Running Head: Frames of Reference and Course Selection Multiple Frames of Reference, Academic Interests, and Coursework Selection in Upper Secondary Schools in Germany

Data of n = 401 students (60.4% female) from German academically oriented secondary schools were collected at the middle of grade 10, at the end of grade 10, and in the middle of grade 12. Parallel measures for mathematics and English were collected, representing school grades, academic interest, academic selfconcept, and coursework selection (advanced vs. basic course). Consistent with a priori predictions, the results show that the extreme domain specificity previously demonstrated with academic self-concept generalizes to academic interest. Specifically, the effects of math interest were positive on math coursework selection but negative on English coursework selection, whereas the effects of English interest were positive on English coursework selection but negative on math coursework selection. This type of pattern was also evident for the effects of math and English self-concepts on math and English interest as well as math and English coursework selection. The findings are theoretically and practically important. Theoretically, predictions from the internal/external self-concept model are extended to academic interest and course selection, underlining the importance of multiple frames of reference when making important academic choices. Practically, the findings highlight the important role of academic interests and self-concepts for important career decisions beyond what can be explained in terms of prior academic achievement. Frames of Reference and Course Selection 3 According to the internal/external (I/E) frame of reference model (e.g., Marsh, 1986, 1990), academic self-concept in a particular school subject is formed in relation to an external (social comparison) reference in which students compare their self-perceived performances in a particular school subject with the perceived performances of other students in the same school subject and an internal (ipsative-like) reference in which students compare their performances in the particular school subject with their own performances in other school subjects. The external comparison process predicts that good math skills lead to higher math selfconcept and that good English skills lead to higher English self-concept. According to the internal comparison process, however, good math skills should lead to lower English self-concept and good English skills should lead to lower math self-concept. In the path model used to test this prediction, the paths leading from math achievement to English self-concept and from English achievement to math self-concept are predicted to be negative. Support for these predictions comes from a large body of research based on a variety of different self-concept measures (Marsh, 1990). In the present investigation we argue that academic interests as well as coursework selection in German upper secondary schools are also influenced by processes of external and internal comparisons in that a higher math self-concept lead to higher math interest and lower English interest, while a higher English self-concept enhances English interest but reduces math interest. The same pattern should be observable for the likelihood of choosing an advanced course in math and English. The internal/external (I/E) frame of reference model To establish a domain-specific self-concept of ability, students have to integrate achievement-related information from various sources. According to Marsh (1986, 1990) they compare their levels of academic ability using two different, but connected, frames of reference. In conducting social comparisons (external frame of reference), students compare self-perceptions of their own ability, for example, in mathematics and English, with the perceived abilities of their classmates, and this external reference serves as a basis for their math and verbal self-concepts. For example, if their math achievement is lower than their classmates’ achievement, then their math self-concept will be lower, too. Because achievements in school subjects are typically positively correlated, these external (social) comparisons lead to the expectation that domainspecific self-concepts are positively correlated as well. The existence of an internal frame of reference, however, helps to explain why typically there are only very small correlation coefficients between different academic self-concepts and why academic self-concepts are domain-specific. According to this internal frame of reference, students evaluate their achievements in any particular subject in relation to their achievements in other subjects. This internal comparison is the second basis for the formation of a self-concept in each subject. According to Marsh, Walker, and Debus (1991), better verbal skills lead to a better verbal self-concept based on external comparisons with classmates’ achievements in verbal domains, but to a poorer math self-concept based on internal comparisons with one’s poorer achievements in math. The relative weight of both comparison processes leads thus to the moderate or nearly zero correlations between math and verbal self-concepts found in a variety of studies with different methodologies carried out in several