Está en la página 1de 19

Knowledge Connectedness in Geometry Problem Solving

Author(s): Michael J. Lawson and Mohan Chinnappan


Source: Journal for Research in Mathematics Education, Vol. 31, No. 1 (Jan., 2000), pp. 26-43
Published by: National Council of Teachers of Mathematics
Stable URL: http://www.jstor.org/stable/749818 .
Accessed: 17/02/2011 17:44

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=nctm. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

National Council of Teachers of Mathematics is collaborating with JSTOR to digitize, preserve and extend
access to Journal for Research in Mathematics Education.

http://www.jstor.org
Journalfor Researchin MathematicsEducation
2000, Vol. 31, No. 1, 26-43

Knowledge Connectedness in
Geometry Problem Solving
Michael J. Lawson, Flinders Universityof SouthAustralia
MohanChinnappan,Universityof Auckland,New Zealand

Ourconcernin this studywas to examinethe relationshipbetweenproblem-solvingperformance


andthe qualityof the organizationof students'knowledge.We reportfindings on the extent to
which content and connectednessindicatorsdifferentiatedbetween groupsof high-achieving
(HA) andlow-achieving (LA) Year 10 studentsundertakinggeometrytasks.The HA students'
performanceon the indicatorsof knowledgeconnectednessshowedthat,comparedwith the LA
group,they could retrievemoreknowledgespontaneouslyandcould activatemorelinks among
given knowledgeschemasandrelatedinformation.Connectednessindicatorswere moreinflu-
ential than content indicatorsin differentiatingthe groups on the basis of their success in
problemsolving. The tasks used in the studyprovidestraightforwardways for teachersto gain
informationaboutthe organizationalqualityof students'knowledge.

Key Words:Assessment; Geometry;Knowledge; Problemsolving; Secondarymathematics;


Teachingpractice

A majoraim of mathematicseducationis to devise ways of encouragingstudents


to take more active roles in acquiring,experimentingwith, and using the mathe-
matical ideas and proceduresthat are includedin the school curriculum.Hiebert
et al. (1996) have recentlyinterpretedthis aim as meaningthat studentsin math-
ematicsclasses"shouldbe allowedandencouragedto problematizewhatthey study,
to define problemsthatelicit theircuriositiesand sense-makingskills"(p. 12). In
statementson mathematicsteaching,teachershave been askedto help studentsto
"developmultiplerepresentationsand connections,and constructmeaningsfrom
new situations"(NationalCouncil of Teachersof Mathematics,1989, p. 125). It
is arguedthat the betterthe quality of the students' problematizingand of their
knowledgeconnections,the morepowerfulwill be the knowledgerepresentations
thatcan be called upon duringa problem-solvingepisode.
Researchon the use of self-explanationprocessesduringthe studyof new infor-
mationin the areasof computing,mathematics,and science provides supportfor
these views, emphasizingthe key role of encodingprocesses in influencingsubse-
quent problem solving (e.g., Bielaczyc, Pirolli, & Brown, 1995; Chi, Bassok,
Lewis, Reimann,& Glaser,1989;Chi,De Leeuw,Chiu,& LaVancher,1994;Renkl,

This research was supportedby grants from the research budgets of the Queensland
Universityof TechnologyandFlindersUniversityandthe AustralianResearchCouncil.The
authorsacknowledge the cooperationand assistance of the staff and studentsof St Peters
College, Brisbane.The commentsof reviewersof the initial version of the articleare grate-
fully acknowledged.
Michael J. Lawson and Mohan Chinnappan 27

1997). These studieshave shown thatthe more detailedandfocused the construc-


tive activity generated by the students' ongoing self-explanations, the more
successful have been their laterproblem-solvingactivities. What this self-expla-
nation researchdoes not provide is informationaboutthe organizationalstate of
the knowledgerepresentationthatcan be drawnuponduringproblemsolving. The
focus of this article is an investigation of the relationshipbetween success in
problemsolving andthe qualityof the knowledgeconnectionsthathavebeen devel-
opedby studentswhentheyhavetriedto makesense of theirmathematics.In partic-
ularwe developedproceduresto illustratethe connectednatureof students'knowl-
edge andexaminethe relationshipbetweenthe extentof knowledgeconnectedness
and problem-solvingperformance.
In contemporarystudies of cognition, the connected natureof memory is well
established (e.g., Anderson, 1995; Derry, 1996). Although there are different
views of the basis for the connectionsin memory,thereis substantialacceptance
of the view that memory is an associative structure.In Anderson's (1990) two-
concept theory of memory, componentsof this associative structureare concep-
tualizedas varyingin both statesof activationandstrength.Activationrefersto the
momentaryavailabilityof a knowledgecomponent,whereasstrengthdescribesthe
durabilityof the knowledge componentover the long term(Anderson,1990). The
two characteristicsof a particularknowledge componentare arguedto vary inde-
pendently.Knowledgecomponentsthatarein a high-activation,high-strengthstate
are expected to be readily accessed during problem solving. Although high-
activationcomponentsshould be easily accessed, high-strengthcomponentsthat
are low in activationmight not be accessed so readily. Componentsthat are low
in both activationand strengthare not likely to be accessed. Anderson's discus-
sion of strengthand activationis useful here because it suggests that researchers
attemptingto gain informationabout the state of organizationof the knowledge
base need to probe the natureof the connections among knowledge components
in a mannerthatprovides opportunitiesfor the studentto access componentsthat
are low in activationlevel. If such probingdoes not occur, statementsabout the
state of connectedness of knowledge may be based on incomplete information
because knowledge thatis availableto the studentmay remaininert and is likely
to be regardedas missing from the student'sknowledge base.
The failure of some studentsto access available knowledge at the appropriate
time duringthe solutionattempthas been discussedby Bransford,Sherwood,Vye,
and Rieser (1986) and by Prawat (1989), and our own work (Lawson &
Chinnappan,1994) providedan example of its existence in mathematicalproblem
solving. We showed thata groupof less successful problemsolvers at high school
level failed to use a substantialbody of theiravailableknowledgeduringattempts
to solve geometryproblems,yet they could access thatknowledgewhen prompted
to do so. Importantcomponents of knowledge remainedinert in these students
until they were given cues by the researcher.This failureto access relevantavail-
able knowledge was less common among the successful problem solvers we
observed.
28 KnowledgeConnectedness

