expectancies, values, and achievement in junior high school

8
Journal of Educational Psychology 1990, Vol. 82, No. 2,319-326 Copyright 1990 by the American Psychological Association, Inc. 0022-O663/9O/S0O.75 Expectancies, Values, and Achievement in Junior High School Thomas J. Berndt and Kristelle E. Miller Purdue University Many theorists have suggested that students' motivation to achieve in school depends on their expectancies for success and the value they attach to success. There are few data, however, on the relation between expectancies and values or their relative contribution to achievement. To examine these issues, we asked 153 seventh graders to complete multiple measures of academic expectancies and values. We used students' report card grades and academic track placements in English and math as indicators of their achievement. Covariance structure analyses showed that students' expectancies were more strongly related to their achievement than were their values. Nevertheless, both expectancies and values made significant, independent contributions to achievement. In addition, the constructs for expectancies and values were positively correlated. Boys and girls had similar expectancies, but boys appeared to value academic success less than did girls. Students* motivation to achieve in school is affected by their expectancies for success and by the value they attach to success in academics. This statement expresses the central proposition in expectancy-value theories of achievement mo- tivation (Atkinson & Feather, 1966; Eccles, 1983; Feather, 1988). Moreover, constructs related to expectancies and val- ues have an important place in other theories of achievement motivation (e.g., Ames & Archer, 1988; Dweck & Leggett 1988; Nicholls, 1984; Weiner, 1985). Previous research has shown that students' expectancies and values are related to their academic achievement and their educational plans (e.g., Eccles, Adler, & Meece, 1984). There is little research, how- ever, on two questions. First, what are the relative contribu- tions of expectancies and of values to students' academic achievement (cf. Dweck & Elliot 1983; Feather, 1988)? Sec- ond, how are expectancies and values related to each other? To clarify these issues, we obtained multiple measures of the expectancies, values, and achievement of junior high school students. Then we examined the relations of expectancies and values to achievement and to each other. Theorists differ in their assumptions about the relative importance of expectancies and values for achievement. On one hand, Dweck (Dweck & Elliott 1983; Dweck & Leggett, 1988) has emphasized the potentially negative impact on achievement of low or unstable expectancies. On the other hand, Eccles (1983) has argued that students' values have a greater long-term effect on their achievement than do their expectancies, because values affect curriculum choices that determine students' exposure to higher-level concepts in aca- demic subjects. The research reported in this article was supported in part by a grant from the Spencer Foundation. We gratefully acknowledge the Foundation's assistance. We also wish to thank the principals and teachers of the schools that cooperated in the research and the students who participated in the studies. Kristelle MiUer is now at the Department of Psychology and Mental Health, University of Minnesota at Duluth. Correspondence concerning this article should be addressed to Thomas J. Berndt, Department of Psychological Sciences, Purdue University, West Lafayette, Indiana 47907. The research findings on the relations of expectancies and values to achievement are limited and inconsistent. Eccles and her colleagues (1984) found that junior high school students' grades in English and math courses were more strongly related to their expectancies than to their values. By contrast, students' plans for future English and math courses were more strongly related to values than to expectancies. In another study (Feather, 1988), college students' course plans were more strongly affected by their perceptions of their math ability, which were viewed as indicators of math expec- tancies, than by the value the students attached to math. Students' course plans were also affected, however, by the value they attached to English courses, but not by their perceptions of their ability in English. No data on students' actual achievement were reported. Platt (1988) related students' achievement during the first semester of college to earlier measures of expectancy of success and effort attributions for success in high school. Previous research suggests that effort attributions can be viewed as indicators of achievement values (Dweck & Elliott, 1983). Platt found that both expectancies and values (or effort attri- butions) were significantly related to achievement. Their rel- ative importance is not clear, however, because Platt reported only the unstandardized parameter estimates from a covari- ance structure (LISREL) analysis. We used covariance structure analysis in this study to examine the relations of expectancies and values to junior high school students' achievement in English and in math. We focused in most analyses on the students' mean level of achievement across both subjects because, as noted below, students' grades in the two subjects were moderately corre- lated and our measures of expectancies and values were not subject-specific. Our first hypothesis, which was based on the research of Eccles et al. (1984), was that students' achievement would be more strongly related to their expectancies than to their values. Our second hypothesis concerned the relation between expectancies and values. Atkinson and Feather (1966) origi- nally proposed that these two constructs are negatively related (see Feather, 1988). Most other theorists have suggested that expectancies and values are positively related. For example, 319

Upload: kristelle-e

Post on 27-Jan-2017

216 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Expectancies, values, and achievement in junior high school

Journal of Educational Psychology1990, Vol. 82, No. 2,319-326

Copyright 1990 by the American Psychological Association, Inc.0022-O663/9O/S0O.75

Expectancies, Values, and Achievement in Junior High School

Thomas J. Berndt and Kristelle E. MillerPurdue University

Many theorists have suggested that students' motivation to achieve in school depends on theirexpectancies for success and the value they attach to success. There are few data, however, onthe relation between expectancies and values or their relative contribution to achievement. Toexamine these issues, we asked 153 seventh graders to complete multiple measures of academicexpectancies and values. We used students' report card grades and academic track placements inEnglish and math as indicators of their achievement. Covariance structure analyses showed thatstudents' expectancies were more strongly related to their achievement than were their values.Nevertheless, both expectancies and values made significant, independent contributions toachievement. In addition, the constructs for expectancies and values were positively correlated.Boys and girls had similar expectancies, but boys appeared to value academic success less thandid girls.

