I 5.
Conclusions
The definition of three-dimensional ants (3D ants) given here results in
ants that build three-dimensional highways. This definition shows that
the rules of Langton's ant are elementary and in some cases universally
applicable. An interesting result is that there is more than one style of
highway. First, classification systems were presented. The researches
stated here are only a small part of the infinite world of 3D ants. But the
search space for ants grows exponentially with the rule length and big
three-dimensional worlds are hard to simulate (with respect to memory,
computing time, and costly visualization). Some additional questions
come up. Are there 3D ants that build arbitrarily complex highways?
Does every 3D ant build a highway? Is there an ant and a number of
steps for every arbitrary shape? The consumption of which resource
(time or space) grows faster by increasing rule length?
A simple software tool to simulate 3D ants made this work possible
and will be made available by the author to those who are interested.
I References
[1] C. G. Langton, "Studying Artificial Life with Cellular Automata," Phys-
ica D, 22 (1986) 120-149.
[2] L. A. Bunimovich and S. E. Troubetzkoy, "Rotators, Periodicity and Ab-
sence of Diffusion in Cyclic Cellular Automata," Journal of Statistical
Physics, 74 (1994) 1-14.
[3] J. Propp, "Further Ant-ics," Mathematical Intelligencer, 16 (1994) 37-42.
[4] D. Gale, J. Propp, S. Sutherland, and S. Troubetzkoy, "Further Travels with
My Ant," The Mathematical Intelligencer, 17(3) (1995) 48-56.
[5] O. Beuret and M. Tomassini, "Behavior of Multiple Generalized Langton's
Ants," Proceedings ofthe Artificial Life V Conference, edited by C. Langton
and K. Shimohara (MIT Press, 1998).
[6] L. A. Bunimovich, "Many Dimensional Lorentz Cellular Automata and Tur-
ing Machines," International Journal of Bifurcation and Chaos, 6 (1996)
1127-1135.
Complex Systems, 14 (2003) 263-268
Read Before You Cite!
M. V. Simkin
V. P. Roychowdhury
Department of Electrical Engineering,
University of California, Los Angeles, CA 90095-1594
We report a method for estimating what percentage of people who cited a
paper had actually read it. The method is based on a stochastic modeling
of the citation process that explains empirical studies of misprint distri-
butions in citations (which we show follows a Zipf law). Our estimate is
that only about 20% of citers read the original.
Many psychological tests have the so-called "lie-scale." A small but
sufficient number of questions that admit only one true answer, such as:
"Do you always reply to letters immediately after reading them?" are
inserted among others that are central to the particular test. A wrong
reply for such a question adds a point on the lie-scale, and when the
lie-score is high, the over-all test results are discarded as unreliable.
Perhaps, for a scientist the best candidate for such a lie-scale is the
question: "Do you read all of the papers that you cite?"
Comparative studies of the popularity of scientific papers has been a
subject of much recent interest [1-8], but the scope has been limited to
citation distribution analysis. We have discovered a method of estimat-
ing what percentage of people who cited the paper had actually read it.
Remarkably, this can be achieved without any testing of the scientists,
but solely on the basis of the information available in the lSI citation
database.
Freud [9] had discovered that the application of his technique of psy-
choanalysis to slips in speech and writing could reveal a lot of hidden
information about human psychology. Similarly, we find that the appli-
cation of statistical analysis to misprints in scientific citations can give
an insight into the process of scientific writing. As in the freudian case,
the truth revealed is.embarrassing. For example, an interesting statistic
revealed in our study is that a lot of misprints are identical. Consider,
for example, a four-digit page number with one digit misprinted. There
can be 104 such misprints. The probability of repeating someone else's
misprint accidentally is 10-4 • There should be almost no repeat mis-
prints by coincidence. One concludes that repeat misprints are due to
copying someone else's reference, without reading the paper in question.
In principle, one can argue that an author might copy a citation
from an unreliable reference list, but still read the paper. A modest
Complex Systems, 14 (2003) 269-274; © 2003 Complex Systems Publications, Inc.

270 M. V. Simkin and V. P. Roychowdhury
100
100
rank
Figure 1. Rank-frequency distribution of misprints referencing a paper, which
had acquired 4300 citations. There are 196 misprints total, out of which 45 are
distinct. The most popular misprint propagated 78 times. A good fit to Zipf's
law is evident.
reflection would convince one that this is relatively rare, and cannot
apply to the majority. Surely, in the pre-internet era it took almost
equal effort to copy a reference as to type in one's own based on the
original, thus providing little incentive to copy if someone has indeed
read, or at the very least has procured access to the original. Moreover,
if someone accesses the original by tracing it from the reference list
of a paper with a misprint, then with a high likelihood, the misprint
has been identified and will not be propagated. In the past decade with
the advent of the Internet, the ease with which would-be nonreaders can
copy from unreliable sources; as well as would-be readers that can access
the original, has become equally convenient. But there is no increased
incentive for those who read the original to also make verbatim copies,
especially from unreliable resources. 1 In the rest of this paper, giving
the benefit of doubt to potential nonreaders, we adopt a much more
generous view of a "reader" of a cited paper as someone who at the
very least consulted a trusted source (e.g., the original paper or heavily-
used and authenticated databases) in putting together the citation list.
As misprints in citations are not too frequent, only celebrated papers
provide enough statistics to work with. Figure 1 shows a distribution
of misprints in citations to one such paper [10] in the rank-frequency
representation, introduced by Zipf [11]. The most popular misprint in
1According to many researchers the Internet may end up even aggravating the copying
problem: more users are copying second-hand material without verifying or referring to
the original sources.
Read Before You Cite!
•
•
271
repeat number
~
~ 0.1
o
li.
0.01
0.001
0.0001
Fi~ure.2. Same data as in Figure 1, but in the number-frequency representation.
Mlspnnts follow a power-law distribution with exponent close to 2.
~ page number propagated 78 times. Figure 2 shows the same data, but
10 a number-frequency format.
~s a preliminary attempt, one can estimate an upper bound on the
ratio of the number of readers to the number of citers R as the ratio of
the number of distinct misprints D to the total number of misprints T.
Cle~rly, .am?ng T citers, T - D copied, because they repeated someone
else s mlspnnt. For the D others, with the information at hand we
~ave no evi~e~ce that they did not read, so according to the presu~ed
mnocent prmclple, we ~ssume that they did. Then in our sample, we
have D readers and T clters, which lead to:
D
R:::::-
T' (1)
Substituting D = 45 and T = 196 in equation (1) we obtain that
R ~ .0.23. :rhi~ estimate would be correct if the people'who introduced
ongmal mlspnnts had always read the original paper. However, given
th~ lo~ value o~ the upper bound on R, it is obvious that many original
mlsprmts were mtroduced while copying references. Therefore a more
careful analysis is neccessary. We need a model to accomplish i~.
O~r model ~or misprint~ propagation, which was stimulated by Si-
mon s explanation of the Zlpf law [12] and the idea of link redirection
by Krapivsky and Redner [4] is as follows. Each new citer finds the
referen~~ to the original in any of the papers that already cite it. With
p~ob.ablhty R he reads the original. With probability 1 - R he copies the
clt~tlon from the paper he found it in. In any case, with probability M
he mtroduces a new misprint.
Complex Systems, 14 (2003) 269-274
 M. V. Simkin and V. P. Roychowdhury Read Before You Cite! 273
272

The evolution of the misprint distribution (here N K denotes the num- and D = N x M x R. As a consequence, equation (1) becomes exact (in
ber of misprints that propagated K times, and N is the total number of terms of expectation values, of course). .
citations) is described by the following rate equations: The preceding analysis assumes that the stationary state had been
reached. Is this reasonable? Equation (5) can be rewritten as:
dN 1 = M _ (1 - R) x (1 - M) x NN1 ,
dN d(~) _ dl N (9)
(K-l)xNK -KxNK M - (~) x (R + M - M x R) - n .
dN
_K=(l-R)x(l-M)x -1 (K>l). (2)
dN N .
As long as M is small it is natural to assume that the first citation was
These equations can be easily solved using methods developed in [4] to correct. Then the initial condition is N = 1; T = O. Equation (9) can be
get: solved to get:
1
'Y = 1 + (1 _ R) x (1 - M)'
(3) T=Nx
R+M-MxR
M x (
1-~~~~ 1)
NR+M-MxR .
(10)

As the exponent of the number-frequency distribution 'Y is related to the
exponent of the rank-frequency distribution a by a relation 'Y = 1 +(1/a),
equation (3) implies that:
(4)
a = (1 - R) x (1 - M).
The rate equation for the total number of misprints is:
dT T (5)
dN = M + (1 - R) x (1 - M) x N'
The stationary solution of equation (5) is:
M (6)
T=Nx R+M-MR'
The expectation value for the number of distinct misprints is obviously
(7)
D=NxM.
From equations (6) and (7) we obtain:
D N-T (8)
R=TxN_D'
Substituting D= 45, T = 96, and N = 4300 in equation (8),.we obt~in
R ~ 0.22, which is very close to the initial estimate obtamed usmg
equation (1). This low value of R is consistent with the "Principle of
Least Effort" [11].
One can ask: Why did we not choose to extract R using equations (3)
or (4)? This is because a and 'Yare not very sensitive to R when it is
small. In contrast, T scales as 1/R.
We can slightly modify our model and assume that original misprints
are only introduced when the reference is derived from the original
paper, while those who copy references do not introduce new misprints
(e.g., they cut-and-paste). In this case one can show that T = N x M
This should be solved numerically for R. For our guinea pig equa-
tion (10) gives R = 0.17. '
Just as a cautionary note, equation (10) can be rewritten as:
T
D
=!x X(l- ~).
NX '
x=R+M-MxR. (11)
The definition of the natural logarithm is:
. aX - l
lna=hm--.
x...o x
Comparing this with equation (11) we see that when R is small (M is
obviously always small):
T
D ~ InN. (12)
This means that a naive analysis using equations (1) or (8) can lead to
an erroneous belief that more cited papers are less read.
One can augment our results with a closer scrutiny of the data. In
order to make sure that misprints have not been introduced by the lSI
as it sometimes happens [13], we explicitly verified a dozen misprinted
citations in the original articles. All of them were exactly as in the lSI
database. There are also occasional repeat identical misprints in papers,
which share individuals in their author lists. Such events constitute a
minority of repeat misprints. It is not obvious what to do with such
cases when the author lists are not identical: Should the set of citations
be counted as a single occurrence (under the premise that the common
co-author is the only source of the misprint); or as multiple repetitions?
However, even if we count all such repetitions as only a single misprint
occurrence, then the number of citation-copiers (i.e., T - D) shall drop
from 151 to 112, bringing the upper bound for R (equation (1)) from
23% up to 29%. However a more detailed analysis via our model
Complex Systems, 14 (2003) 269-274
 M. V. Simkin and V. P. Roychowdhury
274
[14] will bring down the estimate closer to 20%, keeping the original
conclusions unaltered. . d'
We conclude that misprints in scientific citations .shou~d not be IS-
carded as a mere happenstance, but, similar to FreudIan slips, analyzed.
I Acknowledgments
We are grateful to J. M. Kosterlitz, A. V. Melechko, N. Sarshar, H. Muir,
and many others for correspondence.
I References
[1] Z. K. Silagadze, "Citations and the Zipf-Mandelbr~t Law," Complex
Systems, 11 (1997) 487-499; http://arxiv.orglabslphyslcsl9901035.
[2] S. Redner, European Physics Journal B, 4 (1998) 131-134;
http://arxiv.orglabs/cond-matl9804163.
[3] C. Tsallis, and M. P. de Albuquerque, European Physics Journal B, 13
(2000) 777-780; http://arxiv.orglabslcond-matl9903433.
[4] P. L. Krapivsky and S. Redner, Physical Review E, 63 (2001) Art. No.
066123; http://arxiv.orglabslcond-matl0011094.
[5] H. Jeong, Z. Neda, and A.-L. Barabasi,
http://arxiv.orglabslcond-matl0104131.
[6] A. Vazquez, http://arxiv.orglabs/cond-matl0105031.
[7] H. M. Gupta, J. R. Campanha, and B. A. Ferrari,
http://arxiv.orglabslcond-matlOl12049.
[8] S. Lehmann, B. Lautrup, and A. D. Jackson,
http://arxiv.orglabs/physicsl0211010.
[9] S. Freud, Zur Psychopathologie des Alltagslebens (Internationaler psycho-
analytischer Verlag, Leipzig, 1920).
[10] Our guinea pig is the Kosterlitz-Thouless paper (J. M. Kosterli~z a~d
D. J. Thouless, Journal of Physics C,6 (1973) 1181-120.3): The mlsprmt
distribution for a dozen other studied papers look very SImilar.
[11] G. K. Zipf, Human Behavior and the Principle of Lea~t Effort: An Intro-
duction to Human Ecology, (Addison-Wesley, Cambndge, MA, 1949).
[12] H. A. Simon, Models of Man (Wiley, New York, 1957).
[13] A. Smith, New Library World, 84 (1983) 198.
[14] M. V. Simkin and V. P. Roychowdhury, to be published.
Creating Large Life Forms with Interactive Life
William H. Paulsen
Department of Mathematics and Statistics,
Arkansas State University,
State University, AR 72467
This paper demonstrates how very complicated Life forms can easily be
created using the interactive Life program introduced by James Gilbert
[1]. By having control over just a single cell, called the intelligent cell,
a glider gun can be created in under 250 generations. Note that the
standard rules of Life apply to the intelligent cell as well as the other cells.
Ever since John Conway introduced the game of Life in 1970 [2], ama-
teur and professional mathematicians have been obsessed by finding new
complicated structures, and finding new applications for the seemingly
chaotic behavior of Life forms. Perhaps its addictive nature stems from
the simplistic rules: each cell on a square lattice is either on (alive) or off
(dead). The Moore neighborhood of a cell is the eight surrounding cells
(counting diagonals) [3]. If a dead cell has exactly three neighboring
cells which are alive, then the cell becomes alive in the next genera-
tion. On the other hand, a living cell must have either two or three
living neighbors to remain alive, otherwise the cell will die in the next
generation.
Figure 1 shows four generations of the Life form known as the
"glider." Because every four generations, this glider moves one space
diagonally, we say that this glider moves at 1/4 the speed of light. (In
the Life model, light is said to travel at 1 square per generation.)
The first open problem that Conway asked was whether any Life
form could be proven to have unbounded growth. This was solved by
Bill Gosper's discovery of the glider gun, a formation that spews a glider
every 30 generations [4]. Other glider guns have since been discovered,
but none have been as efficient as the original, now referred to as the
"p30 glider gun. "
Using glider guns, one in fact can engineer even more complex Life
forms. Any computer circuit, and even a Turing machine, can be con-
II- II- 11-- 11--
91
- •• • •
Figure 1.
••
Four generations of a glider.
• • •
Complex Systems, 14 (2003) 275-283; © 2003 Complex Systems Publications, Inc.