduced in section
4 and adding a fourth variable to denote fitness. Many readers will also
have noticed the resemblance between the system described by (a)through
(d)and a signaling network (Skyrms 2010). However, a formal treatment
would reveal that conditions (a)through (d)are inadequate in ways that are
hidden by their merely verbal formulation. Ageneral version of the condi-
tions for either inherited or adaptive information requires a measure of
causal influence that does not break down when other variables interact
with C to determine R or interact with R to determine E. Afull treatment
is in preparation.
It is important to recognize that adaptive information is entirely com-
patible with Sheas inherited information. The two relate just as adaptive-
ness relates to adaptation. For example, just as something needs to be
adaptive in the past to be an adaptation in the future, a representation
needs to have contained adaptive information in the past if it is to con-
tain inherited information in the future. Another important parallel is that
when an adaptation is useful in the current environment, this implies it
is still adaptive. Just so, if a system that contains inherited information
produces an adaptive fit between organism and environment, this must
be because it contains the corresponding adaptive information. Possessing
inherited information without adaptive information does not, by defini-
tion, produce an adaptivematch.

[90] The Philosophy of Philip Kitcher

The added value of the idea of adaptive information is that it can feature
in proximate explanations of the operation of living systems. This is easiest
to see when the variable C is in the external environment, for example, the
presence of predators. The statistical and causal relationships among R, C,
and E explain how the organism detects the predators. But adaptive infor-
mation can also feature in proximate developmental explanations, when C is
inside the organism. The statistical and causal relationships among R, C, and
E might, for example, explain how a transcription factor succeeds in relating
a stimulus received by a cell to a developmental response (Calcott 2014).


Since my exchange with Kitcher a decade ago significant progress has

been made in philosophy of biology on the nature of biological informa-
tion. Then I argued that the idea of biological information was a barrier
to understanding gene-environment interaction. Here Ihave argued that
biological information can be a powerful tool with which to characterize
gene-environment interaction and to implement Kitchers principle of
causal democracy.
In sections 4 and 5 Idescribed a new theory of causal/proximate bio-
logical information. This is in the spirit of Sterelny and Kitchers (1988)
difference-making analysis of the sense in which there are genes for
phenotypes. The new theory combines information theory with the inter-
ventionist view of causal explanation to develop a quantitative measure of
In section 6 Idescribed a new theory of ultimate, evolutionary informa-
tion based on an ahistorical teleosemantics. The new theory defines tele-
osemantic intentional information in terms of the current functioning of
organisms and makes the presence or absence of this information a poten-
tial causal difference-maker in development.
Both these accounts of information are in principle equally applicable to
genetic and nongenetic causes in development. They are powerful resources
for the patient, empirical exploration of the relative importance of differ-
ent causes in the development of phenotypes that Kitcher recommended
a decadeago.4

4.This publication was made possible through the support of a grant from the
Templeton World Charity Foundation. The opinions expressed in this publication are
those of the author and do not necessarily reflect the views of the Templeton World
Charity Foundation.

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [91]
Reply toGriffiths

The genomic research of the past two decades has yielded increasingly com-
plicated pictures of both development and evolution. Along with Karola
Stotz and John Dupr, Paul Griffiths has been at the forefront of attempts
to bring new discoveries about the dynamic genome to the attention of phi-
losophers. As Pauls illuminating essay reveals, even those who campaigned
in the 1980s and 1990s against the oversimplified scenarios often spun in
discussions of the biological basis of human behavior, tales woven by
biologists as well as by philosophers, underrepresented the wide variety of
ways in which epigenetic and environmental factors enter into proximate
and ultimate causation. Even the most ardent interactionists were unaware
of the intricacies of the interactions.
Griffiths sees how biologically informed philosophy can make far more
precise the principle of causal democracy I advocated in my critique of
three prominent ventures in articulating general models of development
and evolution that hoped to rival (and displace) the dominant gene-centric
approaches. Susan Oyama, Richard Lewontin, and Griffiths (along with
Russell Gray) aimed at a novel picture of ontogeny and of evolution, one
that would prevent the hasty and unwarranted forms of genetic determin-
ism, seemingly arising with monotonous and potentially harmful regular-
ity. Although I shared these thinkers concerns, I believed that no such
general account was needed. Recognizing causal democracy would suffice.
The principle of causal democracy was intended, as Griffiths sees, to
offer equal opportunity to a wide variety of potential factors. It allows dif-
ferent investigators to focus on a particular type of cause and to explore the
effects as other factors vary. It also encourages them to consider a variety
of potential causes. Perhaps at the end of the day, when different processes
are analyzed in terms of different types of causes, it will turn out that DNA
sequences play the most important role in a plurality of cases. Or perhaps
notsome other factor might dethrone the gene. Alternatively we might
discover that no single type of cause is the most fundamental across
any significant range of developmental and evolutionary phenomena. The
whole idea that some particular biological factor is more important than all
others might turn out to be a mistake.
Griffithss views have evolved since our debate of a decade agoand
so have mine. Although Icontinue to believe that the principle of causal
democracy offers valuable methodological counsel in the design of biologi-
cal research and the appraisal of biological claims, Inow think my earlier

[92] The Philosophy of Philip Kitcher

objections to the ventures in general modeling were too harsh. Although
there are good reasons to doubt whether any fully general model can be
precise enough to be helpful, the models Oyama, Lewontin, and Griffiths
tried to elaborate are useful tools for explaining families of biological phe-
nomena, and, by demonstrating that, its possible to destabilize the mon-
ocular fixation on genes. Ishould have recognized the potential worth of
some vivid examples of causal democracy in action.
In the rest of this reply I want to consider four debates in which the
principle of causal democracy proves useful. The first is the controversy at
the heart of Griffithss essay:Are some causes privileged in virtue of their
character as bearers of information? The second is the disagreement about
strategies in biomedical research. The third is the ancestral home of my
own worries about genetic determinism (and consequent enthusiasm for
causal democracy):the continued emphasis on genetic bases for complex
human behavioral traits. The fourth is an oddity of contemporary analytic
philosophy:the uninformed enthusiasm for attempts to show that evolu-
tion debunks this or that philosophical position.
A gene-centered view of the biological world often takes flight from
the thought that the genes provide a set of instructions for the develop-
ing organism. As gene-centrists willingly concede, gene action is possible
only when other ancillary factors are in place. The genes, however, are
the teachers, and nobody reflecting on education should deny that teach-
ers play a more fundamental role than their aides or the kitchen staff or
the janitors or the classrooms or the blackboards. Educational democrats
might protest that evaluation, and both Griffiths and Nicholas Shea offer
an analogous rebuttal in the biological domain. They are both concerned to
show that other causal factors can be information-bearing. Hence there is
no basis for singling out the genes as privileged.
Theres no doubt that appeal to information has allowed useful for-
mulations of research questions at some points in the history of molecular
inquiry; it did, after all, inspire the search for the genetic code. Griffiths and
Shea want to sanction the usage anddemocraticallyextend the class
of potential information-bearers. Their different programs depend on the
success of teleosemantics. As a naturalistically inclined philosopher, Id be
delighted were some version of teleosemantics to succeed, but the formula-
tions available so far face well-known difficulties. Im inclined to suspect that
any fully adequate semantic theory will outrun the conceptual resources
present-day theorists deploy. Iappreciate the possibility that some future
semantics will endorse many types of biological cause as information-
bearing. But Idont want to leave attacks on the thesis that genetic causes

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [93]
are privileged because the genes are the sole information-bearing biological
entities hostage to debates about the future of semantics.
So Iprefer a different strategy. Sterelny and Kitcher (1988) and Kitcher
(2003a) eschew talk of information in terms of a more austere causal idiom;
as Griffiths rightly points out, those articles look at the differences causes
make, an approach elaborated and much improved in the recent work of
James Woodward and Kenneth Waters. Iregard the talk of information as
a metaphor, useful when it helps to make some biological issue vivid, but
potentially misleading, especially when users forget that the coinage is not
literal. The ultimate test of a use of the metaphor is to translate back into
the primitive austere idiom and to frame questions in terms of difference-
making causation. Causal democracy recognizes that there are all kinds
of difference-making causes: sometimes allelic differences are important
across a wide range of environments; on other occasions an environmental
factor (the absence of an important source of some nutrient, say) makes
a large and uniform difference across a spectrum of genotypes. The basic
causal idiom provides no basis for singling out some causes as always more
crucial than others. Treating informational language as metaphor thus
diagnoses gene-centrists as people who have allowed a figure of speech to
run riot in their thinking.
Griffiths touches on my second debate in his optimistic assessment of
the ways biomedical research is broadening its menu of options for explor-
ing the causation of disease. Iam less hopeful. In my judgment the narrow
focus on genes continues to dominate. Asymptom of that dominance is the
current fascination with personalized medicine. Many universities with
prominent schools of medicinemy own among themare giving very
high priority to the establishment of centers in which researchers will use
information about DNA sequences in attempts to identify drugs and other
forms of treatment that can be beneficial for people who carry a particular
sequence (or some member of a family of sequences). This form of gene-
centered inquiry is often heralded as the next phase in the progress of
Who could complain? Surely it would be better if physicians knew in
advance which of a collection of medical regimes would be optimal for an
individual patient. Yet its worth asking who the beneficiaries are likely to
be. The universities who rush to achieve eminence in personalized medi-
cine are not driven purely by a laudable wish to ameliorate the human con-
dition. Personalized medicine and translational research are terms that often
appear in close proximityand the juxtapositions should remind us of a
basic fact:theres money in it. Visions of profitable patents already dance
before administrative eyes. When the universities receive the returns on

[94] The Philosophy of Philip Kitcher

their wise investments, its easy to predict the patients who will benefit
most from the new advances. They will be those whose medical insurance
will cover the costs of expensive drugs and procedures. The rich will con-
tinue to get healthier.
The more resources devoted to personalized medicine, the less money
and talent will flow to other biomedical research initiatives. Perhaps there
are some genetic determinists who believe that the health gap between rich
and poor is to be explained in terms of the prevalence of different geno-
types in the two groups, but determinists of this stripe are (fortunately)
rare. Higher rates of asthma, diabetes, hypertension, and heart disease
among poor people are likely explicable as effects of environmental factors,
some of them already knownbut not easy to eradicateand others that
remain to be identified. Well-ordered science, used as a diagnostic tool to
pick out promising ways of improving overall human health, would prob-
ably recommend a turn away from gene-centrism toward programs aimed
at discovering more of the harmful environmental factors or developing
better strategies for eliminating those alreadyknown.
At this point the principle of causal democracy enters discussions of
biomedical research policy. So long as gene-centrists can insist on the
fundamental role of genes in biological processes, the search for molecu-
lar causes will seem the most efficacious strategy. The principle of causal
democracy challenges this response. Griffiths and Shea are not embroiled
in a purely academic debate when they seek different ways of vindicating
causal democracy. Iprefer a different line of objection, but we agree on the
goals:first, the dethronement of genes as privileged causes and, in conse-
quence, the broadening of medical research so that it is less skewed toward
further attending to those who are already relatively welloff.
Turn now to my third debate. Griffiths correctly points out how con-
temporary behavioral genetics is beginning to absorb the lesson of causal
democracy. Yet a form of genetic determinism continues to infect some
of the enterprises that march under the banner of evolutionary psychol-
ogy. Here too there are encouraging exceptions:Griffiths alludes to stud-
ies of the effects of maternal nutrition, and there have been some subtle
and rigorous investigations of how the nutritional deficiencies of pregnant
females affect aspects of the behavior of their daughters (see the work of
Patrick Bateson and Daniel Nettle). Some of the investigators (Bateson,
for example) were long-standing causal democrats avant la lettre. Their
approach contrasts with that of the most prominent style of evolution-
ary psychology, beloved of journalists needing a sexy science story:the
Santa Barbara paradigm.

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [95]
The principle of causal democracy descended from my efforts to expose
the flaws in the ambitious style of human sociobiology practiced in the
1970s and early 1980s (Kitcher 1985). The mainstream evolutionary psy-
chology of today rarely commits all the errors of its sociobiological ances-
tor, but theres often a simplified way of thinking about evolution and
development, akin to those Icriticized, that is essential to the attention-
grabbing conclusions. Were told that a particular behavioral propensity
the tendency of young females to find attractive those males who can
supply resources, saywould have been advantageous on the savannah.
This serves as the basis for a hypothesis about a modular psychological
capacity, which can then be seen as operative across a range of experimen-
tal or survey data. Contemporary womens behavior is thus interpreted
as expressing their savannah-selected propensity to be attracted to well-
provided (typically older) men. Add the thesis that selection can operate
only where there is genetic variation, and the way is open for the conclu-
sion that this propensity is hard-wired. At this point, of course, the jour-
nalists pick up thestory.
Even from the far more limited perspective of the 1980s, there were
many evolutionary and developmental alternatives. Today, with our richer
awareness of genomic complexities, the route to genetic determinist con-
clusions is even more crowded with rival scenarios, not to be dismissed out
of hand. If it is to fulfill its ambitions, evolutionary psychology must go
beyond the simple narratives that dominate manybut not allventures
under the Santa Barbara paradigm. It should emulate the approach of
Bateson and Nettle, with its thorough incorporation of causal democracy.
The simplifications of parts of evolutionary psychology are, however, as
nothing compared with the recent fad for evolutionary debunking argu-
ments in philosophy. Since the publication of Street (2006), meta-ethics
in particular has succumbed to a deluge of articles debating whether, if
moral realism is correct, a human capacity for morality could have evolved.
(Interestingly Streets original article is more sophisticated about evolution
than all those Iknow that have come after it.) Underlying the stream of
papers lies a common trio of ideas:evolution means evolution by natu-
ral selection; natural selection favors or frowns upon very specific traits,
things like a capacity for detecting moral truths; and underlying those
traits are genes forthem.
To hold any of these ideas you have to be very innocent with respect to
contemporary evolutionary theory. First, when the animal in whose evo-
lution you are interested is Homo sapiens, cultural transmission and cul-
tural selection can play a not inconsiderable part. Thanks to Robert Boyd
and Peter Richerson (1985), its been known for thirty years that cultural

[96] The Philosophy of Philip Kitcher

selection can lead to outcomes different from those to which natural selec-
tion alone would tend and that the pertinent regime of selection can be
maintained under natural selection. More recently Boyd and Richerson
(2005) have offered a new folk theorem: under cultural selection, just
about anything can evolve. That seems close to a debunking of debunking
Second, the idea that any psychological trait that happens to strike a phi-
losophers fancy would be a target of selection is unwarranted speculation.
Gould and Lewontin (1979) reminded biologists, as well as some interested
philosophers, that the relations between genes and traits are many-to-
many, not one-to-one. Assuming we have a single capacity for detecting
moral truth, its overwhelmingly likely that its genetic basis consists of
several (possibly many) loci, and that the alleles at these loci influence a
whole spectrum of other characteristics. Natural selection would favor the
best combination of such characteristics, and the capacity for moral knowl-
edge might be part of the package quite independently of its direct effects
on survival and reproduction. Matters are even worse for the debunkers if
theres no single capacity but rather the environmentally (culturally) con-
tingent interaction among several distinct psychological abilities.
Finally, the general thesis that evolution is all about the selection of
underlying genes is belied by the upsurge of recent work in evo-devo,
by niche construction theoryand by the earlier ventures pioneered by
Oyama, Lewontin, Griffiths, and Gray. Like the earliest sociobiologists, the
debunkers know nothing of cultural selection, nothing of developmental
constraints and the perils of adaptationism, and, unfortunately, nothing of
causal democracy either.

P r o x i m at e a n d U lt i m at e I n f or m at i o n i n B i ol o g y [97]

Bringing Real Realism BackHome

A PerspectivalSlant


When it comes to debates on realism in science, Philip Kitchers

(2001a) Real Realism: The Galilean Strategy (henceforth abbrevi-
ated RR) occupies its well-deserved place among my top five must-read
articles published in the past forty years or so on the topic, alongside
Putnams (1975) What Is Realism?; Boyds (1991) Realism, Anti-
Foundationalism and the Enthusiasm for Natural Kinds; Laudans
(1981) A Confutation of Convergent Realism; and Psilloss (2000)
The Present State of the Scientific Realism Debate. Personal as this
top-five list may be, there is no doubt that Real Realism has ushered
in a silent revolution. Without much fanfare it has shown that realism
is hard to resist because it begins at home and it never ventures into
the metaphysical never-never-lands to which antirealists are so keen
to banish their opponents (RR 191). Kitcher has taught us how real-
ism began with homely considerations such as those used by Galileo
to persuade the Venetians about the reliability of his telescope to spot
ships approaching the harbor. The following step, from being a reliable
naval instrument to being a reliable instrument, in generalcapable
of revealing the craters of the Moon, the satellites of Jupiter, and the
phases of Venuswas a short one.
The Galilean strategy that Kitcher has so admirably defended in Real
Realism against both empiricism and constructivism (in their respective
semantic and epistemic forms) entices us to a homely line of thought and
warns us against any Grand Metaphysical Conclusions. Its impact cannot
be underestimated. We all stand on Galileos shoulders with our defiant
trust in science and technology to give us access to nature and its inner-
most secrets (pace empiricists intimations against 1-kilogram mortar and
King Kongs ability to break it). More to the point, we all stand on Kitchers
Galilean grid in thinking of realism as a homely enterprise, where a divide
et impera strategy of working posits and idle wheels can guarantee to
the selective realist a cornucopia of past scientific results. Where to go
Closer home is my reply. Whose home? The very home from which
Kitcher (1981) began his intellectual journey in the early 1980s, with his
reflections on explanatory unification as the battleground of two grand tra-
ditions: the Aristotelian tradition, whereby scientists aim to fathom the
order of being, an order that is typically opposed to the order of knowing;
and the Humean tradition (continued by Mach, Duhem, and the logical
empiricists), which, on the contrary, argued for no joints at which nature
can be carved, no objective necessities, no mind-independent causal con-
nections (Kitcher 1986, 202). Against both traditions Kitcher defended a
via media, leading out from Kants writings on the methodology of science
onto the philosophy of science. Central to the Kantian project envisaged
by the early Kitcher was an analysis of scientific knowledge and objective
understanding that does not depend on any mind-independent notions of
causation, natural necessity, or natural kind (204). Yet fifteen years later,
in Real Realism, Kitcher took a stance against the Kantian traditionin
its epistemological constructivist outfitholding that the realists world
is an inaccessible realm of noumena (RR 188).
In this essay Isuggest bringing real realism closer home, namely back to
its Kantian roots. The very same roots that make real realism a homely
kind of realism, against any Grand Metaphysical Conclusions about the
world, its causal necessities, and natural kinds. In particular Isuggest rein-
terpreting a key aspect of real realismthat is, the notion of success at
stake in working positsalong more homely lines, lines that acknowl-
edge historical continuity, conceptual nuances, and our role as epistemic
agents in assessing success and inferring truth. (For some preliminary
reflections, see Massimi 2012, 2014.) The result is a form of perspectival
realismto adopt Ron Gieres (2006, 2013)terminologywhich is, how-
ever, already at a distance from what Giere himself intends by this term

B r i n g i n g R e a l R e a l i s m B ac k H o m e [99]
(see Massimi 2015a). Hence my very own (loosely Kantian-inspired) per-
spectivalist slant to real realism.
Key to the Galilean strategyas I see it through Kantian lenses (see
Massimi 2010)is not just to deploy the telescope to overcome fictitious
boundaries (i.e., those between sea and land, Venice and Amsterdam,
Heaven and Earth) but also to approach nature through principles of rea-
son in one hand and experiments thought out in accordance with these
principles in the other hand, yet in order to be instructed by nature not
like a pupil, who has recited to him whatever the teacher wants to say, but
like an appointed judge who compels witnesses to answer the questions he
puts to them (Kant [178187] 1997, Bxiiixiv). This is how Kant famously
portrayed Galileos contribution to bringing natural science onto the secure
path of knowledge after groping about for so many centuries. It is this
further Galilean strategy that Iturn my attention tohere.
I cannot do justice to the breadth of the philosophical arguments that
Kitchers real realism has put forward. And much as Id like to discuss
Kitchers articulated response both to the epistemological empiricism
of van Fraassenean flavor and to the epistemological constructivism of
Kantian descent,1 Ihave to leave those for another occasion. For here Icon-
centrate on Kitchers influential response against the blockish holism
of epistemological empiricism in its historical form (best expressed by
Laudan 1981), which seems to assume that a theory is false because it is
not entirely true (RR 170). In reply real realism insists that the past suc-
cesses stem from parts of the theories that are approximately correct, (RR
170)namely from those hypotheses that are genuinely put to work (i.e.,that
characterize working posits), and are as such approximatelytrue.
In section 2 Ireview Kitchers famous distinction between working pos-
its and idle wheels in the context of his realist defense against the challenge
coming from the history of science. In section 3 Ifocus on the notion of sci-
entific success and distinguish between two variants:success from within
and success from above. In section 4 Isuggest a perspectivalist take on
real realism in the form of a notion of success from within, able to assess
success from a human vantage point and to capture truth across scientific
perspectives. Iconclude by considering possible objections and replies to
the perspectival view canvassed in section4.

1. Against the epistemological constructivists of Kantian descent invoking a dis-

tinction between objects-as-experienced and objects-in-themselves (RR 189), real
realism responds that the objects we claim to represent accurately are not mysterious
noumena but, in many cases, the things with which we interact all the time(189).

[100] The Philosophy of Philip Kitcher


A powerful line of argument against realism has traditionally rehearsed a

seemingly compelling historical point against the success to truth infer-
ence:that similar inferences made by our predecessors would have issued
in conclusions we now take to be quite wrong (RR 168). Let us leave aside
whether the list of past successful yet false theories reflect historical records
or abide instead by the antirealist inclination to inflate examples. There are
undoubtedly prominent cases from the history of science in which views
we now take to be false were genuinely successful by anyones standards
(168). Not surprisingly perhaps, Fresnels wave theory of light is one such
favorite example. No matter how false the ether theory is, Fresnels ability
to use his mathematical equations to predict a bright spot in the middle
of a dark shade won skeptics like Poisson in the Paris Academy of Sciences
and belies epistemological empiricists objection to realism. To the eyes of
real realists, Fresnels wave theory of light is successful not in virtue of a
tenuous distinction between structure and substance (pace structural real-
ists). Instead its success was achieved via approximately true descriptions
of some of the features of light waves (the mathematical accounts) while
being wrong about others (RR 170)(i.e., how light waves propagate in the
ether). The real realist sees Fresnel as employing many tokens of light
wave to refer to electromagnetic waves and as saying a large number of
approximately true things about the properties of electromagnetic waves
of the appropriate type, despite the false opinion about the propagation of
the waves through an elastic ether (RR170).
Against the blockish holism of the antirealist that would invite us to
regard as false a past theory that is no longer true by our own standards,
the real realist recommends a divide et impera approach. Working posits are
approximately true as long as they explain why past theories were suc-
cessful (to the extent that they were), while idle wheels are theoretical
excrescences that are incorrect (RR 170)and often entangled with work-
ing posits. Fresnel might not have distinguished between the two, but it is
not a foregone conclusion that it would have been impossible for him to do
so. Contemporary selective realists have made their own the real realists
distinction between working posits and idle wheels. And the distinction
continues to be, in my view, one of the most persuasive replies against anti-
realist challenges coming from the history of science.
However, a difficulty still awaits. For the objection against the struc-
tural realistthat the structure/substance dichotomy cannot easily be

B r i n g i n g R e a l R e a l i s m B ac k H o m e [101]
exported to other examplescan similarly be leveled against the real
realist. Consider, for example, Aristotles theory of free fall as accelerated
motion toward a natural place. The theory was undoubtedly successful
by its own lights at the time, and it provided a springboard for medieval
commentators (from Simplicius to Hipparchus and the Arabic commenta-
tors), whose views fed into the impetus theory of Buridan and Oresme, and
ultimately into Galileos early Pisan studies on free fall (see Massimi 2010,
2015b). What are the working posits in Aristotles theory of free fall? And
where do theoretical excrescences begin? Was Aristotles hypothesis that
bodies get heavier nearer the Earth, an idle wheel? Well, it provided an
explanationin Aristotles own scientific perspectiveof why free-falling
bodies accelerate (as opposed to decelerate or move with constant speed)
when moving toward their natural place (where, he assumed, bodies would
regain their form). Moreover it suggested that there might have been
forces acting on the body and pulling it either toward its natural place or in
some different direction (what Avicenna and Abl-Barakt called natural
and violent mail and what Buridan called impetus, as an intrinsic force due
to a natural gravity, which was in turn the ancestor of the early Galileos
gravitas as a weight-related concept and ultimately of Newtons gravita-
tional mass). Was there anything approximately (or even remotely) true in
Aristotles theory? Or should we conclude that Aristotles theory was quite
simply false? What has gone wrong with this example?



Here is a possible diagnosis. In replying to Laudans challenge to convergent

realism, the real realist has himself employed too stringent a criterion in
the divide et impera strategy: a criterion of success from above rather than
from within. In Fresnels case, current electromagnetic theory provides
the criterion of success to discern between the working posits of Fresnels
theory (i.e., the equations for polarization by reflection, which still bear his
name) and the idle wheels of the ether theory (long gone from contempo-
rary textbooks). In Aristotles case, the criterion of success from above
cannot similarly be deployed to distinguish between working posits and
idle wheels. Too many centuries separate the Galilean-Newtonian theory of
free fall from Aristotles, and all the conceptual nuances, small theoretical
steps, and turning corners that the notion of free fall underwent in that
span have long been forgotten (were it not for the assiduous work of dedi-
cated historians of science).

[102] The Philosophy of Philip Kitcher

It may well be that Aristotles theory is as promising as Fresnels when it
comes to identifying parts of the theory that are essential to success (work-
ing posits), and hence approximately true. Scientists in Hipparchuss time
(or even in Buridans time) might still have been able to identify such parts.
But we no longer are, because two millennia separate us from Aristotle.
Thus if my diagnosis is correct, there is nothing wrong with the real realists
divide et impera strategy. What has gone wrong instead in the example of
Aristotles free fall is the real realists tacit appeal to the scientific/conver-
gent realists criterion of success from above, that is, from our very own
current vantage point as if that vantage point were the best one to assess the
past, or the one that provides a royal road to Truth with capitalT.
The perspectival realist (of Kantian leaning) enters the scene. For the
perspectival realist (of the kind Ilike) would rejoin that there is no privi-
leged vantage point from which to assess scientific claims of the past. Our
current vantage point is not a disguised Nagelian view from nowhere, pro-
viding special epistemic standards for assessing the past or a privileged
access to the ontology of nature. Our current scientific perspective is only
one among many others that our ancestors have happened to occupy and
from which failure and success can be evaluated. Homely perspectival con-
siderations of this kind invite us to embrace a more modest criterion of
success from within when it comes to discerning between working posits
and idle wheels.
The perspectival realist may adopt a Galilean strategy, namely the one
adopted by Galileo in his early treatment of free fall in the Pisan trea-
tise De Motu antiquiora (ca. 1590s), before he discovered the law of free
fall (s:t2). Against Aristotles cause of motion (i.e., motion toward a natu-
ral place), Galileo looked for the true cause (vera causa) of accelerated
motion in an Archimedean theory of buoyancy that could explain why
bodies move up or down. But the analogy with Archimedean buoyancy
could explain only uniform (not accelerated) motion. Hence Galileo had to
resort to the medieval impetus theory of Buridan and Oresme in thinking
of a weight-related concept of gravity (gravitas) as an internal static force
that would decay during the free fall. Galileos momentum gravitatis (some-
times also referred to as impeto) is already at a distance from medieval
impetus theory, as it is from Newtons gravity, understood as an external
impressed force acting at a distance between two bodies. Galileos gravity
is still a weight-related internal force, compared to Newtons thoroughly
dynamical concept of gravity. Yet Galileo had to rethink the medieval con-
cept of an internal force and make it obey indubitable principles so as to
demonstrate the law of free fall. (For full details of this story, Irefer the
reader to Massimi2010.)

B r i n g i n g R e a l R e a l i s m B ac k H o m e [103]
Galileos kinematic studies exemplify the perspectivalist strat-
egy of engaging with the past from within (rather than from above).
Working with the Aristotelian tradition that goes from Hipparchus to
the Arabic commentators and Buridan and Oresmes impetus theory,
Galileo could operate within well-trodden paths. He could resort to
Archimedess buoyancy and Hipparchuss theory of free fall and intro-
duce gradual changes to key concepts. For example, the change from
impetus as an internal force propelling a body to momento (momen-
tum gravitatis) as an internal force that, after having propelled the
body, would gradually decay, causing the body to acquire degrees of
speed (celeritatis momenta) in its descent. Galileos breakthrough about
free fall did not happen by debunking the Aristotelian tradition (pace
Galileos own rhetoric against Simplicius in Two New Sciences). Nor did
it happen by selecting working posits in the Aristotelian tradition, for
even Archimedean buoyancy and Hipparchuss theory were inextrica-
bly entangled with idle wheels and not amenable to being imported
tout court into the Galileanstory.
Instead the Galilean kinematic strategy consisted in small theoretical
steps and subtle conceptual nuances that ultimately allowed Galileo to turn
the corner from the Aristotelian tradition. Galileos ability to interrogate
nature with principles of reason on the one hand (i.e., the indubitable prin-
ciples from which he demonstrated the law of free fall) and with experi-
ments thought out in accordance with these principles on the other hand
(i.e., both thought experiments with chords and real experiments with
inclined planes) made the revolutionary shift possible. The perspectival
realist can appeal to this Galilean strategy to bring the real realists notion
of success back home:from above to within.


But how should the perspectivalist notion of success from within be under-
stood? So far Ihave simply suggested that it should not be understood as
the ability of inquirers to identify parts of a theory that are essential to
success and hence approximately true. But this can hardly be enough to
understand the perspectivalist move Iam suggesting for real realism. We
need to unpack the slogan.
In what follows Itake my cue from broader discussions on perspectiv-
alism in contemporary epistemology to propose that success from within
should be understood as success with respect to standards of performance
adequacy appropriate to the scientific perspective of the inquirer when

[104] The Philosophy of Philip Kitcher

assessed from the point of view of another (either diachronically subsequent or
synchronically rival) scientific perspective.2
Given the Aristotelian-Archimedean perspective and the available evi-
dence for free fall, Galileo could conclude that the Aristotelians failed to
satisfy standards of performance adequacy appropriate to their own per-
spective in the explanation of the phenomenon. For example, Aristotles
theory could not explain the precise mechanism through which equal
degrees of speed accrued during the descent, and hence why motion
toward a natural place was uniformly accelerated motion. The Aristotelians
could not obviously be blamed for having believed what they believed
about free fall (i.e., that it was motion toward a natural place) given their
own scientific perspective. And the proposition that free fall was motion
toward a natural place cannot be regarded as relatively true (i.e., true for
the Aristotelians but false for Galileo), on pain of abandoning realism alto-
gether for alethic relativism.
Scientific perspectives, I suggest, provide contexts of assessments for
scientific claims. Galileo could assess the Aristotelian claims about free
fall and find them lacking in satisfying what, from Galileos own per-
spective, were the standards of performance adequacy appropriate to the
Aristotelian epistemic context (e.g., Why is free fall as motion toward a
natural place accelerated motion and not uniform motion, as one should
expect from the analogy with Archimedean buoyancy?). In answering
these questions Galileo came eventually to establish a new scientific per-
spective, from which it became possible to evaluate new claims about
free fall. His scientific perspective, in turn, can be found lacking in sat-
isfying what, from our own current perspective, are the standards of

2. See, for example, Sosas perspectival coherentism (part of his virtue perspectivism
in Sosa 1991), where the justification for beliefs is a matter of perspectival coherence.
Along similar lines, on perspectival justification for beliefs, please see Haack (1993)
and Rosenberg (2002, 149):The reason that we correctly judge that S does not know
that p is that, given our richer informational state, we recognize that what we are (stip-
ulatively) entitled to take to be Ss epistemic circumstances demand a higher level of
scrutiny than we are supposing S himself to have exercised. S therefore, has not satis-
fied what, from our perspective, are the standards of performance-adequacy appro-
priate to his epistemic circumstances, and hence, from our epistemic perspective, we
judge that, despite his not having acted irresponsibly given the information available
to him (judged from his own legitimate perspective on his epistemic circumstances),
he has not justifiably come to believe that p. In what follows Ilatch onto and expand
upon Rosenbergs appeal to standards of performance adequacy, but in a different con-
text and with a different purpose in mind. My goal is not to elaborate a perspectivalist
theory of belief justification but instead to elaborate a perspectivalist notion of success
from within that can serve the purpose of success-to-truth-inferences in the realism

B r i n g i n g R e a l R e a l i s m B ac k H o m e [105]
performance adequacy appropriate to the Galilean-Newtonian epistemic
context (e.g., How to think of Galilean free fall when sense impressions
about the free mobility of rigid bodies and paths of light rays get called
into question, as they were with Helmholtzs mirror sphere thought
experiment, for example? What becomes of Galilean-Newtonian gravity
in a non-Euclidean space?).
Success from within is then the ability of a theory to perform adequately
with respect to standards that are appropriate to the theorys wider
epistemic contextor scientific perspective, as I prefer to call itwhen
assessed from the point of view not just of the scientific perspective at stake but,
crucially, from the point of view of other scientific perspectives. Building on
recent important work in epistemology,3 I suggest the following definition.
A scientific claim (SC) meets the criterion of success from withiniff:

(a) SC expresses a proposition p at scientific perspectiveSP1

(b) p is true (i.e., corresponds to states of affairs in nature) and meets stan-
dards of performance adequacy in SP1 when assessed from other scien-
tific perspectives SP2, SP3,SP4

This definition of success from within vindicates the real realists expecta-
tions for successful posits to track truths in nature (via the first part of
premise (b)). Yet it is perspectival in giving up on both a Nagelian view from
nowhere and a convergent realists/real realists view from here now: it does
not take our currently successful scientific claims as the gold standard for
assessing past failures and successes.
Success from within is kosher to the Kantian spirit of perspectivalism
in giving due consideration to epistemic agents (or, I should say, scientific
communities) commitment to scientific claims (without dismissing them
out of hand as sheer errors of the past). Success from within does justice to
historians anti-W higgish plea for judging past theories in their own terms
and by their own standards (not by ours) when assessed from the point of view
of other (diachronically subsequent or synchronically rival) scientific perspec-
tives. At the same time, it avoids the perils of truth relativism by anchoring

3. Here Iwant to latch onto the helpful distinction between context of use and con-
text of assessment in discussions on relativized truth and faultless disagreement. See
MacFarlane (2005, 2009)and Marques (2014), among many others. By contrast with
MacFarlane, I will not be using this distinction to defend any notion of relativized
truth. Instead Imake use of MacFarlanes distinction between context of use and con-
text of assessment to provide a notion of success in science that does not beg the ques-
tion for scientific realism (i.e., that does not judge past theories on the basis of our
current successful theories).

[106] The Philosophy of Philip Kitcher

success to the truth of perspective-independent states of affairs. That free
fall is accelerated motion is a perspective-independent state of affairs that
either holds in nature or does not. But that accelerated motion is, in turn,
motion toward a natural place (as opposed to motion due to a force of grav-
ity) is a scientific claim that can be assessed only within a given epistemic
context, with its standards of performance adequacy, and so forth. At the
same time, standards of performance adequacy in and of themselves cannot
be entrusted with the goal of delivering success from within.
For example, it is not enough for Aristotelians to be satisfied with their
own theory of free fall (and their associated standards of performance ade-
quacy) for it to count as successful. Scientific communities cannot ratify
their own success if their practices are not deemed successful by other com-
munities of inquirers. Yet other communities cannot in turn smuggle in
their own standards of performance adequacy when evaluating other (past
or rival) theories. Ascientific claim proves inadequate (and hence unsuc-
cessful) when the content of the claim is false and it fails to meet its own
standards of performance adequacy when assessed from another perspective.4
It was possible for Galileo to assess Aristotles theory of free fall by
Aristotles own standards (expressed by Simplicius in Two New Sciences)
and conclude about its inadequacy. As it was possible for William Thomson
(later known as Lord Kelvin) in 1847 to assess Carnots cycle by Carnots
own standards (which included conservation of caloric) and conclude
about its inadequacy (when combined with Joules claim that a quantity of
heat proportional to the mechanical work produced must be consumed in
a paddle-wheel experiment).5

4. On closer reflection, this is what is to be expected from Kuhnian anomalies and

periods of crisis. Anomalies reveal cracks in well-established and well-trodden para-
digms by revealing the inability of the paradigm to handle in its own terms an increasing
number of persistent problems.
5. Following up on Thomson, in 1850 Rudolf Clausius laid the foundations of ther-
modynamics (let us call it SP2) by reconciling Carnots cycle with Joules ideas. For
Clausius envisaged that it was possible to retain Carnots idea that heat passes from a
hot reservoir to a cold one whenever mechanical work is done in a cyclic process, while
also abandoning Carnots additional claim about conservation of caloric. The second
law of thermodynamics was born:in any cyclic transformation of thermal energy into
mechanical energy, a portion of heat gets dissipated irreversibly (pace caloric theory).
Subsequent perspectives, such as Maxwell-Boltzmann statistical mechanics (let us call
it SP3) were still able to evaluate the truth of Carnots cycle and its ability to meet stan-
dards of performance adequacy in its own time (i.e., measuring engines efficiency in
producing mechanical work). Yet in the light of the richer informational content avail-
able to Maxwell and Boltzmann (after Clausiuss introduction of entropy), Carnots
overall claim was deemed as requiring a higher level of scrutiny (about conservation
of caloric and the nature of heat) than Carnot himself could have possibly exercised in
the early nineteenth century.

B r i n g i n g R e a l R e a l i s m B ac k H o m e [107]
These examples show important features about the definition of success
from within Ijustgave:

1. The relevant standards of performance adequacy for scientific claims are

settled in the original context of use, that is, in the scientific perspective
in which the claim is first formulated and advanced.
2. Subsequent perspectives provide contexts of assessment from which it
is still possible to evaluate past scientific claims by their own original
3. Given the richer informational content available to subsequent perspec-
tives, it may be possible for later assessors to regard the performance
adequacy of past claims as lacking in some respects; hence it is possible
for later assessors to either retain or withdraw (in whole or in part) past
scientific claims on the basis of their continuing performance adequacy.

Success from within then becomes a commitment that a community of

epistemic agents undertakes to retain past scientific claims when their per-
formance adequacy continues to be regarded as satisfactory from the point
of view of later scientific perspectives. Success from within bears important
similarities with the real realists working posits. Both react against the
blockish holism of epistemological empiricism. Against scientific realism
they both invite us to a more nuanced reappraisal of past theories. Against
structural realism they both refrain from cashing out success in terms of
structure versus substance. And both equally stress what might be called
the enactive nature of scientific success: success is whatever works or con-
tinues to perform adequately.
Yet real realists and perspectival realists differ when it comes to the
notion of success. For real realists deploy working posits to identify
hypotheses that are approximately true by the criterion of success from
above. Whereas perspectival realists of the kind I like would urge to deploy
success from within to identify scientific claims thatby being justifiably
retained in the shift from the original perspective/context of use to another
perspective/context of assessmentwe have reasons for thinking of as
true (to the best of our knowledge). The perspectival slant I am offering to
the real realists working posits is then in terms of commitment of a scien-
tific community to assess and justifiably retain past scientific claims when-
ever their performance adequacy continues to be deemed satisfactory by
their own original standards when assessed from the vantage point of other
On this perspectival reading truth is not an ex post facto explanation
of the ongoing success of some scientific claims. Instead truth is built into

[108] The Philosophy of Philip Kitcher

the aforementioned definition from the ground up, with the first conjunct
of premise (b):p is true. It is the truth of the propositional content of
a scientific claim together with the ability of the claim to meet standards
of performance adequacy at SP1 (when assessed from other scientific per-
spectives) that ultimately ground success from within. And Ido not mean
explanatorily ground it. I mean instead that ontologically ground it. If
the propositional content of the claim were false by realist lights (i.e., if
there were no such a thing as p in nature), even if the claim were hypotheti-
cally able to meet standards of performance adequacy when assessed from
other perspectives, the claim would not qualify as successful under the cri-
terion of success from within.
Imagine a scientific community in the eighteenth century that could have
built a perfectly consistent scientific system around caloric to advance vari-
ous claims about the production of mechanical work, thermal expansion,
and matters states of aggregation, among others. Our best eighteenth-
century scientists failed to distinguish among states of aggregation as
physical in nature (and went on to identify water as a liquid chemical sub-
stance; see Kuhn 1990). But let us assume that our hypothetical commu-
nity can do better than our own Lavoisier, Dalton, and Carnot and come up
with a perfectly good system of knowledge around caloric that meets their
own standards of performance adequacy at thetime.
For example, such system proves consistent when offering explanations
in terms of caloric for matters states of aggregation and the production of
mechanical work; it gives simple and elegant accounts of how caloric (by
being released and absorbed) underlies all these phenomena; it seems accu-
rate with respect to the evidence available to the community at the time;
and so on. Should we not judgefrom our own current perspectivesuch
a community as having met its own standards of performance adequacy?
Should we not assess its scientific claims as being successful, despite their
propositional contents (in terms of caloric) being false? More to the point,
who are we to conclude that their propositional content is indeed false? Are
not we reintroducing a much-dreaded view from nowhere to reach such a
cross-perspectival Grand Metaphysical Conclusion?


Not so fast. What needs be considered in this imaginary case is whether

positing caloric can indeed give rise to such a perfect system of knowl-
edge able to meet all the aforementioned standards by the light of the
hypothetical eighteenth-century community. I contend that it cannot.

B r i n g i n g R e a l R e a l i s m B ac k H o m e [109]
Consistent explanations first. Assuming caloric is an imponderable fluid
as eighteenth-century scientists did and our hypothetical scientists would
presumably also dowould immediately pose severe challenges to any
attempt to provide a consistent explanation of mechanical work and states
of aggregation. Mechanical work would require caloric to be consumed (pace
conservation of caloric), as much as turning water into ice would require
removing caloric (qua a shell of imponderable fluid surrounding waters
particles) and yet expanding the overall volume. How can waters particles
lose part of their volume (by releasing caloric), while also expanding their
overall volume? Caloric does not seem to license consistent explanations.
Simplicity next (a notoriously slippery standard, if any). Would caloric
provide a simple and elegant account of various phenomena? Caloric could
be squeezed out of particles volumes (assuming a Daltonian model) and
get reattached to them at ease. As simple as that? Well, assuming some
mechanism was in place to explain what held caloric attached to the parti-
cles of matter, what had the power to detach it from matter and reattach it
at will, and so forth. Perhaps some attractive and repulsive forces might do
the trick. Or perhaps electrical fluids. Or some ethereal substratum (along
the lines of Kants matter of heat). Simplicity is not within easy reach.
Acomplex story would have to be told about the mechanisms underlying
calorics behavior in all these phenomena, mechanisms that can potentially
be at odds with eachother.
Perhaps accuracy with the available evidence fares better than consis-
tency and simplicity when it comes to standards of performance adequacy.
Let us assume our hypothetical community has produced a system of scien-
tific claims that are accurate by the experimental standards available to the
community at the time. Such claims must surely be regarded as successful
(no matter how false caloric is from our current vantage point). An analogy
may help here. Suppose I have an accurate story about hedgehogs living
in my garden and creeping out at night to collect the mulberries that have
fallen on the ground. My story is so accurate that it tells me with precision
that hedgehogs come at night, between 1 and 2 a.m., from the far right cor-
ner of the garden, behind the hedge, and collect only the juiciest mulberries
they can get their spiky claws on. So my available evidence of red mulber-
ries on the ground seems to support the accuracy of mystory.
But is accuracy such a malleable standard? Surely, even my garden
hedgehogs would have to respond to some mundane questions: Do they
come out every night? From 1 a.m. or from 2 a.m.? What about the purple
spots on the ground that look like old juiciest mulberries getting moldy?
Accuracy (be it the accuracy of a measurement or the accuracy of a scientific
claim) comes always in tandem with other standards, such as consistency,

[110] The Philosophy of Philip Kitcher

fruitfulness, and explanatory power. Our imagined eighteenth-century
caloric supporters would have to tell a pretty convincing story about how
their scientific claims involving caloric were accurate over and above fitting
a sample of observed regularities (especially if such sample proved in con-
flict with others, and the caloric mechanisms envisaged in each case were
in contradiction with each other and hard to pindown).
Whatever the standards of this hypothetical eighteenth-century com-
munity could have been (the list above, of Kuhnian flavor, is only illustra-
tive and is not meant to be exhaustive), the examples should make it clear
that building a system of scientific claims on an ontologically false ground
is not going to go very far. The system of claims would soon fail by its very
own standards of performance adequacy. And we do not have to resort to
hypothetical scenarios. Real historical communities that entertained stan-
dards similar to the ones listed above came to realize the inadequacy of the
caloric theory in the nineteenth century.
Let us take stock. Acritic was envisaged that challenged the criterion
of success from within on the ground that as long as a community can
justifiably be regarded as meeting its own standards of performance ade-
quacy, the scientific claims advanced on behalf of such standards should
count as successful (despite their propositional contents being false). This
objection attacks the realist component in my definition of success from
within, namely the first conjunct in premise (b):p is true. For it would
seem possible for p to be false and yet still meet standards of performance
adequacy in a given scientific perspective so that scientific claims about p
would count as successful (despite p being false). In response Ihave shown
how if p were false, it would prove in practice impossible to justifiably
meet standards of performance adequacy in a given scientific perspective.
And for good reasons too:ex falso quodlibet. An ontologically false ground
(e.g., caloric) cannot possibly license scientific claims that are arguably
consistent, simple, accurate, and so on (unless inconsistency, inaccuracy,
and so forth are themselves acceptable standards within a particular sci-
entific perspective).6 Thus Iconclude that if the propositional content of
the scientific claim were false by realist lights (i.e., if there were no such a
thing as p in nature, and hence the first conjunct of my premise (b)did not
hold), it would in practice be impossible for the claim to meet standards of

6. Here a relativist may come to the fore and make this kind of rejoinder. (One is
reminded of the familiar story about the Azande and their witchcraft and how stan-
dards of adequacy vary from one epistemic community to another. See Kusch 2002 for
a helpful discussion.) Adiscussion of relativism would lead me into territory farther
afield from the topic of my essay here, and as such Iwill not pursueit.

B r i n g i n g R e a l R e a l i s m B ac k H o m e [111]
performance adequacy in its own scientific perspective (and even more so
when assessed from other perspectives). In other words, it is not the case
that p is false and nonetheless meets standards of performance adequacy
in a given perspective. Ascientific claim of this kind would not satisfy the
criterion of success from within.
A different kind of worry may be raised at this point. Isnt the truth of
the propositional content p enough to secure success from within? Arent
the standards of performance adequacy themselves idle wheels, not neces-
sary to secure success? Here a different critic is envisaged, who may retort
that a real realists working posits ultimately underpin the truth of p, and
my definition of success from within collapses onto the real realists success
from above at a closer inspection. The critic may insist that Fresnels theory
worked and proved successful not because it met standards of performance
adequacy in Fresnels time (e.g., it was fruitful in predicting novel phenom-
ena; it seemed accurate in explaining polarization by reflection; and so
forth) but because Fresnels light wave referred to electromagnetic waves
of high frequency. Or better, Fresnels theory met those standards because
its working posits (i.e., electromagnetic waves) were true. And to empha-
size the idleness of the standards of performance adequacy themselves,
one could easily invoke consistency with the ether theory as an example.
(Yes, Fresnels theory was consistent with popular ether theories at the
time, yet consistency in and of itself does not cut any ice for the success of
Fresnels theory.)
In reply one may consider what would happen to a lone researcher who
gets it right without yet meeting the standards of performance adequacy
of her community at the time (perhaps because such community has not
quite gotten to the stage of precisifying standards able to capture the truth
of what the lone researcher has just discovered). Should we conclude that
the researcher has been successful? Here Icannot help but share Richard
Boyds negative answer to this question,7 although I give a perspectival
gloss to what he portrays as the social dimension of scientific inquiry. That
p is true is not sufficient by itself to ontologically ground success, unless p
also meets standards of performance adequacy at SP1 when assessed from
another scientific perspective.
Consider the astronomer V.M. Slipher, who, at the Lowell Observatory
in Arizona throughout 191217, was able to measure with precision the
radial velocity of galaxies and to empirically establish that galaxies were

7. Boyd (2010, 21718) describes the hypothetical scenario of the lone researcher
who gets it right, but she does not make any contribution to the reliability of our sci-
entific practice unless her success is also recognized as such by a community.

[112] The Philosophy of Philip Kitcher

expanding a decade before Hubble found the law for this phenomenon,
and at a time when Einstein was introducing his cosmological constant in
the equations of general relativity to secure a static universe.8 Slipher was
the lone brilliant experimentalist who got it right in a scientific perspec-
tive dominated by general relativity with Einsteins and de Sitters inter-
pretation of the field equations as implying static solutions. It was only in
1924 that Friedmann first, and then Lematre in 1927 introduced models
of general relativity that implied non-static solutions to the field equa-
tions, with Hubble introducing the law to measure the redshifting of gal-
axies in 1929. Once the idea of an expanding universe became a live option
for scientists, it also became possible to go back to Sliphers experimen-
tal findings and to assess them as meeting the standards of performance
adequacy of the scientific perspective of the time. For example, their
consistency with non-static solutions of Einsteins field equations that
Friedmann and other cosmologists were bringing to the fore in the 1920s.
Using Supernova Ia techniques, current cosmologists can still assess the
performance adequacy of Sliphers findings, despite his pioneering work
being overlooked for a long time by his own peers, who had not yet precisi-
fied the standards of performance adequacy appropriate to their scientific
To conclude, success from within does not fall back onto success from
above because ontologically true grounds in and of themselves (without also
meeting standards of performance adequacy at the time) are necessary but
not sufficient to license success. Scientific success is what a community of
epistemic agents acknowledges and welcomes as such at any given time.
The truth of the propositional contents of our scientific claimsthe first
conjunct in my premise (b)by itself would grant only a view of success
from nowhere, a view that no epistemic community (either here now or back
then) would recognize as its own.
Success from within has both a realist and a perspectival component.
Correspondence with perspective-independent states of affairs and meet-
ing perspectival standards of performance adequacy (which can be assessed
by other agents across perspectives) are both key to the success of our sci-
entific claims (of today and of the past). Success from within is not the suc-
cess of those who historically happened to be the winners. It is instead the
success of those who were responsible for the scientific findings and their
ongoing performance adequacy as still assessed by ustoday.

8. Here Idraw on John Peacocks account of this episode in Massimi and Peacock

B r i n g i n g R e a l R e a l i s m B ac k H o m e [113]

Fifteen years after Real Realism we are all heirs of Galileos strategy. We
learned from Kitchers real realism how to tell truth from falsehood, how
to discern bits that work from idle wheels, and most of all how to believe in
the reliability of the deliverances of our instruments. Empiricists and con-
structivists of all stripes owe us an argument for maintaining a justifiable
degree of skepticism about science and its success. More to the point, they
owe us an argument for justifiably retreating into metaphysical never-
never-lands on the face of so many homely arguments for being realists
about the things with which we interact all thetime.
For myself and for my generation real realism has enticed us to explore
new avenues and encouraged us to appraise success and failure across the
history of science in a careful way. Maybe success from above should leave
room for success from within. We stand on Galileos shoulders by acknowl-
edging our continuity with the past and our ability to assess past scientific
claims by their own lights and from our current vantage point, a vantage
point that is neither metaphysically nor epistemically privileged. That is
how, in my view, a perspectival slant can help us bring real realism back to
the Kantian home, to which it naturally belongs.


I thank the editors for giving me the opportunity to contribute to this

volume and for careful editorial comments. I am very grateful to Philip
Kitcher for reading earlier versions of this essay and providing illuminat-
ing comments on the homely arguments that make perspectival realism
akin to real realism, as well as on their points of departure. This project
has received funding from the European Research Council (ERC) under
the European Unions Horizon 2020 research and innovation programme
(grant agreement European Consolidator Grant H2020-ERC-2014-CoG
647272 Perspectival Realism. Science, Knowledge, and Truth from a Human
Vantage Point).

Reply toMassimi

Michela Massimi offers a subtle response to my real realism, in the spirit

of the valuablebut these days all too raretradition of history and

[114] The Philosophy of Philip Kitcher

philosophy of science. She begins with what she identifies as a problem for
my views and proceeds to develop a form of perspectivism aimed at fur-
ther articulating real realism. Iagree that real realism might profitably be
elaborated by incorporating perspectivist ideas and that this might enable
a more adequate treatment of the science of thepast.
It is helpful to start with a brief review of the difficulty that my notions
of working posits and idle wheels were intended to solve. In his classic
paper, Larry Laudan (1981) claimed that the confident realist defense of
the approximate truth of mature sciences, by appeal to the success of
those sciences, was vulnerable to a straightforward historical challenge.
According to Laudan, the scientific past is full of theories that are now
regarded as false but that were taken to be true on the basis of their sup-
posed successes. What right do we have to hold that our predicament is
different? We look around and see success everywherebut so did our pre-
decessors, even though their theories werefalse.
In my 1993a book Ioffered a double reply to Laudan. First, Icharged that
his list of false theories successful in their own time was inflated. Second,
Iproposed that, in the instances of genuinely successful past practices, the
theories in question were not entirely falseand the successes are to be
explained by noting that they depend only on claims that remain, by our
lights, true. Hence the thought that success depends on truth is restored.
Real Realism (Kitcher 2001a) goes further by trying to locate the success-
to-truth inference in the homely background of which both Massimi and
The divide-and-conquer strategy is to pick out the working posits of
past science, the true claims and the genuine bits of nature to which our
predecessors referred, seeing those as responsible for the past successes.
These working posits are distinguished from the idle wheels, the constit-
uents of past theories that played no positive role in successful practice.
The example of Fresnels wave theory of light is exemplary. Fresnel was
correct to view light as having a wavelike aspect and to characterize wave
propagation as he did. He erred in supposing that any wave motion must
have a medium in which the waves are propagated and thus introducing an
One of Massimis worries is that this line of solution will not apply broadly
enough. She cites the example of theories of free fall in the Aristotelian
tradition. On her account real realism faces a dilemma: either one must
import elements from later science into Aristotles framework, or its nec-
essary to dismiss his theory as thoroughly false. The former option distorts
the content of Aristotelian claims; the latter dismisses Aristotelianism too

B r i n g i n g R e a l R e a l i s m B ac k H o m e [115]
I reply by trying to specify the exact claims at issue. In the Aristotelian
tradition the terms we translate as free fall pick out the motions of bodies
after their release as they tend toward the earths surface. Some contexts
will fix the reference of those terms by importing theoretical ideas from
Aristotle; for example, free fall will be taken to be a motion toward a natural
place. On other occasions, however, thinkers in the Aristotelian tradition
refrain from any such theoretical imposition:free fall is just the downward
motion of a body that has been released. Call these the strong and the
weak senses, respectively. Consider now the statement Bodies in free fall
accelerate as they approach the earths surface. When Aristotelians make
claims we properly translate by this statement (and, of course, they dont
use accelerate!), everything turns on the sense we assign to free fall. In the
strong sense, if the logical form is taken to be a universal generalization,
the claim is vacuously true, since nothing satisfies the antecedent. In the
weak sense, however, the claim is a true empirical generalization, one that
might be used to make some modest predictive successes. On this read-
ing Aristotelians are claiming (correctly) that, when released, heavy bodies
gain velocity in their motion toward the earth. Its thus possible for real
realism, in its original form, to endorse parts of the Aristotelian view as
correct and as predictively successful in consequence of their correctness.
The approach just sketched appears to make progress with a problem
Massimi raises for me, but it remains historically crude, unable to make
sense of the subtle conceptual shifts that lead from Hipparchus through
Buridan to Galileo (early and mature). But before Ioutline some measures
for refining the historical treatment, introducing a form of perspectivism
akin to Massimis, its important to see how the problem of understand-
ing the conceptual shifts is different from that of responding to Laudans
actual skeptical argument.
Historians, and historically sensitive philosophers, often writhe when
they encounter judgments about the truth of particular claims made in
the scientific past, especially when that past is quite distant. By what right
do we adopt a view from nowhere, judging the correspondence of (say)
Aristotles words with nature? The answer is that real realists dont adopt
a view from nowhere but a view from right here, and they do so because its
forced on them by the skeptical challenge. Laudan and the antirealists who
follow him introduce that view. Heres the challenge:You realists believe
that success indicates truth. But there are many past successful scientific
theories that you regard, by your own lights, as false, so success cant reli-
ably indicate truth. Evaluating bits and pieces of past science as correct
from the perspective of contemporary science is an intrinsic part of any
response to Laudans celebrated argument.

