See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/234585889
The Test of Logical Thinking
Article · January 1984 CITATIONS
25
READS
13,939 2 authors, including:
Kenneth Tobin The Graduate Center, CUNY 394PUBLICATIONS 11,516CITATIONS
SEE PROFILE
All content following this page was uploaded by Kenneth Tobin on 06 August 2015.
413
THE DEVELOPMENT AND VALIDATION
OF
A GROUPTEST OF LOGICAL THINKING
KENNETH G. TOBIN’
University of Georgia
WILLIAM CAPIE
University of Georgia
The paper describes the development of the Test of Logical Thinking (TOLT) to measure five modes of formal reasoning: con- trolling variables, proportional reasoning, combinatorial reasoning, probabilistic reasoning, and correlational reasoning. Each of the 10 items
requires participants
to select a correct response andjustifica-
tion from a number of alternatives. Analysis of data from 682 stu- dents from grades 6 through college indicated high test reliability
(coefficient
α = .85) and provided confirmation that the test mea-sured one major underlying dimension termed formal thought. Evi-
dence of criterion-related validity was obtained from a study in
which 88 students from
grades
10through
college were assessed onthe TOLT and on five interview tasks. A correlation of .80 (p <
.0001) suggested a strong relationship between the two measures of formal reasoning.
1981, 41
INITIALLY studied
by Piaget
and hiscolleagues,
thedevelopment
offormal
reasoning ability
has beenextensively
researched in adoles- cents and adults(e.g., Arlin, 1975; Chiappetta, 1976; Farrell, 1969;
Lovell, 1961).
In themajority
of cases, clinical interviews based onprotocols
describedby
Inhelder andPiaget (1958, 1975)
have beenused to assess formal
reasoning ability.
Twoimportant
trends thathave
emerged
from research are that many adolescents and adults arelimited in their
ability
to use formalmodes
ofreasoning
and that for-1 Senior author has returned to the faculty at Mount Lawley College, Western Australia
6050.
Copyright @ 1981 by Educational and Psychological Measurement
AND PSYCHOLOGICAL MEASUREMENT
mal
reasoning ability
is animportant
mediator ofcognitive
achieve-ment
(e.g.,
Cantu and Herron,1978;
Goodstein and Howe,1978).
As aconsequence, researchers have
emphasized
theimportance
of modi-fying
instructionalobjectives, materials,
and activities so thatthey
aresuited to the
cognitive development
of learners.Concomitantly they
have
urged
thatpriority
begiven
to thedevelopment
of formal rea-soning ability
of middle andhigh
school studentsthrough
the use ofappropriate
curriculum materials.Action on either of these concerns
requires
that formalreasoning ability
be assessed in a valid and reliable manner. The clinical inter- viewprocedure
is not suited forwidespread
administration because of the timerequired
to administer a set of tasks to a group of students and because of the level ofexpertise required
of the interviewer. In aresearch context an additional
problem
derives from thesubjectivity
of the
procedures
used. Theadvantages
ofsimultaneously testing large
groups ofsubjects
have led a number of attempts todevelop
valid group measures of formal
reasoning ability (e.g., Bumey, 1974;
Lawson,
1978;
Staver andGabel, 1979;
Tisher andDale, 1975).
Itemsfrom most of these tests were based on content and
logical
processes derived from the works of Inhelder andPiaget (1958, 1975).
Ease of administration
accompanied by objective scoring
proce- dures is an inducement for researchers todevelop
a valid and reliablepencil
and paper measure of formalreasoning ability.
Amajor
diffi-culty, however,
is to ensure thatsubjects
use formalreasoning ability
to solve the items on the test. Whether
subj ects
use formalreasoning
or not may be ascertained from their reasons for
developing
or choos-ing
a response. A characteristic of the clinicalprocedures
that has gen-erally
not beenincorporated
into apencil-and-paper
tests is the neces-sity
forsubjects
tojustify
their solution to aproblem. Exceptions
tothis
requirement
were the test items describedby
Lawson(1978)
andLawson, Adi,
andKarplus (1979),
whichrequired subjects
toprovide
written
justifications
for their solutions toproblems.