We contended (Chinnappan& Lawson, 1994) that failure to access available


knowledge might arise from three aspects of students' processing activity: the
students'dispositionalstates,the strategicnatureof theirmemory-searchactivity,
and the quality of organizationof the knowledge relevantto the problemsbeing
considered.Thus, by way of illustration,access failure might result from one or
more of the following problems:lack of persistencewith the solutionattemptdue
to low self-efficacy, ineffective use of cues providedin the problemstatement,or
lack of stronglyconnectedknowledgerelevantto the problem.In this researchwe
set out to gather evidence that could be used to examine the last of our three
contentions.We reportthe resultsof a study thatwas designed to providefurther
informationaboutthe relationshipbetween indicatorsof the knowledge states of
students and their problem-solvingperformance.This relationshipis examined
throughcomparisonof the performanceof groups of high-achieving and low-
achievingstudents.In the design of the studywe have made a distinctionbetween
performancemeasuresthatindicatewhatproblem-relevantknowledgeis available
to the studentand measures that allow us to draw inferences about the state of
connectednessof that knowledge. We refer to these measuresas content indica-
tors andconnectednessindicators,respectively;we next providea rationalefor the
use of these indicators.

ContentIndicators
The most commonly used indicatorsof the states of students'knowledgebases
are what studentsdo and what they say during a problem-solvingepisode. The
students' writtenand verbal actions provide informationabout availableknowl-
edge, andclassroomteachersandresearchersuse these actionsto makeinferences
aboutknowledge states.At all levels of education,teachers'analyses of problem-
solving behavior depend heavily on evidence gatheredfrom students' written
actions.If studentsmakeappropriatemoves in theirwrittenactions,we makejudg-
ments that "they know this" or "they can use this procedure."These judgments
might be made aboutrelatively simple knowledge schemas such as a right-angle
schema,for which we mightuse the markingof a rightangle as an indicatorof the
schema, or they might be aboutmore complex schemas.
A student's verbal statementsduring a solution attemptcan also be used as
evidence of availableknowledge, thoughthese statementsare often not available
to theteacher.A teacherwho is markinga student's homeworkor examinationscript
mightat timeswish thatthe studentwerepresentto explaina particularmove ("Why
did you do this?")because the verbal explanationmight reveal somethingmore
aboutthe student'sknowledge state thancan be identifiedin the writtenactions.
However, when most markingis being done, studentsare absent.
Researcherscan moreeasily gain access to students'verbalactionsby requiring
studentsto talk while they solve problems.Althoughuse of such think-alouddata
is not unproblematic(see Payne, 1994), these data do provide a rich source of
knowledge for makingjudgmentsaboutknowledge states (e.g., "Becausethat is
Michael J. Lawsonand Mohan Chinnappan 29

a right-angledtriangleand I know the length of side BC, I can now calculate the
lengthofAC"). Inferencesaboutcontentarealso basedon students'recallorrecog-
nition of particularknowledge components, either without assistance from the
teacheror in response to a cue or hint. A studentmight be asked to freely recall
what is known about a theoremor a proof in geometry or to identify key terms,
partsof a diagram,or possible solution paths.
Although these tasks provide information about the knowledge that can be
accessed by the student,they are of limited use as indicatorsof how that knowl-
edge is organized.Usually, the recall or recognitiontasks provide evidence only
aboutthe student'sknowledge in a discreteform and do not requirethe studentto
show relationshipsbetweena knowledgecomponentandotherrelatedcomponents.
In this situationthe researcheragain lacks direct and extensive evidence of how
the student'sproblem-relevantknowledge is organizedand differentiated.More
sensitive procedures,throughwhich one can examine the organizationalrelation-
ships among knowledge components and knowledge schemas, are needed. We
discuss examples of these in the next section.

Connectedness Indicators
A range of procedureshave been used to representthe structurednatureof
knowledge, andthese have demonstratedthe positive relationshipbetweenknowl-
edge organizationand problem-solvingperformance.Deese (1962) reviewed the
use of word-associationproceduresthathadbeen used to illustratethe associative
structureof verbalmemory. The majorinterestin this work was to representthe
frequencyandthe patterningof verbalassociations.Patternsof responsewererepre-
sented by Deese throughuse of factoranalysis. In this work therewas an explicit
concernto representthe organizationalstructureof verbalmeaning,andthese same
procedureswere used by others(e.g., Johnson, 1965) to examine the relationship
between knowledge organizationand problem-solvingperformance.Johnson's
researchhas relevanceherebecausethe organization-performance relationshipwas
examined within the domainof physics problemsolving and, althoughtherewas
some variationin the patternof results,word-associationperformancewas related
to the level of problem-solvingperformance.
A numberof mappingprocedureshave also been used to representfeaturesof
knowledge organization.Concept-mappingprocedures,such as those developed
by Novak and Gowin (1984) and by McKeown and Beck (1990), have been used
for this purpose,principallyto establishthe existence of, and labels for, the links
that studentshave establishedamongknowledge components.Attemptsto repre-
sent the structureof these concept maps in quantitativetermshave not been very
successful (Lawson, 1994). Othermappingprocedureshave more readilyyielded
quantitativeinformation.Shavelson (1972) used digraphsas the basis for gener-
ating distancematricesthatwere used in the analysis of changes in relatednessof
students'cognitive structuresfollowing instruction.Naveh-Benjamin,McKeachie,
Lin, andTucker(1986) arguedagainstthe use of distancematricesanddeveloped
30 KnowledgeConnectedness