[116] The Philosophy of Philip Kitcher

Massimis perspectivism steps sideways from Laudans actual challenge
although, as Ill explain shortly, theres something important that moti-
vates her. Before attending to that motivation, however, I want to point
out an interesting feature of her perspectivism. Her condition (b) seems
to make a stronger claim than is pertinent to Laudan-style skepticism:
the first conjunct (the truth of the proposition) is enough. Why, then, add
the second, which might hold all pieces of past science hostage to some
thoroughly misguided and regressive scientific perspective, one so badly
informed that it judges all the rest to fail in their own terms? Perhaps the
condition might be more plausible as a disjunction rather than a conjunc-
tionan interpretive idea inspired by her remark about lack of success.
(A scientific claim proves inadequate (and hence unsuccessful) when the
content of the claim is false and it fails to meet its own standards of per-
formance adequacy when assessed from another perspective. The negation of
(b), as it stands, should be a disjunction.)
But this is to miss the insight motivating Massimi. She (reasonably)
wants (b) to approach truth-claims via using other perspectives to judge
the performance adequacy of the focal perspective by its own lights (avoid-
ing the view from nowhere and denying any privileged role to the view
from here). As Ive said, to proceed along those lines fails to mesh with
Laudans skeptical argument as actually posed. On Massimis view, as I
understand it, Laudans deep interest in destabilizing our contemporary
views should lead him to be suspicious of the notion of success he actually
adopts in his critique of convergent realism: he should not be content with
evaluation from above but should ask the realist to show that success
depends on truth, when the truth of elements of past science are assessed
from within. The convergent realist (or the real realist) ought to show that
the evaluation is stable as one proceeds from the original practice through
the sequence of practices that succeed it. In this way both the critic and
the realist avoid taking any standpoint as privileged. Massimi claims that
Laudans challenge should have asked for something stronger, namely the
stability of the judgments about success and truth under evaluation from
within. Combining my crude original approach to Aristotelian language
with the more subtle schemes of translation available from perspectives
intermediate between Aristotle and us (e.g., those of Hipparchus, Buridan,
and others), I think the realist can meet this stronger challenge.
In my view the deep motivation for introducing the idea of evaluation
from within stems from Kuhns seminal reflections on the languages of past
science. When you examine the subtle ways in which later Aristotelians
discussed free fall (reviewed in some of Massimis illuminating historical
studies), the translations Ioffered in evading her original dilemma seem

B r i n g i n g R e a l R e a l i s m B ac k H o m e [117]
remarkably blunt. Following Kuhn (especially 2000), we might declare the
central claims of Aristotelians to be literally untranslatable. Id prefer to
put the Kuhnian point differently:we can arrive at approximative transla-
tions, highly context-dependent and requiring preliminary glosses to show
the ways Aristotelian and modern terms cut across one another. Neither
the Kuhnian version nor my preferred alternative vitiates the basic point
that Aristotles successors achieved an insight we can best capture with the
formulation Bodies in free fall accelerate as they approach the earths sur-
faceand its precisely by attributing that insight to them that Laudan-
style skepticism is answered (in the particular instance).
Real realism needs extension, in my view, because it should appreci-
ate the limitations of the translations advanced in combating skepticism.
Massimi is right to suppose that, through a sequence of theoretical devel-
opments of a large perspective (say, Aristotelianism), the more immediate
descendants are more able to reconstruct and explain the ideas of their pre-
decessors than those who come later:once the working posits and idle wheels
have been identified (by us!), Buridan is a better interpreter of the terms
used by Hipparchus in characterizing them than is Galileo, and Galileo, in
turn, does better than we can. If we can attribute a core insight about accel-
eration toward the earth (as Ithink we can), we must also recognize that
the Aristotelian terms in which that insight is expressed are alien to us
and that they are less strange to those who are closer to the Aristotelian
Recognition invites the idea of a sequence of perspectives, distinguished
by ways of conceptualizing the phenomena (in this instance, phenomena
of motion), in which close successors are better able to capture the claims
of their predecessors. To rebut Laudans skepticism its enough to show
that where theres success theres an underlying use of correct ideas, often
expressed in what much later scientists see as highly peculiar ways. A much
deeper understanding of the phenomena of success, and how its won,
requires the historian-philosopher-of-science to reconstruct the perspec-
tives of the past, tracing their continuities with the present.
I read Massimi as aiming to avoid any privileging of the contemporary
standpoint, the view from here. For me, the principal motivation for per-
spectivism stems from the shortcomings of the translations used in recon-
structing the successes of past science. In the rest of this response Ill trace
a route to my preferred version of perspectivism.
Massimi draws from Kant (as Ionce did). My sources these days are the
classical pragmatists. On my interpretation of Peirce, James, and Dewey,
none of them rejects my favorite (post-Tarskian) version of correspon-
dence truth for scientific statements (see Kitcher 2012c, ch. 5). Yet James

[118] The Philosophy of Philip Kitcher

and Dewey (especially) are struck by the plurality of frameworks in which
scientific truths can be embedded. Both of them view the worldin the
rich sense in which the world contains determinate objects divided into
kindsas partially constructed by the community of inquirers, with some
features of that construction reflecting aspects of human psychology and
others responding to our evolving interests. So at different stages in the
development of a science, alternative languages will be articulated, specify-
ing different spatiotemporal boundaries for objects, different groupings of
things into kinds, different termini for processes, and different standards
for normality. None of these languages is privileged in the sense of con-
forming to the intrinsic structure of an independent reality. Conceived as
what is independent of the subject, reality doesnt come with that much
structure. With respect to different human purposes, however, some lan-
guages may function better than others.
A thorough evaluation of some past perspective would start with recon-
structing its language, exposing the ways its categories cut across those
of the present. It would proceed to delineate the goals at which inquiry
aimed, what questions investigators selected as especially significant, and
what standards were adduced for answering them. On this basis it would
explore the extent to which those investigators succeeded in attaining their
goals, why they were successful when they were, and why they failed when
So far the historian-philosopher-of-science is using the contemporary
perspective to study and to appreciate a past perspective in its own terms.
Mindful, however, of the large degree to which the languages cut across one
another, with consequent inadequacy of translation, a historically sensitive
philosopher may explore the intermediate perspectives, using the more
immediate descendants as superior guides to the structure of the ances-
tor. In the end the full sequence of perspectives may serve as the most
adequate standard against which ancestral success is judged. If Iam right,
that conforms in important respects to the perspectivism Massimi prefers.
It is, however, couched in a different philosophical idiom and directed at a
different philosophical problem.
In my view real realism fights on many fronts. As my 2001a paper tries
to show, real realism opposes several varieties of empiricism and construc-
tivism. Laudans skepticism is akin to one strain of empiricist argument,
and it is among the objections I have been concerned to rebut. My cur-
rent view, however, is that Kuhns ideas about conceptual incommensu-
rability, particularly as they were elaborated in his later writings (Kuhn
2000), offer an equally important challenge, demanding of real realists
that they offer a more nuanced view of scientific success, past and present.

B r i n g i n g R e a l R e a l i s m B ac k H o m e [119]
The neopragmatist perspectivism Ihave sketched attempts to take up that
In the end, Ithink, a perspectivalist real realism is doubly motivated
and we dont have to choose which rationale is more important. Massimi
views Laudans use of evaluation from above as betraying one of his cen-
tral insights; thus she introduces a more probing account of success and
elaborates real realism with respect to it. Ihave been more troubled by a
Kuhnian challenge. But we come out in much the same place. This is not
so much because of the resurgence of my lapsed Kantianism as through
celebration of a characterization Kuhn came to relish:he was redoing Kant
with movable categories. So too, Ibelieve, were the classical pragmatists.
Real realism should continue the enterprise.

[120] The Philosophy of Philip Kitcher


Unificationism, Explanatory
Internalism, and Autonomy


Philip Kitcher has made an extraordinary number of distinguished contri-

butions to philosophy of science and to many other areas of philosophy.
Here Ifocus on just one of his projects in philosophy of science, although
one that has been very influential: his development of a unificationist
account of explanation. Although I will engage critically with some of
Kitchers ideas on this subject, I also want to roam more widely, taking
up some broader issues about the role of unification in explanation, the
contrast between internalist and externalist approaches to explana-
tion (in the sense of Kim 1994), and claims about the autonomy of the
special sciences and how these interact with ideas about unification. Iwill
explore these issues against the background of the interventionist account
of explanation I have defended elsewhere (Woodward 2003). The reader
should thus think of what follows as an investigation of some Kitcherian
themes regarding explanation rather than as a critical study that focuses
just on Kitcherswork.
The remainder of this essay is organized as follows. I begin (section
2)with a brief summary of Kitchers ideas about the role of unification in
explanation. I then (section 3) turn to an overview of some of the prin-
cipal claims for which Iwill argue in the hope that this will help to guide
the reader through what follows. Section 4 takes up some issues regarding
internalism, and section 5 defends the importance of incorporating inter-
nalist as well as externalist considerations in models of explanation.
Sections 6 and 7 distinguish two kinds of explanatory projects connected
to unification.


Kitcher describes himself as a deductive chauvinist; he retains the

Hempelian idea that explanation involves constructing deductively valid
derivations of explananda from true premises, although not Hempels
idea that one of these premises must be a law. Kitcher adds to Hempels
account constraints having to do with unification:the derivation must be
an instance of an argument pattern that is more unifying than alternative
patterns. An argument pattern is a schematic argument (the result of tak-
ing a deductively valid argument and replacing some or all of its nonlogical
vocabulary with dummy variables) together with a set of instructions spec-
ifying various permissible ways of instantiating or filling in the dummy
variables. Argument patterns can differ in their stringency in the sense of
imposing more or less strong restrictions on the arguments that instanti-
ate the patterns. Roughly speaking, Kitchers guiding idea is that explana-
tion is a matter of deriving descriptions of many different phenomena by
using as few and as stringent argument patterns as possible over and over
againthe fewer the patterns used, the more stringent they are, and the
greater the range of different conclusions derived, the more unified our
explanations. He summarizes this idea as follows: Science advances our
understanding of nature by showing us how to derive descriptions of many
phenomena, using the same pattern of derivation again and again, and in
demonstrating this, it teaches us how to reduce the number of facts we
have to accept as ultimate (Kitcher 1989,432).
An important part of Kitchers strategy for defending this account
involves showing that the derivations we regard as good explanations are
instances of patterns that, taken together, score better according to the
criteria just described than the patterns instantiated by the derivations
we regard as defective explanations. For example, our present explanatory
practicescall these Pare committed to the idea that derivations of a
flagpoles height (h) from the length of its shadow (l) are not explanatory.
Kitcher compares P with alternative systemizations in which h is derived
from premises that include l. According to Kitcher (1989, 485), P includes
the use of a single origin and development (OD) pattern of explanation,
according to which the dimensions of objectsartifacts, mountains, stars,

[122] The Philosophy of Philip Kitcher

organismsare traced to the conditions under which the object origi-
nated and the modifications it has subsequently undergone. Now consider
the consequences of adding to P an additional pattern S (the shadow pat-
tern) that permits the derivation of the dimensions of objects from facts
about their shadows. Since the OD pattern already permits the derivation
of all facts about the dimensions of objects, the addition of the shadow pat-
tern S to P will increase the number of argument patterns in P but will not
allow us to derive any new conclusions. On the other hand, if we were to
drop OD from P and replace it with the shadow pattern, we would have no
net change in the number of patterns in P but would be able to derive far
fewer conclusions than with OD, since many objects do not have shadows
(or enough shadows) from which to derive all of their dimensions. Thus OD
achieves a high degree of unification in comparison with alternatives, and
this is why we regard it as an acceptable part of our explanatory practice.
Asimilar justification is provided for other familiar features of explanatory
practice; for example, our dissatisfaction with explanations that contain
irrelevancies is understood in terms of the idea that such explanations are
less unifying than alternatives not containing irrelevancies.
What is the role of causation in this account? Kitcher (1989, 477)claims
that the because of causation is always derivative from the because of
explanation. That is, our causal judgments simply reflect the explanatory
relationships that fall out of our (or our intellectual ancestors) attempts to
construct unified theories of nature. There is no independent causal order
over and above this that our explanations must reflect.
Kitcher deploys his unificationist model in support of the idea that there
are autonomous levels of explanation in the special sciences and to argue
against the reductionist view that the claims made in upper-level theories
are always best explained by some lower-level reducing theory.1 Roughly
speaking, this is because the upper-level theory may do a better job of unify-
ing than the lower-level theory and thus can provide superior explanations.
For example, according to Kitcher, the upper-level theory of classical genetics
and associated generalizations, such as the law of independent assortment,
are not explained (or at least not best explained) by the lower-level theory of
molecular biology. This is because phenomena that appear heterogeneous

1. Talk of levels has come in for a good deal of well-deserved criticism recently. My
view is that there is a relatively innocuous way of understanding this notion:think of
it as a way of capturing the idea that certain factors (within some range of variation) do
not make a difference to other factors or relationships. When this is the case the latter
can be regarded as at a different level than the former. It is this understanding that
Iadopt in this essay. Levels sometimes but by no means always track differences in the
spatial or temporal scale at which processes occur; seebelow.

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [123]
or disunified from the point of view of molecular biology are treated in a
much more unified fashion in classical genetics, with the same argument
patterns, formulated in the vocabulary of that theory, being used repeat-
edly to derive a range of different results. This unified pattern would be lost
if we relied solely on derivations from molecular biological premises.



As should be apparent from the preceding summary, Kitcher thinks of uni-

fication as central to understanding current science and as a regulative ideal
to which science aspires. Ihave considerable sympathy for this idea (or at
least something in its neighborhood),2 siding in this respect with Kitcher
against others (e.g., Cartwright 1999; Dupr 1993) who instead stress
the disunity of science. Ialso fully agree with Kitcher that unification is
important to explanation, in the sense that it seems central to how at least
some paradigmatic scientific explanations work that they unify. However,
as recent discussion (see especially, Morrison 2000) has made clear, the
relationship between unification and explanation is a complicated one. To
begin with, there are a number of different activities and achievements in
science that in some sense have to do with unification, and only some of
these are plausibly associated with explanation. For example, there are
many cases in which scientists devise ways of representing phenomena
that previously were described in diverse and unrelated ways within a com-
mon classificatory or representational schemehence achieving a kind of
unification. Moreover, knowing the place of some item in the classificatory
scheme, one may be able to derive various additional facts about the item in
question, so that the unificationist ideal of deriving a lot from a more lim-
ited number of premises is arguably satisfied. However, such schemes are
often regarded as merely descriptive rather than explanatory. Examples
include schemes for biological classificationknowing an animal is a mam-
mal and a primate allows one to predict a number of its other properties
but arguably does not explain why it possesses those properties. Asimilar
observation holds for schemes for the classification of stars, such as the
MorganKeenan system.

2. Although Ithink Kitcher is right to insist on the importance of the establishment

of connections and relationships between parts of scientific knowledge, Idont think
these always take the form of unification via deduction; many other forms of integra-
tion and constraint are also common.

[124] The Philosophy of Philip Kitcher

A related point can be made regarding discoveries that different phe-
nomena can be modeled within a common mathematical framework;
molecular Brownian motion and aspects of the behavior of stock prices
can be modeled using the same mathematics, but this does not amount to
construction of a unified explanation of phenomena in these two domains.
Discussions of explanatory unifications in physics typically emphasize that
this requires much more than finding formal or mathematical connections
or analogies between the phenomena being unified or subsumption within
a common formal framework (Maudlin 1996; Morrison 2000). Something
moreelusively described as the discovery of physical connections or rela-
tionships between the phenomena being unifiedis also required. This is
one reason, as Ishall argue, an appeal to argument patterns by itself does
not seem to fully capture what is going on in successful explanatory uni-
fications; something having to do with external relationships out there
in the world (and which the argument patterns track or represent) is also
required. One of the main tasks of an account of explanatory unification
should be to elucidate the distinction between those cases of unification
that are explanatory and those that are not. Iwill advance some brief and
very incomplete suggestions about this (connected to interventionist
ideas), but Ithink that we are very far from having an adequate treatment.
In addition to these considerations, Iwill argue that it is important to
distinguish (at least) two different kinds of explanatory undertakings in
which a connection between unification and explanation is present. One
sort of undertakingcall this explanatory unification1 (EU1)involves
explaining a large number of different phenomena in terms of just a few
causes or explanatory factors. Newtons unification of terrestrial and celes-
tial motions due to the action of the single cause of gravity is a paradigm
of this sort of achievement. In such cases unification is achieved in the
sense that the many apparently different phenomena are shown to depend
on a small number of explanatory factors or relationships; the latter are
(perhaps unexpectedly, prior to the construction of the unifying theory)
shown to be explanatorily relevant to the former. Kitchers emphasis on the
repeated use of a small number of argument patterns in achieving unifica-
tion seems aimed at capturing cases of this sort. As we shall see, however,
EU1 is often tied to (or used to motivate) successful reduction, which makes
it a somewhat problematic vehicle for establishing antireductionist conclu-
sions of the sort to which Kitcher is sympathetic.
By contrast, a second sort of explanatory projectcall it EU2also
involves (what can be thought of as a) kind of unification, but here, in
contrast to EU1, the establishment of facts about the (relative or partial)
irrelevance or independence or autonomy of certain relationships from others

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [125]
plays a central role. (We might say that the guiding focus on EU1 is relevance
or dependence, and that of EU2 irrelevance or independence.) In one very
common kind of case, EU2s explain or demonstrate or at least make use of
or exploit the independence of various upper-level relationships that figure
in the special sciences from lower-level microdetails about their realizers;
they show or make it understandable why those upper-level relationships
turn out to be stable or invariant across various changes or variations in
other sorts of factors, including those involving microdetails. To employ
an illustration discussed in more detail below, renormalization techniques
explain (in the sense of EU2) why materials of many different sorts, dif-
fering in microphysical details, exhibit similar generic behavior near their
critical points. As another illustration, the method of arbitrary functions
and its elaborations explain why gambling devices of different design and
material composition exhibit similar behavior with respect to the relative
frequencies they exhibit. Kitchers claims about the irrelevance of (many
of) the underlying molecular details to the generalizations of classical
genetics (independent assortment of nonhomologous chromosomes, etc.)
can, Ibelieve, be naturally assimilated to cases of this sort. Because of this
focus on the irrelevance of microdetails, EU2 projects are often bound up
with antireductionist themes about the relative autonomy or independence
of the relationships that are the subject matter of the special sciences. Such
independence can enable or make possible theorizing that seems correctly
describable as having a unificatory aspect or feel to it, since it involves
generalization across or abstraction from irrelevant microdetails. However,
the focus of this sort of unificatory achievement seems different in impor-
tant respects from what is achieved in EU1 projects and, I will suggest,
involves features that are perhaps not so well captured by Kitchers offi-
cial theory of unification, although they are fairly well captured by various
other, more informal observations ofhis.


Any theory of explanation needs to provide a characterization of the

explanatory relation (or relations) R between explanans and explanan-
dum such that the former explains the latter. One fundamental contrast
is whether R is characterized in internalist or externalist terms, in the
sense of Kim (1994). As I will discuss in more detail, Kitchers account
appears to be, at least in some respects, internalist, in the sense that it
makes use of comparisons that are internal to our corpus of knowledge

[126] The Philosophy of Philip Kitcher

(in particular, considerations having to do with the comparative unifying
power of different possible argument patterns) in characterizing R. By con-
trast, Ifavor taking (what Iregard as) a different notionthat of difference-
makingas the starting point for understanding the explanatory relation
R. I advocate understanding difference-making in interventionist terms,
which in turn yields an externalist characterization of R, as having to do
with relationships in the world external to our knowledge. Iwill suggest
that there are a number of features of explanatory practice that require for
their explication a notion of difference-making that is independent of the
notion of unification.
In support of taking difference-making as a point of departure, Iwould
argue that, whatever else an explanation should do, it should convey
information about factors and relationships that make a difference to its
explanandum and that it should not represent as difference-makers fac-
tors that are non-difference-makers or irrelevant. This focus on difference-
making (rather than unification) as a starting point is not meant to suggest
that unification is unimportant in understanding how explanatory practice
in science works. Rather Ihold that we should use the notion of difference-
making in order to elucidate the role that the various explanatory enter-
prises associated with unification play in science. As Isee it, EU1s involve
appeal to difference- making relationships with certain additional fea-
tures:difference-making relations in which the same kind of factor figures
as a difference-maker for many different phenomena. Similarly difference-
making also plays a central role in characterizing EU2s, but here it is the
absence of certain kinds of difference-making relations (i.e., the irrelevance
of certain factors to others) that is crucial.
Although Ihold that an adequate account of explanation must have an
externalist component, Ialso think (and here Iside with Kitcher against
Kim [1994] and against Salmon [1984]) that such an account must include
internalist (or epistemic) components as well. Thus my view is that the
most adequate model of explanation will show how both of these elements
work together cooperatively. In particular such characteristically epistemic
concerns as the character of the representations we employ when we con-
struct explanations and how these track or capture difference-making
relationships, highlighting certain of these and backgrounding others,
are of central importance in explanation. So are computational consid-
erations:whether we can actually carry out and exhibit certain computa-
tions and derivations. Ithink it is an important virtue of Kitchers work
on explanation that it is more sensitive to these considerations than exclu-
sively externalist or ontic approaches.

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [127]

Earlier I said that difference-making or dependency relations are central

to explanation; when one factor or variable X plays a role in the expla-
nation of a second factor or variable, variations in some of the values
of X will make a difference for values of Y in some background circum-
stances. More precisely, in some background circumstances B there will
be a pair of values of X, x and x< x, and a pair of values for Y, y and
y y, such that when X = x, Y = y, and when X = x, Y = y. It seems uncon-
troversial that many difference-making relations are causal; a cause is natu-
rally understood as something that makes a difference for its effect, at least
when other conditions are appropriately controlled for. When difference-
making relations are causal, my preferred explication is in terms of what
happens under interventions; that is, X makes a difference for Y when there
is a possible intervention that changes the value of X such that under that
intervention the value of X is different in some background circumstances B.
(Here an intervention on X with respect to Y is an unconfounded change
in X that changes Y, if at all, only through this change in X and not in some
other way. For more detail, see Woodward 2003.) If, as I am inclined to
think, there are noncausal forms of explanation (or explanations that
embody noncausal features), it is likely that we will need to understand
the notion of difference-making appropriate to them in some other way
besides via appeal to the notion of an intervention, but I advance no pro-
posals here about how to do this.
I have observed elsewhere that difference-making/dependency relation-
ships between X and Y may differ in their degree of stability or invariance
or in the extent to which they are independent of changes in other condi-
tions. At one extreme, interventions on X may be associated with changes
in Y under some very narrow range of background conditions or for some
very narrow range of changes in X and Y, but this relation may not hold at
all outside of these conditions. At the other extreme a difference-making
relationship between X and Y may be such that it continues to hold over
a large range of changes in other conditions. Difference-making relation-
ships currently known in the special sciences, including biology, typically
are at best stable under some range of background conditions and not oth-
ers, rather than holding universally in the sense of being stable under
all physically possible conditions. Other things being equal, we prefer (for
explanatory purposes) generalizations that describe difference- making
relations that are stable or invariant under a relatively wide range of varia-
tions in other factors; these will be generalizations having the kind of (rela-
tive) independence or autonomy discussedabove.

[128] The Philosophy of Philip Kitcher

A key feature of difference-making/dependency relations, as Iconceive
them, is that they can occur at different levels of generality and abstract-
ness. Moreover difference-making relations at different levels will be rel-
evant to (figure in the explanation of) different explananda. For example,
if our target explanandum E is the present exact position and momentum
configuration C of all the molecules making up a mole of gas, the difference
makers for E will include an exact specification of the momenta and posi-
tions of all component molecules at earlier times and much more besides;
vary any of these factors, and C very likely would be different. Suppose
instead the explanandum in which we are interested is E*, the volume V
of a gas in a cylinder with a piston that is movable in a vertical direction
and on which a weight W rests. If the gas is placed in a heat bath at fixed
temperature T and allowed to expand isothermally, then the value of V will
depend on just a very few macroscopic parameters, including T and W, with
the dependence relation in question being given by macroscopic thermody-
namic relationships like the ideal gas law. The reason for this is roughly that
the variations in the position and momenta of the component molecules
that are consistent with the values of the thermodynamic parameters like
P and T are irrelevant to (are not difference-makers for) the final volume
of the gas. On an account of explanation according to which we explain
by citing difference-making factors and relations but not those that are
non-difference-making, we should not cite these molecular details if what
we want to explain is the volume of the gas.3 Note that, in this case, the
ideal gas law nicely combines the features of independence from (or sta-
bility across changes in) lower-level detail regarding some range of varia-
tion in the position and momenta of the component molecules of the gas
with accurate information about difference-making relations about such
upper-level macroscopic variables as P, V, and T. This is what we want in a
relatively autonomous upper-level relationship.
As another illustration, discussed in more detail below, the behavior
of single neurons can be explained and modeled at many different levels
of detail and abstraction, from detailed studies of the behavior of indi-
vidual dendritic currents in the neuron to models of the behavior of the
whole neuron and firing patterns in response to overall synaptic input.
The Hodgkin-Huxley (H-H) model of the generation of the action poten-
tial in a certain class of neurons shows how the shape of this potential
depends on certain generic features of the circuitry of the neuron: the

3. Or, more weakly, we need not cite these non-difference-making details, and we
should not represent them as difference-making when they arenot.

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [129]
capacitance across the neural membrane, the existence of physically sepa-
rated voltage-dependent ionic currents across the membrane with differ-
ent time courses, and so on. Any neuron with this circuitry conforming to
the differential equations characterizing the H-H model will generate an
action potential under the appropriate conditions, independently of such
matters as the particular ions making up the ionic currents or the particu-
lar molecular mechanism involved in the transport of those ions. These
latter factors are not difference-makers (given that the neuron has the
generic features described above) if what we want to explain is the over-
all shape of the action potential. On the other hand, if what we want to
explain is the opening and closing of particular ion channels and the fac-
tors affecting the transport of ions through them, such molecular details
are relevant difference-makers.


Both of these examples illustrate another important theme:often models

at different levels incorporate difference-making factors that are relevant
to different explananda (e.g., the action potential in the case of the H-H
model and the opening and closing of particular ion channels in the case of
the molecular model). Because their explananda are different, these models
need not be viewed as competing with each other. Thus in order to vindi-
cate the explanatory credentials of the upper-level (e.g., H-H) model, there
is no need to argue that it provides better or more unifying explana-
tions than the lower-level model. For this reason, in the remarks that fol-
low, Iplace less emphasis than Kitcher on the role of competition among
different candidate unificatory patterns in vindicating the autonomy of
upper-level special science generalizations. For example, although Iagree
with Kitcher about the explanatory status of Mendels laws and the relative
autonomy of the explanations in which they figure from molecular details
(section 7), Idont think that Mendels laws are explanatory because they
are more unifying than explanations provided by molecular biology
that is, because they are winners in a competition with molecular biology
with respect to unification achieved. Vindicating the explanatory creden-
tials of Mendels laws or the H-H model or phenomenological thermody-
namics only requires showing that these describe stable difference-making
relationships for the explananda they are intended to explain, not that
they are more unifying than theories and generalizations directed at dif-
ferent explananda.

[130] The Philosophy of Philip Kitcher


In exploring Kitchers ideas about explanatory unification and comparing

them to alternative accounts, a useful entry point is Kims (1988 [2010])
contrast between externalist (or realist or objective) and internalist
(or irrealist) accounts of explanation. As Kim observes, this distinction
closely parallels Salmons (1984) well-known distinction between ontic
and epistemic conceptions of explanation, but I focus on Kim since in
some respects he is more detailed. Kims (1988 [2010], 150)distinction is
as follows:

What matters to [explanatory] realism is that the truth of an explanation

requires an objective relationship between the events involved....
Explanatory irrealism, on the other hand, would be the view that the relation
of being an explanation for, as it relates to C and E within our epistemic corpus,
is not and need not be grounded in any objective relation between events c and
e.It is solely a matter of some internal relation between items of knowledge.

According to Kim (1994, 58), Hempels version of the DN model (at

least in its official statements) is internalist. This is because whether the
requirements of the DN model are satisfied depends on factors internal to
a body of knowledge, not on what goes on in the worldexcept of course
for the truth of the statements comprising the explanans (57). By fac-
tors internal to a body of knowledge, Kim presumably has in mind, in the
case of the DN model, the presence of a relation of deductive entailment
between explanans and explanandum; his point is that one can determine
whether this relation holds simply by the operation of inspecting the prop-
ositions making up the explanans and explanandum, and this is a matter
that is internal to the body of knowledge that these propositions makeup.
The relevance of this to unificationism is that, according to Kim,
Kitchers account of explanation (as well as Michael Friedmans 1974
related account) is also internalist:What makes these derivations [that is,
derivations that unify in the manner described by Kitcher and Friedman]
explanatory is their relationship to other items in our epistemic sys-
tem, not some objective facts about external events or phenomena (Kim
1994, 63). In Kitchers case the relevant internal relations have to do with
the number of argument patterns employed and their stringency rather
than any objective relations holding for events or phenomena involved in
the putative explanations (64). Kim holds that a satisfactory account of

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [131]
explanation should be externalist and criticizes both Hempel and Kitcher
for providing purely internalist models of explanation.
When initially encountered, the internalist/externalist (or epistemic/
ontic) contrast can seem puzzling or at least not entirely perspicuous, since
it is clear that supposedly internalist models like Hempels and Kitchers
also contain commitments that look ontic. For example, on Hempels ver-
sion of the DN model, the requirement that the explanans be true requires
the holding of various external, worldly facts, as Kim himself recognizes.
Moreover, on a natural construal of Hempels views, there must be a corre-
sponding external relation in the world in which the explanandum event is
subsumed under or instantiates the regularity described in the explanans.
Similarly in connection with Kitchers model, although it is true that
which argument patterns are instantiated by a derivation, how unifying
these are, and the stringency of the derivations are matters internal to an
investigators corpus of beliefs, Kitchers model also requires that the deriva-
tion itself appeal to true premises and that some appropriate subsumption
relation be present. It is hard to see how to make sense of these require-
ments without supposing that facts about the way the world is constrain
which are the most unifying and stringent derivations. I might undertake
to construct a theory that unifies true propositions from the theory of juve-
nile delinquency, the astrophysics of the early universe, and the molecular
genetics of C. elegans, but given the regularities that actually obtain in the
world, and the fact that the explanatory generalizations to which my pur-
ported unification appeals need to reflect these regularities, the resulting
theory is unlikely to score very high along the dimensions of successful uni-
fication emphasized in Kitchers model. Similarly one would think it is facts
about the world and the nature of the gravitational force that make it pos-
sible to construct a theory that unifies, according to the criteria described
by Kitcher, the motion of terrestrial and celestial bodies. To the extent that
this is so, why shouldnt we think, contra Kim, of Kitchers theory as having
an externalist as well as an internalist component?
Although these observations seem correct, as far as they go, there is an
important insight behind the distinction that Kim and Salmon are attempt-
ing to draw and that has important implications for Kitchers proposals
about unification. One way of bringing this out is to ask the following
question about a model of explanation:When it comes to characterizing
the explanatory relation(s) R, which (if either) is primary and which is
derivative(i) internal (e.g., deductive) relationships among propo-
sitions or (ii) external or worldly relations? Over and above any truth
requirement we impose on the explanans, can we characterize the explana-
tory relationship R (just) in terms of (i), or does the characterization of R

[132] The Philosophy of Philip Kitcher

require appeal to (ii) in a form that is independent of (i)? On one natural
interpretation of many of Hempels remarks in support of the DN model, he
seems to conceive of the explanatory relation R, as it is in the world, as just
whatever corresponds to or is represented by the DN relation of deductive
entailment via true premises; it is this entailment relation that is primary,
and the external features of R are characterized by reference to it. This is
reflected in the fact that Hempel does not seem to allow that we have any
access to or purchase on the explanatory relationship, as it is in the world,
independently of whatever is captured or represented by the DN entail-
ment relation (or the subsumption relation that automatically accompa-
nies it). Because there is, in this sense, no possibility of a gap or failure of
correspondence between the explanatory relation R conceived externally
and the internal representations of R within the DN framework, we cannot
even raise the question of whether the latter adequately captures or tracks
the former. In other words, when given the strongly internalist construal
described above, there appears to be no room to say that a DN derivation is
explanatory because or to the extent that it traces or represents some inde-
pendently existing relationship R in the world that is relevant to explana-
tion, where this is understood in a way that carries with it the possibility
that some DN derivations may fail to represent or capture this relation. We
see this in Hempels willingness to regard DN derivations running from
effects to causes as explanatory; there is nothing in his official framework
that allows for the possibility that a sound DN running in the wrong direc-
tion may fail to track an independently existing explanatory relationship.
Given this conception of what is at stake in the contrast between inter-
nalism and externalism, is Kitchers model internalist? Im not entirely
sure (and would be interested to hear what Kitcher has to say on this score),
but his claim that the because of causation is entirely derivative from the
because of explanation, with the latter understood entirely in terms of
the comparative unificatory merits of various argument patterns, perhaps
suggests an affirmative answer. As Ihave said, my own view, by contrast,
is that an adequate account of the explanatory relation R must be charac-
terized (at least in part) externally rather than purely internally. Achar-
acterization of R along interventionist lines provides such an external
characterization:R has to do with what would in fact happen in the world
to Y if an intervention on X were to occur, where this involves a worldly
change in X, and the question is whether a similarly real change occurs in Y.
This characterization allows us to frame questions about whether various
representations we may employ asserting the existence of a difference-
making or dependency relation between X and Y are correct or accurate
in the sense of truly describing how Y responds under interventions on X.

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [133]
The internal relations between X and Y (whatever they may be) do not auto-
matically provide a correct answer to this question.
As an illustration that is particularly relevant to both Hempel and
Kitcher, suppose that we are presented with a derivation of some explanan-
dum from premises specified in a candidate explanans, where the deriva-
tion has a DN structure (the derivation is deductively valid, the premises
are true, at least one is a law essential to the derivation, etc.) We may then
ask, within an interventionist framework, whether the (nonnomic) fac-
tors cited in the candidate explanans are such that there are interventions
on those factors that would change the explanandum phenomenon in the
way described by the nomic premises in the derivation; if so, the deriva-
tion can be thought of as tracking or representing the difference-making/
dependency relations between the factors cited in the explanans and the
explanandum phenomenon. However, whether the derivation satisfies this
interventionist requirement depends on what nature is like, and this is not
settled just by whether the derivation meets the official DN requirements.
For example, a derivation running from the height h of a flagpole and the
angle of the sun on the horizon to the length s of the shadow it casts
identifies factors such that interventions changing the value of those fac-
tors will change the value of s; in this sense the derivation can be thought
of as providing information about how the value of s depends on the value
of h and and as demonstrating that (and how) h and are difference-
makers for s. However, a parallel claim is not true regarding a derivation
of the value of h from s and a point that can be established by, for
example, varying (or observing its variation) and seeing whether there
are changes in the value of h.4 Within an interventionist framework this

4. There is a good deal more to be said about the elucidation and identification of
explanatory asymmetries. One consideration is that when one gets the direction of
explanation wrong, this is often reflected in the apparent presence of unexplained
coincidences or correlations that are not present when one gets the direction right.
(To put the same idea in a slightly different way, one gets violations of the requirement
that, in the absence of some special reason for supposing otherwise, the independent
or cause variables in a purported explanatory relationship should be capable of varying
independently of each other and should not exhibit any particular stable correlation.)
In the example above, if one alters the value of (e.g., by tilting the angle between the
pole and the ground) or even just observes the naturally occurring variation over the
course of the day, one will observe a corresponding change in the value of s that occurs
in such a way that the value of h appears to be constant; that is, the values of s and
are correlated, adjusting in just the way that is required to maintain the same value
for h. Furthermore the envisioned explanation provides no explanation of this cor-
relation. By contrast, when the direction of explanation is from h and to s, no such
mysterious correlation is present:the independent variable h is a constant and hence
is uncorrelated with any variations in . Related procedures are used in statistics and
machine learning to identify causal direction in nonexperimental contexts.

[134] The Philosophy of Philip Kitcher

difference underlies our willingness to regard the first derivation but not
the second as explanatory.
This treatment of explanatory asymmetries contrasts with Kitchers
account of their origin. As I noted, Kitcher traces these asymmetries to
differences in the degree of unification achieved by different possible argu-
ment patterns:an account in which s is derived from h and belongs to
an argument pattern that is more unified and better satisfies stringency
requirements than an account in which h is derived from s and . Iwill not
try to argue that this claim about comparative unification is mistaken, but,
as Isee it, the alternative account of the basis for explanatory asymmetries
associated with interventionism has certain advantages:for one thing, it
appeals to considerations that at least in a number of cases are straightfor-
wardly empirically accessible and that do not involve complex comparative
judgments about the stringency and unifying power of different argu-
ment patterns, which of the many patterns instantiated in any particular
case are the appropriate ones to consider, and so on. For example, there
are straightforward experiments that can be done to determine whether,
for example, intervening to alter s will alter h. (Of course since we already
know how these experiments will turn out, we dont think it worthwhile
to do them.) And when we cant do experiments, other sorts of empirical
testing or inference procedures such as those described in note 4 may be
available that are connected to intervention-based considerations and that
can be used to settle questions about causal direction.
A second, closely related point is that once one thinks in terms of an
external, independently characterized relation R, ones conception of the
role or function of whatever internal features F one holds must be present
in an explanation changes; one thinks of their role as having to do with
tracking or representing the independently holding explanatory relation R,
and it is not automatic that they will successfully dothis.
A third point is this:once the (or at least a) goal of successful explana-
tion is viewed in the way just described (representing difference-making/
dependency relations between explanans and explanandum, construed
along interventionist lines, the more stable and invariant the better) there
is no particular reason to believe this goal can be achieved only by the use
of just one kind of representational structure. In particular, if explanations
have to do with the accurate representation of difference-making or depen-
dency relationships, representations besides those involving deductive
relationships among propositions of the sort emphasized by the DN and
kindred (including unificationist) models may also achieve this goal. For
example, directed graphs (including various elaborations and extensions of
these) are an alternative device that is used for the purpose of representing

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [135]
dependency relations in many areas of science, both in the social and
behavioral sciences and in the biological sciences. In such graphs an arrow
drawn from one variable to another (XY) represents that Y depends in
some way on X, but without specifying the exact functional form or param-
eterization of the dependence. For certain explananda this may be all that
is needed for successful explanation.



So far my discussion has emphasized the attractions of thinking of explan-

atory relationships in realist or external terms. Iwant now to suggest,
however, as urged in section 3, that this should not come at the expense
of the neglect of the role of more internal or epistemic considerations
and that it is a great virtue of Kitchers version of unificationism that it
recognizes this. Indeed if one holds (as Ido) that a good explanation needs
to represent or track explanatory relationships and to exhibit (preferably
in detail) how the explanandum phenomenon depends on factors cited in
its explanans, this immediately leads to a focus on internal or epistemic
considerations having to do with the characteristics of the representations
we employ in constructing explanations. For example, if our candidate
explanation employs systems of equations to represent dependency rela-
tions, it matters crucially whether one can actually write down and solve
(at least approximately or via simulation) these equations, do calculations
that result in actual numbers characterizing the behavior of the explanan-
dum phenomenon (or failing that, achieve some qualitative insight into the
behavior of these equations), and so on. It is also often crucial for explana-
tory purposes that one be able to provide arguments justifying the neglect
of or exclusion of certain factors that one might otherwise think are sig-
nificant difference-makers and for focusing on others that represent the
dominant or most important difference-making factors. Particularly in the
case of complex systems with microstructures possessing many degrees of
freedom, it is crucial to find representations that allow for the aggregation
of these microvariables into macrovariables possessing (for the explana-
tory purposes at hand) far fewer degrees of freedom. Principled arguments
for neglecting these many degrees of freedom at the microlevel (arguments
that many of the lower-level details dont matter) are also crucial to success-
ful explanation and are a major concern of model builders in many areas of
science. This is also an enterprise with an important internal or epistemic
dimension. Thus while Ido not agree that all explanations have to take the

[136] The Philosophy of Philip Kitcher

form of deductive arguments or that all there is to successful explanation
is the instantiation of the right sort of internally characterized deductive
structure, Iside with Kitcher and Hempel in holding that internal features
of the representations we employ should be a central concern in any model
of explanation. What we need to think about is how those internal features
work to capture or track external relationships.
To illustrate some of the difficulties that arise when internal features
associated with explanatory representation and derivational structure are
neglected, consider Salmons suggestion, in the course of defending his
ontic theory of explanation, that the mark of a successful explanation is
that it shows how the phenomenon to be explained fits into the causal
structure of the world, where the latter is understood as a (vast) network
of individual causal processes and their interactions, as described in his
Scientific Explanation and the Causal Structure of the World (1984). Return
to the example in which a mole of gas in a cylinder with a movable pis-
ton is placed in a heat bath and allowed to expand isothermally until it
reaches an equilibrium volume V*, which is what we want to explain. Here
the relevant causal processes and intersections are apparently the trajecto-
ries of the individual molecules and their collisions, but even putting aside
the point that many of the details of these trajectories are not difference-
makers for V*, there is obviously no possibility of writing down and solving
the 61023 body problem of the molecular interactions. Instead we need to
find some more tractable way of representing the initial state of the gas
and its subsequent developmenta macroscopic characterization in terms
of just a few variables or degrees of freedom. As Isee it, a central problem
with Salmons talk of showing how the explanandum phenomenon fits into
the causal nexus is that it gives us no guidance about what features such
a macroscopic theory or representation should possess if it is to count as
explanatory of V*. Indeed the natural construal of Salmons view is that the
real explanation of the behavior of the gas (the underlying ontic story)
is at the level of the individual molecular trajectories and interactions; it
is at this level that the causal processes and interactions are to be found.
This construal is reinforced by Salmons repeated claims, contrary to what
is asserted in the textbooks on thermal physics, that macroscopic thermo-
dynamic generalizations like the ideal gas law are nonexplanatory because
In my view this focus on a privileged level of ontic description and lack
of concern about or resources for capturing upper-level dependency rela-
tions in complex systems (and understanding when and how lower-level
detail does not matter) is a common (and perhaps unavoidable) feature of
theories that neglect the internal or epistemic dimension of explanation.

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [137]
For example, this feature is also present in Carl Cravers (2014) recent
defense of an ontic approach to explanation, even though he does not adopt
Salmons specific view of what the ontic involves. For Craver the relevant
ontic facts in the case of biological explanations are apparently (or at least
usually) characterized in molecular/chemical terms. As an illustration,
Cravers version of an explanation of the generation of the action poten-
tial (the spiking activity of a single neuron), which is in the ontic mode, is
the following: The flux of sodium (Na+) and potassium (K+) ions across
the neuronal membrane explains the action potential (31). This is favor-
ably contrasted with models of the generation of the action potential such
as the H-H model, which abstracts away from the molecular details of the
processes by which ions are transported across the neural membrane and
which instead exhibits how the action potential depends on macroscopic
variables characterizing the whole neuron such as the capacitance across
the neural membrane, the total ionic currents, the membrane potential,
and so on. Craver regards the H-H model as at best an explanation sketch
or as merely phenomenological (and hence defective qua explanation)
because of its neglect of molecular detail; he associates the fact that the
generic shape of the action potential is a solution to the H-H equations for
certain parameterizations of those equations with the claim that the model
functions in DN-like fashion to show the action potential is nomically
expectable under certain conditions, but takes this to be a consideration of
merely epistemic rather than ontic significance, since nomic expectability
is an epistemic rather than an ontic notion.
Rather than illustrating the advantages of an ontic approach to explana-
tion, Cravers discussion illustrates the disadvantages of neglecting the epis-
temic or internal dimension of explanation. One relevant consideration is
that different types of neurons exhibit different firing patterns in response
to synaptic input; some exhibit the characteristic firing pattern associated
with generation of the sort of action potential described by the H-H model,
and some do not. Whether a neuron exhibits this firing pattern does not
depend just on whether there are fluxes of Na+ and K+ ions across the cell
membrane but rather on the way these are organized (that is, on the circuit
diagram for the whole neuron, including the physical separation of the Na+
and K+ channels and the fact that these operate according to different time
courses, the fact that the cell membrane is sufficiently insulated to act as a
capacitor, and so on).5 Aneuron that is not organized in this way will not

5. Kitchers discussions in his 1984 and 1999 papers also emphasize the importance
in explanation of considerations having to do with system-level spatial organization
that are often neglected in philosophical defenses of reductionism.

[138] The Philosophy of Philip Kitcher

exhibit action potentials with the shape described above, even if there are
fluxes of Na+ and K+ ions across the cell membrane. Moreover the action
potential occurs in neurons with channels involving other sorts of ions,
such as Ca++ ions, as long as these conform to the requirements of the H-H
model. For these reasons little or no explanation of the action potential is
provided simply by citing the ontic information Craver describes. Nor, con-
trary to what Craver claims, is much of an explanation of the generation
of the action potential provided by supplementing this with information
about the details of molecular mechanisms involved in the opening and
closing of the ion channels. What we want for an explanation of the action
potential is an exhibition of factors on which the action potential depends
(and an exhibition of the overall pattern of dependence) in the sense that
variations in these factors make a difference for whether one gets an action
potential with its characteristic shape rather than some other neuronal
response. The H-H model satisfies this demand; information about the
details of molecular transport mechanisms for ions by itself does not, both
because such mechanisms can be (and are) present in neurons that gener-
ate action potentials and in those that do not, and because, as noted above,
this information by itself leaves out many of the factors that are difference-
makers for the action potential. An important part of the explanatory force
of the H-H model is a demonstration that the function describing the gen-
eral shape of the action potential is a solution to the H-H equations; if such
a demonstration is regarded as of merely epistemic significance (because
it involves a derivation or inference), this simply vindicates the importance
of the epistemic dimension in explanation.
One common objection (made particularly by those who favor exclusive
focus on the ontic dimension) to this emphasis on the epistemic is that it
commits us to a notion of explanation that is too dependent on merely
pragmatic factors or on factors that are too closely tied to human abili-
ties to gather information, measure, deduce, and calculate. Thus it may
be argued that while, as a practical matter, we could never gather infor-
mation about the precise positions and momenta of all of the component
molecules in a mole of gas and use this to calculate these positions and
momenta in the future, this merely reflects a fact about us and our limi-
tations. If we claim that an explanation of the macroscopic behavior of the
gas framed in terms of thermodynamic variables is superior to (or even
as acceptable as) the molecular explanation just described, we are allow-
ing facts about what we can or cant do or know to infect our theory of
explanation, thus relativizing it to anthropocentric facts about our human
epistemic predicament. Asimilar conclusion might be advanced regarding
the H-H model as an account of the action potential:that our need to rely

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [139]
on it merely reflects our parochial intellectual limitations. On this view of
the matter the underlying reality is that the action potential is the product
of facts about the component molecules and atoms making up the neuron
and their relationships, and it is this underlying reality and nothing more
that is relevant for the purposes of explanation. What we can calculate or
measure is irrelevant.
In my view this objection draws on the mistaken idea that internalism
and externalism about explanation are mutually exclusive alternatives and
that the internal features of explanation are entirely disconnected from the
external onesideas rejected above. Although it is true that we lack the
computational ability to derive facts about the behavior of a macroscopic
sample of gas from facts about the trajectory of individual molecules, it
is also true that it is a fact about the world (and not a fact about our com-
putational limitations) that many aspects of the macroscopic behavior of
a dilute gas can be captured by means of relationships among a few mac-
roscopic parameters that are to a very large degree independent of (stable
across) variations in molecular details. Similarly for models of neuronal
behavior that abstract away from molecular details. Thus these models
capture ontic (or worldly or external) facts about macroscopic dependency
relations in these systems. The correct way to think about the relationship
between these ontic facts and epistemic considerations is that the ontic
facts provide us with computational and derivational opportunities that are
then reflected in epistemic or internal features of the models and repre-
sentations we construct.6 In other words, it is because of the existence of
certain stable upper-level patterns (ontic facts) that we are sometimes able
to construct tractable models (reflecting internalist considerations) of the
behavior of the systems that interest us. But this does not mean that these
stable upper-level patterns somehow spring into existence just as a result
of our interests or our computational limitations or as a result of our cogni-
tive organizing activities, or that they are mere projections of our inter-
est in finding tractable models with epistemically pleasing features. Rather
we find or discover preexisting relationships in the world that fit with and
enable the application of our limited cognitive and calculational abilities in
the construction of explanations.
On this view of the matter there are lots of difference-making and
dependency relations in nature; these occur at different levels or scales

6. This picture of nature as providing us with calculational opportunities that we

incorporate into our models is emphasized in Wilson (forthcoming), to which I am
greatly indebted.

[140] The Philosophy of Philip Kitcher

and also differ in the range of circumstances over which they are stable.
Anthropocentric or human-centered considerations may enter into the
explanations we construct in the sense that facts about what we can mea-
sure, calculate, and manipulate, as well as considerations having to do with
what we find interesting and important, can lead us to focus on some of
these relationships and not others and also to focus on some explananda
and not others. However, we should not find this sort of anthropocentrism
disturbing; it is fully compatible with reasonable versions of externalism
or realism regarding the status of explanatory relationships themselves, as
existing independently of our inferential/epistemic activities. We are thus
led to a view of explanation that involves a synthesis of externalist and
internalist themes: external relationships provide opportunities for the
construction of explanations that would not be available in their absence,
but it matters too that the explanations we construct have an internal
structure that tracks or represents those external dependency relations.
Applied to issues about the role of unification in explanation, this sug-
gests an approach that retains Kitchers emphasis on the importance of
epistemic factors but supplements these with ontic elements, understood
along difference-making lines, seeing the latter as supporting (rather than
as an alternative to) the former.



I noted in section 3 that there seem to be several different explanatory

projects or activities that are related to unification. In addition to EU1 proj-
ects exhibiting the dependence of a range of different phenomena on some
small set of explanatory factors, there are also EU2 projects having to do
with claims about (and/or the provision of explanations for) the relative
autonomy and independence of various upper-level dependency relation-
ships, across variations in other factors (relative because such autonomy
is typically partial rather than complete). It is a very general fact about
nature that such independence is rather common (perhaps more common
than one might have expected) and that understanding and recognizing why
and when it occurs, when it might expected, and how it might be exploited
in theory construction is very important in building explanations.
A striking example, discussed in detail by Robert Batterman (e.g., 2001),
is provided by the universal behavior exhibited by a wide variety of dif-
ferent materials, including fluids of different material composition and

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [141]
magnets near their critical points, both characterized by the same criti-
cal exponent b. In the case of fluids, for example, behavior near the criti-
cal point can be characterized in terms of an order parameter S given by
the difference in densities between the liquid and vapor forms of the fluid
S=liq vap. As the temperature T of the system approaches the critical
temperature Tc, S is found to depend upon a power of the reduced tem-
perature t= TTc/T:
where b is the critical exponent referred to above. Remarkably the same
value of b characterizes not just different fluids but also the behavior of
magnets in the transition from ferromagnetic to paramagnetic phases.
Suppose one is interested in explaining why some particular kind of fluid
has the critical point that it does. Since different kinds of fluids have differ-
ent critical points, the value of Tc for any particular fluid will indeed depend
on microphysical details about its material composition.7 However, if one
is instead interested in explaining the universal behavior just described
(the phenomenon or generic fact that S ~ |t|b with fixed b for many differ-
ent materials), then information about the differing microphysical details
of different fluids is irrelevant:within the framework for thinking about
explanation defended above these details are non-difference-making fac-
tors. In other words, the universality of this behavior shows us that its
explanation must be found elsewhere than in details about the differences
in material composition of different fluids. Instead the explanation for this
universal behavior is provided by renormalization group techniques, which
in effect trace the behavior to very generic qualitative features (e.g., certain
symmetries) that are shared by the Hamiltonians governing the interac-
tions occurring in each of the systems, despite the fact these Hamiltonians
differ in detail for each system.
In this case we have a kind of unification since we are shown why a
variety of very different systems exhibit a common or unified pattern of
behavior near their critical points. Isuggest, however, that the kind of uni-
fication achieved seems somewhat different from the sort of unification
(EU1) that is achieved when a number of (apparently) different phenomena
are attributed to the same general type of causal factor. In the case of EU1
we begin with a variety of apparently different phenomena (the orbits of

7. This illustrates the notion that the autonomy of upper-level behavior in such sys-
tems is only partialholding with respect to some explananda but not others.

[142] The Philosophy of Philip Kitcher

different planets, the trajectories of comets, the trajectory of projectiles
near the surface of the earth, etc.). Initially it is not recognized that these
are unified in the sense of being due to the operation of a single type of
causal factor; unification is achieved when this fact is recognized. Moreover
the unification proceeds by recognizing the differences among the different
phenomena explained and then showing how these differences result from
gravitational forces of different magnitudes, conforming to the same gen-
eral law but operating on different initial conditions.
By contrast, in EU2 there is a commonality or universality in the behav-
ior of different systems that is recognizable independently of the discov-
ery of the explanation for this commonality. At least in the case discussed
above, this universality is not (described by a generalization that is) part of
some unifying explanans but rather is seen as the target explanandum:it is
something that is itself explained by appeal to the renormalization group.
Moreover the explanation proceeds by showing that the features of the
individual systems that make them different from one another (e.g., dif-
ferences in chemical composition) are irrelevant to this common behavior,
rather than focusing (or focusing only), as an EU1 explanation would, on
how differences in the behavior of different materials near their critical
points depends on some single type of explanatory factor that operates on
different initial conditions.
Explanations exhibiting the general features just described (EU2s) are
very common in many areas of science. They may be invoked, for example,
when one wants to understand why systems governed by deterministic laws
exhibit stable relative frequencies in coarse-grained behavior. Consider the
behavior of a properly made roulette wheel. If the wheel exhibits the appro-
priate macroscopic symmetries, then, as shown by a series of arguments
initiated by Poincar (the method of arbitrary functions) and continued by
such writers as Hopf and Engel, for a very large class of different possible
dynamics governing the wheel (as long as these satisfy certain very generic
conditions) and for almost any set of macroscopic interventions performed
by the croupier in spinning the wheel (again as long as these satisfy very
weak general conditions), stable relative frequencies will result. Further
details having to do with differences in the materials from which the wheel
is constructed, the precise dynamics governing its behavior, or the behav-
ior of the croupier are irrelevant to (make no difference for) the frequencies
with which various outcomes are generated. Again we have an explanation
for a kind of universality in behavior that involves showing that certain
details are irrelevant given other, far more generic details that are relevant.