Students* motivation to achieve in school is affected bytheir expectancies for success and by the value they attach tosuccess in academics. This statement expresses the centralproposition in expectancy-value theories of achievement mo-tivation (Atkinson & Feather, 1966; Eccles, 1983; Feather,1988). Moreover, constructs related to expectancies and val-ues have an important place in other theories of achievementmotivation (e.g., Ames & Archer, 1988; Dweck & Leggett1988; Nicholls, 1984; Weiner, 1985). Previous research hasshown that students' expectancies and values are related totheir academic achievement and their educational plans (e.g.,Eccles, Adler, & Meece, 1984). There is little research, how-ever, on two questions. First, what are the relative contribu-tions of expectancies and of values to students' academicachievement (cf. Dweck & Elliot 1983; Feather, 1988)? Sec-ond, how are expectancies and values related to each other?To clarify these issues, we obtained multiple measures of theexpectancies, values, and achievement of junior high schoolstudents. Then we examined the relations of expectancies andvalues to achievement and to each other.

Theorists differ in their assumptions about the relativeimportance of expectancies and values for achievement. Onone hand, Dweck (Dweck & Elliott 1983; Dweck & Leggett,1988) has emphasized the potentially negative impact onachievement of low or unstable expectancies. On the otherhand, Eccles (1983) has argued that students' values have agreater long-term effect on their achievement than do theirexpectancies, because values affect curriculum choices thatdetermine students' exposure to higher-level concepts in aca-demic subjects.

The research reported in this article was supported in part by agrant from the Spencer Foundation. We gratefully acknowledge theFoundation's assistance. We also wish to thank the principals andteachers of the schools that cooperated in the research and the studentswho participated in the studies.

Kristelle MiUer is now at the Department of Psychology and MentalHealth, University of Minnesota at Duluth.

Correspondence concerning this article should be addressed toThomas J. Berndt, Department of Psychological Sciences, PurdueUniversity, West Lafayette, Indiana 47907.

The research findings on the relations of expectancies andvalues to achievement are limited and inconsistent. Ecclesand her colleagues (1984) found that junior high schoolstudents' grades in English and math courses were morestrongly related to their expectancies than to their values. Bycontrast, students' plans for future English and math courseswere more strongly related to values than to expectancies.

In another study (Feather, 1988), college students' courseplans were more strongly affected by their perceptions of theirmath ability, which were viewed as indicators of math expec-tancies, than by the value the students attached to math.Students' course plans were also affected, however, by thevalue they attached to English courses, but not by theirperceptions of their ability in English. No data on students'actual achievement were reported.

Platt (1988) related students' achievement during the firstsemester of college to earlier measures of expectancy of successand effort attributions for success in high school. Previousresearch suggests that effort attributions can be viewed asindicators of achievement values (Dweck & Elliott, 1983).Platt found that both expectancies and values (or effort attri-butions) were significantly related to achievement. Their rel-ative importance is not clear, however, because Platt reportedonly the unstandardized parameter estimates from a covari-ance structure (LISREL) analysis.

We used covariance structure analysis in this study toexamine the relations of expectancies and values to juniorhigh school students' achievement in English and in math.We focused in most analyses on the students' mean level ofachievement across both subjects because, as noted below,students' grades in the two subjects were moderately corre-lated and our measures of expectancies and values were notsubject-specific. Our first hypothesis, which was based on theresearch of Eccles et al. (1984), was that students' achievementwould be more strongly related to their expectancies than totheir values.

Our second hypothesis concerned the relation betweenexpectancies and values. Atkinson and Feather (1966) origi-nally proposed that these two constructs are negatively related(see Feather, 1988). Most other theorists have suggested thatexpectancies and values are positively related. For example,

319

Page 2: Expectancies, values, and achievement in junior high school

320 THOMAS J. BERNDT AND KRISTELLE E. MILLER

Dweck and Leggett (1988) argued that students who attributetheir failure on a task to low ability (and thus have a lowexpectancy of success) often show a lack of interest in the task(cf. Nicholls, 1984). Eccles (1983) speculated that studentslower their estimate of a task's value if they believe theycannot succeed on the task. This speculation also implies thatexpectancies and values are positively related. Eccles (1983)reported positive correlations between adolescents' expectan-cies and values; both Feather (1988) and Platt (1988) reportedpositive correlations among measures that reflect the expec-tancies and values of college students. We hypothesized thatconstructs defined by multiple measures of junior high schoolstudents' expectancies and values would also be positivelyrelated.

