Title: RESPONSE FORMAT IN WRITING SELF-EFFICACY ASSESSMENT: GREATER DISCRIMINATION INCREASES PREDICTION ,  By: Pajares, Frank, Hartley, James, Valiante, Giovanni, Measurement & Evaluation in Counseling & Development, 07481756, Jan2001, Vol. 33, Issue 4

RESPONSE FORMAT IN WRITING SELF-EFFICACY ASSESSMENT:
GREATER DISCRIMINATION INCREASES PREDICTION


Contents
METHOD
Participants
Procedure
RESULTS
DISCUSSION
REFERENCES

Writing self-efficacy scales differing in response format were compared. A scale with a 0-100 format was psychometrically stronger than a scale with a traditional Likert format.

According to Bandura's (1986) social-cognitive theory, students' beliefs about their academic capabilities, or self-efficacy beliefs, are good predictors of their academic achievement and of their subsequent career choices and decisions (Betz & Hackett, 1997; Fouad & Smith, 1996; Hackett, 1995; Lapan, Adams, Turner, & Hinkelman, 2000; Lapan, Boggs, & Morrill, 1989; Lent & Hackett, 1987). As a result of this predictive utility, the construct of self-efficacy has become increasingly prominent in the counseling literature and its measurement a primary concern (Betz, 1992; Fouad, Smith, & Enochs, 1997; Hansen, 1997; Lent & Hackett, 1987; Lent & Maddux, 1997).

In the area of writing, researchers have confirmed that students' confidence in their writing skills is related both to writing competence and to academic motivation variables such as writing self-concept, writing apprehension, achievement goals, and the perceived value of writing, as well as their writing competence (McCarthy, Meier, & Rinderer, 1985; Meier, McCarthy, & Schmeck, 1984; Pajares & Johnson, 1994; Pajares, Miller, & Johnson, 1999; Pajares & Valiante, 1997; Shell, Colvin, & Bruning, 1995; Shell, Murphy, & Bruning, 1989). Following guidelines outlined by Bandura (1997), social-cognitive researchers have typically assessed writing self-efficacy by asking students to provide judgments of their capability to successfully perform various writing skills appropriate to their academic level. Students provide these judgments on a rating scale that ranges from 0 to 100 (e.g., Pajares et al., 1999; Pajares & Valiante, 1997; Shell et al., 1995; Shell et al., 1989).

Researchers have questioned whether a more traditional Likert-type measurement can be used as an alternative to the typical 0-100 format (Maurer & Pierce, 1998). In part, this question is based on researchers' concerns that children are unable to make the discriminating judgments required of a 100-point scale. Self-efficacy researchers point out that thinking in 0-100 terms is congruent with the manner in which students are typically graded in school. Consequently, scales with response formats that range from 0 to 100 are appropriate and should result in greater discrimination than should scales with narrower response options. Moreover, as Bandura (1997) observed, "scales that use only a few steps should be avoided because they are less sensitive and less reliable" (p. 44; see also Streiner & Norman, 1989).

The purpose of this study was to investigate whether the 0-100 format of assessing writing self-efficacy beliefs differed in empirical qualities from a traditional 1-6-point Likert-type format. We sought to discover whether the two measures differed in (a) their factor structure, (b) their internal consistency, (c) their relation to motivation constructs, and (d) their prediction of achievement indexes.

METHOD

Participants

Participants consisted of 497 middle school students in Grades 6-8 (ages 11-14) enrolled in language arts classes in one middle school located in the northeastern United States. The socioeconomic status of the school and of the area that the school served was largely middle class, and students were primarily White. Students at the school were enrolled separately in reading and in writing (called language arts) classes. Regular education, gifted, special education, and English-as-a-second-language students took language arts classes separately; our sample consisted only of students in the regular education classes. Each student completed a 10-page instrument tapping writing attitudes and motivation variables. These included writing self-concept, writing apprehension, value of writing, self-efficacy for self-regulation, and achievement goals in writing (task, performance-approach, performance-avoid). Instruments were group administered in individual language arts classes during one period. Students completed the instrument one item at a time as that item was read aloud by the administrator. Grades in language arts class, as well as teachers' ratings of students' writing aptitude, were obtained from the students' teaching team. Procedures were similar to those used by writing self-efficacy researchers (e.g., Pajares et al., 1999; Shell et al., 1995; Shell et al., 1989).