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [143]
Depending on the system under investigation, there are many differ-
ent reasons (in addition to those operative in the cases mentioned above)
why various factors (including facts about microstructure) may be irrel-
evant to overall patterns in their behavior. In a large and important range
of cases the irrelevance of certain factors or processes for certain depen-
dency relationships follows from considerations having to do with the
differences among the spatial or temporal or energy scales that are rel-
evant to the behavior of those factors. For example, a process or influence
may either occur so quickly (in comparison with the dependency relation-
ships in which we are interested) or so slowly that we may safely regard it
as irrelevant. Or the influence may operate at length or energy scales that
make it irrelevant to the phenomena we are trying to explain, as when the
details of the behavior of the strong and weak force (which are very short-
ranged) are justifiably ignored in explaining chemical behavior. In particle
physics, processes operative at very high-energy scales are thought to be
irrelevant to many processes operative at lower energy scalesirrelevant
in the sense that many different alternative high-energy theories are con-
sistent with the same low-energy behavior, so that variations in these
make no difference for low-energy behavior. (Various decoupling theo-
rems provide results about the extent of this independence.) This fact
makes particle physics, as currently practiced, possible, enabling the con-
struction of so-called effective theories since really high-energy behavior
is (currently) unobservable. (If finding an adequate low-energy theory
required identifying which high-energy theory is correct, physics would
be stuck.)
Similar considerations (with separations of scale motivating claims
of independence and irrelevance) are very likely operative in connection
with many biological phenomena, although there has been less systematic
exploration of such cases than in physics. For example, because biologi-
cal processes occur on quite different time scales, it is sometimes possi-
ble to treat processes that are slow relative to the process one wants to
understand as approximately constant, hence warranting the assumption
that there are no actual variations in the slow process that are relevant
to the faster process. Similarly it is sometimes reasonable to assume that
certain processes occur very quickly and reach a steady-state equilibrium
relative to some process of interest; again this justifies treating the former
as approximately constant (Voit 2013, 10). Alon (2007, 1011) provides
illustrations:inputs change the activities of gene transcription factors on
a subsecond scale; in contrast binding of the active transcription factor to
its DNA site reaches equilibrium in seconds. Transcription and translation

[144] The Philosophy of Philip Kitcher

of the target gene takes minutes and the accumulation of the protein prod-
uct can take many minutes to hours.



I remarked earlier that in cases like the explanation of universal behav-

ior near the critical point, this universality is treated as an explanandum
rather than as part of an explanans. Of course, as Ialso noted, in the case
of some dependency generalizations it may be possible to both (a)establish
that they are relatively stable and independent of various other factors and
then (b)exploit this stability feature in using the generalization to explain
in the fashion of EU1s. The generalizations of thermodynamics have this
character, and it is arguable, following Kitcher, that generalizations like
Mendels laws also have this status:they are both relatively independent
of certain molecular details and such that they identify factors on which
a range of different phenomena in population genetics depends, so that
they figure in EU1s. Note, though, that the independence (when it obtains)
of these upper-level generalizations from underlying details is an empiri-
cal facta matter of which dependency or relevance relations exist in the
world. This sort of independence does not seem to be something that can
be established (or explained) just on the basis of considerations having to
do with the number and stringency of argument patterns or how much can
be derived from their repeated use. In particular, although once it has been
established that some generalization G is independent in the right way
from variations in other factors, including various low-level details, one
can sometimes then use G to achieve an EU1 treatment of a range of differ-
ent phenomena, showing that G is a premise in such EU1s does not by itself
establish that G has this sort of independence or stability across variation
in microlevel details. For example, although it is arguable that (a)Mendels
laws have a kind of independence from various sorts of molecular details
and although it is also true that (b)Mendels laws figure in EU1s of a range
of phenomena in population biology, it is not the case that (b)provides a
justification for (or explanation of) or basis for belief in (a); (a)is not true
because (b)is. Rather the truth of (a)is something more like a presupposi-
tion for the use of Mendels laws in an EU1. Put differently, explaining why
(or how it can be the case that or justifying the claim that) Mendels laws
have the sort of independence described is an EU2 project that is different

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [145]
from showing that Mendels laws figure in the derivation of a variety of dif-
ferent phenomena, which is an EU1 project.
I conclude with a related observation: I suggested earlier that reduc-
tive explanations seem to be naturally viewed as cases of EU1 or as in part
motivated by the aspiration to construct EU1s. For example, the reduction
of the thermodynamics of dilute gases achieved by statistical mechanics
makes use of the laws of Newtonian mechanics, which of course figure in
the EU1 of many other phenomena. Similarly many of the physical and
chemical generalizations employed to explain aspects of the behavior of
biological systems can also be used to provide EU1 explanations of the
behavior of inorganic systems. Thus if we should prefer those EU1s that
best satisfy Kitchers criteria for explanatory unification, it is not obvious
that we can resist the contention that we should prefer reductive explana-
tions that appeal to statistical mechanics over those that appeal to phe-
nomenological thermodynamics or explanations of biological phenomena
that appeal to more fundamental principles of physics and chemistry over
those that do not. For this reason appeals to EU1 do not seem to be an
entirely convincing way of defending antireductionist theses about biology
or the other special sciences. Isuggest that a better strategy is to appeal to
the sorts of considerations that underlie EU2s:the behavior of many of the
systems that are the subjects of the special sciences simply do not depend
on the factors that figure in underlying microtheories of those systems. To
the extent this is the case, we have all of the justification we need for treat-
ing the sciences of those systems as relatively autonomous.

Reply toWoodward

Besides the qualities he shares with other distinguished philosophers of

scienceclarity, rigor, and originalityJim Woodward is well known for
two distinctive virtues. First is Jims enviably wide scientific range, mani-
fest in his discussions of issues and examples in many fields of natural and
social science. Second is his unusual ability to read sympathetically. Jim
intends to learn from the sources he probes. His goal is not to ascertain
a weakness and move to a swift refutation but to uncover ideas that can
be worked into a more comprehensive vision. He exemplifies philosophical
patience. Along with the more common qualities, these special virtues are
evident in the essay Jim has written for this volume.

[146] The Philosophy of Philip Kitcher

Woodwards (2003) superb book Making Things Happen already contains
a penetrating critical discussion of the unification approach to explana-
tion proposed in my essays Explanatory Unification and Explanatory
Unification and the Causal Structure of the World (Kitcher 1981, 1989).
I have not replied to his objections, and there is no need for me to take
up that task here, for since the 1980s my thinking about explanation has
evolved considerably, largely under my increasing attachment to (some
might call it an infatuation with) pragmatism. Ino longer believe that there
is any general account of explanation. Explanations are of many kinds, and
the things counted as successful explanations embody a variety of virtues.
Sometimes what makes a proposed explanation a good one is the fact
that it specifies a cause that makes a difference (as Woodward suggests).
Sometimes it is its ability to unify a host of disparate phenomena. And
there are several other distinctive qualities.
Bas van Fraassen (1980, ch. 5) took a step in the right direction in
his pragmatic theory of explanation. Wesley Salmon and I(Kitcher and
Salmon 1987)argued that the stated version led to trivialization:anything
can explain anything. We saw the trouble as arising from van Fraassens
failure to specify the relevance relation (the relation that holds between
an explanatory answer and the why question it answers). In later work
my coauthor saw further than I did, recognizing two distinct styles of
scientific explanation (Salmon 1990). During the 1990s it began to dawn
on me that the right way to approach explanation was to allow a number
of different relevance relations (more than two) while insisting that not
any relation counts (the vast majority do not). Reflection on historical
explanation then led me to abandon the thesis that explanation-seeking
questions can be identified with why questions. Consider the following
sample of scientific questions: What are the constituents of eukaryote
cells? Will the universe continue to expand indefinitely? Is there intel-
ligent life elsewhere in the universe? When did human language evolve?
How many species of australopithecines were there, and how were they
related? To what extent can one form a range of silicon compounds that
rivals the diversity of carbon compounds? What is the natural host organ-
ism for the Ebola virus? Can nonhuman animals count? Iview all these
as interesting questions, quite apart from any practical benefits answers
to them might provide. Successfully addressing these questions would
advance our understanding. Hence Iregard them as explanation-seeking
questions, and the answers to them as explanations. Explanation focuses
on a diverse set of questions and provides answers standing in a variety
of relevance relations.

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [147]
How did philosophy of science get sucked into thinking that there must
be some single type of explanation (explaining why) governed by a single
pattern of successful explanation? The answer lies deep in the Humean past
of logical empiricism. One great attraction of Hempels covering-law model
lay in its avoidance of any suspicious reliance on a prior notion of causation.
Indeed logical empiricists often wanted to reconstruct causal talk by iden-
tifying the causes with the factors identified in successful explanations. By
the 1970s it was evident that the Hempelian attempt to characterize scien-
tific explanation had broken down. Those of us who continued to harbor
Humean scruples cast about for new ways of characterizing explanation
independently of causationsometimes explicitly thinking of causation
as the projection of explanatory structure onto the world (Kitcher 1986).
Unification seemed to me an especially attractive possibility for three main
reasons:it promised to provide an account of the theoretical explanation
(systematization) of laws; it resonated with the ways many prominent sci-
entists defended their explanatory proposals; and it offered an embryonic
account of the explanatory relations within mathematics, a topic on which
Ihad made some timid forays (Kitcher 1975; for bolder work in this area,
see Steiner 1978, and now Lange forthcoming).
I tried to explicate unification in terms of the repeated use of patterns
of argument to generate a broad set of conclusions. There is something to
this idea, and Iwould still claim that it offers a useful way of reconstruct-
ing some parts of theoretical science, offering a hybrid between the so-
called semantic conception of theories and the idea of a family of kindred
methods of problem solution that lies at the heart of Kuhns much-abused
notion of a paradigm. But my efforts to specify the notion of unification so
as to solve problems of causal asymmetry led to difficult (probably intrac-
table) issues of defining technical notions. Fortunately conversations with
Nancy Cartwright dislodged my Humean scruples, and Ibegan to see the
notion of unification for what its worth:a useful device for some exercises
in philosophical reconstruction of scientific theories (e.g., Darwins) and a
virtue possessed by some, but by no means all, explanations.
So far a confessional preamble to taking up Woodwards major points.
Woodward distinguishes two notions of unification. The first, the concept
Iattempted to characterize, sees unification as consisting in showing how
apparently different phenomena depend on a small number of explanatory
factors. The second, which Woodward takes to figure in some of my antire-
ductionist arguments, consists in establishing that some facts are inde-
pendent of certain others. So, for example, in defending the autonomy of
classical genetics, Iattempt to show that the transmission of genes at loci

[148] The Philosophy of Philip Kitcher

on chromosomes from different pairs is independent of the specific details
of the molecular structures.
Woodward also recalls a useful distinction of Jaegwon Kims between
those explanatory relations that are internal to our corpus of knowledge
and those that depend on a connection to something beyond, paradig-
matically parts of nature. He uses his two notions of unification and the
internal/external distinction to explore the potential explanatory uses of
explanation, the prospects for internal explanations, and the ways that
unification and his own preferred approach via difference-making treat dif-
ferent scientific examples.
I want to take up four specific questions raised by Woodwards essay.
(1)Is an emphasis on unification at odds with the disunity view defended
by Cartwright and Dupr? (2) Do antireductionist arguments hinge on
Woodwards second notion of unification? (3)Is a unificationist approach
to explanation internalist? (4) Are there instances in which a particular
scientific example can be approached from both the unificationist and the
difference-making perspective, and, if so, do these perspectives inevitably
(1) In my work from Science, Truth, and Democracy (Kitcher 2001b) on
Ive endorsed the Cartwright-Dupr picture of a bundle of sciences, meth-
odologically diverse and focused on different aspects of nature. The prag-
matism that stems from William James ([1907] 1975)and is taken up by
Dewey ([1925] 1981) insists that inquiry is inevitably selective and that
the idea of some complete unified science is an illusion. Appreciating this
insight doesnt compel us to glory in the mess or to overlook the virtues
of obtaining whatever order we can find; as Cartwright recognizes, nomo-
logical machines may be rare, but we should welcome them where we can
find them (or create them). Order might consist in the existence of general
laws that govern a particular domain of phenomena. More likely Ibelieve
(and, again, this is to follow Cartwright from her 1983 book on) is that
true exceptionless laws are hard to find and that our ordering of the phe-
nomena is principally achieved by specifying idealizing models that can be
applied again and again to answer a large family of questions. Some phi-
losophers of science concentrate on the models and their features. Ifocus
on the unifying work that is done by applying the same problem-solving
pattern again and again. In this Iattempt to render explicit the tacit knowl-
edge Kuhn ascribes to his normal scientists. The Darwinian evolutionary
theorist recognizes how to instantiate the Darwinian patterns across a
range of instances (Kitcher 1993a, ch. 2). Identifying unifying patterns of
this sort is a major scientific achievement, preparing the way for an exten-
sion of normal science, or even a new normal scientific tradition. But these

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [149]
traditions are local, specific to the major branches of the sciences or to sub-
fields within them, or even to subfields of subfields. Local unification is
compatible with a lot of difference and disorder across the space of areas of
inquiry. My explication of Kuhns important point, and my endorsement of
local unification, can coexist with the Cartwright-Dupr thesis of disunity.
Unification is a regulative ideal:enjoy it where you can find it, but dont
suppose that it can be achieved on any global scale. Woodward sides with
me against Cartwright and Dupr, but Isee no need to take sides.
(2) Woodward cites my discussion of the supposed reduction of clas-
sical genetics to molecular biology as exemplifying his second type of
unification. He is quite right to recognize that central to my antireduction-
ist thesis is a claim that the details of the molecular structures and rear-
rangements are irrelevant to the patterns of transmission, specifically to
the independence of assortment of alleles on different chromosome pairs.
Iwant to suggest, however, that unification in the first sensethe sense
Ihave attempted to explicateplays an important supporting role in the
judgment of irrelevance. Precisely because we have a unifying account of
the transmission of genes on nonhomologous chromosomes, we see that
the underlying molecular details dont matter.
Imagine our predicament if we had no such explanation, either in this
case or in the comparable case of the sex ratio at birth. Under these circum-
stances wed be in the predicament often affecting medical researchers who
study complex diseases: in this instance, the etiology involves this com-
bination of factors, in that something quite different, and so on and on.
Without some explanation to bind them together, wed view the apparent
regularity as a giant coincidence, something that comes about on the basis
of quite diverse causal antecedents and that might well break down as we
sample further. Unless, of course, like Dr.Arbuthnot, the first observer of
the preponderance of male births, we were ready to chalk the whole thing
up to divine providence.
So, I suggest, its the unifying power of seeing gene transmission as
a type of pairing-selection process or of understanding birth sex ratios
from the perspective of Fishers evolutionary argument that generates the
judgment that the lower-level factsthe gory detailsare irrelevant.
Unification of the first kind warrants our ascribing unification of the sec-
ond kind. (See (4) for Woodwards potential response to this.)
(3) As Ioriginally proposed it, Itake the unification approach to explana-
tion to be internalist in exactly the same way Hempels covering-law models
were. Hempel requires true covering laws; Idemand that the instantiations
of the unifying patterns be true statements. Equally Hempel avoids any
ontic commitment to causal relations by supposing that explanatory

[150] The Philosophy of Philip Kitcher

relevance is a matter of derivation using covering laws. If theres a place
for causation in Hempels reconstruction of science, the causes are the fac-
tors picked out by (certain types of) explanatory relationships. Thus the
explanatory status of a relation between some factor and an event doesnt
depend on any prior causal fact in nature; rather the causal fact is merely
the consequence of a particular internal relation in the system of (true)
beliefs (derivability of the sort allowed by Hempels models). By exactly the
same token, the criterion for the explanatory worth of particular patterns
of derivation is their belonging to a store that overall unifies our corpus of
true beliefs. As my 1986 and 1989 essays make explicit, causation is simply
a relation that holds between C and E, when the best unification of true
beliefs allows the derivation of E from an instantiation of some unifying
pattern in which C figures. The forms of internalism are identical, and, in
both cases, internalism is the expression of a commitment to honoring
Humes blackballing of real causal connections.
(4) But suppose you took the liberal view (to which my preamble con-
fessed) of supposing that its entirely legitimate to forget Humes critique,
to invoke causal connections in nature, and even to talk of causes that
make a difference. Is there any role for unification still to play? Does all
explanation reduce to some form of causal explanation?
In my view there are important types of noncausal explanation. Some
of them occur in mathematics and in the theoretical reaches of the natu-
ral sciences. Others crop up at less ethereal levels, as in the recognition of
equilibria in sex ratio theory. Sometimes, although by no means always, the
explanatory value can be identified with the unification achieved. (Some
mathematical explanations work through unification, although Lange has
convinced me that mathematics is more diverse with respect to its explana-
tions than Ioriginally thought.)
Yet even when it is reasonable to offer a Woodward-style explanation in
terms of causes that make a difference, Im inclined to think that unifica-
tion of my original type can play a role. Woodward might reply to my dis-
cussion under (2)by suggesting that our understanding of the irrelevance
of the molecular facts (the gory details) doesnt depend on our having a
unifying explanation but rather on our seeing which factors make a differ-
ence. In my ecumenical old age, Im happy to talk his language:indeed we
do see that it isnt the molecular structures that make the difference. But
how do we arrive at that judgment? By being able to adopt a particular
unifying!perspective on the phenomena we seek to understand. Were
deploying unification as part of a methodological rule that assists us in our
search for local order. Ironically that methodological rule is not only akin to
the standards invoked by eminent scientists (the scientists who originally

U n i f i c at i o n i s m , E x p l a n at or y I n t e r n a l i s m , a n d Au t o n o m y [151]
encouraged me to think that unification is important in explanation) but
also to the precepts issued by philosophers:Hume, recall, appends a list of
rules by which we should judge causation, and, in my reading at least, they
endorse unification as a regulativeideal.


It may seem that Ihave treated Jims probing essay in a highly selective
way, picking up a question here and another there. But in the spirit of my
opening report on how my views have evolved, Iwant to propose that he
and Iare both now engaged in a different type of philosophical project with
respect to explanation. The name of the game is no longer to enunciate
the Final and Complete Analysis of Explanation in All Forms. Rather the
philosophical task is to develop and refine some tools for specific recon-
structions and analyses of bits and pieces of explanation across the range
of diverse sciences. As I read Jims rich discussion, hes recognizing that
both difference-making and unification might be valuable tools (concepts),
good for working on different problems, or even in combination on some
occasions. Hes showing some ways those tools might be put to use. Ive
tried to sketch a little in the samevein.

[152] The Philosophy of Philip Kitcher


Special-Science Autonomy and

theDivision ofLabor

H ow reductionisms star has fallen. Once it lived a blessed life in a grand

logical empiricist mansion high on a philosophical hill; now it wanders
the streets below, its face drawn and its clothes ragged, carrying under its
arm a tired manifesto that no one wants to read. Among the revolution-
ary leaders responsible for its overthrowamong such firebrands as Jerry
Fodor and John Dupris the sweet and reasonable voice of Philip Kitcher.1
Kitchers 1953 and All That (1984) advances three arguments against
the reducibility of classical genetics to molecular genetics:

1. Classical genetics does not contain the kind of general laws required by
Nagels (1979) canonical account of intertheoretical reduction.
2. The principal vocabulary of classical genetics cannot be translated into
the vocabulary of lower-level sciences; nor can the vocabularies be con-
nected in any other suitable way (that is, by bridge principles).
3. Even if the reduction were possible it would not be enlightening, because
once you have the cytological explanation of genetic phenomena, the
molecular story adds nothing of further interest.

This essay takes issue with the third of these arguments, contending that a
robust explanatory reductionism can coexist with the sort of explanatory

1. See for example Fodor (1974); Kitcher (1984); Dupr (1993).

autonomy that Kitcher considers to be its manifest refutation. In the spe-
cial sciences, then, we can have it all: a thoroughgoing explanatory auton-
omy in the lab and the field along with a severe reductionist philosophy of
explanation in the all-seeing armchair.
I will not offer anything so immodest as a comprehensive case in favor
of reductionism. Idiscuss only one form of explanatory reductionism, and
Idefend it against only one objection, the argument from autonomy, while
giving no positive considerations in its favor. Reductionism will not be
returned to its place of former splendor, thenbut I do hope to give it
some sustenance, a steady job, and a place to hang itshat.


Kitchers argument from autonomy in 1953 and All That turns on the
explanation of an enhanced version of Mendels secondlaw:

Independent assortment: Genes on nonhomologous chromosomes assort


A satisfactory explanation of independent assortment, Kitcher asserts,

describes meiosis, the process driving assortment, at the level of cell bod-
ies and their interactions:

Cytology provides the answer. At meiosis, chromosomes line up with their

homologues. It is then possible for homologous chromosomes to exchange
some genetic material, producing pairs of recombinant chromosomes. In the
meiotic division, one member of each recombinant pair goes to each gamete,
and the assignment of one member of one pair to a gamete is probabilistically
independent of the assignment of a member of another pair to that gamete.
Genes which occur close on the same chromosome are likely to be transmitted
together (recombination is not likely to occur between them), but genes on non-
homologous chromosomes will assort independently.(347)

This is, of course, a standard textbook explanation of the Mendelian effect.

Surely it can be enhanced, though, by describing the molecular mech-
anisms that implement the processes in question? By explaining at the
chemical level how chromosomeslong strands of DNAare reassembled
and assigned to gametes, will we not see still more deeply why genes assort

[154] The Philosophy of Philip Kitcher

Kitcher allows that some further details at the cytological level will
deepen our understanding, such as a description of the formation of the
spindle and the migration of chromosomes to the poles of the spindle just
before meiotic division, which will allow us to see that the chromosomes
are not selectively oriented toward the poles of the spindle (348). But
there he draws the line. Molecular detailschemical detailsof the work-
ing of the spindle, even those that bear directly on the equiprobability of
orientation, contribute nothing to the explanation.
That is not to say that the molecular details do not add to our total
explanatory knowledge. They constitute an explanatory extension of clas-
sical genetics, Kitcher writes, but he maintains that it does not follow that
the explanations provided by the [classical] theory can be improved by
attaching these extensions (365). In other words, we can have lower-level
understanding of (at least parts of) high-level theories, but this lower-level
knowledge has no legitimate place in the explanations offered by the high-
level theories. For the purpose of understanding independent assortment,
then, its irrelevant whether the genes are made of nucleic acid or of Swiss
cheese (1999, 200). Indeed specifying the molecular implementation of
meiosis would decrease the explanatory power of the cytological explana-
tion because it would disguise the relevant factor (1984,348).
In what follows Iput aside the claim about disguise and decrease, which, as
Kitcher himself concedes in the same passage, may be too subjective, turn-
ing as it does on our own limited cognitive powers. What is important is the
claim that the molecular details, because they are irrelevant, do not increase
the power of the explanation. That is where reductionism comes togrief.
In The Hegemony of Molecular Biology (1999), Kitcher introduces a
further example to make roughly the same point. How to explain the strik-
ing fact noted by John Arbuthnot, that in every year between 1628 and
1709 the number of males born in London exceeded the number of females?
Aschematic explanation puts together the fact that the sex ratio of humans
and many other animals is 1:1 at sexual maturity with the fact of greater
male infant mortality:though more boys are born, more die before puberty,
so that the ratio at puberty is in large populations just about exactly1:1.
This explanation benefits considerably from a fleshing out. It is aug-
mented in particular, as Kitcher observes, by the evolutionary explana-
tion of the 1:1 sex ratio so influentially advocated by R.A. Fisher (1930).2

2. On the prehistory of the explanation in the work of Darwin and later writers, see
Edwards (1998).

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [155]
According to Fishers story, the sex ratio has become fixed at 1:1 because
the even ratio is a stable and unique equilibrium. And equilibration occurs
at 1:1 because, in a population with more females than males, individuals
with a propensity to produce more males than females will have a higher
expected number of grandchildren and vice versa. Why a higher expected
number of grandchildren? Your expected number of grandchildren is pro-
portional to your expected number of children and your childrens expected
number of matings. Since matings require exactly one male and one female,
a males expected number of matings will increase, relative to a females, as
the proportion of males in a population decreases.
If explaining an explanations explainers deepens the original explanation
if explanatory relevance is transitivewhy not go further? Why not detail
the biological mechanics in virtue of which, for example, successful repro-
duction always and only involves a single male and a single female?
Here Kitcher might have said: As with molecules in the Mendelian case,
I draw the line at these gritty goings-on. As long as you see that procre-
ation, if not copulation, is strictly one on one, it is irrelevant whether the
protagonists are made of meat or Swiss cheese.
In fact he is somewhat more circumspect, more choosy about the ingre-
dients of the explanatory sandwich. Certain details of implementation are
worth investigating, he remarks. Presumably he allows that even quite
low-level details might appear in an explanatory extension of Fishers
model. But they do not, merely because they extend the model, count as
explanatorily relevant to the things that the model explains, such as the
sex ratio, and if they are sufficiently low level they certainly do not count
as relevant.
Indeed in the sex-ratio case and almost everywhere else, Kitcher appears
to hold the following view:moving down the levels of potential explanation
from ecology to physiology to cytology to chemistry to fundamental phys-
ics, there is some point beyond which further unpacking of mechanisms
becomes entirely irrelevant. Thus, for example, deriving the one-on-one
nature of procreation from quantum mechanics adds nothing whatsoever
to our understanding of Arbuthnots observation. Thanks to some aspect
of the nature of scientific explanation, the facts of implementation cease to
explain when the scale reaches so fine a level of description.
What explanatory principle is it that undercuts the transitivity of
explanatory relevance, severing the link in certain cases between the
explainers of explainers and the explanandum?
In 1953 Kitcher accounts for failures of transitivity by proposing that,
when transitivity falls through, it is because categories essential to explain-
ing high-level phenomena cannot be ascribed explanatory relevance by

[156] The Philosophy of Philip Kitcher

lower-level explanations due to their not constituting natural kinds from
the lower-level point of view (Kitcher 1984,349).
This is a peculiar argument to use against the central target of the paper,
reductionism as understood by the logical empiricist tradition:the empiri-
cists had little use for the notion of a natural kind, except perhaps as an hon-
orific bestowed on especially useful categories at the end of inquiry (Quine
1969). Most modern philosophers of science will also, Ithink, doubt that
there is some notion of naturalness that precedes and constrains the facts
about explanatory relevance. Certainly Imyself reject this thesis unreserv-
edly.3 Something like it has gained prestige in recent years thanks to the
work of David Lewis (1983), but Lewiss metaphysics does not supply what
Kitchers argument requires:a notion of high-level naturalness that cross-
cuts physical naturalness.
Perhaps it is better to understand Kitchers appeal to naturalness from the
perspective of his own unificationist account of explanation (Kitcher 1981;
1989). The permissible content of an explanation is dictated, the unification-
ist says, by the argument patterns that appear in the explanatory store
that is, in the set of most unifying patterns. Among the constraints imposed
by such patterns are limits on the properties that may be mentioned in an
explanatory argument. If we understand a propertys being natural at a cer-
tain level as its being allowed into the characteristic argument patterns of
that level, and if a certain principle of conjunction is deniedthat concat-
enating two permissible argument patterns always produces a third permis-
sible argument patternthen we can make sense, in a unificationist context,
of Kitchers claim that a factor A may be a natural explainer of B, and B a
natural explainer of C, without As being a natural explainerofC.
For the purposes of this essay, however, Iwill put aside the question of
why Kitcher treats explanatory relevance as intransitive, of why he thinks

3. And Ireject the underlying theses:that such a notion of naturalness constrains

what may count as a law of nature and that explanation requires laws of this sort. These
ideas about the importance of naturalness, also found in Fodor (1974), are, Ipresume,
supposed to allude to Hempel and Oppenheims (1948) search for a notion of lawlike-
ness to constrain what may count as a law of nature and so play a role in deductive-
nomological explanation. Hempel and Oppenheims strategy is to rest lawlikeness on
the notion of a purely qualitative predicate, which they attempt to precisify in terms
of formal logic. Though by their own admission they do not entirely succeed, they
would surely have regarded an appeal to a metaphysics of naturalness with unalloyed
horror. In any case there is nothing in their conception of a purely qualitative predi-
cate that would rule out the use, in molecular biology, of the sort of functional defini-
tions that give rise to multiple molecular realizability (which is what Fodor, Kitcher,
and others take to preclude the corresponding properties naturalness). Hempel and
Oppenheims concern is with predicates that are overly specific; they are aiming for,
not avoiding, generality.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [157]
that the correct story about an explainers implementation does not always
contribute to the explanation. Let me simply take it as given that Kitchers
claims about relevance are correct. It is a fact, Iwill suppose, that scientists
of the high level regard much information about implementation as irrel-
evant to their researchthe sociological fact of explanatory autonomy.
Assuming that those scientists are not wholly mistaken, a philosophical
fact follows immediately: details of implementation often are explanato-
rily irrelevant, and below a certain level of description, in many or perhaps
even all cases, every detail of implementation is irrelevant.
That irrelevance looks to be flatly incompatible with explanatory reduc-
tionism. Or as Fodor (1974, 11213) memorably writes:

Reductivism... flies in the face of the facts about the scientific institution:the
existence of a vast and interleaved conglomerate of special scientific disciplines
which often appear to proceed with only the most token acknowledgment of
the constraint that their theories must turn out to be physics in the longrun.

The next section outlines a notion of explanatory reductionism that pro-

claims exactly what Kitcher and Fodor deny: the explanatory relevance,
always and everywhere, of fundamental physics. Ithen show how to make
that reductionism fit with the sociological and philosophical fact of explan-
atory autonomy.


The physical world, there is ever more reason to think, is the only world we
have. Everything is made of physical stuff, and everything that happens,
happens because of the way physical laws push physical stuff around. This
is the doctrine of physicalism. It is deniable, but ever ascendant.
From physicalism it follows that any state of affairs or pattern of behav-
ior we find in the world, no matter how high level or abstract, can be derived
from fundamental physical facts and laws. It also seems plausible (though
it does not strictly follow) that any state of affairs or behavior at a given
level can be derived from facts and regularities about entities at the next
level down. Thus economic regularities can be derived from psychological
facts and regularities, psychological regularities from physiological facts
and regularities, and so on through cytology, molecular biology, and chem-
istry to fundamental physics. (The relevant lower-level facts will concern to
a great degree the arrangement of and relationships between lower-level
entities:we do not get psychology from the study of neurons in isolation

[158] The Philosophy of Philip Kitcher

but from the study of the neural structure of the brain as a whole.) This
picture, reminiscent of Oppenheim and Putnams (1958) unity of science
manifesto, is a little too simple:sometimes a derivation will have to pull
facts from different levels (combining, say, physiological facts about organ-
isms with physical facts about climate), and indeed the picture of the world
as organized into levels corresponding to university departments is surely
a caricature. All of this may be allowed without compromising the basic
assumption of derivability from the lowerlevel.
Now let me go further. Some of these derivations are also explanations;
further, for any higher-level phenomenon there is at least one derivation
from the lower level that is explanatory. (There are also, in my view, many
such derivations that are not explanatory, for reasons I give in the next
The fact of explanatory autonomy is quite consistent with everything
Ihave said so far. Indeed Kitcher himself may well agree with most of it; his
favored explanation of independent assortment, a regularity identified by
classical genetics, takes the form of a derivation from the cytological level,
and his favored explanation of the 1:1 sex ratio at maturity takes the form
of a derivation from facts about individuals grandchildrens prospects
hence properties of a population are derived from individuals propensities
to prosper. More generally Kitchers notion of an explanatory extension
turns on the availability of many such derivations, though perhaps not a
derivation for absolutely every high-level state of affairs.4
A final step, however, will secure a collision with Kitcher in particular
and autonomy in general and will transform the view Iam describing into
something that is clearly reductionist. It is to endorse the transitivity of
explanatory relevance: explain something that explains a phenomenon,
Ipropose, and you have added something relevant to, and so enriched, your
original explanation. The cytological explanation of independent assort-
ment is good, but it can be made even better by explaining the cytological
explainers in turnby giving a molecular explanation of the relevant cytol-
ogy. And better still by providing the chemistry of the relevant molecular
facts. Best of all is to derive, from the fundamental physics, the relevant
chemical facts. At that point you can descend no further; you have an expla-
nation that is in one sense maximally good because maximally reductive.5

4. Kitchers skepticism about bridge principles for general categories such as gene is,
as he himself notes, compatible with the derivability of all observed behaviors of genes,
as the behaviors may be derived piecewise, that is independently for particulargenes.
5.If particular matters of fact appear in the explanation, it might be further
improved by tracing back in time the genesis of those facts. But this proposal (Strevens
2008, 4.31) has nothing to do with reductionism.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [159]
Give this thesis a name:explanatory reductionism. An explanatory reduc-
tionist, then, holds that an explanation is always improved by giving a
lower-level explanation of its partsof the initial conditions, regularities,
and structural facts that it citesand that such further explanations are
available for every nonfundamentalpart.
How is this compatible with the sociological fact of autonomywith
the fact that working scientists consider many and in some cases all lower-
level details to be irrelevant? That looks like a simple question; the answer,
clearly, is that the fact of autonomy refutes explanatory reductionism.
But no: it is a complex question, and compatibility is possible after
all. The key is to identify more than one sense of explanatory relevance.
Explanatory reductionism is true of one sense, autonomy of theother.
To make this dichotomizing plausible, in the next section Iwill intro-
duce a reductionist account of explanationmy own kairetic theory
and flesh out the notion of explanatory relevance at its heart. Agreat deal
of lower-level detail turns out to be irrelevant to high-level phenomena.
But not all detailas must be the case, since the kairetic theory is a form
of explanatory reductionism. The discussion thus opens the door to, with-
out entirely achieving, the synthesis of reductionism with autonomy. That
goal requires the postulation of the second kind of explanatory relevance
in section4.


The kairetic theory of explanation is a causal theory. It begins with physi-

calism, not only about things and laws but about causation itself:the raw
material of causality is the fundamental-level relation of causal influ-
ence. Newtonian force is the paradigm of such an influence relation. In
the Newtonian worldview force is responsible for all changes of velocity
of all objects, microscopic or massive, and so, along with inertial motion,
for everything that happens. It is described completely by fundamen-
tal physics.6 The question whether there is causal influence to be found
in modern physics is more fraught, but Iask you to put it entirely to one
side and to suppose that physics does indeed give us such a relation, as

6. To say that influence is a fundamental-level relation is not to say something about

its metaphysical foundation, but only about its relata. It leaves open the possibility
that the facts about influence are determined by other facts, even high-level facts such
as the direction of entropy increase or the structure of human causal concepts.

[160] The Philosophy of Philip Kitcher

many philosophers of causation have argued (Dowe 2000; Lewis 2000;
Reichenbach 1956; Salmon 1984).7
The kairetic account holds that an event or phenomenon is to be
explained by showing how it was produced by the aggregate causal influ-
ence of other states of affairs and the laws in virtue of which they exerted
their influence. In a Newtonian world, for example, an event is explained
by exhibiting whatever objects and events pushed around certain constitu-
ents of the world in a way that realized the occurrence of the event, along
with the laws in virtue of which the pushing around occurred, that is, the
laws in virtue of which the relevant forces were brought tobear.
It would be a mistake, however, to proffer as the explanation of an
event the entirety of the web of antecedent causal influence in which it is
embeddedso says the kairetic account. What explains the event is only
those aspects of the web that made a difference to the events occurrence.
Various familiar philosophical accounts can be given of difference-making.
Astatistical relevance theory holds that an aspect of the web of influence
made a difference to an explanandum event if it raised the probability of
the event. Acounterfactual theory requires of difference-makers that, had
they not been present, the explanandum would not have occurred.
The kairetic account of difference-making provides the following recipe
for determining difference-makers:Take a comprehensive causal model of
the production of the explanandum, that is, a complete description of the
web of influence leading up to but not including the explanandum. Assume,
for simplicitys sake, that the explanandum is deterministically produced
and that the description (therefore) entails its occurrence. Now make the
description as abstract as possible without either undermining the descrip-
tions entailment of the explanandum or undermining the descriptions
status as a causal model. What remains in the description after this process
of abstraction are specifications of difference-making factors. (Needless to
say, a fully adequate presentation of the recipe requires more than a few
sentences; for the official version, including the criterion for a descriptions
constituting a causal model, see Strevens2008.)
To explain, for example, the Broad Street cholera epidemic of 1854,
begin with a complete description of the epidemic:the leaching of the chol-
era bacteria into the water supplying the Broad Street pump, the carry-
ing of the water to various residences, its ingestion by various people, the

7. As this list makes plain, there are many ways to give a metaphysical theory of
causal influence. There is no need, in what follows, to choose amongthem.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [161]
course of the disease in each victim, its transmission to others. Notionally
the process is described at a maximal level of detail:the position of every
bacterium and the disposition of every drop of water. Now take away what
you can without invalidating the descriptions entailment of the explanan-
dum, that is, of the fact of the epidemic. The precise positions of individual
micro-organisms do not matter at all; that information can be deleted,
leaving only a specification of the approximate density of the organisms
in the water retrieved from the pump. The time of day (I assume) that
the water is pumped is also irrelevant; what matters is only that a certain
amount was consumed. Likewise the course of death need not be charted
in excruciating detail; the rough facts about the degree of dehydration and
its inevitable physiological effects is enough to entail an upward step in the
statistics of mortality. Throughout this process of information removal,
nothing is added; rather, more detailed specifications of the causal web are
replaced with strictly less detailed specifications. What you are left with is a
description of the same causal web with which you began, but an extremely
abstract description: bacteria in considerable quantities leaked into the
water supply; the water was consumed by a significant number of people;
given the prevailing conditions, also specified at a high level of generality,
they went on to contract cholera.
More or less the same recipe applies to the explanation of regularities.
Why does Marss orbit around the sun conform approximately to Keplers
laws? Begin with a complete causal model for Marss orbita complete
specification of the causal influences on the planets trajectory over the
course of a Martian year, along with the relevant physical laws. Your model
will predict every minor twist and turn in Marss movement. But with or
without these perturbationswith or without the other stars, the planets,
the interstitial rubblethe model predicts Keplerian behavior. The kairetic
criterion therefore orders the perturbers removal, or more exactly, it tells
us to replace the painstaking specification of the distribution of mass with
something as abstract as possible having the same net implications for the
explanandumin this case a specification that the total gravitational force
due to objects other than the sun did not exceed some (small) upper bound.
What remain are the difference-makers: the physics of gravitation; the rela-
tive size, position, and velocity of Mars and the sun; and the aforemen-
tioned upper limit.
Note that to determine explanatory relevance it is not necessary to con-
struct complete causal models. We can see that certain things are not going
to count as difference-makers without going through the rigmarole, so it is
possible to have knowledge of difference-makers while having only a very
rough knowledge of the underlying causal web. The kairetic recipe is the

[162] The Philosophy of Philip Kitcher

ultimate criterion for the correctness of claims about difference-making,
but here, as so often in life, we do not deploy the ultimate criterion in day-
to-day cognition.
For the purposes of this essay two things matter about the kairetic
account. First, for many high-level or coarse-grained explananda (epidem-
ics, approximate orbital trajectories), it declares vast amounts of physical
detail to be explanatorily irrelevant, even when that detail has some causal
connection to the course of events to be explained. Second, the kairetic
account is nevertheless a species of explanatory reductionism. This is
because its specifications of difference-makers, thus its explanations, are
descriptions of the web of causal influence, hence of properties of the fun-
damental laws of physics and the physical states of affairs, whose aggre-
gate causal impact determined that the explanandum holds or occurred.
A complete kairetic explanation of Keplerian behavior does not have to
specify in any level of detail the distribution of mass in the solar system,
but it does have to describe what that mass is made of and how, at bottom,
gravitational attraction works: curved space-time, geodesics, stress-energy
tensor, and all.8 Conversely an explanation that does not have something
to say about the fundamental laws or configurations of things has not said
everything that an explanation should say. To add what is missing will
improve the explanation.
Putting it another way, an explanation that bottoms out at, say, the cyto-
logical level implies that there is something about the web of fundamental-
level influence that makes certain cytological claims true but does not
specify what that something is. According to the kairetic account, it is
obliged to do so. The most natural and straightforward way to discharge
the obligation is to give an explanatory derivation of the relevant cytologi-
cal facts and generalizations from the physical facts and generalizations.
There is much that such a derivation will not say about the physical real-
ization of the cytological facts:it specifies only those things that make a dif-
ference to the cytological facts holding. For example, it will omit pointless
enumerations of the precise vibrational energies of various components of,
and for that matter the precise positions of, various important molecules.
But it will specify approximate positions and lay out in physical terms
the relevant properties of DNA, of telomeres, of the spindle structure in

8. The question of exactly what details can be removed from a description of the
gravity mechanism is determined by the kairetic accounts cohesion constraint.
Rather than attempting to describe the workings of the constraint in this essay, Isim-
ply spell out its reductionist consequenceswhich reductionism is what matters for
the purposes of the discussion.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [163]
meiosis, describing the fundamental physical basis of the molecular inter-
actions that make the whole machine work the way it does. It will, that is,
give precisely the details, and only the details, you would expect if you were
to ask a molecular biologist to explain the relevant features of the cytology
and he were in turn to ask a physicist to explain the relevant features of the
molecular biology.
A reductionist theory of explanation can, as the kairetic account
shows, prescribe very abstract explanations, but they must be at the
same time physical explanations. There, it seems, is the rub:the fact of
explanatory autonomy is simply the fact of the irrelevance of these phys-
ical explanations. Their high degree of abstraction is no palliative. The
kairetic account says that the physical structure of the spindle must be
specified, albeit not in great detail. Autonomy says it does not. Thus the
kairetic account, like all reductionist accounts of explanation, is falseso
you might conclude.
What next? One strategy is to give something like the kairetic account
a stronger criterion for difference-makingstrong enough to imply the
absolute irrelevance of all facts below a certain level (Franklin-Hall forth-
coming). Another is to find a different kind of explanatory irrelevance,
orthogonal to the irrelevance diagnosed by the kairetic criterion, and to
show that autonomy is all about that other kind of relevance. That is the
route I will take.


If a detail makes no difference to a phenomenon, say that it is objectively

irrelevant to the explanation of that phenomenon. It is compulsory, when
explaining, to ignore objectively irrelevant factors. There is another rea-
son for ignoring details that is not compulsory in quite the same way but
that allows for the existence of explanatory models that say nothing about
lower-level implementation; Icall it contextual irrelevance.
The notion of contextual irrelevance is not proprietary to the
difference-making approach to explanation: any philosopher of expla-
nation, whether reductionist or antireductionist, whether allied to the
difference-making, the unificationist, or some other view, should recog-
nize the existence and importance of irrelevance of this sort. As I will
eventually show, however, the notion of difference-making does have
something important to contribute to our understanding of the systema-
ticity of contextual irrelevance.

[164] The Philosophy of Philip Kitcher

4.1.Contextual Irrelevance

Suppose that a team of archeologists in the distant future excavate a Ford

factory and attempt to reverse-engineer the internal combustion engine
to construct an explanation of how the thing works. They might reason-
ably pursue their task by dividing the engine into (apparent) functional
units:the piston assembly, the starter motor, the cooling system, and so
on. Ateam would be assigned to each unit with the task of determining
how that unit contributed to the functioning of the whole. The advantage
of such a division of labor is, of course, that one part of the engine can be
analyzed without any detailed understanding of the other parts. Although
the other parts cannot be ignored altogether, they can be treated as black
boxes with specified inputs and outputs but no internal details. The team
working on the piston assembly needs to know that there is a cooling sys-
tem and something about the parameters of that systems operationhow
it reacts to increasing engine temperature, how fast it cools, and so on
but that isall.9
The full explanation of the engines workings is attained by bringing
together the teams and replacing each black box with the proper internal
causal model of its workings (omitting objective non-difference-makers).
Until that time no individual fully understands the engine. This gap in each
investigators knowledge is all to the good; it would be a mistake, in a prac-
tical sense, for the piston people to start thinking about cooling, as it would
unravel the efficiencies gained by the division of epistemic labor. Thus the
prime directive or principal norm governing each teams subproject is as
follows:The details of any other components workings are to be considered
irrelevant to your explanation; in deciding what does and does not matter
for explanatory purposes, pay them noheed.
This is what Imean by contextual irrelevance in explanation. The work-
ings of the cooling system are objectively relevant to explaining the engines
behavior but they are contextually irrelevantirrelevant, that is, accord-
ing to the procedural rules laid down for the archeologists investigation
of the piston assembly. Contextual relevance is quite different from objec-
tive relevance. Whereas the facts about objective relevance are decided
by the world, the facts about contextual relevance are decided by us, the

9. Does a black box specify causal relationships (without specifying a mechanism),

counterfactual relationships, actual-world regularities, or something else? In this essay
Ileave the question open; since black boxes function only as placeholders for explana-
tory parts and not as explainers themselves, it does not much matter.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [165]
organizers of inquiry into the world. Whereas including objectively irrel-
evant factors in an explanation is an intellectual error, including contextu-
ally irrelevant factors is a social or practical error. Nevertheless the success
of an explanatory investigation may hinge just as much on the norms of
contextual relevance as on the norms of objective relevance. Investigators
must take both equally seriously. (I should add that the norms for both
kinds of relevance govern what goes into an explanation, but they do not
govern what the explanatory investigators may contemplate. Asuccessful
explainer will of necessity have to spend rather a lot of time thinking about
non-difference-makers, just so as to be able to recognize them assuch.)

4.2.The Division ofScientif ic Labor and Autonomy

That significant philosophical consequences follow from sciences divi-

sion of cognitive labor among competing approaches to a single problem
or domain is one of Philip Kitchers enduring contributions to our field
(Kitcher 1990, 1993a). Science also divides its labor among complementary
fields of research, in twoways.
First, like the archeologists with their engine, it implements what might
be called functional compartmentalization, as when in the study of the
physiology of the human body the immune system specialists black-box
the parts of the brain controlling the respiratory system, or when in evolu-
tionary biology some biologists black-box the mechanisms of inheritance
while others black-box the phenotypic causes of differential reproduction
Second, science implements what might be called functional stratifica-
tion, the individuation of the explanatory pursuit by levels:the physicists
study the fundamental particles, the chemists the molecules made up by
those particles, the cytologists the cells made up by those molecules, and so
on. Typically investigation at one level black-boxes everything that goes on
at lower levelsnot ignoring lower-level goings-on altogether, but intro-
ducing them by way of functional specifications, as when chemists (though
not quantum chemists) talk in terms of bonds or nuclei without asking
how a chemical bond is implemented or a nucleus constructed, or popula-
tion ecologists talk in terms of predators and prey without attempting to
represent the mechanics of predation.
The notion of a black box is the same whether you are doing functional
compartmentalization of an engine or functional stratification of an eco-
system. But what you do with the black boxes is somewhat different: in
compartmentalization you plug black boxes into your systems inputs

[166] The Philosophy of Philip Kitcher

and outputs; in stratification you build your model of the system with
I will focus, naturally, on stratification. One motivation for functional
stratification is objective irrelevance:if the details of a lower-level process
do not make a difference to a high-level explanandum, they should be omit-
ted from the explanation. But not all black-boxing of the lower level can be
accounted for in this way. Consider, for example, a population ecologists
explanation for why a boom in the predator population can lead, down
the line, to a boom in the population of prey:the predator boom causes
the prey population to crash; the paucity of prey then causes the predator
population to crash; the prey population, having a higher rate of reproduc-
tion or shorter generation length, subsequently recovers more quickly and,
without a substantial predator presence, grows much faster than it nor-
mally would. Such an explanation would typically account for the popula-
tion change in each of these phases using a standard Lotka-Volterra model,
which represents the rate of predation as increasing with the number of
predators but supplies no further information about the way predators
hunt for prey (Roughgarden 1979). Such a model black-boxes predation:it
specifies the effect of predator numbers on prey numbers but says nothing
about the mechanism responsible for the effect. The mechanism is, how-
ever, not irrelevant to the explanation in the objective sense: if you are
trying to understand the relation between population booms and busts,
and it turns out that the relation between predator population and preda-
tion rate plays an important role (as it clearly does in the case at hand),
it is far from irrelevant to ask Why is the rate of predation proportional
to the number of predators? (Holling 1959; Lotka 1956). If the predation
mechanism does not appear in the explanation, then, it must be because
the population ecologists have been excused, by some principle of division
of labor, from having to supplyit.
To turn this into a general proposition:the principle by which the econ-
omists black-box psychology, the psychologists black-box neuroanatomy,
the evolutionary biologists and population ecologists black-box various
aspects of physiology, and so on, is the principle of contextual irrelevance.
It divides the explanatory enterprise into many different parts, in this case
layers or strata, for the sake of more efficiently producing the components
of complete explanations.
The efficiency exists for two reasons. First, puny human intellects think
better when relieved of the burden of bearing in mind the big picture.
Second, many explanatory models overlap at higher levels of organiza-
tion. Predator-prey ecosystems that look quite different at the organismic
levelpelagic fish versus woodland animals, saymay and often do share

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [167]
certain mathematical properties when viewed in the abstract; further,
these high-level similarities have important ecological consequences. It
would be a waste of many scientist-hours to have the fish ecologists and
the mammal ecologists independently derive the consequences; better to
have a single group working on the high-level properties while the fish and
mammal specialists confine their attention to those lower-level phenom-
ena where scales and fur comeapart.
The explanatory pieces so efficiently produced are made for a higher pur-
pose, to be sewn together into complete explanations, that is, explanations
in which all objective difference-makers, whether described at the level of
physics, cytology, psychology, or whatever, are brought together into a sin-
gle explanatory model. It is one of the great glories of modern science that
we have, in many cases, the necessary materials for something approaching
this full understandingthat we have, in other words, the ability to sketch
answers to the chain of explanatory questions leading all the way down
to the fundamental level. But of course no one person is custodian of this
understanding. It is spread across the sciences, with (in the predator-prey
case) the population ecologists leading off, the cognitive ethologists then
taking up the baton, and so on, all the way to the particle physicists.
It would be madness to organize explanatory inquiry in any other way.
The creation of explanations is therefore a fragmented process; fragmenta-
tion, however, is merely a means to a unitary end that can be achieved only
by science as awhole.
The conception of explanation I have proposed marries reductionism
and autonomy. On the one hand, autonomy: the work of explanatory
inquiry is divided among many domains, each of which is not merely per-
mitted but required to black-box the explanatory models generated by
other domains. Within a domain an explanatory model is complete if it
omits only details that are either objectively or contextually irrelevant.
On the other hand, reductionism: Explanatory models within a domain
are a means to a greater end, namely explanations that replace all black
boxes with substantive models accounting for the explanandum in physical
terms, that is, relating what goes on to the fundamental-level causal web,
as the kairetic account requires. Such an explanation is complete if it omits
only details that are objectively irrelevant.
Two complementary senses of explanatory completeness figure in this
scheme of things:contextual completeness, that is, completeness relative
to the standards by which explanatory labor is divided among domains, and
what you might call absolute completeness. The ultimate end of explana-
tory inquiry is the reductionist goal of completeness in the absolute sense;
however, the means to this end, given the division of explanatory labor, is

[168] The Philosophy of Philip Kitcher

contextual completeness, and so it is completeness in the contextual sense
that dominates scientific practice day to dayfor which reason you find in
every domain black-boxing explanations that are rightly called complete.
That is no reason, Ihope Ihave demonstrated, to embrace antireduction-
ism; practical considerations explain black-boxing equally well on either a
reductionist or an antireductionist approach to explanation.

4.3.The Systematicity ofContextual Irrelevance

There is a systematicity to the facts about objective relevance, that is, to the
facts about difference-making. It is not that stardust makes a difference to
the orbits of some planets but not to others, or that the vibrational modes
of a telomere make a difference to meiosis in some types of cell but not
in others:the non-difference-making, the irrelevance, is across the board.
High-level explainers may consequently ignore certain kinds of detail as a
matter of general policy; they need not treat each case on a custombasis.
Approaching the question of orbits, the celestial mechanic has good
reason in advance to abstract away from planetary constitution and the
existence of interplanetary pebbles and dust. For explanatory purposes she
need not see her system as made up of molecules at all; it is enough to see
it as made up of planetsdiscrete spherical objects of great mass and inde-
terminate composition. The facts about difference-making, then, provide
the explainer, by way of abstraction, with a high-level taxonomy of the sys-
tem in question from which certain aspects of the underlying fundamental
physics have disappeared altogether.
Likewise the cytologist is apt to think in terms of telomeres rather than
in terms of DNA sequences, letalone in terms of the individual atoms that
make up such sequences, each with its own particular position, velocity,
modes of vibration, and so on, and the population ecologist will naturally
think in terms of populations or standard types rather than actual organ-
isms in all their individuality, letalone in terms of the organisms molecu-
lar makeup.
Contextual explanatory relevance is also systematic: there are long-
lasting, across-the-board rules determining what is and is not contextually
relevant within an explanatory domain.
That may sound dubious. What is contextually relevant is determined
by the way the explanatory enterprise is parceled out among the members
of the scientific community. These allocations are nothing more than epis-
temic heuristics; why expect such matters of practical policy to be uniform
over time, from place to place, or across a discipline?

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [169]
If research strategies were simply a matter of fashion, systematicity
would be surprising. But given their goal of investigative efficiency, they
are strongly constrained by the world out there: there are certain divi-
sions of explanatory labor that are very efficient, and many that are grossly
Consider again the investigation of the internal combustion engine.
One way to divide the work of understanding the engine is for one team
to take the left half, one team the right half. The left-hand team might
find themselves, then, trying to understand the workings of the left-hand
side of a piston while black-boxing the pistons right-hand side. This is not
impossible in principle, but the black box in question, which must specify
all behavior of the right-hand side that makes a (relevant) difference to the
behavior of the left-hand side, will have to contain an extremely detailed
description of the right-hand sideso detailed that it will hardly be a black
box atall.
More generally, black-boxing is useful only when the system to be com-
partmentalized or stratified is somewhat modular, in the sense that it can
be divided into units or strata, each of which makes a difference to the oth-
ers in ways that can be specified compactly and tractably.
The sense of difference-making relevant to determining the proper level
of detail in a black boxs functional specification is not quite identical to
the objective difference-making discussed in the previous section but turns
on the same core idea and is equally objective. Here, in outline, is a kai-
retic characterization of the notion (though there are no doubt other viable
approaches to characterizing functional difference-making).10
Take some proposed division of your system into putative black boxes,
either at the same level of description (compartmentalization), or at dif-
ferent levels (stratification), or both. The aim is to determine, for any such
division, which aspects of the proposed black boxes are difference-makers
for other boxes and which are not.
To this end, associate with each proposed black box a complete func-
tional specification of the boxs inputs and outputsnot just the appar-
ently relevant inputs and outputs but every way the box interacts with its
surroundings. This specification will therefore be a comprehensive map-
ping from environment to behavior. The functional specifications, together
with whatever aspects of your explanatory model are not black-boxed,

10. Perhaps the most important difference, on the kairetic approach, between the
kind of difference-making that applies to causal factors and the kind of difference-
making that applies to functional specifications is that the latter allows for difference-
making properties that are radically multiply realizable.

