Bandura (1986, in press) has cautioned researchers attempting to predict academic outcomes from students' self-efficacy beliefs that, to increase accuracy of prediction, they would be well advised to follow theoretical guidelines regarding specificity of self-efficacy assessment and correspondence with criterial tasks. This caution has often gone unheeded in educational research, resulting in self-efficacy assessments that reflect global or generalized self-perceptions of competence and that bear slight resemblance to the criterial task with which they are compared. Often, no criterial task is identified, as researchers aim to discover simply the nature of the interplay between motivational variables in the absence of performance attainments. In still other studies, judgments of "confidence" that bear passing resemblance to self-efficacy judgments are used instead of more appropriate particularized measures. The result is often confounded relationships and ambiguous findings that obfuscate the potential contribution of self-efficacy beliefs to the understanding of academic performances.
I hope to accomplish three objectives during this presentation. First, I want to briefly highlight some of the guidelines provided by Bandura (in press) regarding self-efficacy assessment. Second, I want to clarify the issue of specificity versus generality of measurement. And last, I want to report on preliminary results of a study demonstrating that, when efficacy beliefs assessed do not reflect with specificity the academic tasks with which they are compared, their predictive value is diminished. Conversely, prediction of academic outcomes is enhanced as self-efficacy and performance more closely correspond.
First, let me highlight what I believe to be particularly useful guidelines presented by Bandura (in press) regarding the specificity and correspondence of self-efficacy and performance assessment [Overhead 1, Table 1]. The broadest, most general, self-efficacy assessments would consist of an omnibus-type instrument that attempts to measure a general sense of "confidence." Such omnibus measures create problems of predictive relevance and are obscure about just what is being assessed. Omnibus tests that aim to assess general self-efficacy, for example, provide global scores that decontextualize the self-efficacy/behavior correspondence and transform self-efficacy into a generalized personality trait rather than the context-specific judgment Bandura (1986, in press) suggests it is. After all, generalized self-efficacy instruments basically assess "people's general belief that they can make things happen without specifying what [these things] are" (Bandura, in press). Even domain-specific omnibus measures are problematic if composite multiscale scores drawn from differing subsections of the domain are used. It is not altogether easy, for example, to see what value composite scores provided by multiple-scale instruments such as the often used Mathematics Self-Efficacy Scale may have, especially if one wishes to predict relatively discrete mathematics outcomes. The same might be said of a scale that would regard "writing self-efficacy" as the combined score of one subscale assessing a student's confidence to accomplish writing tasks of varying difficulty and another subscale assessing confidence to perform various composition skills.
Various researchers have assessed general academic self-perceptions of competence. The basic problem with such assessments is that students must generate these judgments without a clear academic activity or task in mind. As a result, they generate the judgments by in some fashion mentally aggregating related perceptions that they hope will be related to imagined tasks. Domain-specific assessments, such as asking students to provide their confidence to learn mathematics or writing, are more explanatory and predictive than omnibus measures and preferable to general academic judgments, but they are inferior to task-specific judgments because the subdomains differ markedly in the skills required.
Academic domain-specific assessments of self-efficacy are especially common in educational research in part because the criterial outcome tasks such as semester grades or achievement test results that are often used do not lend themselves well to particularized self-efficacy assessment. The typical strategy of researchers in this regard is to use multiple items to restate different facets (or even similar facets differently phrased) of the same academic subject. For example, it is not unusual for a mathematics self-efficacy scale to be populated with items such as "I am confident about my ability to do the work in this class;" "I am certain I can understand the math presented in this class;" and "I am confident I can perform as well or better than others in this class." Although high internal consistency can be counted on, such an assessment primarily provides a redundant measure of the general domain of mathematics.