a representationof cognitive structureusing an ordered-treeprocedurethatgener-


ated measuresof organization,levels of organization,and similarity.In all these
mapping procedures,measures of organizationwere positively associated with
students'achievement.
Naveh-Benjaminet al. (1986) criticizedthe use of distancematriceson the basis
thatthey requiredthe use of measuresthatwere somewhatremovedfrom the data
generated by the student, ignored the inert-knowledgeproblem, and failed to
reflectthe dynamicnatureof knowledgeorganization.However,it is not clearthat
use of the ordered-treerepresentationprovides an adequatesolution for all these
problems.In Naveh-Benjaminet al.'s procedure,the datawere generatedfromuse
of a set of discreteconcepts. This choice seems to limit the potentialfor the tech-
nique to provide rich informationabout the dynamic structureof knowledge,
particularlythe structureof knowledge while it is being used. In addition,how the
ordered-treeprocedure directly addresses the problem of inert knowledge is
unclear, inasmuch as the student's control over activation of knowledge was
removedwhen the studentwas providedwith a list of concepts. For these reasons
otherindices of knowledge connectednesswere used in this study.
Response time has a long historyof use as an indicatorof knowledge organiza-
tion. In general it is assumedthat the longer the response time, the less strongly
related and the less accessible is the problem-specific knowledge (Anderson,
1990).In this sense,thismeasurecanbe arguedto providemorespecificinformation
aboutthe stateof particularknowledgecomponentsthanthe accessible/notacces-
sible informationprovidedby the contentindicators.Response-timeinformation
does allow for more sensitive comparisonsbetween characteristicsof knowledge
components,anda response-timemeasurewas developedfor this study.The focus
of this measure was on time needed for recognition of knowledge components
presentedin diagrams,such contexts being centralto the students'use of knowl-
edge in geometry.However,on theirown, response-timemeasurescanprovideonly
an incomplete indication of knowledge organization.Although response times
provide informationaboutthe ease of access to schemas and so move us beyond
absence/presencejudgments,they do not focus directlyon the details of relation-
ships in particularknowledge configurations. Other measures can be used to
providemoreinformationaboutthe connectionsbeing constructedby the learner.
Even thoughthe contentindicatorsdescribedabove can provideinformationof
what knowledge is in a high state of activation,they are unlikely to indicate the
extent of knowledge that is not highly active, that might remain inert. It is this
knowledge that may be accessed by students during systematic prompting.
Campione,Brown, and Ferrara(1982) developed a gradedhinting task that was
designedto providea measureof whichknowledgeschemascouldbe accessedand
used on transfertasks.We (Lawson& Chinnappan,1994) adaptedthis taskfor use
in examiningthe levels of connectednessof knowledge schemas used in solving
simple geometryproblems.By providingstudentswith increasinglevels of cueing
support,one can use the hinting task to index the level of what Mayer (1975)
referredto as the internalconnectednessof a schema.A schemawith components
Michael J. Lawson and Mohan Chinnappan 31

thatareeffectively organizedis one for whichminimallevels of cueingarerequired


for activation.When a greaterlevel of hinting supportis needed for access, we
arguedthatthe knowledge schema is eitherless extensive or less well-connected.
Use of this measureprovidesinformationaboutthe relatednessof componentsin
a knowledgeschema,informationthatis not availableusingpresence/absenceindi-
cators.When studentsin ourstudywere systematicallypromptedwith gradedhints
after a solution attempt, they accessed a further substantialbody of problem-
relevantknowledge thathad not been accessed duringthe solution attempt.
In our study of geometryproblemsolving (Lawson & Chinnappan,1994), we
developed two other tasks to examine relationshipsamong knowledge schemas,
or what Mayer (1975) referredto as external connectedness. Both tasks were
designedto examine students'use of relevantschemas.The firstrequiredstudents
to use a schemato developa sampleproblem,the solutionof whichwouldcall upon
use of that schema. This task requiredstudentsto move beyond simple accessing
of a schemato embed the complete schemain an appropriateproblemframework.
This applicationtask did not, however, requirethe studentsto directlyrelate the
targetschemato otherrelatedschemas.Evidenceof connectionsamonggeometry-
relatedschemaswas requiredin the elaborationtask.Withthistaskwe investigated
how differenttheorem schemas that were relevantto a particularproblemcould
be related,one to the other. In both these external-connectednesstasks, students
were required to work on their own to make connections within and among
schemas.
The recognition,hinting,application,andelaborationtasksprovideinformation
thatmoredirectlyindexesthe stateof organizationof knowledgethando thecontent
indicators generated from observations of problem-solving and recall perfor-
mance.The fourformertypes of taskscan be used to provideindicatorsof the ease
of access of differentknowledge components,of the amountof supportrequired
to facilitatethat access, and of the internaland externalconnectednessof knowl-
edge schemas. We contend that effectively organizedknowledge will be readily
accessed and more richly connected,internallyand externally.If this assertionis
correct, then differences between less successful and more successful problem
solvers on organizationindicatorsare likely to be substantial.We designed this
study to provideevidence relevantto a test of these arguments.
In this studywe comparedthe performanceof groupsof high- andlow-achieving
studentson sets of contentandconnectednessindicators.The comparisonbetween
groups differing in level of problem-solvingperformancewas set up to facilitate
the investigationof the influence of the two sets of indicatorson problem-solving
performance.On the basis of previous research (e.g., Lawson & Chinnappan,
1994), we predictedthatthe groupswould differ on both sets of indicatorsbut that
the differencesbetween them would be greateron the connectednessindicators.
32 KnowledgeConnectedness