A secondary purpose of our study was to examine possiblesex differences in expectancies and values related to academicsuccess. Dweck (1986) proposed that many girls, especiallygirls high in achievement, have low or unstable expectanciesfor academic success. Most researchers have not found sexdifferences consistent with this proposal (Eccles et al., 1984;Feather, 1988; Wigfield, 1988). More often, researchers havefound sex differences in measures related to achievementvalues. In some samples, boys report lower interest in school*work (Wigfleld, 1988) and less effort on academic activitiesthan do girls (Ryckman & Peckham, 1987). These findingsare consistent with claims that many boys value academicsuccess less than girls do (Dweck & Elliott, 1983). Yet becauseof the mixed findings in past research, we did not advancespecific hypotheses about sex differences.

In summary, this study tested two hypotheses: (a) students'achievement is more strongly related to their expectanciesthan to their values, and (b) expectancies and values arepositively related. The study included multiple measures ofstudents' expectancies and values. Expectancies are oftenestimated from students' ratings of their ability (e.g., Feather,1988), because the two types of measures are strongly corre-lated (Dweck & Elliott, 1983; Eccles, 1983; Nicholls, 1984).In addition, students' ability self-ratings are correlated withtheir attributions of past successes and failures to ability(Marsh, Cairns, Relich, & Debus, 1984). Thus, we assumedthat measures of perceived ability and of ability attributionsfor success and failure could be used as multiple indicators ofstudents' expectancies. Similarly, the value that students at-tach to academic success is related, in theory, to their involve-ment in their schoolwork and their attributions of their aca-demic successes and failures to effort and interest (Dweck &Elliott, 1983; Eccles, 1983). Thus we assumed that measuresof these behaviors and attitudes could serve as multiple indi-cators of students' values. Our data analyses provided tests ofthese assumptions.

Method

informed consent forms more often than boys did. Nearly all studentswere White. Their families represented a range of socioeconomicstatus groups.

Achievement Motivation

All measures of students' motivation to achieve in school are listedin Table 1. In addition, the table lists the reliability (coefficient alpha)and the number of items on each measure. All measures wereincluded in questionnaires administered to small groups of students.The measures were presented in the order listed in the table. A briefdescription of each measure follows.

Involvement. As noted earlier, we assumed that students' involve-ment in their schoolwork is one indicator of the value they attach toacademic success. To assess school involvement, we adapted theinvolvement subscale from the Classroom Environment Scale (Moos& Trickett, 1974). The original items referred to the attitudes andbehavior of "the kids in your class." The revised items referred tostudents' own attitudes and behavior in the classroom and out ofschool. For example, students answered the question, "How often doyou take part in classroom discussions?" They responded to eachitem on a 5-point scale, with higher scores indicating greater involve-ment.

School value. Eccles et al. (1984) used a measure of the value ofschool that we modified for this study. Students reported their beliefsabout the utility of school learning, the importance of school success,and their interest in their schoolwork. In contrast to the items ofEccles et al- (1984), which referred specifically to English or mathclasses, our items referred to schoolwork in general. Students wereasked, for example, "How valuable do you think your education willbe in getting the job you want?" They responded on a 5-point scale.

Perceived scholastic competence. Because students' perceptionsof their ability reflect their achievement expectancies (e.g.. Feather,1988), we included the subscale for perceived scholastic competencefrom Harter's (1985) Self-Perception Profile for Children as an indi-cator of the expectancy construct. A sample item reads "Some kidsoften forget what they learn BUT other kids remember things easily."Students first decide which type of child is more like them, and thenwhether the statement is "sort of true" or "really true" for them. Eachitem is scored on a 4-point scale with higher scores for greaterperceived competence.

Attributions. Students made attributions for their academic suc-cesses and failures by rating the importance of various causes ofspecific outcomes (cf. Marsh et al., 1984). For example, the studentsrated four possible reasons for a good grade on a report card: (a) they

Table 1Coefficient Alpha Estimates of Reliability and Number ofItems for Each Measure

Subjects

The sample included 153 seventh graders {mean age = 13 yearsand 3 months) from a junior high school in a small town. There weremore girls (99) than boys (54) in the sample, partly because therewere more girls in the seventh grade and partly because girls returned

Measures

InvolvementSchool valuePerceived scholastic competenceAttributions of success to

AbilityEffortInterestExternal causes

Attributions of failure toAbilityEffortInterestExternal causes

a

.83

.84

.81

.72

.77

.70

.38

.78

.70

.77

.63

Items

12186

5555

5555

Page 3: Expectancies, values, and achievement in junior high school

EXPECTANCIES, VALUES, AND ACHIEVEMENT 321

have high ability (e.g., "you are smart"), (b) they put a lot of effortinto their work (e..g, "you studied hard"), (c) they were interested inthe subject (e.g., "you were interested in the subject"), or (d) therewere external causes of their success (e.g., "your teacher helped you").The attributions to external causes always referred to teachers' atti-tudes or actions (e.g., "the teacher doesn't like you"). Students ratedeach reason on a 5-point scale ranging from not at all important (0)to very important (4).