Procedure

We selected academic motivation scales that have been extensively used in various studies. Achievement goals were assessed using the scale provided by Middleton and Midgley (1997), which was derived from the Patterns of Adaptive Learning Survey. The Self-Efficacy for Self-Regulated Learning Scale was adapted from Bandura's (1995) Children's Multidimensional Self-Efficacy Scales, which assess students' judgments of their capability to learn in various academic courses and to use various self-regulated learning strategies (see Zimmerman & Martinez-Pons, 1990). Writing self-concept was assessed with six items adapted from Marsh's (1990) Academic Self-Description Questionnaire. Writing apprehension was measured with a scale adapted from Daly and Miller's (1975) Writing Apprehension Test and subsequently used by Pajares et al. (1999). Finally, value of writing consisted of items assessing perceived importance, interest, and enjoyment of writing (see Eccles, 1983; Meece, Wigfield, & Eccles, 1990; Seegers & Boekaerts, 1996). All of the motivation variables were assessed in a manner consistent with their typical use, that is, with a traditional Likert format. In this study, and to be consistent with the manner in which the Likert version of one of the self-efficacy scales was assessed, we provided students with a response option consisting of 6 points on the Likert scale.

Included in the complete instrument were two versions of a writing self-efficacy scale that differed only in the manner in which students could provide their responses. Each scale consisted of the same 10 items tapping students' judgments of their confidence that they possessed composition, grammar, usage, and mechanical skills appropriate to their academic level (see Table 1). The skills listed were identified by the students' language arts teachers as the writing skills appropriate for middle school students in the school in which the investigation took place. In Version 1 of this scale, students were asked how sure they were that they could perform the writing skills on a scale ranging from 0 (no confidence at all) to 100 (completely confident). In Version 2 of this scale, students were asked to respond on a traditional Likert format consisting of six options ranging from 1 (no confidence at all) to 6 (completely confident). Half of the students completed Version 1 first (on page 2 of the instrument) and Version 2 second (on page 10); the other half completed Version 2 first (on page 2) and Version 1 second (on page 10). Students completed each scale as items were read aloud by the administrator, and they could not return to a page they had already completed. Hence, when they got to the end of the instrument to complete the second self-efficacy scale, they could not review how they had responded earlier. The instrument took approximately 40 minutes to complete.

RESULTS

We first conducted exploratory factor analyses of the two scales. We used the maximum-likelihood method of extraction (Joreskog & Lawley, 1968) because this method is believed to produce the best parameter estimates (Pedhazur, 1982). We used a variety of criteria to determine the number of common factors to retain and analyze, including Cattell's (1966) scree test, the percentage of common variance explained by each factor using the weighted reduced correlation matrix, and the interpretability of the rotated factors. Because we expected any factors that emerged from the analyses to be intercorrelated, we chose the oblimin method of oblique rotation. All analyses were conducted using the SAS system's FACTOR procedure (SAS Institute, Inc., 1989).

The factor analyses revealed that both the 0-100 scale and the Likert scale were composed of two factors. Table 1 shows the factor structure coefficients from the rotated pattern matrix and the percentage of variance explained for the two-factor solutions for each of the two scales. Factor structure coefficients from the pattern matrix demonstrate the relationship between an item and a factor when all other items are held constant. Factor structure coefficients of .40 or higher were considered strong enough to demonstrate that the item indicated the common factor. Factor 2 comprised the first five items on each scale, and Factor 1 comprised the second five items. A quick look at the items shown in Table 1 reveals that Factor 1 tapped advanced composition skills (e.g., constructing paragraphs and relating main ideas), whereas Factor 2 tapped basic grammatical and usage skills (e.g., spelling and punctuation). Factor structure coefficients were similar for each scale, as was the proportion of variance accounted for by each factor. Interfactor correlations were .65 for the 0-100 scale and .62 for the Likert scale. Cronbach's alpha coefficients were also similar for each factor.