METHOD

Participants
The participantswere 36 Year 10 male studentsfrom a privatecollege in metro-
politanBrisbane,Australia;these studentsvolunteeredto participatein the study.
In this college, studentswere streamedinto differentclasses on the basis of their
performancein Year 9 and Year 10 mathematicstests. The college curriculum
requiredthat all studentscomplete a topic involving trigonometryand geometry
duringYears 8, 9, and 10. At the time of this study,all the studentshadcompleted
this topic. High-achievingstudents(HA: n = 18) came from the uppertwo Year
10 streams.The low-achieving students(LA: n = 18) came from the threeclasses
of the lower streams.

Procedure
All studentsparticipatedindividuallyin two 60-minutesessions. Duringthe first
session, studentswere requiredto complete four tasks: the Free Recall Task, the
ProblemSolving Task,the GeometryComponentsTask,andthe HintingTask.The
structureof these tasks andthe proceduresused to score students'responseswere
the same as those we had used earlier(Lawson & Chinnappan,1994). The Free
Recall Task (Recall) requiredstudentsto identify known geometrytheoremsand
formulas.Studentswere askedto recallanygeometrytheoremsthattheyknew, and
they were told that they could identify the theoremsby verbal and writtenstate-
mentsor throughuse of diagrams.If high-achievingstudentsdevelop more effec-
tively organizedgeometricschemas,we should expect them to be able to retrieve
more extensive bodies of within-schema knowledge in a free-recall situation.
Performanceon this task will not, however, isolate the reason for this outcome:
Better recall performancecould reflect the existence of either more extensive
available knowledge or more effective recall of available knowledge. The score
for this task was the numberof theoremsrecalled or demonstrated.
The ProblemSolving Task consisted of four plane geometryproblemsthatcan
be solved by the use of theoremsand formulasthat are taughtin the first 3 years
of the high school mathematicscurriculum.One of the problems is shown in
Figure 1. The task provideda sample of students'problem-solvingperformance
duringwhich theiraccessingof problem-relevantknowledgewouldbe cued by the
problem statementsand by their own problem-solvingactions. This observation
of performancewas necessaryto providean estimateof students'knowledgeacti-
vation when they workedunaidedon typical problems.A student'sperformance
on the problemswas scored using a 3-point scale (2, 1, or 0 points scored); the
middlescorereflectedpartialcreditfor a solutionattemptthatinvolvedappropriate
moves but was incomplete.
The GeometryComponentsTask was developed to examine students'knowl-
edge of partsof geometricfigures andof the theoremsor rules thatarerepresented
by these figures. Studentswere shown figures related to the problem shown in
Michael J. Lawsonand Mohan Chinnappan 33

AEis a tangentto the circle,centreC.


AC is perpendicularto CE,and angleDCE B
has a measure of 300.
The radiusof the circleis equalto 5 cm.
FindAB.
C E

Figure1. Problem4 usedin theProblemSolvingTask.

Figure 1 and were requiredto identify the parts of the figure (Forms) and to
producea rule or theoremthatwas illustratedby the figure (Rules). Studentssaw
one figure at a time and were shown five figures duringthis task, one of which is
shown in Figure 2. The rule or theorem associated with this figure was "The
tangentto a circle is perpendicularto the radiusat its point of contact."The score
for this task was the numberof correctidentificationsof forms and theorems.

B ABCis a straightlinetouchingthe
circleat B. 0 is the centre.

Task.
Figure2. Figureusedin theGeometryComponents

In the HintingTask studentswere providedwith a sequence of gradedhints on


the basis of a commonly adoptedsolution path for three of the problems. (Hints
were not given for the otherproblembecause the elements of that problemwere
34 KnowledgeConnectedness

used as cues in the programfor the Recognition Task.) When studentsfailed to


producethe complete solution for one of these problems,they were requiredto
attemptto solve thatproblemwith the help of hintsgiven by the investigator.Each
hintwithina sequenceprovideda studentwithan increasedlevel of assistance.The
initialhintsdrewthe student'sattentionto a partof the problem,andthe researcher
waited to see if that would lead the student to generate any furtherproblem-
relevant information.If students needed furtherhints, they might be asked to
attendto the markspresenton specific lines in a diagram;the final hint would be
a "give-away"hintthatshowed a methodof solutionfor the problem.The number
of hints requiredconstituteda student'sscore on this task. Any studentwho was
providedwith the give-away hint was regardedas not havingfunctionalaccess to
thatparticularknowledge component.An exampleof a sequenceof hints is given
in Table 1.

Table1
Exampleof a Sequenceof Hints Used in the Hinting Task
Level Hint
1. Whatdo younoticeabouttriangleABC?
2. Whatdo younoticeaboutlinesACandBC?
3. Whatdoesthequestionstatement tellyouaboutlinesACandBC?
4. LinesACandBCareof equallength.
5. WhatcanyousayabouttriangleABC?
6. TriangleABCis isosceles.AnglesBACandABCareequal.