[170] The Philosophy of Philip Kitcher

if any, will entail the explanandum (as in the previous section, I assume
Now delete everything from a boxs functional specification that can be
removed without invalidating the entailment of the explanandum, using
the same sense of removal or abstraction that the kairetic account applies
to determine objective explanatory relevance. What remains are the differ-
ence-making elements of the specification.
An efficient division into black boxes is one that allows the removal of
as much detail as possible, so enabling the researchers working on each
subunit in a compartmentalization, or each level in a stratification, to pro-
ceed knowing relatively little about the other researchers work because
they have in the short black box specifications everything they need to
complete their own allotted task.11 Science seeks an efficient division of
explanatory labor, hence an efficient black-boxing scheme, for purely prac-
tical reasons. But the facts about a black-boxing schemes efficiency are not
pragmatic or observer-relative; they depend ultimately (according to the
kairetic account) on the entailment relation, which resists all attempts at
persuasion and intimidation and is oblivious to intellectual fashion.
In short, contextual irrelevance depends on the efficient allocation of
cognitive labor, which depends on the opportunities for compact black-
boxing on offer, which depend in turn on a kind of functional difference-
making very closely related to the objective causal difference-making
discussed in the previous section. The facts about causal difference-making
are, Ihave suggested, systematic; for the same reasons, Ipropose, the facts
about functional difference-making are also systematic. As before Iam gen-
eralizing speculatively from a small if wide-ranging set of examples:popu-
lation ecology will, for most purposes and in most populations, be able
to get by with relatively brief functional specifications of its organisms;
economics of its rational actors; chemistry of itsbonds.
The division of explanatory labor will inherit such systematicity. The
same details of physical implementation will be ignored for a wide range of
explanatory tasks in a given high-level domain; consequently the domain
will acquire a certain explanatory taxonomy and a certain distinctive
explanatory style. In this way the unabashedly reductionist kairetic theory

11. As always with matters determined by difference-making, the complexity of the

explanatorily relevant functional specification is relative to the explanandum. If you
are trying to explain why the engine weighs what it does, having one team determine
the weight of the right half and the other team determine the weight of the left half
will not create ludicrously complex black boxes, since for the purpose of understand-
ing the right halfs weight nothing at all need be known about the behavior of the left
halfthe one half makes no difference to what the other half weighs.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [171]
of explanation explains why the sciences form a mosaic of explanatory sub-
cultures, autonomous units each only peripherally and sporadically con-
cerned with the work going on nextdoor.


The reductionist can, Ihave shown, explain the fact of autonomy. But is his
explanation as good as the antireductionists? Two antireductionist argu-
ments may suggestnot.

5.1.The Argument fromParsimony

I use two notions of relevance, objective and contextual, to make sense of

autonomy, whereas the antireductionist perhaps needs only one. Should an
autonomy theorist inclined to parsimony opt for the antireductionist story
with its unified conception of relevance?
Hardly. First and most briefly, an antireductionist account of autonomy
is likely to posit far more in the way of local rules, cultures, ontologies, and
epistemologies than a reductionist account, which sees scientific practice
as (at bottom) unified.12 If parsimony points somewhere, it is toward unity
and reduction.
Second, the notion of contextual relevance has many uses within the
philosophy of science: it occurs whenever a potentially relevant factor is
put into what Icall the explanatory framework (Strevens 2008, 5.3). Such
factors are typically objectively relevant; their being placed in the frame-
work, however, renders them contextually irrelevant, hiding them by fiat
from the explanatory spotlight. There are numerous practical reasons to
put something into the framework and thereby to put it, explanatorily
speaking, off limits. The division of labor is one; the pragmatics of conver-
sation is another; personal interests are another still. As long as explana-
tory discourse is responsive to these forces, there will be frameworking.
Frameworking and contextual relevance can be used in addition to found
the distinction between causes and background conditions, to make sense
of apparently invidious distinctions among the explanatory relevance of
various absences, to interpret talk of prevention, and to account for the

12.Exception: an account on which the explanatory boundaries between the

disciplines are dictated by a single interdisciplinary criterion, as in Franklin-Hall

[172] The Philosophy of Philip Kitcher

apparent failures of the transitivity of difference-making. (See Strevens
2008, respectively 6.1, 6.3, 6.4, 6.5.) Contextual relevance is therefore
very likely to remain with us whatever the fate of reductionism. To use it
to explain autonomy requires nothing that we were not carrying already.
Third, there is something of a test to diagnose contextual as opposed
to objective relevance, and it shows that there is plenty of both. The test
begins with the idea that, because what is contextually relevant varies
with the explanatory framework, assertions of irrelevance will appear to
be more secure when the framework is clearly fixed, best of all by explicit
specification using the words given that, assuming that, or similar. Consider,
for example, the following two claims:

The molecular implementation of meiosis is irrelevant to under-

standing the independent assortment of genes on nonhomolo-
gous chromosomes.
Given that the molecular implementation of meiosis ensures that
individual chromosomes are not selectively oriented toward one
pole or other of the spindle, the workings of the implementation
are irrelevant to understanding independent assortment.

The status of the first claim is at best somewhat murky; the second, by con-
trast, seems quite reasonable. Its rightness is palpable because the given
that locution specifies that the objectively relevant parts of the molecular
implementation are being frameworked, hence excluded from the explana-
tory picture. The same locution has no effect when dealing with objective

Rubble in the Kuiper belt is irrelevant to Marss orbits conforming

approximately to Keplers laws.
Given that Kuiper belt rubble exerts only a minuscule gravitational
force on Mars, it is irrelevant to Marss orbits conforming approxi-
mately to Keplers laws.

Given that does not improve the irrelevance claim at all; indeed it gives
the impression of sloppy writing. Because is far better. Put because in
the claim about independent assortment, however, and youget:

Because the molecular implementation of meiosis ensures that indi-

vidual chromosomes are not selectively oriented etc., the workings
of the implementation are irrelevant to understanding indepen-
dent assortment.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [173]
That seems malformed, indeed perverse, whether or not you are an explan-
atory reductionist.
There are two kinds of irrelevance claim, then: the kind that sounds
good with a given that rationale and bad with a because rationale and
the kind for which the situation is reversed (unless given that is heard as
meaning because). Now a sweeping claim:Take all of the antireduction-
ists claims of irrelevance for lower-level details that the kairetic theory
counts as difference-makers. Apply the given that/because test. Their
rationales will sound better with given that. Do the same with lower-
level details that the kairetic theory counts as non-difference-makers,
hence as objectively irrelevant, and the rationales will sound better with
because. That is just what this essays distinction between objective and
contextual irrelevance predicts: in the former case the lower-level facts
do something explanatorily important, but their contribution is frame-
worked; in the latter case the lower-level facts are explanatorily otiose.
The antireductionist who collapses the two kinds of relevance can make
no senseofit.

5.2.The Argument fromIncredulity

The antireductionist argument from incredulity goes like this: I want to

explain the Broad Street cholera epidemic. Isee that some of the details are
important. But explanatory reductionism implies that a description of the
process at the level of quantum mechanicsparticle by particle, potential
by potentialwill somehow help me to understand what happened. How
can that possibly be? It seemscrazy.
Such arguments make two mistakes (intentionally perhaps). First, by
emphasizing minutiaeparticle by particle, potential by potential
they presume that a physical description is necessarily a fully concrete
description, that to go physical is to relinquish abstraction. But that is not
the case. Even an incontestably physical notion such as center of mass is
highly abstract and therefore extraordinarily multiply realizable:a galaxy,
a gorilla, and a gallium atom may have exactly the same center of mass.
The kairetic account pushes for abstraction wherever it is possible, while
insisting that the tie to physical implementation is preserved. It is quite
possible to have plenty of both. Physical does not entail detailed, and
most kairetically complete explanations of high-level phenomena, though
physical, are not detailed atall.
Resist, I am tempted to say, every antireductionist argument that
avidly deploys the word details. The strategy is to put before your mind

[174] The Philosophy of Philip Kitcher

as paradigms of physicality objectively irrelevant minutiae and to insinu-
ate that the reductionist is committed to giving such trifles an explanatory
role. Not so. (Kitcher [1984,348] uses the word throughout his writing on
the topic:In neither case are the molecular details relevant. Indeed, add-
ing those details would only disguise the relevant factor. The last use in
that paper is a resounding gory details [370].)
Second, the argument from incredulity trades on the sheer psycho-
logical impossibility of our entertaining even a highly abstract quantum-
mechanical description of a high-level phenomenon, running together
a failure to comprehend the description with a failure of the description
to explain. (Kitcher [1984, 348] uses this strategy in the passage just
quoted, but then acknowledges the reductionist riposte and puts it aside.)
Attempting to grasp in its entirety a complete explanation for something
like independent assortment in sexually reproducing life on earththe
explanation assembled in the scientific congress at the end of time from
the contributions of geneticists, cytologists, molecular biologists, chem-
ists, and physicistsin an attempt to eyeball its explanatory value is sim-
ply not something we are capable ofdoing.
What we can do without cerebral overload is to ask whether it is worth
our explanatory while to trace particular lines of implementation down to
their physical foundation. Do we understand independent assortment bet-
ter when we grasp the molecular-level reasons for chromosomes lack of
selective bias toward the spindles poles? Do we understand it better still
when we grasp the physical reasons for the symmetries of intermolecular
forces that underlie the lack of bias? In this almost purely defensive essay
Ihave not argued the case for an affirmative, reductionist answer to these
questions, but the issue is surely a live one; the affirmative answer is both
reasonable and plausible.
I will finish with another version of the antireductionist argument
from incredulity, a one-liner by Fodor (1974, 1034): What is interest-
ing about monetary exchanges is surely not their commonalities under
physical description. Asensible reductionism does not (as Fodor implies)
deny the importance of a high-level, functional characterization of mon-
etary exchanges. But it asserts the interest in addition of the reasons such
exchanges instantiate the patterns that they doof the psychological
principles that drive spending and saving, borrowing and lending, gam-
bling and charitable giving, and then in turn of the neurological reasons
that the psychology takes the form it does. Most of the gory details of the
neurology will be objectively irrelevant, but some abstract neural facts will
genuinely contribute to our understanding of an economys monetary ebb
and flowthat is the reductionists explanatorybet.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [175]

For their philosophical insights and critique, thanks to Laura Franklin-

Hall, Dmitri Gallow, Angela Potochnik, Alex Rosenberg, Kyle Stanford,
Brad Weslake, and the audiences at the Lorentz Center 2010 workshop
Understanding and the Aims of Science and the University of Cologne 2010
workshop Types of Explanation in the Special Sciences.

Reply toStrevens

If any contemporary philosopher of science has raised a sweet and reason-

able voice in support of the central commitments of logical empiricism, it is
surely Michael Strevens. Much of Michaels recent work has been devoted
to tidying, or at least containing, the disorder identifiedand sometimes
celebratedby critics who have viewed the great midcentury tradition in
philosophy of science as oversimplifying complex phenomena. Michael has
been concerned to show that there is less disarray than post-Kuhnian his-
toricists or champions of disunity (like Nancy Cartwright and John Dupr)
have claimed. In his elegant and characteristically witty essay, he continues
his efforts at clearing up themess.
As indicated in my reply to Jim Woodward, Im all in favor of finding
order where we can; unification serves as a regulative ideal, sometimes
realizable within domains or subdomains. But Idont think theres a single
model of explanation, and even though Strevenss sophisticated kairetic
approach may be more inclusive than others, Iview explanation as a mot-
ley affair. Not all explanations are causal (think of mathematics, of theo-
retical unification in many fields, and those historical explanations that
show us what it was like to live in a particular place at a particular time).
Moreover, even among causal explanations, I dissent from the structure
Strevens wants to impose. The center of my dissent is concern about a prin-
ciple of transitivity of explanation:

(T) If C explains E, and C* explains some elements of C, then CC*, obtained by

replacing in C the elements in question by C*, explainsE.

Here Itake C and C* to be some corpus of statements and E to be a single

statement. (T)may need refinement to capture Strevenss exact intentions,
but Idont think anything Isay will depend onthat.

[176] The Philosophy of Philip Kitcher

Before elaborating my concerns about (T), I want to acknowledge an
important accomplishment of Strevenss discussion. He separates what
he takes to be the objective explanatory order from the organization of
explanation-seeking inquiry. His discussion of the division of labor among
sciences is aimed at defusing a line of objection to his preferred version of
reductionism, but even antireductionists can welcome its methodological
insights about the value of black-boxing.
What is wrong with (T)? From my perspective it is broken-backed in
several ways. First, Iwould suggest that explanations are directed at many
different sorts of questions: why questions, how questions, how possibly
questions, what questions, what is it like questions, and others besides. If
we start with one sort of question (Why E?), answer it by referring to
some mechanism M, and then consider an explanation that answers some
different type of question about parts of M, inserting that second explana-
tion into the first is not likely to improve our understanding of the phe-
nomenon recorded in E. Second, (T)allows for the production of odd hybrid
accounts. If we have available a causal account of part of the mechanism
(and we almost never have a causal account of the whole), the allegedly
deeper explanation CC* is a mixture of different levels of depth. Strevenss
reductionist thesis is most plausible if we imagine a sequence of full substi-
tutions, in which all the elements at a particular level of explanation give
way to a succession of uniform accounts at more fundamental levels.
Van Fraassens valuable account of explanations and questions sug-
gests another general difficulty. Continuing to suppose that explanations
answer why questions, lets add van Fraassens (1980, ch. 5; see also Dretske
1972)suggestion that explanation-seeking questions are contrastive. Now
its evident that (T)will be problematic when the contrast class shifts. We
start asking Why E rather than ?; C* addresses a question with a differ-
ent contrast class *; why should inserting C* in place of the elements it
explains enhance our answer to the original question?
All this is to air some rather abstract concerns. Let me make them more
concrete by considering examples. Ill start with a mundane instance from
a domain that has concerned many antireductionists:psychological expla-
nation. Ill work back toward the two cases Ive deployed (from transmis-
sion genetics and sex-ratio theory), on which Strevenss critical discussion
At breakfast there was only a scant supply of milk in the fridge, and
we had to be careful in rationing it. I promised to pick some up on the
way home after my morning meeting. But in the wake of my return, there
comes a wail from the kitchen, the cry of someone longing for a proper cup
of Earl Greyand Imust confess that Iforgot. Ihasten to explain:The

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [177]
meeting raised so many questions that I was distracted. I was thinking
about them all the way home, and simply walked past the grocery store.
Now there are occasions on which a simple explanation of this sort can
be deepened by introducing biological causes. Perhaps (although I hope
fervently that this is not so) there have been recent changes in my brain
that manifest themselves in increased forgetfulness. Or perhaps some neu-
robiological details would explain my lack of attention to the milk short-
age:Ifailed to remember later because my earlier registration of the need
for milk was superficial. Suppose, however, that factors of this sort are
entirely absent. My forgetting is just what Itook it to be:a product of my
tendency to become absorbed in the interesting points raised in a discus-
sion and consequent absentmindedness. Under these circumstances, Isug-
gest, the descent into neurobiological detail sheds no further light on why
Ibehaved as Idid. We know all there is to know about why Iwalked past the
store rather than turning in and fulfilling my promise. (Here, Ithink, the
abstract point about contrast classes is made concrete.)
Move now to a more important type of case, one in which the explanation
traces an etiology:Why did the English win the Battle of Agincourt, even
though they were exhausted, hungry, and seriously outnumbered? Military
historians typically dont answer the question by citing Henry Vs inspiring
words on St. Crispins Eve. They point to environmental conditions (the mud),
to brilliant tactics (the placement of stakes and bowmen), and to the skill of
the English archers. Of course the etiology could be easily extended:Why was
the ground muddy? How did the stakes get formed and placed? How did the
archers come to acquire their skills? We could amplify the causal story in any
number of directions and even lead it back through previous days, months,
years, centuries, millennia, epochs, all the way to the Big Bang. (T)supposes
that any and all of these extensions deepen the original explanation. Of
course Strevens will insist that his kairetic account doesnt suppose that the
extensions he envisages involve all the gory details; he emphasizes that his
preferred explanations abstract and filter. How that is to be done in examples
of this type isnt entirely clear to me. But as I envisage tracing Agincourt
into the remote past, Ifind it very hard to understand how any appeals to
causal preconditions or to basic physical mechanisms add to the explana-
tions military historians standardly give. Even if Iimagine something more
local and more promisingan account of where the wood used for the stakes
came from or a meteorological history that explains why the ground was so
muddyno advance in understanding seems to accrue. Once we see how a
small group of skilled archers could create havoc among the French cavalry,
the question of why a hopelessly outnumbered group of weary soldiers could
prove victorious is completely resolved.

[178] The Philosophy of Philip Kitcher

With this in mind, lets turn to the case of Dr.Arbuthnot and the male
years. What sex-ratio theory does for us is to redirect our explanation-
seeking away from asking etiological questions:we dont continue to treat
the data Arbuthnot examined as a sequence of historical events, each to be
understood by reconstructing the causal chains that led to them. The kai-
retic approach can account for this by proposing that Fishers 1:1 equilib-
rium at sexual maturity captures the causally relevant fact. But Strevenss
reductionism commits him to contending that our understanding will be
deepened by embedding the argument for the 1:1 ratio in our species (and
in many others) in a more general account of the ways sex ratios depend
on the genetic and environmental details. Explanations of those facts will
lead back into the emergence of diploidy and of environmental conditions
that generate the selection pressures Fisher takes for granted. Once again
we are launched on an inquiry into the causes of the fundamental features
of life on our planet. Its not clear to me that theres some definite point
in this extensive sequence of explanations at which we lose any power to
illuminate further the regularity that provoked Arbuthnots interest. But
Im confident that the returns are diminishing, and fade to zero well before
were confronting the Big Bang and the formation of the first elements.
Now, finally, to the principal example Ive previously used to cam-
paign against reductionism. Why do alleles on nonhomologous chromo-
somes assort independently? My 1984 paper contrasts a preferred mode
of explanationviewing genetic transmission as a pairing-segregation
process with an envisaged gory details explanation that identifies
genes with particular molecular structures and then traces the sequence of
reactions out of which the descendant gametes emerge. Strevens doesnt
treat this rival reductionist account as the only one, or the proper alterna-
tive. Insightfully his reductionism concedes that the pairing-segregation
approach is the right first-level explanation but that it can be deepened by
embedding it in a molecular perspective.
Using van Fraassens framework we can recognize that the contrast class
for the initial why question includes the thought that particular pairs of
nonhomologous chromosomes might have some relationship that makes
alleles on them more likely to be transmitted together. (If you think that
such chromosomes might be related so as to decrease the chances of joint
transmission, note that, in that case, there would be a different pairing
with increased probability of cotransmission.) Knowing that the segrega-
tion mechanism involves the poles of the spindle, we might follow up the
original why question by asking Why is it that the members of nonho-
mologous chromosome pairs segregate randomly? Thats a question that
can be addressed in terms of the mechanisms of constructing the spindle

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [179]
and drawing one chromosome from each pair to each pole. Cytology offers
an answer that abstracts from many molecular details, and extending the
original account in terms of pairing and segregation by introducing the
cytological details of spindle formation and chromosome migration really
does deepen the answer to the original question. This specific instance of
(T)is correct. Furthermore its easy to see why. The original why question
generates a more sophisticated why question, and one that retains the
same contrastclass.
Recognizing that feature of the situation helps us see why (T)is often
violated and why Strevenss vision of a long chain of ever-deeper explana-
tions leading back to fundamental physics is misguided. You have the cyto-
logical story about spindles and migration, and you see why (in the usual
case) the segregation is random. What questions do you ask next? There
are many. You might wonder about the etiology of the spindle or about the
selective history that has given rise to this intracellular mechanism. That
direction of extension would lead to the types of causal explanations weve
seen in previous examples, for which (T)plainly fails. Strevens surely wants
to probe the molecular bases of the mechanism invoked in the cytological
storybut what exactly is the explanation-seeking question to which the
more fundamental analysis provides the answer? Im inclined to think
that it isnt a why question at all, but something more like How does it
work? or What is going on at the level of the constituents? Even if you
try to frame a why question, it seems highly dubious that it will retain the
contrast class. We already know from the cytological story what is to be known
about why the segregation is random rather than biased to some particular asso-
ciation of nonhomologous chromosomes. Recall that the kairetic approach was
supposed to abstract from the details of particular instances and bring out
the crucial common feature. Thats already been done, and its been done by
cytology. Theres nothing more to deliver along theselines.
How does this relate to the argument originally advanced in my 1984
paper? Strevens chides me (gently) for a diagnosis in terms of natural
kinds, pointing out that this was hardly a concept beloved by those who
defended the unity of science hypothesis. That is literally correct. Its
worth pointing out, however, that something in the vicinity was a stan-
dard part of logical empiricist thinking from the late 1940s on, after
Nelson Goodman forced philosophers of science to come to terms with
the distinction between predicates that are projectible and those that are
not. So I recapitulate my diagnosis here, with a different emphasis: The
reductionists problem is that higher-level explanations turn on identify-
ing kinds that cut across the kinds featuring at lower levels. When the
advocate of kairetic explanation does his abstraction and filtering, he is,

[180] The Philosophy of Philip Kitcher

perforce, driven back to the level at which the important common factor
As my reply to Woodward indicates, I currently think about explana-
tion in a highly pragmatist way. The sciences are various, scientific explana-
tions are of many types and styles, and philosophical reconstruction ought
to honor the differences. Strevenss mansion on the hill is, in the end, a
beautiful folly, decorative but unfit for scientific habitation. Irecommend
instead building some functional housing in which real bits of scientific
practice can be athome.

Sp e c i a l - S c i e n c e Au t o n o m y a n d t h e Di v i s i o n of L a b or [181]

Toward a Political Philosophy


T here was a time when philosophers of science treated science simply as

a means of accumulating truth, or at any rate justified belief. Its central
question was how it achieved this excellent result and what differentiated
it from more benighted human practices with less respectable claims to
knowledgereligion or philosophy, for instance.
Nowadays, though most philosophers still see science as the preemi-
nent source of knowledge of the world, enthusiasm is often more nuanced.
Skepticism is common concerning the limits of scientific truth; perhaps
there are important areas of human concern that science cannot reach,
and perhaps there are still important things to be said for other kinds of
knowledge or wisdom. Arelated concern that is one focus of this essay is
whether, even if science delivers truth, this is enough to count it as a good
thing. This is one of the questions to which Philip Kitcher has turned his
attention over the past fifteen years or so, and his pathbreaking work on
this and related questions have helped to bring ethical and political issues
concerning science to the forefront of philosophical attention.
A good entry point into the topic is Kitchers (2001b) distinction between
truth and significant truth. There is an infinite number of truths:Ihave x
hairs on my head, Ihave fewer than x +1 hairs on my head, Ihave fewer
than x + 2 hairs on my head. For some x all of these are true, but none is
very significant. And of course there are many facts that have much greater
significance than this for somethe dates of my childrens birthdays, say,
are important facts for methat are of very little interest to science. There
is surely not time enough for science to enumerate all the truths there are,
and even if there were, there are some truths that we would very much like
to know sooner rather than later. Some of these are among the significant
truths. How do we decide which these are and thus direct our finite sci-
entific effort to finding out the things that we have some reason to want
This question perhaps seemed less pressing when many philosophers
believed that science was a unified whole. This belief was often interpreted
in terms of what has come to be known as the layer cake model, in which
sciences were ordered in a hierarchy, with physics as the most fundamen-
tal, chemistry to be derived from the laws of physics, biology from the laws
of chemistry, and so on through psychology, sociology, and the rest. The
most significant truths were fairly obviously the laws that articulated this
hierarchical structure, and the more fundamental, that is, the nearer to
physics, the more significant. Also highly significant were the descriptions,
for example of chemical entities in terms of physical constituents and their
relations, which made possible the derivations from more to less funda-
mental levels.
I think this view of a unified science is no longer defensible (Cartwright
1999; Dupr 1993), a position that is also central to Kitchers recent work.
Kitcher belongs to a growing number of philosophers of science who see sci-
ence as pluralistic, as consisting of particular theories or models designed
to address particular classes of question. Scientific models abstract from
the complexity of nature and aim to focus on a relatively small set of prop-
erties that are more or less decisive in generating a certain kind of phe-
nomenon. Unlike the unified picture, pluralism offers no internal account
of which truths are significant; this is something that must be decided by
some parallel process. Moreover for a pluralist there is little reason to sup-
pose there is any limit to the possible topics that might be pursued sci-
entifically. New interests will make possible new sciences, and there is no
reason to anticipate any limit to the interests we might acquire. Indeed
there is little reason to assume that the sciences we have allow only a finite
number of truths to be discovered. So deciding what questions to address
becomes an unavoidable part of the scientific process, and how these deci-
sions are to be made is something on which philosophers of science have
had little to say. To the extent that science is a public enterprise, funded
by the public for the general good, this must be a political problem. Hence,
Kitcher argues, we need a political philosophy of science.
The question of significant truth provides a point of entry into a political
philosophy of science. Once we have made this entry, however, other issues

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [183]
inevitably arise. One of the things that makes truth significant is that we
can do things with it:cure disease, grow more food, understand our place
in nature. How are these to be evaluated or compared? How and when are
scientific truths to be applied to individual or political decisions? Several
such questions will arise as this essay progresses.


Well-ordered science is the concept Kitcher (2001b) introduced to refer to

the way that institutions for the collection and storage of knowledge should
fit into the organization of a democratic society. In addition to questions
already mentioned about what scientific research should be supported, this
includes questions about certification (when a scientific claim should be
taken to have been established) and access (who should have access to what
parts of the accumulated, certified scientific knowledge within the state).1
And, especially important, how should scientific knowledge be applied to
making decisions of policy?
It is clear that current democratic states have yet to achieve a well-
ordered science. At least it seems evident that neither individuals nor
states reliably act in ways that seem clearly mandated by sound scientific
knowledge. Children die from measles because one thoroughly discred-
ited scientific paper has persuaded millions that a safe and effective vac-
cine causes autism; in the most affluent nation on earth a majority of the
population reject overwhelming evidence that we evolved over hundreds
of millions of years from simpler organisms through natural processes and
believe that our existence is to some extent the reflection of the intentions
of an all-powerful supernatural being.2 It is hard to argue that the alloca-
tion of scientific effort is optimal. Vast resources are devoted to amelio-
ration or cure of the diseases that afflict the old in the richest countries,
while little research is done on possibly quite simple measures that might
massively reduce the devastating impact of infectious diseases, gener-
ally on the young, in the developing world. Finally, even where scientific

1. These further questions are taken up in detail in Kitcher (2011b).

2. According to a recent survey by the Pew Research Centre, 60percent of Americans
believe that humans have evolved over time, as opposed to having existed from the
beginning of time in their present state. However, only 32 percent of Americans
believe that this happened through natural processes rather than as part of Gods
means to the creation of humans (Publics Views on Human Evolution, Pew Research
Center, December 30, 2013,
on-human-evolution/, accessed May 17,2014).

[184] The Philosophy of Philip Kitcher

knowledge has been acquired with obvious political implications, it is not
always appropriately applied. An overwhelming scientific consensus pre-
dicts catastrophic climatic changes resulting from carbon emissions, yet
little is done to reverse this disastrous process.
Kitchers (2001b, 2011b) Science, Truth, and Democracy and Science in a
Democratic Society represent a systematic attempt to address these vital
issues. Here Ifocus especially on two of the central issues:How is it to be
decided what scientific research should be undertaken? How should demo-
cratic decisions be made about the application of science to public policy?


What scientific research should be undertaken? Many scientists probably

believe this is something they are best able to decide and that the ideal
situation would be for all scientists to be free to address whatever ques-
tions they consider important and perhaps tractable. Whether or not this
is right, and Ishall say a bit more about it below, it is pretty clearly not very
relevant. In a world of finite resources in which much of science is paid for
by states, public decisions will need to be made about which projects are
funded. How should this bedone?
A modified version of the nave first answer is that the relevant deci-
sions be made by the community of scientists. In fact this seems close to
what happens in practice in most democratic states. Scientists send pro-
posals for research projects to funding agencies, and peers evaluate them
and decide which should be supported. Unfortunately this just moves the
problem up a level, to communities of scientists. If someone submits a pro-
posal for funding on the mating behavior of the Spangled Drongo, it will
be assessed, if not by experts on the Spangled Drongo, at least by experts
on bird behavior or perhaps experts on animal behavior generally. These
experts may decide the project is less worthy than one on, say, foraging
behavior among wombats. But they are unlikely to decide that less or no
funding should be allocated to any study of the behavior of wild animals,
a decision requiring an implausible kind of professional suicide. Whether
limited resources are better devoted to animal behavior or inorganic chem-
istry, say, does not seem to be something for which any particular kind of
scientist has particular expertise.
Kitcher, in fact, argues that some scientific projects are better not done
at all, and certainly should not be publicly funded. The example he con-
siders in detail in Science, Truth, and Democracy (2001b) is that of racial
science, the exploration of differences between people of different races.

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [185]
He discusses The Bell Curve, the notorious book by Richard Herrnstein and
Charles Murray (1994) in which they argue, first, that economic class gen-
erally reflects talent: people are poor because they are relatively dumb; and,
more notoriously, that the great overrepresentation of African Americans
among the poor reflects the fact that African Americans are on average less
intelligent than Americans of European descent. Suppose this is true. What
would be the benefits of knowing it? As Kitcher argues, there are few obvi-
ous benefits and some obvious harms. Centrally the position of a widely
disadvantaged group is likely to be substantially worsened as those who
have continued a centuries-old tradition of discrimination against African
Americans, for example in employment, feel justified in their discrimina-
tion and are (even) less motivated to end the practice. Worsening the posi-
tion of the already badly off is widely agreed to be a very bad thing.
There will no doubt be many who will respond to this argument by
insisting that we should want the truth whether or not it hurts. After all,
the discovery claimed by Herrnstein and Murray really may show that
alleged discrimination in employment isnt what it seems. Perhaps it is just
a reflection of the systematically lower qualifications of African Americans.
Just as women are (appropriately, it may well be claimed) underrepre-
sented in occupations requiring upper body strength, so African Americans
are appropriately underrepresented in jobs that require exceptional
At this point we need to be a little more critical of the assumption
that the research in question is likely to generate truth. In the first place,
research of this kind assumes that there is something being measured,
intelligence, that is somehow a purely biological property, independent
of upbringing and education. This is highly contentious. If, on the other
hand, we recognize that measured intelligence reflects the outcome of a
developmental process influenced at least as much by education as by any
natural endowment, then we see that the outcomes of intelligence tests
are likely to be a symptom of inequality rather than a measure of the cause
of inequality. Moreover this misinterpretation, if taken seriously, is likely
to justify the continuation of the unequal treatment that in fact causes
the perceived differences. Thus this research may not only be taken to jus-
tify unequal treatment, but it may help to perpetuate the phenomena that
form the basis of its misguided interpretation.
It is, in addition, highly debatable whether a further premise of the most
notorious aspect of the research in question is justified, namely that there
are two kinds of peopleAmericans of African descent and Americans of
European descentthat may turn out to differ systematically in their prop-
erties. As is well-known, genetic diversity is much greater within standardly

[186] The Philosophy of Philip Kitcher

distinguished racial groups than between groups, and given the extent of
human interbreeding it is difficult or impossible to provide sharp defini-
tions of who belongs to which group. Hence we find such bizarre conven-
tions as the one drop of blood rule that defined anyone with detectable
African ancestry as black. There is at least serious debate about whether
racial categories have any ontological validity.
These concerns about the categories in which this research is framed
raise a further concern. Quite apart from ethical questions about undertak-
ing the research in question, values are in fact already embedded in these
categories. Consider intelligence. Typical intelligence tests address various
competencies, for example the ability to handle and manipulate numbers,
to visualize the relationships between shapes, or to deploy vocabulary
effectively. Why these rather than, say, the ability to recognize the makes
of cars or to play fast-reaction video games? If such tests are not, as is often
said, merely a measure of peoples abilities to do intelligence tests, it must
be supposed that they are correlated with other skills, perhaps those sup-
posed to be useful in succeeding in human life. Such correlations could no
doubt be investigated. But what is success here? Clearly at some point this
is going to require a normative decision. This is not a problem of objectiv-
ity:there could be a perfectly objective measure of the ability to recognize
cars or shoot down virtual vampires. We choose to measure something we
call intelligence because we value outcomes that this supposed capacity
helps us to achieve.
Similar issues arise even more obviously for racial categories. There is
nothing ontologically defective with the category of people with at least
one drop of African blood (or, let us say, one ancestor native to Africa
within the past four hundred years). But why anyone would be interested
in this as a category for scientific research is another matter. Only a norma-
tive explanation, whether based on racial hostility or historical reparation,
seems possible. At any rate, the fact that social, political, or ethical values
are embedded in much research from the outset produces even more press-
ing questions about the desirability of the research.
I dont want to make a detailed argument that research on psychological
differences between races is epistemically misguided, though Ithink it is.
The point is rather that the premises that underlie this research are at best
controversial, and hence so are the results of the research. The combination
of research the outcome of which is likely to be both dubious and harmful
provides a paradigm for research it would be better nottodo.
I dont assume that both of these conditions must be present to make
research undesirable. There is a lot of epistemically sound research that
should not be carried out for obvious ethical reasons. It is argued that some

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [187]
Nazi research on issues such as hypothermia, while sound in principle, is
so morally repugnant in its methods that it has been intensely debated
whether it is even morally acceptable to make use of its outcomes (Moe
1984). Perhaps exploding atomic bombs in earthquake faults would be a
good way of learning about tectonics, but few people would advocate pur-
suing this line of inquiry.
The preceding remarks illustrate a central theme in Kitchers work:the
traditional idea of science as value-free is indefensible. Values are unavoid-
ably implicated not only in decisions about what topics we decide to inves-
tigate but in the concepts in terms of which we formulate the questions
we try to answer. These questions, in turn, cannot be answered apart from
decisions about the goals we would like our science to serve. This is not, as
is still sometimes supposed, an argument that science is subjective or that
the acceptance of its results is a matter of taste. It is an argument that we
cannot understand science properly without attending to these fundamen-
tal normative aspects.3
It is, at any rate, uncontroversial that there should be limits on what
scientific research should be undertaken at all, letalone publicly funded,
and Kitchers contribution in his discussion of the example of racial dif-
ference is valuable especially for exploring the wide range of important
respects in which this research may be highly undesirable even if from a
scientific point of view, in terms of its likelihood to discover truths, it were
The question then inevitably arises, how we should decide what research
should be undertaken. Kitchers answer, very roughly speaking, is that
such decisions should be made democratically. The democracy he has in
mind, however, is not the vulgar democracy of popular referenda but
a more Millian conception that recognizes the importance and value of
expertise. Ishall consider some aspects of this solution in the next section.
Here I note only that Kitcher does not advocate the simple and obvious
solution of banning research that is deemed undesirable, though presum-
ably the processes that he advocates for decision making would at least
make public funding of, for example, the research just discussed on racial
difference very unlikely.
I will confess, in passing, that Iam somewhat tempted to a more coercive
view. Democratic decision making, vulgar or sophisticated, is likely mainly
to affect questions of public funding. The Millian perspective that Kitcher

3. The role of values in science has been quite widely discussed in recent philosophy
of science. See, for example, Douglas 2009; Kincaid etal.2007.

[188] The Philosophy of Philip Kitcher

largely endorses sets a high bar against limiting the intellectual activities of
private individuals, and this certainly applies to scientific inquiry. However,
concerns about harmful science today apply not to private individuals in
their garages or back rooms but to vast corporations sometimes dwarfing
nation-states in their resources. This is not the place to consider in detail
whether such corporations are in fact engaged in research that violates the
constraints of well-ordered science. But to the extent that they are, it is
hard to see how they could be restrained from doing so other than by ban-
ning relevant domains of inquiry. If a well-ordered democratic state would
decide that research on racial differences should not be publicly funded, Im
not sure why it should not prevent such research being done outside the
confines of state institutions.
Kitcher does have an argument for not banning research:to avoid the
impression that research was banned for fear that it would produce results
that would do harm. In the present case banning research on racial dif-
ference may lead people to assume that the research is banned because it
is supposed that it will demonstrate that black people are less intelligent
than white people (Kitcher 2001b, 1057). But though this argument is
plausible enough, it is only one consideration among many that need to
be balanced. As Kitcher regretfully notes, this argument may well apply as
much to removal of public funding as to an outright ban. My suggestion
is just that a democratic process that effectively, and without overwhelm-
ing negative consequences, decided that research of a certain kind should
not, for ethical reasons, be publicly supported might very well have equally
good grounds for saying that it should be altogether proscribed.


The most important symptom of an ill-ordered science is the failure to

employ science to improve individual lives and social policy. Often the
relevant boundaries between the individual and the social here are open
to debate. In the case of vaccination it is widely understood as a paradig-
matic case for individual choice what medical technologies people choose
to employ for themselves and their children. Yet vaccination is only the
most obvious technology that problematizes such a view. The situation of
many people deciding not to treat their children with the MMR vaccine is
already causing serious health risks not only to their children but to the
children of others. One solution would be for states to enforce mandatory
vaccination. But while this may be justifiable in the end, it would seem
much preferable if people were able to understand and respond to sound

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [189]
scientific information so as to take sensible precautionary health measures
without coercion. To do this they must have either the ability to assess the
value of scientific research or have a high level of trust in scientific experts.
But the first option seems unrealistic in the foreseeable future, and the
second seems both frequently absent and anyhow problematic in various
ways. This brings us to the heart of the problem Kitcher (2011b) addresses
in most detail in the second of the books under discussion here, Science in
a Democratic Society:What is the proper relation between democracy and
Kitcher approaches the problem through what he calls the division of
cognitive labour. Various people, including Immanuel Kant, have been
described as the last person to know everything worth knowing, but there
is no doubt that that is a feat far beyond the reach of anyone currently liv-
ing. PubMed, an index of biomedical publications, contains almost 24mil-
lion citations at the time of writing, and a new one is added about every
minute. Even if 95percent of these have nothing very interesting to say,
this still leaves a million or so worth reading, and this is just one major area
of scientific knowledge. Hundreds of other databases can be found listing
tens or hundreds of thousands of resources on topics from Japanese his-
tory to gardening, from astrophysics to philosophy. (PhilPapers now lists
over one million books and articles.) Of course only a fraction of all this
should probably count as worth knowing, but then the problem is to find
out what fraction. The only solution to this problem is a division of cogni-
tive labor: for many different areas of knowledge there are some people
who know a good deal about what is known or credibly believed; these are
the experts. If we need to know something about an area on which we are
not an expert, rather than dive into this ocean of more or less reliable ver-
biage, we find an expert.
So far, so good. Now return to the central questions for political phi-
losophy of science:How do we decide what science should be done, or at
least funded, and how do we apply science to real practical problems? Two
problems arise. First, there is wide consensus that decisions on public
policy should be to some extent democratic. Placing decisions wholly in
the hands of an elite caste, whether they be politicians, philosopher kings,
priests, or scientific experts, notoriously leads to abuse and oppression.
On the other hand, democracy is likely sometimes to lead to policies that
scientific evidence suggests will be disastrous. Second, there are areas in
which, arguably, no expertise exists. Ihave already suggested that there are
no scientific experts on which science should be supported, as all scientists,
by virtue of the cognitive specialization that is their job, are bound to be
biased on this issue. Kitcher (2011a, 286)adds an additional interesting if

[190] The Philosophy of Philip Kitcher

controversial claim:that there are no experts in ethics. If ethics is a tech-
nology for promoting altruism and social consensus, there can perhaps be
facilitators of this process, but it is up to the citizenry at large to reach
How, then, do we steer between the Scylla of epistemic equality and the
Charybdis of Platonic authoritarianism in making these judgments about
the content and application of science? Since these are, of course, norma-
tive questions, it will be relevant to mention Kitchers (2011a) account of
ethics. Ethics is, for him, a social technology. One may or may not be con-
vinced by the evolutionary story he tells about the emergence of technology
as a solution to the social coordination problems faced by early humans,
but the conclusion to which this leads, that ethics is constituted by a set
of social practices and institutions to regulate behavior in the interest of
social harmony, seems compelling.
Implicit in Kitchers treatment of the present issue is his treatment of
science and democracy as social technologies in the same sense as ethics
(see DAgostino 2013). Surely this perspective provides the right perspec-
tive on the questions under consideration. Democracy, very crudely, is a
technology for pubic decision making, which more or less aims to reflect
the views of citizens, or of the majority of citizens, in decisions about pub-
lic policy. As a technology it is constituted by procedures such as voting,
parliamentary government, referenda, and, importantly, various more
informal channels through which citizens come to express their views in
ways that can become known to politicians and influence their behavior.
Iemphasize this last feature because it is arguably the quite recent prolif-
eration of channels of communication, notably on the Internet, that has
brought this clash between democracy and science to the fore. Citizens
today have access to quantities of information inconceivable a few decades
ago, though perhaps not accompanied by comparably effective tools for
assessing its quality. They also have multiple means of expressing their
opinions:through blogs, social media, email petitions, andsoon.
Science as a social technology has been studied intensively by sociolo-
gists and by scholars in Science, Technology, and Society (STS). Important
elements include university departments, laboratories, training programs,
academic journals, and hierarchies of power and authority. It is a technol-
ogy of knowledge production but also for the production of practical tech-
nologies for the satisfaction of human needs and wants and for interacting
with the world. It is also the home of expertise, of people taken to be quali-
fied to provide authoritative opinions on a wide range of issues.
It is easy to see how these institutions can come into conflict in the con-
text of public policy decisions. Consider, for instance, the debate in the UK

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [191]
and elsewhere in Europe on genetically modified (GM) foods. Emerging
from a remarkable explosion of knowledge of and technical capacity in
interaction with genomes, these were developed as products intended to
meet human needs for food.4 Within the scientific community they were
widely agreed to be an excellent pathway to produce many desirable char-
acteristics of cropsreduced competition from weeds, resistance to insect
attack, extra nutrients, and so onand hence to a more efficient and pro-
ductive agriculture. Though a fair amount of due diligence was seen as nec-
essary, as in any major modification of the human food chain, scientists
generally saw little reason to anticipate serious risks to health and claimed
possible benefits for the environment, including reduced need for herbi-
cides and pesticides.
Public reaction was another matter. Under the inspired label of
Frankenfoods, GM crops were widely portrayed as a God-like interference
with the natural order, likely to involve unacceptable threats to human
health and the environment. This perspective seemed quite unaffected
by declarations by scientific experts that such fears were groundless, and
political pressure driven by the strength of public hostility led to the large-
scale abandonment of the technology. In some sense this must presumably
be seen as a success for democracy:the public on balance rejected this tech-
nology, and the democratic government responded to their preferences.
This appears also, however, to be a departure from well-ordered science.
Assuming the experts are rightand there seems little compelling reason
to doubt it in this casean opportunity for a valuable technology, not to
mention a leading position in the development of the technology as an
internationally marketable commodity, appeared to have been passed up.
Why did the UK public, and publics in much of the rest of Europe, take
so strongly against this technology? The story is, unsurprisingly, very
complex, involving a variety of interests and arguments.5 There is a his-
tory that goes some way to explain the background of public suspicion of
scientists, for example the then recent fiasco in the United Kingdom over
the management of the bovine spongiform encephalopathy (mad cow
disease) outbreak, in which scientists had informed the government that
there was no risk to public health, something that turned out to be false.
There was widespread and understandable suspicion of the motives of the

4. They were also, of course, technologies intended for the enrichment of corpora-
tions and their shareholders, a factor that no doubt played some part in generating the
public suspicion directed towardthem.
5.Much more detailed discussion of some of these arguments can be found in
Barnes and Dupr (2008, chs. 6 and7).

[192] The Philosophy of Philip Kitcher

corporations, notably Monsanto, that were major funders of GM research.
It is also clear that this outcome was highly contingent, as no comparable
public opposition emerged to GM technology in the United States, where
millions of acres of GM corn and other crops are cultivated without anyone
being much concerned.
A disturbing issue raised by this case for well-ordered science is the
question of how informed public perceptions of science are even possi-
ble given the communications revolution alluded to earlier. Consider, for
instance, the average well-educated reader of a news story about the bac-
terium Klebsiella planticola. According to an account that circulated widely
a few years ago, this is an organism that had been genetically modified
with the perhaps excellent intention of helping to turn plant waste into
alcoholthis at a time when growing motor fuel in fields still seemed a
plausibly good idea. According to the story, however, a heroic amateur sci-
entist, Elaine Ingham, discovered that this organism had the potential to
provide a film over the roots of plants that would generate a lethal dose of
ethanol. Potentially, the story went, this could wipe out plant life across
the planet, and we would of course follow soon enough. If the reader were
concerned whether there might be environmental risks to genetic modifi-
cation of organisms, this might seem to be compelling evidence for a pes-
simistic conclusion.
About ten years ago I encountered this story and looked up Klebsiella
planticola on Google. Hundreds of websites reported this horrifying brush
with disaster, and none that Icould find raised any doubts about the cred-
ibility of the research. Indeed it is still widely cited today in support of the
alleged riskiness of genetic modification. Repeating this search today (May
14, 2014), the large majority of hits still report this result, often with dra-
matic headlines such as The Bacterium That Nearly Ate the World. There
are also a few blogs that report the discrediting of the research. With some
perseverance a source that is likely to be more impressive to the scientific
reader than a blog of unknown provenance, Nature Biotechnology (Fletcher
2001), reports that Ingham subsequently apologized to the New Zealand
government for submitting false claims about the ecological impact of
genetically modified organisms, claims that were backed up by a scientific
reference that didnt in fact exist. Note, however, that this is a resource
behind a paywall that is unlikely to be passable by the nonacademic reader.
In fact the modification that caused this furor was the addition of several
copies of a plasmid, a mobile genetic element taken from another bacte-
rium (Tolan and Finn 1987). Plasmids move quite freely between bacterial
species, so this was the kind of event that happens frequently in the wild.
If this gene were beneficial for wild strains of this organism, it is highly

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [193]
likely that it would have been found in the wild, though killing the organ-
isms with which it associatesall or most plantsprobably would not be
a selectively advantageous strategy. Moreover the extrapolation from the
ability of this modified bacterium to kill a plant in a confined environment
in a laboratory to its likely spread across the planet is, to put it politely, bio-
logically unlikely. But, and this is my main point, to the reader with limited
scientific knowledge and limited access to scientific resources (should he or
she even think of looking for them) the conclusion that GM technology had
almost wiped out life on Earth might seem perfectly reasonable.
Parallels with the case of the MMR vaccine are obvious. The notorious
paper by Andrew Wakefield et al. (1998) has been fully discredited in the
scientific community, disowned by Wakefields collaborators, and with-
drawn by the Lancet. Dr. Wakefield has been struck off the medical reg-
ister by the British Medical Council. Yet again, consulting Google on this
topic it is clear that in the wider public a debate still rages. Many suggest
that Wakefield was smeared by the medical establishment, certainly one
possible interpretation of what has happened to him. A particularly telling
comment in the Internet discussion is the following: Why dont you let
parents just make their own decisions? Do your research and make what-
ever decisions you think are best for your OWN children. Isnt that the very
essence of parenting? This seems to reflect the ideological reality in which
many parents in fact decide not to vaccinate their children and the conse-
quent rising incidence of disease.


In the previous section Iportrayed the current much discussed flood of

information as presenting a problem for anyone, except perhaps the rel-
evant expert, in deciding what to believe. One aspect of the well-ordered
science that Kitcher (2011b) advocates in Science in a Democratic Society
is a public institution of certification through which established results
get into the repository of socially accepted knowledge. Clearly Kitcher
would like to see the consensus of climate scientists on global warming, of
biologists on safe ways of improving crops, our best evolutionary theory,
and so on, certified for the repository, and the views of climate change
deniers, GM scaremongers, and intelligent design theorists excluded.
Iagree with him. But how is this to be done in a way that is democratic
and broadly acceptable to a democratic citizenry? Reflection on the broad
(democratic?) discussion made possible by the Internet is not altogether

[194] The Philosophy of Philip Kitcher

Kitcher does not endorse a nave or vulgar democracy. Certainly deci-
sions on scientific matterswhat science is to be funded, what results
are to be certified, and how they should be appliedare not to be decided
by referendum. Kitcher does base his answers to these questions on dis-
cussion and deliberation, but not of the anarchic kind that the Internet
represents. What is wrong with vulgar democracy? One problem is the
venality of some participants. Resistersclimate change skeptics, intel-
ligent design theorists, and otherswho oppose rationally indisputable
scientific results are in fact operating on the basis of different, and gener-
ally concealed, values. Climate science deniers are often associated with
the fossil fuel industry; scientists arguing for the safety of cigarettes were
frequently found to be funded by the tobacco industry. The proper kind of
transparent discussion will reveal these values, and the democratic major-
ity at least will find that they do not share them. It is not that the compe-
tent scientists do not have values of their own, but generally these will be
values that, again with proper explanation and transparency, the citizenry
will endorse.
Kitcher actually has two very different conceptions of the relevant con-
versation, the conversation that will in typical cases lead to convergence
between the informed views of experts and the views at least of well-
intentioned members of the public. The one that ultimately drives the
argument is a Rawlsian conversation of ideal deliberators who aim to reach
consensus in a reflective equilibrium. This is not an actual conversation but
a thought experiment, one designed to elucidate the conditions required
for a democratically and scientifically acceptable decision on matters of
certification as well as of application. Acentral feature of such an ideal con-
versation is that it will eliminate values that are not sustainable in such a
context. The greed of tobacco companies or oil companies, for instance, is
unlikely to be recognized as a good basis for public decision making.
On the other hand, although a thought experiment may clarify the cor-
rect decision on matters of scientific authority and application, it does
little to address the practical problems posed by ignorance and venality.
So Kitcher also proposes actual conversations between experts and repre-
sentative lay participants. The latter, selected to meet various criteria of
representativeness, are to be thoroughly briefed by the former and given
ample time to raise whatever concerns they may have, and will then be able
to reach informed decisions subsuming sustainable values. It should also
be noted that while the former conception appears to offer a criterion for
what the normatively correct (ideal) outcome should be, this outcome must
presumably be aligned with the naturalistic account of ethics as an actual
social technology that evolved from actual discussions between agents

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [195]
concerned to resolve social discord. It is not easy to see how this alignment
is to be achieved.
A somewhat cynical view is that the notion of ideal deliberators often
seems like little more than a philosophical delivery van for Kitchers policy
proposals (Brown 2013, 395). Slightly less cynically, it is difficult to see
what would make a deliberator ideal if, as Kitcher insists, there are no ethi-
cal experts.6 Not cynically at all, while Ihave a lot of sympathy for the model
of an ideal conversation as an intellectual tool, it seems to me somewhat
tangential to what is arguably the most interesting feature of Kitchers
work in this area, the aim of understanding and addressing the discordance
between two social technologies that Kitcher, and Iexpect most of his read-
ers, admire:science and democracy. For this is a technological problem, a
problem, that is to say, in social technology. Here it seems we should be
concerned with a social process, perhaps Kitchers actual proposed conver-
sations, rather than a theoretical, meta-ethical account of what would be
the normatively desirable outcome of a well-ordered science.
This may seem wrong for the following reason. Surely we need some
explanation of why we thought science was disordered in the first place,
and doesnt this require an account of value against which the present situ-
ation can be judged wanting? Ithink this worry does point to a real tension
in Kitchers writings on this topic. His concerns do certainly begin with
the firm conviction that things are amiss, for example the unwillingness
of democratic states to adopt serious measures to combat climate change.
And I dont doubt that there are good reasons for this. Indeed it might
well be sufficient to formulate these reasons in terms of evidence and
self-evidently appalling consequences regardless of the views that would
be reached by ideal deliberators. But of course both evidence and conse-
quences do involve normative assumptions.
At any rate, if one is serious about democracy, even a sophisticated
democracy free of all the familiar forms of vulgarity, one cannot prejudge
the question of what any suitable democratic process will decide about any
actual policy issue. Surely it cannot be a condition on an adequate account
of democracy that it always reach the correct decision as judged by some
external normative standard. Kitcher is an optimist, and he clearly believes
that a proper, nonvulgar democratic system will produce decisions more
or less of the sort that sensible people like he and Iwould prefer. But it
is conceivable, for example, that a democratic society might reach a fully

6. This is because Kitcher (2011b, 4950) sees the ethical project as fully egalitarian,
in which democratic processes are designed to produce consensus. It is the process, not
some privileged access to the proper result, that is important.

[196] The Philosophy of Philip Kitcher

articulated and thoroughly debated view that without major changes in
consumption patterns the world will be uninhabitable in two hundred
years, that Bangladesh will be under water in a hundred, and so on, and
democratically agree that they would rather keep their gas-guzzling cars
and air-conditioned homes. One may conclude, So much the worse for
our descendants and the Bangladeshis, or So much the worse for democ-
racy, but Idont think the problem can be solved by yoking democracy to
a theoretical account of what should be done. It might be, of course, that
there are ethical experts after all, and Kitcher is one of them. But in that
case shouldnt we try to devolve our decisions to Kitcher and his colleagues
rather than to an unreliable democratic process?
This brings me to the more practical kind of discussion that Kitcher
considers. Part of the practical resolution between democratic and expert
opinion is to be achieved by something like citizens juries, including both
a range of relevant experts and representatives of the public, the latter
selected to cover as wide a range as possible of the diversity of perspectives
within a population, and especially representing groups most likely to be
affected by the decisions being taken. The problems of vulgar democracy
will be addressed by making sure that all the facts as seen by experts are
presented, and time has been allowed to raise all questions or doubts in the
minds of the lay members of the panel. If all goes right, the panel will reach
some consensus on the matter at hand, and the process will have sufficient
general legitimacy that its decisions will command widespread respect.
It is important to say that this is the right kind of solution: it is an attempt
to mesh the two conflicting social technologies in a way that will reconcile
their conflicts. But I must confess to being a bit skeptical as to whether
such a system is likely to achieve the benefits Kitcher hopes for. Will the
fundamentalists, racists, climate change deniers, Christian Scientists, and
alien abductees have a privileged place at the table as groups likely to be
most affected by the reasonable consensus of these committees? Will they
be convinced? Even if they are convinced, will they not be perceived as trai-
tors by their wider communities? More generally will these committees be
perceived as genuinely democratic or merely as an opportunity for the sci-
entifically minded to generate propaganda for their views with the help of
ambitious stooges from the wider public? Given the experience that many
people are happy to draw their opinions from sources wholly opposed to
almost universal scientific consensus, it is hard to know why they should
change their views in the face of even the most well-meaning quango.7

7.In the United Kingdom a quango is a quasi-

autonomous nongovernmental

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [197]
All this is not to deny that groups of this kind can be a good idea. Indeed
they are demonstrably so. One highly effective quango, the UK Human
Fertilisation and Embryology Authority (HFEA), which regulates research
involving human gametes and embryos, is a very widely respected group, in
many respects very much the kind of entity Kitcher advocates.8 This body
includes scientists, doctors, women who have experienced fertility treat-
ments, a bishop, a lawyer, and even a philosopher. Unfortunately its deci-
sions do not appear to change the minds of those who oppose them. For
example, when the HFEA first licensed therapeutic human cloning in 2004,
there was outrage from a range of pro-life groups. Perhaps most telling
was the comment attributed to Josephine Quintavalle, of the pro-life
group Comment on Reproductive Ethics:It is very worrying indeed. We
have decisions of this magnitude being taken by an unelected government
quango.9 The absence of vulgar democracy may thus be used as a weapon.
I have argued that there may be a problem in harmonizing the ideal
and practical conceptions of conversation. This is most obvious in rela-
tion to the question of the scope of the problems that are to be addressed.
Recall that the kind of social technology that is the model for addressing
ill-ordered science, the ethical project, is taken to have originated in discus-
sions within small proto-human tribes. The problem of climate change, in
contrast, is global. The possibility of small groups sitting around a table to
hammer out a relatively local problem is appealing; perhaps even the adult
male citizenry of a Greek polis might gather in the agora to similar effect.
Representation of all the peoples of Earth, and all the different perspec-
tives within each nation, is a different matter and perhaps accessible only
to the ideal conversation. One problem with the ideal conversation when
it is more than some kind of rationalization of a practical process is that it
seems suspiciously like a job for an expert ethicist, of whom, Kitcher has
told us, there arenone.
The difficulties with a conversation over climate change do not end
here. Those most affected, in all but the most pessimistic scenarios, have
yet to be born. Perhaps there are many possible people who will not be
born unless we do something serious to address this problem. Who will
speak for the unborn or the possibly never to be born? Kitcher thinks that
among peoples central life goals are the well-being of their children and

8. Sufficiently well-respected that it survived the so-called bonfire of the quangos,

in which the Cameron administration attempted to dispose of as many such bodies as
9. See Scientists Given Cloning Go-Ahead, BBC News, August 11, 2004, http://, accessed May 20,2014.

