Professional Documents
Culture Documents
ability: Harris, 1978; Maccoby & Jacklin, construct validity of spatial ability (e.g.,
1974; FBI: Saarni, 1973; formal reasoning Harris, 1980; McGee, 1979). A variety of
ability; Karplus, Karplus, Pormisano, & "spatial abilities" appear to exist.
Paulsen, 1977; Lawson, 1978; Linn & Pulos, Recently, Snow and his collaborators have
in press; course taking experience: Fennema analyzed the overlap among a wide range of
& Sherman, 1977,1979). It seems likely that spatial ability tests. Snow, Lohman, Mar-
no dingle explanation is sufficient to account shalak, Yalow, and Webb (Note 1) at-
for all the observed differences. Those we tempted to separate Cattell (1971) and Horn
selected for study represent all prevalent and Cattell's (1966) construct of general fluid
ideas about aptitude and experience influ- ability (as measured, for example, by letter
ences on gender differences in scientific series) from spatial visualization as mea-
reasoning. sured, for example, by Embedded Figures
Difficulties in establishing the relationship (Oltman, Raskin, & Witkin, 1971). Snow et
between aptitudes and gender differences al. (Note 1) could not separate General fluid
reflect the multiple processes involved in ability from Spatial visualization, so they
each. Aptitude or experience measures identified a combined dimension, which they
(such as spatial ability) are often thought to labeled General fluid visualization (Gfv).
contribute to male-female differences in Gfv was defined by tests requiring mental
scientific reasoning because (a) males and manipulation of figural or nonfigural mate-
females frequently differ in performance on rial and is represented by Paper Folding,
them and (b) they have high correlations Embedded Figures, and Letter Series
with measures of scientific reasoning. (French, Ekstron, & Price, 1963) in this
However, these two conditions are not suf- study.1
ficient to show that aptitude measures ex- Linn and Kyllonen (1981) clarified the
plain gender differences in scientific rea- relationship between Gfv and Witkin and
soning. For example, gender differences iii Goodenough's (1981) concept of FDI.
spatial visualization mean scores could re- Witkin and Goodenough (1981) had identi-
flect females' reluctance to guess on visual- fied two FDI dimensions. One, Cognitive
ization items, whereas high correlations be- Restructuring, was measured by Embedded
tween spatial visualization and scientific Figures. The other, Perception of the Up-
reasoning for both sexes could reflect ability right, was measured by the Rod-and-Frame
to maintains image in memory. Thus, the Test. Linn and Kyllonen (1981) combined
mean gender differences for spatial visual- measures of Gfv and both aspects of FDI.
ization and scientific reasoning could reflect They identified two dimensions. One, in-
different processes and therefore not con- cluded the Cognitive Restructuring tests but
tribute to the correlation between the two did not differ from what Snow et al. (Note 1)
measures. In this article we examine labeled Gfv< The other was characterized by
whether differences in aptitude and experi- Perception of the Upright but also included
ence can explain the gender differences in Perception of Horizontally and the Wes-
scientific reasoning. chler Picture Completion test; they labeled
Of the aptitudes associated with gender this Familiar field (Ff). The Ff dimension
differences, spatial ability is the most prev- was hypothesized to measure strategy se-
alent and the most difficult to define. In lection in familiar situations when compet-
many formulations it includes tests labeled ing strategies were available. Cognitive re-
FDI by others. Maccoby and Jacklin (1974) structuring was represented by a version of
identified gender differences in spatial embedded figures, and Ff was represented
ability on a number of different tests in- by Water Level (DeAvila & Havassy, with
cluding FDI tests such as Embedded Figures Pascual-Leone, 1976) in this study.
(Witkin, Moore, Goodenough, & Cox, 1977) Besides spatial ability in all its guises,
and the Rod-and-Frame Test (Oltman, some research suggests that gender differ-
1968). They did not specify a spatial con- ences in Piagetian formal reasoning might
struct. Most subsequent investigators have
catalogued gender differences on tests la- 1
We combined these measures, as described below,
beled "spatial" but have not studied the in some analyses to avoid problems of colinearity.