Duringthe second session, studentswere requiredto complete threetasks: the


RecognitionTask,the GeometryApplicationTask(Application),andthe Geometry
ElaborationTask(Elaboration).The RecognitionTaskdevelopedfor thisstudywas
computer-presentedand computer-controlledand was based on HyperCardsoft-
ware that recordedthe time taken by a studentto correctlyidentify a particular
geometricformor relationship.Detailsof theprogramareprovidedby Chinnappan,
Lawson,andGardner(1998). A samplescreenfromthe programis shownin Figure
3. This task involved students in identifying the names of selected features of
geometricalforms thatwere displayedon the computerscreen.Studentsindicated
recognitionof a componentby clicking on thatcomponent,andthen they typedin
the nameof the component.The timetakenfortypingwas not includedin therecog-
nition time. Studentswere instructedto work quickly and accuratelyin making
recognitiondecisions. Each display in this HyperCardprogramwas developedto
representa geometric schema commonly taught in the classroom, for example,
right-angledtriangleand its properties.Figures with multiple componentswere
cycled throughthe presentationformat,with the recognitiontime being recorded
from time of presentationof the figureuntil the studentsignaledrecognition.Only
times for correctrecognitionswere used in the analysis for this article.The scores
Michael J. Lawson and Mohan Chinnappan 35

used for this indicatorwere the mean recognitiontimes for correctlyrecognized


components.Studentsdid not receive any assistancefrom the programdisplay or
from the researcherduringrecognitionof these components.

rightangle

Figure3. SamplescreenfromRecognitionTask.

In the ApplicationTask studentswererequiredto reportor illustratein a diagram


an example of use of each of five theoremsor formulas.In undertakingthis task,
the students were requiredto display knowledge of links among components
within schemas associatedwith the theoremsor formulas.If studentswere unable
to producean example,they were providedwith simpleproblems(like the problem
in Figure4) to solve. Responses were scored on a scale from 0 to 5: A score of I
was awarded for a spontaneous though incorrect application, 2 for a correct
response aftera cue, 3 for a correctspontaneousresponserelatedto the rule, 4 for
a single, correctnovel application,and a score of 5 was awardedif more thanone
correctnovel applicationwas generated.

O O is the centreof the circle.AB is a


tangent to the circle at B. Angle OAB = 300.
FindangleAOB.

Figure4. Figureusedin Application


Task.
36 KnowledgeConnectedness

In the ElaborationTask,the investigatorpresentedpairsof theoremsor formulas


to the students,one pair at a time. Studentswere requiredto generatea problem
thatinvolveduse of boththeorems.One pairof problemsused in this taskis shown
in Figure 5. This task was designed to provide an estimateof the extent to which
students could establish and exploit connections among related schemas. The
score for this taskrangedfrom0 to 4: A score of 1 was given for a partiallycorrect
connection, 2 for a correct"basic"connection, 3 for a correctnovel connection,
and a score of 4 was awardedif more than one correct novel connection was
providedby the student.

Theorem1: The perpendicular drawnfromthe centreof a


circleto its chordbisects the chord.
Theorem2: Pythagoras'theorem.

Figure5. Figureusedin Elaboration


Task.

RESULTS

Scores for the two groupsof studentson each of the tasks areshown in Table 2.
As was expected, given the design of the study, the groupsdiffered significantly
in performanceon the ProblemSolving Task.
The performancesof the two groupson each of the sets of contentand connect-
edness indicatorswere comparedusing separateone-way multivariateanalysesof
variance.Because the F values for both sets of indicatorswere significantbeyond
the .05 level, the initial analyses were followed up with univariatet tests of the
differences between group means on each indicatorwithin a set. In judging the
significance of the individual univariatecomparisons, we made a Bonferroni
adjustment,so thatthe alphalevels set for significancewere .017 and .012 for the
contentand connectednessindicators,respectively (Stevens, 1996, p. 160). The t
values and effect sizes for each comparisonare also shown in Table 2.

ContentIndicators
The differencebetween the groupsfor the set of contentindicatorswas statisti-
cally significant(MultivariateF(3, 32) = 3.72,p < .03), suggesting,in generalterms,
that the HA group was able to spontaneouslyaccess a wider range of problem-
relevantknowledge. The univariatecomparisonssuggest thatit was performance
on the Rules Task thatcontributedto the multivariatesignificantdifferencefound
Michael J. Lawsonand Mohan Chinnappan 37

Table2
DescriptiveStatisticsand UnivariateTest Resultsfor All Indicators
Highachieving Lowachieving Univariate
Task (n = 18) (n = 18) t value
(Possiblescore) M SD M SD (pvalue) Effectsize
Problems 4.72 2.02 2.39 1.91 3.55
(8) (.001)
Content
indicators
Recall 10.83 4.48 7.06 5.77 2.19 0.65
(Open) (.036)
Forms 14.17 2.23 12.78 3.59 1.39 0.39
(24) (.172)
Rules 4.67 0.76 3.56 1.25 3.22 0.89
(5) (.003)
Connectedness
indicators
Hinting 13.06 9.74 21.06 10.25 3.06 -0.78
(Open) (.004)
Application 20.67 5.57 16.89 3.80 2.38 0.99
(25) (.024)
Elaboration 8.06 4.09 4.39 2.89 3.10 1.27
(12) (.004)
Recognition 7.87 2.50 11.63 5.80 2.52 -0.65
time (.017)
(seconds)

between the groups on these indicators.Comparisonof the effect sizes indicates


thatthe differencebetweenthegroupson the RulesTaskwas greaterthanthateither
for theirfree recall performanceor for the accuracyof recognitionof geometrical
forms.
The patternof performanceon these tasks suggestedthatthe differencebetween
the groups in terms of content was not simply in ability to recognize the simple
geometrical forms that provide the basis of knowledge relevant to this area of
problemsolving. Instead,differencesbetween the groupswere more apparentin
the more complex relationshipsrepresentedin the Rules Task.