Anotherimpor-
tant aspect of the
procedure employed by
Lawson was the use of dem-onstrations to
provide
a realistic context for theproblems
to be solved.The purpose of this
study
was todevelop
a group test of formal rea-soning ability
that wouldrequire
students to solveproblems
and tojustify
the solutions obtained.Development
of the Test ofLogical Thinking (TOLT)
commenced with a selection of ten itemspreviously reported by
Lawson(1978)
and Lawson et al.(1979).
Procedures
In the sections that
follow, procedures
are described for the devel- opment of theTOLT,
andinvestigations
of internalconsistency,
con-struct
validity,
and criterion-related are detailed.TOBIN AND CAPIE
Development of
TOLTItems that had been used in
prior
research(Lawson, 1978;
Lawsonet
al., 1979)
wereemployed
as a basis fordeveloping
an initial version of TOLT. Theadoption
of thisprocedure
assured that TOLT would contain items that had beenpreviously reported
as valid measures offormal
reasoning ability.
Two items were selected to measure each offive modes of formal
reasoning: controlling variables, proportional reasoning, probabilistic reasoning,
correlationalreasoning,
and com-binatorial
reasoning.
The testincorporated
demonstrations toprovide
a context for the items. Students selected a correct response for a num-
ber of alternatives and
provided
writtenjustification
for their selec- tion.Although
thereliability
estimate of the TOLT wasreasonably high (a = .74),
several factors wereapparently reducing validity.
For ex-ample,
manyhigh
school students were unable to formulate a clear writtenjustification
forselecting
aparticular
response.Rescoring
these responses
suggested
many inconsistencies inscoring.
Because ofthese
problems,
the initial version of TOLT was modified so that mul-tiple justifications
wereprovided
as well asmultiple
solutions for eachproblem.
Reasons that had been volunteeredby subjects
on the firstversion of TOLT were modified and
incorporated
into a revised ver-sion of the test. In the revised test a correct solution
required
selectionof the correct response and the best
justification
for the response. Asample
item is included inFigure
1 to illustrate the item format.The revised test also utilized a color
video-tape
to present the con-text for each
problem
and to standardize administrationprocedures.
Adequate
time wasprovided
for each item to becompleted.
The timeFigure 1. A sample item from the TOLT.
416
required
for the test to be administered wasapproximately
40 min-utes.
Investigations of
TOLT:Validity
andReliability
Three
samples
were used in aninvestigation
of thereliability
andvalidity
of the TOLT. The TOLT was administered to asample
of 353students in middle school
grades 6,
7 and8;
to asample
of 82physics
and
chemistry
students fromgrades
11 and12;
and to 247 students en- rolled incollege
science coursesIntact classes were
employed
in all cases to obtain data from relia-bility
and constructvalidity investigations.
In such studies the purposewas not to differentiate
subjects
at differentgrade levels,
but to iden-tify
the extent to which test items measure a common dimension. Anadequate investigation
of these test characteristicsrequires
a range inability
of examinees on the constructbeing
measured. As a con-sequence,
subjects
were selected from a number ofgrade
levels.Prior research
(Renner
andGrant, 1978)
had shown thedevelop-
ment levels of
physics
students to be distributeddifferently
from thedevelopmental
levels of all students. Ahigher proportion
ofphysics
than of other students was found to be
operating
at a formal opera- tional level. As a consequence, students ingrades
11 and 12 fromphysics
andchemistry
classes were selected in an endeavor to obtain agreater range in formal
reasoning.
Thesample
was not chosen to berepresentative
of thepopulation
ofgrade
11 and 12 students.The internal
consistency
of TOLT was assessed from thecomplete
data set
(n
=682) by using
coefficient a(Cronbach, 1951).
Two sepa- rate factoranalyses
were conducted in aninvestigation
of the under-lying
structure of theperformance
data. In thefirst,
the data to be an-alyzed
consisted ofperformance
on each of the fivehypothesized
modes of formal
reasoning (each
moderepresenting performance
ontwo
items).
In the secondanalysis, performance
on each item was ana-lyzed.
In each factor
anlaysis
aprincipal
axisprocedure
was used to ex-tract common factors. In this
procedure, diagonal
elements of the cor-relation matrix of the five modes of
reasoning
werereplaced
withcommunality
estimates. The initial estimates were thesquared
mul-tiple
correlation ofperformance
on each mode ofreasoning
with thaton the
remaining
mode ofreasoning.
A Scree test(Cattell, 1966)
andthe Kaiser varimax criterion
(Kaiser, 1960)
werejointly applied
to ob- tain a final solution to the factoranalysis.
Rotation forinterpretation
was not
required,
as one-factor solutions were obtained in each case.417
Criterion-Related
Validity of
TOLT Thesamples
for thisinvestiga-
tion consisted of 25 students enrolled in
college
science educationcourses and of 63
high
school students fromgrades
10through
12. TheTOLT was administered to students followed
by
abattery
of five clin- cial interviews selected from those described in Inhelder andPiaget (1958,1975)
toprovide
a measure of each of the formal modes of rea-soning
assessedby
TOLT. On the basis ofperformance
on each taskstudents were rated one if
they
demonstrated formalthought
in solv-ing
aproblem
or zero ifthey
did not.Ratings
for each task weresummed to
provide
a measure ofperformance
on the clinical inter- views.Results
Coefficient a for the TOLT which was based on the total
sample
of682
students,
was .85. The internalconsistency
estimate of each two- item subtestranged
from .56 to .82.Descriptive
data related to the items and subtests of the TOLT are contained in Table 1. Item diffi- cultiesranged
from .18 to .41 with an average of .30. Item discrimina- tion indicesranged
from .39 to .71 with an average of .55.When the data were
separately analyzed
for eachgrade
level agradual
increase inperformance
was evident fromgrade
6 tocollege
level. A
frequency
distribution for eachgrade
level in thesample
isprovided
in Table 2.Performance on the five modes of
reasoning
wasmoderately
inter-TABLE I
Descriptive Data for the TOLT
EDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT TABLE 2
Distribution of Formal Reasoning Ability
correlated. The correlation coefficients included in Table 3 demon-
strate the interrelated nature of
performance
on the five modes of for- malreasoning incorporated
into the TOLT.The factor
analysis
of the intercorrelations of measures, each of which washypothesized
to reflect one of the five modes ofreasoning, produced
a one-factor solution which accounted for 43% of the com- mon variance. Each measured mode ofreasoning
washighly
corre-lated with the
factor,
with the factor structureloadings ranging
from.60 to .72. The factor structure
loadings
andcommunality
estimatesfor this
analysis
are contained in Table 4In the second factor
analysis
of the intercorrelations of items a one-factor solution which accounted for 38’ percent of the common vari-
ance was obtained. In this case the factor structure
loadings,
whichTABLE 3
Intercorrelations among Hypothesized Modes of Formal Reasoning
on the TOLT (n = 682)
aTwo items were included for each Mode.
TOBIN CAPIE
a Two items are included for each Mode.
ranged
from .49 to .73(Table 5),
were indicative of a common(uni- dimensional)
structureunderlying performance
on each item.Criterion-Related
Validity
The correlation between
performance
on the interviews and scores on the TOLT was .80.Descriptive
data for the clinical interviews areprovided
in Table 6.Predictive
Validity
Results from concurrent
investigations
at theUniversity
ofGeorgia (Table 7)
indicatedsignificant relationships
between TOLTperform-
ance and that on other variables. These studies
provided
an indicationof the
predictive validity
of the TOLT.In a
study involving
353 students fromgrades
6through
8(Tobin
and
Capie, 1980),
35 percent of the variance on a test ofintegrated
sci-ence
processes
was attributable to variation inperformance
on theTOLT. Attenuation correction
suggested
avalidity
coefficient of .74.TABLE 5
Factor Structure Loadings for the TOLT Items
EDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT TABLE 6
Description Data for Clinical Interviews
a Two items are included for each Mode.
Tobin, Capie
andBradley (1980)
also found that theperformance
of 150
college
students wassignificantly
correlated withintegrated
sci-ence process achievement
(r
=.49,
p <.0001). They
alsoreported
acorrelation of .49
(p
<.0001 )
with asample
ofhigh
school students.Bradley (1980) reported
asignificant relationship
between TOLTperformance
for asample
ofcollege
students and two measures of vi-sualization : the paper
folding
test(r
=.61,
p <.0001 )
and the surfacedevelopment
test(r
=.55,
p<.0001). Significant relationships
wereTABLE 7
Predictive Validity Data for the TOLT
TOBIN CAPIE
also
reported
for scores obtained on theCollege
Board ScholasticAp-
titude Test(SAT) (Educational Testing Services, 1948-1980).
Yeany, Helseth,
and Barstow(1980)
citedsignificant relationships
between TOLT
performance
and(a)
achievement incollege biology
and
(b)
SAT scores.Discussion
The TOLT furnished a reliable means of
assessing
formalreasoning ability.
The ten-item test has ahigh
internalconsistency
and several of the subtests exhibit sufficientreliability
to allow decisionmaking
at thesubtest level. The
reliability
coefficients are of sufficientmagnitude
toenable the test to be used in
diagnostic
assessment, in a research con-text, or in studies
designed
to promotespecific
formalreasoning
abili-ties. The
magnitudes
of intercorrelations among modes of formal rea-soning
weresuggestive
of a commonunderlying
unidimensionalstructure. When the coefficients were corrected for the attenuation that occurs because of the
unreliability
of thesubscales,
the inter- correlations of the subscalesranged
from .46 to .70.Each of the measures
reflecting
differenthypothesized
modes of the formalreasoning
washighly
correlated with the one-factor in the so-lution obtained from the factor
analysis
of subtest scores. This resultcan be
interpreted
in terms of each mode ofreasoning contributing
toa common
underlying
one-factor structure.A similar
interpretation
isapplicable
to the factoranalysis
of the separate item data. These resultsprovided
support for the constructvalidity
of the TOLT. The datasuggested
that each item affords a measure on oneunderlying
dimension which is defined to be formalreasoning ability.
Although
acomparatively
smallsample
was involved in the crite- rion-related validation of TOLT, ahigh
non-chancerelationship
wasestablished between TOLT
performance
and clinical interview per- formance. A correlation of .80 between the two variables was in- dicative of similarperformance
on each variable. The evidence sug- gests that the same processes that allowsubjects
to solveproblems correctly
in the clinical interview are involved in the solution of TOLT items.The data obtained in the
investigation
of criterion-relatedvalidity
suggest that TOLT ismeasuring
formalreasoning ability.
Additionalstudies of this type are
required using larger samples
that encompass the entire range of formalreasoning ability.
The initial purpose of
developing
TOLT was to construct an in-strument that could be used in studies of
teaching
andlearning.
Onthe basis of
psychometric properties
described in the paper, the use ofEDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT
the TOLT is advocated to obtain measures that differentiate student formal
reasoning ability.
Data sogenerated
would be useful in relat-ing
formalreasoning ability
toachievement;
ininvestigating possible
interactions of formal
reasoning ability
with teachervariables;
or instatistically controlling
for variations in formalreasoning
so that theeffects of other teacher and student variables can be determined. In these respects the TOLT can be used to obtain continuous or cate-
gorical
measures of formalreasoning ability.
The datapresented
inTable 7 show that the TOLT has
already
beenemployed
for the three purposesjust specified.
Although
thedevelopment
of the TOLT grew frompractical
con-cerns
relating
to research anddiagnostic teaching,
themethodology
used to
develop
the TOLT isapplicable
to furtherinvestigations
in thenature of formal
reasoning. Specific
tests for each mode ofreasoning
could bedeveloped,
orexisting
TOLT items could be modified to en-able additional information to be obtained from each item. For ex-
ample,
a set offollow-up questions
could be constructed. Theprin- ciple
ofusing
amultiple-choice
format for thejustification
of eachanswer offers the
potential
fordeveloping
measures of formal reason-ing
that are valid and reliable. With these instruments based on asound clinical interview
experience,
researchers would have efficient tools tostudy developmental
patterns amonglarge
numbers of sub-jects.
Conclusion
Evidence suggests that the TOLT does measure formal
thinking.
The
reliability
data are indicative ofhigh
internalconsistency
and thevalidity
data are diverse andsupportive
of an effective group test of formalthought.
The TOLT
provides
a means ofassessing
formalreasoning ability
as a
diagnostic
aid for teachers or as data for researchersinvestigating
the nature of
learning.
The test is suitable for administration on agroup basis to students from
grade
6through college.
REFERENCES
Arlin,
P.Cognitive development
in adulthood: A fifth stage.Develop-
mental
Psychology, 1975, 11,
602-606.Bradley,
C. F. Theeffects of imagery stimulation,
visualization skill andcognitive development
level on the science achievementof college
stu-dents
.
(Doctoral dissertation, University
ofGeorgia, 1980).
Burney,
G. M. The construction and validation of anobjective
formalreasoning
instrument.(Doctoral dissertation, University
of North-ern
Colorado, 1974.)
Dissertation Abstracts International,1975, 35,
4535-B.
(University
Microfilms No.75-05, 403)
Cantu, L. and
Herron,
D. Concrete and formalPiagetian
stages and science concept attainment. Journalof
Research in Science Teach-ing, 1978, 15,
135-143.Cattell,
R. B. The Scree test for the number of factors. Multivariate BehavioralResearch, 1966,
1, 245-276.Chiappetta,
E. A review ofPiagetian
studies relevant to science in- struction at thesecondary
andcollege
level. ScienceEducation, 1976, 60,
253-261.Cronbach,
L. J. Coefficientalpha
and the internal structure of tests.Psychometrika
, 1951, 16,
297-334.Educational
Testing
Service.College
Entrance Examination Board ScholasticAptitude
Test.Princeton,
NewJersey,
1948-1980.Farrell,
M. A. The formal stage: A review of the research. Journalof
Research and
Development
inEducation, 1969, 3,
111-118.Goodstein,
M. and Howe, A. The use of concrete methods in second- arychemistry
instruction. Journalof Research
in ScienceTeaching,
1978,15, 361-366.
Inhelder,
B andPiaget,
J. Thegrowth of logical thinking from
child-hood to adolescence. New York:
Basic,
1958.Inhelder,
B andPiaget,
J. Theorigin of
the ideaof
chance in children.New York: N. W.
Norton,
1975.Kaiser,
H. F. Theapplication
of electronic computers to factoranaly-
sis. EDUCATIONAL AND PSYCHOLOGICAL
MEASUREMENT, 1960, 20,
141-151.
Lawson,
A. E. Thedevelopment
and validation of a classroom test of formalreasoning.
Journalof
Research in ScienceTeaching, 1978,
15, 11-24.
Lawson, A.
E., Adi,
H. andKarplus,
R.Development
of correlationalreasoning
insecondary
schools: Dobiology
courses make a differ-ence ? The American
Biology Teacher, 1979, 41,
420-425.Lovell,
K. Afollow-up study
of Inhelder andPiaget’s
thegrowth
oflogical thinking.
British Journalof Psychology, 1961, 52,
142-153.Renner, J. and
Grant,
R. Can students graspphysics concepts ?
TheScience
Teacher, 1978,
45, 30-33.Staver, J. R. and
Gabel,
D. L. Thedevelopment
and construct valida-tion of a
group-administered
test of formalthought.
Journalof
Re-search in Science
Teaching, 1979, 16,
535-544.Tisher, R. P. and
Dale,
L. G.Understanding
in science test. Victoria:Australian Council for Educational
Research,
1975.Tobin,
K. G. andCapie,
W.Relationships
between formalreasoning ability,
locus ofcontrol,
academic engagement, andintegrated
process skill achievement. Journal
of Research
in ScienceTeaching, 1981,
18(in press).
Tobin,
K.G., Capie,
W. andBradley,
C. F. Therelationship of formal reasoning ability
and process skill achievement(Tech. Rep. 39), University
ofGeorgia,
October 1980.Yeany, R., Helseth,
E. H. andBarstow,
W. Interactive instructionalvideo-tapes,
scholasticaptitude, congitive development
and locusof
control as variables
influencing
science achievement. A paper pre- sented at the annualmeeting
of the National Association for Re- search in ScienceTeaching, Boston, April,
1980© 1981 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.http://epm.sagepub.com at Mina Rees Library/CUNY Graduate Center on January 17, 2008 Downloaded from