88 MARCIA C. LINN AND STEVEN PULOS
explain gender differences in scientific rea- Howe & Shayer, 1981). We investigated
soning (e.g.; Lawson, 1978). Formal rea- how weight related expectations influence
soning also suffers from poor construct def- performance on Predicting Displaced
inition (e.g. Martorano, 1977), possibly due Volume.
to the influence of experience with specific
item content on performance (e.g. Pulos & Method
Linn, 1981). Thus, as for spatial ability, we
employ several measures of Piagetian formal Subjects
reasoning and assess their independent and
combined relationship with our measure of Participants were 778 students in the 7th (159 fe-
scientific reasoning. males and 145 males), 9th (77 females and 136 males),
and llth grades (139 females and 122 males) in three
Experience differences for the sexes are districts. The school districts differed in location and
well established (e.g. Newcombe, Bandura, socioeconomic status (SES). Location and SES were
& Taylor, in press), but their impact on sci- (a) lower-middle-class semirural; (b) middle-class
entific reasoning is difficult to assess. Ex- urban; and (c) upper-middle-class suburban.
perience differences for males and females
include course taking, role expectations, Predicting Displaced Volume Test
media depictions, and so on. In this study,
we investigate the role of science and math Our Predicting Displaced Volume test was group
administered. A sample item is shown in Figure 1.
course taking. Each item pictured a cylinder half full of water and two
metal blocks. The relative size and weight of each block
Expectation-Based Strategy Usage was indicated both in the drawing and in the printed
question. Four types of items were used. Each item
had either equal or unequal volume and equal or un-
To clarify male-female differences in equal weight. Only the practice item combined equal
scientific reasoning, we selected Predicting volume and equal weight. Others showed unequal
Displaced Volume, which allowed specifi- volume, unequal weight, or unequal weight and
cation of expectation-based strategies used volume.
for correct responses and for incorrect re- Instructions were as follows: "All the blocks sink and
are completely covered by the water. I took one of the
sponses. The incorrect strategies in this blocks and put it in the water and then took it out.
task reflected inaccurate expectations con- Next, I took the second block, put it in the water, and
cerning which variables influence perfor- took it out Which block made the water go up higher?
mance. Analysis of the relationship between Block A, Block B, or did both blocks make the water go
up the same amount?"
strategy usage and aptitude enables us to After a practice item, subjects had five minutes to
clarify gender differences in performance. respond to eight items. All subjects completed the
Research on many scientific reasoning eight items before the time was up. Subjects appeared
problems suggests that students consistently to solve without much deliberation.
use the same accurate or inaccurate strategy Research on Predicting Displaced Volume (e.g., Pi-
aget, 1951a, 1951b) suggested that inaccurate strategies
for similar problems (e.g., Proportions: would incorporate some role for weight. Our Predicting
Karplus, Karplus, Formisano, & Paulsen, Displaced Volume items were designed so that each of
1977; Siegler, 1976; Subtraction: Brown & the inaccurate strategies we identified and the correct
Burton, 1978). As the study mentioned strategy would be reflected in a different pattern of
responses. Responses of subjects using strategies dif-
above by Linn & Swiney (1981) suggests, ferent from the four we identified would not be cate-
these strategies may be based on expecta- gorizable in our system. The four strategies we iden-
tions about the variables. tified were as follows:
How do expectations influence perfor- 1. Weight-only strategy. The amount of water
mance on Predicting Displaced Volume? displaced by an object immersed in water depends only
on the weight of the object.
Piaget (1951a, 1951b) and Piaget and In- 2. Weight-except-when-equal strategy. The
helder (1941) report interviews of subjects amount of water displaced by an object immersed in
asked to predict displaced volume. Piaget water depends on the weight of the object, except that
and others report that subjects commonly when two objects weigh the same, the one with the
greatest volume displaces the most water.
expect the weight of the object to influence 3. Volume-except-when-equal strategy. The
the volume displaced even when they know amount of water displaced by an object immersed in
the object sinks (e.g., Hobbs, 1973, 1975; water depends on the volume of the object, except when
PREDICTING DISPLACED VOLUME 89
I. Blocks A_ and £ are the same size. Block £ weighs more than Block A_.
10 oz. 20 oz.
Block A
Block £ '
two objects have the same volume then the one with the Piaget by DeAvila et al. (1976). Score is number correct
greatest weight displaces the most water. on eight trials.
4. Volume-only strategy. The amount of water
displaced by an object immersed in water depends only Formal Reasoning Measures
on the volume of the object. This strategy is correct.