countries (Bong, 1998; Faber, 1992; Köller, Klemmert, Möller, & Baumert, 1999; Marsh, 1986, 1990; Marsh et al., 1991; Marsh & Yeung, 1998; Skaalvik & Rankin, 1992, 1995; Tay, Licht, & Tate, 1995; Williams & Montgomery, 1995; for an exception see Skaalvik & Rankin, 1990). The effects of combined internal and external comparisons can be illustrated by the students’ reactions when receiving achievement feedback (see Marsh, 1990). A student may accurately perceive himself as above average in math and English, compared to his peers (external), but as being better in math than in English (internal). The external comparison leads to a high self-concept in both subjects, but the internal comparison may moderate the impact of the positive social comparison on the students’ self-concept. The internal comparison results in a higher self-concept in math (above average with regard to both external and internal frames of reference) and reduces the self-concept in English (below average with regard to an internal frame of reference). If the internal comparison is dominant, this student may have an average or even a below average self-concept in English. As another consequence, the I/E model predicts in path Frames of Reference and Course Selection 4 analyses negative effects of achievement in one subject on self-concept in the other subject (see Marsh, 1986, 1990; Marsh et al., 1991, for empirical support). The competency/affect distinction in self-concept research Recent papers by Bong and Clark (1999) and Zimmerman (2000) have contrasted self-efficacy and academic self-concept research, arguing that one important distinction between the two constructs refers to their composition. While self-efficacy primarily deals with cognitively perceived capability of the self, academic self-concept, particularly in the research tradition of Marsh and Shavelson (e.g., Marsh, 1986; Shavelson, Hubner, & Stanton, 1976), contains both cognitive and affective aspects. Self-concept questionnaires developed by Marsh and colleagues (e.g., Marsh, Barnes, Cairns, & Tidman, 1984; Marsh & O’Niell, 1984) were specifically designed to reflect competency (or cognitive) and affect components of domain-specific academic self-concept. Self-perceived competency was defined by items asking students whether they were good at, learned things quickly in, had got good marks in, and found work to be easy in different school subjects. Affect was defined by items asking students whether they were interested in, looked forward to, liked, and enjoyed work in different school subjects. Putting both components together in research on the I/E model is typically justified by empirical findings suggesting that both the affective and the cognitive aspect is influenced by internal and external frames of reference (e.g., Marsh, 1986). In a recent factor-analytical work, however, Marsh, Craven, and Debus (in press) argued for a separation of the two components. Using large samples of students, the authors analyzed responses to the self-description questionnaire (SDQ-I). Confirmatory factor analyses revealed clearly distinguishable affective and cognitive factors for verbal, math, and school academic self-concept. The average correlation between competence and affect components was about .75 for reading and math and slightly lower for school. The separation of the competency and affect components is also supported by studies carried out with a German adaptation of the SDQ-I (Tänzer, 1996), and a Norwegian adaptation of the SDQ-II (Skaalvik, 1996). Consequently, an increasing number of researchers are turning toward a narrower definition of academic self-concept such as self-concept of ability (Brookover, Thomas, & Paterson, 1964; see also Bong & Clark, 1999). Wigfield and Karpathian (1991) noted results consistent with this work: They found that recent formulations of self-concept tend to emphasize competence perceptions as the key aspect of academic self-concept. In Eccles’ expectancy-value model of achievement and academic choice (e.g., Eccles, 1994; Eccles et al., 1983; Wigfield & Eccles, 1992, 2000) self-concepts of ability and task value are clearly distinguished. Although based on somewhat different constructs, the work by Eccles and her colleagues focuses on how students’ competency-related beliefs in specific activities (e.g., how good they are) combine with task value (e.g., interest, enjoyment, usefulness, importance) to influence academic choice. Of particular importance for the present investigation is that Eccles et al. posit a causal path from the self-concept of ability to the task value, meaning that perceived competence within a domain influences all aspects of the task value, including interest. Summarizing the research mentioned above, there is currently a strong tendency to separate cognitive and affective components of academic self-concept. This separation is justified both by empirical findings and by theoretical frameworks such as the Eccles model. Conceptualization of academic interests Current research on interest, at least in Germany, is mainly based on the theoretical framework provided by H. Schiefele and colleagues (e.g., Krapp, Hidi, & Renninger, 1992; Prenzel, 1988; Prenzel, Krapp, & H. Schiefele, 1986; U. Schiefele, 1992, 1996a, b). These authors conceptualize interest as a