ConnectednessIndicators
The multivariatetest of differencebetweenthe groupson the connectednessindi-
catorswas also significant(MultivariateF(4, 31) = 4.52, p < .01). We interpretthis
differenceas pointingto a superiorityin organizationof the knowledgeof the HA
group.This superiorityin organizationwas reflected in relative performanceson
each of the indicatorsdesigned to reflect the facility and extent of connectedness
of the students'knowledge bases relevantto this areaof geometricalknowledge.
The effect sizes relatedto the comparisonsof the groupson these indicatorswere
generallylargerthanthosefor thecontentindicators.Whenthe individualunivariate
comparisonswere considered,the t values for both the Hinting and Elaboration
38 KnowledgeConnectedness

comparisons were significant at the adjusted alpha level. The t value for the
Recognitioncomparisonwas slightlyoutsidethis adjustedalphalevel. In each case
the HA group performancereflected use of a knowledge base that was charac-
terizedby betterqualityknowledgeconnections.The HA grouprequiredless assis-
tance in the form of graded hints to access relevant knowledge. This group
requirednot only fewer hints to access such knowledge but also fewer give-away
hints (HA: 0.2 hints; LA: 1.9 hints), though this difference was not statistically
significant (t(34) = -1.96, p > .05).
The HA studentsalso showedgreaterevidenceof externalconnectednessamong
schemas in the ElaborationTask. Their performanceon the Recognition Task
suggestedthatthey mightalso be able to morequicklyactivateknowledgecompo-
nentsthatwere relevantto the selected areaof problemsolving. The HA groupnot
only had lower meanrecognitiontimes but also made a greaternumberof correct
recognitions.

DiscriminantAnalysis
A differentperspectiveon the influenceof the two sets of indicatorson theperfor-
manceof the groupscanbe gainedthroughuse of descriptivediscriminantanalysis.
In this case the purposeof the analysis was to gain informationaboutwhich indi-
cators were most importantin predicting the membership of the two groups
observed in this study. In particularthe focus of interesthere was in the relative
contributionto predictionsof groupmembershipof the contentandconnectedness
indicators.Forthis analysisall the contentandconnectednessindicatorswere used
as predictorsandwere enteredinto a directdiscriminant-analysisprocedureusing
the SPSS Discriminantprogram.Because the use of the seven indicatorswith the
availablesamplesize is close to the minimumrecommendedcase/variableratiofor
discriminantanalysis, the results should be taken only as suggestive of possible
strengthsof influence of the indicators.
Discriminantanalysis producedone significantdiscriminantfunction (Wilks's
Lambda= 0.57, X2(7,N= 36) = 16.95,p < .02). The structurecoefficients(discrim-
inantloadings)andstandardized weightsforeachvariableareshownin Table3. The
structurecoefficientsarecorrelationsbetweenthevariablesandthediscriminant func-
tion, similarin natureto factorloadingsin factoranalysis,andaregenerallyfavored
as indicatingthe contributionof each variableto the discriminantfunction(Stevens,
1996). Thompson(1998) arguedthatboth structurecoefficients and standardized
weights mustbe inspectedto makea judgmentaboutthe contributionof a variable
to the discriminantfunctionbecause a standardizedweight nearzero (see weights
for HintingandRecall) does not necessarilyindicatethata variableis unimportant.
Theresultsin Table3 suggestthateachof thepredictorsmadea contribution in differ-
entiatingthe two groupsof studentson the basis of theirproblem-solvingperfor-
mances,althoughthecontribution of theFormsscorewas lowestin thisanalysis.Apart
fromtheRules score,the connectednessindicatorscontributemorestronglyto sepa-
rationof the groupsthando the remainingcontentindicators.
Michael J. Lawson and Mohan Chinnappan 39

Table3
DiscriminantAnalysis Coefficients
Discriminantanalysiscoefficients
Indicator Structurecoefficient Standardized weight
Rules .64 .71
Elaboration .62 .54
Hinting -.61 -.02
Recognition -.50 -.58
Application .47 .15
Recall .44 -.01
Forms .28 -.48

CONCLUSIONAND DISCUSSION

Ourconcernin this study was to develop a detailedpictureof the natureof the


knowledge representationsand connections that are developed by studentsas a
result of study in the area of geometry. Such a descriptionshould be of use to
teachers and researcherswho are seeking furtherunderstandingof the reasons
behindeffective, andineffective, problem-solvingperformance.The thrustof our
approachwas to seek to examinethe relationshipbetweenproblem-solvingperfor-
mance and the quality of the organizationof students'knowledge. To study this
relationship, we investigated the influence of a range of measures of content
knowledge and knowledge organization on students' problem-solving perfor-
mance. We were interestedin examiningthe extent to which these sets of content
and connectednessindicatorswould differentiatebetween groupsof studentswho
differed in levels of achievementin high school geometry. Of particularinterest
was investigation of the prediction that the groups would be differentiatedby
students' performanceson the tasks designed to provide informationabout the
qualityof knowledge organization.
In both free-recallandpromptedsituations,the HA studentswere able to access
a widerbody of knowledgeof geometryfacts andtheoremsthanthe LA group.The
groups did not differ in their recognitionof geometricforms, but they did differ
significantlyin theirspontaneousaccessingof geometricrules.The lowerfrequency
of give-awayhintsprovidedto the HA groupsuggeststhat,relativeto the LA group,
these studentshad a widerbody of problem-relevantknowledgeto call upon. This
resultwas also found in the comparisonof HA and LA groupsin our earlierwork
(Lawson & Chinnappan,1994).
However, other results suggest that this difference in access was not the result
of just a differencein the extent of knowledge availableto the studentsin the two
groups.The resultsof the HintingTaskshowed thatthe LA studentsrequiredmore
assistanceto access relevantknowledgethathadnot been accessed spontaneously.
Consistentwith our earlierfindings (Lawson& Chinnappan,1994), the LA group
appeared to have knowledge available that was not accessed until they were
providedwith cues thatfacilitatedmemorysearch.This patternof performanceis,
40 KnowledgeConnectedness

