Learning and Individual Differences 28 (2013) 9–19

Contents lists available at ScienceDirect

Learning and Individual Differences

j ourna l homepage: www.e lsev ie r .com/ locate / l ind i f

Further evidence on the structural relationship between academicself-concept and self-efficacy: On the effects of domain specificity

Ronny Scherer ⁎Humboldt-Universität zu Berlin, Humboldt Graduate School, Unter den Linden 6, 10099 Berlin, Germany

⁎ Tel.: +49 176 6593 5151.E-mail address: ronny.scherer@chemie.hu-berlin.de.

1041-6080/$ – see front matter © 2013 Elsevier Inc. All rihttp://dx.doi.org/10.1016/j.lindif.2013.09.008

a b s t r a c t

a r t i c l e i n f o

Article history:Received 7 May 2013Received in revised form 26 July 2013Accepted 6 September 2013

Keywords:Chemistry educationConfirmatory factor analysisConstruct validityDomain specificitySelf-conceptSelf-efficacy

Given the importance of students' competence beliefs in science learning, many researchers have focused on theinterplay between self-concept andperformance in various domains.However, little research has beenundertak-en on the structure of competence beliefs and the domain specificity in scientific subjects such as chemistry. Thisstudy, consequently, aims to analyze the structure of competence beliefs by taking into account components ofself-concept and self-efficacy as well as domain and construct effects. By using the data of 459 German high-school students of grade levels 10 to 13, it was found that structural models, which distinguish between generalself-concept, chemistry self-concept and chemistry self-efficacy, represented the data reasonably well. The re-sults provide evidence for (1) the empirical distinction between self-concept and self-efficacy within the domainof chemistry; (2) significant differences between general academic and domain-specific self-concept; and (3)substantial relationships among students' competence beliefs and school achievement. Furthermore, teachers'orientations towards hands-on inquiry activities and students' enjoyment in science were strongly related toself-concept and self-efficacy. Based on present competence-oriented curricula, it was possible to clarify therelationship among self-concept and self-efficacy in chemistry.

© 2013 Elsevier Inc. All rights reserved.

1. Introduction

Students' competence beliefs can be regarded as essential predictorsof performance, motivation, and learning. Especially in scientificdomains, these constructs play an important role when it comes toachieving learning goals, developing adequate epistemic beliefs, andsolving problems (Mason, Boscolo, Tornatora, & Ronconi, 2013; Tsai,Ho, Liang, & Lin, 2011). Competence beliefs are broadly defined as“children's cognitive representations of how good they are at a givenactivity” (Freiberger, Steinmayr, & Spinath, 2012, p. 518). In this context,they refer to self-concepts and self-efficacy which have been extensive-ly studied for general academic, mathematical, and verbal domains(e.g., Van Dinther, Dochy, & Segers, 2011). For instance, Marsh,Walker, and Debus (1991) were able to show that academic self-concept comprises different components within a hierarchical structureand that self-concept and self-efficacy are related constructs. However,research suggests that there is evidence on the empirical distinctionbetween self-concept and self-efficacy (e.g., Bong & Skaalvik, 2003;Ferla, Valcke, & Cai, 2009; Marsh et al., 1991). Wagner, Göllner,Helmke, Trautwein, and Lüdtke (2013) and Marsh and Scalas (2010)pointed out that students' perceptions of competences or classroom-based aspects such as instructional quality can be regarded as multidi-mensional and domain-specific. However, little research has been pro-posed on whether or not this distinction holds for scientific domains

ghts reserved.

such as chemistry. In this context, the conceptual approach of analyzingdomain specificity, which was proposed by Brunner (2008), could pro-vide a reasonable tool to address this shortcoming. Additionally and dueto a lack of appropriate assessments, little is known about how studentsevaluate their competences according to the demands of specific curric-ular standards. In light of recent developments on establishing nationalstandards in scientific subjects in Germany, it is of interest to assess stu-dents' specific competence beliefs in order to use these as sources of in-dividual feedback and predictors of achievement (Köller & Parchmann,2012; Marsh & Martin, 2011).

The present study, consequently, aims to analyze the relationship be-tween chemistry-specific self-concept and self-efficacy and intends tocheck whether or not these two constructs are related and could bedistinguished from general academic self-concept. In this context, amethodological approach is presented which allows researchers to ob-tain evidence on domain and construct specificity (Marsh et al., 2013).Besides analyzing the structure of students' competence beliefs (internalvalidation), their relationshipswith further constructs are investigated inorder to externally validate the assessment (Messick, 1995). In thisstudy, methods of structural equation modeling are applied to a sampleof 459 German high-school students.

1.1. Literature review

1.1.1. Construct definitionsIn general, academic self-concept has been defined as students'

perception of themselves within the academic environment (Marsh,

10 R. Scherer / Learning and Individual Differences 28 (2013) 9–19

1990; Marsh & Scalas, 2010). Self-concept is continually formed byexperience and interaction with the environment (Bandura, 1997;Spinath & Steinmayr, 2012). In educational sciences, this construct hasbeen extensively studied and, up to now,much is known about the gen-eral structure of self-concepts. For instance, research has shown thatthere is evidence for a multidimensional and hierarchically organizedstructure of the construct (e.g., Arens, Craven, Yeung, & Hasselhorn,2011;Marsh, 1990;Marsh et al., 1991). On the first level, one can distin-guish between academic and non-academic self-concepts (Marsh &Martin, 2011; Marsh & Scalas, 2010). The second level comprises sub-jects as further factors which are indicated by perceptions of one's com-petence in these subjects (Marsh et al., 1991). This model has becomeknown as the Marsh/Shavelson model (Brunner, Keller, Hornung,Reichert, & Martin, 2009; Marsh, 1990).

In educational research, academic self-efficacy refers to students' per-ceptions on their ability tomaster given tasks or develop specific compe-tences (Bong, 2001; Gallagher, 2012; Van Dinther et al., 2011).Consequently, these self-beliefs are task- and future-oriented (Bong &Skaalvik, 2003). Hoffman and Schraw (2009) pointed out that students'self-efficacy in science plays an important role in solving problems,which demand a higher level of working memory capacity in science.Also, Chen and Usher (2013) argued that self-efficacy strongly affectsstudents' general abilities and competences in science. This argumenta-tion was supported by further studies which systematically analyzedthe effects of different aspects of self-efficacy on students' achievement(e.g., Freiberger et al., 2012; Hoffman & Schraw, 2009). Chen and Usher(2013) further investigated the sources of science self-efficacy andproposed a framework with four determining factors: observation of ac-tivities, verbal and social persuasions from others, interpretation of pastperformance andmaster experience, and affective states such as anxiety.

1.1.2. On the empirical relationship between self-concept and self-efficacyPajares andMiller (1994) showed that self-efficacywas an important

predictor of achievement in math whereas students' gender mediatedthis relationship in favor of males. Similar results were obtained byPietsch, Walker, and Chapman (2003) who, additionally, studied therelationship among self-concept and self-efficacy. In their study, they ar-gued that students' self-efficacy and their general perceptions of compe-tences were empirically distinct. Further research suggests that bothconstructs are positively and substantially correlated (Bong & Skaalvik,2003). On a conceptual level, both refer to students' perceptions oftheir competences and their evaluations of mastering tasks and prob-lems in the academic environment. Consequently, they are mainlybased on mastery experience, performance, and behavior of avoidance(Mason et al., 2013) and focus on perceived competences which areoften referred to as “competence beliefs”. Based on these competence be-liefs, performance goals could be developed (Komarraju & Nadler, 2013;Mason et al., 2013; Spinath & Steinmayr, 2012). Moreover, there is evi-dence that both constructs are domain-specific and multidimensionalin nature (Bong, 2001; Bong & Skaalvik, 2003; Marsh & Scalas, 2010).

On the other hand, self-concept and self-efficacy are also different, asthey differ in the context of students' evaluations. Self-efficacy largelyrefers to context-specific judgments whereas self-concept mainly relieson aggregated and global perceptions (Bandura, 1997; Bong & Skaalvik,2003). Additionally, there has been empirical support for the structuraldistinction of both constructs which wasmainly based on language andmath learning (Bruning, Dempsey, Kauffman, & McKim, 2013; Marshet al., 1991). For instance, Ferla et al. (2009) found a moderate correla-tion for math (ρ = .37), meaning that students who perceive their spe-cific competences in math as high are more likely to regard themselvesas generally competent in this subject. This result appears reasonable ifself-efficacy is defined as a more specific competence belief (e.g.,Gallagher, 2012; Van Dinther et al., 2011). Consequently, the constructsof self-concept and self-efficacy are both components of students' com-petence beliefs. However, only a few studies systematically analyzed

this relationship for scientific domains (Bruning et al., 2013; Ferlaet al., 2009; Lewis, Shaw, Heitz, & Webster, 2009; Lin & Tsai, 2013).

Regarding the covariates of self-efficacy and self-concepts, there isa great variety of common factors. For instance, in many studies, re-searchers found significant and substantial effects of interest(Freiberger et al., 2012), gender (Velayutham, Aldrige, & Fraser, 2012),anxiety (Ferla et al., 2009), achievement as indicated by grades(Brunner et al., 2009; Velayutham et al., 2012), epistemological beliefs(Tsai et al., 2011), and cultural differences (Lee, 2009; Marsh et al.,2013). These findings enable researchers and teachers to predict and in-fluence students' competence beliefs inmany different settings (Huang,2011; Van Dinther et al., 2011). Lin and Tsai (2013) pointed out thatthese relationships could also be used to obtain evidence on constructvalidity in terms of an external validation (see also Messick, 1995).

1.1.3. Assessing and modeling self-concept and self-efficacyBy extending theMarsh/Shavelsonmodel, Brunner et al. (2009) pro-

posed a newmeasurement perspective which did not only take into ac-count the different factors of self-concept but also the issue of domainspecificity. In theirmodel, they suggested a nested structure of students'academic self-concepts (Fig. 1) and assumed that,first, self-concepts aredomain specific and, second, there is a general academic factor whichshares variance with specific self-concepts (see also Brunner et al.,2010). From a statistical point of view, the Correlated-Trait-(Method-Minus-One) model represents this structure within a confirmatory fac-tor analysis framework (CTC[M-1] model; Eid, 2000). However, thismodel has not yet been applied to self-concept and self-efficacy inchemistry in order to address domain and construct specificity.

As Tsai et al. (2011) suggested, valid assessments of specific compe-tence beliefs (self-efficacy) can be designed by using specific descrip-tions of competences required to achieve a learning goal or to solve ascientific problem. Such operationalizations are necessary in order tocapture students' individual convictions on mastering academic tasksaswell as their test-taking efforts and perseverance in specific problems(Bandura, 1997; Liu, 2010; Pajares & Miller, 1994). By using statementsof self-perception regarding different competences requires an appro-priate rating scale which is used to measure the outcome on differentlevels. Liu (2010), thus, proposed developing items with Likert-typescales to capture these perceptions.

Based on a multidimensional framework, Lin and Tsai (2013) wereable to develop a rating test on science self-efficacy with considerablysubstantial evidence on reliability and validity. The resulting factorialstructure referred to different competences in science classrooms: (1)conceptual understanding, (2) higher-order cognitive skills, (3) practi-cal work, (4) everyday application, and (5) science communication.This approach contrasted unidimensional assessments of self-efficacyin science (Glynn, Brickman, Armstrong, & Taasoobshirazi, 2011;Glynn, Taasoobshirazi, & Brickman, 2009; Hoffman & Schraw, 2009)and stressed the importance of defining an appropriate conceptualframework of the competences, used as indicators of self-efficacy(Bong & Skaalvik, 2003; Lin & Tsai, 2013; Marsh et al., 1991). As thecurricular specifications of scientific literacy play a crucial role inimplementing science education (Köller & Parchmann, 2012), thesecould form the basis for developing assessments of students' compe-tence beliefs in amore specific context, providing feedback towards par-ticular standards. In Germany, for instance, there are four competenceswhich take into account different skills and abilities in science class-rooms. These are mainly based on the concept of scientific literacy(see Table 1; Neumann, Fischer, & Kauertz, 2010). This frameworkdoes not only refer to the acquisition of domain knowledge but also in-cludes inquiry, communication, and decision-making skills. Due to thisbroad operationalization, which is specified for chemistry as a scientificdomain, the framework was used to establish standards in chemistryeducation for 10th graders in Germany. Although there is some overlapwith the multidimensional conceptualization of science self-efficacyproposed by Lin and Tsai (2013), which was developed for science as

Fig. 1. Nested structure of academic self-concepts. Note. SC = self-concept, A = academic, M = mathematics, V = verbal.Brunner et al. (2009) p. 390, modified.

11R. Scherer / Learning and Individual Differences 28 (2013) 9–19

an integrated subject, these standards take into account chemistry-specific abilities and skills, especially within the context of understand-ing particular concepts.

1.2. The present study

The present study, consequently, aims to investigate the distinctionbetween self-concept and self-efficacy in chemistry, which are regardedas two important predictors of students' achievement (Arens et al.,2011; Bong, Cho, Ahn, & Kim, 2012; Lin & Tsai, 2013). Moreover, thedomain and construct specificity of competence beliefs are analyzed.Finally, this study is aimed at:

1) Developing a test on students' self-efficacy which exemplarilyrelates to curricular demands of competences in chemistry (testdevelopment)

2) Analyzing the structure of students' competence beliefs by takinginto account general academic self-concept, chemistry self-concept,and chemistry self-efficacy (internal structure)

3) Modeling the specificity of competence beliefs (domain and constructspecificity)

4) Investigating the relationships among competence beliefs andcovariates (external validity).

2. Methods

2.1. Participants

Participants were 459 high-school students of grade levels 10 to 13who attended a chemistry course in one of five selected schools in theGerman federal states of Berlin and Brandenburg. These schools were

Table 1Description of the four competenceswithin the German standards of chemistry education.

Competence Description

Domain knowledge Students apply their content knowledge to solveproblems and tasks. They understand the basicconcepts in chemistry such as thestructure-properties-relationship.

Scientific inquiry Students acquire knowledge by formulating,testing, and evaluating hypotheses inexperimental settings.

Decision-making Based on multiple perspectives and scientificdata, students choose among different problemsolutions and evaluate their appropriateness.

Scientific communication Students present chemistry-related topics byusing appropriate scientific terms and multiplerepresentations.

selected according to their willingness to participate in this study.53.4% of the studentswere female and 78.5% had a German background.Students' mean age was 16.6 years (SD = 1.2 years) and ranged be-tween 14 and 21.

2.2. Measures

In order to facilitate achieving the study's research goals, differentaspects of students' competence beliefswere assessed. These aspects re-ferred to general academic and chemistry self-concept as well as chem-istry self-efficacy. Additionally, covariates were taken into account inorder to check for external validity.

2.2.1. Students' self-conceptsIn this study, two scales of self-concept were used: The first refers to

general academic self-concept; the second refers to domain-specific self-concept for chemistry. The appendix contains item examples of each ofthese three scales.

2.2.1.1. General academic self-concept. Students' general academic self-concept was assessed by using the German version of the PISA studentquestionnaire, which contained four items on general self-perceptionsin school (OECD, 2009). These items were answered on a 4-point Likertscale ranging from 0 = I disagree to 3 = I totally agree. The underlyingconstruct of general academic self-concept was used to obtain evidenceon the domain specificity of chemistry self-concept, following the em-pirical approach by Brunner (2008).

2.2.1.2. Chemistry self-concept. Students' chemistry self-concept wasassessed as another dimension of competence beliefs in educationalsettings (Liu, 2010; Marsh, 1990). This construct strongly referred toevaluations on competences and knowledge in chemistry in a generalcontext. In this study, the PISA 2006 self-concept questionnaire wasused (OECD, 2009), which contained 6 itemswith a rating scale rangingfrom 0 = I disagree to 3 = I totally agree. These items were adapted tothe domain of chemistry. Since previous research on academic self-concepts has focused on students' self-concept in science, physics, orbiology (e.g., Arens et al., 2011; Brunner et al., 2010; Marsh & Martin,2011), the proposed study consequently extends this construct to thedomain of chemistry.

2.2.1.3. Chemistry self-efficacy. In this study, the descriptions of compe-tences within the German notion of educational standards were usedto develop items which refer to different aspects of scientific literacy(e.g., Köller & Parchmann, 2012). In this context, four competenceswere distinguished: (1) the application of domain knowledge, (2) the

12 R. Scherer / Learning and Individual Differences 28 (2013) 9–19

acquisition of domain knowledge (scientific inquiry), (3) decision-making and evaluation, and (4) scientific communication. Based onthese four competences, specific descriptions of scientific skills were de-veloped and used as items of the self-efficacy scale. The resulting test onchemistry self-efficacy comprised 19 items which had to be rated on a4-point Likert scale (0 = I disagree to 3 = I totally agree). The appendixshows item examples. It is noted that the resulting test was presumablybased on the curricular framework of chemistry in Germany. Conse-quently, the present study exemplarily shows how to develop such atest by referring to existing curricula.

2.2.2. Covariates of competence beliefsIn this study, different covariates of competence beliefs were

measured to obtain evidence on external validity (Messick, 1995).These variables were chosen after a review of existing research andliterature (e.g., Chen, Yeh, Hwang, & Lin, 2013; Ferla et al., 2009;Freiberger et al., 2012; Jonkmann, Becker, Marsh, Lüdtke, & Trautwein,2012; Marsh & Scalas, 2010). The resulting predictors can be classifiedas achievement-based,motivational, classroom-based, andperceptionalvariables. In accordance with existing research, these variables functioneither as factors or sources of competence beliefs (Bong & Skaalvik,2003; Chen & Usher, 2013).

2.2.2.1. Grades in chemistry. Students' grades in chemistrywere collectedas indicators of domain-specific achievement. According to the Germangrading system, a grade of 1 represents a very good and outstandingachievement, whereas “6” refers to an insufficient achievement. The rela-tionship between the numerical value of the grade and performance is,thus, expected to be negative.

2.2.2.2. Enjoyment in science. Students' general enjoyment in sciencewasassessed by the JOYSCIE scale of the PISA 2006 student questionnaire(OECD, 2009). In this scale, students' task was to evaluate 5 statementson a 4-point Likert scale (0 = I disagree to 3 = I totally agree).

2.2.2.3. General value of science. In order to take into account theperspec-tives from which students evaluate science and scientific contents, ashort scale on the general value of science was administered. Again,this scale was taken from the PISA 2006 questionnaire and contained5 itemswhich had to be evaluated on the same Likert scale as the state-ments of the previously described enjoyment scale (OECD, 2009). Forinstance, the statement of “Science is important in order to understandthe natural world” was part of this scale.

This scale could be regarded as an indicator of students' epistemo-logical beliefswhich are closely related to their understanding of the na-ture of science (Liu, 2010). As Schwartz and Lederman (2008) pointedout, the nature of science refers to values and assumptions towardsthe development of scientific knowledge. Moreover, previous researchsuggested that students' beliefs in their own skills and processes ofknowledge acquisition are affected by the way in which they value orregard science (e.g., Urhahne, Kremer, & Mayer, 2011). Furthermore, ithas been found that understanding the nature of science is quite generaland similar in different contexts or science subjects (Schwartz &Lederman, 2008; Urhahne et al., 2011). Consequently, students' valuesof science have been assessed for science as an integrated subject. Asvalues and views of science are implicit and, thus, difficult to assess byusing explicit statements, internal consistencies of such scales usuallyrange between .50 and .70 (Muis, Bendixen, & Härle, 2006).

2.2.2.4. Hands-on activities in chemistry classrooms. In order to addressstudents' experience with inquiry-based activities in chemistry lessons,the PISA 2006 scale SCHANDS (hands-on activities in science) was ad-ministered (OECD, 2009). In this four-item scale, students had to statehow often specific activities such as conducting experiments and ob-serving the results occurred in their chemistry lessons (0 = never oralmost never, 3 = in almost every lesson).

From a science educator's perspective, teaching chemistry throughhands-on activities could lead to a more positive attitude towards thissubject which could, subsequently, enhance students' competence be-liefs (Freedman, 1997; Usher & Pajares, 2009). From an assessment per-spective, the degree to which such activities are present in chemistryclassrooms provides more information on the validity of students' rat-ings of their skills and competences. However, this scale does not reflectthe entire process of scientific inquiry which is demanded in curricula.Instead, it refers to selected competences related to this construct(Frey et al., 2009).

2.2.2.5. Students' background. Students' background referred to theirmother language. As Niehaus and Adelson (2013) found significanteffects of language background on students' self-concept, this dummy-coded variable has been incorporated in the questionnaire (0 = non-German and 1 = German).

2.3. Procedure

In this survey, a cross-sectional design was chosen in order to de-scribe students' competence beliefs at one measurement occasion. Thetests on competence beliefs and covariates were administered byusing a paper-and-pencil assessment which was formatted as the PISAstudent questionnaires. Students completed this questionnaire in classwithin 20 min. In all parts of the survey, students were able to returnto previous items and correct their answers if necessary. The resultingdata were coded in SPSS (Version 19, IBM, 2010).

2.4. Statistical analyses

In order to analyze the structure, domain and construct specificity ofstudents' competence beliefs, confirmatory factor analyses with andwithout nested factors were conducted in Mplus (Version 6, Muthén &Muthén, 2010). Especially the analysis of domain and construct speci-ficity used a model which could be compared to the Correlated-Trait-Correlated-Methods-Minus-One (CTC[M-1]) approach presented byEid (2000). In these analyses, the robust maximum likelihood estimatorwas used and evaluated by taking into account the following goodness-of-fit statistics: the Satorra–Bentler scale corrected χ2 value (SB-χ2),the Comparative-Fit-Index (CFI), the Root-Mean-Square-Error-of-Approximation (RMSEA), and the Standardized-Root-Mean-Square-Residual (SRMR) (Brown, 2006). Common guidelines for a substantialmodel fit require a CFI value above .90, an RMSEA below .08, and aSRMR below .09 (Hu & Bentler, 1999). However, the statistical signifi-cance of the SB-χ2 value strongly depends on the sample size and can,thus, reveal an insignificant statistic although the model fits the data(Brown, 2006). In order to compare competing and nested models,the χ2-difference test with a Satorra–Bentler correction was applied(Bryant & Satorra, 2012). In all analyses, missing data were handledby using the Full-Information-Maximum-Likelihood (FIML) procedure,implemented in Mplus (Enders, 2010). This procedure assumes thatmissing data patterns are at least missing at random (MAR). By usingLittle's MCAR test (missing completely at random), the data providedevidence for MCAR (χ2(2193, N = 459) = 2282.04, p = .09), whichlegitimizes the FIML procedure because an even stronger assumptionon missings holds (Little, 1988). Since missings are likely to followthis mechanism, a multiple imputation procedure which creates com-plete data sets yielded the same values of internal consistencies as theincomplete data set (Enders, 2010).

3. Results

3.1. Descriptive statistics and test development

Descriptive statistics of the test scales are shown in Table 2. The testson self-concepts revealed high values of reliability and acceptable to

Table 2Descriptive statistics of the test scales (N = 459).

Scale Ni M SD Min Max α PISA αa

Academic self-concept 4 8.00 2.52 1 12 .84 .90Self-concept in chemistry 6 9.52 4.07 0 18 .90 .90Self-efficacy in chemistry 19 33.11 9.38 0 54 .90 .82b

Enjoyment in science 5 9.68 3.54 0 15 .88 .92General value of science 5 10.31 2.91 0 15 .78 .81Hands-on activities in chemistrylessons

4 7.12 2.13 0 12 .65 .71

Note. Ni = number of items, α = Cronbach's Alpha based on the complete data set aftermultiple imputation with predictive mean matching (20 imputations; Enders, 2010).

a PISA α = value of Cronbach's Alpha for the PISA 2006 scales for science in Germany(N = 4891; Frey et al., 2009).

b Note that this scale referred to specific problems of science which were describedwithin 8 items.

13R. Scherer / Learning and Individual Differences 28 (2013) 9–19

good item-to-total correlations. Especially, the newly developed scale ofchemistry self-efficacy showed sufficient test characteristics (α = .90,rit = .36–.67). In light of these results, all items remained in furtheranalyses.

Considering the tests on enjoyment in science and the general valueof science, both scales were sufficiently reliable. The internal consisten-cy of the hands-on activities in science scale was low compared to theother constructs. As teachers reported, this might be due to the factthat students were only able to plan and perform scientific experimentsand hands-on activities in a few science courses which could havecaused a lack of variability in the data set. Furthermore, the internal con-sistencies were quite similar to the values obtained in PISA 2006 for theGerman sample.

3.2. Structure and dimensionality of students' competence beliefs

In order to analyze the structure of students' competence beliefs inchemistry, a series of nested models was specified which representeddifferent theoretical assumptions. Table 3 shows the goodness-of-fitstatistics of these CFA models.

First, it was assumed that competence beliefs comprise a single la-tent factor without further differentiations into self-concept and self-efficacy (model 1dim). This unidimensionalmodelfitted the data poorlyandwas, thus, rejected. In a second step, a two-factormodelwas appliedwhich distinguished between self-concept and self-efficacy, withouttaking into account the domain-specific or domain-general characterof self-concepts (model 2dim). Again, this model was rejected due to apoor goodness-of-fit. A three-factor model assuming general academicself-concept, chemistry self-concept, and chemistry self-efficacy asthree correlated but distinct constructs showed a substantial but stillunacceptable model fit (model 3dim). Based on empirical modificationindices, which were obtained in Mplus, correlations among item resid-uals have been introduced (model 3dim corr). This approach yielded asubstantial and acceptable model fit. By using the corrected χ2 differ-ence test, this model fitted the data significantly better than the three-dimensional model without correlated residuals (SB-χ2(8) = 301.9,p b .001). From a theoretical point of view, introducing correlated

Table 3Goodness-of-fit statistics of structural models of competence beliefs.

Model SB-χ2 (df) CFI RMSEA CI90 SRMR

1dim 2095.6 (377)⁎⁎⁎ .71 .09 (.09, .10) .082dim 1792.1 (376)⁎⁎⁎ .76 .09 (.08, .09) .083dim 1263.3 (374)⁎⁎⁎ .85 .07 (.06, .07) .063dim(gf) corr 811.1 (366)⁎⁎⁎ .92 .05 (.05, .06) .05

Note. corr = model with correlated residuals among items, gf = model with a second-order general factor, CI90 = 90% confidence interval of the RMSEA.⁎⁎⁎ p b .001.

residuals reflects the relatedness of different science competences.For instance, in the context of scientific inquiry, students' competencesof planning a scientific experiment are likely to affect subsequentsteps such as the interpretation or communication of results (seeAppendix A). Consequently, residual dependencies occur in their ratings(see also Marsh et al., 2013).

Due to the assumption that each of the three scales is related to com-petence beliefs, a second-order factor was introduced leading to amodel with an acceptable fit (model 3dimgf corr; Fig. 2). To summarize,this model was chosen as the final measurement model, because it re-flects the theoretical nature of self-concept and self-efficacy as compo-nents of competence beliefs.

Furthermore, the latent correlations among the three scales of com-petence beliefs showed low values for the relationships between gener-al academic self-concept and chemistry self-concept, and betweengeneral academic self-concept and chemistry self-efficacy. In contrast,a high correlation was found for chemistry self-efficacy and chemistryself-concept (see Table 4).

3.3. Investigating domain and construct specificity

Based on the results on the structure of competence beliefs, the ef-fects of domain and construct specificity were analyzed by using twonested models (see Fig. 3). In the first model, a general factor of stu-dents' competence beliefs and two correlated factors of self-conceptand self-efficacy in chemistry were assumed (nested model 1, Fig. 2a).This model takes into account the domain of chemistry as a combiningelement between self-concept and self-efficacy and controls forstudents' general competence beliefs. Regarding the goodness-of-fit sta-tistics, the resulting model revealed an acceptable and substantialgoodness-of-fit and could, therefore, be accepted (SB-χ2(343, N =459) = 838.1, p b .001, CFI = .92, RMSEA = .05, CI90 = (.05, .06),SRMR = .05). The correlation between chemistry self-concept andself-efficacy showed a high value of ρ = .78, indicating that both con-structs affect each other.

Following the methodological approach proposed by Brunner(2008), domain specificity was investigated by using a nested factormodel with a general factor of students' competence beliefs and thetwo self-concept factors (nested model 2, Fig. 2b). By applying thismodel, the relationship between the self-concepts can be estimatedafter controlling for the second-order factor. Again, the resultingmodel showed a substantial and acceptable goodness-of-fit and was,thus, accepted (SB-χ2(358, N = 459) = 854.1, p b .001, CFI = .92,RMSEA = .05, CI90 = (.05, .06), SRMR = .05). The latent correlationbetween general academic and chemistry self-concept revealed no sig-nificant linear relationship among these two factors.

Regarding the amount of variance explained by the general factor ofcompetence beliefs in these two models, it was found that in nestedmodel 1 less variance could be explained. In this model, the mean com-munality of items was M = .14 (SD = .18, Min = .01, Max = .67),whereas model 2 showed a higher mean (M = .31, SD = .14, Min =.06, Max = .51). This difference was statistically significant with amedium effect (t(28) = 3.41, p b .01, d = .63). It is noted that thetwo models with a nested factor did not contain three traits for reasonsof model identification (Eid, 2000).

3.4. Effects of covariates

In order to obtain evidence on external validity of the compe-tence belief scales, latent regression analyses were performed byusing general academic self-concept, chemistry self-concept, andchemistry self-efficacy as dependent variables. The resulting modelswere run separately and revealed different predictors, as shown inTable 5.

Regarding students' general academic self-concept, grades in differ-ent subjects such as science, math, and German were significant

Fig. 2. Structure of competence beliefs with three traits (3dimgf corr). Note. ACSC = academic self-concept, SCCHEM = chemistry self-concept, SECHEM = chemistry self-efficacy,***p b .001.

14 R. Scherer / Learning and Individual Differences 28 (2013) 9–19

predictors. Also, students' value of science showed strong effects on theoutcome in self-concept. In this regressionmodel, 37% of variance couldbe explained.

For the chemistry self-concept scale, which can be regarded as adomain-specific component of competence beliefs, grades in chemistry,students' enjoyment in science, and hands-on activities in their chemis-try classrooms revealed significant effects, yielding 40% of variance ex-planation. These predictors were also found for the construct ofchemistry self-efficacy. Additionally, there was a moderate gender ef-fect on this scale, favoring males. The resulting model explained 47%of variance in self-efficacy. Interestingly, students' grade in chemistrywas a substantial predictor across all competence belief scales.

4. Discussion

The present study provided evidence on the empirical distinctionbetween self-concept and self-efficacy in chemistry and revealed thatthere are effects of domain and construct specificity on students' com-petence beliefs. Furthermore, it exemplarily showedhowanassessmentof chemistry self-efficacy could be based on curricular demands. In thissection of the paper, the results are discussed in light of previousresearch.

4.1. Test development

This study was aimed to develop a test on students' self-efficacyin chemistry, which was based on a curricular definition of compe-tences. By referring to four major components of chemistry compe-tence (e.g., Köller & Parchmann, 2012), items were developed asstatements of specific skills and task requirements. This designfeature could be used to distinguish between self-concept andself-efficacy because the first refers to global perceptions and

Table 4Latent correlations among the three competence belief scales.

Academicself-concept

Chemistryself-concept

Chemistryself-efficacy

Academic self-concept 1.00 .40⁎⁎⁎ .42⁎⁎⁎

Chemistry self-concept – 1.00 .82⁎⁎⁎

Chemistry self-efficacy – – 1.00

⁎⁎⁎ p b .001.

statements of competences within a subject or domain (Marsh,1990). The procedure followed the conceptualization of self-efficacy which was proposed by various researchers (Gallagher,2012; Marsh et al., 1991; Van Dinther et al., 2011). Using a differ-ent framework of assessing self-efficacy might also be appropriatebut could lead to slightly different results (Bong & Skaalvik,2003). However, this study shows how such assessments could bedeveloped and, furthermore, be used to validate performance as-sessments which specifically refer to these standards (Köller &Parchmann, 2012). This might result in providing students andteachers with assessments that capture their perceptions on com-petences as tools to foster monitoring and evaluation skills in class-rooms (Bong & Skaalvik, 2003; Liu, 2010). The resulting measureprovides specific feedback for students and teachers regarding spe-cific competences (e.g., Lin & Tsai, 2013).

The newly developed scale of self-efficacy in chemistry showed anacceptable internal consistency. However, compared to the self-concept scales, the value of Cronbach's α was slightly lower although ahigher number of items have been administered. This finding indicatesthat the newly developed scale shows heterogeneity for some of theitems. As Liu (2010) discussed, assessing students' self-perceptions ofmany different competences ismore difficult than assessing their globalcompetence beliefs. Furthermore, the PISA 2006 scale on science self-efficacy, which consisted of 8 items referring to competences in differ-ent contents of natural and environmental sciences, also showed alower value of internal consistency (α = .82) than the scientific self-concept scale (α = .90). It might be argued that the construct of self-efficacy itself is more heterogeneous than students' self-concepts(Bong & Skaalvik, 2003).

Additionally and in light of existingdiscussions on test efficiency, it isdesirable to develop a shorter version of the existing scale (Liu, 2010).But due to the broad framework of the German standards of chemistryeducation which differentiates four competences, sub-competences,and concepts, the self-efficacy scale could not be shortened in this study.

To summarize, it was possible to develop a new measure onchemistry-specific self-efficacy which sufficiently met the criterion ofreliability and content validity.

4.2. Evidence on construct validity

In order to obtain evidence on construct validity of the test on stu-dents' competence beliefs, different psychometric analyses wereconducted. First, the internal structure was investigated and the empir-ical distinction between self-concept and self-efficacy was checked.

Fig. 3. Nested factor models referring to two different measurement perspectives. Note. ACSC = academic self-concept, SCSCIE = chemistry self-concept, SESCIE = chemistry self-efficacy, ns = statistically insignificant, ***p b .001.

15R. Scherer / Learning and Individual Differences 28 (2013) 9–19

Second, the effects of domain and construct specificity were modeledwithin a nested factor model and, third, the relationships among com-petence beliefs and covariates were taken into account. These aspectsrefer to Messick's (1995) concept of how to obtain evidence on con-struct validity.

4.2.1. Internal structure of competence beliefsConcerning the structure of students' competence beliefs, it was

found that three factors could be distinguished: (1) general academicself-concept, (2) chemistry self-concept, and (3) chemistry self-efficacy. This finding confirmed the multidimensional structure ofcompetence beliefs. Following the argumentation of Marsh (1990)and Marsh et al. (1991), the hypothesis is apparent that self-concept ishierarchically ordered and multidimensional as well. The proposed

Table 5Results of latent regression analyses with three scales of competence beliefs as dependent vari

Academic self-concept Chemistry self-concept

Predictors β (SE) p Predictors β (SE

GENSCIE .23 (.06) .000 JOYSCIE .38Grade Bio − .13 (.05) .015 SCHANDS .18Grade Che − .15 (.06) .008 Grade Che − .35Grade Ger − .20 (.06) .000Grade Mat − .20 (.06) .000R2 .37 (.05) .000 R2 .40

Note. Bio = biology, Che = chemistry, Ger = German, GENSCIE = general value of science,Gender was dichotomously coded as 0 = males and 1 = females.

study supports this hypothesis because academic and chemistry self-concept were empirically distinguishable. In light of the resulting latentcorrelations, self-concept and self-efficacy can also be regarded as dis-tinct constructs. The proposed measurement model, thus, confirms thefindings of Ferla et al. (2009) as well as Peterson and Whiteman(2007)whowere able to show this relationship for the domain of math.

As the model with correlated factors was empirically not distinctfrom the second-order model, they showed the same goodness-of-fitindexes (Brown, 2006). However, these models represent differenttheoretical assumptions on the structure of the construct: In the corre-latedmodel, the three scaleswere assumed to be correlated. In contrast,the second-order model contained a general trait which underlay thethree traits. Besides this methodological discussion and from a theoret-ical point of view, a common factor of competence beliefs was also

ables.

Chemistry self-efficacy

) p Predictors β (SE) p

(.05) .000 JOYSCIE .46 (.05) .000(.05) .001 SCHANDS .23 (.06) .000(.05) .000 Grade Che − .27 (.06) .000

Gender − .12 (.04) .003

(.04) .000 R2 .47 (.05) .000

SCHANDS = hands-on activities in chemistry lessons, JOYSCIE = enjoyment in science.

16 R. Scherer / Learning and Individual Differences 28 (2013) 9–19

reasonable, as the existence of a higher-order factor indicated that thereis variance which could be explained by a common construct beingshared (Brown, 2006; Brunner, Nagy, & Wilhelm, 2012). In the contextof self-efficacy and self-concepts, this construct could be interpretedas students' competence beliefs which are omnipresent in various do-mains and subjects (Brunner et al., 2010; Marsh & Martin, 2011) or,more precisely, these refer to students' perceptions about their skillsand processes of knowledge acquisition in the academic environment(Bandura, 1997). As the higher-order structure of competence beliefswas empirically confirmed and theoretically reasonable, the presentstudy supports the findings of Lee and Stankov (2013), which indicatedthat self-concept and self-efficacywere parts of a second-order factor inmath.

Furthermore, the nested model of students' self-concepts wassupported in different ways (Brunner et al., 2009): First, it was possibleto extend the nested structure to competence beliefs which contain afactor of self-efficacy as well. Second, the proposed results supportedthe empirical distinction between general academic and chemistry-specific self-concepts. Third, the specificity of competence beliefs inchemistry has been shown.

4.2.2. Domain and construct specificity of competence beliefsBased on the modeling approaches which were conducted by

Brunner et al. (2009), Marsh (1990), and Eid (2000), two differentnested models have been specified in order to analyze the domain andconstruct specificity of the proposed three factors. The resulting modelsrevealed substantial goodness-of-fit and were accepted.

After controlling for the general factor of students' competence be-liefs within the first nested model, the relationship between chemistryself-concept and self-efficacy was still very high (ρ = .78). This indi-cates that both constructs do not only share common variance in a strictstatistical sense, but are also affected by another factor, which cannot beexplained by the general factor of competence beliefs. More precisely,both constructs are highly related. In this context, the author interpretsthis relationship as an effect of construct specificitywithin a domain andconcludes that Brunner's model of self-concept (Brunner et al., 2009)could be extended to this modeling situation, in which self-conceptand self-efficacy are involved. In light of referring to the same domain,which is chemistry in this case, self-concept and self-efficacy arehighly related, but empirically distinct, as they share approximately61% of variance. Taking into account the theoretical definitions andoperationalizations; efficacy describes students' beliefs on specificcompetences in science lessons, whereas self-concept refers to a moregeneral description of competences (Bong & Skaalvik, 2003). Thus,they are highly related but not the same.

In the second nestedmodel, the general factor of competence beliefswas controlled and two self-concept dimensions were introduced tothis model. In this situation, the modeling approach primarily focusedon domain specificity and aimed to answer the question of whether ornot there is evidence for an empirical distinction between generalacademic and chemistry self-concept. The results suggest that the rela-tionship between both constructs disappears if the general factor ispartialled out. Thus, academic and chemistry self-concept can also beregarded as empirically distinct constructs. By referring to Marsh(1990) and Marsh and Martin (2011), it can be concluded that there isevidence on the domain specificity of academic self-concepts.

These results do not only contribute to knowledge on how tomeasure or specify students' self-concepts and self-efficacy. They alsoshow that, besides global beliefs of one's competence within a domainor, more generally, within the school environment, specific beliefs oncurriculum-based competences are also important in order to fosterstudents' self-concept (Tsai et al., 2011).

The explanation of variance for the second-order competence belieffactor indicated significant higher communalities for the second model(nested model 2). In nested model 1, both constructs share a largeamount of variance which cannot be explained by the nested factor. In

this context, this finding indicates that there might be a further factorwhich explains the high correlation amongboth constructs beyond gen-eral competence beliefs. In contrast, almost the entire relationship be-tween academic and chemistry self-concept could be explained by thegeneral factor, supporting the previous argumentation. This result canbe regarded as further evidence on domain specificity.

Additionally, the results obtained in the present study replicated andextended the findings of Ferla et al. (2009) as well as Lee (2009) on thedistinction between self-concept and self-efficacy for the domain ofchemistry. In the present research, a more complex approach was cho-sen in order to investigate the effects of domain and construct specificityaswell. However, this study does not provide evidence on domain spec-ificity in further domains such as biology, physics, or earth science. Sub-sequent research could, therefore, systematically address this issue infurther academic domains (e.g., Marsh & Scalas, 2010; Marsh et al.,1991).

Besides the methodological contribution of this study, the empiricaldistinction between domain-specific and general academic self-concepts has also practical implications. For instance, if teachers intendto foster students' self-perceptions they should not only refer to stu-dents' general convictions in the academic environment but also devel-op appropriate evaluations within a specific subject (Brunner et al.,2009). This approach could subsequently lead to a better understandingof one's own capabilities on a meta-cognitive level which positively af-fects performance goals and the actual school achievement (Huang,2011; Marsh & Martin, 2011). Furthermore, test developers shouldtake into account the distinctive but related nature of self-efficacy andself-concept when developing surveys within specific contexts, do-mains, and subjects (Arens et al., 2011; Bong & Skaalvik, 2003). Conse-quently, such assessments could be beneficial in order to developskills of monitoring and evaluating one's competences within a domain.These skills are ofmajor importance, especially as predictors of academ-ic achievement (Ferla et al., 2009; Lee & Stankov, 2013; Pajares &Miller,1994; Pietsch et al., 2003).

4.2.3. External validityIn this study, latent regression modeling was applied in order to

check for external validity. The resulting models revealed significantpredictors of students' competence beliefs with substantial variance ex-planations and obtained evidence on construct validity of the newly de-veloped test on chemistry self-efficacy.

4.2.3.1. School achievement. As expected, therewere significant effects ofstudents' grades in different school subjects on general academic self-concept. This finding is in line with previous research on self-conceptand validates the measure (e.g., Brunner et al., 2009; Marsh, 1990;Marsh et al., 1991). Interestingly, students' grade in chemistry was theonly achievement-based predictor of the domain-specific scales ofself-concept and self-efficacy with significant regression coefficients.This result provides further evidence of the domain specificity. Further-more, it extends the argumentation of Marsh et al. (1991) as well asBrunner et al. (2009) who were able to show similar effects for otherdomains except for chemistry. In light of discussions on the importanceof students' grades within a domain on their ability perceptions, thepresent results underpin the substantial relationships among compe-tence beliefs and school achievement (Marsh & Martin, 2011; Pietschet al., 2003).

4.2.3.2. Motivational covariates and the general value of science. In thisstudy, students' enjoyment in science revealed effects on chemistryself-concept and self-efficacy but not on their general academic self-concept. These findings support previous research on the influence ofscience motivation on personality related constructs (e.g., Glynn et al.,2011; Lewis et al., 2009). For instance, Spinath and Steinmayr (2012)pointed out that competence beliefs have a mediating role withinthe relationship between performance and intrinsic motivation.

17R. Scherer / Learning and Individual Differences 28 (2013) 9–19

Consequently, these constructs provide substantial information on per-formance and motivation (Bong et al., 2012).

The effects of students' general value of science on their competencebeliefs supported the importance of this construct for the academic self-concept (Mason et al., 2013). Urhahne et al. (2011) argued that a betterunderstanding of the nature of science would lead to a better perfor-mance in assessments of scientific literacy and, consequently, yieldhigher values of competence beliefs. This argumentation was mainlybased on the influences of performance on self-concepts but did nottake into account reciprocal effects (Marsh, Trautwein, Lüdtke, Köller,& Baumert, 2005). Unexpectedly, only students' general values in sci-ence were significantly correlated with their general academic self-concept but not with chemistry-specific self-concept and self-efficacy.This findingmight be due to the generalizability of the nature of scienceconcept across the different science subjects (e.g., Schwartz &Lederman, 2008; Urhahne et al., 2011), showing no specific results forchemistry. However, further studies are necessary which systematicallyaddress this issue, first, by using amore differentiated framework of thenature of science and, second, by investigating causes and effects of thisrelationship.

These findings provide practical implications for chemistry class-rooms, as they show the importance of developing an appropriatevalue of science and enjoyment in this subject. Consequently, fosteringstudents' views on and interest in science could determine their compe-tence beliefs and, subsequently, their performance in specific tasks(Bandura, 1997; Spinath & Steinmayr, 2012).

4.2.3.3. Classroom-based activities. Students' reports on how often theyconduct scientific experiments in chemistry revealed significant ef-fects on students' competence beliefs except for the general academ-ic self-concept. On the one hand, this result can be interpreted asfurther evidence on domain specificity, since the SCHANDS scalewas operationalized for chemistry. As Hofstein and Lunetta (2004)argued, the implementation of inquiry-based classroom activitiesin science has been quite rare. In light of the reciprocal effectsmodel and the effects of hands-on activities on students' science at-titudes, this could lead to a lower self-concept and self-efficacy inchemistry (e.g., Freedman, 1997; Marsh et al., 2005). In the proposedstudy, it was found that incorporating experiments and hands-on ac-tivities in science classrooms is positively related to students' self-concept and self-efficacy in chemistry. Unfortunately, the methodol-ogy applied in this study does not allow for causal interpretations butobtains evidence on the existence of this relationship. In future as-sessments, it would be interesting to examine these effects moreprecisely. However, these findings support the theoretical frame-work of Chen and Usher (2013) which incorporated specific sourcesof self-efficacy, which takes into account classroom-based factors.Consequently, chemistry-specific competence beliefs are also deter-mined by inquiry-based activities in the classroom. In light of this re-sult, it appears reasonable that, at least to some degree, teacherscould contribute to the development of students' competence beliefsin chemistry classrooms.

4.2.3.4. Gender effects. The effect of gender on chemistry self-efficacysupported previous findings from Glynn et al. (2011) who analyzed sci-ence and nonsciencemajors' motivation and self-efficacy in science andfound a gender effect favoring boys. This result obtains further evidenceon the scale's construct validity and indicates that, after controlling forstudents' performance in chemistry, male students perceive themselvesasmore competent than female students (Ferla et al., 2009). In contrast,Chen and Usher (2013), and Louis and Mistele (2012) found that pro-files of science self-efficacy were not influenced by gender. In order toexplain this contradiction, future research should model gender differ-ences by using measurement models which provide more detailed in-formation on construct changes and measurement bias across gendergroups (e.g., Velayutham et al., 2012).

Taken together, the results on the effects of different covariatesprovide evidence on construct validity indicating domain and con-struct specificity, as there were different groups of predictorsacross constructs but similar predictors within a domain. Addition-ally, the amount of variance explained by these factors was consid-erably high. In this context, the present study also providedevidence on the factors of students' competence beliefs in chemis-try. These factors also showed that classroom-based factors alsoplay an important role. Furthermore, the regression models ofchemistry-specific self-concept and self-efficacy were quite similarwhich was also indicated by the high correlation of these con-structs. Again, an interpretation of this finding refers to domainspecificity. However, another explanation takes into account thatboth represent students' perceptions of their chemistry-relatedcompetences (Lin & Tsai, 2013).

However, as research on personality traits, students' career motiva-tion, and retention suggests (Glynn et al., 2009; Jonkmann et al., 2012;Peters, 2013; Sawtelle, Brewe, & Kramer, 2012), further factors whichare related to personality traits and the impact of the Big-Fish-Little-Pond-Effect as an example of a classroom-based factor could be takeninto account in future studies.

Moreover, in the proposed study, cause–effect relationshipsbetween competence beliefs and school achievement could not beinvestigated, as the cross-sectional design did not allow for causalconclusions. Hence, future research could systematically address theserelationships by conducting studies with a longitudinal and cross-lagged design. Furthermore, it would be interesting to analyze thesources of students' competence beliefs (Chen & Usher, 2013; Usher &Pajares, 2009). However, the present study did not aim to answer thequestion of causes and effects in competence beliefs. Instead, the rela-tionships among the different scales of self-concept, self-efficacy, andcovariates were interpreted as evidence for external validity (Messick,1995).

4.3. Conclusion

The results of this study indicate that there is evidence for theempirical distinction between self-concept and self-efficacy withinthe domain of chemistry. Supporting previous works of Ferla et al.(2009) and Lee (2009) for math and the argumentation of Bongand Skaalvik (2003), this finding shows that these personalitytraits are conceptually and empirically distinct. Taking into ac-count previous discussions of Brunner et al. (2009) on whetheror not domain-specific components of competence beliefs repre-sent different factors than domain-general components, the em-pirical distinction can be regarded as evidence of domainspecificity which was also proposed by different researchers(Bong, 2001; Marsh, 1990; Marsh et al., 1991). The present studycontributed to an extension of the nested Marsh/Shavelsonmodel (e.g., Brunner et al., 2010) to competence beliefs in the do-main of chemistry. Additionally, the proposed methodological ap-proach of analyzing domain specificity could be transferred toother constructs and domains (Brunner et al., 2009).

This study also provided an assessment tool of students' com-petence beliefs which is based on specific competences in chemis-try. It could, consequently, be used to for individual feedback andself-evaluation in lessons which could, subsequently, lead tohigher and more precise self-perceptions (Bong et al., 2012; VanDinther et al., 2011). According to Lee and Stankov (2013) andLiu (2010), these could yield higher values in students' perfor-mances on specific chemistry tasks and standardized assessments.Together with Huang (2011), the author emphasizes the impor-tance of interventions which systematically combine the enhance-ment of self-perceptions and the development of competences inclassrooms.

18 R. Scherer / Learning and Individual Differences 28 (2013) 9–19

Appendix A

References

Arens, A., Craven, R., Yeung, A., & Hasselhorn, M. (2011). The twofold multidimen-sionality of academic self-concept: Domain specificity and separation betweencompetence and affect components. Journal of Educational Psychology, 103(4),970–981. http://dx.doi.org/10.1037/a0025047.

Bandura, A. (1997). Self-efficacy: The exercise of control. New York, NY: Freeman.Bong,M. (2001). Between- andwithin-domain relations of academicmotivation amongmid-

dle and high school students: Self-efficacy, task-value, and achievement goals. Journal ofEducational Psychology, 93(1), 23–34. http://dx.doi.org/10.1037//0022-0663.93.1.23.

Bong, M., Cho, C., Ahn, H. -S., & Kim, H. (2012). Comparison of self-beliefs for predictingstudent motivation and achievement. The Journal of Educational Research, 105(5),336–352. http://dx.doi.org/10.1080/00220671.2011.627401.

Bong, M., & Skaalvik, E. (2003). Academic self-concept and self-efficacy: How different arethey really? Educational Psychology Review, 15(1), 1–40. http://dx.doi.org/10.1023/A:1021302408382.

Brown, T. (2006). Confirmatory factor analysis for applied research. New York, NY: TheGuilford Press.

Bruning, R., Dempsey, M., Kauffman, D. F., & McKim, C. (2013). Examining dimen-sions of self-efficacy for writing. Journal of Educational Psychology, 105(1),25–38. http://dx.doi.org/10.1037/a0029692.

Brunner, M. (2008). No g in education? Learning and Individual Differences, 18, 152–165.http://dx.doi.org/10.1016/j.lindif.2007.08.005.

Brunner, M., Keller, U., Dierendonck, C., Reichert, M., Ugen, S., Fischbach, A., et al. (2010).The structure of academic self-concepts revisited: The nested Marsh/Shavelsonmodel. Journal of Educational Psychology, 102(4), 964–981. http://dx.doi.org/10.1037/a0019644.

Brunner, M., Keller, U., Hornung, C., Reichert, M., & Martin, R. (2009). The cross-culturalgeneralizability of a new structural model of academic self-concepts. Learning andIndividual Differences, 19(4), 387–403. http://dx.doi.org/10.1016/j.lindif.2008.11.008.

Brunner, M., Nagy, G., & Wilhelm, O. (2012). A tutorial on hierarchically structuredconstructs. Journal of Personality, 80(4), 796–846. http://dx.doi.org/10.1111/j.1467-6494.2011.00749.x.

Bryant, F., & Satorra, A. (2012). Principles and practice of scaled difference chi-squaretesting. Structural Equation Modeling, 19(3), 372–398. http://dx.doi.org/10.1080/10705511.2012.687671.

Chen, J., & Usher, E. (2013). Profiles of the sources of science self-efficacy. Learningand Individual Differences, 24, 11–21. http://dx.doi.org/10.1016/j.lindif.2012.11.002.

Chen, S. -K., Yeh, Y. -C., Hwang, F. -M., & Lin, S. (2013). The relationship between academicself-concept and achievement: A multicohort–multioccasion study. Learning andIndividual Differences, 23, 172–178. http://dx.doi.org/10.1016/j.lindif.2012.07.021.

Eid, M. (2000). A multitrait–multimethod model with minimal assumptions.Psychometrika, 65(2), 241–261. http://dx.doi.org/10.1007/BF02294377.

Enders, C. (2010). Applied missing data analysis. New York, NY: The Guilford Press.Ferla, J., Valcke, M., & Cai, Y. (2009). Academic self-efficacy and academic self-concept:

Reconsidering structural relationships. Learning and Individual Differences, 19(4),499–505. http://dx.doi.org/10.1016/j.lindif.2009.05.004.

Freedman,M. (1997). Relationship among laboratory instruction, attitude toward science,and achievement in science knowledge. Journal of Research in Science Teaching, 34(4),343–357. http://dx.doi.org/10.1002/(SICI)1098-2736(199704)34:4b343::AID-TEA5N3.0.CO;2-R.

Freiberger, V., Steinmayr, R., & Spinath, B. (2012). Competence beliefs and perceivedability evaluations: How do they contribute to intrinsicmotivation and achievement?Learning and Individual Differences, 22(4), 518–522. http://dx.doi.org/10.1016/j.lindif.2012.02.004.

Frey, A., Taskinen, P., Schütte, K., Prenzel, M., Artelt, C., Baumert, J., Blum,W., Hammann, M.,Klieme, E., & Pekrun, R. (Eds.). (2009). PISA 2006 Skalenhandbuch. Münster: Waxmann.

Gallagher, M. (2012). Self-efficacy. In V. Ramachandran (Ed.), Encyclopedia of humanbehavior, Vol. 2. (pp. 314–320). Oxford, UK: Elsevier.

Glynn, S., Brickman, P., Armstrong,N., & Taasoobshirazi, G. (2011). Sciencemotivation ques-tionnaire II: Validation with science majors and nonscience majors. Journal of Researchin Science Teaching, 48(10), 1159–1176. http://dx.doi.org/10.1002/tea.20442.

Glynn, S., Taasoobshirazi, G., & Brickman, P. (2009). Science motivation questionnaire:Construct validation with nonscience majors. Journal of Research in Science Teaching,46(2), 127–146. http://dx.doi.org/10.1002/tea.20267.

Hoffman, B., & Schraw, G. (2009). The influence of self-efficacy and working memory ca-pacity on problem-solving efficacy. Learning and Individual Differences, 19(1), 91–100.http://dx.doi.org/10.1016/j.lindif.2008.08.001.

Hofstein, A., & Lunetta, V. (2004). The laboratory in science education: Foundations forthe twentyfirst century. Science Education, 88, 28–54. http://dx.doi.org/10.1002/sce.10106.

19R. Scherer / Learning and Individual Differences 28 (2013) 9–19

Hu, L. -T., & Bentler, P. (1999). Cutoff criteria for fit indexes in covariance structure anal-ysis: Conventional criteria versus new alternatives. Structural EquationModeling, 6(1),1–55. http://dx.doi.org/10.1080/10705519909540118.

Huang, C. (2011). Self-concept and academic achievement: A meta-analysis of longitudinalrelations. Journal of School Psychology, 49, 505–528. http://dx.doi.org/10.1016/j.jsp.2011.07.001.

IBM (2010). SPSS statistics 19. Armonk, NY: IBM.Jonkmann, K., Becker, M., Marsh, H., Lüdtke, O., & Trautwein, U. (2012). Personality traits

moderate the Big-Fish-Little-Pond-Effect of academic self-concept. Learning andIndividual Differences, 22(6), 736–746. http://dx.doi.org/10.1016/j.lindif.2012.07.020.

Köller, O., & Parchmann, I. (2012). Competences: The German notion of learning out-comes. In S. Bernholt, K. Neumann, & P. Nentwig (Eds.),Making it tangible — Learningoutcomes in science education (pp. 165–185). Münster: Waxmann.

Komarraju, M., & Nadler, D. (2013). Self-efficacy and academic achievement: Why do im-plicit beliefs, goals, and effort regulation matter? Learning and Individual Differences,25, 67–72. http://dx.doi.org/10.1016/j.lindif.2013.01.005.

Lee, J. (2009). Universals and specifics of math self-concept, math self-efficacy, and mathanxiety across 41 PISA 2003 participating countries. Learning and IndividualDifferences, 19, 355–365. http://dx.doi.org/10.1016/j.lindif.2008.10.009.

Lee, J., & Stankov, L. (2013). Higher-order structure of noncognitive constructs and predic-tion of PISA 2003 mathematics achievement. Learning and Individual Differences, 26,119–130. http://dx.doi.org/10.1016/j.lindif.2013.05.004.

Lewis, S., Shaw, J., Heitz, J., & Webster, G. (2009). Attitude counts: Self-concept andsuccess in general chemistry. Journal of Chemical Education, 86(6), 744–749.http://dx.doi.org/10.1021/ed086p744.

Lin, T. -J., & Tsai, C. -C. (2013). A multi-dimensional instrument for evaluatingTaiwanese high school students' science learning self-efficacy in relation totheir approaches to learning science. International Journal of Science andMathematics Education. http://dx.doi.org/10.1007/s10763-012-9376-6.

Little, R. (1988). A test of missing completely at random for multivariate data withmissing values. Journal of the American Statistical Association, 83(404), 1198–1202.http://dx.doi.org/10.1080/01621459.1988.10478722.

Liu, X. (2010). Using and developing measurement instruments in science education. Char-lotte, NC: Information Age Publishing.

Louis, R., & Mistele, J. (2012). The differences in scores and self-efficacy by student genderin mathematics and science. International Journal of Science and MathematicsEducation, 10(5), 1163–1190. http://dx.doi.org/10.1007/s10763-011-9325-9.

Marsh, H. (1990). The structure of academic self-concept: The Marsh/Shavelsonmodel. Journal of Educational Psychology, 82(4), 623–636. http://dx.doi.org/10.1037/0022-0663.82.4.623.

Marsh, H., Abu-Hilal, F., Xu, M., Abduljabbar, A., Morin, A., Leung, K., et al. (2013). Factorial,convergent, and discriminant validity of TIMSS math and science motivation mea-sures: A comparison of Arab and Anglo-Saxon countries. Journal of EducationalPsychology, 105(1), 108–128. http://dx.doi.org/10.1037/a0029907.

Marsh, H., & Martin, A. (2011). Academic self-concept and academic achievement:Relations and causal ordering. British Journal of Educational Psychology, 81,59–77. http://dx.doi.org/10.1348/000709910X503501.

Marsh,H., & Scalas, L. (2010). Self-concept in learning: Reciprocal effectsmodel between ac-ademic self-concept and academic achievement. In P. Peterson, E. Baker, & B. McGaw(Eds.), International encyclopedia of education (3rd ed.), Vol. 6. (pp. 660–667). Oxford,UK: Elsevier.

Marsh, H., Trautwein, U., Lüdtke, O., Köller, O., & Baumert, J. (2005). Academicself-concept, interest, grades, and standardized test scores: Reciprocal effects modelsof causal ordering. Child Development, 76, 397–416. http://dx.doi.org/10.1111/j.1467-8624.2005.00853.x.

Marsh, H., Walker, R., & Debus, R. (1991). Subject-specific components of academicself-concept and self-efficacy. Contemporary Educational Psychology, 16(4), 331–345.http://dx.doi.org/10.1016/0361-476X(91)90013-B.

Mason, L., Boscolo, P., Tornatora, M., & Ronconi, L. (2013). Besides knowledge: Across-sectional study on the relations between epistemic beliefs, achievementgoals, self-beliefs, and achievement in science. Instructional Science, 41, 49–79.http://dx.doi.org/10.1007/s11251-012-9210-0.

Messick, S. (1995). Validity of psychological assessment: Validation of inferences frompersons' responses and performances as scientific inquiry into score meaning.American Psychologist, 50(9), 741–749. http://dx.doi.org/10.1037/0003-066X.50.9.741.

Muis, K., Bendixen, L., & Härle, F. (2006). Domain-generality anddomain-specificity in per-sonal epistemology research. Educational Research Review, 18, 3–54. http://dx.doi.org/10.1007/s10648-006-9003-6.

Muthén, B., & Muthén, L. (2010). Mplus 6. Los Angeles, CA: Muthén & Muthén.Neumann, K., Fischer, H., & Kauertz, A. (2010). From PISA to educational standards: The

impact of large-scale assessments on science education in Germany. InternationalJournal of Science and Mathematics Education, 8(3), 545–563. http://dx.doi.org/10.1007/s10763-010-9206-7.

Niehaus, K., & Adelson, J. (2013). