You are on page 1of 7

Journal of Educational Psychology

1973, Vol. 65, No. 3, 364-370

A MULTIVARIATE ANALYSIS OF ACADEMIC ABILITIES AND


STRATEGIES FOR SUCCESSFUL AND UNSUCCESSFUL COLLEGE
STUDENTS IN DIFFERENT MAJOR FIELDS1

ROY D. GOLDMAN 2 AND DAVID J. HUDSON


University oj California, Riverside

A sample of 256 freshmen college students, representing different


major fields of study, were compared on a set of ability measures and
a set of academic strategy measures. Multivariate analysis of variance
was employed to compare students in a 5 X 2 X 3 (Major Field X
Sex X Grade Point Average) design. There were significant main ef-
fects for major field and for sex on both ability and strategy measures.
There were significant differences in group centroids for grade point
average groups only for the strategy measures. A multivariate analysis
of covariance, using ability measures as covariates, yielded significant
main effects for strategy measures. The results strongly suggest that
there are major field differences in academic strategy, which are in-
dependent of ability differences. These major field differences are prob-
ably due to differences in the "task structure" of the different
curricula.

It has been demonstrated that often a ally determinate, then prediction of task
given task can be performed through the performance by ability measures should be
use of different operations (strategies) and quite good. The abilities necessary for suc-
that success on the task is partly deter- cessful task performance would be entirely
mined by the particular choice of a strategy task determined. Similarly, if a task were
(Bruner, Goodnow, & Austin, 1956). Fur- very indeterminate, prediction of perform-
thermore, the choice of a strategy deter- ance by abilities would be poor, since dif-
mines which abilities will be used on a ferent individuals could employ different
given task (French, 1965), and task struc- abilities to perform the task. Thus, the
ture determines the range of efficient strate- strategy employed in a given task may be
gies (Frederiksen, 1969). Frederiksen's determined by both task parameters (e.g.,
model can be used to explain many abil- the nature or "structure" of the task) and
ity-strategy relationships in terms of task by organismic parameters (including the in-_
structure. The model begins with the asser- dividual's ability profile).
tion that tasks that can be solved in more If college classes are considered as tasks,
than one way are "functionally indetermi- it is likely that they have different task
nate"; the wider the range of potentially structures. Different major fields present ,
effective strategies, the more indeterminate different types of problems to the student. \
the task. By selecting a strategy, the indi- Clearly, there is little surface resemblance I
vidual "restructures" the task; thus strate- between the solution of mathematical e q u a - J
gies serve as moderators between abilities tions and the writing of literary critiques.
and performance. If a task were function- When students from different major fields
of study have been compared, strategy dif-
1
This research was supported in part by U. S. ferences have often been detected (e.g.,
Public Health Service Grant MH-21119-01 from Biggs, 1970; Brown & Dubois, 1964; Cro-
the National Institute of Mental Health. pley & Field, 1969). In the Biggs study,
2
Requests for reprints should be sent to Roy
D. Goldman, Department of Psychology, Uni- science students were compared with liberal
versity of California, Riverside, California 92502. arts students on the study behavior ques-
364
ABILITIES AND STRATEGIES OF COLLEGE STUDENTS 365

tionnaire. Science students were found to be is unclear whether students in different