[198] The Philosophy of Philip Kitcher

grandchildren and, perhaps by extension, future generations generally. Im
not sure. Ithink most people care a lot about the well-being of their chil-
dren, and many, but by no means all, care about their grandchildren. But
future generations long after their own death? Im much less sure. Perhaps
again they will need experts on ethics to speak on their behalf.


I have been somewhat skeptical about the proposals Kitcher sketches for
reconciling science and democracy. He is, Ihave said, an optimist, and in
this domain Iam more pessimistic. Kitcher thinks that scientists discredit
themselves by making excessive claims for their expertise and by acting on
values other than those that should legitimately underlie their professional
work. Iagree. If institutions could be constructed that would expose these
flaws and open their work to rational and civilized discussion, these flaws
could be removed and well-ordered science would regain the deserved trust
of the citizenry.
I fear that the sources of these defects go deeper and that they will be
even harder to remove than Kitcher supposes. Ihave described Kitchers
project as a political philosophy of science, and I wholeheartedly agree
that that is something we need. The politics, however, is largely limited
to the commitment to an admittedly sophisticated conception of democ-
racy. This is, no doubt, a good thing to be committed to. But many of the
problems that Kitcher is concerned with arise not merely from failures of
democracy but also from the intrinsic problems with the liberal, or increas-
ingly neoliberal, framework within which most current democracies exist.
In a social system that foregrounds competition between individuals it is
hard to imagine scientists who dont have their own agendas, even if these
are often no worse than the quest for personal success by doing good sci-
ence. More problematically, a system that encourages the accumulation of
wealth in large competitive corporations, and encourages these corpora-
tions to fund scientific research, will inevitably produce research infected
with the values of the corporate funders. These are issues, I fear, that
can be addressed only at a more systemic level than even the most well-
constructed institutional add-ons for enlightened public debate. They are,
that is to say, political rather than ethical problems.
Having expressed these doubts, Imust nonetheless reiterate my endorse-
ment of the importance of Kitchers project and the gratitude we should
feel for the work he has done to open up the philosophy of science to these
absolutely fundamental questions. The two books Ihave been considering

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [199]
on well-ordered science, together with his account of the ethical project,
constitute a systematic attempt to address the political question of the role
of science in society, and his account is full of valuable insights that should
remain part of this debate.
My suspicion, however, is that the attempt to construct a democratic
science may ultimately be impossible without more integration into the
problem of constructing a democratic society. In 2012 the worlds one hun-
dred richest people became $241 billion richer. They are now worth $1.9
trillion:just a little less than the entire output of the United Kingdom.10
The problem that Kitcher is discussing is democracy of voice, not equality
of resources. But as resources become ever more unequal, democracy of
voice becomes ever more unrealistic. It would be nice to see a democratic
socialist account of well-ordered science, though perhaps, given the dis-
tance we are from democratic socialism, Kitchers account, despite inevi-
table weaknesses, will be more useful.


Many thanks to Regenia Gagnier for discussion and helpful comments on

an earlier draft. The research leading to these results has received fund-
ing from the European Research Council under the European Unions
Seventh Framework Programme (FP7/20072013)/ERC grant agreement

Reply toDupr

For the past thirty-five years John Dupr has been one of my closest philo-
sophical interlocutors, someone from whom Ihave learned more than from
almost anyone else. His lucid intelligence, his wide-ranging knowledge, and
his deep humanity are apparent in the essay he has written herealong
with the wit so familiar to his friends. As has so often happened in recent
decades, his reservations about some of my ideas crystallize concerns Iam
already beginning to formulate, helping me to become more explicit about

10. See Bloomberg News, Worlds 100 Richest People Got $241 Billion Richer in
2012, Los Angeles Times, January 3, 2013,
03/business/la-fi-billionares-gain-20130103, accessed May 20, 2014. The figures are
derived from the Bloomberg BillionairesIndex.

[200] The Philosophy of Philip Kitcher

the next steps Idimly foresee. Ihope that John and Iwill continue to walk
together along ever more convergent philosophical pathways.
In 2001 I introduced well-ordered science as an ideal, shrugging off
the problem of implementing it as a task for others. By 2011, thanks in
no small measure to sympathetic critiques from Dupr, Ihad come to rec-
ognize the need to say more, to specify possible ways of moving toward a
better-ordered science. In this Ihave been inspired by the work of James
Fishkin (2009) on deliberative polling. As Dupr points out very clearly,
the gap between the ideal and the actual may not be so easily closed. With
respect to the important issues that concern meclimate change prom-
inent among themactual deliberators may be so far from meeting the
ideal conditions that attempts to set up the procedures Ienvisage may fail,
possibly with devastating consequences.
I like Duprs way of framing the problem:there are two social technol-
ogies, science and democracy, in a state of discord. Well-ordered science
might be viewed as saying what harmony between them would be like. The
challenge, one might then suppose, is to specify measures for turning dis-
cord into harmony. Fishkin and others interested in reforming and refin-
ing democracy might be hailed as the much-needed tuners. But on Duprs
reading of our current situation, the cacophony is extreme enough to war-
rant pessimism about the forms of tuning envisaged. Moreover, given my
theoretical commitments in thinking about ethics, specifically my qualms
about ethical expertise, Ihave deprived myself of some alternative ways of
avoiding disastrous consequences (an uninhabitable world two centuries
To make headway with these issues, Ithink we should start with Duprs
metaphor of discord but not follow the argumentative trajectory of the
previous paragraph. In the past Ive not been completely clear on two
important issues: the proper role of ideals and the concept of an ethical
expert. Duprs diagnosis prompts me to offer some clarifications, and also
to offer, on that basis, an outline account of social progress. If we take that
account seriously, we come, Ishall claim, to the judgment about our politi-
cal predicament that informs his conclusion.
People who specify ideals may do so on the basis of two quite dis-
tinct attitudes. Utopians think of ideals as identifying states we should
try to achieve, and for them the problem of implementation consists in
specifying steps that will lead us from our current position to the desired
goal. Many readers of my discussions of well-ordered science havenot
unreasonablytaken me to be a utopian. Consequently they have viewed
my proposals about implementation as bedeviled by severe difficulties,
obstacles Iought to have recognized. But in fact, though Ihavent confessed

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [201]
it before, and havent even made the distinction about underlying atti-
tudes, Iam not a utopian but a pragmatist. Pragmatists, Dewey prominent
among them, dont have to abandon ideals. They can see those ideals not
as characterizing a goal-state to be achieved but as marking a direction in
which to move. Ideals are to point us to ends-in-view (Dewey [1925] 1981,
ch. 3). As we move toward the current end-in-view we take stock of our
situation and continue to modify our aims. The principal contribution of
ideals is thus to help us in diagnosing the deficiencies of our current state,
alerting us to what is most problematic about it, so that we can start an
attempt to address the difficulties we face. Well-ordered science should be
viewed as just this type of diagnostictool.
My claims about ethical expertise need a similar sort of clarification. The
ideal of a conversation that involves representatives of all perspectives, fully
informed by the best available knowledge and mutually engaged (deter-
mined to take other perspectives seriously and to work for a solution all can
accept), is not something we can ever expect to realize. Nevertheless it draws
attention to my thesis that ethical authority is collective, and to the features
we should try to approximate in our collective decision making. None of us
has ultimate authority in the ethical conversation, but that doesnt mean
we are all on a par as potential deliberators. Some people are more igno-
rant than others, or more dogmatic than others, or uninterested in trying
to accommodate rival points of view. When such people participate in joint
discussions about what to do, things are not likely to go well. On the other
hand, there arefortunatelysome who have broad and deep knowledge,
who are genuinely concerned to listen to alternative perspectives, who try
to find ways of addressing the concerns of people who occupy those perspec-
tives, and who also have a talent for facilitating discussion. The late John
Rawls was an obvious exemplar. Our deliberations about vexed questions
would surely go better if he, and others like him, were in the room. Yet even
Rawls, wise as he was, should not have the last word on ethical matters, for,
as Ihave emphasized, ethical judgment is collective.
Pragmatism favors an approach to progress that abandons teleol-
ogy. From Dewey ([1909] 2007) on, pragmatists have found inspiration
in Darwin, thinking of progress from rather than progress to. (This theme
is also sounded in the last chapter of Kuhn 1962 and elaborated in Kuhn
2000.) Of course there are contexts in which a teleological notion of prog-
ress is apt. When we are traveling we typically measure our progress by
the decreasing distance to our destination. In many areas, however, a very
different conception is in order. Medicine does not progress by diminish-
ing the gap between human lives and some ideal state of health. What
would that ideal be? Rather it progresses by solving particular problems,

[202] The Philosophy of Philip Kitcher

and medical researchers (when medicine is reasonably well ordered) make
progress by finding ways to tackle the most debilitating diseases.
We ought to think about social progress in similar terms. The task at
hand is to identify the most urgent problems and take steps to resolve
them. Democracy should be viewed as a work in progress, something that
will never be finishedand certainly not something we can congratulate
ourselves on having achieved through the introduction of particular, rather
simple procedures. The discord Dupr discerns comes about because of the
flaws and limitations of two admirable institutions, and the function of my
notion of well-ordered science is to diagnose the troubles requiring imme-
diate attention. The attitude is meliorism:our task is to make things better,
not to make them perfect, and, in particular, to make improvements where
the defects are most dangerous.
Ill articulate these thoughts by focusing on the scientific question that
most worries both Dupr and me, the issue of climate change. As he points
out, it is possible that a democratic society might reach a fully articulated
and thoroughly debated view that without major changes in consump-
tion patterns the world will be uninhabitable in two hundred years, that
Bangladesh will be under water in a hundred, and so on, and democrati-
cally agree that they would rather keep their gas-guzzling cars and air-
conditioned homes. Id go further. Duprs bleak scenario seems to me
quite probable unless certain problems of contemporary democracies and
the place science occupies within them are recognized and addressed.
Climate science has already established that the world is warming (and
the oceans acidifying) and that, without serious and immediate reductions
in our use of fossil fuels, the lives of our descendants, a century or two
hence, stand a good chance of being intolerable. Thinking in terms of the
ideal of well-ordered science, failing to consider policies that phase out
carbon-based fuels is a clear violation of the envisaged mode of delibera-
tion. Either the people who will suffer in the future are not represented
or, if represented, there is no attempt to engage with their likely predica-
ments. Comparing the ways citizens of democratic societies actually
make their judgments with those of ideal deliberators, we can identify a
social pathology. Vast numbers of people are acting in the fashion of those
whom Dupr sees as abusing the privilege of entering the conversation, the
various types of fanatics and dogmatists who insist on realizing their pre-
ferred ends, come what may. What has gone wrong, and why? How might
the problem be solved?
Part of the trouble is that democracy is a much-abused termhence
my insertions of scare quotes. As I argue in my 2011b book, elections
enhance human freedom only when those who vote can recognize how

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [203]
their choices will advance (or retard) their central ends. When elections
are dominated by monied interests, when time at the public microphone
can be purchased, the voices of experts are drowned out by those of slick
spokesmen hired to trumpet the corporate view, and the Millian ideal of
free speech disintegrates. Those people who are centrally interested in
leaving a habitable planet for their children and grandchildren, and more
generally for those who will come after them, are confused or conned into
voting against that central interest. Democracy degenerates into statistical
plutocracy, a form of government in which the wealthy few invest so as
to increase the chances that voters will elect representatives who support
the well-heeled minority. In effect the electorate becomes a set of cogs in
a probabilistic machine, designed so the right results will flow from elec-
toral choices. Unfortunately the plutocracy is still statistical. Sometimes
the machine is imperfect. (A black man, concerned at least with climate
change, is elected.) But no doubt the plutocrats look forward to the day
when the bugs are all fixed and voters obediently traipse to the polls to
endorse the policies on which the wealthy rulers have agreed.
This, however, cannot be the whole story. Dupr is skeptical of my
thought that citizens are centrally concerned with future generations. Ive
come to see his skepticism as probably warranted. Yet again its worth ask-
ing why peoplequite likely a significant fraction of themwould violate
the conditions of ideal deliberation, not attending to the perspectives of
those who are distant in time and space. The closing pages of Duprs essay
contain the materials for an answer. Contemporary capitalism, with its
relentless emphasis on unbridled competition, generates another patholog-
ical condition. Many people are in no position to fashion their life plans
or central desires in the ways philosophers view as ideal. They lack educa-
tional opportunities that would open up a wide array of careers for them,
their chances of achieving genuine community with others are severely
limited, and, most obviously, they are faced with constant challenges to the
most basic tasks of securing their lives and those of their immediate fam-
ily. The pronounced inequalities, across the globe and within even the most
affluent democracies, constitute a condition under which its eminently
likely that the subtraction of unnecessary material goods will be viewed as
an unacceptable sacrifice:once a gas-guzzler has been acquired, its steer-
ing wheel will have to be pried from the owners cold, dead fingers. Myopia
prevails. There is no serious sense that life becomes valuable through con-
tributions to a larger human project.
Duprs conclusion expresses his own conviction that this social
pathology must be remedied. Well-ordered science, with its attendant
views about ethical authority, generates the same diagnosis. It marks out

[204] The Philosophy of Philip Kitcher

problems not only with the institution of science but also in the democra-
cies we celebrate and in the ways the neoliberal framework narrows the
vision of those whose deliberations should shape policies. So Iagree with
Dupr that, as things stand, we cannot expect that deliberative polling or
citizen juries will amend our predicament. To misquote a famous apologist
for the most prominent statistical plutocracy:youd have to do the delib-
erative polling with the deliberators youve got. More radical reforms are
required. Pragmatically construed, the framework of my 2011a and 2011b
books deliver the diagnosis.
But what exactly is to be done? Ihope that ideas still matter and that
philosophical clarification and diagnosis, possibly repeated in variant
forms, can arouse enough support to spark a political movement. So Iclose
by amending a line Benjamin Britten borrowed from Wilfred Owen as
the epigraph to his War Requiem:all a philosopher can do today is warn.
Perhaps John Dupr and Ican join forces in the work of warning.

T o wa r d a P ol i t i c a l P h i l o s op h y of S c i e n c e [205]

Kitcher onScience, Democracy, and

Human Flourishing


In his book Science in a Democratic Society, Philip Kitcher (2011b,

100)declares, Not much time, if any, has been devoted to wondering how
public knowledge might be shaped so as to be good for democracy. We lack
any convincing theoretical conception of how Science contributes to valu-
able goals. Other recent works by Kitcher (1993a, 1996, 2001b, 2011a)
address related themes, albeit with different emphases and varying levels
of generality, and it seems fair to say that the challenge of reconciling sci-
ence and democracy, or more broadly science and human values, has been
one of his major philosophical preoccupations of the past twenty years. In
this essay Ireflect on Kitchers ongoing project (his own favored word, for
reasons Ishall return to presently) from the standpoint of a historian of
science:What, if anything, can history tell us about the prospects of sci-
ence serving the commonweal in a democratic polity?
Kitcher (1998) had already raised these themes in A Plea for Science
Studies. Although the mid-1990s debates about realist versus social con-
structionist positions that framed the Science Wars now have an antique
ring to them,1 one section of the 1998 essay remains as urgent today as it

1. Hacking (1999) offers a cogent and nuanced account of the debate.

was then. Under the rubric The Real Challenges, Kitcher threw down a
gauntlet to all scholars involved in the study of science:Reflective people
(whether scientists or not) want to know whether research in various areas
is skewed by the values of particular groups and, at the broadest level, how
science bears on human flourishing (46). In Kitchers view neither phi-
losophy of science nor science studies had acquitted itself with much glory
to date, although he gave the latter credit for at least trying. Recent bra-
vura work in science studies, such as Edwards (2010), Jasanoff (2012), and
Hecht (2012), is proof positive of the power of that fields finest analyses
to illuminate the questions and quandaries Kitcher underscored:What are
the politics of contemporary science? What is the science of contemporary
politics? Is either, both, or neither promoting the public good? His own
subsequent work has approached these problems from the standpoint of a
philosopheras a philosopher of science, first and foremost, but also as a
political theorist, an ethicist, and (probably the label he would most readily
embrace) an engaged citizen. The aim is to provide a framework of values
epistemic, ethical, and politicalfor thinking about current decisions at
which science and politics intersect and for imagining better arrangements
to guide future policy.
Since the bulk of the current discussion about science and the public
good is understandably aimed at the present and the futurewe must
decide now about whether and how to respond to climate change or how to
apportion how much collective wealth to which research areas if we want
to have any kind of livable futureit would be reasonable to ask what the
contribution of history could be. Doubts abound. Why look backwards
when the direction we must chart is forward, fast forward? Isnt it intrin-
sic to the very nature of science, as the motor of modernity, to transform
itself and its ambient society repeatedly and dramatically? Isnt modern
science the factor most responsible for disrupting any induction based on
the premise that the future will be like thepast?
There are at least three reasons for nonetheless training a historical
lens on Kitchers proposals for how science might better serve the cause of
human flourishing:the first is quite general; the second is more specific to
this subject matter; and the third is more specific still to this thinker. First,
history in general limbers up the imagination for the thinkable and the pos-
sible. Present arrangementspolitical, intellectual, socialoften appear
inevitable, not only the way things happen to be but also the way things
must be. Although the past is rarely, if ever, a blueprint for the future, it
can at least undermine the apparent necessity of the presentand thereby
clear the way for alternatives. Second, the history of science in particular
sheds light on how science has functioned (and malfunctioned) in a variety

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [207]
of cultural and political contexts. Since many debates about the advisability
of bringing science and democracy into a different alignment hinge on the
risks of thereby corrupting both science and democracy, it is useful to have
a trove of past examples and counterexamples against which to test claims
made by all sides. (The same is of course true, mutatis mutandis, for the
history of the polities and values with which science is to be brought into
alignment, but other, better qualified scholars will have to pronounce upon
these.) Third, Kitchers own work in the philosophy of science, beginning
with his earliest publications on mathematics, has been deeply informed
by historynot just the usual toy examples retailed at secondhand but
detailed, thorough analyses based on immersion in the primary sources.
This is a philosopher who thinks with history. In the case of his recent work
on science and the public good, this historicism has seeped into the concep-
tualization of the problem:he conceives of both science and democracy as
under construction, not as finished achievements but as projects in the
making. On Kitchers account these institutions are not so much like bro-
ken artifacts in need of repair as evolving organisms with contingent pasts
and open-ended futures.
In what follows I bring all three historical perspectives to bear on what
I take to be the most fully developed version of Kitchers (2011b) views
on science and human flourishing to date. I begin with a brief summary
of what seem to me the principal points in Kitchers argument and then
examine how his analysis might promote human flourishing, mostly from
the standpoint of a historian. My conclusion returns to Kitchers vision of
the project.


Kitchers (2011b, 20)point of departure is specific to a place and time:in

and- now debates about evolutionary theory, stem cell research,
genetically modified organisms, climate change, and any number of other
recent controversies about the public role of science in the United States,
the authority of Science has been eroded. Despite abundant evidence of
scientific successes in all areas of daily life, from nanotechnology to bio-
medicine, public distrust of expert scientific opinion is rising. As Kitcher
notes, the tension between, on the one hand, egalitarian polities that
emphasize transparency and discussion open to all and, on the other, small
groups of experts with knowledge accessible only to a few is nothing new.
Moreover the past involvement of such elites in perfecting new and more
hideous weapons, promoting dubious social programs such as eugenics, or

[208] The Philosophy of Philip Kitcher

conducting ethically indefensible medical experiments gives some ground
for suspicion even among disinterested and well-educated citizens. Even
if they believe the science, they distrust the motives and judgment of the
scientists. Yet the United States, like all other industrialized societies, is
saturated with science and science-based technology, which affect the pros-
perity, security, health, and future prospects (not to mention the daily sur-
vival) of every citizen. Never has the country been in more pressing need of
the scientific counsel that a significant segment of the citizenry and their
representatives in elected office appear toshun.
What is to be done? Kitchers therapy flows in large part from his diag-
nosis of the ailments, which afflict American democracy as much as they
do science. Democracy suffers from mistaken conceptions of transparency
and free speech; science is hobbled by an ideology of freedom from values
that is false to scientific practice and therefore always a potential trump
in the hands of those who seek to discredit this or that scientific consen-
sus. What Kitcher proposes are not concrete practical measures to alleviate
these ills (though he surely is not averse to these) but rather a theoretical
framework for thinking our way out of the impasse between science and
democracy that would require the transformation of both parties to the
conflict. Three ideas (or ideals) guide his analysis:the ideal conversation,
the division of epistemic labor, and well-ordered science.
The ideal conversation aims to make existing democracy more, not less
participatory and egalitarian and could serve as a regulative political prin-
ciple even for debates that have nothing to do with science. Ideal conversa-
tions are inclusive, overflowing the boundaries of nation-states to embrace
not only the worlds current population but also future human generations
(and perhaps even some other species); they are egalitarian, at least as far
as ethical matters and life chances are concerned; and they are governed by
the norms of what Kitcher calls mirroring. Mirroring is a complex notion
that comes in both primitive and extended forms, but its core intuition
is simple and Smithian (Smith [1759] 1976):participants in the ideal con-
versation should observe a sympathetic reciprocity, desiring to weight the
egoistic preferences of others equally with their own. More advanced con-
versationalists filter their egoistic preferences through the sieves of the
ethically permissible and the factually possible in order to minimize the
clash of incompatible desires. Ideal conversationalists form their sympa-
thetic desires by extended mirroring of the desires of others, achieving the
desires they judge to be the best balance among the varying assessments
(indefinitely iterated) made by fellow participants (Kitcher 2011b,52).
Whether even approximations of such conversations are possible, and
if possible, desirable, is a question that Ishall postpone for the moment.

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [209]
Note only how radically Kitchers vision of democracy diverges from cur-
rent reality and also from more familiar liberal ideals:sympathetic mirror-
ing is worlds away from John Stuart Mills ([1859] 1975)jostling, agonistic
free marketplace of ideas, in which rival positions clash in gladiatorial
combat, each armed to the teeth with its strongest arguments. It is equally
at odds with Isaiah Berlins ([1950] 1979, 14760) conviction that there
are irreconcilable differences among communal visions of the good, each
the product of a distinct history. Kitcher shares Mills commitment to fair-
ness (everyone has a voice on matters upon which no one is an expert)
and Berlins respect for pluralism (there are multiple visions of the good
life). But Kitchers subtle but strenuous requirements of mirroring and
mutual engagement demand that the conversationalists modify their
desires in light of those of all the others, striving toward a kind of prear-
ranged harmony through sympathy.
If the ideal conversation expands participation in some discussions, the
division of epistemic labor narrows it in others:

Consider the entire range of questions pertinent to public life, all matters about
what the society should aspire to and how it might realize whatever aims are
set. These topics are partitioned, divided into non-overlapping sets, and for each
set in the partition except one [ethical matters], a particular group of people is
designated as authoritative with respect to that set. For the remaining set, epis-
temic equality holds. (Kitcher 2011b,21)

Therefore, not everyone will be qualified to pronounce upon all issues,

although the circle of discussants can be expanded if citizens are willing
to be tutored in the esoteric knowledge involved in this or that particular
decision (114). Kitcher acknowledges that the partition may be controver-
sial and that its exclusionary implications may seem incompatible with
democratic openness:Why shouldnt, for example, creationists enjoy equal
rights with evolutionary biologists to debate which account of the origins
of life will be taught in public schools or guide publicly funded research?
Kitchers reply grants the role of broad values in scientific research; val-
ues per se are not an epistemic disqualification. But values that endorse a
dogmatic adherence to some scripture are:Broad schemes of values can
play a legitimate part in scientific practice, but they are required to be sus-
tainable in an ideal conversation (60). The chimeric epistemologies
part doctrine, part common sense, part scienceof religious opponents
of Darwin violate the cognitive conditions of mutual engagement (60).
More generally Kitcher would extend this ban to any appeal to the super-
natural in the context of both public knowledge and the ideal conversation

[210] The Philosophy of Philip Kitcher

concerning how to use that knowledge. If, however, dogmatic religion ille-
gitimately undermines the division of epistemic labor, ethics transcends
it altogetherand rightly so, according to Kitcher. There are no ethical
experts; all people have an equal right to participate in discussions on how
to lead a worthwhile life and equal opportunitiesserious equal opportu-
nities to realize such a life(50).
Well-ordered science would be regulated by the ideals of mutual
engagement and the division of epistemic labor working in tandem: A
society practicing scientific inquiry is well-ordered just in case it assigns
priorities to lines of investigation through discussions whose conclu-
sions are those that would be reached through deliberation under mutual
engagement and which expose the grounds such deliberation would
present (Kitcher 2011b, 114). This is a tall order, perhaps an impossible
one:the affective (extended mirroring among the entire worlds present
population, plus representatives of the interests of future generations)
and cognitive (comparative weightings of the significance, costs, feasi-
bility, and prospects for success of all competing investigative strate-
gies) conditions are so demanding that Kitcher concedes that any actual
conversation of this type is impossible (115). Nonetheless he persists
in the assertion that embracing such an admittedly unattainable ideal
would at least orient conduct and inspire citizens to achieve ever better
Realizing that both scientists and citizens are likely to query whether
the ideal of well-ordered science is desirable (not just whether it is pos-
sible), Kitcher steers between two likely criticisms:citizens might protest
that all that is needed is a majority vote on competing areas of scientific
research, in the same way that other competitive claims to scarce collec-
tive resources are settled (why should the decision between, say, cancer
versus climate research be decided any differently than that between guns
versus butter?); scientists will object that to compromise the autonomy of
science is to risk its destruction. Even practical applications will come to a
grinding halt when nonscientists set sciences goals, not just the advance
of theoretical understanding. (Lysenkoism will inevitably be mentioned,
and perhaps also the meager therapeutic harvest of the hugely expensive
war on cancer.) Kitcher counters that democracy by majority vote is vul-
gar, permanently prejudicial to the desires and aspirations of the minority
and arguably also to those of the majority, since it tends to flatten and
abridge deliberation. (Contrast the conditions for decision making by a jury
aiming at consensus versus a legislature striving only for a bare majority.)
As for scientific autonomy, Kitcher briskly asks to see the statistical evi-
dence that science in the service of externally imposed aims really has been

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [211]
less fruitful of results and advances in understanding than blue sky or
curiosity-driven research.
However utopian the ideal of well-ordered science may seem, Kitcher
contends that it might still serve as a pole star by which to guide future
attempts to improve the forms of democracy and science we have inherited
from the vicissitudes of histories driven more by accident than by design
(just as, by analogy, an ideal of true equality might guide political and social
reform, even if it is an asymptote never reached). He asserts that even
partial attempts to realize the ideal would be superior to current arrange-
ments (e.g., majority rule in politics, increasing privatization in science)
and offers suggestions concerning specific dilemmas likely to arise (e.g.,
what principles might adjudicate competing claims as to which diseases
most urgently deserve research efforts).
Although a strict observance of the division of epistemic labor may seem
to restrict citizen participation in decisions about science-informed public
policy, Kitcher is particularly concerned with how the circle of those able at
least to appreciate the factual constraints on and probabilities of success of
alternative lines of scientific investigation might be widened to include more
than just specialists. Some citizens might be taken behind the scenes to
observe relevant research firsthand or tutored in the details of a burning
question; more scientists might dedicate more of their time to explaining sci-
ence clearly and engagingly to the general public; scientific education might
be improved for all. Similarly (but more vaguely) citizens might be educated
to be both aware of and sympathetic to the plight of others:One part of
the remedy would take seriously the idea that part of education consists in
the encouragement and expansion of altruistic tendencies (Kitcher 2011b,
130). Even achieving a first approximation of Kitchers ideal of a new social
contract between democracy and science would seem to require substantial
modifications of what citizens can be expected to know and to feel. It is, to
use Kitchers own word, a project, and a highly ambitious one atthat.


What reason do we have to think that such a project could succeed? And
even if we were persuaded of its bare possibility, at least in some approxima-
tion, what reason do we have to think that Kitchers ideals for science and
democracy would be enough of an improvement to warrant the Herculean
effort of striving to realize them? Kitchers answers to both of these chal-
lenges appeal to history, but history in different registers and enlisted to
serve differentends.

[212] The Philosophy of Philip Kitcher

Central to Kitchers arguments in favor of the ideal of well-ordered sci-
ence is the claim that both democracy and science as we now know them
are the results of haphazard histories that have optimalized neither. There
is therefore no reason to be resigned to or complacent about the status
quo; pace what politicians and scientists may tell us, we do not live in the
best of all possible worlds. The historical developments that have produced
contemporary science and democracy were contingent (things could have
turned out otherwise and need not remain as they are) and the results
imperfect (things should not remain as they are). Note that the could
and should arguments are in principle independent of one another, as
an evolutionary analogy makes clear. Contingencies of every sort (random
mutations uncorrelated with environmental changes) govern the evolu-
tion of species, not some telos toward which the species is striving or a
design imprinted at its origins. Yet the helter-skelter processes of evolu-
tion can nonetheless produce adaptations so ingeniously fitted to the spe-
cies way of life that they give the illusion of being masterfully (though not
optimally) engineered. This illusion is the departure point for the argu-
ment from design, and although Darwin ([1859] 1964, 18694) convinc-
ingly refuted it, even he felt its force. My point is not to revive the illusion
of design but rather to point out that haphazard history can and does
produce impressively (not perfectly) well-adapted institutions, just as
haphazard evolution produces impressively (not perfectly) well-adapted
organisms. So haphazard products cannot be reliably inferred from hap-
hazard processes.
What about the case of modern science? Kitcher is concerned with a
particular maladaptation, which he attributes to sciences contingent his-
torical development. Because science has evolved by happenstance, its
position within various societies shaped by contingent events and the
opportunities they offered, its compatibility with other institutions, nota-
bly democracy, is a matter of luck rather than reflection (Kitcher 2011b,
100). Noting that contemporary science preserves elements from differ-
ent stages of its long history (theoretical contemplation from Aristotle
and his disciples, autonomy from the private gentlemanly science of the
seventeenth-century virtuosi, the union of teaching and research from the
nineteenth-century German university and its imitators worldwide, public
authority and prestige from its spectacularly successful twentieth-century
applications, from antibiotics to atomic bombs), Kitcher suggests that at
least some of these fossilized ideals and practices have become dysfunc-
tional. The enshrinement of theoretical contemplation or autonomy as
core values of science may have made sense when inquiry was largely a pri-
vate affair with little or no public impact, but in an age in which science and

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [213]
science-based technology have repeatedly transformed society, for good or
ill, these commitments are at best antiquated and at worst dangerous.
For the historian Kitchers account raises two questions:Is the account
accurate where it matters to Kitchers theses? If it is in the main accurate, is
it a legitimate use of history? It would be pedantic to criticize what is meant
to be a sketch as if it were a treatise, but there is one consequential aspect
of Kitchers depiction that most historians of science would query:the ori-
gins and pervasiveness of scientific autonomy. In cultures that have insti-
tutionally supported systematic inquiry since the Renaissance, examples
of autonomy as an ideal are rare, and actual examples are rarer still. All
the luminaries of early modern science Copernicus, Kepler, Galileo,
Descartes, Boyle, Hooke, Newton, Leibniznot only vaunted the utility of
their scientific pursuits; they also practiced what they preached. Whether it
was Kepler (1615) estimating the volume of wine barrels for the city of Ulm,
Galileo marketing a proportional compass for military engineers (Valleriani
2010, 2740), or Leibniz corresponding with alchemists about artificial
phosphors (P. H. Smith 1994, 24855), these natural philosophers were
profoundly committed to science in the service of worldly goals. Thomas
Shadwell ([1676] 1966)and other Restoration wits might have poked fun
at the Royal Societys investigations of the blue of plums and luminescent
lamb shanks, but the Fellows pursued such topics with practical applica-
tions in mind. The Royal Societys French counterpart, the Acadmie royale
des sciences in Paris, was from its inception charged with the responsibility
of offering technical advice to the state on everything from large engineer-
ing projects to smallpox inoculation (Hahn 1971). If anything, the cult of
public utility only intensified during the eighteenth century. Even with the
rise of the prestige of pure science (reine Wissenschaft) in the latter half
of the nineteenth century in the context of the German research univer-
sity (Daniels 1967), German scientists of the first rank, such as Justus von
Liebig (Rossiter 1975)and Hermann von Helmholtz (Cahan 1989), prided
themselves on their contributions to the advancement of agriculture and
electromagnetic technology.
Far from being a holdover from an earlier era, the insistence on the
autonomy of science seems to have first emerged in twentieth-century
democracies, usually as part of a defense of public funding for pure or
basic or fundamental science without immediate prospect of applica-
tion (but usually with the promissory note of future practical benefits).
Although such pronouncements date at least as far back as the 1920s
(Pielke 2012), Kitcher is surely correct to identify Vannevar Bushs ([1945]
1980)ScienceThe Endless Frontier as the locus classicus of the view that the
autonomy of basic science guarantees the fruitfulness of applied science. As

[214] The Philosophy of Philip Kitcher

the chief scientific administrator of the Manhattan Project, Bush asserted
confidently, Statistically it is certain that important and highly useful dis-
coveries will result from some fraction of the undertakings in basic science;
but the results of any one particular investigation cannot be predicted with
accuracy (19). However, the key points here are that Bushs articulation
and defense of the value of scientific autonomy (often conflated with basic
research) were largely the creation of the twentieth century and a response
to massive increases in the public funding of science. Whereas scientists in
earlier centuries (e.g., Francis Bacons [(1620) 1960,6] dedicatory epistle to
King James Iin the Novum organum) had sought such support on a grand
scale by promising all manner of practical benefits, and governments had
in isolated instances financed expeditions (e.g., to observe the Transits of
Venus in both the eighteenth and nineteenth century [Woolf 1969]) or
ambitious projects (e.g., Charles Babbages plans to build the Difference
and Analytical Engines [Swade1991]), the sustained public funding of sci-
ence on a vast scale after World War II, especially in the United States, was
unprecedented and transformative (Greenberg 2001, 7888). But he who
pays the piper calls the tune:once on the payroll, the scientists were under
new and intense pressure to comply with research directives and regula-
tions stipulated by their paymaster.
Does this alternative history of scientific autonomy matter to Kitchers
theses? In some ways it is fuel for his fire: a new situation calls for new
values. If the public funding of science has increased so steeply in the past
half-century, then the public has every right to demand a louder voice in
the kind of research pursued. But novelty cuts both ways. Assertions that
past curtailments of scientific autonomy in the service of political goals
have or have not had baleful effects on the quality of science are irrelevant
if past and present are no longer comparable. (Kitcher counters warnings
about Lysenkoism with a call for a more systematic survey of the historical
evidence.) In effect this strengthens the case of the advocates of autonomy,
who can plausibly argue that current interventions at all levels of scientific
research, from choice of topic to design of investigation to presentation of
results, amount to a historically unprecedented level of interference that
threatens the very existence of science.2 Either way, rethinking the history
requires rethinking the arguments.
Turning to the second question:Can history provide grounds for believ-
ing that the ideals of mutual engagement and well-ordered science can be

2.Scientists see regulation of research methods as a particularly potent threat


K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [215]
at least approximately realized and that both science and democracy would
be better off if they were? In the case of mutual engagement, Kitchers use
of history is both variegated and subtle. He musters examples from both
actual and speculative history to make his case that over the course of cen-
turies and even millennia, human beings have shown themselves capable of
widening their circle of sympathies to include ever more people entitled to
the full dignity of personhood. The gradual de jure abolition of slavery all
over the world (de facto abolition, alas, has yet to be achieved) is an impres-
sive example for Kitchers case:most citizens of modern polities recoil at
the intuitions that Aristotle and his ancient Athenian contemporaries con-
sidered self-evident on this subject, however much his writings on other
ethical and political topics may still resonate. With such examples in mind,
Kitcher (2011b, 48) formulates a different version of moral progress: not
progress toward any ethical truth but rather progress from, measured in
terms of our ability to solve problems.3 If humans are capable of so remark-
able an expansion of their sympathies in the teeth of their economic and
other egoistic interests, why is it not realistic to hope and strive, Kitcher
asks, for a still broader embrace of the predicaments, perspectives, and aspi-
rations of others, embracing even unborn generations and other species?
So far, so good: the cautious historian may point out the glacial pace
of such conceptual, affective, and institutional transformations and their
costs in blood, wealth, and stability; the cautious anthropologist may won-
der how far the radius of human sympathy can stretch without snapping
but neither would challenge the relevance and force of Kitchers appeal to
actual historical analogy. However, both may falter over his use of spec-
ulative history. A keystone of Kitchers notion of mutual engagement is
the ethical equality of the participants in the ideal conversation. You and
Imay differ concerning how well we understand the facts and probabili-
ties of the matter at hand, and the epistemic division of labor dictates that
Ishould defer to you if you are better informed. But there is no deference
concerning whose interests in the possible outcomes should prevail. Why
should this be so? There are many historical precedents of societies that
have institutionalized deference to various sources of authority:religious
saints, wise elders, powerful chieftains, rich bankers, hereditary princes,
charismatic politicians, violent warlords. Hierarchies are the rule, not the
exception in the history of human societies.
Of course Kitcher knows all this. However, he interprets historyor
rather prehistory, and therefore speculative historyas offering an earlier

3. Ethics as problem solving is developed at much greater length in Kitcher (2011a).

[216] The Philosophy of Philip Kitcher

and (therefore?) superior alternative. Drawing upon descriptions of con-
temporary hunter-gatherers, Kitcher (2011b, 43)imaginesthat

our Paleolithic predecessors sat down to decide on the precepts for governing
their group life.... All adult members of the band are to be heard, and the wishes
of each must be considered. To diverge from egalitarianism of this sort would
risk the survival of the group, for all had to pull together on occasion, to meet
the challenges of the environment.

Subsequently more complex societies formed based on a more refined divi-

sion of labor, and hierarchy supplanted equality as the model for political
decision making. Kitcher proposes to undo the distortion wrought by his-
tory and renew the ethical project by returning to the putative egalitari-
anism of our Palelolithic ancestors, but this time replacing the small band
by the entire species, recognizing the webs of causal interaction that link us
to people who live at great distances from ourselves (4950).
A swarm of objections come to mind:How plausible is this speculative
history? Could the model of the small egalitarian band numbering in the
scores of members be scaled up to billions? Would it be an improvement
over the status quo? Kitcher anticipates most of these critical questions
and attempts replies. But one objection he does not address concerns the
form of the argument: even if his speculative history of the early social
arrangements of our species were correct in its essentials (and Kitcher is
suitably cautious about his claims on its behalf), why would the histori-
cal precedence of egalitarian arrangements entail their moral precedence
over alternatives? Note that he draws no such inferences from the earliest
recorded forms of recognizably scientific inquiry:ancient Greek ideals of
theoretical contemplation most certainly would not alone satisfy his stan-
dards for well-ordered science. Whereas the history of science is dismissed
as too haphazard to serve as a normative guide, the prehistory of political
deliberation seems to fit thebill.
There are conceivable arguments that might be invoked in favor of the
latter position, but Kitcher notably does not appeal to themin my view
with good reason. Any claims that the earliest hominid arrangements
are somehow truer to the nature of the species would place more weight
on his possible story of how the ethical project got started than it could
bear. (Kitcher is careful to assert only that such a naturalistic account is
possible and at least plausible, acknowledging that there is no available
evidence for asserting anything more.) Moreover the demonstrable exis-
tence of many other arrangements, attested by the annals of history and
anthropology, offer prima facie evidence that the species can prosper under

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [217]
nonegalitarian regimes. Indeed the weight of the statistical evidence would
seem to buttress the conclusion that hierarchy is more natural than equal-
ity to Homo sapiens. Kitcher (1985) has elsewhere written brilliantly of the
snares and pitfalls of such partial arguments concerning human nature and
does not renege on those principles hereagain, in my opinion, rightlyso.
To conclude: Kitcher invokes two kinds of argument from history in
order to fortify his case that present forms of both democracy and science
are neither inevitable nor incorrigible. One kind of argument uses history
to broaden the horizon of the possible:no necessary trajectory of events
has led us to our current arrangements; alternatives are both conceivable
and instantiated by history; history provides at least analogies to the sorts
of cognitive and affective transformations Kitcher holds to be preconditions
for realizing improved versions of democracy and science. This is history
in the service of the imagination, and its message is that there have been
(and could be) more things in heaven and earth than have been heard of in
our current philosophy. As Ihave argued, the details of this history matter
in ways consequential for Kitchers larger aims, but his general strategy of
using history to enlarge the realm of the thinkable and to fortify hope that
reform is possible seems to me to succeed admirably. Iam more skeptical
about the second kind of argument, which uses history to ground norma-
tive claimsespecially because it is deployed asymmetrically in the case of
science and politics. History can persuade us only that certain forms of sci-
ence and politics are possible (and this is well worth knowing, given the lazy
tendency of our presentist culture to naturalize its own arrangements). But
history alone cannot suffice to show that one alternative that has existed in
the past is better than another, nor to make us desire still other alternatives
enough to make significant sacrifices in order to realize them. Only a com-
pelling vision of a better way of life can mobilize those energies, and that is
why Kitchers notion of human flourishing lies at the heart of his proposals.


What would it mean for both democracy and science to promote human
flourishing? Kitcher (2011b, 55) does not pretend to have an answer to
the age-old philosophical question about what constitutes the good life
(though he offers some observations, e.g., making a positive difference in
the lives of others), and it would be contrary to the spirit of his refusal to
admit ethical expertise to privilege the answers of philosophers over those
of other people. However, he does have views about the conditions most
likely to further the goal of providing, for the entire population, equal and

[218] The Philosophy of Philip Kitcher

serious chances for a worthwhile life (54). It is these chances, not their
specific realizations in the lives of individuals, that constitute his vision
of human flourishing through the pursuit of the ethical project. Mutual
engagement and well-ordered science are means to ends, not the ends in
themselves. Nor are they guarantees that the ends will be realized: all that
such measures can achieve is to equalize and increase everyones chance
of living a worthwhile life, not the certainty of doing so. We still dwell, as
Locke said, in the twilight of probabilities.
But given how mighty the efforts required to attain even a first approxi-
mation of the ideals of mutual engagement and well-ordered science, and
how long that might take, it is legitimate to ask whether and how the means
themselves might promote human flourishing. Just as Aristotelian ethics
encompasses becoming virtuous as well as the goal of being virtuous (we
become brave by doing brave acts, just by doing just acts), so one might
ask how participation in the ideal conversation and well-ordered science
might have a value in se. This value could never be entirely divorced from its
ultimate telos, no more than becoming virtuous can be severed from being
virtuous (or seeking the truth from finding it). Yet dedication to an ideal
that demands costly sacrifices (of egoistic desires, of the freedom to follow
ones own curiosity, in the case of Kitchers proposals for democracy and
science) would seem to promise some reward in and of itself if it is not to
become a teeth-clenched exercise in self-abnegation.
Kitcher himself seems to recognize at least implicitly that process mat-
ters as well as product. Although the value that clearly animates his vision
of the ideal conversation and well-ordered science is justice on a global
scale, he does not seem to consider the duty to see justice done to be suf-
ficient. It is not enough that the participants in the ideal conversation per-
ceive that justice demands that an equal hearing be given to the desires of
others; they must adjust their emotions so as to desire to give equal weight
to the desires of others and modify their own desires accordingly. This is to
be accomplished by the virtuoso exercise of sympathy of extended mirror-
ing, described earlier. At first glance this seems to be an attempt to merge
Kantian ends with Smithian means, because Kitchers focus on projects
commits him to a concern with process. It is not enough to will justice; we
must desire to will justice. Empathy (the cognitive ability to think the world
from anothers perspective) is necessary but not sufficient; sympathy (the
affective ability to feel the world from anothers perspective) must also be
engaged. Empathy and sympathy are distinct:a shrewd strategist may be a
master of empathetic divination of the enemys desires and aims without
feeling a dram of sympathy for them. Kitchers ideal conversation demands
both empathy and sympathyand both developed to a heroiclevel.

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [219]
Would such a process of mutual adjustment of desire promote human
flourishing in se? The nobility of the end ennobles the means, but is that
enough? The question is made more pressing because of the uncertainty
of the outcome at every stage:participants in the ideal conversation can
only try to find the best balance among the varying assessments (indefi-
nitely iterated) made by fellow participants (Kitcher 2011b, 52); even if
they succeed, they can equalize only chances (not certainties) of leading a
worthwhile life; the well-ordered science guided by the ideal conversation
can never promise to deliver the ideal results. It is a grand game, played in
earnest for the highest stakes, but a game all the same. Is the game worth
the candle?
I think that there are at least two reasons for doubt in the case of the
ideal conversation. The first is intrinsic to the process itself: even if the
extraordinary levels of cognitive and affective perspectival suppleness
Kitcher calls for are humanly possible, cultivating them may weaken or
even extinguish traits rightly deemed essential to human flourishing. How
would spontaneity fare in the deliberative vertigo of the hall of extended
mirroring? Would we not be permanently sicklied over with the pale cast
of thought?4 Deliberation famously cools the passions, but might not the
indefinitely iterated deliberation of extended mirroring freeze all desire?
The particular kind of deliberation required, infused by sympathy, poses
still more troubling threats to values of character and integrity. Firmness
of purpose and resolve, constancy of traits, peculiarities of personality and
taste might withstand deference to duty, however arduous the task of self-
mastery required. But could they withstand the modification of the very
structure of desire through sympathy with the desires of others, all others?
The hypersympathetic psyche might resemble the chameleon of Woody
Allens film Zelig (1983), who involuntarily took on the tincture of what-
ever personality he encountered. Such a pliant self might raise worries as
to whether it was a self atall.
The second reason for doubt concerns the objects rather than the act of
sympathy:the desires of others. In primitive mirroring these are egois-
tic desires; in extended mirroring, desires filtered to eliminate impos-
sible and incompatible options. But they remain individuals desiresnot
shared ideals of the collective good, not common visions of a better way
of life. Of course nothing in Kitchers scheme prevents individuals from

4.Elster (1984, 40) makes a similar point in the context of cost-benefit analy-
sis:Even if the deliberations do succeed in modifying the behavior of the character in
the desired way, the very act of deliberating can modify the character for the worse, and
in ways judged even more important, through the stultifying effects on spontaneity.

[220] The Philosophy of Philip Kitcher

desiring to serve the commonweal, but nothing promotes it either. The unit
of analysis is the individual and the individuals conception of a worthwhile
life. Participants in the ideal conversation are placed in the curious predica-
ment of exercising selfless means (heroic sympathy) to attain selfish ends
(each persons version of a worthwhile life, modulo everyone elses desire).
Insofar as history provides examples of sustained and selfless striving, they
are mostly in the service of objectives that transcend the individual or even
a whole community or generation. To labor to build cathedrals (or catalog
hundreds of thousands of stars) over centuries is to labor in the service of a
human community that transcends an individual lifetime; sacrifices of life,
riches, and opportunity made in the name of a nation, a creed, or an ideal
can be found in almost every peoples history.
Such transcendent, unifying visions of the collective good might emerge
in the aggregate of individuals who practice extended mirroring, and Iread
Kitcher as positing exactly that: a democracy enriched by the terms of
mutual engagement ultimately aims at justice for all. But that is not the
banner under which individual citizens will march in most cases. Instead
extended mirroring will engage them in a gigantic accounting operation
to arrive at an equilibrium among individual desires mutually adapted by
sympathy with the desires of othersnot a collective commitment to a
shared ideal. Kitchers aggregate is neither a Hobbesian state of nature nor
a neoclassical free market:individuals who participate in the ideal conver-
sation are neither at war nor in competition with one another. Nonetheless
the harmonization of desires through extended mirroring has more than a
whiff of methodological individualism about it,5 and it is unclear whether
this process would promote one of the most intense forms of human
flourishingthe exaltation of serving a cause greater than oneself, shoul-
der to shoulder with dedicated fellowsanywhere near as well as a holistic
ideal enlivened more by solidarity than sympathy.


One of the most original and compelling features of Kitchers vision of

science, democracy, and science in a democracy is the unfinished, quest-
ing, open-ended nature of the endeavor. Neither democracy nor science

5. Kitchers metaphors are optical (mirroring) rather than musical (harmoniza-

tion), yet this reader at least felt that the experience of singing in a choir or playing
in an orchestra better captured the constant and exquisite (and exhilarating) mutual
adjustments of the ideal conversation.

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [221]
is progressing toward some ultimate goal, be it perfect justice or absolute
truth; both are instead progressing from some anterior, unsatisfactory
state of affairs that must be remedied. How the remedy is to be sought
is as important as what the remedy is; the process of achieving the ideal
conversation and well-ordered science teaches us more about both than
any solution handed down ever could. Dissatisfaction is never-ending, and
therefore so is the search for remedies. This is a vision suffused with the
lessons of history, both the history of science and the history of democracy.
Both are works in progress and destined to remainso.
This is a vision more in tune with the self-image of science than with that
of democracy (at least American democracy). Because the latter is bound to
interpretations of its Constitution, consequential decisions about the let-
ter and spirit of democracy look to the past rather than the future or even
the present. Not all interpretations of the Constitution have been funda-
mentalist (e.g., attempts to divine and abide by the original intentions of
the eighteenth-century framers), but all necessarily must build analogi-
cal bridges between past principles and present case (a structure of argu-
ment reinforced by common law appeals to precedent). Even jurists who
disavow literalist interpretations of the Constitution bear the burden of
demonstrating the continuity of the present decision with past precepts
although the Supreme Court has proved itself capable of rare moments of
remarkable and consequential creativity.6
In contrast, modern science is largely oriented toward the future.7 This
future fixation has sometimes taken on a melancholy cast, when scien-
tists confront the likelihood that they and their work will be forgotten in
a generation and that much of what they believed to be true will inevi-
tably be revised by their successors (Daston 2001). But the open-ended
dynamism of science has also exerted a powerful magnetism, and not just
upon scientists. Since the early nineteenth century science has been held
up as a model of knowledge that liberates (Secord 2014) and of a polity
that reconciles progress and stability (Jewett 2012). Kitchers account of
well-ordered sciences contribution to human flourishing rightly concen-
trates on products: therapies for cruel diseases, hardy crops for hungry
people, accurate accounts of everything from the causes of species deple-
tion to the evolution of pathogens. However, there is some evidence that

6.For example, in the 2013 Supreme Court decision United States v. Windsor
7. However, some sciences, especially those that study phenomena that unfold on
a superhuman time scale, also look to the past to supply archives of essential data

[222] The Philosophy of Philip Kitcher

a growing number of citizens are also enthralled by the process of science.
The growing number of popularizations by scientists applauded by Kitcher
(2011b, 12829) have found a growing number of readers and viewers. Still
more encouragingand surprising, given the depressingly low number
of students wishing to pursue careers in science and engineering in afflu-
ent democracies (Sjberg and Schreiner 2010)is the boom in Internet-
based citizen science, which enlists volunteers to document new species
of insects, fold proteins, recover historical weather data, classify galaxies,
observe biotopes, and track solar storms (Citizen Science Alliance 2014).
For the scientists the volunteers are badly needed reinforcements to help
deal with the flood of data, but it is less clear what motivates the volun-
teers, some of whom are spending hundreds of hours on their projects in
return for neither fame nor money. Some of the projects aim at eventual
practical benefits, but as in the case of most scientific research, there are
very few short-term payoffs. To all appearances it is the process of research
itself that attracts volunteers:in a survey of volunteer galaxy classifiers, the
strongest motivation named by both men and women was to contribute to
original scientific research (Raddick etal. 2013, 25). Being part of the proj-
ect is a part of human flourishingfor science as well as for democracy.