The students responded to five situations that referred to academicsuccesses and five situations that referred to academic failures. Foreach situation, students rated the importance of ability, effort, interest,and external causes. These four types of attributions were examinedbecause young children often mentioned them in previous research(Elig & Frieze, 1979;Nicholls, 1984; Wigfield, 1988). Most significantfor this study were students' attributions to ability and to effort,because these attributions are central to expectancy-value theories ofachievement motivation (cf. Ames & Archer, 1988; Dweck & Leggett,1988; Eccles et al., 1984). The alpha coefficients for the measures ofexternal attributions were relatively low (see Table 1), as in previousresearch (Marsh et al., 1984), so results for these measures must beinterpreted cautiously.

Academic Achievement

Each student's English and math teachers reported the grade thatthe student had received in these subjects on the last report card. Thestudents' class schedules provided information about their academictrack placements in English (regular or basic) and math (advanced,regular, or basic). Students' mean grades in English and math werecorrelated with a mean score for their track placements (r = .49, p <.001). The positive correlation implied that track placements couldbe used along with mean grades as measures of academic achieve-ment. There were also positive correlations between students' gradesin English and in math (r = .59, p < .001) and between students'track placements in the two subjects (r = .41, p < .001). Thecorrelations suggested that separate analyses of the measures for eachsubject would yield largely redundant information. Therefore,achievement scores averaged across both subjects were used in mostanalyses.

Results

Intercorrelations of the Measures

Table 2 shows the correlations among the measures ofmotivation and achievement. There were strong or moderatecorrelations among the presumed measures of achievementvalue: involvement, school value, and attributions of successand failure to effort and interest. Strong or moderate corre-lations also existed among the presumed measures of achieve-ment expectancy: students' perceptions of their scholasticcompetence and ability attributions for success and failure.In addition, significant correlations that varied in size werefound among all types of success attributions, and among alltypes of failure attributions.

Most of the motivation-related measures were significantlycorrelated with report card grades and school track place-ments. Grades were most strongly correlated with students'perceptions of their scholastic competence. Track placementswere most strongly correlated with students1 attributions offailure to lack of ability. These correlations suggest that aca-

demic achievement was more closely Linked to students' ex-pectancies than to their values. Further analyses were done toexamine this conclusion more systematically.

Table 2 also shows the correlations of each measure withsex. The variable for students' sex was coded so that positivecorrelations indicate higher scores for girls than for boys. Girlsreported greater classroom involvement and placed a highervalue on their schoolwork than boys did. Girls were morelikely than boys to attribute their successes to their effort; theywere less likely than boys to attribute their failures to a lackof interest and external causes. Boys and girls did not differin their perceived scholastic competence or in their abilityattributions. In other words, girls scored higher than boys onthe presumed measures of achievement values; they did notdiffer from boys on the presumed measures of achievementexpectancies. Actual achievement judged by report cardgrades was higher for girls than boys.

Finally, the intercorrelations of the measures were exam-ined for the two sexes separately. Only 6 of the 78 correlations(or less than 8%) differed significantly for the two sexes. Inaddition, a test of the covariance matrices for boys and girlswith Box's M-statistic (Winer, 1971) did not yield a significantdifference, x2(91) = 113.9, p > .05. Thus the relations amongthe measures were largely comparable for the two sexes.Consequently, further analyses were done with the entiresample.

Structural Equation Models of the Relations ofMotivation to Achievement

In the next set of analyses, we examined the relations ofmotivation to achievement in a more integrated and theoret-ically guided way. In a first step, each measure was provision-ally linked to a latent construct for expectancies, values, orachievement. Then we used the LISREL VI program (Joreskog& Sorbom, 1984) to test a model of the relations of themeasures to the constructs and the relations among the con-struct. Once a model that adequately fit the data was obtained,additional models were examined to identify the measures ofexpectancies and values that related most strongly to achieve-ment.

To help specify the measurement model for the initialLISREL analysis, we examined the relations among the moti-vation measures in a principal-components factor analysis.This analysis yielded three factors with eigenvalues greaterthan 1.0 that accounted for 70% of the variance. As expected,one factor had high loadings (>.45) for the measures ofperceived scholastic competence and ability attributions forsuccess and failure. In the initial structural equation model,shown in Figure 1, these three measures were treated asindicators of the latent construct for achievement expectancy.