more intrinsically motivated and more dog- fields employ different strategies because of
matic but less tolerant than liberal arts stu- the nature of the tasks they face or because
dents. Brown and Dubois compared engi- of ability differences. This question can be
neering students with liberal arts students subdivided into a set of more specific ques-
on several scales from the Minnesota Mul- tions.
tiphasic Inventory and found engineering 1. Are there strategy differences between
\students to be more hardworking, energetic, major fields?
Vnd tending toward conformity. In these 2. Are there ability differences between
studies, the nature of the tasks facing the major fields?
students was cited as a possible explanatory 3. Is there an association between abili-
factor for the group differences in scholastic ties and strategies?
strategies. 4. Do strategy differences between major
Elton and Rose (1967) employed discrim- fields exist independently of ability differ-
inant function analysis to compare stu- ences?
dents in different fields. This multivariate Since the present investigation also dealt
technique provides considerable advantage with sex and grade point average, answers
over a series of univariate comparisons. to other questions could also be sought.
Often one series of univariate comparisons 5. Are there male-female differences with
is likely to magnify the differences between regard to strategies and/or abilities?
groups, since dependent variables are likely 6. Are sex differences in strategies (if ]
to be correlated. Such intercorrelation of any) independent of sex differences in abil- '
variables makes it difficult to interpret the ities? I
meaning of group differences. The multi- 7. Are there strategy differences between \
variate comparison of group centrcnds pro- successful and unsuccessful (in grade point <
vides an alpha level that is not increased by average) students? /
the use of multiple dependent variables. Fur- 8. Are there ability differences between \
thermore, there is considerable heuristic successful and unsuccessful students? —*"^
benefit in the use of multivariate methods, The above questions are amenable to in-
for discriminant function analysis permits vestigation through the use of multivariate
the assessment of the loci of differences analysis of variance and covariance as well
among groups. It is also possible to employ as canonical correlation analysis.
discriminant function analysis or "map" the Another question that must be answered
groups under study (Cooley & Lohnes, concerns the interaction of academic success
1971). In a recent investigation Goldman with major field. This last question asks if
and Warren (1973) plotted group centroids the differences between successful and un-
for physical science, biological science, so- successful students are different (for differ-
cial science, and humanities students in a ent fields of study).
two-dimensional discriminant space. One The purpose of the present investigation
dimension, reflecting "formal" (mathemati- was to seek answers to the above questions,
cal-logical) reasoning, arrayed groups in a and in this endeavor, provide a description
science-nonscience continuum; whereas the of the abilities and strategies of college stu-
second function, which reflected "applied" dents.
thinking, arrayed the groups differently.
This mapping of major fields in the hy- METHOD
perspace of student's study strategies is a
method for investigating the task structure Subjects
of those fields. Unfortunately this method is Subjects were 256 freshman students at the Uni-
indeterminate, since it leaves several ques- versity of California, Riverside. They were part of
tions unanswered. One such question con- a sample randomly selected from the registrar's
list and recruited by mail. Since subjects were paid
cerns the etiology of strategy differences be- for their participation, there was a relatively high
tween students in different major fields. It rate of response to this method of recruitment
366 ROY D. GOLDMAN AND DAVID J. HUDSON

(approximately 70%). In terms of grade point Mnemonics (Scale 8). Mnemonics is defined by
average, our sample was very similar to the entire items that indicate the use of memorization as a
freshman class (2. 89 vs. 2.84). strategy of study (e.g., I memorize specific facts
for use in exams, I use mnemonic devices [memory
Procedure tricks] with specific problems or terms).
Planfulness (Scale 4)- Planfulness contains items
The design was correlational in nature with all that reflect concern for punctuality and fore-
subjects receiving a battery of tests designed to planning (e.g., I take especially complete notes;
measure selected abilities and academic strategies. I attend all or nearly all class meetings; I keep up
Another measure was grade point average, com- with the assigned readings week by week; I do
puted at the end of the second quarter of study. most of the assigned reading at the end of the
Student grade point average was trichotomized to quarter—in other words, I cram [scored in the
yield a 5 X 2 X 3 design (Major Field X Sex X reverse direction]).
Grade Point Average). Formal thinking (Scale 6). Formal thinking
contains items related to logical and mathematical
Ability Measures thinking (e.g., I enjoy mathematical reasoning; I
The ability measures which were employed try to find a logical reason for all information; I
were selected from the Kit of Reference Tests for am good at syllogistic [logical] reasoning). —
Cognitive Factors (French, Ekstrom, & Price, Note taking (Scale 6). Note taking is defined by
1963). The tests that were used represented the items that reflect the behavior of writing while
factors of vocabulary, syllogistic reasoning, general studying (e.g., I extract the important points from
mathematical reasoning, induction associative the assigned reading and write them down; I out-
memory, spatial scanning, and number facility. line the reading and review the outlines; I take
The specific tests used to represent these factors notes on the reading while doing the reading).
were Advanced Vocabulary Test (V-4), Inference Transformation and application (Scale 7).
Test (Rs-3), Mathematic Aptitude Test (R-2), Transformation and application reflect active in-
Letter Sets Test (I-I), Object-Number Test (Ma- tegration of scholastic information (e.g., I read
2), Choosing a Path (Ss-2), and Division Test the same material from three or four different
(N-2). sources and then combine all specific views into
one large view; I find interrelationships in the
Strategy Measures material covered as a memory aid). This scale also
reflects application and relevance of scholastic
The study techniques questionnaire (Gold- material (e.g., I try to find practical application
man & Warren, 1973) was administered to all of the material I read; I try to find personal
subjects. Responses were used to form seven sub- relevance in lecture and reading material so I can
scales. These subscales were designed to reflect con- look at it subjectively rather than objectively).
ceptually focused dimensions that had resulted In sum this scale reflects some altering by the in-
from an earlier factor analytic study (Goldman & dividual of the original input of scholastic in-
Warren, 1972). Items that had high loadings on a formation into a more amenable form.
factor were included in the computation of the
new subscale. Thus, the subscales used in the RESULTS
present study were not orthogonal. It was felt that
the above approach more closely reflects the reality A multivariate analysis of variance was
of study habits in that scholastic strategies seem performed, employing as dependent varia-
likely to be correlated. The seven subscales are bles (a) ability measures, (b) strategy
briefly described below. All items were answered
on a 5-point scale of agreement-disagreement. measures, (c) ability measures with strat-
egy measures partialed out as covariates,
Description of Subscales and (d) ability measures with strategies
~~ Clerical diligence (Scale 1). Clerical diligence partialed out. Cell sizes in the 5 x 2 X 3
contains items that reflect considerable effort in design were not proportional with regard to
preparation for classes (e.g., I spend more time sex and major field. There were proportion-
studying than most people; I review my class ally fewer female science majors and male
notes before each class session; I do research at humanities majors. To obtain conservative
the library on class material even if it is not re-
quired). F ratios, treatment sums of squares and
Academic "savvy" (Scale 2). Academic savvy cross-products were calculated by subtrac-
contains items that reflect a sensitivity to the tion in stepwise fashion. The following
academic "game" (e.g., I try to find out what the method was used: Each treatment sum of
instructor expects and study with that in mind;
I try to learn from errors on graded material; I squares and cross-products (SSCP) matrix
get individual help from the professor and TAs of interest was obtained by subtraction
[teaching assistant] whenever necessary). from the total treatment SSCP matrix after
ABILITIES AND STRATEGIES OF COLLEGE STUDENTS 367

TABLE 1 also important discriminators between


COMPARISON OP MALES AND FEMALES
ON ABILITY MEASURES
major fields. Since only one discriminant
function was significant, only one set of dis-
criminant function coefficients is displayed
7
Discriminant
Variable Univariate F P function
coefficient
in Table 2.
Group Comparisons on Strategy Measures
Vocabulary <1 ns -.13
Inference <1 ns -.06 There were significant main effects due to
Math 15.28 .0002 .68 major field (F = 2.03, df = 28/792, p <
Induction <1 ns -.58 .001), sex (F = 3.36, df = 7/220, p < .002),
Spatial 24.9 .0001 .75 and grade point average (F = 2.03, df =
Memory
Division
1.76
2.86
.18
.09
-.25

Note. Multivariate F = 7.75, df = 7/220, p <


.09 14/440, p < .01). The loci of differences
between major field groups were centered in
planfulness and transformation-application
1
1.000. (see Table 3). The differences between
males and females were centered in note
subtracting the other SSCP treatment mat- taking, transformation-application, and
rices. This provides a conservative esti- formal reasoning (see Table 4). The differ-
mate of the significance of the difference
between group centroids. TABLE 3
Group Comparisons on Ability Measures COMPARISON OF MAJOR FIELDS ON
STRATEGY MEASURES
Using the .05 level of significance, there
were significant differences between cen- Univariate
Discrimi-
nant
troids due to sex (F = 7.75, df = 7/220, p Variable F P function
coefficient
< .0001) and major field (F = 1.51, df =
28/794, p < .04). There was no significant Diligence <1 ns -.10
effect due to grade point average nor were Savvy 1.75 .13 -.17
there significant two- or three-way interac- Mnemonics 1.76 .13 .01
tions. The loci of the difference between Planfulness 1.54 .19 -.07
Formal reasoning 6.29 .0001 1.15
sexes appears to be centered in mathemati- Note taking <1 ns .17
cal and spatial reasoning (see Table 1). Transformation 2.10 .08 -.62
The loci of the difference between major
field centroids are less clear, but the discrim- Note. Multivariate F = 2.03, df = 28/794, p <
.01.
inant analysis and pattern of univariate F
ratios shown in Table 2 seem to indicate
that numerical facility and vocabulary are ences between grade point average levels
occurred largely in the use of planfulness
TABLE 2 and formal reasoning (see Table 5). There
COMPARISON OP MAJOR FIELDS ON was also a significant interaction between
ABILITY MEASURES sex and major field, which would imply that
male-female differences in strategies are
Variable Univariate F *
Discriminant
function
moderated by major field.
coefficient
Covariation Between Sets of Ability
Vocabulary 1.05 .37 .57 and Strategy Measures
Inference <1 .51 .01
Math 2.00 .09 -.27
To assess the degree of association be-
Induction <1 .59 .12 tween ability and strategy measures, the
Spatial 2.24 .06 -.43 technique of canonical correlation was em-
Memory 1.36 .24 -.28 ployed. This technique describes both the
Division 3.42 .01 -.65
degree and the nature of the relationship.
Note. Multivariate F 1.51, df •= 28/794, p < The canonical correlation coefficient be-
.04. tween ability and strategy measures was .40
368 ROY D. GOLDMAN AND DAVID J. HUDSON

TABLE 4 Group Comparisons on Strategy Measures


COMPARISON OF MALES AND FEMALES with Abilities Partialed Out
ON STRATEGY MEASURES
The multivariate analysis of covariance
Discrimi- using strategy measures as dependent var-
Variable Univariate nant iables and ability measures as covariates re-
F P function
coefficient vealed significant main effects for all fac-
tors (sex, major field, and grade point aver-
Diligence <1 ns .23
age). The discriminant function coefficients
Savvy <1 ns -.36
Mnemonics <1 ns -.16 for these contrasts are displayed in Tables
Planfulness <1 ns .15 7-9. No interactions were significant.
Formal reasoning 3.97 .04 .45
Note taking 8.51 .01 -.95 DISCUSSION
Transformation 2.47 .11 .60
Note. Multivariate F = 3.36, df = 7/220, p < Comparison of Major Field Groups
.01. Major field groups were found to differ
significantly in both abilities and strategies.
(x2- 86.24, df = 49, p < .001). Only one
canonical root was significant, indicating TABLE 6
only one common orthogonal dimension be- COMPAEISON OF MALES AND FEMALES ON
tween the two domains of measures. ABILITIES WITH STRATEGIES
PAHTIALED OUT
Group Comparisons on Ability Measures
Discriminant
with Strategy Measures Partialed Out Variable Univariate F P function
coefficient
The sex comparison on ability measures
with strategy measures removed as covar- Vocabulary 1.72 .19 -.22
iates (see Table 6) indicated a highly sig- Inference <1 ns -.09
Math 8.31 .01 .62
nificant difference between males and Induction 3.65 .05 -.6]
females (F - 6.73, df = 7/213, p < .0001). Spatial 18.64 .0001 .72
However, the ability differences between Memory 2.47 .11 -.27
major fields was not significant when Division 1.55 .21 .08
strategy measures were partialed out (F = Note. Multivariate F = 6.73, df = 7/213, p <
1.21, df = 28/769, p > .20). No other effects .0001.
,~.were significant.
Number facility was the best ability dis-
TABLE 5 criminator among major fields, while formal
COMPARISON OF Low- , MIDDLE- , AND HIGH- reasoning and transformation-application
GRADE-POINT-AVERAGE GROUPS ON were the best strategy discriminators. It ap-
STRATEGY MEASURES pears that the science majors excel in num-
Discrimi-
ber facility and make more use of formal
Variable Univariate
F P nant
function
reasoning, while nonscience fields make
coefficient more use of transformation-application. It
further appears that there is a significant
Diligence <1 ns .09 association between abilities and strategies,
Savvy 2.01 .13 -.07
Mnemonics <1 ns .30 as indicated by canonical correlation. When
Planfulness 4.20 .01 -.66 abilities are partialed out, the strategy dif-
Formal reasoning 4.27 .01 -.75 ferences between major fields become even
Note taking 1.57 .21 .44 clearer. However, when strategy measures
Transformation <1 ns .02 are partialed out, the ability difference be-
Note. Multivariate F = 2.03, df = 14/440, tween major fields is no longer significant.
< .01. These results suggest that strategy differ-
ABILITIES AND STRATEGIES OF COLLEGE STUDENTS 369

ences between major fields exist independ- TABLE 8


ently of ability differences. The fact that COMPARISON OP MAJOR FIELDS ON STRATEGY
formal reasoning is the major discriminator MEASURES WITH ABILITIES PARTIALED OUT
suggests that this strategy may be consider- Discrimi-
ably more important in science (as opposed Variable Univariate
F P
nant
function
to nonscience) fields. Thus, students in sci- coefficient
ence fields make greater use of this tech-
nique of study. Thisjmggests that the struc- Diligence <1 ns .00
Savvy 1.72 .14 -.03
ture of science curricula presents problems Mnemonics 1.55 .18 -.05
that are most efficiently solved through for- Planfulness 1.60 .17 -.24
ttml-rcaaoningT- Formal reasoning 5.48 .001 1.12
Note taking <1 ns .15
Sex Differences Transformation 1.64 .16 -.52
It appears that males and females differ Note. Multivariate F = 1.71, df = 28/766, p <
in ability centroids. Most of this difference .01.
is manifested in mathematical and spatial
and (6) strategies with abilities partialed
TABLE 7 out, it was discovered that significant dif-
COMPARISON OF MALES AND FEMALES ON ferences remained. It appears, therefore,
STRATEGY MEASURES WITH ABILITIES that males and females differ in their use of
PARTIALED OUT study strategies independently of their dif-
ferences in abilities. The same may be said
Discrimi-
Variable Univariate
P nant for the sex difference in abilities.
P function
coefficient
Comparison of High-, Middle-, and
Diligence \ ns .22
Low-Grade-Point-Average Groups
Savvy I ns -.39 It appears that the grade point average
Mnemonics 1 ns .07 groups differed significantly on strategies
Planfulness 1 ns .20
Formal reasoning 1.68 .19 .21 but not on abilities. Planfulness and formal
Note taking 6.02 .01 -.99 reasoning best discriminated among grade
Transformation 2.87 .09 .72 point average groups. When abilities were
partialed out, the clarity of the difference
Note. Multivariate F = 2.56, df = 7/213, p <
.01. among grade point average groups was en-
hanced.
reasoning. The present results show sex dif- TABLE 9
ferences in ability, which were not simply COMPARISON OF Low- , MIDDLE- , AND HIGH-
an artifact of different numbers of males GRADE-POINT-AVERAGE GROUPS ON STRATEGY
and females in different major fields. The MEASURES WITH ABILITIES
PARTIALED OUT
method of calculating F contrasts precluded
such artifactual findings. When males and Discrimi-
females were compared on self-reported Variable Univariate nant
F P function
strategy measures, there were again signifi- coefficient
cant differences. These differences reflected
the reported use of note taking, formal rea- Diligence <1 ns .14
r Savvy 1.93 .14 .05
soning, and transformation-application. It Mnemonics <1 ns .26
, appears that males claim greater use of for- Planfulness 3.39 .03 -.72
mal reasoning and lesser use of transforma- Formal reasoning 3.94 .02 -.95
tion and note taking than do females. When Note taking <1 ns .16
Transformation <1 ns .28
multivariate analysis of covariance was
employed to compare males and females on Note. Multivariate F = 1.73, df 19/426, p <
(a) abilities with strategies partialed out .04.
' '
370 ROY D. GOLDMAN AND DAVID J. HUDSON

. The findings in the present study lend success for high ability freshmen men. Personnel
support to the idea that strategies may be and Guidance Journal, 1964, 42, 603-607.
BRUNER, J., GOODNOW, J., & AUSTIN, G. A study oj
more fundamental determinants of aca- thinking. New York: Wiley, 1956.
demic success than are abilities. This state- COOLEY, W., & LOHNES, P. Multivariate data analy-
ment is supported not only by the strategy sis. New York: Wiley, 1971.
differences among grade point average CBOPLEY, A., & FIELD, T. Achievement in science
groups but also by the lack of ability differ- and intellectual style. Journal of Applied Psy-
ences. Naturally, the range of abilities rep- chology, 1969, 53, 132-135.
ELTON, C., & ROSE, H. Significance of personality
resented in our sample is truncated by the in the vocational choice of college women. Jour-
college selection process. It is possible that nal of Counseling Psychology, 1967,14, 293-298.
a wider range of abilities would have pro- FREDERIKSEN, C. Abilities, transfer and information
duced strikingly different results. However, retrieval in verbal learning. Multivariate Be-
havioral Research Monographs, 1969, 2, 1-82.
it appears that there may be considerable FRENCH, J. Social class and the relationship of
benefit to teaching study strategies to a problem solving styles to the factor composition
^college population. of tests. Educational and Psychological Measure-
There was an absence of interaction be- ment, 1965, 25, 9-28.
tween grade point average and major field FRENCH, J., EKSTROM, R., & PRICE, L. Kit of
reference tests for cognitive factors. Princeton,
; for either abilities or strategies. This finding N.J.: Educational Testing Service, 1963.
would suggest, in answer to the last ques- GOLDMAN, R., & WARREN, R. Configuration in dis-
tion posed earlier, that strategies necessary criminant space: A heuristic approach to study
for academic success are relatively consis- techniques. Paper presented at the meeting of
tent across major fields. the Western Psychological Association, Portland,
April 1972.
' REFERENCES GOLDMAN, R., & WARREN, R. Discriminant analysis
BIGGS, J. Faculty patterns in study behaviour. of study strategies connected with college grade
Australian Journal of Psychology, 1970, 22, 161- success in different major fields. Journal of Edu-
174. cational Measurement, 1973, 10, 39-47.
BROWN, F., & DCBOIS, T. Correlates of academic (Received September 21, 1972)

You might also like