To assess formal reasoning we used four tests similar
General and Spatial Ability Measures to those used by Inhelder and Piaget (1958):
1. The balance puzzle. This is a measure of pro-
We measured general ability and spatial ability in its portional reasoning (Linn & Pulos, in press). Score is
various guises. We measured general ability to disen- total correct of 13 items.
tangle its effects from other abilities in the regression 2. Controlling variables. This is a measure of con-
analysis. Tests used were as follows: trolling variables, a locally developed test comparable
1. Vocabulary. We measured general ability fol- to Springs (Linn & Rice, 1979). Score is number correct
lowing Cattell's (1971) description of General crystal- of 6 items.
lized ability (Gc), using a locally developed adaptation 3. Permutations. This is a locally developed test
of several multiple choice vocabulary tests (to ensure requiring the subject to generate the permutations of
a wide range). Score is sum of performance on two 2- four elements. The score is number of unique permu-
minute sections, tations generated in 2 minutes.
2. Letter series. We measured Cattell's (1971) 4. Combinations. This is a locally developed test
General fluid ability (Gf), to assess Snow et al.'s (Note requiring the subject to generate combinations of three
1) findings for spatial visualization using a local modi- elements. The score is number of unique combinations
fication of the French, Ekstron, & Price (1963) version. generated in 2 minutes.
Score is a sum of number correct on two 2-minute sec-
tions. Experience Measures
3. Find a Shape Puzzle (FASP). To measure Cog-
nitive Restructuring as defined by Witkin & Goode- Experience with math and science was assessed by
nough (1981) we used aversion of embedded figures in asking students to indicate how many years of math and
which the simple and complex shapes are on the same science courses they had taken.
page (Pulos & Linn, Note 2). Score is number of simple
shapes located in>4 minutes. Test Administration
4. Paper folding. A measure of spatial visualization
as studied by Snow et al. (Note 1); from the French et The measures were administered to class groups
al. (1963) battery was used. Score is performance on (15-28 students) in two 40-minute sessions.
two 2-minute sections.
5. Water level. A measure of what Linn & Kyllonen
(1981) called Familiar field and related to what Witkin Results ,
& Goodenough (1981) called Perception of the Upright
was used. The test measured perception of the hori- The anticipated male-female differences
zontal in tilted bottles of water and was adapted from in Predicting Displaced Volume emerged as
90 MARCIA C. LINN AND STEVEN PULOS
Males 63
Of the aptitude and experience measures,
.82 (Vocabulary, Letter Series, FASP, Paper
Females 49 folding, Water level, course taking) males
out-performed females only on Water Level
Males 74 .86
(Ff). Means and standard deviations are
Females 55 given in Table 2. It should be noted that
contrary to some earlier studies reported by
111
Maccoby and Jacklin (1974), we detected no
Males 82 male-female differences on FASP, our
.88
Females 62 measure of Embedded Figures. Other re-
cent studies (Petersen, Note 3; Fennema &
Sherman, 1979) have also found no male-
shown in Table 1. Sex accounted for 6% of female differences for measures similar to
the variance in 7th grade (r = .24, p < .01), Embedded Figures. In addition, we de-
8% in 9th grade (r = .29, p < .01), and 6% in tected no male-female differences in number
llth grade (r = .24). of years of math and science courses taken,
paralleling the findings of Armstrong (1979).
In contrast to our findings, differences in
Reliability of Predicting Displaced
Volume Table 2
Mean, Standard Deviation and r 2 for Each
The eight-item test of Predicting Dis- Aptitude and Experience Measure by Sex
placed Volume was reliable using Cronbach's
alpha (Cronbach, 1951), as shown in Table Males Females
1 (coefficient alpha was .82 in seventh grade,
Measure M SD M SD r2a
and .88 in eleventh grade). All but the eas-
iest items contributed their share to the re- Aptitude
liability. (On average, the reliability with
each item deleted was about .03 lower than Vocabulary (Gc) 26.56 6.80 26.06 7.27 00
Series (Gfv) 13.79 5.66 14.93 5.72 01
the reliability in Table 1.) FASP (Gfv) 9.11 5.11 8.86 5.08 00
To determine whether our strategy as- Paper Folding
signments were reliable, we conducted a (Gfv) 10.10 4.26 9.38 4.05 01
scalability analysis (Mclver & Carmines, Water Level (Ff) 17.04 6.87 14.42 7.08 05*
1981). The coefficient of scalability was .99. Proportions
(formal) 6.32 2.49 5.27 2.07 05*
The coefficient of reproductability was .98. Controlling
Thus, strategy usage was very consistent (formal) 5.50 3.25 4.72 3.47 01
across items. Permutations
(formal) 15.30 5.62 15.87 4.57 00
Combinations
Validity of Predicting Displaced Volume (formal) 11:15 3.80 11.41 4.70 00
Table 4
Predicting Displaced Volume: Correlations of Aptitude and Course Experience by Grade and
Sex (Decimals Omitted)
Grade 7 Grade 9 Grade 11
Measure M M M
Aptitude
Vocabulary (Gc) 21 -07 28 36 26 28
Series (Gfv) 20 13 27 28 20 24
FASP (Gfv) 13 23 36 21 34 20
Paper Folding
(Gfv) 30 21 13 36 28 31
Water Level (Pf) 21 26 32 35 23 33
Proportions
(formal) 19 -08 31 10 16 21
Controlling
(formal) 18 -04 20 19 11 -08
Permutations
(formal) 15 -13 06 12 01 11
Combinations
(formal) 09 -05 07 05 19 17
Course Experience
Math 05 15 26 20 20 20
Science 08 -06 10 -05 23 26
Note. M = male; F = female; Gc = General crystallized ability; Gfv = General fluid visualization; Ff = Familiar
field; FASP = Find a Shape Puzzle.