we argue,indicativeof a less effectively organizedandless well-managedknowl-


edge base. A reviewerof this articlearguedthatthe differencein Hintingperfor-
manceis a biasedindicatorinasmuchas the HA studentswere highly likely to need
fewer hints than the LA studentsbecause of their betterproblem-solvingperfor-
mance. Although this bias exists, the results of the Hinting Task are not without
value. Use of this task allowed us to make two importantjudgments about the
students.First,we could provideevidence thatsupportedthe expectationthat the
more successful studentswould suffer less from the problemof inertknowledge.
Second, and more important,by probingsystematicallyfor students'knowledge
in this HintingTask, we were able to make reasonableclaims aboutthe quantum
of knowledge available to the students in this area of geometry. Without this
evidence, derivedfromthe frequencywith whichgive-awayhintswere needed,we
would have less justificationfor claiming that the differencebetween the groups
was not simply a matterof quantityof knowledge.
The response-timedataprovidedfurtherevidence suggestive of the moreeffec-
tive organizationalstateof the knowledgebases of the HA students.These students
wereableto correctlyrecognizerelevantknowledgecomponentsmorequickly.This
finding also suggests thatsome featureof the state of organizationof theirgeom-
etry knowledge, possibly strength,allowed more rapidaccess to this knowledge.
Othermeasuressupportthe view thatthis morerapidrecognitionis associatedwith
more effective connectednesswithin and among knowledge schemas.
The resultsof the ApplicationandElaborationTasksaddressthe issue of knowl-
edge organizationmore directly.In these cases it is not so much the influence of
searchproceduresas the state of connectednessof knowledge that is of concern.
We contendthatstudentswith high scores on these tasks show evidence of being
able to activatewider networksof geometryknowledge. The performanceof the
HA studentsindicates that they had a richer set of connections among schemas
relatedto this areaof geometry,a finding thatin termsdiscussed earliersuggests
the presenceof more links among relatedknowledge components.
The differencesin patternsof connectionwithinandamongschemasfor the HA
and LA groupsthat we have attemptedto demonstratehere might also be related
to the diagram-configurationmodel of geometry-theoremproving developed by
Koedinger and Anderson (1990),1 who argued that expert geometry-problem
solvers organizetheirgeometryknowledgein clustersof facts "thatareassociated
with a single prototypical geometric image" (p. 518). Although we have not
attemptedto develop a formalmodel of specific schemas,the procedureswe have
used in this study seem to be ones that could access problem solvers' geometric
"perceptualchunks."Ourresultssuggestthattheperceptualchunksof the HA group
are of betterqualitythanthose of the LA students.
The resultsof this studyprovidefurtherinformationaboutwhy high-achieving
studentsareable to producebettersolutionoutcomesthanlow-achievingstudents.

1Ananonymousreviewerof this articledrewthe workof KoedingerandAnderson(1990) to ouratten-


tion.
Michael J. Lawson and Mohan Chinnappan 41

We have arguedthat,among otherfactors,the organizationalquality of students'


geometric knowledge is associated with better problem-solving performance.
Superiorsolution attemptsof high-achieving students appearto be driven by a
geometric-knowledgebase that is more extensive and betterstructuredthan that
of low-achieving students. The HA students' performanceon the indicatorsof
knowledge organizationused in this study showed that, comparedwith the LA
group,they (a) were able to retrievemoreknowledge spontaneouslyand (b) could
activate or establish more links among given knowledge schemas and related
information.Thus,the resultsof this studysuggestthatsuccessfulproblem-solving
performanceis associated with a knowledge base that is better organized and
more extended, supportingthe views expressed by Prawat (1989) and Larkin
(1979).
The researchreportedhere has two importantimplicationsfor the way teachers
of high school mathematicsteach and assess students'understandingin the area
of geometry.First,our resultsshowed thatthe organizationalqualityof geometry
knowledge constructedby the high achieversconfers on them an advantageover
the low achieversin the solutionof problems.The challengeformathematicseduca-
tors and classroom teachers is to devise strategies for helping all students to
improve the state of connectednessof their knowledge bases, but particularlyto
assist the less effective problemsolversto exploitmoreof the knowledgethey have
acquired.More effective connectionsare importantboth within specific schemas
and among relatedschemas. In the terms discussed by Koedingerand Anderson
(1990), betterqualityconnections allow studentsto "thinkat a largergrain size"
(p. 547). Given the active, constructivenatureof students' study practices, we
believe thatclassroominstructiontime shouldbe allocatedto display and discus-
sion of the schemas that students develop for topics within their mathematics
programs.The findings of the present study suggest that less effective problem
solvers mightneed extratime anddiscussionto set up the types of connectionsand
representationsthatlead to effective accessing of knowledge.
A second majorimplicationof this study concernsassessmentof school math-
ematics, especially geometry.In a recentarticleon assessment,Senk, Beckmann,
and Thompson(1997) found thathigh school teacherstendedto assess students'
understandingsfrom a narrowbase of standardizedtests and arguedfor the need
to use more open-endedtasks. The tasks thatwe have developed and used in this
study,especially the ElaborationandApplicationTasks,appearto providea wider
and possibly more productiveenvironmentin which studentscould display their
geometricalknowledge.These tasksrequirethe studentto retrieveanduse connec-
tions that might not be activatedin other ways. With access to this information,
the teacheris likely to have a broaderpictureof the stateof a student'sknowledge
on which to base decisions aboutany difficultybeing experiencedby the student.