Reasonably precise judgments of capability matched to a specific outcome afford the greatest prediction and offer the best explanations of performance outcomes, for these are typically the sorts of judgments that individuals use when confronted with behavioral tasks (Bandura, 1986). To this end, if the purpose of a study is to achieve explanatory and predictive power, self-efficacy judgments should be consistent with and tailored to the domain of functioning and/or task under investigation. This is especially critical in studies that attempt to establish causal relations between beliefs and outcomes. All this is to say that capabilities assessed and capabilities tested should be the same or similar capabilities. When these guidelines regarding correspondence between belief and outcome are not followed, the resulting loss of predictive power is ensured and the influence of self-efficacy is minimized. Some researchers, for example, have operationalized mathematics self-efficacy as students' judgments of their capabilities to solve math problems, perform math-related tasks, and succeed in math-related courses--the three subscales of the Mathematics Self-Efficacy Scale--although these judgments of math capabilities are substantively and conceptually quite different. If, as Bandura (1986) argued, self-efficacy assessment should be consistent with the criterial task to be useful and predictive, what criterial task is consistent with a composite score that comprises judgments of confidence to succeed in mathematics courses as diverse as geometry and accounting, complete math-related tasks as disparate as filling out an income tax form and figuring out how much material to buy so as to make curtains, and solve algebra and geometry problems of varying difficulty?
When these differing judgments of mathematics capability are compared with differing math-related outcomes--ability to solve the problems on which self-efficacy is assessed and math-relatedness of academic majors--results confirm that Bandura's (1986) guidelines regarding the match of self-efficacy and performance assessment are well founded (see Pajares & Miller, 1995). Students' confidence to solve mathematics problems is a more powerful predictor of their ability to solve those problems than is their confidence to perform math-related tasks or their confidence to earn high marks in math-related courses. Similarly, their confidence to succeed in such courses is more predictive of their choice of majors that required them to take many of the math-related courses on which they express that confidence. One might also question the practical utility of administering a 52-item instrument when greater prediction may be had from a shorter instrument more closely matching the performance task.
Let me also note that the skills required to accomplish the performance attainments that form the outcome assessment should be clear to the participant. When students do not know with any degree of accuracy what it is they are expected to do, the judgments on which they will base their capability to do it will be nebulous at best. When criterial tasks are unclear, what little prediction is obtained is primarily due to the similarity of tasks and the powers of generalizability individuals can bring to bear on their self-perceptions. In these cases, a researcher can do little more than predict from a domain-specific measure that taps common demands.
Multon, Brown, and Lent's (1991) meta-analysis of self-efficacy studies revealed that the effect sizes of self-efficacy on performance outcomes depended on specific characteristics of the studies, notably on the types of efficacy and performance measures used. The strongest effects were obtained by researchers who compared specific efficacy judgments with basic cognitive skills measures of performance, developed highly concordant self-efficacy/performance indices, and administered them at the same time. Significant relationships are obtained even with generalized self-efficacy indices, a phenomenon that Multon et al. described as reinforcing the theoretical and practical value of self-efficacy but that also tends to produce the confounded and misleading results to which I earlier referred. In fact, if global and generalized self-efficacy assessments can predict performances that are not specifically related, the relationship between properly assessed self-efficacy and performance should certainly increase.
Despite the optimal benefits that result from particularized efficacy and performance assessments, many academic outcomes of interest are not as particularized as, say, one's capability to solve specific mathematics problems, the typical level of specificity at which self-efficacy judgments are most predictive of academic performances. Lent and Hackett (1987) rightly observed that specificity and precision are often purchased at the expense of external validity and practical relevance. However, to be both practically useful and predictive, the level of specificity of an efficacy assessment should depend on the complexity of the performance criteria with which it is compared. Judgments of competence need not be so microscopically operationalized that the assessment loses all sense of practical utility (see Table 2 for sample self-efficacy items). If the criterial task should be a particularly relevant one such as choice of intention to enroll in math-related courses, self-efficacy judgments can be tailored to this level and still remain highly predictive. Lent, Lopez, and Bieschke (1993) showed how this can be accomplished when they compared students' confidence to succeed in math-related courses with three career-related outcomes--intention to take the courses listed on the instrument, grades obtained in math-related courses that students took during the subsequent term, and interest in the math courses listed on the instrument. Self-efficacy beliefs were predictive on each account, and such judgments offer information not afforded by broader judgments of competence.