PREDICTING DISPLACED VOLUME 93
measures clarified what Predicting Dis- ferential experiences? Our measure of math
placed Volume measures for both sexes and and science experience correlates with Pre-
clarified the Ff dimension. Third, our ap- dicting Displaced Volume but does not ex-
titude and experience measures did not ac- plain male-female differences in perfor-
count for male-female differences in strategy mance.
usage, although they correlated substantially Course taking may no longer be an effec-
with performance for both sexes. tive measure of math and science experience,
Female enrollment in advanced courses has
Expectation-Based Strategy Usage increased (e.g., Fennema & Sherman, 1977;
Armstrong, 1979). Recent efforts to en-
Over 90% of the subjects used one of the courage, females to continue in math and
expectation-based strategies we anticipated. science may result in poorly prepared fe-
Females used weight-based inaccurate males taking advanced courses. This could
strategies more frequently than did males. occur either because females did not take
The strategies based on weight rather than appropriate prerequisite courses or because
volume may reflect inaccurate or overgen- they were treated differently in the courses
eralized expectations about weight. Thus, they took.
male-female differences in Predicting Dis- Are females treated differently? Females'
placed Volume may be due to differential experiences differ from those of males
expectations about weight. throughout the life span. Many females
Weight is irrelevant in Predicting Dis- describe themselves as math-anxious (e.g.,
placed Volume if the objects sink. Why Tobias, 1978), as unable to solve math
then do subjects consider weight? They. problems (e.g., Covington & Omelich, 1979),
may confuse floating and sinking with dis- and as "confused" or "unable to reason"
placement and consider weight because it about machines (e.g., Linn, 1980b). Class-
influences floating and sinking. (Some room experiences for males and females
subjects mentioned that heavy things sink differ, e.g. teachers praise females less than
faster than light things of the same volume.) they do males (e.g., Brophy & Good, 1970),
In addition, subjects may generalize their and females receive more negative feedback
knowledge of weight and expect weight to for the intellectual quality of their work
influence displacement just as it influences (Dweck & Reppucci, 1973; Dweck, Davidson,
other phenomena such as balance beams or Nelson, & Enna, 1978). Research on free-
bending rods. choice environments reveals that females
Expectations about weight influence choose different experiences than males in
performance on another complex reasoning math and science (e.g., Linn, 1980a; Rice &
task—the pendulum, (e.g., Linn, 1977, and Linn, 1978). Thus, other experience dif-
Pulos & Linn, 1978). Most subjects expect ferences not reflected in our measure of
weight to influence pendulum oscillation just course taking may influence male-female
as they expect weight to influence displaced differences in Predicting Displaced
volume. Linn (1977) attempted to alter Volume.
expectations about the influence of weight
on pendulum oscillation by providing some Aptitude Clarification of Predicting
experimental evidence that weight was not Displaced Volume
important in pendulum oscillation. She
found that subjects correctly interpreted the By establishing the relationship between
experimental evidence but continued to ex- aptitudes and Predicting Displaced Volume,
pect weight to influence pendulum oscilla- we clarified what Predicting Displaced
tion in subsequent trials. Thus, subjects Volume measured. The aptitudes we mea-
have strong expectations about the influence sured accounted for about 25% of the vari-
of weight that are not easily altered. It ap- ance in Predicting Displaced Volume.
pears that for Predicting Displaced Volume, The single strongest predictor of Pre-
males have more accurate expectations dicting Displaced Volume was Ff, an aspect
about weight than do females. of FDI that also accounted for a significant
Are expectation differences due to dif- portion of the variance after Gc and Gfv were
PREDICTING DISPLACED VOLUME 95
entered in the regression. This finding single aptitude explanation is not sufficient
combined with the expectation-based to account for gender differences in spatial
strategy analysis of Predicting Displaced ability, formal reasoning, and scientific
Volume is congruent with Linn and Kyllo- reasoning.