On a second factor, the measures of involvement, schoolvalue, and attributions of failure to effort, interest, and exter-nal causes had high loadings. These measures were treated asindicators of a latent construct for achievement values asso-ciated with explanations for academic failures (see Figure 1).On the third factor, the measures of involvement and schoolvalue again had high loadings, along with attributions of

Page 4: Expectancies, values, and achievement in junior high school

322 THOMAS J. BERNDT AND KRISTELLE E. MILLER

Table 2Intercorreiations of the Motivation-Related and Achievement Measures

Measure

1. INV2. SVAL3, SCHL4. SABL5. SEFF6. SINT7. SXT8. FABL9. FEFF

10. FINT11. FXT12. GRDS13. TRK14. Sex

1

42***.28***.65***56***.36***

-.26**—.54***-^64***-.35***

.32***

.24**

2

.31**"

.27**

.65**'

.59**11

.29**'- . 1 9 *- .49** '- .58** '- .36** '

,41**-.27**

3

<.57***

> 26**" 29**** .08

— A i • • •"~.D 1' -.22**

' -.24**K .53***

.39***.40*** - . 0 9

4

34***.46***.27**

-.38***.00

-.07-.14

,34***.24**

-.03

5

.52***

.03-.36***-.40***-.11

.26**

.16gA***

6

38***-.13-.26**

-~^25**.24**.13,11

7

.23**-.19*-.19*-.04

.10-.04

.10

8

.17*

.22**44***

_;36***-.48***

.08

9

.59***

.21*-.11-.09-.10

10

.45***™. 7 8 * * *

™ 14

™ 24**

11 12 13

- .2 !*-.31*** .49***-.28*** .17* .15

Note. INV = involvement; SVAL - school value: SCHL - perceived scholastic competence; SABL = success attributions, ability; SEFF =success attributions, effort; SINT = success attributions, interest; SXT = success attributions, external; FABL = failure attributions, ability;FEFF « failure attributions, effort; FINT = failure attributions, interest; FXT = failure attributions, external; GRDS - report card grades;TRK » school track. Sex was coded 2 for girls and 1 for boys, so positive correlations indicate higher scores for girts.*p<.05. **p<.01. ***p< .001.

success to effort, interest, and external causes. These measuresserved as indicators of a latent construct for achievementvalues associated with explanations for academic success.1 Tocomplete the measurement model, report card grades andschool track placements were treated as indicators of academicachievement (see Figure 1).

The arrows connecting the latent constructs in Figure 1define what Joreskog and Sorborn (1984) call the structuralmodel. Evidence on the relations among these constructs wasmost relevant to our major hypotheses.

The initial run of the program showed that the model inFigure t did not fit the data well, x2 (57, N = 153) - 188.58,p < .001, goodness-of-fit index (Joreskog & Sorbom, 1984;Tanaka, 1987) = .84. The fit of the model was substantiallyimproved by allowing for nonzero covariances (or correlatederrors) for 7 of the 55 elements in the matrix of measurementerror covariances. All instances of correlated errors involvedthe attribution measures. Correlated errors were allowed be-tween ability attributions for success and the other three typesof success attributions, between external attributions for fail-ure and effort and ability attributions for failure, and betweenfailure attributions to ability and success attributions to effortand external causes. The correlated errors apparently reflectthe effects of assessing all types of attributions with the samesituations and response format.

The model with correlated errors yielded a substantiallyreduced but still significant chi-square, x2 (50, N = 153) =91.4, p < .001. The goodness-of-fit index was .92, whichindicates a reasonably adequate fit (Tanaka, 1987). In addi-tion, the most important parameter estimates were com-parable in the models with and without correlated errors.More specifically, the standardized estimates for the six pa-rameters linking the latent constructs differed by an averageof only .05 (range - .00-. 11) in the two models. Figure 1shows the standardized parameter estimates for the finalmodel

Of the three motivation-related constructs, the one forexpectancies was most strongly related to achievement. The tvalue for the parameter indicating the expectancy-achieve-ment relation was 5.54, p < .001. The parameter linkingsuccess-related values to achievement was weaker but stillsignificant, t = 1.97, p < .05. The construct for failure-relatedvalues was not significantly related to achievement,t - 0.06, All parameters indicating the relations among thethree motivational constructs were significant, ts > 3.31, ps <.01.

At first glance, these results imply that students' attributionsof failure to lack of effort, lack of interest, and external causeshave no effect on academic achievement. This conclusionwould have considerable theoretical significance, becausemost attribution-retraining programs try to increase students'tendencies to attribute their failures to a lack of effort (Fors-terling, 1985). The conclusion may be incorrect, however,because the measures of involvement and school value alsoserved as indicators of the latent construct for failure-relatedvalues.

To examine more precisely the relations of value-relatedattributions to achievement, we performed a second analysiswithout the measures of involvement and school value. Thatis, attributions of success and failure to effort, interest, andexternal causes served as the only indicators of the values'constructs. The indicators of the expectancy construct werenot changed.

! Double loadings like those for the involvement and school-valuemeasures are not typical in structural equation models. Yet, as Table2 shows, the double loadings are required because these two measureswere correlated with both success and failure attributions. Moreover,structural equation models in which involvement and school valuewere treated as indicators of a single latent construct for achievementvalue did not fit the data satisfactorily. Indeed, these models failed toconverge.

Page 5: Expectancies, values, and achievement in junior high school

EX

PEC

TA

NC

IES,

VA

LU

ES, A

ND

A

CH

IEV

EM

EN

T323

Page 6: Expectancies, values, and achievement in junior high school

324 THOMAS J. BERNDT AND KRISTELLE B. MILLER

Once again, a model that allowed correlated errors fit thedata better than one with no correlated errors (goodness-of-fit indices = .85 and .95, respectively), but in both models theexpectancy construct again was strongly related to achieve-ment (p$ < .001). In both models, the two constructs repre-senting effort, interest, and external attributions for successand failure were not significantly related to achievement (ps> .20). Note that this analysis estimated all relations amongthe variables (i.e., all parameters) simultaneously. Thus theresults show that students' attributions to effort, interest, andexternal causes were not significantly related to achievementonce their relation to ability attributions was taken intoaccount.

What, then, accounts for the significant association inFigure 1 between success-related values and achievement?This association might be due to the two measures of achieve-ment values not included in the second set of analyses, thosefor involvement and school value. To test this hypothesis, weexamined another model in which involvement and schoolvalue were the only indicators of the values' construct. Thismodel, shown in Figure 2, fit the data satisfactorily withoutcorrelated errors, x3(H,iV= 153) = 22A,p< .05, goodness-of-fit index = .96. The standardized parameter estimate forexpectancies was nearly twice as large as that for values, butboth constructs contributed significantly to achievement, ;s =5.27, p < .001, and 3.15, p < .01, respectively. Moreover,

there was a significant relation between the expectancy andvalue constructs, / = 4.71, p < .01.

Finally, to examine the contributions of expectancies andvalues to achievement in English and in math, we tested themodel shown in Figure 2 in separate analyses with measuresof achievement in each subject. When math grades and trackswere used as indicators of achievement, the model fit thedatawell, x

2 ( n , i V - 153)= \9.65,p> .05, goodness-of-fitindex ~ .97. Again, expectancies were more strongly relatedto achievement than were values (parameter estimates = .61and .30, respectively), but both estimates were significant, is= 4.18, p < .001, and 2.38, p < .05, respectively. Again,expectancies and values were moderately related (estimate =,46U = 4.70,/>< .001.

An exactly parallel analysis could not be done for Englishbecause there was little variance in track placements {i.e.,there were only two tracks and few students were in the lowertrack). Therefore, report card grades were used as the soleindicator of achievement in English. To achieve identificationof the model, we estimated the reliability of grades from thecorrelation between English and math grades. This model alsofit the data well, x

2 (7, N = 153) = 7.39, p > .05, goodness-of-fit index — .98. Once again, expectancies were morestrongly related to achievement than were values (estimates= .75 and .40, respectively) but both estimates were signifi-cant, ts = 5.31, p < .001, and 3.02, p < .01, respectively.

INVOLVEMENT

SCHOOL VALUE .31

2.48GRADES

. 4 6 | ACHIEVEMENT

SCHOLASTIC

COMPETENCE

ABILITY:

SUCCESS

ABILITY:

FAILURE* > *

.60

.42

t

( ACHIEVEMENT

I EXPECTANCY

.57SCHOOL TRACK

Figure 2, Final structural equation model of the relations among constructs for expectancies, values,and achievement. (The numbers linking the measures to the latent constructs and linking the latentconstructs to each other are standardized maximum-likelihood [LISREL] parameter estimates.)

Page 7: Expectancies, values, and achievement in junior high school

EXPECTANCIES, VALUES, AND ACHIEVEMENT 325

Expectancies again were positively related (.46) to values, t =4.65(p < .001. Thus the relations of expectancies and valuesto achievement and to each other were comparable whenachievement was measured for different academic subjects.

Discussion

The results support both of our major hypotheses. Theyshow, first, that students' achievement is more strongly relatedto their expectancies of success than to the value they attachto success. Second, students' expectancies and values arepositively related. In addition, the correlations among themotivational measures were largely consistent with our origi-nal assumptions that these measures are indicators of expec-tancies or of values. Nevertheless, a detailed examination ofthe findings suggests that these hypotheses and assumptionsrequire some qualification or careful interpretation.

Most of the motivation-related measures were significantlycorrelated. A factor analysis and later covariance structureanalyses supported the assumption that students' perceptionsof their scholastic competence and their ability attributionsfor success and failure were indicators of their achievementexpectancies. The same analyses suggested that the othermotivational measures related not to a single construct forachievement values, but to two distinct constructs. The indi-cators of one construct included success attributions to effort.interest, and external causes; the indicators of the other con-struct included the corresponding failure attributions. Marshet al. (1984) obtained comparable results for attribution meas-ures with a similar response format. In their study and inours, students rated several causes of the same academicoutcome (e.g., success). This response format may haveheightened the relations among different types of successattributions and reduced the relations between success andfailure attributions. Nevertheless, these effects did not seri-ously blur the theoretically significant distinction betweenachievement expectancies and achievement values.

In the first covariance structure analysis of the relationsamong the constructs for expectancies, values, and achieve-ment, expectancies were strongly related to achievement. Thevalue construct defined by students' involvement, schoolvalue, and attributions of success to effort, interest, andexternal causes was less strongly but still significantly relatedto achievement. The value construct defined by involvement,school value, and the corresponding failure attributions wasnot significantly related to achievement.

A second analysis showed nonsignificant relations ofachievement to constructs based entirely on effort, interest,and external attributions. This result was surprising becausemany successful attribution retraining programs are based onthe premise that students' achievement can be enhanced bychanging their attribution patterns. In particular, achievementimproves when students attribute their failures to a lack ofeffort rather than a lack of ability (Forsterling, 1985). Thischange enhances motivation because it makes students feelthey can succeed if they try harder.

Because attribution retraining programs attempt to increaseeffort attributions and decrease ability attributions for failure,they imply that ability and effort attributions are negatively

related. In natural settings, however, these types of failureattributions are positively correlated. Moreover, both typescorrelate negatively with achievement (see Table 2 and Marshet al., 1984). In other words, low-achieving students often saythat they have little ability and that they don't work very hardat school. Moreover, actual achievement is more stronglyrelated to ability attributions than to effort attributions. Thus,when both relations are assessed simultaneously in a structuralequation model, constructs based on ability attributions relatesignificantly to achievement and constructs based on effortattributions do not.

In our final structural equation model, the involvementand school-value measures were indicators of a single con-struct for achievement values. The analysis of this modelshowed that expectancies related more strongly to achieve-ment than did values, although both relations were significant.Eccles et al. (1984) reported comparable results. These find-ings could be taken as support for theories that emphasize theeffects of expectancies on achievement (e,.g, Dweck & Leggett,1988). Yet because our study was correlational, conclusionsabout the causal influences of expectancies and values onachievement are unjustified. Students' expectancies might bemore influenced by their past achievements than their valuesare (cf. Bohrnstedt & Felson, 1983). If so, our results couldbe viewed as evidence that values have an influence on futureachievements that is equal to or greater than that of expectan-cies. These hypotheses could be evaluated directly with alongitudinal design in which measures of expectancies, values,and achievement were obtained at multiple times.

The evidence for our second hypothesis, that expectanciesand values are positively related, replicates previous researchwith adolescents (Eccles, 1983) and college students (Feather,1988; Platt, 1988). Thus, students who are more confidentabout their academic success are more interested in theirschoolwork and more convinced of the usefulness of theireducation. These findings imply that expectancies and valuesinfluence each other, although the direction of influence isdifficult to judge. Students who attach little value to academicsuccess may exert little effort on their schoolwork and thenlower their expectancy for success (cf. Covington & Omelich,1979). Conversely, students with low expectancies for successmay devalue academic activities (Dweck & Leggett, 1988).Again, these hypotheses could be examined in a longitudinalstudy.

The importance of the distinction between expectanciesand values was further demonstrated by the sex differences instudents' responses. As in previous research (e.g., Eccles et al.,1984; Wigfield, 1988), boys and girls did not differ signifi-cantly in their perceived scholastic competence or their attri-butions about their ability. Girls reported greater involvementin school and valued school more than did boys. In addition,girls more often attributed their successes to effort and lessoften attributed their failures to a lack of interest than boysdid. These findings suggest that adolescent boys and girlsdiffer less in their achievement expectancies than in theirvalues. The sex differences were not entirely consistent, how-ever. For example, girls and boys did not differ in their effortattributions for failure. Previous researchers also found incon-sistent patterns of sex differences (e.g., Ryckman & Peckham,

Page 8: Expectancies, values, and achievement in junior high school

326 THOMAS J. BERNDT AND KRISTELLE E. MILLER

1987; Wigfield, 1988). Therefore, firm conclusions about theexpectancies and values of adolescent boys and girls cannotbe drawn.

Even the evidence supporting our major hypotheses mustbe interpreted cautiously because of the limitations of thestudy. The sample was relatively small, especially for covari-ance structure analyses. We attempted in the same study toidentify measures of our key constructs and to assess therelations among these constructs. The measures of students'expectancies were indirect, focusing on students' perceptionsof their ability rather than their predictions about futureperformance. Although measures of perceived ability and ofexpectancies are strongly related (with correlations above .90in Eccles, 1983), research with direct measures of expectancieswould be desirable. Most important, we assessed only oneeducational outcome, academic achievement. In the future,other outcome measures, such as course plans and educationalaspirations, should be investigated.

Finally, the theoretical and practical implication of ourresults should be mentioned briefly. Current theories ofachievement motivation emphasize variables related to stu-dents' expectancies, such as their confidence in their abilityor their ability attributions (see Ames, 1987; Dweck, 1986).Our findings suggest that theorists need to give more emphasisto students' values. On a practical level, interventions toincrease students' motivation may be most effective whentargeted at the component of motivation that specific studentslack. Students who underestimate their chances for academicsuccess should be encouraged to raise their expectancies.Students who have little interest in school should be led tosee its value. Nevertheless, because of the positive relationbetween expectancies and values, educational interventionsthat focus on both components of achievement motivationmay be most helpful. This dual focus would also be desirablein future research on academic achievement motivation.

References

Ames, C. (1987). The enhancement of student motivation. In M. L.Maehr & D. A. Kleiber (Eds.), Advances in motivation and achieve-ment: Vol. 5. Enhancing motivation (pp. 123-148). Greenwich,CT: JAl Press.

Ames, C, & Archer, J. (1988). Achievement goals in the classroom:Students' learning strategies and motivation processes. Journal ofEducational Psychology, 80, 260-267.

Atkinson, J. W., & Feather, N. T. (Eds.). (1966). A theory of achieve-ment motivation. New York: Wiley.

Bohrnstedt, G. W., & Felson, R. B. (1983). Explaining the relationsamong children's actual and perceived performances and self-esteem: A comparison of several causal models. Journal of Person-ality and Social Psychology, 45, 43-56.

Covington, M. V., & Omelich, C. L. (1979). Are causal attributionscausal? A path analysis of the cognitive model of achievement

motivation. Journal of Personality and Social Psychology, 37,1487-1504.

Dweck, C. S. (1986). Motivational processes affecting learning. Amer-ican Psychologist, 41, 1040-1048.

Dweck, C. S., & Elliott, E. S. (1983). Achievement motivation. In E.M. Hetherington (Ed.), Handbook of child psychology, Vol. 4 (pp.643-691). New York: Wiley.

Dweck, C. S., & Leggett, E. L. (1988). A social-cognitive approach tomotivation and personality. Psychology Review, 95, 256-273.

Eccles, J. P. (1983) Expectancies, values, and academic behaviors. InJ. T. Spence (Ed.), Achievement and achievement motives (pp. 75-146). San Francisco: W. H. Freeman.

Eccles, J. P., Adler, T., & Meece, J. L. (1984). Sex differences inachievement: A test of alternate theories. Journal of Personalityand Social Psychology, 46, 26-43.

Elig, T. W., & Frieze, I. H. (1979). Measuring causal attributions forsuccess and failure. Journal of Personality and Social Psychology,37, 621-634.

Feather, N. T. (1988). Values, valences, and course enrollment:Testing the role of personal values within an expectancy-valenceframework. Journal of Educational Psychology, 80, 381-39!.

Forsterling, F. (1985). Attributional retraining: A review. Psycholog-ical Bulletin, 98, 495-512.

Harter, S. (1985). Manual for the self-perception profile for children.Denver: University of Denver.

Joreskog, K. G., & Sorbom, D. (1984). LISREL: Analysis of linearstructural relationships by the method of maximum likelihood.Chicago: SPSS, Inc.

Marsh, H. W., Cairns, L., Relich, J., & Debus, R. L. (1984). Therelationship between dimensions of self-attribution and dimensionsofself-concept. Journal of Educational Psychology, 76, 3-32.

Moos, R. H., & Trickett, E. J. (1974). Classroom environment scalemanual. Palo Alto, CA: Consulting Psychologists Press.

Nicholls, J. G. (1984). Achievement motivation: Conceptions ofability, subjective experience, task choice, and performance. Psy-chological Review, 91, 328-346.

Platt, C. W. (1988). Effects of causal attributions for success on first-term college performance: A covariance structure model. Journalof Educational Psychology, 80, 569-578.

Ryckman, D. B., & Peckham, P. D. (1987). Gender differences inattributions for success and failure. Journal of Early Adolescence,7, 47-63.

Tanaka, J. S. (1987). "How big is big enough?": Sample size andgoodness of fit in structural equation models with latent variables.Child Development, 58, 134-146.

Weiner, B. (1985). An attributional theory of achievement motivationand emotion. Psychological Review, 92, 548-573.

Wigfield, A. (1988). Children's attributions for success and failure:Effects of age and attentional focus. Journal of Educational Psy-chology, 80,1'6-81.

Winer, B. J. (1971). Statistical principles in experimental design. NewYork: McGraw-Hill.

Received April 6, 1989Revision received October 31, 1989

Accepted November 21, 1989