Reply toDaston

During the academic year 20078, which I spent as a Visiting Fellow in

Lorraine Dastons Abteilung of the Max Planck Institute in Berlin, Ilearned
something very important. Theres an alternative to the Anglophone philo-
sophical ide fixe that serious discussion requires contestants to shred
any proposal displaying the slightest sign of vulnerabilityall with the aim
of discovering the smartest guy in the room. Raines weekly colloquium
diverges fundamentally from these familiar displays by always fostering a
cooperative effort to help the speaker elaborate the best version of his or
her ideas. Those constructive hours in Boltzmannstrasse were not needed
to demonstrate that Raine is one of the premier historians of science in
the world today (that Ialready knew). But they did teach me that she is an
incomparable mentor.
Dastons essay views many themes in my recent work through the lens
of history, and her historical expertise allows me to correct some errors and
to articulate other ideas more adequately than Ihave previously done. As
she rightly notes, Ithink with history. Indeed much of my philosophical

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [223]
work is motivated by viewing history as providing clues to the character
of areas of human practice. Our understanding of mathematics, science,
ethics, and democracy is limited if we fail to probe the complex histories
out of which our current commitments in these domains have emerged.
For my philosophical purposes details of history sometimes matterand
sometimes they do not. Dastons thoroughly sympathetic exploration of
my views leads her to see where overhasty historical reading has interfered
with my philosophical purposes. She doesntcarp.
As she recognizes, my account of the autonomy of early modern sci-
ence (drawn, Ifear, from attending too much to the Thomases, Sprat and
Shadwell) is defective, and the deficiency affects my argument. The lumi-
naries she cites (Galileo, Boyle, and company) certainly engaged in proj-
ects they took to be socially useful. Isuspect that at least some of them
were not so committed to being free and unconfind that they would have
resisted directives from the broader society to engage with particular areas
of inquiry. So my history of a long-held dedication to autonomy needs revi-
sion. One consequence, drawn by Daston herself, is an added emphasis
on justifying claims to scientific autonomy in an age of radically increased
public funding. But, as she sees, the argument I offer against those who
defend autonomy by gesturing at famous instances of government inter-
ference with sciencethe Lysenko affair being the favorite exampleis
undermined once the history is corrected. My proposal to replace the
appeal to anecdote with a statistical survey of the effects of outside direc-
tion of research topics will not do if the current level of interference is his-
torically unprecedented. Fortunately the negative point about appeal to
anecdotes survives. Moreover even if a more accurate and richly variegated
history were to raise questions about the potentially harmful effects of
diminishing scientific autonomy, the appropriate stance (envisaged in my
reply to Nancy Cartwright and Alexandre Marcellesi) would be to engage
in cautious experimentation rather than insisting that a handful of loosely
analyzed examples demonstrate that scientists knowbest.
A deeper challenge focuses on the history Ioffer in an attempt to liber-
ate us from the constraints usually taken to limit the options in thinking
about ethics. As Daston rightly notes, my history involves reconstructing
events and processes that occurred long before the invention of writing
and thus is inevitably speculative. But claims about prehistoric human
life, even prehistoric human social life, dont all involve the same degree of
conjecture. The bones and artifacts left by our precursors enable archaeolo-
gists to defend relatively well-grounded hypotheses about the size of ances-
tral bands at various moments in the pastand to infer the less firmly
established but still not completely speculative conclusion that, until about

[224] The Philosophy of Philip Kitcher

fifteen thousand years ago, hominid social life involved the same sorts of
small groups, mixed by age and sex, in which our closest evolutionary rel-
atives, the chimpanzees and bonobos, assort themselves. Studies of the
materials used in tools and the distances between the sites of use and the
nearest available sources provide evidence for the existence of trading net-
works in the late Paleolithic. An attempt to understand the status of ethics
by reconstructing the history of ethical practice contains a mixture of how
actually and how possibly explanations, and I have tried to mark the
distinction carefully.
Dastons central concern is not that my history is too speculative to
be useful but that its significance for normative conclusions must be care-
fully explained. As she puts it, Why would the historical precedence of
egalitarian arrangements entail their moral precedence over alternatives?
With small reformulations this seems to me to be the crucial question.
Entailment isnt the issue; the issue is:How does understanding the his-
tory of our ethical practices help us to see how to continue the work of
revising those practices? Effectively, how do we learn from history?
In previous discussions with one another, Daston and I have both
lamented the absence of an analog of philosophy of science that focuses
on the arts and the humanities (and Iowe to her the point that this was
not always so, and that thinkers like Dilthey aimed at identifying the
methodologyor the methodologiesof the Geisteswissenschaften as well
as the Naturwissenschaften). Daston and I agree that we can learn from
history (and also from ethnography and from literature). One mode of
learning consists in the enlargement of the realm of the thinkable (her
excellent phrase). We liberate ourselves from the narrow space of options
that has hitherto confined our reflections.
I want to go further than this. In light of an improved historical
understanding, people can come to see what they are doing from a new
perspectiveor with a different Gestalt. Examples of this occur all the
time. When Iwas very young, the family living next door owned a black
cat, to whom they had given the name Nigger. For me, as for the little
girl with whom I played in the 1950s south of Englandand indeed for
our parents toothat name was entirely unproblematic. Only as Istarted
to learn about the history of slavery and the role the word played within it
did Istart to regret my own past linguistic usages. When Irecall our unen-
lightened speech, Icringe. Istrongly suspect that the woman who is the
present-day descendant of that little girl (like me, she moved to the United
States) cringestoo.
History can reorient us, leading us to modify our commitments and
our ways of behaving. That isnt a matter of entailment but rather of our

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [225]
coming to feel or perceive or judge (its not obvious what the right verb is)
things we have previously taken for granted to be no longer acceptable. The
historical narrative has exposed what we were doing, and our new aware-
ness is embodied in a feeling of profound discontent (shame or guilt) and
perhaps that a particular form of change is the way to set thingsright.
How does this apply to the narrative offered in The Ethical Project? The
mix of how possibly and how actually explanations generatesat
least in mea sense of the norms and structures of shared human lives
as emerging from a long series of attempts to wrestle with a deep prob-
lem in the human predicament. Isee members of our species as drawn to
live together with one another, without biologically based psychological
adaptations for making that shared life go smoothly. Our responsiveness
to others is limited, and ethical life grows from a technology for amplifying
that responsiveness. The pioneers who began the ethical project found the
obvious solution to that deep problem. They deliberated on terms of rough
equality, listening in the cool hour to the claims issued by all members of
the band. Out of their efforts has emerged a far more complex way of life,
one in which human beings experience thoughts and emotions that were
inaccessible to their ancestors. Along the way the old egalitarianism has
been severely compromised. Many voices, many claims are no longer heard.
Yet the old problem of limited responsiveness endures, manifesting itself
in new forms and causing continued suffering.
Perhaps Ishould have written differently, posing questions rather than
trying to defend theses:Here is a narrative of how we arrived at our cur-
rent ethical practices. Do you see how the same old problem reemerges in
new guises? Doesnt that problem deserve to be seen as a deep feature of
our predicament, because it endures and because its effects are so large and
damaging? Isnt the ancestral solution, of mutually engaged conversation,
the obvious way to address that problem? What else could we do today
besides trying to scale up the old solution, difficult though that will surely
be? When you reflect on the history, what possible ways of going on do
The questions Ive posed could be viewed as opening a conversation
with Daston, Smith, Srinivasan, and others. Out of the conversation might
come something entirely new, a consensus on a different set of emotional
reactions and revisions of commitment. We would have learned from his-
tory, but not quite in the way I originally thought. History would have
sparked some individual reactions, and interactions with others would
have refinedthem.
But what are the principles that guide inferences from premises about
the history of ethical practice to conclusions about how we should modify

[226] The Philosophy of Philip Kitcher

our norms and commitments? In response to that natural question, Ican
say little more than I have already said. Perhaps some future extensions
of logic (broadly construed) will systematize these types of inference in
ways analogous to those in which Frege and his successors have taught us
to formalize deductionor perhaps the thesis that reasonable change of
belief must always submit to an exact analysis is a shibboleth. For the pres-
ent, and possibly for all time, we shall have to understand some forms of
justified inference in terms of judgment, what Duhem thought of as le bon
sens. To make that concession is, of course, to invite the objection that my
terminologylearning from historyis self-deception. Yet in a range of
examples, extending from my embarrassed reflections on the naming prac-
tices in which my childhood self acquiesced to the more ambitious conclu-
sions about continuations of the ethical project, Iam more convinced of
the phenomenon of reasonable inference than Iam of any specific way of
characterizing it formally.
A post-Kuhnian understanding of the resolution of large episodes of sci-
entific change reinforces my confidence. The kinds of judgments Ihope to
license are also exemplified by episodes in which the character of repeated
efforts to tackle recalcitrant problems justifies modifying the goals toward
which scientific effort is directed. Ascientific community reflects on a slice
of recent historythe attempts to use the concept of phlogiston to sys-
tematize a growing corpus of chemical reactions, the efforts to understand
features of life and its history in terms of episodes of special creationand
reconciles itself to giving up certain kinds of questions (about the qual-
ity shared by combustible substances, about the purposes manifested in
the living world). Old goals are abandoned, new ones adopted. Judgments
of what should be achieved are modified by attending to what history has
shown (Kitcher 1993a, chs. 6 and 7; Laudan 1984). If the inferences Ifavor
in understanding the reasonable evolution of the ethical project prove
problematic, there will be fallout for domains traditionally (and, Ithink,
correctly) viewed as the home turf of reason and evidence.
I shall close with some remarks about human flourishing, the topic of
Dastons penultimate section. As my replies to Cartwright and Marcellesi
and to John Dupr emphasize, the ideal of well-ordered science is to be a
diagnostic tool, one that orients us to ends-in-view, not to an unattainable
utopia. It should inspire us to refashion ethical life through greater inclu-
siveness and through mutual engagement with those whom we ignore.
To be so inspired does not entail that all our decision making should be
explicitly guided by trying to simulate within our imaginative conscious-
ness some ideal discussionany more than the Kantian is committed to
supposing that every action should be preceded by an attempt to figure

K i t c h e r o n S c i e n c e , De m o c r a c y, a n d H u m a n F l o u r i s h i n g [227]
out whether our maxim could be universalized or the utilitarian doomed
to impossibly complex calculations. We often proceed, justifiably, by taking
our habits (Deweys good word) for granted, interrogating them according
to our ethical standard only under conditions of live doubt (Peirces impor-
tant counter to Cartesianism). Neither Hamlet nor Zelig is our destiny.
I entirely agree with Daston on the importance of individual projects to
human flourishing. Following Kant, Humboldt, and Mill, Itake it to be cru-
cial that our life themes should be freely chosen, and an important mode of
ethical progress would consist in spreading more broadly and more securely
the conditions under which people can freely choose the shape they intend
their lives to have. Yet I add another condition, one that may renew an
Aristotelian theme. Valuable life themes must involve connections among
lives. Thus the individual must enjoy the possibility of living in commu-
nity with others, of participating in ventures that affect a larger group than
the single self. So Ishare Dastons sense that flourishing lives contribute
to something bigger than themselves (Kitcher 2013, 2014). Whether my
emphasis on this form of community extricates me from a narrow method-
ological individualism is a matter Ileave to my critics.
Among these larger ventures might be an attempt to achieve pure under-
standing of nature, without thought of solving practical human problems.
Moreover parts of that attempt might be endorsed by the participants in
an ideal conversation. Suppose, however, that were not so. At a particular
historical moment the scientific community forges ahead with its search
for deeper understanding of the universes first few microseconds, at a cost
to the health and the lives of a large group of people. Imagine that a com-
prehensive, informed, engaged conversation would resolve that the neglect
of those peoples welfare was an ethical mistake. In that context the scien-
tific communitys declaration that they were pursuing an important mode
of human flourishing would ring hollow. They would be like imaginary
British mathematicians who rejected the call to Bletchley in favor of con-
tinuing their work on Fermats Last Theorem. Flourishing is important, but
there are occasions on which your preferred mode of flourishing must be
given up for the larger good. With this judgment Ihope Raine wouldagree.

[228] The Philosophy of Philip Kitcher


Deliberating Policy
Where Morals and MethodsMix

P hilip Kitcher has long been a model and an inspiration. He has not
taken the easy way:philosophy for philosophys sake, speaking to our
own history, pursuing topics and ideas that wear their philosophical worth
on their sleeve, ideas that advance the problems they are grappling with
in an easily discernible fashion. Instead Kitcher has pioneered what Otto
Neurath called philosophy for life. His philosophy serves the problems we
all, as members of society, grapple with, problems that a facile application
of what our philosophical history provides will not help, where new phi-
losophy must be built, molded to the problems of life it aims to serve, and
to do so it must be serious, deep, and detailed. This is a kind of philosophy
that few of us can emulate, but we can venerate it. That is our purpose in
contributing to this volume in honor of Kitcher.
Throughout his career Kitcher (2001b, 65) has taught us that moral
and social values [are] intrinsic to the practice of the sciences. We agree
wholeheartedly. And we think this lesson is of special importance when the
sciences in question are directly relevant to policy issues such as child wel-
fare. And like Kitcher, we also think that values should not enter scientific
practice willy-nilly. As we will illustrate, when morals and methods mix in a
way they should not, undesirable consequencesregarding the welfare of
victims of child abuse, for instancecan follow.
In 2004 in the London borough of Haringey, seventeen-month-old Peter
Connelly was found dead in his crib. The child had suffered fractured ribs
and a broken back after months of abuse at home. His mother, her partner,
and a lodger were jailed for his death. Peter had been seen by health and
social services professionals from Haringey Council sixty times in the eight
months before hedied.
There were two kinds of government responses to this that we shall dis-
cuss. First, Minister of Education Ed Balls sacked the director of Childrens
Services in Haringey, Sharon Shoesmith, with immediate effect in a live
press conference on television. Shoesmith (2011) defended herself and the
Haringey Services in an interview with BBC Radio 4:We should not be put
into blame; it does not result in anything productive and obscures the
bigger picture. The journalist interviewing her argued to the contrary. If
nobody accepts the blame, he asked, then how can we stop this happening
A second response came from Prime Minister Tony Blair (2006) in a lec-
ture delivered on September 5, 2006. He argued that the government can
make children and young people safer by identifying at-risk families and

Let me summarise my argument. I am not talking about . . . trying to make

the state raise children, or interfering with normal family life. Iam saying that
where it is clear, as it very often is, at [a]young age, that children are at risk of
being brought up in a dysfunctional home where there are multiple problems,
say of drug abuse or offending, then instead of waiting until the child goes off
the rails, we should act early enough, with the right help, support and disci-
plined framework for the family, to prevent it.... It may be the only way to save
them and the wider community from the consequences of inaction.

According to Blair, We can predict. We can then, in the jargon, intervene.

Both these responses are morally questionable. We do not mean by this
that they are wrong policies; we mean only that they are open to moral
debate. Look first at Blairs. Blairs program is intended to identify at-risk
families and offer help. There is evidence that applying a stigmatizing label
such as being an at-risk family or child will have an impact on those
labeled and on how they are treated by the many professionals who would
know of that label. And experience from studies of labeling theory would
suggest that some of this impact will be negative (see, e.g., Goffman 1959).
As a memorandum of the Association for Improvements in the Maternity
Services (AIMS 2009, 6) reports:

Many of our families reported feeling bruised, and some were seriously trauma-
tised, by their encountershowever briefwith social workers, emerging with

[230] The Philosophy of Philip Kitcher

lowered self esteem, decreased confidence as parents, and diminished trust in
all professionals. Our own direct observation of some encounters showed that
even with a supporter and observer present, the behaviour described by parents
was replicated.

So the program advocated by Blair may cause harm overall.

The question of parents rights and family autonomy also looms. As
Elizabeth Brake and Joseph Millum (2013, 5.1) put it, Parents have
moral and legal rights regarding their children. They have the liberty to
make decisions on behalf of their children regarding matters such as diet,
schooling, association with others and the right to exclude others from
such decision-making. Seana Shiffrin (2000, 216) argues that a govern-
ment acts paternalistically when it aim[s]to take over or control what is
properly within the agents own legitimate domain of judgment or action.
In giving directions to parents the government replaces their judgment of
how to raise their child with its own judgment. So all this considered, even
if the interventions will produce the predicted benefits, there remains a
question about whether such interventions are justified.
As to blame: blame is retributive; it is often vindictive; it attacks the
moral character of the culprit, not the deed, it vilifies the culprit, and as
Garrath Williams (2003, 433) explains, There is clear evidence from
social psychology that blame is frequently and inappropriately attributed
to individuals in modern Western societies. There is also a tendency to
overestimate the extent to which peoples behaviour is due to internal,
dispositional factors, and to underestimate the role of situational factors
(Aronson etal. 1999, 128). Blaming a person is more than grading him or
her negatively. As R.Jay Wallace (1994) argues, blaming makes the person
the object of negative emotions such as resentment and indignation and
subject to adverse treatment:avoidance, reproach, scolding, denunciation,
remonstration, and punishment.
Here is a case from the United States. Psychologist Robert Hare
developed a test to identify characteristics of psychopaths, such as lack
of empathy, lack of remorse, and lack of guilt. One study, in which the
test was administered to prisoners, found that those who did not have
the characteristics identified by Hare were reconvicted within five years
about 20 to 25percent of the time. Those who did have the characteristics
were reconvicted within five years about 80percent of the time (see, e.g.,
Hare 1991). Once this study was released, parole boards across the United
States and Canada began using the test when considering the release of
prisoners. In several US states parole boards are even mandated by law to
do so. Because the psychopathy test predicts a high rate of recidivism, it

De l ib e r at i n g P ol i c y [231]
is unlikely prisoners with Hares characteristics will be paroled, consider-
ing the risk to the publicand the political risk:if a person with known
psychopathic characteristics reoffends, the parole board is in trouble.
So there is little incentive for parole boards to release those with Hares
Yet many moral and religious points of view would hold that people
should not be kept in prison because of predictions about their future
behavior. Prison is for people who have committed a crime (in the past
tense). Those up for parole have already served the minimum sentence for
their crime. Yet for those prisoners with Hares characteristics, other con-
siderations favoring their release on parole are overridden by the results
associating those characteristics with a high recidivism rate. Furthermore,
while the test predicted reconviction rates fairly accurately under the con-
ditions his lab was using, Hare (1998) is concerned that the test is not
as accurate when it is not administered under the same conditions. For
instance, court records suggest that the results of the test are biased by
the person paying to have the test administered: tests administered on
behalf of the prosecution see scores, on average, 8 points higher than those
administered on behalf of the defense, where the higher the score on the
test, the higher the predicted reconviction rate. And there are no guaran-
tees that the test is being administered by a trained psychologist.
Here is another example of a morally questionable policy reported on
the front page of the NewYork Times on March 3, 2013. In an effort to cut
down on robbery, the New York City Police are aggressively intervening
in the lives of certain young people. As the police describe it, they aim to
make them [the young people] radioactive and thus isolate them from
their friends. Which young people? Those destined for trouble and most
likely to commit these crimes (Ruderman2013).
All of these policy responses are morally questionable, and all promise
success based on what we can claim toknow.
A good policy decision always requires a mix of considerations: Who
benefits? Who suffers? Who pays? How much? What are possible good side
effects? Bad ones? Will the effects last? And so on. Central among these are
issues of effectiveness and issues of legitimacy:Will the policy achieve the
desired ends? Is it morally, politically, culturally acceptable? Few policies
will be all to the good for all concerned; few have only a moral upside and
no moral downside. So it is inevitable that a balance be struck. In particular
it can be perfectly acceptable to adopt a policy that is morally questionable
or that has morally negative aspects if we can be sure it will achieve good
ends, so long as the balance is reasonable and we operate within a range of
what is at least morally permissible.

[232] The Philosophy of Philip Kitcher

Our concern is with cases where we get the balance wrong because we are
overconfident in our predictions about what the policy results will be. This
in turn generates concerns about the current drive for evidence-based policy
(EBP), which is advocated, indeed often mandated, across policy domains,
from medicine to education, crime, and economic development. Of course
in general we will get more reliable predictions of policy outcomes if we take
into account the evidence than if we ignore it. But there is the promiseor
perhaps just the hopeof far more certainty than the evidence can deliver.
The danger of these mistakes is that they encourage an unjustified degree of
optimism about how effective our policies can be, in which case we are likely
to get policy deliberation wrong in the delicate balance between consider-
ations of effectiveness and considerations of legitimacy.
Here, then, is the basic problem we want to underscore:morals and meth-
ods ought to mix when it comes to policy deliberation. Thats because
leaving costs asidewhether a particular policy should be implemented

1. Whether the policy will be effective, that is, will produce the expected
2. Whether the policy is morally, socially, politically, and culturally

But these requirements dont always mix in the way they should, because
we often focus on (1)and discount the importance of (2). Aplausible expla-
nation for this is that we think we have methods, for instance random-
ized controlled trials (RCTs), that will provide us with objective and fairly
certain answers to the effectiveness question. By contrast, we do not have
methods that can give us objective and certain answers to the moral ques-
tion; here things are much muddier and open to debate.
This slidefrom

A. The methods we have for ascertaining (1)are better than the ones we have
for ascertaining(2)


B. We should give (1)more weight than (2)in policy deliberations

is eased along by the prestige that EBP enjoys. The point of EBP is to ensure
that effectiveness predictions are based on sound evidence. But the expec-
tations EBP creates can lead us to give this evidence and the predictions it

De l ib e r at i n g P ol i c y [233]
supports greater weight than concerns of moral acceptability. Effectiveness
can become the cardinal value in policy deliberations, as we think we are
now seeing in UK political calls for coercive interventions into families that
seriously change the state-family relationship.
In the rest of this essay we want to hack away some mistaken philo-
sophical stances that can make effectiveness considerations loom larger
than they should. These involve a circle of mistaken ideas about objectivity,
certainty, and causality:

1. We bank on certainty.
2. We suppose objectivity is the path to certainty because we assume that
objectivity amounts to the elimination of the subject and of judgment,
via the use of methods that have manuals that fix correct procedures.
3. We assume that causality is linear and that it is God-given.

EBP champions objectivity and certainty in social policy deliberation. It

insists that, for policy evaluation and prediction, we rely only on objec-
tive methods like RCTs that promise to provide certainty. An ideal RCT
(one where the net effect of other causal factors is indeed balanced between
treatment and control) can clinch causal claimsand they are advertised
to do so without the intrusion of subjective judgment. From this posi-
tion we slide easily into our third problematic assumption:that causality is
linear and God-given.
Look at linearity first. The slide is easy hereand easy not to notice.
Thats because, looking through the lens of RCTs, complex causal webs
get projected onto a line. Figure9.1 depicts the causal process as RCTs tell
it. There are two different senses of linear involved in this image, and
we tend to suppose both. First, what we label the causethe policy
whose results we aim to predict or the actions we want to blame for some

Figure9.1 Domino causation:the causal process as RCTs tell it (by CharlieGee).

[234] The Philosophy of Philip Kitcher


Work Student
feedback motivation

Consistent Student
lessons ability

Study Supportive
space family

Figure9.2 Acausal cake for the contribution of homework to academic achievement.

disastrous outcomeis seldom enough on its own to produce the effect in

question. It needs help. Generally a whole team of support factors must act
together with the highlighted cause, or no contribution to the effect will
be produced. Epidemiologists illustrate this with what they call causal pie
diagrams. Figure9.2 provides an example based on Harris Coopers work
on the effectiveness of homework (see e.g. Cooper1989).
We refer to these not as causal pies but as causal pancakes (or just
cakes). To make pancakes you need flour, milk, eggs, and baking powder,
and you need them all. With just three of the four ingredients, you dont
get three-fourths of a pancakeyou dont get a pancake at all. Similarly for
the ingredients in the causal cake diagrams:all the ingredients are required
or you will not get the expected contribution to the effect.
The cake diagrams make vivid a crucial point. But they also make it look
too simple. For most policies the connection between cause and effect is not
immediate: there is a long chain of steps in between, and each one has to
occur at the appropriate time to lead on the next. Consider the diagram for
the Nurse-Family Partnership in Figure9.3; it is more complicated than our
simple domino image since the policy initiates not just a single causal chain
but three different policy actions that lead by interwoven chains of intermedi-
ate effects to the targeted outcomes:less child abuse and fewer young people
arrested. We assume that causation is linear in a second sense when we
ignore the multiple pathways linking our policies to the outcomes of interest.
Focus on the bottom line in Figure9.3, which looks like a straightforward lin-
ear sequence. But to describe it thus is to miss the point about support factors

De l ib e r at i n g P ol i c y [235]

Home visits weekly the first month following program Pregnant women display improved
enrollment, then every other week until birth of infant.
Nurse address:
Effects of smoking, alcohol and illicit drugs on fetal growth,
health behaviors.
cigarette smoking
pregnancy-induced hypertension
and assist women in identifying goals and plans for
IMPROVE reducing cigarette smoking, etc.; use of community resources
pregnancy outcomes Nutritional and exercise requirements during pregnancy
by helping women and monitor and promote adequate weight gain; Helping First-Time Parents Succeed
improve prenatal Other risk factors for pre-term delivery/low birth weight
health Newborns are 37 weeks gestation
(e.g., genitourinary tract infections, pre-eclampsia);
Preparation for labor and delivery/childbirth education;
& weigh 2500 grams or more.
pre-term delivery among smokers
Nurse-Family Partnership
Basics of newborn care and newborn states;
Family planning/birth control following delivery of infant; birth weight among young teens
(<17 years)
Theory of Change Logic Model
Adequate use of office-based prenatal care; and
Referrals to other health and human services as needed. neurodevelopmental impairment

Parents demonstrate sensitive and

Home visits weekly postpartum period, every 2 weeks until competent caregiving for infants and
toddler is 21 months, monthly until child is 2 years. Nurses: toddlers. Early Childhood (46 yrs):
Educate parent on infant/toddler nutrition, health, growth, child rearing beliefs associated with safety hazards in home
development and environmental safety; child maltreatment (Bavolek AAPI) stimulating home environment - Adolescence (15 yrs):
Role model PIPE activities to promote sensitive parent-
IMPROVE child interactions facilitative of developmental progress;
verified cases of child abuse & neglect HOME score state-verified reports
child health and Assess parent-child interaction, using NCAST sleeping and incidents of child injuries or ingestions incidents of injuries & ingestions of child abuse and
development teaching scales and provide guidance as needed; stimulating home environments, i.e., noted in medical records neglect from 015 years)
by helping parents Assess infant/toddlers developmental progress at increase in appropriate play preschool Languages Scale scores arrests and adjudication
provide sensitive and selected intervals using Ages and Stages Questionnaire or materials (HOME Inventory) executive Functioning Composite for incorrigible behavior
competent caregiving DDSII, and provide guidance as needed; scores (e.g. truancy, destroying
Promote adequate use of well-child care; problems in clinical range on property)
Guidance to new parents in building and fostering social Child displays age and gender
appropriate development. Achenbach CBCL
support networks;
Guidance assessing safety of potential/actual child care language & cognitive/mental delays
arrangements; and more responsive in interactions with
Referrals to other health and human services as needed. mothers (NCAST)/less distress to fear

Later parental life course

Home visits weekly during postpartum period, every 2 weeks (13 yrs following program
until toddler is 21 months, monthly untill child is 2 years. Parents have developed plans for completion):
IMPROVE Nurses: economic self-sufficiency.
parental life-course Facilitate decision-making regarding planning of future additional pregnancies
children and selection of birth control to achieve goals; subsequent pregnancies and live births
by helping parents interval between 1st and 2nd child Early parents life course (34 yrs
develop a vision for their Assist parents to set realistic goals for education and work, spacing between 1st
and identify strategies for attaining goals; number of months women employed following program completion): and 2nd child
future, plan subsequent
pregnancies, continue Coaching parents in building and fostering relationships during childs 2nd year additional pregnancies and live births months on AFDC and
their education and with other community serivces; months on welfare months on AFDC and Food Stamps Food Stamps
find work Parents family planning, education and work goals; and father involvement in child care rates of living with father of child arrests and convictions
Referrals to other health and human services as needed. and support rates of marriage days in jail

Figure9.3 Acausal diagram for the Nurse-Family Partnership. Each step in the diagram corresponds to a causal cake depicting the requisite support factors.
and causal cakes:what we picture as the cause typically cannot produce the
effect on its own but needs the help of a whole team of support factors. Thats
going to be true for each step in this causal sequence. Theres not just one
causal cake here but a different causal cake for eachstep.
If we want to identify the support team necessary for the initial causes
in the Nurse-Family Partnership program to produce the targeted out-
comes, we have to gather all the members of all the support teams from
each stage and graph them together in one huge causal cake. Recall that the
point about causal cakes is that all their ingredients have to be in place or
you dont get the effect. To the extent that the presence of any of the neces-
sary ingredients is uncertain, so too is that of the final outcome. But look
at our circle of problems. We bargain for certainty. The simple linear causal
model makes this seem to be a far better bargain than it generally is.So:

We often expect results that cant be achieved, which leads to wasted

money and effort and to heartbreak and dashedhopes.
We dont put in place the support factors that can help make our policies
work because we havent noticed the need forthem.
We blame perfectly good policies for failing that could achieve better
results in better circumstances.
We despair of doing anything because we cannot find the miracle

The linear model and the omission of support factors also predispose us to
focus efforts on eliminating harmful causes at the head of a sequence, such as
family drug and alcohol abuse, which can be a tall order. But it can be just as
effective to remove support factors anywhere along the causal path. Consider
the growing body of research on resilience factors. Resilience describes the
product of a combination of mechanisms for coping in the face of adversity.
Evidence from longitudinal studies suggests that many children can recover
from short-lived childhood adversity with little detectable impact in adult
life. Encouraging resilience is important because resilient children are better
equipped to resist stress and adversity, to cope with change and uncertainty,
and to recover faster and more completely from traumatic events.
Linear models also dont have cycles in them. But cycles can mat-
ter. Consider the UK governments 2011 Munro Review of Child Protection
(Munro 2011), which notes that policies, even good ones, can figure in neg-
ative cakes alongside positive ones. The negative cakes diminish the good
effects of the policy and can even, if they are strong enough, outweigh the
good effects. This is just the trouble that the Munro Review pinpoints for
one of the big UK child welfare policies.

De l ib e r at i n g P ol i c y [237]
The policy in question was intended to improve welfare outcomes
in children and young people (CYP in Figure 9.4) by providing stricter
guidelines for what social workers must do in dealing with children and
families and by better monitoring of what they are doing:by ensuring that
specific mandated facts about the family and the child are ascertained and
recorded properly and that all required meetings take place. But this policy,
the Munro Review argues, can have serious negative effects alongside the
intended positive ones. How so? Through various negative feedbackloops.
Have a look at a diagram from the Munro Review (Figure9.4). Two nega-
tive loops are pictured, R1 and R2. Both start in the same way. Increasing
the amount of prescription imposed on social workers, you can reduce their
sense of satisfaction and self-esteem. In R1, this increases staff sickness
and absence rates; in R2, it increases staff turnover rates. Both these effects
tend to result in an increase in average social worker caseload, which leads
to social workers spending less time with each of the children and young
people and their families. This in turn reduces the quality of the social work-
ers relationships with the children and their families, which then reduces
the quality of the outcomes. So the policy may produce bad unintended
consequences. Worse, these negative effects can become amplified via the
feedback loops. When the outcomes are regularly too unsatisfactory, this
reduces social workers sense of self-esteem and personal responsibility,
and the negative cycle is set in motionagain.
Besides the habit of taking causality as linear, we take it to be God-given.
But the kinds of causal principles we rely on for policy prediction are not
God-given. They depend on intricate underlying structures. And our focus
on objectivity and certainty takes our attention away from these underly-
ing structures. To make this point vivid we use an example not from social
policy but from Nancys own daily policies. Nancy often writes her lectures
on paper with a sharp pencil. She sharpens her pencils by putting a kite out
her study window. She can do that because her study was designed by Rube
Goldberg. You can view this Rube Goldberg machine in (Cartwright and
Hardie 2012, 77). Pulling the kite string via a double pulley opens a small
door in a cage containing moths, which are thereby released. The moths
eat a flannel shirt sitting on a scale, and so forth, till finally a woodpecker
becomes free to peck the pencil sharp. Putting a kite out the window is
a very effective policy for Nancy to get nice sharp pencils. Still we dont
advise you to fly a kite to sharpen your pencils. Kite flying undoubtedly
figures in the causal principles that govern pencil sharpening in Nancys
study. It would, for instance, pass any rigorous RCT. Put the kite out the
window on randomly chosen days and you will certainly get more sharp
pencils when you put it out than when you dont. But that principle is local;

[238] The Philosophy of Philip Kitcher

Variety of needs
of CYPs
prescription of
Quality of social
O worker relationship Increased prescription of
Quality of practice especially ICS
outcomes for CYPs with CYP and family
Scope for
dealing with S
variety of CYP
Time spent working
needs with
with CYPs and families
expertise S R1 O
Sense of satisfaction Average C&FSW
self-esteem and personal caseload
Staff absence or
sickness rate S S

O Staff turnover rate

An arrow linking variable A to variable B should be read as a change in the value of A produces a change in the value of B. The qualitative nature of
the link is indicated by a link polarity. These should be read as:
S: the variables move in the same direction ceteris paribus, so a change in variable A produces a change in variable B in the same
direction: if A goes up, B goes up.
O: the variables move in the opposite direction ceteris paribus, so a change in variable A produces a change in variable B in the
opposite direction: if A goes up, B goes down.
double bars on a link indicate a particularly long delay in the causal connection.

Note that the link polarity says nothing about the size, or quantity of the change. The indication of the effect is qualitative only. Moreover, there is no
presumption of a linear relationship between the two variables.

Figure9.4 The complex impact of increased prescription of social work practice. Source:Munro 2011,136.
it depends on the underlying structure of her study. The causal role played
by kite flying in Nancys study is not God-given:it depends on a complex
pattern of interactions in an intricate underlying structure.
Of course this is not a typical social policy case. Social policies suppose
principles like Burnout causes turnover in child welfare service work-
ers or Age does not cause turnover in child welfare service workers. Or
Apathetic-futile mothers are more likely to maltreat their children. These
are clearly not God-given either. And surely it is implausible to suppose
that getting a good social regularity, as these are purported to be, depends
less on the details of the underlying structure than on getting regularities
between pure physical quantities.
Note too that we are not supposing that there are no human universals,
that people in Bangladeshi villages are essentially different from those in
NewYork high-rises, nor across the three hundred language groups and more
than fifty nonindigenous communities of London. In fact our Rube Goldberg
example of a local causal principle works in just the opposite way:by relying
on other causal principles that hold widely. The pencil sharpener depends
on a number of fairly universal principlesfrom the laws of the lever and
the pulley to the familiar fact that moths eat flannelto ensure that the
arrangement and interaction of the components result in the causal prin-
ciple Nancy uses to sharpen her pencils. So the fact, if it is one, that there are
a large number of universal truths about human behaviors, emotions, and
reactions does not demonstrate that the kinds of causal principles we rely on
in typical social policies will be anything other than verylocal.
Our aspirations for certainty divert our attention to these kinds of local
causal principles since they are ones that we can nail down with objective
methods, like RCTs:Flying kites in Nancys study sharpens pencils. Or from
the website of the MIT-based Jameel Poverty Action Lab, after a study in cer-
tain Indian villages:Informing villagers of poor teaching in their villages and
raising awareness of accountability mechanisms had no impact on teacher
attendance.1 Our efforts are taken away from the more difficult study of the
underlying structures that make these causal principles possible.
When it comes not to prediction but to evaluationlooking back to see
what was responsible for an outcomethe focus on linear causal princi-
ples, with their objective certifying methods, leads to skewed views about
human error and individual responsibility. In policy evaluation, just as in
policy prediction, methods and morals often mix in a way they ought not

1. Abdul Latif Jameel Poverty Action Lab, Healthcare Provider Attendance, http://
(accessed December 16,2105).

[240] The Philosophy of Philip Kitcher

to. As Eileen Munro (2005, 377)explains, when a tragedy like the death of
Peter Connelly occurs, the standard response is to hold an inquiry, looking
in detail at the case and trying to get a picture of the causal sequence of
events that ended in the childs death.... We are tracing a chain of events
back in time to understand how it happened. She adds, Unlike the police
investigation, which focuses on the perpetrators of the homicide, these
inquiries focus primarily on how the professionals acted, judging them
against the formal procedures for working with families and principles of
good practice (377). Where does this backward tracing stop? As Munro
argues, the events that bring the investigation to a halt usually take the
form of human error. Practitioners did not comply with procedures or
lapsed from accepted standards of good practice (37778).
But as a UK Department of Health (2000, 20)pamphlet explains:

There are two ways of viewing human error:the person-centred approach and
the system approach. The [person-centered]... approach focuses on the psycho-
logical precursors of error, such as inattention, forgetfulness and carelessness.
Its associated countermeasures are aimed at individuals rather than situations
and these invariably fall within the control paradigm of management. Such
controls include disciplinary measures, writing more procedures to guide indi-
vidual behaviour, or blaming, naming and shaming.

This is just what we saw in the case of Sharon Shoesmith. As is also noted
in the pamphlet, however,

aside from treating errors as moral issues, [the person-centered approach] iso-
lates unsafe acts from their context, thus making it very hard to uncover and
eliminate recurrent error traps within the system....
The system approach, in contrast, takes a holistic stance on the issues of fail-
ure. It recognises that many of the problems facing organisations are complex,
ill-defined and result from the interaction of a number of factors. (2021)

This is just as in our Rube Goldberg pencil sharpener example.

The same worry is studied in Kohn etal.s (2000, ix) To Err Is Human:
Building a Safer Health System, a report commissioned by the US National
Academy of Sciences:

The title of this report encapsulates its purpose. Human beings, in all lines of
work, make errors. Errors can be prevented by designing systems that make it
hard for people to do the wrong thing and easy for people to do the right thing.
Cars are designed so that drivers cannot start them while in reverse because that

De l ib e r at i n g P ol i c y [241]
prevents accidents. Work schedules for pilots are designed so they dont fly too
many consecutive hours without rest because alertness and performance are

The report urges, The focus must shift from blaming individuals for past
errors to a focus on preventing future errors by designing safety into the
system (5). To put this in the terms we have been using, we should be
less concerned with the easier-to-certify causal sequences that start with
human error and end with disastrous consequences and far more con-
cerned with understandingand restructuring the underlying struc-
tures that make this kind of causal sequence likely.
As Munro (2005, 378)notes, When society is shocked and outraged by
a childs terrible tale of suffering, there seems a basic human desire to find
a culprit, someone to bear the guilt for the disaster and to be the target
of feelings of rage and frustration. This puts us squarely in the business
of finding these local linear causal principles, and, with Tony Blair, we can
feel morally and epistemically safe in doing sowe are not likely to cast
blame in the wrong placesbecause these are the kinds of claims about
which, with due care, our objective methods can deliver reasonable cer-
tainty. But the kinds of preventative measures this leads torecall the UK
Department of Health examples:disciplinary actions; writing more proce-
dures to guide individual behavior; or blaming, naming, and shamingare
often unlikely to stop these kinds of sequences occurring. As Munro urges,
Child protection is a systems problem (375). So too are a good many other
social problems, from poor child nutrition in Bangladesh and poor school
attendance by teachers in Indian villages to crime, education, health, and
climate change adaptation almost anywhere. Our thirst for certainty and
our admiration for methods that can be run by rules must not lead us to
buy cheap knowledge that cant serve ourneeds.
For a timely illustration of linear causal thinking in child protection,
consider the investigation into the death of Daniel Pelka, a four-year-
old who died at the hands of his mother and stepfather in March 2012
in Coventry, UK. His death provoked a massive outcry across the United
Kingdom in large part because, according to the Final Overview Report of
Serious Case Review re Daniel Pelka, commissioned by the Coventry Children
Safeguarding Board (CSCB 2013), the social workers (as well as teachers
and police officers) who had been in contact with Daniel and his family
in the months leading up to his death missed twenty-six opportunities to
help him and to act in a way that would have prevented hisdeath.
Because the Final Overview Report blames individuals and it blames
them for failing to behave in certain ways, the lessons it draws are about

[242] The Philosophy of Philip Kitcher

the ways those individuals should have behaved. And the Report ties dif-
ferences in behavior to differences in attitude toward available evidence;
for instance, lesson 15.4 states, Domestic abuse/violence is always a child
protection issue and must always be approached with this as the mind-set
of professionals, and lesson 15.13 states, Professional optimism about a
family and of their potential to change or improve their parenting must be
supported by objective evidence and that any contra indicators have been
fully considered prior to any optimistic stance being taken (CSCB 2013,
72). These lessons make it sound as if the social workers, teachers, doc-
tors, and police officers involved looked at the evidence with rose-colored
glasses, which we do not think the evidence reviewed in the Final Overview
Report supports. This seems to be the view of Munro as well:in a public
interview she said that she cant claim she would have done better on the
evidence presented in the Final Overview Report (Pemberton2013).
In the same vein the Final Overview Report castigates the professionals
involved along theselines:

In this case, professionals needed to think the unthinkable and to believe and
act upon what they saw in front of them, rather than accept parental versions
of what was happening at home without robust challenge. Much of the detail
which emerged from later witness statements and the criminal trial about the
level of abuse which Daniel suffered was completely unknown to the profession-
als who were in contact with the family at the time. (CSCB 2013,6)

Blaming the professionals involved for failing to think the unthinkable

in this way is odd. Apediatrician should not be blamed for not including
abuse as a likely diagnosis when a mother brings a child to him concerned
that her child is losing weight. Deliberate starvation as the cause is highly
improbable, and any number of physical illnesses are far more probable.
Likewise what the social workers saw in front of them clearly wasnt that
Daniel was being abused in horrific ways by his parents. The sordid details
of Daniels abuse emerged only after a full-blown criminal investigation
that involved searching the house and the parents cell phones, which is
not, as we understand it, something the Coventry social workers or even
the police could have done at an earlier stage. These social workers, by con-
trast, had access to concrete evidence supporting the view that Daniel was
not being abused by his parents, as the Final Overview Report itself states.
Besides the morally questionable aspects of blaming individual social
workers, the Final Overview Report seems to suppose that the way to pre-
vent harmful outcomes like Daniels death is to turn the professionals
involved into better and better detectors of abuse. Better in this case

De l ib e r at i n g P ol i c y [243]
means fewer false negatives (and correlatively, it seems safe to assume,
more false positives, with more families subjected to nasty investigations
and sometimes losing their children at a lower level of evidence). From the
evidence in the Final Overview Report itself, the Daniel Pelka case has the
earmarks of a systems issue:it is the result of many minor flaws in practice
plus the basic incompleteness of information rather than one devastating
individual act. Each failure of each individual was just one of many ingre-
dients in the causal cake that led to the failure to see how much danger
Daniel was in. So it is not obvious that improving the detectors of abuse in
that system is the best way to improve child welfare and reduce the prob-
ability of such abuse occurring again. Perhaps it would be better to redesign
the system so it is easier to do it right and harder to do itwrong.
It should also be noted that the lessons the Final Overview Report pro-
poses to turn into rules of best practice are inherently local in character,
so that following these rules might in fact be harmful. The Report seems
to have used the following process to arrive at a recommendation for best
practice:Professionals are to be blamed for failing to act in way W when
doing so would have prevented harmful outcome O in the Pelka case.
Therefore professionals should, as a rule and whenever relevantly similar
situations arise, act in way W.Note, moreover, that since the failure of the
professionals to act in way W is perceived to be a moral failure, the rule
stating that one should act in way W becomes a moral rule. The issue here is
that a rule the application of which might have prevented harmful outcome
O in the Pelka case might not contribute to preventing the occurrence of
harmful outcomes of the same kind in other contexts. In fact if applied
widely it might contribute to promoting the occurrence of such outcomes.
Consider as an illustration of this worry the case of lesson 15.8 (CSCB
2013, 72):Any facial injuries to a child must be viewed with concern, with
physical abuse needing to be actively considered as a possible cause, and clear
records, interventions or referrals made accordingly. As Munro remarks
(Pemberton 2013), bruises to the head arent that uncommon in four-year-
olds, which means that strictly following 15.8 might lead to a great number
of false positives, that is, of cases in which parents are wrongly suspected of
abuse. Further, wrongly stigmatizing parents as abusing their children may
decrease the welfare of children in the population at large, for instance if a
large number of false positives leads families to distrust social workers.
The potential harmful effects of lesson 15.11 (CSCB 2013, 73)are even
more striking:

For professionals from Childrens Social Care or the Police to defer to medical
staff for the provision of the primary evidence to confirm or otherwise whether

[244] The Philosophy of Philip Kitcher

an injury to a child was the result of abuse or not, could be unhelpful, particularly
when no definitive view one way or the other can be given. To do so could lead to
any following investigation being inappropriately downgraded and implies that
other aspects of the child[s] life are less significant for the purposes of assessing
the existence of childabuse.

Taken at face value, the recommendation is simply that social workersas

a rulebe skeptical toward medical opinions regarding the likely causes
of physical injuries in children. But such skepticism can lead to a general
distrust between social workers and doctors. And if families are aware that
social workers may disregard medical diagnoses in establishing abuse, this
might lead them in turn to distrust social workers. And this is before one
even takes into account the fact that for social workers to proceed in this
way may lead to a greater number of false positives.
We chose to discuss the Daniel Pelka case to illustrate worries about lin-
ear models of causation. Our final worry about trusting that good objec-
tive methods can deliver a high degree of certainty is that these methods
only deliver answers in the language in which we ask our questions, and
that is often not the language in which the system under study operates.
Very often we dont know what we are testing with these methods. In fact
believing we do know exactly what we are testing may be the problem.
Our best objective methods for testing causal claims, like RCTs, require a
precise characterization of both the cause and the effect. This is a crucial
part of the study protocol. We must ensure that everyone in the treatment
group gets the same treatment and that there are strict criteria for deciding
whether the effect has occurred. Otherwise the validity of the conclusion
is impugned. This in turn means that treatment and effect descriptions are
couched in concrete operationalterms.
Theres the rub. For policy prediction we need causal principles that hold
widelyat least widely enough to cover both the study situation and the
target. But these kinds of principles often relate not concrete concepts of
the kind operationalized in a good study but far more abstract concepts,
such as the lever and pulley that matter in the Rube Goldberg pencil sharp-
ener. This kind of problem looms large for social policy because the same
thing doesnt mean the same in different social settings. We give two exam-
ples below, taken from Cartwright and Hardie (2012).
Consider, first, the Integrated Nutrition Program that provides nutrition
education to mothers to improve the health of their infants. This program
was a success in Tamil Nadu, but not in Bangladesh. According to the World
Bank post hoc analysis, thats because in many rural Bangladeshi households
fathers, not mothers, do the shopping, and mothers-in-law are in charge of

De l ib e r at i n g P ol i c y [245]
food distribution. In Tamil Nadu there was a generalizable cause of improved
infant health:the nutritional education of a person (i)who is responsible
for household food selection, and (ii) who is responsible for food distribu-
tion, and (iii) who holds the infants welfare paramount in carrying out these
responsibilities. The Tamil Nadu study asked about mothers. And in Tamil
Nadu mothers refers to a class of people with the requisite characteristics to
improve infant health via nutritional education. But not in Bangladesh. If we
wanted more generalizable results, we were asking the wrong question.
The second example is another case from UK child-welfare policy, not
one that we currently have strong evidence for but where there seems to be
cause to worry. In many cases a childs caregivers, though not legally com-
pelled, are encouraged to, even badgered into attending parenting classes.
This includes fathers. But what constitutes a father? Is father instanti-
ated by biological father, or male partner of the mother who lives in the
household with the child, or maybe male caregiver? Maybe the policy will
be effective if the male caregivers or men living with the mother are tar-
geted, but not biological fathers. If so, to focus on being a father would be
to move to too high a level of abstraction since only the more specific male
caregiver or male partner of mother who shares the childs household
enters into a reasonably reliable principle. On the other hand compelling
the male caregiver to attend classes can be too concrete. Different cultures
in the United Kingdom have widely different views about the roles fathers
should play in parenting. Compelling fathers to attend classes can fall under
the more abstract description ensuring caregivers are better informed
about ways to help the child, in which case it could be expected to be posi-
tively effective for improving the childs welfare. But it may also instantiate
the more abstract public humiliation, in which case it could act negatively.
And of course it can fall under both at once. In any case, if the two more
abstract features pull in opposite directions, there will be no reliable prin-
ciple to formulate at the more concrete level involving fathers. Nor is this
pull in opposite directions an unrealistic hypothesis. We know from empiri-
cal research that there are varying outcomes associated with compelling or
strongly encouraging parents to attend parenting classes and also that these
are correlated with varying motivations (see, e.g., Barlow etal.2006).
Let us begin now to tie matters together. We have been criticizing three
mistaken, mutually supporting philosophical stances that contribute to
bad policy decisions:

1. We bank on certainty.
2. We do not trust the kind of open, multimethod, theory-infested scien-
tific investigations it takes to uncover the structures of these underlying

[246] The Philosophy of Philip Kitcher

systems but prefer objective methods, like RCTs, that deliver only sur-
face relations.
3. We take causal relations to be linear and God-given and do not tend suf-
ficiently to the underlying systems that make these relations possible.

Nancys mother maintained that a little learning is a dangerous thing.

Thats what we see here. It has even become encoded in our language,
in a familiar expression now used everywhere from newspaper reports
to private conversations to policy deliberations:Studies show But
studies show very little. No matter how good a study is, it can deliver
facts only about the population studied. To go beyond the boundaries of
the population enrolled in the study it takes theory, and to have a war-
rant for doing so it takes good, well-supported theory. Theres no getting
around that. Sometimes we seem to act as if we believe in induction by
simple enumeration:Swan 1 is white, swan 2 is white so all swans are
white. Study population 1 does x, study population 2 does x so all
populations do xbut with the additional drawback that we would usu-
ally be generalizing from a very small inductive base indeed, not tens of
thousands of white British swans but one or two studies or, in the best
of cases, a handful. And a visit to Sydney, Australia soon puts paid to the
idea that all swans arewhite.
It is as if we have forgotten the lessons about simple induction that
have been rehearsed generation after generation for eons. Recall Bertrand
Russells chicken:She infers, on a very good basis, that when the farmer
comes in the morning, he feeds her. That inference serves her well until
Christmas morning, when he chops off her head for Christmas dinner. Of
course the chicken did not base her inference on a randomized controlled
trial. But had we conducted one we would have obtained exactly the same
results. Her problem was not her study design but rather that she was
studying surface relations. She did not understand the underlying socio-
economic system that gave rise to the causal relations she observed. We
often act as if the methods of investigation that served the chicken so badly
will do perfectly well forus.
We have just one sentence in conclusion. Now as then we must take seri-
ously the World War II warning, Loose talk can cost lives.2

2. With the help of Nate Rockwood. Both Nancy Cartwright and Alexandre Marcellesi
would like to thank the UK AHRC project Choices of Evidence: Tacit Philosophical
Assumptions in Debates on Evidence-Based Practice in Childrens Welfare Services
and Eileen Munro in particularfor support for the research and writing of thisessay.

De l ib e r at i n g P ol i c y [247]
Reply toCartwright and Marcellesi

I have been learning from Nancy Cartwrights work, and from conversa-
tions with her, for more than thirty years. Sometimes the readingsand
especially our face-to-face exchangeshave planted seeds that have con-
tinued to flower in my thinking. In the essay they have written for this
volume, Nancy and Alexandre Marcellesi continue that line of interaction,
asking me to consider aspects of the uses to which scientific research is put
on which Ihavent previously focused. This reply will take up the invitation.
In my 2001b book, and more extensively in my 2011b book, Im con-
cerned with Science as an institution that interacts with other institutions
in contemporary societies (those that characterize themselves as democ-
racies). The philosophical project I conceive is melioristic in character: it
surveys the interacting institutions in hopes of improving their joint func-
tioning. (Regrettably Ihavent always been as forthright as Imight have
been about the pragmatism of my approach, inspiring some commentators
to interpret me as being more utopian than Iaim to be; my reply to John
Dupr tries to say some things I should have said earlier.) The center of
my interest has been Science conceived as a system of public knowledge,
potentially available to radiate out into individual and social projects. Ive
been asking what might valuably be done to improve that system, in its
ways of setting the research agenda, in its modes of certifying potential
novel findings, and in the channels through which certified results are
distributed. In her recent work Cartwright and her coauthors have been
probing many instances of a specialand vitally importantclass:public
policy decisions to investigate questions that bear directly on human wel-
fare, to demand particular methods for certifying answers, and to set up
mechanisms for intervening on the basis of the certified results.
One of Cartwrights enduring contributions to our discipline, encapsu-
lated in her brilliant 1999 book, is the recognition that the sciences provide
a bundle of loosely connected domains, in which nature goesor more
usually, can be set up to goreliably and predictably. Cartwrights term
for these is nomological machines; Ill call them spheres of order. She
reminds us, rightly, that these spheres are local, that they cannot be easily
extended to operate smoothly on a wider scale, and that their functioning
depends on a variety of special conditions (of which we often only slowly
and painfully become aware). This aspect of her views elaborates ideas
present in Neurath (to whom she is explicitly indebted) and in Dewey.
Deweys ([1925] 1981, [1929] 1984)account of the role of the sciences in

[248] The Philosophy of Philip Kitcher

thought and in action is that inquiry (his term) constructs spheres of order
in response to our evolving aims. He sees philosophical reflection on that
role as directed toward the potential refinement of a crucial function, as
philosophers consider the pressing questions toward which inquiry ought
to be directed, and how the activity of order-building might more reliably
advance our proper aims. Well-ordered science is one way of articulating
this side of Deweys thought.
The Neurath-Dewey vision recognizes the relevance of values to scien-
tific research and of questions about values as central to the philosophy of
science. That recognition is shared by a growing number of contemporary
philosophers, like Cartwright and others, and Heather Douglas (2009) has
given it the focus it has long needed. Cartwright and Marcellesi prompt me
to think, however, that we havent yet come to terms with the full range of
value questions that ought to be addressed.
Its time to descend from the abstract framing of the issues to the concrete
cases Cartwright and Marcellesi present. Children are abused. Prisoners
released from jail commit new crimes. People with particular profiles
tend to cause trouble. These are the first-order phenomena. Because theres
wide agreement that abuse, crime, and trouble are bad, theres an attempt
to reconstruct these domains, to create spheres of order that regularly and
reliably generate more desirable outcomes: fewer abused children, less
crime, less trouble. We have a second-order domain, one in which research
and public policy interact, that generates modifications of the first-order
domains. Cartwright and Marcellesi point out how the modifications
close monitoring of families, psychological testing of prisoners, profiling
young people who have not yet done anything blameworthy (a class of
cases especially salient for those living in St. Louis, Baltimore, or NewYork
in early 2015)are also ethically compromised. They trace the trouble to
a number of features of that second-order domain in which research and
policy interact. They see an unwarranted faith in RCTs, an overreliance on
linear causal models, a failure to appreciate the many special conditions on
which the achievement of order depends, as well as neglect of the ethical
questions that ought to have been raised. In the grip of an account of sci-
ence that views conclusions obtained by the favored method du jour as firm
and ethical claims as inevitably squishy, policymakers charge confidently
ahead, replacing one disorderly domain with another that is equally prob-
lematic, or possibly even ethicallyworse.
In effect Cartwright and Marcellesi are studying a putative second-order
nomological machine, one set up for the regular and reliable produc-
tion of first-order machines aimed at decreasing the frequencies of child
abuse, crime, and social trouble. In the second-order attempt at order,

De l ib e r at i n g P ol i c y [249]
researchers doing socially relevant science follow certain supposedly privi-
leged methodsfor example, doing RCTs on populations that lend them-
selves to studywhich they pass on to bureaucrats who design broad-scale
policies without considering either the possible unintended effects of the
extrapolation or the ethical issues that inevitablyarise.
So we have a second- order misadventure, generating first- order
attempts at social engineering that repeatedly go awry. Viewing the situ-
ation in those terms poses a third-order question, one for philosophers
of science: How would it be possible to make the second-order domain,
in which researchers and policymakers interact, function better? Ive no
doubt that the status quo might be improved if those involved in shap-
ing the research agenda and pursuing inquiries of the pertinent types were
thoroughly cognizant of the Munro (2011) report, as well as Cartwright
(1999), Cartwright and Hardie (2012), Douglas (2009), and the essay
Cartwright and Marcellesi have written for this volume. Yet Idont think
this is sufficient. In the remainder of this reply Ill draw on my reading of
Cartwrights work, as well as my own ideal of well-ordered science, to make
an attempt at pragmatic meliorism.
The sacking of Sharon Shoesmith illustrates the procedures policymak-
ers tend to follow when terrible effects come to light. Reluctant to ques-
tion the system, officials look for a culpable actor whose behavior caused
the damage. Cartwright and Marcellesi want policymaking to rely less on
taken-for-granted (objectively established) results. Yet if the source of the
trouble is an unwarranted extrapolation from cases where order is achieved
by reliance on unrecognized local conditionsthe kind of fundamental-
ism Cartwright (1999) exposeshow is a better policy to be crafted? Lets
suppose that responses to breakdowns in existing policy are less peremp-
tory than they were in the case of Peter Connelly, that theres a serious and
thorough inquiry to understand the actions of the relevant actors (and the
burdens policy places on them and the limitations of their possible impact
on the circumstances with which they must deal). Lets even suppose that
the inquiry involves participation of people with different viewpoints,
social workers and parents, as well as academic investigators and bureau-
crats. Those who take part have digested the insights of Cartwright and her
coauthors, and they conclude that there is indeed a systematic problem.
Shoesmith and her staff are exculpated.
What next? It might turn out that there are ideas in existing psycho-
logical or sociological studies suggesting what could have gone wrong. In
that case there would be an indicated direction for amendment. Even so,
those appreciative of the Cartwright-Marcellesi critique would quickly
recognize the high probability of those studies also being dependent on

[250] The Philosophy of Philip Kitcher

unacknowledged local conditions. Extrapolation from them is not likely to
work on the scale desired. Moreover when, as is often the case, background
research provides no clues about how to fix the existing system, experi-
menting is the only option. Once you appreciate Cartwrights important
point about fundamentalismespecially apt in the complex causal situ-
ations with which ventures in social engineering are concernedyou see
that the certainty she and Marcellesi question must give way to a tenta-
tive, pragmatic, experimental approach. Trial and error is the only serious
Where there is tentative trial, there will surely be error. Moreover,
as with the system you aim to improve, the errors will be highly con-
sequential. Children will die. Young lives will be permanently blighted.
Injustice will be done. Even when experimentation is cautious and those
who carry out the experiment ethically scrupulous, some effects of the
experiment will almost certainly be regretted by those who look back on
it from a more enlightened perspective. When that fact becomes vivid,
its natural to ask whether societies should even try to achieve order in
these domains; without certainty, social policy becomes ethically suspect.
No hands areclean.
I dont draw that conclusion. Like Dewey, Iview social experimentation
as a principal component of the ethical project. For tens of thousands of
generations, our species has been committed to attempts to structure our
lives together, without advance assurance that those attempts will bring
nothing but good. As we continue, more self-consciously, to experiment,
we cant back away from efforts to promote important shared valuesto
increase the proportion of children whose development prepares them for
happy and worthwhile lives, to decrease the incidence of violence and the
prevalence of injustice. Given the pragmatic naturalist ethics to which Im
committed, the judgment of the permissibility of experimentation in gen-
eral depends on supposing that a fully representative assembly of delib-
erators, well-informed and mutually engaged, would reach consensus on it
(see Kitcher 2011a and my reply to Michael Smith).
That seems a safe bet. There are some effects we attempt to remedy, so
destructive and debilitating that they cannot be ignored, unless we sup-
pose any attempt to prevent them would be hopeless. And though the task
of fathoming complex causal connections is difficultso that sequential
trial is likely to generate plenty of error on the way to successour efforts
are sometimes rewarded. Think of the checkered history of medicine from
ancient times to the present. So Id expect my ideal deliberators to buy into
the project of social experimentation, recognizing in advance that, how-
ever carefully it is conducted, terrible harms will be inflicted on unknown

De l ib e r at i n g P ol i c y [251]
people. (The Rawlsian device of the original position seems to generate the
same result, allowing an institution for undertaking social experiments.)
Yet Id also expect them to want to amend our actual institution
which is, effectively, as Cartwright and Marcellesi make clear, one that
introduces attempts at order without advance knowledge of the range
of their effectsand not only by supplying the existing decision makers
with information about the complexities of causation. Proposals for social
experiments ought to involve those most likely to be affected by the ways
the planned sphere of order is to be constructed. Representatives of fami-
lies from different socioeconomic classes and with different structures
should be involved in the conversation about child welfare, as should the
social workers who will be asked to implement the envisaged policies.
Experimental review should be frequent and attentive to the possibilities
of various kinds of damaging consequences, not merely to the problem
for which the policy seeks a remedy.
Nor is this enough. Many of the examples Cartwright and Marcellesi
present consider policy interventions introduced in isolation from
related problems. This is most evident in relation to crime, where profil-
ing is defended by appeal to statistics, without any consideration for
background circumstances that might lie behind the numbers cited. To
inaugurate or continue a policy of stopping young men of a particular
race, without doing additional experiments to see what might be done
to increase the opportunities and prospects of members of the targeted
group, is to conduct an illegitimate experiment, one that an informed,
mutually engaged, comprehensive deliberation would almost certainly
If, as Ibelieve, we are committed to experimentation aimed at creating
valuable order in social domains, we are also committed to monitoring the
experiments and to recognizing the injustice of conducting certain kinds
of trials in isolation from complementary efforts. To reform the research-
policy institution whose current methodological defects Cartwright and
Marcellesi expose requires systematic attention to the background social
context, recognition that pervasive features of the society compromise
the ethical standing of many widely accepted social experiments. Only in
the context of a cluster of policy attempts may an apparently attractive
individual proposal count as just. If it were subjected to the kind of con-
versation well-ordered science demands, policymaking would be directed
to a package of trials that would better promote justice. Methods and
morals really do mix. We need both the causal insights Cartwright and
her coauthors offer and a more comprehensive conversation.