Correlations between each of the factors and the full scales on the one hand and other motivation and achievement indexes assessed on the other hand revealed that the 0-100 scale had significantly stronger correlations than did the Likert scale with the academic performance variables. The Likert scale tended to be better related to self-efficacy for self-regulation and task-goal orientation. There were no significant differences in correlations between the factors and writing self-concept, writing anxiety, value of writing, or performance-goal orientations. We point out at this juncture that all of the attitude scales were measured with the same 6-point Likert scale format as the Likert self-efficacy scale; hence, method variance can easily increase the correlations between scales using the same form of measurement. We used Williams's (1959) T2 statistic to determine if the correlations were significantly different.

For the multiple regression analyses predicting the achievement indexes, we joined the two factors from each scale to compose one scale reflecting the Likert assessment and another scale reflecting the 0-100 assessment so as to create the full writing self-efficacy scales that would typically be used in self-efficacy studies. Scores for these two scales were computed by adding students' responses to the 10 items on each scale. These analyses were supplemented by a regression commonality analysis to determine the proportion of the explained variance of the dependent variable associated uniquely with each independent variable and with the common effects of each (Rowell, 1996) and by obtaining regression structure coefficients (Thompson & Borrello, 1985). Because the efficacy scales were highly correlated (.81), interpreting regression structure coefficients is especially critical. Structure coefficients are not suppressed or inflated by collinearity.

With both scales as independent variables in analyses predicting grade point average (GPA) in language arts and teachers' ratings of students' writing competence, results revealed that the 0-100 scale predicted both outcomes whereas the Likert scale assessment did not. Moreover, 37% of the explained variance in GPA and 28% of the explained variance in teachers' ratings were associated uniquely with the 0-100 scale; the proportion of explained variance associated uniquely with the Likert scale was negligible for both dependent variables (see Table 2). The 0-100 scale provided structure coefficients of .99 in each model; the structure coefficients for the Likert scale were .79 for the model with GPA and .85 for the model with teachers' ratings.

To ensure that simultaneous entry did not mask actual differences in prediction, we conducted forced-entry hierarchical regression in which we entered the Likert scale in the first step and the 0-100 scale in the second step. For the prediction of GPA, the Likert scale accounted for 11% of the variance; the 0-100 scale entry significantly increased the incremental multiple correlation squared (R2) value to .18 and rendered nonsignificant the predictive value of the Likert scale. For the prediction of teachers' ratings of students' writing capability, the Likert scale accounted for 15% of the variance; the 0-100 scale entry increased the incremental value to .21 and again rendered the influence of the Likert scale nonsignificant. Conversely, when the 0-100 scale was entered first, the subsequent entry of the Likert scale did not account for a significant increase in the prediction of either achievement index.

Finally, we conducted two multiple regression analyses of the sort that self-efficacy researchers would typically conduct in studies of writing attitudes and achievement. Our aim was to predict GPA in language arts. Our independent variables were writing self-concept, writing apprehension, value of writing, self-efficacy for self-regulation, and gender. We did not include the achievement goals in this analysis because achievement goals are not typically presumed to predict writing achievement. In the first multiple regression model, we included the writing self-efficacy scale using the 0-100 assessment; in the second model, we included the self-efficacy instrument using the Likert scale. The model using the 0-100 scale accounted for 23% of the variance in GPA; the model using the Likert scale self-efficacy instrument accounted for 19%. In the model with the Likert efficacy scale, self-efficacy was nonsignificant, although the structure coefficients revealed that its effect on GPA was influenced by its collinearity with the predictor variables. In the model with the 0-100 scale, self-efficacy made an independent contribution to the prediction of GPA (beta = .279, t = 3.7, p < .0001; see Table 3). The results confirm that response scale properties can influence the outcome of an investigation of academic motivation.