person-object relation that is characterized by value commitment and positive emotional valences. Of substantial theoretical importance is the distinction of an affective (feeling-related) and a value-related component, presented in Table 1 (cf. U. Schiefele, 1992). A similar distinction is made in Eccles’ extended expectancyvalue model (see above), in which intrinsic interest and personal importance are aspects of the value component. While the feeling-related component of interest in the sense of U. Schiefele (1992) is quite Frames of Reference and Course Selection 5 similar to intrinsic interest in the expectancy-value model, the value-related component is theoretically close to the personal importance aspect in the Eccles model. -------------------------------------------insert Table 1 about here -------------------------------------------Interests can pertain to objects in the physical or natural environment, to symbolic representations, or to activities (Rheinberg, 1989). Individual interests are conceived of as dispositions that are based on mental schemata associating the objects of interest with positive emotional experiences and the personal value system. During interest-driven actions, the latent disposition (trait) becomes the actualized interest (state). The experience of competence and personal control, the feeling of autonomy and self-determination, and a positive emotional state are all characteristic of this action. In the optimal scenario, there may be an experience of flow when person and object coincide (Csikszentmihalyi & Nakamura, 1989; Csikszentmihalyi & Schiefele, 1993). Actualized interest must be distinguished from situational interest, which is also conceptualized as a state phenomenon. In contrast to actualized interest, however, situational interest is generated by external stimuli -the interestingness of the situation or object. A series of actions eliciting situational interest may lead to the development of dispositional interest. Situational interest has primarily been studied in the domain of text-based learning. Several researchers have reported findings that support the assumption that achievement and/or selfperceived competence affect academic interest (Hacket & Campbell, 1987; Köller, Schnabel, & Baumert, 1998; Lopez, Lent, Brown, & Gore, 1997; Sjoeberg, 1985). This relationship is supported by the theoretical work of Eccles in her expectancy-value model, where self-concept of ability indeed is assumed to affect interest as a part of the task value. Academic self-concept, interest, and academic choices Academic self-concepts as well as interests are often considered as important determinants of course selections in high school. Eccles’ expectancy-value model, which is the most prominent theoretical approach to predicting academic choices, contains both variables, that is, self-concept of ability as an antecedent of the expectancy component, and interest as part of the value aspect. Several studies have shown the important role of both aspects (cf. Meece, Wigfield, & Eccles, 1990; see Wigfield & Eccles, 2000, for an overview). According to this model, academic choices are posited to be a function of task values and expectancies of success. Expectations of significant others (= subjective norm in the sense of Ajzen & Fishbein, 1980), previous academic achievement, and the interpretation of the achievement outcomes influence the self-concept of ability, which then influences expectancies and values. Studies using confirmatory factor analysis show that items measuring self-concept of ability and items measuring expectancies are empirically not readily distinguishable (Eccles & Wigfield, 1995; Wigfield & Eccles, 2000). As a consequence, several studies have shown that there are strong direct paths from academic selfconcept to coursework selection. Eccles et al. (1983) found that self-concept had both a direct effect on coursework selection plans and indirect effects that were mediated by achievement expectancies. Also on the basis of the expectancy-value model, Feather (1988) investigated university course selection. The effects of math self-concept on course selection in mathematics and English self-concept on course selection in English were of substantial size. Interestingly, the effect of self-concept was stronger for math than for English. Similar findings were reported by Marsh (1989). Meece et al. (1990) assessed how 7 to 9 graders’ ability beliefs and performance in math Year 1 predicted their expectancies for success, perceived importance of math, math anxiety, and intentions to continue taking more math in Year 2. The authors found 1 Note, however, that there is a substantial theoretical difference between the two constructs in that the selfconcept of ability is formed by prior experiences in achievement situations, whereas expectancies are futureoriented, that is, focus on anticipated outcomes. Frames of Reference and Course Selection 6 that Year 1 self-concept of math ability directly and positively predicted Year 2 expectancies for success and importance attached to math. Year 2 importance ratings and (less strongly) expectancies for success predicted intentions to continue taking math in school. Marsh and Yeung (1997) examined the role of academic self-concept for subsequent intentions to take courses in nine school subjects. With the exception of one subject, the domain-specific academic self-concept came out as a significant predictor of the intention to take courses in the corresponding subject. Note that the effects of self-concept were significant even when grades were included as additional predictors. To summarize these findings, there is both theoretical and empirical support for the assumption that academic self-concepts and interests are important predictors of academic choices in terms of coursework selection in school. The German secondary school system When investigating coursework selection in German schools, one must consider some special characteristics of the educational system in Germany. The German school system is characterized by early and selective assignment to different secondary school tracks, that is, Hauptschule, Realschule, and Gymnasium. After four years of elementary school, students generally enter one of the three types of secondary school; the choice of track is based substantially on their performance in the first years of school. Typically, Hauptschule and Realschule students (vocational track) complete school grades 5 to 9 or 10 while Gymnasium students (academic track) attend school up to grade 13 and are then expected to attend university. For all students there is a compulsory curriculum of classes that are taken in a fixed order and together with the same classmates. That is, up to grade 10, students are not allowed to drop classes in accordance with their abilities or interests. As a consequence, they cannot avoid classes with a heavy course load. The picture changes at the beginning of secondary level II, however, when students have to choose advanced courses in two or three subjects and basic courses in the remaining subjects. Advanced courses usually comprise five to six lessons per week, whereas basic courses involve two to three lessons per week. Where the choice of advanced courses is concerned, there are certain restrictions that typically prompt students to choose either mathematics and/or science and/or German and/or a foreign language. As a result, about one third of all students choose mathematics as an advanced course, while approximately two thirds take a basic course in mathematics. The same is true for English. The present investigation Central assumptions of the I/E model relating different frames of reference and self-concepts have been tested in several studies (e.g., Marsh, 1986, Marsh & Yeung, 1998; Skaalvik & Rankin, 1992, 1995; Tay et al., 1995). The present study tries to extend the I/E model in that effects of multiple frames of reference on academic interest and coursework selection are investigated. School subjects considered here are mathematics and English (as a foreign language). Bearing in mind recent studies that claim a clear distinction between affective and cognitive components (e.g., Bong & Clark, 1999), the cognitive part of the academic self-concept is conceptualize here as a self-concept of ability, and the affective component is part of the interest construct, which also contains the personal importance (see above). Keeping in line with previous research (e.g., Marsh, 1986) we hypothesize I/E effects on both the self-concept of ability and academic interest. Furthermore, we hypothesize on the basis of Eccles’ expectancy value model (e.g., Wigfield & Eccles, 2000) that domain-specific self-concepts of ability are determinants of academic interest, in that the self-concept increases interest in the same domain. As an extension not explicitly considered in Eccles’ model we also hypothesize effects across domains, that is, self-concept within one domain should reduce academic interest in the other domain. Note that the distinction of an external and internal (ipsativelike) frame of reference is usually also considered by Eccles and colleagues when they measure domainspecific self-concepts of ability. Hence, a typical item wording for the external frame of reference is: “If you were to list all the students in your class from the worst to the best in math, where would you put yourself? (one of the worst / one of the best)”; and for the internal frame of reference: “Some kids are better in one Frames of Reference and Course Selection 7 subject than in another. For example, you might be better in math than in reading. Compared to most of your school subjects, how good are you in math? (a lot worse in math than in other subjects / a lot better in math than in other subjects)” (Wigfield & Eccles, 2000, p. 70, Table 1). To our knowledge, however, there is no study by Eccles and colleagues in which they have investigated the unique effects of both frames of reference on task value or academic choice. In regard to coursework selection, the expectancy-value model suggests self-concept of ability and interest to be central determinants of this academic choice. Again, in an extension of the expectancy-value model, we assume negative effects of grades, self-concept, and interest in one domain on course selection in the other domain. Figure 1 summarizes the expectations of the present study; negative signs indicate negative effects, whereas positive signs represent positive effects. -------------------------------------insert Figure 1 about here ---------------------------------------