[252] The Philosophy of Philip Kitcher


Function and Truth inEthics


I n the first part of The Ethical Project Philip Kitcher (2011a) provides a nat-
uralistic account of the function of ethical practice. Though he conceives
of this as an unashamedly empirical taskhe denies that the function
of ethical practice is to get at the ethical truth or to get people to behave
in accordance with the ethical truthit is not one that he thinks has a
debunking upshot. Instead he thinks that his account of the function of
ethical practice, which is based on his reading of the practices origins in
the behavior of our hominid ancestors, provides us with all of the materials
required for a pragmatic naturalist account of ethical truth. More familiar
a priori attempts to ground ethical truth in the spirit of Kant are given
short shrift, as are the supposedly more empirically adequate noncognitiv-
ist accounts of ethical judgment which hold that such judgments express
emotions, commitments, intentions, desires, or noncognitive attitudes of
some otherkind.
Though there is a great deal to admire in The Ethical Project, especially for
someone like me, who was so antecedently ignorant of the relevant science
and history, Iam afraid that Iremain unconvinced. My discussion is divided
into three main sections. In the first I consider Kitchers account of the
function of ethics and his pragmatic naturalist account of ethical truth. If
ethical truth really were fixed by the function of ethical practice in the way
he suggests, then this would make the truth of ethical claims turn on issues
that seem to me, and Iwould have thought to Kitcher too, to be quite irrel-
evant. It would also mean that many of our ethical beliefs, including some
of those required to maintain ethical practice so understood, are false. This
doesnt entail that his account of ethical truth is mistaken, but it does sug-
gest that something might have gone wrong, and Isuggest what that some-
thing might be. In the second section Iargue that Kitchers account of the
function of ethical practice does not commit him to a pragmatic naturalist
account of ethical truth. Noncognitivism remains an alternative, notwith-
standing his reservations. Since it does a much better of job of shoring up
our ethical commitments than his own pragmatic naturalist account, Isug-
gest that Kitcher would be better off embracing some form of noncognitiv-
ism. In the third and final section Ibriefly consider Kitchers objections to
Kantian approaches to ethical truth, and Iexplain why at least one version
of this approach, the one that Imyself prefer, is not vulnerable to his objec-
tions. This too remains an option for someone who accepts his account of
the function of ethical practice.



According to Kitcher, ethical practice has its origins in the biologically

driven need our hominid ancestors had to live in cooperative communities.
This need led to their having a direct concern not just for their own kin but
also for those who were members of the cooperative coalitions to which
they belonged, as their limited altruistic concern gave rise to the coopera-
tive foraging behavior that enabled the members of these coalitions to do
better than those who lacked such concern.
Since altruistic concern wasnt always strong enough to override other
dispositions the members of these coalitions had to please themselves, a
tendency to engage in elaborate and time-consuming peacemaking behav-
iors developed alongside such concern:

Peace and mutual tolerance are typically hard-won. Precisely because of this,
observations of chimpanzee societies disclose periods of intense social interac-
tion, lengthy bouts of grooming undertaken to reassure friends who have been
disappointed by recent behavior. At times of great tension within a group, chim-
panzees can spend up to six hours a day huddled together, vastly longer than
any hygienic purpose demands. (Kitcher 2011a,73)

Though limited altruistic concern and elaborate peacemaking strategies

were sufficient for cooperation among individuals so long as the groups
remained small, by the time our ancestors were using tools that required
a trade in raw materials that were mined and transported large distances,

[254] The Philosophy of Philip Kitcher

and even more so by the time they lived in cities of a thousand or more,
strategies of peacemaking through face-to-face reassurance [were] no
longer applicable, so there had to be a system of agreed-upon rules for
forestalling potential conflicts and for dealing with people who are relative
strangers (116). Individual behavior thus came under the control of what
Kitcher calls normative guidance (6769).
What is the distinctive feature of normative guidance? According to
Kitcher (2011a, 69), there is normative guidance whenever the cumber-
some peacemaking of our original hominids is replaced by a new device,
one preempting rupture rather than reacting to it, and in principle capable
of operating in a wide variety of contexts. As Iunderstand it, this consists
in the possibility of our having and conveying thoughts with normative
contents, normative thoughts that themselves become a source of motiva-
tion: The simplestand originalform of normative guidance consists
in an ability to transform a situation that would otherwise have been an
altruism failure, by means of a commitment to following a rule:you obey
the command to give weight to the wishes of the other (74). Normative
guidance is thus the hallmark of ethical practice as we know it, as in large
groups cooperation is underwritten by thoughts about the behaviors
required by rules and commands. Ethical practice, so understood, has
been on the scene for much longer than we might have thought:By thirty
thousand years before the present, the enterprise of framing rules for life
together, the ethical project, must have been quite well developed(118).
If the function of ethical practice is to remedy altruism failures, then we
can evaluate particular moments in the history of ethical thinking by that
very standard. For this reason Kitcher sees certain transformations in the
ethical thinking of our ancestors and ourselves as examples of ethical prog-
ress: the move from an eye-for-an-eye conception of retribution, which
allowed for those other than a wrongdoer to bear the burden of punish-
ment for a wrongdoers actions, to a conception that makes the wrongdoer
himself bear that burden; the move from a conception of wrongdoing that
emphasizes a loss of honor to one that emphasizes the common good; the
end of slavery; the extension of legal and political rights to women; and the
partial but continuing shift among heterosexuals toward more accepting
attitudes toward homosexuals (Kitcher 2011a, 13865). In each case those
who were candidates for altruistic concern all along, but were excluded
from it, have been or are being brought within itsscope.
Should we suppose that the ethical judgments that favor extending
altruistic concern in these ways are true? Kitchers (2011a, 246)answer is
that we should suppose that they are true because the concept of ethical
progress can itself be used to define a conception of truth for ethical claims:

Function and Truth inEthics [255]

Descriptive counterparts of ethical rules count as true just in case those rules
would be adopted in ethical codes as the result of progressive transitions and
would be retained through an indefinite sequence of further progressive transi-
tions. There is no prior conception of ethical truth, so that people make ethical
progress when they discover (or stumble on) independently constituted ethical
truths. Progress is the prior notion, and descriptive counterparts of rules come
to count as true in virtue of the fact that they enter and remain in ethical codes
that unfold in a progressive sequence.

We didnt just change the way we treated women; women deserved

to be treated better than they had been treated in the past, and this
propositionthe proposition that women deserved better treatmentis
true. It is true, or so we have reason to believe, because the rule requiring
us to treat women in this way would be retained in every further progres-
sive transition in our ethical thinking.
Attractive though this picture is, there are, Ithink, some hidden prob-
lems. The main source of these problems lies in Kitchers characterization
of the function of ethical practice. To repeat, he tells us that function is to
remedy altruism failures, or alternatively to satisfy the endorsed desires
of all (2011a, 241). But given his own description of ethical practice,
neither of these is quite accurate. A better characterization would be to
remedy those altruism failures that would interfere with cooperation, or to sat-
isfy the endorsed desires of all to the extent required for cooperation. This is
because progress itself isnt characterized ethically but is rather a matter
of ongoing cooperation. Kitcher himself is of course aware of this, but he
doesnt always bear it in mind when he explains how ethical progress has
and willoccur.
Consider, for example, his explanation of why it would count as ethical
progress to extend altruistic concern to nonhuman animals:

Is anything similar available in the case of nonhuman animals? Apparently

so. Just as the late Paleolithic witnessed first occasions of transient associa-
tion among neighboring groups and later increases in band size, so, at the very
end of this period, people began to set up more regular patterns of association
with some kinds of nonhuman animals. The practice of domestication creates
something like a society, one including some nonhuman members. That practice
refines the ethical function of satisfying the endorsed desires of all. Yet, because
the animals newly included differ in some important properties from the people
across the river, obvious questions arise. Does this expansion really create any-
thing like a society? If so, to what extent can the rules adopted within the local
group be carried over to the new members?

[256] The Philosophy of Philip Kitcher

... Just as participants in the ethical project have learned, over tens of thou-
sands of years, to deal first with nongroup members in a limited range of con-
texts, to form larger societies governed by a common framework of rules, and
ultimately to apply an ethical code to people with whom interactions are neg-
ligible or even nonexistent, so too we can understand a progressive sequence
of changes identifying altruism failures in inflicting, or even tolerating, pain in
animals whose anatomy, physiology, and behavioral reactions resemble those
already embraced within the ethical framework....
The idea that it is good to relieve pain, wherever it occurs, is a natural exten-
sion of ideas about the good, progressively elaborated during the evolution of
the ethical project. Accompanying that idea is a related thought about human
beings:it is not good for people to be insensitive to pain, whether it occurs in
other people or in nonhuman animals. Inflicting painor even permitting
itproduces human beings who are debased, whose characters and lives are
less good than they might be. As the ethical project evolves, views of the good
human life become richer, and engaging in conduct that causes unnecessary
pain to others, including nonhuman animals, comes to appear detrimental to
living well. (2011a,3079)

What Kitcher says here sounds like a credible bit of ethical reasoning, but
we arent officially supposed to be in the business of evaluating the thought
that it is good to relieve pain, wherever it occurs, in terms of its intrinsic
plausibility. That would be for progress to be ethically characterized. Instead
we are supposed to evaluate it in terms of its power to serve the function
of ethics, where that function is to fix the content of rules whose adoption
would remedy altruism failures that would interfere with cooperation.
Seen in this light it seems quite implausible to suppose that the rule we
have to adopt, for ethical practice to serve its function, is the injunction
to relieve pain, wherever it occurs. It is quite implausible because we are
evidently able to ignore vast quantities of pain, both human and nonhu-
man, without that having any affect at all on the levels of cooperation we
enjoy. A failure to extend altruistic concern to nonhuman animals quite
generally would thus seem to leave cooperation largely intact. Moreover to
the extent that a failure to extend our altruistic concern to nonhuman ani-
mals in some limited way would have an effect on cooperation, the explana-
tion of this would have to be that a failure to extend such concern would
interfere with cooperation, not the truth of the claim that it is good to
relieve pain wherever it occurs. In the first instance what is important is
the cooperation of other humans, not the alleviation of nonhuman animal
pain. Indeed Itake it that the claim that it is good to relieve pain, wher-
ever it occurs, thus turns out to be false, on Kitchers way of understanding

Function and Truth inEthics [257]

ethical truth. What is true is rather that it is good to relieve pain, wherever
it occurs, to the extent that a failure to do so would interfere with cooperation, a
claim that does not have the same ring of intrinsic plausibility as the origi-
nal. There is thus a huge gap between the rules we need to adopt in order
for ethical practice to serve its function, which is ongoing cooperation, and
the ethical beliefs that many of us have and give voice to insofar as we talk
with each other about these rules, Kitcher included.
A similar problem arises when Kitcher (2011a, 153) explains why the
extension of legal and political rights to women counts as ethical progress,
though in this case he is more upfront about whats atstake:

What was discovered? Factual knowledge advanced:people learned that, under

different conditions of socialization, women wanted things traditionally denied
to them; that they found satisfaction in attaining some of these things; that
fulfillment of the wishes did not thwart desires previously seen as central to
female naturepublic life combined more or less satisfactorily with family
life. Increased factual knowledge proliferated desires for access to public life,
fostering acceptance of the desires as prevalent and no longer pathological.
Recognition of the suppression or frustration of those desires aroused sympa-
thy, recruiting male as well as female allies for the reform movement. Like the
early elaboration of normative guidance, in which particular altruism failures
cause too much trouble, the increase of sexual egalitarianism occurred partly
because, in the end, traditionalists wanted a quieterlife.

As Amia Srinivasan (2012, 18)says in her commentary on this passage:

This may or may not be a sound account of how the feminist movement origi-
nated, but the point is that Kitcher interprets the question Why do we believe
that women and men have equal rights? as a request for causal explanation,
not as a demand for justification. Feminism is justified not because women want
to be treated like men and because it so happens that granting them equality
doesnt cause too much damage to family life:it is justified because of the equal-
ity of men andwomen.

Indeed in the terms in which Iput it earlier, on Kitchers way of thinking

about ethical truth, it turns out to be false that women are the equals of men.
What turns out to be true is rather that women are the equals of men to the
extent that mens failure to treat them as equals would interfere with cooperation.
Once again this claim doesnt have the same ring of intrinsic plausibility.
To repeat, what these examples bring out is the gap that exists between
the rules we have to adopt if ethical practice is to serve the function Kitcher

[258] The Philosophy of Philip Kitcher

says it has, and the ethical beliefs that many of us, Kitcher included, natu-
rally have and express insofar as we take these examples to be examples
of ethical progress. Nor should that be surprising given that the idea of
cooperation that is doing the work in Kitchers account of ethical practice
has to take the empirical facts as given. To stick to the example of relations
between men and women, the explanation of their cooperation must take
the preferences of men and women as given, and it must also take as given
their relative positions in a power hierarchy, where that power hierarchy
might itself be kept in place in part by the fact that certain false ethical
beliefs are widely shared. With that idea firmly in mind, consider a some-
what more cynical explanation of how men came to extend their altruistic
concern towomen.
Lets assume that though men had more power than women and that
though many of them would have preferred to continue to dominate
women in the ways they always had, they did not have sufficient power
to continue to do so while retaining their cooperation. They therefore
had to compromise. How much men needed to compromise is, however,
an open empirical question, a question whose answer is settled by where
the equilibrium points lie in the satisfaction of mens and womens prefer-
ences, given the differences in their power relations. Ethical beliefs take
on a special significance in this context, given the effect they can have on
cooperative behavior. If, for example, both men and women could come to
believe that women are the equals of men simpliciter, and the effect of their
acquiring this belief was to limit the extent to which men had to compro-
mise with womenperhaps women would no longer agitate for all of the
changes that the truth of the belief would demand if they believed (falsely)
that, because men were on board, they would make those changes will-
inglythen widespread possession of the false ethical belief that women
are the equals of men simpliciter, precisely because it would limit the need
for men to compromise with women and allow them to maintain their
superior position in the power hierarchy, would be a crucial part of what
made the ethical progress Kitcher identifies possible. There would, how-
ever, be a very substantial gap between the content of the rules that men
and women adopt and the content of the ethical beliefs that make their
adoption of those rules stably support their cooperation. It would be no
surprise if we so easily fall into a more intrinsically plausible commentary
when we explain ethical progress.
Of course this more cynical story wouldnt be interesting if it didnt
seem to be a more or less accurate description of what has in fact happened
in relations between men and women, insofar as mens altruistic concern
has been extended to women. But precisely because it does seem to be so

Function and Truth inEthics [259]

accurate, the interest of the story is plain. It suggests that if we were to
accept both Kitchers account of the function of ethical practice and his
pragmatic naturalist account of the truth of ethical claims, then his prag-
matic naturalist account of truth may well end up debunking the truth of
many of our ethical beliefs. To function well in remedying failures of altru-
ism to the extent required for cooperation, ethical practice may itself cause
us to have ethical beliefs whose possession is cooperation-conducive, but
ethical beliefs that turn out to be false, given Kitchers account of the truth
of ethical claims. This, at any rate, is what seems to be the case as regards
ethical relations between men and women, and a similar story looks like it
might be right regarding ethical relations between humans and nonhuman
animals aswell.
It is perhaps worth noting that what I have said here is in much the same
ballpark as Kim Sterelnys (2012) criticisms of The Ethical Project. There are,
however, important differences. Like me, Sterelny thinks that an account
of ethics like Kitchers might well make many of our explicit ethical beliefs
turn out to be both crucial for cooperation and false, given that they are so
ripe for manipulation by power elites. However, Sterelnys argument for
this conclusion begins by taking issue with Kitchers claim about the role
played by normative guidance in ethical practice. In his reply to Sterelny
this is the issue on which Kitcher (2012a, 17274) focuses. By contrast,
I have accepted that normative guidance plays the role that Kitcher says
it does in ethical practice; I have argued that this means Kitcher himself
needs to distinguish between the rules that we adopt and act on, on the one
hand, and the explicit ethical beliefs that we express when we give our com-
mentary on these rules, on the other, ethical beliefs that are themselves
cooperation-conducive; and I have argued that, to the extent that norma-
tive guidance plays the role that Kitcher says it does, there will often be a
gap between these two. If ethical truth is defined in terms of the rules we
adopt, in the way Kitcher supposes, it follows that he himself must admit
that the explicit ethical beliefs many of us have, ethical beliefs that may
well be crucial to the stability of ethical practice, could well turn out to be



Is it possible to accept Kitchers account of the function of ethical prac-

tice but reject his account of ethical truth? Or, to put the question in the
slightly different terms suggested by Srinivasan, is it possible to disconnect

[260] The Philosophy of Philip Kitcher

questions of causal explanation from questions of justification and to con-
cede that while Kitcher answers the causal questions correctly, his answers
to the questions of justification are incorrect? If so, then that would leave
us free to admit that though we have in fact internalized certain rules as
the result of transitions in ethical thinking that are cooperation-conducive,
and though these rules arelets stipulaterules that would remain inter-
nalized through an indefinite sequence of further progressive (in Kitchers
sense) transitions, our acting in accordance with those rules might lead us
astray ethically. One obvious way to concede that Kitchers answer to the
causal question is correct, but then reject his account of justification, is by
accepting some kind of noncognitivism.
Consider once again ethical relations between men and women. Perhaps
Inotice something peculiar about myself:Ifind that when Ireflect on rela-
tions between men and women, Ican see no ethically relevant distinctions
between them; women are, Isay, the equals of men simpliciter. But Ialso
notice that the rules we have all internalized and live by, men and women
alike, myself included, lead us to treat women as equals only to the extent
required to elicit their cooperation. Itherefore notice that there is a cooper-
atively stable future in which Iam able to retain my relatively higher posi-
tion in the power hierarchy, and so control the extent to which Igive up
benefits so that additional benefits can flow to women, notwithstanding
my ethical views. With all of this understood, what could Ibe doing when
Ijudge that women are the equals of men simpliciter?
The only answer Kitcher can give is that I am expressing a false ethi-
cal belief. Noncognitivists, by contrast, can say that I am expressing my
endorsement of a norm that requires me to treat women as equals sim-
pliciter. In other words, I am condemning the rules that I have internal-
ized and recommending the internalization of alternative rules, rules that
would lead me to treat women as the equals of men simpliciter. I could
well understand that the transformation Iendorse would lead to a rockier
future, one in which willing cooperation between the sexes may never be
fully forthcoming. Given our stipulation, Icould also have good reason to
believe that that transformation will never come about. But I could still
coherently endorse the transformation, as Icould think that that future,
rocky though it would be, would be a better future than one in which men
and women stably cooperate but only because they dont act on their ethi-
cal belief that women are the equals of men simpliciter.
On this noncognitivist way of thinking, the mere fact that men and
women would still stably cooperate without women being treated as the
equals of men simpliciter puts no pressure on us to suppose that women
are not really the equals of men simpliciter. Similarly, to switch examples,

Function and Truth inEthics [261]

the mere fact that people would stably cooperate with each other without
their internalizing a rule requiring them to relieve the pain of both human
and nonhuman animals whenever they are able to do so puts no pressure
on us to suppose that that rule doesnt capture the ethical truth of the
matter. Nor should it be thought that, as noncognitivists, we would have
to shy away from talking of ethical truth. Allan Gibbard (1990), among
others, has shown how and why we would still treat ethical judgments as
truth-apt even if we were to deny, as many noncognitivists do, that such
judgments are truth-apt in any robust sense. Given that Kitcher himself
seems to have ethical views about the treatment of women, nonhuman ani-
mals, and homosexuals, views that are at some distance from the ethical
truth about these matters; given his pragmatic naturalist account of ethical
truth; and given that these views are more intrinsically plausible than the
views that that account of ethical truth commits him to, the question is
why he doesnt embrace some form of noncognitivism.
Chrysostomos Mantzavinos (2012, 38)asks Kitcher this question in his
commentary on The Ethical Project. Here is Kitchers (2012a, 179)reply:

For non-cognitivism to succeed, however, would require an account of progress

for the emotions taken to be expressedan account of what is meant by saying
that the emotions elicited at later stages are more apt than those appearing at
earlier times. Despite my admiration for Allan Gibbards work, Icannot find any
such account in his writings, nor in those of any other non-cognitivist Iknow.
So, if the challenge is to be met, some serious work will have to bedone.

But once we have distinguished questions of causal explanation from ques-

tions of justification, Kitchers demand that the noncognitivist tells us
what it is for an emotion to be apt can be seen to be ambiguous.
If progress is understood in terms of the removal of impediments to
ongoing cooperation, then an apt emotion is simply one that grounds
our internalization of rules that would be internalized in any transfor-
mation of the rules whose internalization would similarly result in ongo-
ing cooperation. The question thus concerns causal explanation, and the
noncognitivist should borrow Kitchers answer. Importantly, however, the
noncognitivist is not committed to endorsing such ongoing cooperation
remember again our discussion of ethical relations between men and
women. But if progress is understood in terms of aligning our ethical
commitments with the ethical truth, then an apt emotion is simply
one whose expression would result in our making true ethical claims. The
question concerns justification, and the noncognitivist should answer it
in his own terms, something he would presumably do by expressing his

[262] The Philosophy of Philip Kitcher

endorsement of emotions whose expression would result in ethical claims
that are much the same as the ethical claims he makes when he expresses
the emotions he in facthas.
The upshot is that there is no challenge here that the noncognitivist
lacks the resources to meet. It therefore seems to me that Kitcher should
abandon his account of ethical truth and open himself up to the possibility
of embracing some form of noncognitivism. Doing so would allow him to
argue in favor of our internalizing rules without regard to whether or not
the internalization of those rules would result in stable cooperation. Given
that stable cooperation might itself depend on the maintenance of unjus-
tifiable discrepancies in power relations, this is surely just as it shouldbe.



Embracing noncognitivism isnt the only way someone could accept

Kitchers account of the function of ethical practice while rejecting his
pragmatic naturalist account of ethical truth. Another alternative would be
to embrace an account of the truth of ethical claims in the spirit of Kant.
Kitcher (2011a, 27273) himself is, however, no fan of Kantian approaches:

Consider some approaches found in or inspired by Kant. Ethics expresses the

requirements of pure practical reason:to deny the moral law reason generates is
to fall into a mode of irrationality, in which one contradicts oneself. Or it might
be said:ethics consists in a set of principles ideally rational agents would agree
to under ideal circumstances, so failing to abide by its precepts is to violate con-
ditions of rationality. Tough-minded skeptics would hardly be brought to silence
by either of these dicta or by any plausible emendations of them. The skeptic
speaks:You can call the procedures you use to generate the rules you favor pure
practical reason, if you like, and suppose those who dont go along with them
are involved in some sort of contradiction, but the mere label doesnt frighten
me, and the effects you envisage dont appear particularly dreadful. If Ireject
these rules, Iam hardly doing something similar to asserting a statement and
its negationand even if Iwere, its not obvious anything Icare about would
be compromised by doing so. The history of science contains episodes in which
people have worked quite well with internally inconsistent ideas (think of the
Bohr model of the atom), and Ihave no reason to think my practical irrational-
ity will pose difficulties for me. Nor am Imuch moved by the thought of rules
hypothetically ideal people (supposedly better than me) in some fictional situ-
ation would agree to. Why should Ibe bound by what they would decide? There

Function and Truth inEthics [263]

is no argument for thinking the purposes I care about would be ill served by
flouting any such precepts.
. . . Maybe, though, the skeptics answers have exposed something wrong
with him:he has failed to meet ideal conditions of rationality. If that is so, the
criterion for a successful reply to the skeptic is modified. He does not have to be
silenced; one must merely have an account of why his responses are problematic.
Given this understanding, however, pragmatic naturalism can do just as well
as the allegedly superior nonnaturalistic approaches. Where Kantians and con-
tractarians see failures of ideal rationality, pragmatic naturalism diagnoses an
inability to appreciate how central the ethical project is to humanlife.

How should Kantians respond? The first thing Kantians should say is,
Ihope, apparent from the earlier discussion.
According to Kitcher, the Kantians response to the skeptic reduces
to the claim that there is something wrong with him, and this means
that the Kantian and the pragmatic naturalist are on an equal footing as
regards the skeptic. They agree that there is something wrong with him,
but disagree about what that thing is:the Kantian says that the skeptic is
practically irrational, whereas the pragmatic naturalist says that he has an
inability to appreciate how central the ethical project is to human life. The
issue should therefore be decided by figuring out whose account of whats
wrong with the skeptic is better supported by the evidence. But as the
earlier discussion made clear, there isnt obviously anything wrong with
someones having the inability that the pragmatic naturalist identifies. To
be sure, the ethical project, as Kitcher describes it, is central to human life
to the extent that it sets the terms for stable cooperation. But given that
stable cooperation might itself be ethically problematicremember again
our discussion of what might underwrite the stable cooperation we find
between men and womenthat doesnt count in favor of anyones appreci-
ating that kind of cooperation, in the sense of their being disposed to bring
it about. We should therefore be skeptical of the ethical project, insofar as
we agree that it has the function Kitcher identifies.
This leaves us with Kitchers disparaging remarks about Kantian
approaches at the very beginning of the passage. Kitchers skeptic says,
You can call the procedures you use to generate the rules you favor pure
practical reason, if you like, and suppose those who dont go along with
them are involved in some sort of contradiction, but the mere label doesnt
frighten me, and the effects you envisage dont appear particularly dread-
ful. That isnt just the skeptic talkingthats Kitcher talking. As Iunder-
stand it, he is saying that the Kantian doesnt have a plausible story to tell
about what the practical irrationality of the skeptic consists in. The Kantian

[264] The Philosophy of Philip Kitcher

claims that the skeptic is practically irrational, but this claim is pure bluff.
The question is whether anything can be said to convince Kitcher that he is
wrong aboutthis.
The Kantians idea, at least as Iunderstand it, is that we can derive a
substantive account of the reasons agents have, where the substance of
those reasons is recognizably ethical, from the concept of an agent. Such a
derivation is possible, it seems to me, if the following premises are alltrue:

1. That agents qua agents have the function of being desire realizers and
knowledge acquirers (analytic truth about agents).
2. That it follows from (1)that ideal agentsthat is, agents who perform
their function optimallymust have and exercise the capacity to realize
their desires no matter what their content and know what the world is
like, no matter what it is like, at least insofar as the worlds being the
way that it is bears on their realizing their desires (the modal conception
of the ideality of a functionalkind).
3. That it follows from the potential for conflict in the optimal possession
and exercise of these two capacitiesthink of the conflict present in an
otherwise ideal agent who desires to believe that p whether or not it is
truethat ideal agents must have certain dominant conflict-resolving
desires: specifically ideal agents must have a dominant desire not to
interfere with their exercise of their capacities to know their world or
realize their desires in it (on condition that those desires wouldnt lead
them to interfere), whether in the present or the future, and they must
also desire to do what they can to help ensure that they have the capaci-
ties to know their world and realize their desires in it, whether in the
present or the future (novel premise whose truth depends on (2), the
fact that agents are temporally extended, and a painstaking examina-
tion and rejection of alternative ways in which the conflict identified
might be resolved).
4. That there is no consistent way for an ideal agent to restrict his desires
to help and not interfere to the current and future stages of himself, and
hence that ideal agents have expanded versions of these two dominant
desires ranging over all agents:that is, they desire to help everyone and
not to interfere with anyone (a premise inspired by Nagel [1970] and
Parfit [1984] that teases out the implications for the theory of practi-
cal rationality of the fact that each agent is [potentially] just one agent
5. That there is an analytic tie between facts about an agents reasons for
action and facts about which of that agents options his ideal counter-
part has desires for (analytic truth about reasons for action).

Function and Truth inEthics [265]

Though this is not the place to argue that these premises are all true,
Ishould say that Ihave argued for their truth elsewhere (M. Smith 1994,
2011, 2012, 2013). What is crucial for present purposes, however, is not
so much the truth of these premises as a conditional claim. If (1)through
(5) are all true, then all agents have dominant reasons to help and not
interfere, which are substantive reasons for action with recognizably ethi-
cal content, and those agents who fail to help, or who interfere, act in ways
that they have dominant reasons not to act, which is by all accounts one
way an agent could be practically irrational. Though Kantian accounts of
the content of ethical claims and their truth might still be mistakenone
or more of premises (1)through (5)might be falseit thus seems to me
that a plausible story can be told about why the skeptic is practically irratio-
nal. The Kantians claim that the skeptic is practically irrational thus isnt
What should Kantians say about Kitchers account of the function of
ethical practice? Kantians can agree with Kitcher that the function sup-
ported by the practices origins in the behavior of our hominid ancestors is
to preempt and remedy altruism failures that interfere with cooperation;
they can agree that ethical practice achieves this by getting individuals to
internalize certain rules; and they can agree that there are therefore certain
rules whose internalization would, as a matter of empirical fact, lead people
to stably cooperate. But Kantians should insist that it is an open empiri-
cal question whether people have reasons to perform the acts that would
result from their internalization of these rules, even if their internalization
would lead them to stably cooperate. They should say that that depends
on whether the scheme of cooperation is one in which people are helping
and not interfering to the extent that they can. If it is such a scheme, then
they should insist that there is a happy coincidence. But if it isnt, then the
Kantians, like the noncognitivists, should insist that we have good reasons
to reject such a stable scheme of cooperation in favor of a scheme in which
people do help and dont interfere to the extent that theycan.
Here once again we see the relevance of Srinivasans distinction between
questions of causal explanation and questions of justification. If our focus is
on purely causal questions, then there is nothing wrong with supposing that
the function of ethical practice is the function Kitcher identifies. But if we
switch to questions of justification, then, at least according to the version of
the Kantian approach that I myself prefer, we should suppose that the func-
tion of ethical practice should be to get people to believe and behave in accor-
dance with the ethical truth as spelled out in (1) through (5). Alternatively
put, insofar as we think that our function is desire realization and knowledge-
acquisition, and that it follows from this that our function is to be sensitive to

[266] The Philosophy of Philip Kitcher

reasons, we should suppose that the function of ethical practice is to get at
the truth about what the reasons are, and then to get people to behave in
accordance with that truth.

Reply toSmith

Occasionally a critic or commentator finds a way to formulate an issue that

enables the author whose work is under discussion to see the need to clar-
ify a central claim or theme. Even more rarely, if the critic or commentator
is able to connect with other reactions to the focal work, the author can
recognize what other critics have been driving at. Michael Smiths appraisal
of The Ethical Project helps me in both these ways, and I am immensely
grateful tohim.
Smith offers three main objections, of which the first is the most funda-
mental. The Ethical Project views ethical practice as a collective human con-
struction, permanently unfinished, constantly adapting to our evolving
social environments, and pervasively constrained by the fact that we are
social organisms whose evolved psychological capacities do not enable us
smoothly to pursue the kind of social life we have inherited. The objectiv-
ity of ethics, on my account, lies in the fact that our construction of ways
of living and of living together must adjust to that pervasive constraint.
As I put it, the original function of ethical practice was to remedy altru-
ism failures. I would now put it slightly differently: the original function
of ethical practice was to overcome the limits of our responsiveness to
others. (Here I have been helped by Sterelny 2012; Tomasello 2009, 2016.)
The notion of function deployed here is one Ihad introduced in Kitcher
(1993b). Functions are seen in terms of overcoming problems, and those
problems are embedded in problem backgrounds. Viewing ethics as a social
technology, Irecognized that achievements in overcoming some problems
typically generate new problems. In my motivating analogy, as you intro-
duce vehicles that enable you to travel to new places or to travel to previous
destinations more easily and speedily, you must find ways of regulating
traffic, training users, and so on and on. The Ethical Project worries about
the problem of functional conflict. As a series of problems accumulates,
spun off from the original enterprise, how do we weigh the relative impor-
tance of those problems? Might we discover, in the end, that the original
problem no longer matters? Idevote much attention to arguing that the
original function of ethics is still central to the ethical project, and that

Function and Truth inEthics [267]

consequently we can use the criterion of overcoming the limits of our
responsiveness to others as a measure of ethical progress.
Smiths central insight is that the difficulty is deeper than the book
appreciates. Its not simply that an uncontroversial original function (rem-
edying limited responsiveness) might conflict with other functions gener-
ated during the course of the project, but that my thesis about the original
function is itself open to dispute. The Ethical Project consistently talks
about progress in terms of remedying altruism failures, but, in Smiths
view, Im committed to supposing that the mark of progress is remedying-
altruism-failures-insofar-as-that-promotes-cooperation. Substituting his
preferred identification of the original problem generates, as he sees, some
very unpalatable consequences.
Now, of course, limited responsiveness showed up in the lives of our
ancestors as social friction (requiring them to engage in time-consuming
peacekeeping and preventing them from profitable cooperative ventures).
I saw the social friction and the reduced cooperation as symptoms of an
underlying disease. To cope with the predicament is to tackle the disease,
not simply to remove those symptoms that have manifested themselves so
far. So Iclaimed that the original function of ethics was to remedy altruism
failures (or to overcome the limits of our responsiveness to others). Smith
challenges the aptness of this medical analogy.
Heres a blunt reply to the challenge: From Harvey on, people have
declared that the function of the heart is to pump the blood, not to pump-
the-blood-insofar-as-its-necessary-for-maintaining-bodily-organs. The dec
laration persists even when we recognize that a mutant heart, pumping
more efficiently, might be selected against because the developmental
resources used in building it might be deployed to promote greater reproduc-
tive successthe superior pump is overkill. So Imight accuse Smith of abus-
ing normal and scientific talk of functions.
But this is too blunt. Smith is correct, I believe, in taking my char-
acterization of the problem background to need more defense than
I have given. What would justify identifying the problem as one of rem-
edying altruism failures rather than remedying-altruism-failures-insofar-
The identification is something achieved by the analyst who reflects on
the history of the ethical project, striving to understand what might count
as progress in that project. As Ifrequently note, the thought that they were
achieving partial relief from altruism failures was not something avail-
able to the pioneers who began ethical practice, nor to their successors.
Only when we have some narrative that specifies the starting point and

[268] The Philosophy of Philip Kitcher

some parts of the route leading from pre-ethical life to the present is there
any genuine possibility of reaching a considered view about what ethical
progress might be. The Ethical Project attempts to show that, given a narra-
tive informed by the fruits of various fields of inquiry (including, but not
restricted to, the natural sciences), it is hard to take progress to consist in
the recognition of prior ethical truths. Instead it proposes to view progress
as problem solving and (in a Peircean vein) to understand truth as what
stably emerges when you make progress.
Its at this point that the problem uncovered by Smith arises. There
are two rival descriptions of the original problem, and Smiths objection
should not be that Iam committed to the one he prefers (and that, as he
shows, leads to repugnant conclusions) but rather that theres no basis for
preferring the characterization Iconstantly use (remedying altruism fail-
ures) to his qualified version. Yet it seems to me that we do have a basis.
As we reflect on the history of the ethical project, its evident that our
failure to respond to one another takes many different forms, from the
ancestral unwillingness to share scarce resources with familiar band mem-
bers to the current obliviousness to the plight of people distant from us in
space or time. Theres ample evidence for the conclusion that the ways our
limited responsiveness causes trouble can be highly various and that we
cant specify in advance all the symptoms that might express the under-
lying disease. Yet, Smith might suggest, without knowing the details of
the manifestations we can specify the generic characteristics of the syn-
drome:all the pathological effects involve interference with cooperation.
Hence, he would claim, his preferred characterization of the problem is at
least as well justified asmine.
I disagree. Just as doctors typically focus on the underlying condition, so
should the reflective meta-ethicist. In medical practice attempts to circum-
scribe the unfortunate effects of a malformation or infection would seem
unnecessary and unwarranted; so too with our limited responsiveness to
others. On what basis can the reflective analyst judge that all the effects
of our tendency not to respond can be counted in costs to cooperation?
True, as Isuggest, the first effects, the ones that alerted our ancestors to
the existence of trouble, were expressed in cooperative failures. Yet it is also
an important part of my story that the ethical project has given rise to new
emotions and new conceptions of ourselves, and that failure to attend to the
limitations of our responsiveness can have an adverse effect on these, quite
independently of any considerations about cooperation. Indeed the very facts
Smith cites to expose the inadequacies of my view, when construed according to his
preferred characterization of the original function of ethics, reveal that thisisso.

Function and Truth inEthics [269]

I maintain, then, that a reflective analyst, surveying the history of the
ethical project, should prefer my characterization of the original problem
(and, derivatively, my account of progress) to the hedged version to which
Smith claims Im committed. But to see clearly why this is so, its useful
to take up some of his other points:the suggestion that noncognitivism
would better serve my purposes, the concerns about possible conflation
of causation and justification, and the charge that Ihave underrated the
resources of Kantianism. Ultimately well see how the differences between
us may turn on rival conceptions of philosophical method.
Smith chides me for not having offered a more adequate reply to Chrys
Mantzavinoss question about noncognitivism. He is right to do so. Iwor-
ried that noncognitivists can provide no sense of ethical progress. They
would need apparently to find some criterion for declaring that the emo-
tions expressed in judgments made after the progressive transition are
more apt than those previously endorsed. Smith contends that aptness can
be understood in terms of a propensity for yielding true ethical judgments.
But what sense of truth can the noncognitivist provide? Smith supposes
that Allan Gibbard (1990) provides a sense (albeit not a robust one). Here
Iremain unconvinced. Gibbard concludes that normative judgment mim-
ics the search for truth (218), but Ido not see how his complex account of
the endorsement of norms generates an applicable criterion for use in the
cases of ethical change that concern me. Smiths sketch of an accounta
noncognitivist endorses emotions whose expression would result in ethi-
cal claims that are much the same as the ethical claims he makes when he
expresses the emotions he in fact haslooks like a recipe for conserva-
tism. From where you now stand no modification is justified.
I now think noncognitivists can do better than Ionce thought, and bet-
ter than the position Smith assigns them. Noncognitivists should recognize
that the emotional responses they make, and even the emotions available
to them, are dependent on the environments in which they develop and
live. Apt emotions are those developing and expressed in privileged envi-
ronments, and the appropriate privilege is identified by my account of ethi-
cal method:they are the emotions generated after informed and mutually
engaged interaction with the perspectives of all others, equally informed
and mutually engaged. Justified ethical claims are those that flow from
such emotions. True ethical claims are those expressed in apt emotional
responses that continue to remain stable, as the community continues to
participate in comprehensive, informed, mutually engaged interaction.
So noncognitivism can mimic my approach to ethical truth and ethical

[270] The Philosophy of Philip Kitcher

This possibility is worth noting because it helps us to recognize what is
occurring when my reflective analyst specifies the original function of eth-
ics. Suppose the analyst contemplates the possibility that the measure of
ethical progress is remedying altruism failures insofar as they interfere with
cooperation, subjecting them to the kinds of scenarios that rightly worry
Smith (e.g., stable cooperation between the sexes without any agreement
to equality). Based on her emotional reaction, the analyst will reject this
account of ethical progressperhaps in favor of the one Iprefer (remedying
altruism failures, period). Adding a dash of noncognitivism at exactly this
point seems to help my account. Moreover the analyst finds that, in mutu-
ally engaged informed discussion with others, the assessment of this emo-
tional response as apt survives. So the analyst is justified in the judgment.
The suspicion that Iconfuse questions of justification with causal ques-
tions, voiced both by Smith and by Amia Srinivasan, is readily understand-
able, given that I spend considerable time providing causal accounts of
the evolution of ethical practice and I endorse a causal standard for jus-
tification. In some domains (science, common factual judgments) Ithink
of truth in correspondence terms; in others (ethics, mathematics) Iadopt
a Peircean approach. Across the board the basis of my approach is that
beliefs are justified when they are generated and sustained by causal pro-
cesses of types that have a propensity to generate and sustain true beliefs.
(This needs to be qualified and refined in various respects.) Idont suppose
that feminism is justified because it turns out that taking women to be
equal doesnt cause too much trouble. Rather the justification is that mutu-
ally engaged, informed representatives of a wide range of positions would
endorse equality (although if the empirical facts about family life were dif-
ferent, they might arrive at different ways of expressing their commitment
to equality). That counts as justification because the most reliable way of
arriving at true beliefs of this kind is to fashion them through comprehen-
sive, informed, mutually engaged deliberation.
In effect this is deeply Kantian, pledged to a thoroughly naturalistic
reading of the underrated third formulation of the categorical imperative.
The obvious solution to the problem prompting the ethical project is to
begin a conversation in which the involved parties (originally the members
of a small hunter-gatherer band) exchange their claims. Analysts, reflecting
on that problem and its solution, are likely to find Jamess thesis about the
importance of all claims and all claimants and Deweys democratic com-
mitment to equal standing plausible as generalizations of this (historically
original) position. So they achieve a kind of Kantian perspective, seeing
the collective construction of ethics as a project in which almost all human

Function and Truth inEthics [271]

beings act as legislators on equal terms. (Sterelny [2012] is not wrong to see
me as sounding some Kantian themes.)
But Smith is also right when he claims that Igive Kant short shrift. Iam
impatient with simple declarations that lapsing from morality is falling into
practical inconsistency. Smith reminds us that Kantians want to say more,
and his five premises attempt to give substance to the practical rationality
at issue. Yet Iremain unconvinced. All his principles involve concepts that
give me pausethese are simply not my words. More specifically Ienvisage
an agent whose assignment of weights to his various desires violates one
or more of these principles:the itch in his finger is vastly more important
than anything else. The effort to rule him out by asserting analytic truths,
a special sort of ideal agent, a necessarily dominant desire, consistent
restrictions, and analytic ties that link agents to their ideal counterparts
strikes me as dubious force majeure.
I offer something much simpler to the same end:Reflect on the ethical
project. Realize that you are the kind of being you are as the result of a
complex history involving many actors, and that your life is bound up with
the lives of others who are similarly heirs to the past. We have inherited
structures and roles and institutions and norms and face the task of how to
adjust them to our circumstances. As we reflect we see the pervasive fact of
our limited responsiveness to others, the difficulties to which it has given
rise, and some partial attempts to come to terms with the fundamental
condition. What better way of going on do we have than that of attempt-
ing to interact in informed, comprehensive, mutually engaged discussion?
Amia Srinivasan (2012) concludes her review of The Ethical Project by
suggesting that, despite her reservations about several aspects of my posi-
tion, she finds the idea of broadly collective deliberation I recommend
attractive. Iam encouraged by this. As Lorraine Daston sees, history has
always been important to my thinking about philosophical issues:Ibelieve
that we often see how to go on by understanding the trajectory that has
led us to where we are. So, in thinking about ethics or in addressing ethical
skepticism, Idont start by searching for first principleslike Smiths five
premisesthat might settle the issue. Instead Iurge synthetic reflection
on what we know, including the history of our practices, and interchange
of ideas across informed and mutually engaged perspectives. This sepa-
rates me from the many Anglophone philosophers for whom the key is to
use intuitions to ground sweeping generalizations. Perhaps if Ihad been
more explicit about my approach, The Ethical Project might have been more
obviously heterodoxor more attractive?

[272] The Philosophy of Philip Kitcher


What toDo While Religions Evolve

beforeOur VeryEyes

T he term New Atheists has recently emerged as a way to refer to the

explosion into public visibility of a group of atheist critics of religion,
distinguished not just by their outspokenness but by their success in com-
manding attention in the media. The books by Richard Dawkins (2006),
Sam Harris (2004, 2006), Christopher Hitchens (2007), and me (2006)
were among the first of a still-growing list, and someone christened us the
Four Horsemen, a sobriquet that seems to have gone to fixation. (We are
not, as Kitcher [2011c] puts it in Militant Modern Atheism, self-styled
horsemen, but we have acquiesced in the term.) In spite of our differences
in topics, tactics, and manners, we have often been lumped together as a
convenient target for condemnation by the legions of defenders of religion.
In addition to the religious spokespeople and churchgoers who have risen
up to deplore our rude assault on religions are the critics we Horsemen
refer to as the Im an atheist but crowd. Religion is not for them, but in
deference to those who love it and need it, they abstain from criticizing
it and accuse us militants of the secular sin of intolerance. While Kitcher
expresses sympathy for many of the claims of the New Atheists, he (Kitcher
2011c, 2) also agrees with the critics that valuable options are being fore-
closed. Militant Modern Atheism stands out from the rest of the Im
an atheist but literature by attempting to articulate some of the positive
details of ways religions might evolve harmlessly under the protective
mantle of tolerance.
This is important, because it helps to expose a systemic weakness in the
sacrosanct ideal of Freedom of Religion, which we might call, with a bow to
Senator Daniel Patrick Moynihan, benign neglect. Apolicy of maximizing
tolerance for religious diversity can work smoothly only if we are careful
not to scrutinize the boundaries too closely, and Kitchers tightrope act
wobbles tellingly between the philosophical ideal of explicitness and the
theological necessity of smoke and mirrors. Consider the pressing ques-
tions we are wise to postpone, if possible:Exactly which religious practices
are beyond the pale? Human sacrifice for sure, animal sacrifice less cer-
tainly. Is clitoridectomy genital mutilation and hence intolerable, and if so,
is the circumcision of males, at least in infancy or childhood, to be banned
as well? Is it tolerable for a religious sect to deny reading and writing to its
(female) children, or to replace the treatments of modern medicine with
These are profound questions about how to live a human life, and the
principle of religious freedom declares that answering such questions is
so important that it should be left to individuals to the greatest extent
possible, not dictated by the state, not even by majority rule of the citi-
zens. Following that policy to the maximum, however, could in principle
undo the states primary function of protecting its citizenry from harm,
licensing all the crimes it otherwise forbids, under the exemption of reli-
gious freedom. Where does tolerance slide over the cliff edge into anarchic
hypertolerance? Any attempt to legislate an answer would itself be a viola-
tion of religious freedom, it seems. Cant we just depend on the goodwill
of all parties, counting on religious groups to moderate their ways in def-
erence to the sensibilities of others? No, obviously, we cannot. There are
today, and have been for millennia, religious sects and even major religions
that endorse or even enjoin practices that are widely held by others to be
criminal offenses. We may all individually think we know which are the
punishable crimes and which are the permissible practices, but we have
to recognize that there are many thoughtful, serious people who disagree
with us. We would like to use gentle, reasonable persuasion to create a con-
sensus on these fraught issues, but we must also recognize that there is
such a thing as being too diplomatic.
The ethical questions merge with strategic questions:How much pres-
sure is apt to be effective in bringing about the sought-for changes? If
reform at gunpoint worked, there might be occasions to revert to it, and
in fact it has worked in the past. It is worth reminding ourselves that racial
integration was enforced in the United States with the help of thousands
of armed troops, and few today would argue that it would have been wiser
to forgo the troops and rely on diplomacyalone.

[274] The Philosophy of Philip Kitcher

This uneasy balance between principle and strategy is the background
that frames the disagreement between Kitcher and me. We are both athe-
ists who disagree about which truths it is wise to broadcast, a difference
of opinion about strategy but not so stark as whether or not to use force.
Neither Inor any other of the New Atheists has called for armed suppres-
sion of any religious doctrines, so far as Iknow. We are content to try to
achieve our ends using political means:swaying public opinion with forth-
right verbal criticism, and even condemnation on occasion. That is the
extent of our aggression. But Kitcher still thinks it is counterproductive,
and instead of just blasting us for our rudeness the way most of the Im an
atheist but crowd do, he attempts to demonstrate an alternative path that
he expects will have more humane results.
Lets consider his proposal:replacing the belief model of religion with
the orientation model. The belief model is the more or less standard sup-
position that individual religions are distinguished by their different doc-
trines, and that to be committed to a particular religion is to believe the
doctrines constitutive of that religion (Kitcher 2011c, 3). According to
this view, coreligionists share values, aspirations, attitudes, and loyalties
because they share belief in a set of propositions, the defining creed of their
religion. The proposed orientation model inverts this dependence: the
shared values and loyalties are an orientation definitive of a religion, with
the characteristic beliefs tagging along as a typical effect, not the core or
essence of the religion.
The idea is that religious creeds are no longer the protective shields
they once were, uniting the brethren in a shared litany of proprietary
and unquestioned truth; they have become part of the problem, and not
a small part. If religions could just find a way of quietly jettisoning their
creeds while preserving their traditions, rituals, art, and music and honor-
ing the symbolic and historic value of their holy texts, they would actually
shed the source of their greatest vulnerability: the all-but-demonstrable
falsehood of most of the cosmology and history they have heretofore felt
obliged to profess in defiance of evidence and rationality. Some religions
some denominations or at least some congregationsalready occupy this
enlightened niche, so we know it is a possible stable outcome, but are there
steps we can take to get others to transform themselves? If the relentless
harsh light shone by the New Atheists on traditional believers just hardens
their resistance and enmity, is there a kinder, gentler way of fostering the
change of perspective?
There is a tactic that all of us recognize, some of us deplore, and some
of us employ. You would think it would be a matter of considerable con-
troversy, but it is seldom discussed, since even raising the issue tends to

W h at t o D o W h i l e R e l i g i o n s E volv e b e f or e O u r V e r y E y e s [275]
preempt the controversy by taking one side. This is the tactic that dare not
speak its name; it belongs to a family of tactics that spread in a continuum
from indefensible (lying for Christ is an example) to indispensable (respect-
ing the privacy and feelings of others to the greatest extent possible). Some
relatively benign points on this continuum can be justly called diplomacy or
tact, and other points, over near the other extreme, should be called what
they are: intellectual dishonesty at its most culpable. One of the clearest
results of my ongoing research with Linda LaScola on the plight of nonbe-
lieving clergy (Dennett and LaScola 2010, 2013, 2015) is the recognition
that the slippery slope between these extremes has no boundaries to which
one may cling in a principled way. What begins as tact can sour into dissem-
bling that shatters the integrity and blights the lives of many clergy. I have
coined a deliberately mildcomically mildterm for the middle ground,
faith fibbing, which is not so harsh that people find it impossible to confess
to, but not so benign that we can all publicly advocate it as a policy without
undermining it in the process. No one would occupy a pulpit adorned with
the declaration Faith Fibbing Practiced Here.
I went to some lengths in Breaking the Spell (Dennett 2006) to distin-
guish two spells one might consider breaking: the taboo against looking
too closely at religion, holding it up to the same harsh light of rational
probing to which we subject all other important phenomena, and the spell
of religion itself. In my book I declared my intention to break the first spell
and my agnosticism about the wisdom of breaking the secondciting the
very considerations that Kitcher advances more positively. Kitcher ignores
my distinction but in fact is in nearly perfect harmony with my positions
on them. His essay is an example of breaking the first spell: he writes with
unflinching candor about the shaky status of any religion adopted on what
he calls the belief model and uses that spell-broken perspective to look hard
at the prospects for keeping the second spell unbroken by relying on what
he calls the orientation model, supposing that this is perhaps the only sur-
viving mode of religion that can provide the benefits he wants to preserve,
which may just be a necessity of meaningful life for many people. As I noted
in my book, there is a reasonable fear that breaking the first spell will inevi-
tably break the second as well, which fear is the obligatorily tacit standard
justification for not breaking the first. Kitcher vividly illustrates that prob-
lem in his essay, trying to split the difference between being patronizing
on the one hand and uneasy complicity with unacceptable nonsense on the
The point of Kitchers introduction of the orientation model is to give
him a way of reversingmost of the timethe otherwise standard depen-
dence of serious commitments and aspirations on grounded beliefs. The

[276] The Philosophy of Philip Kitcher

orientation-type religionists put commitments first, as the fundamental
landmarks of their lives, and let the expression of (what take the place
of) grounds for these commitments wander somewhat opportunistically
between mythically self-conscious metaphor at one pole and doctrinal
entanglement (flirting with the belief model) at the other, with convenient
vagueness in the middle. (The doctrinally indefinite folks take refuge in
language that is resonant and opaque, metaphorical and poetic, and deny
that they can do any better at explaining the beliefs they profess [Kitcher
2011c, 6].) Whatever floats your boat, as one says. And indeed, if maintain-
ing a religious orientation is the only way for you to have a meaningful
life, you should rely on whatever floats your boat. But then it will just make
matters harder for you if you have to confront Kitchers trio. Tell me, sir or
madam, have you decided to go with mythic self-consciousness, doctrinal
entanglement, or doctrinal indefiniteness? Dont ask! Dont tell! Thats why
many think the first spell should not be broken, but Kitcher and I have both
ignored that admonition.
Kitcher (2011c, 6) is at pains to express his defense of these delicate
options sympathetically: Ill suggest that doctrinal indefiniteness can be
a reasonable expression of epistemic modesty, and that even doctrinal
entanglement can be justified when it is the only way of preserving, in the
socio-cultural environment available, a reflectively stable orientation. But
a somewhat less diplomatic version hovers in the background: Kid yourself
if you have to.
Epistemic modesty sounds like an exemplary character trait, but we
should note that it has its dark side. In many religious settings it serves
handily to unnerve and disarm those who otherwise might be brave
enough to observe that the emperor has no clothes:think twice before you
challenge what the elders say; what do you know? A nice feature of this
ambient attitude is that it can be iterated in a pinch. Pastors can hold off
skeptical challenges by subtly impugning the comprehension, the experi-
ence, the sophistication of a lay challenger, and then, if that doesnt do the
trick, confess their own abject incomprehension in the face of mystery. Not
Who are you to question these sacred and difficult truths? but Who are
we to question these sacred and difficult truths?
One of the troublesome aspects of the books by the Four Horsemen is
that we dont bother wrestling with the intricate arguments and analyses
of theologians, and we dont back off; we dont accept the modest role that
self-styled religious sophisticates try to impose on us. This arrogance of
the New Atheists is often criticized, but, as Inever tire of responding:we,
like scientists and philosophers generally, are forever asking ourselves
But what if Im wrong?a reflection that is not just rare but positively

W h at t o D o W h i l e R e l i g i o n s E volv e b e f or e O u r V e r y E y e s [277]
discouraged among religious spokespeople. There are some famously arro-
gant Nobel Laureates in the scientific community, but Ihave never encoun-
tered one who can hold a candle to the overweening overconfidence, the
smug certainty of the typical self-righteous defender of religion. Their dis-
dain for evidence-seeking and careful argumentation is often breathtak-
ing, amounting to self-disqualification for the role of rational discussant.
Those who view it as positively immoral to entertain alternatives to, or
even objections to, their faith place themselves outside the marketplace
of ideas, incompetent to participate in the serious political conversations
that we ought to be engaging in today. And one of the effects of fostering
epistemic modesty in all matters religious, as Kitcher recommends, is pro-
tecting the social niche in which these influential subverters of reason can
operate largely unchallenged.
I have been asking defenders of sophisticated theology for a recom-
mended reading list, for works they are prepared to defend as intellectually
bracing and honest, and have yet to have my challenge met. Iam tempted
to draw the conclusion that, on closer examination, they recognize that
they have indeed been adopting a double standard and letting pass as deep
thought work that is actually just obscure, and apparently often deliber-
ately obscure. These works do serve a useful purpose for the adopters of
epistemological modesty who can reason as follows:These professors are pro-
fessional thinkers about religion; they are still in the church, so they must have
gone way beyond me in thinking these issues through. I dont get it, but they
do, so I should accept their authority. In other words, these works provide
examples of high-flown rumination that one can confess to finding some-
what incomprehensible but nevertheless deem inspiring and authoritative.
Go read the meticulous arguments of this thinker; they should sweep away your
doubts. (And if they dont, it must be your fault.)
Kitcher (2011c, 9)is well aware of the risks entailed by adoption of the
orientation model, and Im happy to say that his defense of the three con-
veniently smudged alternatives firmly draws the line at letting any of these
options abrogate a commitment to reason when deciding ethical mat-
ters:Someone who makes decisions affecting the lives of others is ethically
required to rely on those propositions best supported by the evidence.
This requirement runs into awkward interference from the orientation
model, however, as Kitcher reveals in discussing one of the central problems
with the Abrahamic religions (Judaism, Christianity, and Islam):Abraham!
Abraham is called upon by God to perform an unthinkable act (to commit an
unspeakable crime, bluntly put). Is this any kind of inspiring example? In a
felicitous phrase Kitcher (2011c, 8)notes that there ought to be no teleo-
logical suspension of the ethical, but then just what is the positive role of

[278] The Philosophy of Philip Kitcher

Abraham as a knight of faith supposed to be, when he so clearly violates
this principle? Kitcher (2011c, 8)says that Abrahams sort of trust is not
legitimateyou cant put it much plainer than thatbut how, then, does
he find a way of endorsing Abrahams (mythical) story as any sort of talisman
for a meaningful life? He doesnt say, and Isuspect that his reticence covers
a conviction that any answer would be intellectual sleight of hand atbest.
Kitcher (2011c, 9) sketches a speculative account of the evolution of
religious phenomena that is, as he says, an alternative to my own specula-
tive conjectureit sees the predominance of religion as explained by its
being (socially) adaptive, not a byproduct of other evolutionary selection
pressuresbut he then draws a misleading contrast:

If you start with the thought that the predominance of religion in human societ-
ies is to be explained by a cognitive deficiency, you will tend [my italics] to see
your campaign for the eradication of myths in terms of a return to intellectual
health. . . . By contrast, if you suppose that the social factors towards which
Ihave gestured have played a non-trivial role in the spread of the worlds reli-
gions, you will wonder [my italics] if there are psychological and social needs that
the simple abandonment of religion will leave unfulfilled.

There may be such a tendency, pulling in opposite directions, but Darwin

(1862, 461)long ago showed us that it should be resisted:It is in perfect
accordance with the scheme of nature, as worked out by natural selection,
that matter excreted to free the system from superfluous or injurious sub-
stances should be utilized for [other] highly useful purposes.
So Kitchers wonder is just as available to me, and, similarly, however
socially adaptive religious phenomena may have been in the past, their util-
ity may have expired. It is simply a mistakebut a very common oneto
seek a theory of the evolution of religious phenomena that would provide
some imagined warrant for your view of the value of religion today. Iworked
hard to keep these issues distinct in my book, and Kitcher should acknowl-
edge that his preferred speculation is logically independent of the main point
for which he is arguing:that religion is valuable today, all things considered.
It may well be. Ifind his most compelling point to be his observation
that Dawkins and Ishould not extrapolate glibly from our own immense
good fortune:we find ourselves in a position not only to understand and
appreciate the glories of the scientific worldview but to have the thrill
no other word will sufficeof playing significant roles in the spreading of
this vision:The vast majority will never be able to recognize themselves as
important participants in any impressive joint enterprise that contributes
to knowledge and enlightenment (Kitcher 2011c,11).

W h at t o D o W h i l e R e l i g i o n s E volv e b e f or e O u r V e r y E y e s [279]
I discuss this in Breaking the Spell (Dennett 2006, 28692), where Inote
that religion has the unparalleled capacity to give people a chance to be, in
Kitchers good phrase, important participants in the world they were born
into. But as Igo on to discuss there, nobody has yet estimated what price
we should be prepared to payin xenophobia, violence, the glorification of
unreason, the spreading of patent falsehoodfor that wonderful sense of
importance religion gives to many people who would otherwise lead lives
without drama, without a point. Kitcher wants to preserve religions (at
least for the foreseeable future, I gather), but I think it would be better
to work constructively on secular institutions that can provide alternative
structures of meaning for everyone, hastening the day of religions demise.
Still we might accomplish this most practically by encouraging existing reli-
gious institutions to evolve into former religions. Some have already
done so, but they are not yet competing very well in the marketplace of
allegiances. That may soon change.
The transparency of information engendered by electronic media has
dramatically changed the epistemological environmentthe environment
of knowledge, belief, error, illusion, confidencethat we all inhabit. It
threatens the security and stability of all institutions that depend on con-
fidence and trust, which includes such disparate entities as newspapers,
banks, hospitals, religions, and universities. If a reliable source of news
loses its reputation for telling the truth, it may be out of business, no mat-
ter how scrupulously it checks the facts it publishes. So a new arms race is
now ensuing, dealing in the manipulation of reputations for truth-telling,
and its campaigns can be detected on all sides. Al Jazeera has an excellent
and deserved reputation for truthful reporting in most of the world out-
side the United States. Will its recent acquisition of Al Gores news website
finally secure its respectability in this country or damage Gores reputa-
tion? Time magazine continues its print edition in the United States, while
Newsweek abandoned its US edition in 2012, choosing to link its fate to
the Daily Beast website. In recent decades both magazines have experi-
mented with the ploy of bolstering sales by running favorable cover stories
on religious topics:the Shroud of Turin, the Dead Sea Scrolls, the Gnostic
Gospels, new interpretations of putative relics and archaeological discover-
ies. They clearly have seen the security of religious institutions as a possible
lifeboat to cling to, but it is not obvious that their design choice was wise,
in retrospect.
Oxford zoologist Andrew Parker (2003) has proposed in his book, In
the Blink of an Eye that the famous Cambrian Explosion of novel life forms
that occurred about 530 million years ago was triggered by a change in
the chemistry of the atmosphere or the seas or both, which increased the

[280] The Philosophy of Philip Kitcher

transparency of these media, letting much more sunlight in, making vision
possible for the first time. Suddenly it paid to have eyesight with which to see
your prey or your predators in the offing, and as eyes evolved, so did all the
devices and tactics that eyes enable. The result was an arms race of new ways
of hiding and seeking, locomoting and defending oneself that drove evolution
into one of its most innovative and revolutionary periods. My suggestion is
that we are entering just such an arms race today thanks to the cultural evo-
lution of all the new ways of obtaining and distributing informationand
misinformation (Dennett and Roy, 2015). Old tactics and defenses no longer
suffice, and whoever doesnt redesign in a hurry is doomed to extinction.
I suspect that the critical element in the information revolution that
jeopardizes traditional religions is the explosive increase in mutual knowl-
edge. It is not just that hundreds of millions of people know about the sex-
ual abuse scandal in the Catholic Church; hundreds of millions of people
know that hundreds of millions of people know about the sexual abuse.
The Church used to be able to hush up scandals locally, moving the offend-
ers to other parishes; there were millions of people who each knew about
one local case or another, or maybe a few, but nobody (one may charita-
bly assume) knew that there were these millions of people who knew this;
nobody knew that there were thousands of cases around the world. Now,
thanks to the media, when everybody knows about the thousands of
priests around the world who have been uncovered as sexual abusers, the
Church is very hard-pressed to recruit new priests. Knowing that all the
parents in ones parish will be warning their children about the dangers
of being alone with a priest must dampen the enthusiasm of any young
Catholic man who is considering entering the priesthood. Ageneration ago
Ireland had three priests for every parish; today it is three parishes for each
priest. The number of Roman Catholic priests in North America is down a
third from a generation ago, and the seminaries dont have enough appli-
cants to replenish the pipeline.
That is a particularly vivid example of how swiftly a reputation prob-
lem can threaten the very survival of an institution that has seemed for
centuries to be as impervious as any mountain. More indirect effects are
worrying the leaders of other denominations. When children learn of the
many alternatives to the creeds they are taught by their elders, otherwise
unthinkable options become real possibilities. Baptists in America today
are baptizing about as many each year as they did back in the middle of
the twentieth century, when the population was less than half what it is
today, a huge erosion of market share. The Evangelical movement antici-
pates shrinking dramatically in the twenty-first century, according to some
studies. According to one, If current trends continue, only four percent

W h at t o D o W h i l e R e l i g i o n s E volv e b e f or e O u r V e r y E y e s [281]
of teenagers will be Bible-believing Christians as adults (Goodstein 2006;
see also Spencer 2009). Not without good reason do defenders of religion
inflate the numbers of their adherents and concealas best they canany
negative trends in attendance. For the same reasons, of course, colleges
and universities neglect to report declining enrollments and applications
while trumpeting any gains but among religions there are few gains to her-
ald. Muslims and Mormons have increasing populations, due mainly to dif-
ferential birth rates, not conversions, but no religion at all is the fastest
growing category, worldwide.
As these and a host of other such facts become more widely knownand
known to be knownit will be hard for religious spokespeople to maintain
their traditional tone of authority. Religions that flourished in the murky
epistemic waters of earlier millennia are going to find themselves increas-
ingly vulnerable to impertinent questions about their practices, their fund-
ing, their creeds. Recent history has shown us that cover-ups that used to
succeed in the past now have a way of imploding spectacularly. Will the com-
mitment model that Kitcher recommends be able to deflect the scrutiny that
would undermine it? Who knows what the near future will bring? Religions
have changed more in the past century than in the past millennium, and
perhaps they will change more in the next decade than in that past century.
Kitcher and Iagree on so much. We agree that public reason must be
thoroughly secular (Kitcher 2011c, 12). We agree that the belief model of
religion is indefensible. We agree that the first spell must be brokenwe
have both broken it. We differ, apparently, only in our assessment of how
to ease the people of the twenty-first century into a more reasonable and
socially benign form of organization. Its like the problem of how to remove
an adhesive bandage: slowly, gently lifting, pulling, pausingor a swift,
well-timed yank. Ifavor the quick shock and its overnot so bad, and now
lets get on with our lives. Kitcher favors the glacial approach, and whatever
there is to be said for it when calculating the costs and benefits, it is impor-
tant not to overlook the suffering of those who get caught in the pulpit,
slowly accumulating a deplorable history of dissembling and obfuscation
in the name of tact. We all indulge in those little white lies, and it would be
heartless to forswear them, but how much is too much? Ithink we should
both admit that we havent yet figured thatout.1

1. Portions of this essay are revised from Kitcher versus Dennett:Is New Atheism
Counterproductive?, Why Evolution Is True, October 7, 2010, https://whyevolu- 2 010/ 10/ 07/ k itcher- versus- d ennett- i s- new- atheism-
counterproductive-2/, and from Dennett and LaScola 2013 and2015.

[282] The Philosophy of Philip Kitcher

Reply toDennett

For four decades now Ive been a great admirer of Dan Dennetts philosoph-
ical work. His concerns that philosophy should be broadly accessible and
that it should focus on questions that interest more than a small coterie of
professionals are sources of inspiration. Dans ability to combine stylistic
elan with clarity and precision sets a model for us all. I have often been
encouraged by the fact that he and Ihave stood shoulder to shoulder on
many issues. As Dan points out, we agree on somuch.
With respect to religion our disagreements resemble those of the
Bolsheviks and the Mensheviksall the more intense, perhaps, because
the differences might appear, to an external observer, to be so small. In the
end, however, Ithink they are more important than simple divergences on
strategy. We are divided not only about the value of religion but also on
some basic questions about belief.
Dennetts characteristically lucid and forceful essay engages with the
argument of my Militant Modern Atheism (Kitcher 2010; also Kitcher
2012c, ch. 12). Dennett responds to my contrast between the belief and
orientation models by arguing that, once the orientation model is made
explicit, those whose religious commitments conform to it are placed in
an uncomfortable position. They have to decide exactly where they stand
on matters of doctrine: are they mythically self-conscious, doctrinally
indefinite, or doctrinal[ly] entangle[d]? Unless they fall into the last cat-
egory, in which case they are lapsing into the belief model, they are forced
to realize that they are kid[ding] themselves. When religion backs away
from full-blooded commitment to doctrines, the devout have to fess up to
faith fibbing.
To my mind this response is revealing, for it shows how Dennett thinks
about belief. His writings about religion tacitly adopt a pair of theses,
shared by the less sophisticated members of the Four Horsemen (I owe
the whole quartet an apology for having mistakenly suggested that they
proposed thisname):

1. When a person properly affirms a statement, there is a precise and defi-

nite content, formulable in language as transparent as the languages of
the sciences.

W h at t o D o W h i l e R e l i g i o n s E volv e b e f or e O u r V e r y E y e s [283]
2. Believing truly is an ultimate value, one that can override other values.
Replacing false belief by true belief brings gains that outweigh whatever
losses are involved in the replacement.

In my later work on religion (Kitcher 2014), Ive called both theses into
question, insisting on the need to recognize that some valuable linguistic
practices stand in need of interpretation and that arriving at a state of clear
and definite true belief may not be worth the sacrifices required to achieve
it. My aim in what follows is to elaborate these points.
Dennett imagines religious believers who are dumbfounded by the
request to situate themselves with respect to the orientation model.
Perhaps he and I move in different circles, but I know many thoughtful
people who would have no trouble with the question. Some, like the late
Robert Bellah, would identify themselves as mythically self-conscious.
They recognize that their creedal professions arent intended to record
matters of historical fact, and they are happy to talk of their doctrines as
myths (Bellah 2011)and even to claim that clashing myths can count
as true. Chapter3 of Kitcher (2014) offers a philosophical reconstruction
of this usage, distinguishing alternative approaches to truth in different
domains (see my replies to Gideon Rosen and to Michael Smith). To put
the point very simply, religious statements are viewed as metaphorical, as
gestures toward a different aspect of realitythe transcendentand valu-
able because of their fruits forlife.
Others, probably the majority of those who profess liberal (or, as
Iwould characterize it, refined) religion, would answer Dennetts question
differently. They would point out, from the beginning, that the statements
in question are complex, in need of interpretation. Some of them might
invoke particular interpretations they favor, interpretations that separate
the content of belief from any literal commitment to supernatural goings-
on. All would deny that the interpretations they can give, or even the inter-
pretations that have been provided in the history of their religion, exhaust
the significance of the doctrinal statements. Many would likely concur with
the attitude recommended by Bellah, seeing all the worlds major religions
(and perhaps the less prevalent ones as well) as gesturing toward an aspect
of reality that literal language cannot capture, an aspect that deserves the
emotional responses William James ([1902] 1982, lecture 2)saw as central
to religion (solemnity and awe, combined with a joyous acceptance). Faced
with the suggestion that they are kidding themselves or faith fibbing,
they would reject the description (possibly indignantly); resorting to meta-
phor or allegory, they would explain, is not the same as lying, particularly
when there is no other way to express important insights.

[284] The Philosophy of Philip Kitcher

Dennett, working with Linda LaScola, has been much concerned to help
a group of lapsed religious people who cannot take the stance I find so
common among the thoughtful devout. The effort to help clergy who have
lost faith deserves sustained applauseeven though, as Ill suggest below,
theres a vastly larger group of vulnerable people to whose predicaments
Dennett often appears blind. Interactions with those who cannot find
any convincing interpretation of the doctrines they are paid to recite and
teach has probably encouraged Dennetts sense that people who grope for
words to explain their religious commitments are hopeless muddleheads.
He issues a challenge:Give me a list of theologians who are worth reading.
Ill bite. James (although not officially a theologian), Martin Buber, Paul
Tillich, George Lindbeck, Nicholas Lash, and Bellah (again, a believer, not
a theologian). Ianticipate Dennetts reply:All these writers descend into
deep-sounding vagueness at crucial moments; their works lack clarity and
precision. Apparently nothing short of an explication meeting Carnaps
high standards will do, either for the believer who is asked to clarify his
version of the orientation model or for the theologians to whose works the
believer might turn. Thesis 1 tacitly underlies Dennetts complaint against
liberal (refined) religion.
Let me be clear. In the end, as Kitcher (2014) argues, even refined reli-
gion should be seen as a way station on the road to secular humanism. My
book (2014) is concerned to show how refined religion evades the most
powerful arguments against (literalist) religions (those committed to the
belief model) and how a philosophical reconstruction of it can be given.
Refined religion fails because it introduces an unnecessary and potentially
diversionary transcendent realm that compromises the human values at
the core of the orientation model and because it narrows the class of narra-
tives that are fruitful for humanlife.
Yet refined religion should not be dismissed out of hand in the way
Dennett and his fellow Horsemen favor. Thesis 1 embodies a narrow sci-
entism. To see this, consider literature generally, and poetry in particular.
Neither King Lear nor The Brothers Karamazov nor Finnegans Wake would
pass the Carnapian explication test. The failure is even more evident with
respect to the poetry of Rimbaud or Rilke or Blake or T.S. Eliot. The closing
section of Little Gidding is plainly religiously inflected, but even a reso-
lute atheist like me can find significant content in Eliots resonant lines.
You dont have to be a believer to sense profound truths, truths you can
partially express but never render fully in the language Dennett demands.
Why should a differenthigherstandard be set for the doctrinally indef-
inite believer or for the writers on mylist?

W h at t o D o W h i l e R e l i g i o n s E volv e b e f or e O u r V e r y E y e s [285]
I turn now to Thesis 2 and its unfortunate consequences. Iagree with
Dennett and Dawkins that it is often good for people to be enlightened, for
their false beliefs to be replaced by true ones. From my Science, Truth, and
Democracy (Kitcher 2001b) on, however, Ive been arguing that not every
truth is worth having. The aim of our epistemic ventures is significant truth.
Moreover the significant truths are those that contribute to human well-
being. Practical significance is the most obvious species: we value some
truths because they help doctors ameliorate diseases or protect vulnerable
people from threats to their security. Yet we shouldnt overlook the value
of truths that simply advance our understanding of the world. Dawkins is
often eloquent on the joy that comes from viewing nature clearly and accu-
rately. He rightly laments the blindness that prevents many people from
experiencing the uplift enlightenment brings.
Nevertheless once you make the shift from the value of truth (period)
to the value of significant truth, Thesis 2 becomes problematic. Because
significance depends on the conditions of human lives, the predicaments
different individuals and communities face, significant truth is thrown into
a broader mix of values, so that its benefits may be swamped by the losses
entailed in achieving it. No doubt if we lived in some rather different world,
it would be good to learn about the degrees to which genetic differences
affect the talents on which our current societies place most emphasis. But
if our only method for acquiring the knowledge would be to breed identical
twins, rearing them apart in controlled environments, we rightly forgo the
benefits of understanding the precise genetic contribution. And even if we
have alternative routes to knowledge, we might be obligated not to take
them if we were convinced that the likely damage from future uses of the
information would be toogreat.
My resistance to sweeping away all religion as noxious rubbish stems
from combining the pragmatist sense of truth as one value among others
(the point of the previous paragraph) with reflections on the lives Ihave
known. For anyone like me, for whom a youth in the lower reaches of the
British class system brought acquaintance with many people whose lives
were made bearable by their local church (liberal, by todays standards) and
the sense of community it provided, the losses of religious involvement can
easily outweigh the gains of enlightenment:Ihave known too many people
who have said, sincerely and accurately, Without my faith Ijust couldnt go
on. Their declaration stems, of course, from failures of the ambient societ-
ies. They are probably at least as common among those who live just north
of me in Harlem and who turn to one of the local churches or synagogues
or mosques for comfort, for support, for community, and sometimes for
a joint campaign for social justice. Ideally we could attend to the causes

[286] The Philosophy of Philip Kitcher

of injustice and provide meaningful communities for those who are strug-
gling. In practice we are at a far remove from being able todoso.
Dennett and I focus on different groups. Salient for him (and for
LaScola) are the lapsed clergy, whose chosen professions require them
publicly to proclaim things they privately reject. Prominent for me are the
worlds poor and vulnerable, whose religious practices give shape and value
to their lives. The task of building a fully secular world in which their needs
will be properly accommodated strikes me as vastthat world wont be
built in a day or even in a generation. Much as Iadmire the humanist orga-
nizations that have tried to assume the functions of religious communities,
their efforts strike me as early experiments to find surrogates for the ritu-
als, devotions, and forms of joint life that the major religions have honed
over centuries.
For Dennett the main difference between us is one of strategy. He prefers
to pull off the bandage quickly and sees me as favoring the glacial approach.
His image captures some features of our quarrel, but its important to pose
a question:Who, or what exactly, feels the pain? The most obvious subject
is a society, a particular nation, perhaps, or the entire human population.
Society, after all undergoes a transformation, from a state in which religion
is prevalent, even dominant, to another state from which religion has van-
ished. The envisaged benefits are clear:no more obfuscation to cloud policy
discussions, no more people with false religious conceptions, no more clergy
crossing their fingers as they preach. Society, however, doesnt feel the pain.
Individual people doand especially those whose otherwise drab or miser-
able lives are transformed by their religious practices.
The potential anguish caused by the quick removal of the bandage
seems to me vast enough to outweigh the benefits. But even if the utilitar-
ian calculation goes differently, computing a net gain in aggregate levels
of well-being, theres no shirking the obvious standard objection. Alarge
number of lives would be blighted in order to secure the well-being of oth-
ers. Better, Ibelieve, to develop secular humanism as a positive position,
showing clearly how a meaningful life after faith is possible, and, probably
more important, to combat social injustice and fashion functional substi-
tutes for religious institutions. My recent work on religion is intended to
point theway.
The rate of secular transformation does not have to be glacial. Ihope it
goes smoothly and quickly. But Idont think the task of building a secular
world will be forwarded by cheery assurances about quick tugs of the ban-
dage or by neglect of the positive comforts the worlds religions bring. So
for all my admiration for Dan, all my agreement with his endorsement of
the secular life, Iremain a soft atheist, an unrepentant Menshevik.

W h at t o D o W h i l e R e l i g i o n s E volv e b e f or e O u r V e r y E y e s [287]

AIMS. 2009. Memorandum from the Association for Improvements in the Maternity
Services (AIMS) to the House of Commons Select Committee for Children,
Schools and Families.
Alon, Uri. 2007. An Introduction to Systems Biology:Design Principles of Biological Circuits.
Boca Raton, FL:Taylor and Francis.
Alston, William. 1958. Ontological Commitments. Philosophical Studies 9:817.
Amundson, Ron, and George Lauder. 1994. Function without Purpose. Biology and
Philosophy 9 (4):44369.
Aronson, Elliot, Timothy Wilson, and Robin Akert. 1999. Social Psychology. 3rd ed.
Bacon, Francis. (1620) 1960. The New Organon. Ed. Fulton H.Anderson. Upper Saddle
River, NJ:PrenticeHall.
Balaguer, Mark. 1998. Platonism and Anti-Platonism in Mathematics. Oxford: Oxford
Barlow, Allison, Elena Varipatis-Baker, Kristen Speakman, Golda Ginsburg, Ingrid
Friberg, Novalene Goklish, Brandii Cowboy, Pauline Fields, Ranelda Hastings,
William Pan, Raymond Reid, Mathuram Santosham, and John Walkup. 2006.
Home-Visiting Intervention to Improve Child Care among American Indian
Adolescent Mothers:ARandomized Trial. Archives of Pediatrics and Adolescent
Medicine 160 (11):11017.
Barnes, Barry, and John Dupr. 2008. Genomes and What to Make of Them.
Chicago:University of ChicagoPress.
Batterman, Robert. 2001. The Devil in the Details:Asymptotic Reasoning in Explanation,
Reduction and Emergence. Oxford:Oxford UniversityPress.
Becker, Jo. 2014. Forcing the Spring: Inside the Fight for Marriage Equality.
Bellah, Robert. 2011. Religion in Human Evolution. Cambridge, MA: Harvard
Benacerraf, Paul. 1965. What Numbers Could Not Be. Philosophical Review 74
. 1973. Mathematical Truth. Journal of Philosophy 70:66179.
Benacerraf, Paul, and Hilary Putnam. 1964. Philosophy of Mathematics: Selected
Readings. Englewood Cliffs, NJ:PrenticeHall.
Benson, Etienne. 2012. Endangered Science:The Regulation of Research by the U.S.
Marine Mammal Protection Act and Endangered Species Act, 19711978.
Historical Studies in the Natural Sciences 42:3061.
Berlin, Isaiah. (1950) 1979. Concepts and Categories. NewYork:Penguin.
Bigelow, John, and Robert Pargetter. 1987. Functions. Journal of Philosophy
Blair, Tony. 2006. Our Nations FutureSocial Exclusion. Our Nations Future lec-
ture series, September 5.
Boorse, Christopher. 1976. Wright on Functions. Philosophical Review 85:7086.
Boyd, Richard. 1991. Realism, Anti-Foundationalism, and the Enthusiasm for Natural
Kinds. Philosophical Studies 61:12748.
. 2010. Realism, Natural Kinds, and Philosophical Method. In The Semantics
and Metaphysics of Natural Kinds, ed. Helen Beebee and Nigel Sabbarton-Leary,
21234. London: Routledge.
Boyd, Robert, and Peter Richerson. 1985. Culture and the Evolutionary Process.
Chicago:University of ChicagoPress.
. 2005. Not by Genes Alone. Chicago: University of Chicago Press.
Brake, Elizabeth, and Joseph Millum. 2013. Parenthood and Procreation. In The
Stanford Encyclopedia of Philosophy (Winter 2013), ed. Edward N.Zalta. http://
Brandon, Robert N. 2013. A General Case for Functional Pluralism. In Functions:
Selection and Mechanisms, ed. Philippe Huneman, 97104. Dordrecht: Springer.
Brown, Mark B. 2013. Review of Philip Kitcher, Science in a Democratic Society. Minerva
Burgess, John P. 1983. Why IAm Not a Nominalist. Notre Dame Journal of Formal
Logic 24 (1):93105.
Burgess, John, and Gideon Rosen. 1997. A Subject with No Object. Oxford: Oxford
Bush, Vannevar. (1945) 1980. ScienceThe Endless Frontier. NewYork:ArnoPress.
Cahan, David. 1989. An Institute for Empire:The Physikalisch-Technische Reichsanstalt,
18711918. Cambridge, UK:Cambridge UniversityPress.
Calcott, Brett. 2014. The Creation and Reuse of Information in Gene Regulatory
Networks. Philosophy of Science 81 (5):87990.
Carnap, Rudolph. 1950. Logical Foundations of Probability. Chicago: University of
Cartwright, Nancy. 1983. How the Laws of Physics Lie. Oxford:ClarendonPress.
. 1999. The Dappled World. Cambridge, UK:Cambridge UniversityPress.
Cartwright, Nancy, and Jeremy Hardie. 2012. Evidence-Based Policy:APractical Guide to
Doing It Better. NewYork:Oxford UniversityPress.
Chihara, Charles. 1973. Ontology and the Vicious-Circle Principle. Ithaca, NY: Cornell
Christensen, Wayne. 1996. A Complex Systems Theory of Teleology. Biology and
Philosophy 11 (3):30119.
Citizen Science Alliance. 2014. Why Citizen Science? http://www.citizenscience Accessed April 26,2014.
Cooper, Harris. 1989. Homework. White Plains, NY:Longman.
Craver, Carl F. 2007. Explaining the Brain. Oxford:ClarendonPress.
. 2014. The Ontic Account of Scientific Explanation. In Explanation in the
Special Sciences, ed. Marie I.Kaiser, Oliver R.Scholz, Daniel Plenge, and Andreas
Httemann, 2752. Dordrecht:Springer.

Crick, Francis H.C. 1958. On Protein Synthesis. Symposia of the Society for Experimental
Biology 12:13863.
CSCB. 2013. Final Overview Report of Serious Case Review re Daniel Pelka. Coventry,
UK:Coventry Safeguarding ChildrenBoard.
Cummins, Robert. 1975. Functional Analysis. Journal of Philosophy 72:74165.
. 2002. Neo-Teleology. In Functions:New Essays in the Philosophy of Psychology
and Biology, ed. Andr Ariew, Robert Cummins, and Mark Perlman, 15773.
Oxford:Oxford UniversityPress.
Curry, Haskell B. 1951. Outlines of a Formalist Philosophy of Mathematics.
DAgostino, Fred. 2013. Review of Philip Kitcher. Science in a Democratic Society.
Analysis 73:59394.
Daniels, George. 1967. The Pure- Science Ideal and Democratic Culture. Science
Darwin, Charles. (1859) 1964. On the Origin of Species. Cambridge, MA: Harvard
. 1862. On the Various Contrivances by Which British and Foreign Orchids Are
Fertilised by Insects, and on the Good Effects of Intercrossing. London:John Murray.
Daston, Lorraine. 2001. The Historicity of Science. In Historicization-Historisierung,
vol. 5 of Aporemata:Kritische Studien zur Philologiegeschichte, ed. Glenn W.Most,
20121. Gttingen:Vandenhoeck & Ruprecht.
. 2012. The Sciences of the Archives. Osiris 27:15687.
Davidson, Nicholas O. 2002. The Challenge of Target Sequence Specificity in CU
RNA Editing. Journal of Clinical Investigation 109 (3):29194.
Dawkins, Richard. 2006. The God Delusion. NewYork:Norton.
Dennett, Daniel C. 2006. Breaking the Spell: Religion as a Natural Phenomenon.
Dennett, Daniel C., and Linda LaScola. 2010. Preachers Who Are Not Believers.
Evolutionary Psychology 8 (1):12250.
. 2013. Caught in the Pulpit:Leaving Belief Behind.
. 2015. Caught in the Pulpit: Leaving Belief Behind. Expanded ed. Durham,
Dennett, Daniel C., and Deb Roy. 2015. Our Transparent Future. Scientific American
312 (3): 6469.
Detlefsen, Michael. 2005. Formalism. In The Oxford Handbook of Philosophy of
Mathematics and Logic, ed. Stewart Shapiro, 236317. Oxford: Oxford University
Dewey, John. (1909) 2007. The Influence of Darwin on Philosophy. In The Influence
of Darwin on Philosophy and Other Essays, ed. Larry Hickman, 312. Carbondale:
University of Southern Illinois.
. (1925) 1981. Experience and Nature. Vol. 1 of Later Works, ed. Jo Ann Boydston.
Carbondale:University of Southern Illinois.
. (1929) 1984. The Quest for Certainty. Vol. 4 of Later Works, ed. Jo Ann Boydston.
Carbondale:University of Southern Illinois.
Douglas, Heather. 2009. Science, Policy, and the Value-Free Ideal. Pittsburgh:University
of PittsburghPress.
Dowe, Phil. 2000. Physical Causation. Cambridge, UK:Cambridge UniversityPress.
Dretske, Fred. 1972. Contrastive Statements. Philosophical Review 81:41137.

References [291]
Dummett, Michael. 1991. Frege:Philosophy of Mathematics. Cambridge, MA:Harvard
Dupr, John. 1993. The Disorder of Things. Cambridge, MA:Harvard UniversityPress.
Edwards, Anthony W.F. 1998. Natural Selection and the Sex Ratio:Fishers Sources.
American Naturalist 151:56469.
Edwards, Paul N. 2010. A Vast Machine:Computer Models, Climate Data, and the Politics
of Global Warming. Cambridge, MA:MITPress.
Elster, Jon. 1984. Ulysses and the Sirens: Studies in Rationality and Irrationality.
Cambridge, UK:Cambridge UniversityPress.
Field, Hartry. 1980. Science without Numbers. Princeton, NJ:Princeton UniversityPress.
. 1989. Realism, Mathematics and Modality. Oxford: Oxford University Press.
Fisher, Ronald A. 1930. The Genetical Theory of Natural Selection. Oxford: Oxford
University Press.
Fishkin, James. 2009. When the People Speak. NewYork:Oxford UniversityPress.
Fletcher, Liz. 2001. New Zealand GMO Debacle Undermines Green Lobby. Nature
Biotechnology 19:292.
Fodor, Jerry A. 1974. Special Sciences. Synthese 28:97115.
Forrest, Peter, and David M.Armstrong. 1987. The Nature of Number. Philosophical
Papers 16:16586.
Fox, Charles W., Monica S. Thakar, and Timothy A. Mousseau. 1997. Egg Size
Plasticity in a Seed Beetle: An Adaptive Maternal Effect. American Naturalist
149 (1):14963.
Franklin-Hall, Laura R. Forthcoming. The Causal Economy Account of Scientific
Explanation. In Philosophical Perspectives on Causal Reasoning in Biology, ed.
C. K. Waters and J. Woodward, vol. 21 of Minnesota Studies in the Philosophy of
Science. Minneapolis: University of Minnesota Press.
. 2015. Explaining Causal Selection with Explanatory Causal Economy:
Biology and Beyond. In Explanation in Biology: An Enquiry into the
Diversity of Explanatory Patterns in the Life Sciences, ed. P. A. Braillard and
C. Malaterre, 413438. Dordrecht: Springer.
Frege, Gottlob. (1903) 2013. Basic Laws of Arithmetic. Ed. Philip A. Ebert and Marcus
Rossberg. Oxford: Oxford University Press.
. (1918) 1984. Der Gedanke: Eine Logische Untersuchung. In Beitrge zur
Philosophie des deutschen Idealismus 1 (191819):5877. Translated as Thoughts,
in Collected Papers on Logic, Mathematics and Philosophy. Trans. Max Black,
V. H.Dudman, Peter Geach, Hans Kaal, E. H.W. Kluge, Brian McGuinness, and
R. H.Stoothoff. Ed. B. McGuinness. Oxford:Basil Blackwell.
French, Steven. 2012. The Resilience of Laws and the Ephemerality of Objects. In
Probabilities, Laws, and Structures, ed. Dennis Dieks, Wenceslao Gonzalez,
Stephan Hartmann, Michael Stoltzner, and Marcel Weber, 187200. Dordrecht:
Springer Science and BusinessMedia.
Friedman, Michael. 1974. Explanation and Scientific Understanding. Journal of
Philosophy 71:519.
Gibbard, Allan. 1990. Wise Choices, Apt Feelings. Cambridge, MA:Harvard University
Giere, Ronald. 2006. Scientific Perspectivism. Chicago:University of ChicagoPress.
. 2013. Kuhn as Perspectival Realist. Topoi 32:5357.
Gdel, Kurt. 1964. What Is Cantors Continuum Problem? In Philosophy of
Mathematics:Selected Readings, ed. Paul Benacerraf and Hilary Putnam, 25873.
Englewood Cliffs, NJ:PrenticeHall.

Godfrey-Smith, Peter. 1993. Functions:Consensus without Unity. Pacific Philosophical
Quarterly 74:196208.
. 1994. A Modern History Theory of Functions. Nous 28:34462.
Goffman, Erving. 1959. The Presentation of Self in Everyday Life. NewYork:AnchorBooks.
Gonzalez, Wenceslao J., ed. 2011. Scientific Realism and Democratic Society: The
Philosophy of Philip Kitcher. Amsterdam:Rodopi.
Goodstein, Laurie. 2006. Evangelicals Fear the Loss of Their Teenagers. New York
Times, October6.
Gould, Stephen Jay, and Richard C. Lewontin. 1979. The Spandrels of San Marco
and the Panglossian Paradigm: A Critique of the Adaptationist Programme.
Proceedings of the Royal Society B 205:58198.
Greenberg, Daniel S. 2001. Science, Money, and Politics: Political Triumph and Ethical
Erosion. Chicago:University of ChicagoPress.
Griffiths, Paul E. 2006a. The Fearless Vampire Conservator: Philip Kitcher, Genetic
Determinism and the Informational Gene. In Genes in Development:Re-reading
the Molecular Paradigm, ed. Christoph Rehmann-Sutter and Eva M.Neumann-
Held, 17598. Durham, NC:Duke UniversityPress.
. 2006b. Function, Homology and Character Individuation. Philosophy of
Science 73 (1):125.
. 2009. In What Sense Does Nothing Make Sense Except in the Light of
Evolution? Acta Biotheoretica 57 (12):1132.
. 2013. Lehrmans Dictum: Information and Explanation in Developmental
Biology. Developmental Psychobiology 55 (1):2232.
Griffiths, Paul E., and Russell D.Gray. 1994. Developmental Systems and Evolutionary
Explanation. Journal of Philosophy 91 (6):277304.
Griffiths, Paul E., Edouard Machery, and Stefan Linquist. 2009. The Vernacular
Concept of Innateness. Mind and Language 24 (5):60530.
Griffiths, Paul E., Arnaud Pocheville, Brett Calcott, Karola Stotz, Hyunju Kim, and Rob
D.Knight. 2015. Measuring Causal Specificity. Philosophy of Science 82:52955.
Griffiths, Paul E., and Karola Stotz. 2013. Genetics and Philosophy: An Introduction.
NewYork:Cambridge UniversityPress.
Griffiths, Paul E., and James G.Tabery. 2013. Developmental Systems Theory:What
Does It Explain, and How Does It Explain It? Advances in Child Development and
Behavior 44:6594.
Haack, Susan. 1993. Evidence and Inquiry: Towards Reconstruction in Epistemology.
Hacking, Ian. 1999. The Social Construction of What? Cambridge, MA: Harvard
Hahn, Roger. 1971. The Anatomy of an Institution:The Paris Academy of Sciences, 1666
1803. Berkeley:University of CaliforniaPress.
Hamer, Dean. 2002. Rethinking Behavior Genetics. Science 298:7172.
Hamer, Dean H., S. Hu, V. L.Magnuson, N. Hu, and A. M.L. Pattatucci. 1993. A Linkage
between DNA Markers on the X Chromosome and Male Sexual Orientation.
Science Communication 261:32026.
Hare, Robert. 1991. The Hare Psychopathy ChecklistRevised. Toronto: Multi-Health
. 1998. The Hare PCL-R:Some Issues concerning Its Use and Misuse. Legal and
Criminological Psychology 3 (1):99119.
Harris, Sam. 2004. The End of Faith: Religion, Terrorism and the Future of Reason.

References [293]
. 2006. Letter to a Christian Nation. NewYork:Knopf.
Hecht, Gabrielle. 2012. Being Nuclear:Africans and the Global Uranium Trade. Cambridge,
Hempel, Carl G., and Paul Oppenheim. 1948. Studies in the Logic of Explanation.
Philosophy of Science 15:13575.
Herrnstein, Richard J., and Charles Murray. 1994. The Bell Curve:Intelligence and Class
Structure in American Life. NewYork:Simon and Schuster.
Hitchens, Christopher. 2007. God Is Not Great: How Religion Poisons Everything.
Holling, Crawford S. 1959. The Components of Predation as Revealed by a Study of
Small-Mammal Predation of the European Pine Sawfly. Canadian Entomologist
Huneman, Philippe. 2013. Introduction to Functions: Selection and Mechanisms.
James, William. (1902) 1982. The Varieties of Religious Experience. Harmondsworth,
. (1907) 1975. Pragmatism. Cambridge, MA:Harvard UniversityPress.
Jasanoff, Sheila. 2012. Science and Public Reason. London:Routledge.
Jewett, Andrew. 2012. Science, Democracy, and the American University:From the Civil
War to the Cold War. Cambridge, UK:Cambridge UniversityPress.
Kant, Immanuel. (178187) 1997. Critique of Pure Reason. Cambridge, UK:Cambridge
Kepler, Johannes. 1615. Nova stereometria doliorum vinariorum. Linz, Austria:
J. Plancus.
Kim, Jaegwon. 1988 [2010]. Explanatory Realism, Causal Realism, and Explanatory
Exclusion. In vol. 12 of Midwest Studies in Philosophy, 22539. Reprinted in
Kim, J. (2010). Essays in the Metaphysics of Mind, 14866. Oxford: Oxford
. 1994. Explanatory Knowledge and Metaphysical Dependence. Philosophical
Issues 5:5169.
Kincaid, Harold, John Dupr, and Alison Wylie, eds. 2007. Value-Free Science: Ideals and
Illusions. New York: Oxford University Press.
Kitcher, Philip. 1973. Fluxions, Limits, and Infinite Littlenesse. Isis 64:3349.
. 1975. Bolzanos Ideal of Algebraic Analysis. Studies in the History and
Philosophy of Science 6:22971.
. 1980. Arithmetic for the Millian. Philosophical Studies 37:21536.
. 1981. Explanatory Unification. Philosophy of Science 48:50731.
. 1982. Abusing Science:The Case against Creationism. Cambridge, MA:MITPress.
. 1983. The Nature of Mathematical Knowledge. NewYork:Oxford UniversityPress.
. 1984. 1953 and All That: A Tale of Two Sciences. Philosophical Review 93:
. 1985. Vaulting Ambition: Sociobiology and the Quest for Human Nature.
Cambridge, MA:MITPress.
. 1986. Projecting the Order of Nature. In Kants Philosophy of Physical Science,
ed. Robert Butts, 20135. University of Western Ontario Series in Philosophy of
Science. Dordrecht:D. Reidel.
. 1989. Explanatory Unification and the Causal Structure of the World. In
Scientific Explanation, ed. Philip Kitcher and Wesley Salmon, 410505, vol. 13
of Minnesota Studies in the Philosophy of Science. Minneapolis: University of

. 1990. The Division of Cognitive Labor. Journal of Philosophy 87:521.
. 1993a. The Advancement of Science. NewYork:Oxford UniversityPress.
. 1993b. The Evolution of Human Altruism. Journal of Philosophy 90:497516.
. 1993c. Function and Design. Midwest Studies in Philosophy 18:37997.
. 1996. The Lives to Come: The Genetic Revolution and Human Possibilities.
NewYork:Simon and Schuster.
. 1998. A Plea for Science Studies. In A House Built on Sand: Exposing
Postmodernist Myths about Science, ed. Noretta Koertge, 3256. NewYork:Oxford
. 1999. The Hegemony of Molecular Biology. Biology and Philosophy 14:195210.
. 2001a. Real Realism:The Galilean Strategy. Philosophical Review 110:15197.
. 2001b. Science, Truth, and Democracy. NewYork:Oxford UniversityPress.
. 2003a. Battling the Undead: How (and How Not) to Resist Genetic
Determinism. In In Mendels Mirror: Philosophical Reflections on Biology, 283
300. Oxford:Oxford UniversityPress.
. 2003b. Developmental Decomposition and the Future of Human Behavioral
Ecology. In In Mendels Mirror: Philosophical Reflections on Biology, 30120.
Oxford:Oxford UniversityPress.
. 2007a. Joyces Kaleidoscope: An Invitation to Finnegans Wake. Oxford: Oxford
. 2007b. Living with Darwin: Evolution, Design, and the Future of Faith.
Oxford:Oxford UniversityPress.
. 2011a. The Ethical Project. Cambridge, MA:Harvard UniversityPress.
. 2011b. Science in a Democratic Society. Amherst, NY:PrometheusBooks.
. 2011c. Militant Modern Atheism. Journal of Applied Philosophy 28:113.
. 2012a. Afterthoughts:Reply to Comments. Analyse & Kritik 1:16789.
. 2012b. Mathematical Truth? In Preludes to Pragmatism, 166 91. New
York:Oxford UniversityPress.
. 2012c. Preludes to Pragmatism. NewYork:Oxford UniversityPress.
. 2013. Deaths in Venice:The Cases of Gustav von Aschenbach. NewYork:Columbia
. 2014. Life after Faith. New Haven, CT:Yale UniversityPress.
Kitcher, Philip, and Wesley Salmon. 1987. Van Fraassen on Explanation. Journal of
Philosophy 84: 31530.
Kohn, Linda, Janet Corrigan, and Molla Donaldson, eds. 2000. To Err Is Human:Building
a Safer Health System. Washington, DC:National AcademyPress.
Kuhn, Thomas. 1962. The Structure of Scientific Revolutions. Chicago: University of
. 1990. Dubbing and Re-dubbing: The Vulnerability of Rigid Designation. In
Scientific Theories, ed. C. Wade Savage, 298318, vol. 14 of Minnesota Studies in
the Philosophy of Science. Minneapolis:University of MinnesotaPress.
. 2000. The Road Since Structure. Ed. James Conant and John Haugeland.
Chicago:University of ChicagoPress.
Kusch, Martin. 2002. Knowledge by Agreement. Oxford:Oxford UniversityPress.
Lange, Marc. Forthcoming. Because without Cause. NewYork:Oxford UniversityPress.
Laudan, Larry. 1981. A Confutation of Convergent Realism. Philosophy of Science
. 1984. Science and Values. Berkeley:University of CaliforniaPress.
Lewis, David. 1983. New Work for a Theory of Universals. Australasian Journal of
Philosophy 61:34377.

References [295]
. 2000. Causation as Influence. Journal of Philosophy 97:18297.
Lewontin, Richard C. 1983. The Organism as the Subject and Object of Evolution.
Scientia 118:6582.
Linquist, Stefan, Edouard Machery, Paul E.Griffiths, and Karola Stotz. 2011. Exploring
the Folkbiological Conception of Human Nature. Philosophical Transactions of
the Royal Society B 366:44453.
Lotka, Alfred J. 1956. Elements of Mathematical Biology. 2nd ed. New York: Dover.
Originally published as Elements of Physical Biology in1925.
MacFarlane, John. 2005. Making Sense of Relative Truth. Proceedings of the
Aristotelian Society 105:32139.
. 2009. Nonindexical Contextualism. Synthese 166:23150.
Maddy, Penelope. 1997. Naturalism in Mathematics. Oxford:Oxford UniversityPress.
Mantzavinos, Chrysostomos. 2012. The Ethical Project:ADialogue. Analyse & Kritik
Marques, Teresa. 2014. Relative Correctness. Philosophical Studies 167:36173.
Massimi, Michela. 2010. Galileos Mathematization of Nature at the Crossroad
between the Empiricist Tradition and the Kantian One. Perspectives on Science
. 2012. Scientific Perspectivism and Its Foes. Philosophica 84:2552.
. 2014. Natural Kinds and Naturalised Kantianism. Nos 48:41649.
. 2015a. Walking the Line:Kuhn between Realism and Relativism. In Kuhns
Structure of Scientific Revolutions: 50 Years On, ed. Alisa Bokulich and William
Devlin. Boston Studies in the Philosophy of Science. Springer. DOI: 10.1007/
. 2015b. Working in a New World: Kuhn, Constructivism and Mind-
Dependence. Studies in History and Philosophy of Science 50,8389.
Massimi, Michela, and John Peacock. 2014. The Origins of Our Universe: Laws,
Testability and Observability in Cosmology. In Philosophy and the Sciences for
Everyone, ed. Michela Massimi, 1432. London:Routledge.
Matan, Adee, and Susan Carey. 2001. Developmental Changes within the Core of
Artifact Concepts. Cognition 78 (1):126.
Maudlin, Tim. 1996. On the Unification of Physics. Journal of Philosophy 93:12944.
Maynard Smith, John. 2000. The Concept of Information in Biology. Philosophy of
Science 67 (2):17794.
Mayr, Ernst. 1961. Cause and Effect in Biology. Science 134 (3489):15016.
McLaughlin, Peter. 2001. What Functions Explain: Functional Explanation and Self-
Reproducing Systems. Cambridge, UK:Cambridge UniversityPress.
Mill, John Stuart. (1859) 1975. On Liberty. Ed. David Spitz. NewYork:Norton.
Millikan, Ruth Garrett. 1984. Language, Thought and Other Biological Categories.
Cambridge, MA:MITPress.
Moe, Kristine. 1984. Should the Nazi Research Data Be Cited? Hastings Center Report
Morrison, Margaret. 2000. Unifying Scientific Theories: Physical Concepts and
Mathematical Structures. Cambridge, UK:Cambridge UniversityPress.
Munro, Eileen. 2005. Improving Practice: Child Protection as a Systems Problem.
Children and Youth Services Review 27:37591.
. 2011. The Munro Review of Child Protection: Final Report. London: UK
Department of Education.
Nagel, Ernest. 1979. The Structure of Science. Indianapolis:Hackett.
Nagel, Thomas. 1970. The Possibility of Altruism. Princeton, NJ: Princeton

Nanney, D.L. 1958. Epigenetic Control Systems. Proceedings of the National Academy
of Sciences 44 (7):71217.
Neander, Karen. 1991. Functions as Selected Effects. Philosophy of Science 58:16984.
. 1995. Pruning the Tree of Life. British Journal for the Philosophy of Science 46
. 2002. Types of Traits: The Importance of Functional Homologues. In
Functions: New Essays in the Philosophy of Biology and Psychology, ed. Andr
Ariew, Robert Cummins, and Mark Perlman, 390 415. New York: Oxford
. 2015. Functional Analysis and the Species Design. Synthese. DOI: 10.1007/s
Neander, Karen, and Alex Rosenberg. 2012. Solving the Circularity Problem for
Functions:AResponse to Nanay. Journal of Philosophy 109 (10):61322.
Okasha, Samir. 2009. Causation in Biology. In The Oxford Handbook of Causation,
ed. Helen Beebee, Peter Menzies, and Christopher Hitchcock, 707 25.
NewYork:Oxford UniversityPress.
Olby, Robert Cecil. 2009. Francis Crick:Hunter of Lifes Secrets. NewYork:Cold Spring
Harbor LaboratoryPress.
Oppenheim, Paul, and Hilary Putnam. 1958. Unity of Science as a Working
Hypothesis. In Concepts, Theories, and the Mind-Body Problem, ed. H. Feigl,
M. Scriven, and G. Maxwell, 336, vol. 2 of Minnesota Studies in the Philosophy of
Science. Minneapolis:University of MinnesotaPress.
Oyama, Susan. 2000a. Causal Democracy and Causal Contributions in Developmental
Systems Theory. Philosophy of Science 67 (s1):S332S347.
. 2000b. The Ontogeny of Information: Developmental Systems and Evolution.
2nded., revised and expanded. Durham, NC:Duke UniversityPress.
Parfit, Derek. 1984. Reasons and Persons. Oxford:Oxford UniversityPress.
Parker, Andrew. 2003. In the Blink of an Eye:How Vision Sparked the Big Bang of Evolution.
Cambridge, MA:Perseus.
Pearl, Judea. 2009. Causality: Models, Reasoning, and Inference. 2nd ed. New York:
Cambridge UniversityPress.
Pemberton, Camilla. 2013. Eileen Munro on Daniel Pelka Case:I Cant Claim IWould
Have Done Better. The Children's Services Blog. Community Care. http://www. services-blog/2013/09/eileen-munro-
Pielke, Roger, Jr. 2012. Basic Research as a Political Symbol. Minerva 50:33961.
Potochnik, Angela. 2010. Explanatory Independence and Epistemic
Interdependence:ACase Study of the Optimality Approach. British Journal for
the Philosophy of Science 61:21333.
Psillos, Stathis. 2000. The Present State of the Scientific Realism Debate. British
Journal for the Philosophy of Science 51:70528.
Putnam, Hilary. 1971. Philosophy of Logic. NewYork:Harper andRow.
. 1975. What Is Realism? Proceedings of the Aristotelian Society 76:17794.
Quine, W.V. 1961. Word and Object. Cambridge, MA:MITPress.
. 1969. Natural Kinds. In Ontological Relativity and Other Essays.
NewYork:Columbia UniversityPress.
Raddick, M.Jordan, Georgia Bracey, Pamela L.Gay, Chris J.Lintott, Carie Cardamone,
Phil Murray, Kevin Schawinski, Alexander S.Szalay, and Jan Vandenberg. 2013.
Galaxy Zoo: Motivations of Citizen Scientists. Cornell University Library. Accessed April 26,2014.
Reichenbach, Hans. 1956. The Direction of Time. Berkeley:University of CaliforniaPress.

References [297]
Resnik, Michael. 1980. Frege and the Philosophy of Mathematics. New York: Cornell
Rosen, Gideon. 1999. Review of Maddy (1997). British Journal for the Philosophy of
Science 50:46774.
. 2014. Abstract Objects. In The Stanford Encyclopedia of Philosophy (Fall 2014),
ed. Edward N. Zalta.
Rosen, Gideon, and John P.Burgess. 2005. Nominalism Reconsidered. In The Oxford
Handbook of Philosophy of Mathematics and Logic, ed. Stewart Shapiro, 51535.
Oxford:Oxford Univer