DISCUSSION

Lent, Brown, and Hackett (1994,1996) provided a model aimed at explaining how students' career beliefs develop and how these beliefs influence their career choices. In this model, self-efficacy beliefs are viewed as being central to the process in which students engage as they select their academic paths and subsequent careers. Lent et al. argued that self-efficacy beliefs play a critical role in the development of students' career interests and intentions. For this reason, they urged teachers and counselors to be cognizant of their students' self-efficacy beliefs and to ensure that attention to these beliefs becomes foundational to their counseling practices and to intervention strategies carried out in classrooms and schools (also see Betz & Hackett, 1997; Young & Chen, 1999). For example, counselors can use results of such assessments both to evaluate the effectiveness of academic programs and intervention strategies and to monitor students' progress through such programs (Fouad et al., 1997). A critical concern in this regard is that academic self-efficacy beliefs be assessed with instruments that possess sound psychometric properties.

Some researchers have suggested that teachers and counselors should pay as much attention to students' perceptions of competence as to actual competence because the perceptions may more accurately predict students' motivation and future academic choices (e.g., Betz & Hackett, 1997). Assessing students' self-efficacy beliefs can provide counselors with important insights about students' academic motivation, behavior, and future choices. For example, unrealistically low self-efficacy perceptions--not lack of capability or skill-can be responsible for maladaptive academic behaviors, avoidance of courses and careers, and diminishing school interest and achievement (Bandura, 1997; Betz & Hackett, 1997; Hackett, 1995; Hackett & Betz, 1989). Students who lack confidence in skills they possess are less likely to engage in tasks in which those skills are required, and they will more quickly give up in the face of difficulty. In such cases, in addition to continued skill improvement, schools must work to identify their students' inaccurate self-beliefs and design and implement interventions to challenge them. To accomplish this, counselors must have confidence that scores obtained from their students' responses to self-efficacy instruments provide a reliable assessment of their academic confidence.

Our results led us to the conclusion that Bandura's (1997) guidelines regarding self-efficacy assessment are empirically well-grounded. Results of the factor and reliability analyses showed that a writing self-efficacy scale with a 0-100 response format was psychometrically stronger than a traditional Likert format scale. Moreover, children at the middle school level can indeed make a discriminating judgment using a 0-100 scale. We should also note that the mean for the 0-100 scale was 75.07 (or 4.5 when converted to a 6-point Likert scale) and the mean for the Likert scale was 4.4 (or 73.17 when converted to a 0-100 scale). These results revealed that neither scale artificially inflated or deflated confidence judgments. The fine-grained discrimination of the 0-100 scale provided an assessment of self-efficacy that was not only more strongly related to the achievement indexes with which it was compared but also predictive of achievement in a regression model, whereas the less discriminating scale using the Likert format was not.

It bears keeping in mind, of course, that reliability is a characteristic of scores from a specific sample of individuals administered an instrument at a particular time in particular situations. We urge researchers to continue to refine writing self-efficacy scales, such as the one we used in this study, with the aim of further enhancing their psychometric properties and their applicability to other student populations and settings. We also urge researchers to examine whether scores on this writing self-efficacy scale are related to behavioral indexes such as course and career information selection, career intentions, and eventual selection of careers and life paths. In the area of mathematics and science, researchers have already well established the connection between mathematics self-efficacy and these behavioral indexes (see Hackett, 1995; Lent et al., 1994, 1996; Lent & Maddux, 1997).

We concur with Bandura's (1997) warning that "including too few steps loses differentiating information because people who use the same response category would differ if intermediate steps were included" (p. 44). We recommend that researchers including measures of self-efficacy in studies of academic motivation do so with attention to Bandura's assessment guidelines. More important, we urge teachers and counselors wishing to assess their students' self-efficacy beliefs about writing (indeed, about any academic subject) to do so using scales that are likely to provide the greatest predictive utility. In the case of writing self-efficacy as assessed by the scales we tested, neither scale was more difficult or longer than the other. Consequently, using a format that adds predictive utility is especially warranted.

TABLE 1

Rotated Factor Pattern Coefficients and Factor Structure Coefficients of Likert and 0-100 Self-Efficacy Scales With Alpha Reliability Coefficients

Legend for Chart:

A - Item
B - Factor 1, 0-100
C - Factor 1, Likert
D - Factor 2, 0-100
E - Factor 2, Likert

A                                           B                  C
                                            D                  E

1. Correctly spell all
   words in a one-page
   story or composition.             .01(.51)           .01(.45)
                                     .77(.78)           .71(.72)

2. Correctly punctuate
   a one-page story or
   composition.                     -.01(.54)          -.06(.48)
                                     .85(.84)           .87(.83)

3. Correctly use all
   parts of speech in
   a written composition.            .18(.60)           .16(.55)
                                     .65(.77)           .64(.74)

4. Write simple sentences
   with good grammar.                .30(.58)           .25(.55)
                                     .43(.63)           .49(.64)

5. Correctly use
   singulars and plurals,
   verb tenses, prefixes
   and suffixes.                     .24(.58)           .18(.53)
                                     .52(.68)           .56(.67)

6. Write a strong
   paragraph that has
   a good topic sentence
   or main idea.                     .82(.82)           .83(.81)
                                     .00(.53)          -.04(.47)
7. Structure paragraphs
   to support ideas in
   the topic sentences.              .73(.81)           .83(.82)
                                     .12(.59)          -.02(.50)

8. End paragraphs with
   proper conclusions.               .62(.71)           .66(.73)
                                     .14(.54)           .11(.52)

9. Write a well-organized
   and sequenced paper
   with good introduction,
   body, and conclusion.             .84(.86)           .59(.72)
                                     .03(.57)           .21(.58)

10. Get ideas across in
    a clear manner by
    staying focused
    without getting off
    the topic.                       .74(.80)           .58(.70)
                                     .10(.58)           .20(.56)

Proportion of variance                    .88                .86
                                          .12                .14

Reliability coefficients                  .86                .85
                                          .90                .87

Note. Factor structure coefficients are presented in parentheses.

TABLE 2

Multiple Regression Results Predicting GPA in Language Arts and Teacher Ratings of Students' Writing Competence and Commonality Analysis Summary

Legend for Chart:

A - Result
B - GPA, 0-100
C - GPA, Likert Scale
D - Teacher Ratings, 0-100
E - Teacher Ratings, Likert Scale

A                             B          C          D          E

Multiple regression results

b                         .0026     -.0026      .0024      .0052
Standard error            .0004      .0074      .0004      .0071
t                          6.44       -.35       6.17        .73
p                         .0001      .7254      .0001      .4641
beta                       .449      -.024       .422       .050
S                          .999       .793       .998       .848
R2                        .1843                 .2149

Commonality analysis

Unique to 0-100 scale     .0685         --      .0605         --
Unique to Likert scale       --      .0002         --      .0009
Common to both            .1156      .1156      .1535      .1535
Total                     .1841      .1158      .2140      .1544

Note. GPA = grade point average; S = the structure coefficient. For GPA, F(2,494) = 55.79, p < .0001; for teacher ratings, F(2,494) = 67.60, p < .0001.

TABLE 3

Multiple Regression Results Predicting Writing GPA Using Either 0-100 or Likert Self-Efficacy Scales

Legend for Chart:

A - Variable
B - Using 0-100 Self-Efficacy Scale, t
C - Using 0-100 Self-Efficacy Scale, beta
D - Using 0-100 Self-Efficacy Scale, p
E - Using 0-100 Self-Efficacy Scale, S
F - Using Likert Self-Efficacy Scale, t
G - Using Likert Self-Efficacy Scale, beta
H - Using Likert Self-Efficacy Scale, p
I - Using Likert Self-Efficacy Scale, S

A                                   B        C        D        E
                                    F        G        H        I

Writing self-efficacy            5.04     .279    .0001     .896
                                 1.46     .092    .1439     .775

Writing self-concept             2.77     .188    .0059     .762
                                 3.74     .267    .0002     .831

Writing apprehension            -0.93    -.044    .3526    -.511
                                -1.38    -.066    .1677    -.557

Value of writing                -3.07    -.168    .0023     .367
                                -3.02    -.169    .0027     .401

Self-efficacy for
self-regulation                  2.60     .130    .0096     .701
                                 3.20     .176    .0014     .765

Gender                          -2.34    -.095    .0199    -.356
                                -2.61    -.109    .0095    -.388

R2 .2294 .1929

Note. See Table 2 Note. Gender was coded -1 for girls and 1 for boys.

REFERENCES

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall.

Bandura, A. (1995). Guide for constructing self-efficacy scales. (Available from Frank Pajares, Division of Educational Studies, Emory University, Atlanta, GA 36322)

Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.

Betz, N. E. (1992). Career assessment: A review of critical issues. In S. D. Brown & R. W. Lent (Eds.), Handbook of counseling psychology (2nd ed., pp. 453-484). New York: Wiley-Interscience.

Betz, N. E., & Hackett, G. (1997). Applications of self-efficacy theory to the career assessment of women. Journal of Career Assessment, 5, 383-482.

Cattell, R. B. (1966). The scree test for the number of factors. Multivariate Behavioral Research, 1, 245-276.

Eccles, J. (1983). Expectancies, values, and academic behaviors. In J. T. Spence (Ed.), Achievement and achievement motives: Psychological and sociological approaches (pp. 75-146). San Francisco: Freeman.

Daly, J. A., & Miller, M. D. (1975). The empirical development of an instrument to measure writing apprehension. Research in the Teaching of English, 9, 272-289.

Fouad, N. A., & Smith, P. L. (1996). A test of a social cognitive model for middle school students: Math and science. Journal of Counseling Psychology, 43, 338-346.

Fouad, N. A., Smith, P. L., & Enochs, L. (1997). Reliability and validity evidence for the Middle School Self-Efficacy Scale. Measurement and Evaluation in Counseling and Development, 30, 17-31.

Hackett, G. (1995). Self-efficacy in career choice and development. In A. Bandura (Ed.), Self-efficacy in changing societies (pp. 232-258). New York: Cambridge University Press.

Hackett, G., & Betz, N. E. (1989). An exploration of the mathematics self-efficacy/mathematics performance correspondence. Journal for Research in Mathematics Education, 20, 261-273.

Hansen, J. C. (1997). The measurement of self-efficacy. Measurement and Evaluation in Counseling and Development, 30, 2-3.

Joreskog, K. G., & Lawley, D. N. (1968). New methods in maximum likelihood factor analysis. British Journal of Mathematical and Statistical Psychology, 21, 85-96.

Lapan, R. T., Adams, A., Turner, S., & Hinkelman, J. M. (2000). Seventh graders' vocational interest and efficacy expectation patterns. Journal of Career Development, 26, 215-229.

Lapan, R. T., Boggs, K. R., & Morrill, W. H. (1989). Self-efficacy as a mediator of investigative and realistic general occupational themes on the Strong-Campbell Interest Inventory. Journal of Counseling Psychology, 36, 176-182.

Lent, R. W., Brown, S. B., & Hackett, G. (1994). Toward a unifying social cognitive theory of career and academic interest, choice, and performance. Journal of Vocational Behavior, 45, 79-122.

Lent, R. W., Brown, S. B., & Hackett, G. (1996). Career development from a social cognitive perspective. In D. Brown, L. Brooks, & Associates (Eds.), Career choice and development (3rd ed., pp. 373-422). San Francisco: Jossey-Bass.