REFERENCES
Anderson,J. R. (1990). Cognitivepsychology and its implications(3rd ed.). New York:Freeman.
Anderson,J. R. (1995). Cognitivepsychology and its implications(4th ed.). New York:Freeman.
42 KnowledgeConnectedness

Bielaczyc, K., Pirolli, P. L., & Brown, A. L. (1995). Trainingin self-explanationand self-regulation
strategies:Investigatingthe effects of knowledgeacquisitionactivitieson problemsolving. Cognition
and Instruction,13, 221-252.
Bransford,J., Sherwood, R., Vye, N., & Rieser, J. (1986). Teaching thinkingand problem solving.
AmericanPsychologist, 41, 1078-1089.
Campione,J. C., Brown, A. L., & Ferrara,R. A. (1982). Mentalretardationand intelligence. In R. J.
Sternberg(Ed.), Handbookof humanintelligence (pp. 392-490). Cambridge,England:Cambridge
UniversityPress.
Chi, M. T. H., Bassok, M., Lewis, M. W., Reimann,P., & Glaser,R. (1989). Self-explanations:How
studentsstudy and use examples in learningto solve problems.CognitiveScience, 13, 145-182.
Chi, M. T. H., De Leeuw,N., Chiu,M. -H., & LaVancher,C. (1994).Elicitingself-explanationsimproves
understanding.CognitiveScience, 18, 439-477.
Chinnappan,M., & Lawson, M. J. (1994). School geometry: Focus on knowledge organization.
AustralianMathematicsTeacher,50, 14-18.
Chinnappan,M., Lawson, M. J., & Gardner,D. (1998). The use of microcomputersin the analysis of
mathematicalknowledgeschemas.InternationalJournalof MathematicalEducationin Science and
Technology,29, 805-811.
Deese, J. (1962). On the structureof associative meaning.Psychological Review,69, 161-175.
Derry, S. J. (1996). Cognitive schema theory in the constructivistdebate.EducationalPsychologist,
31, 163-174.
Hiebert,J., Carpenter,T. P., Fennema,E., Fuson, K., Human,P., Murray,H., Olivier, A., & Wearne,
D. (1996). Problemsolving as a basis for reformin curriculumand instruction:The case of mathe-
matics. EducationalResearcher,25 (4), 12-22.
Johnson, P. E. (1965). Word relatedness and problem solving in high-school physics. Journal of
EducationalPsychology, 56, 217-224.
Koedinger,K. R., & Anderson,J. R. (1990). Abstractplanningandperceptualchunks:Elementsof exper-
tise in geometry. CognitiveScience, 14, 511-550.
Larkin,J. H. (1979). Processinginformationfor effective problemsolving. EngineeringEducation,70,
285-288.
Lawson, M. J. (1994). Concept mapping.In T. Husen & N. Postlethwaite(Eds.), Internationalency-
clopedia of education (2nd ed., Vol. 2, pp. 1026-1031). Oxford,England:Elsevier Science.
Lawson, M. J., & Chinnappan,M. (1994). Generative activity during geometry problem solving:
Comparisonof theperformanceof high-achievingandlow-achievinghigh school students.Cognition
and Instruction,12, 61-93.
Mayer, R. E. (1975). Informationprocessing variables in learning to solve problems. Review of
EducationalResearch,45, 525-541.
McKeown, M. G., & Beck, I. L. (1990). The assessment and characterizationof young learners'
knowledge of a topic in history.AmericanEducationalResearchJournal, 27, 688-726.
NationalCouncilof Teachersof Mathematics.(1989). Curriculumand evaluationstandardsfor school
mathematics.Reston, VA: Author.
Naveh-Benjamin,M., McKeachie,W. J., Lin, Y. -G., & Tucker,D. G. (1986). Inferringstudents'cogni-
tive structuresand their developmentusing the "orderedtree technique."Journal of Educational
Psychology, 78, 130-140.
Novak, J. D., & Gowin, D. B. (1984). Learning how to learn. Cambridge,England: Cambridge
UniversityPress.
Payne, J. W. (1994). Thinkingaloud:Insightsinto informationprocessing.Psychological Science, 5,
241, 245-248.
Prawat,R. (1989). Promotingaccess to knowledge, strategy,and disposition in students:A research
synthesis.Review of EducationalResearch,59, 1-41.
Renkl, A. (1997). Learningfrom worked-outexamples:A study on individualdifferences. Cognitive
Science, 21, 1-29.
Senk, S. L., Beckmann,C. E., & Thompson,D. R. (1997). Assessmentandgradingin high school math-
ematics classrooms.Journalfor Research in MathematicsEducation,28, 187-215.
Michael J. Lawson and Mohan Chinnappan 43

Shavelson, R. J. (1972). Some aspects of the correspondencebetween contentstructureand cognitive


structurein physics instruction.Journal of EducationalPsychology, 63, 225-234.
Stevens, J. (1996). Applied multivariatestatistics for the social sciences (3rd ed.). Mahwah, NJ:
Erlbaum.
Thompson,B. (1998, April).Five methodologyerrorsin educationalresearch:Thepantheonofstatis-
tical significance and other faux pas. Paper presented at the annual meeting of the American
EducationalResearchAssociation, San Diego, CA.

Authors
Michael Lawson, Associate Professor, School of Education,Flinders University, GPO Box 2100,
Adelaide 5001, Australia;mike.lawson@flinders.edu.au
Mohan Chinnappan, Lecturer,MathematicsEducationUnit, Departmentof Mathematics,University
of Auckland,Auckland,New Zealand;chinnap@math.auckland.ac.nz