Moreover, researchers have demonstrated that self-efficacy perceptions are also good predictors of reasonably generalized performances such as obtained grades (Bandura, 1993; Zimmerman, Bandura, & Martinez-Pons, 1991) or choice of academic majors (Hackett & Betz, 1989), and it bears repeating that the optimal level of specificity of any efficacy assessment should ultimately depend on the complexity of the performance criteria with which it is compared (see Lent & Hackett, 1987).
Let me close by briefly reporting on results of a study demonstrating that prediction of academic outcomes is enhanced as self-efficacy and performance more closely correspond [Overhead 2, Figure 1]. Colleagues and I are currently engaged in a study that in part attempts to investigate the relationship between mathematics self-efficacy beliefs assessed at three levels of specificity and their corresponding outcomes. To this end, we ask 8th grade students to report their confidence (a) to correctly solve each of 20 problems on a high-stakes mathematics exam, (b) in the various letter grades they will earn on the exam after completing the instructional unit but without seeing the specific exam questions, and (c) in the class grade they will receive for math at end of term. The self-efficacy assessments and the exam take place during the first week of a 9-week term. You will, I know, permit me to paint the canvas of this preliminary report with broad strokes as data input and collection is still ongoing.
Preliminary findings from our first assessment reveal that, when belief assessed and criterial task are matched, prediction is enhanced. Moreover, the strongest prediction is obtained from the most particularized match. In this study, when confidence to solve specific problems is compared to the number of problems solved (.57), the relationship is stronger than, say, when confidence in the overall mathematics test grade is compared to the grade on the test itself (.42). Also, note that the relationship between matched beliefs and outcomes is generally stronger than between unmatched variables even when the efficacy assessment is particularized. For example greater prediction of term grade is available from self-efficacy beliefs about term grades (.45) than from self-efficacy to solve problems (.34). As Bandura (in press) has outlined, the issue is one of specificity and correspondence. The Williams T2 statistic was used to determine that the correlations were significantly different
These findings demonstrate that the optimal level of specificity
of any efficacy assessment depends on the complexity of the
performance criteria with which it is compared, and that judgments of
competence need not be so microscopically operationalized that the
assessment loses all sense of practical utility. When a criterial
task of interest is relatively broad, such as term grades, self-efficacy judgments can be tailored to these levels and still remain
highly predictive. Various forms of self-referent thought measured at
various levels of specificity can also prove useful outside the
research arena as diagnostic and assessment tools--they can provide
teachers and counselors with information regarding students'
dispositions, and results may be useful in helping to understand
affective influences on performances that do not easily lend
themselves to microanalytic analysis.
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice Hall. Bandura, A. (1993). Perceived self-efficacy in cognitive development and functioning. Educational Psychologist, 28, 117-148. Bandura, a. (in press). Self-efficacy: The exercise of control. New York: Freeman. 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. Lent, R. W., & Hackett, G. (1987). Career self-efficacy: Empirical status and future directions. Journal of Vocational Behavior, 30, 347-382. Lent, R. W., Lopez, F. G., & Bieschke, K. J. (1993). Predicting mathematics-related choice and success behaviors: Test of an expanded social cognitive model. Journal of Vocational Behavior, 42, 223-236. Multon, K. D., Brown, S. D., & Lent, R. W. (1991). Relation of self-efficacy beliefs to academic outcomes: A meta-analytic investigation. Journal of Counseling Psychology, 38, 30-38. Pajares, F., & Miller, M. D. (1995). Mathematics self-efficacy and mathematics outcomes: The need for specificity of assessment. Journal of Counseling Psychology, 42, 190-198. Williams, E. J. (1959). The comparison of regression variables. Journal of the Royal Statistical Society, Series B, 21, 396-399. Zimmerman, B. J., Bandura, A., & Martinez-Pons, M. (1992). Self-motivation for academic attainment: The role of self-efficacy beliefs and personal goal setting. American Educational Research Journal, 29, 663-676.