Lent, R. W., & Hackett, G. (1987). Career self-efficacy: Empirical status and future directions. Journal of Vocational Behavior, 30, 347-382.

Lent, R. W., & Maddux, J. E. (1997). Self-efficacy: Building a sociocognitive bridge between social and counseling psychology. The Counseling Psychologist, 25, 240-255.

Marsh, H. W. (1990). The structure of academic self-concept: The Marsh-Shavelson model. Journal of Educational Psychology, 82, 623-636.

Maurer, T. J., & Pierce, H. R. (1998). A comparison of Likert scale and traditional measures of self-efficacy. Journal of Applied Psychology, 83, 324-329.

McCarthy, P., Meier, S., & Rinderer, R. (1985). Self-efficacy and writing. College Composition and Communication, 36, 465-471.

Meece, J. L., Wigfield, A., & Eccles, J. S. (1990). Predictors of math anxiety and its influence on young adolescents' course enrollment and performance in mathematics. Journal of Educational Psychology, 82, 60-70.

Meier, S., McCarthy, P. R., & Schmeck, R. R. (1984). Validity of self-efficacy as a predictor of writing performance. Cognitive Therapy and Research, 8, 107-120.

Middleton, M. J., & Midgley, C. (1997). Avoiding the demonstration of lack of ability: An underexplored aspect of goal theory. Journal of Educational Psychology, 89, 710-718.

Pajares, F., & Johnson, M. J. (1994). Confidence and competence in writing: The role of writing self-efficacy, outcome expectancy, and apprehension. Research in the Teaching of English, 28, 313-331.

Pajares, F., Miller, M. D., & Johnson, M. J. (1999). Gender differences in writing self-beliefs of elementary school students. Journal of Educational Psychology, 91, 50-61.

Pajares, F., & Valiante, G. (1997). The predictive and mediational role of the writing self-efficacy beliefs of upper elementary students. Journal of Educational Research, 90, 353-360.

Pedhazur, E. J. (1982). Multiple regression in behavioral research: Explanation and prediction (2nd ed.). New York: Harcourt Brace.

Rowell, R. K. (1996). Partitioning predicted variance into constituent parts: How to conduct regression commonality analysis. In B. Thompson (Ed.), Advances in social science methodology (Vol. 4, pp. 33-43). Greenwich, CT: JAI Press.

SAS Institute, Inc. (1989). SAS/STAT users' guide (Version 6, 4th ed., Vol. 1). Cary, NC: Author.

Seegers, G., & Boekaerts, M. (1996). Gender-related differences in self-referenced cognitions in relation to mathematics. Journal for Research in Mathematics Education, 27, 215-240.

Shell, D. F., Colvin, C., & Bruning, R. H. (1995). Self-efficacy, attributions, and outcome expectancy mechanisms in reading and writing achievement: Grade-level and achievement-level differences. Journal of Educational Psychology, 87, 386-398.

Shell, D. F., Murphy, C. C., & Bruning, R. H. (1989). Self-efficacy and outcome expectancy mechanisms in reading and writing achievement. Journal of Educational Psychology, 81, 91-100.

Streiner, D. L., & Norman, G. R. (1989). Health measurement scales: A practical guide to their development and use. Oxford, England: Oxford University Press.

Thompson, B., & Borrello, G. M. (1985). The importance of structure coefficients in regression research. Educational and Psychological Measurement, 45, 203-209.

Williams, E. J. (1959). The comparison of regression variables. Journal of the Royal Statistical Society, Series B, 21, 396-399.

Young, R. A., & Chen, C. P. (1999). Annual review: Practice and research in career counseling and development--1998. The Career Development Quarterly, 48, 98-141.

Zimmerman, B. J., & Martinez-Pons, M. (1990). Student differences in self-regulated learning: Relating grade, sex, and giftedness to self-efficacy and strategy use. Journal of Educational Psychology, 82, 51-59.