nen's (1981) definition of Ff as a measure of
strategy selection from among salient com- Reference Notes
peting strategies, and supports Witkin &
Goodenough's (1981) assertion that aspects 1. Snow, R. E., Lohman, D. P., Marshalek, B., Yalow,
E., & Webb, N. Correlational analyses of reference
of FDI measure a unique aspect of reasoning. aptitude constructs (Tech, Rep. No. 5). Stanford,
Similar findings by Linn (1978) and Linn Calif.: Stanford University School of Education
and Swiney (1981) lend further support to Aptitude Research Project, September 1977.
this hypothesis. Ff may measure ability to 2. Pulos, S., & Linn, M. C. The find a shape puzzle
select from among expectation-based strat- (FASP). Berkeley, Calif.: Lawrence Hall of
Science, 1979,
egies. 3. Peterson, A. Sex related differences in proportional
Is formal reasoning a unique aptitude? A reasoning: Spatial visualization, affective, social,
variety of studies suggest that Gc, Gfv, and and other factors. In M. C, Linn (Chair), Sex dif-
Ff overlap substantially with formal rea- ferences in scientific reasoning. Symposium pre-
soning performance (e.g., Linn, Pulos, & sented at the Sixth Annual Conference on Research
on Women and Education of the American Educa-
Gans, 1981; Cloutier and Goldschmid, 1976). tional Research Association, Pacific Grove, Califor-
Some have suggested that the overlap is nia, December 7-9, 198(5.
complete (e.g., Humphreys & Parsons, 1979),
whereas others think there may be unique References
variance in formal reasoning tasks (e.g., Linn
& Swiney, 1981). Our study suggests that Armstrong, J. M. A National Assessment of achieve-
ment and participation of women in mathematics.
formal reasoning does not contribute unique Education Commission of the States, Boulder, Col-
variance to Predicting Displaced Volume orado, 1979.
performance after Gc, Gfv, and Ff and that Brophy, J. E., & Good, T. L. Teachers' communication
formal reasoning aptitude does not influence of differential expectations for children's classroom
performance: Some behavioral data. Journal of
male and female performance differently. Educational Psychology, 1970,61, 365-374.
Brown, J. S., & Burton, R. R. Diagnostic models for
Aptitude Explanations of Gender procedural bugs in basic mathematical skills. Cog-
nitive Science, 1978, 2,155-192.
Differences Cattell, R. B. Abilities: Their structure, growth and
action. Boston: Houghton Mifflin, 1971.
Male-female differences in Predicting Cloutier, R., & Goldschmid, M. Individual differences
Displaced Volume remained unexplained in the development of formal reasoning. Child De-
after measures of spatial ability, FDI, formal velopment, 1976, 47,1097-1102.
reasoning, and course-taking experience 'Covington, M., & Omelich, C. Effort: The double-
edged sword in school achievement. Journal of
were entered in the regression. All com- Educational Psychology, 1979, 70,169-182.
monly suggested aptitude explanations of Cronbach, L. J. Coefficient alpha and the internal
male-female differences were represented structure of tests. Psychometrika, 1951, 16,
in our measures. In particular, spatial vi- 297-334.
sualization Was represented by several dif- DeAvila, E. A,, & Havassy, B., with Pascual-Leone, J.
Mexican-American Schoolchildren: A neo-Piage-
ferent tests. Several measures that also tian analysis. Washington, D.C.: Georgetown
yielded male-female differences (e.g., pro- University Press, 1976.
portions, Water Level) were used. These Dweek, C. S., Davidson, W., Nelson, S., & Enna, B. Sex
measures correlated with Predicting Dis- differences in learned helplessness II, The con-
tingencies of evaluative feedback in the classroom and
placed Volume performance but did not ac- HI, An experimental analysis. Developmental
count for gender differences. The lack of Psychology, 1978,14, 268-276,
relationship between male-female differ- Dweck, C. S., & Reppucei, N. D. Learned helplessness
ences on these tasks suggests the existence and reinforcement responsibility in children.
of multiple explanations for gender differ- Journal of Personality and Social Psychology, 1973,
25, 109-116.
ences. No evidence for a single factor ex- Fennema, E., & Sherman, J. Sex-related differences
plaining male-female differences on these in mathematics achievement, spatial visualization
tasks emerged. These data suggest that a and sociocultural factors. American Educational
96 MARCIA C. LINN AND STEVEN PULOS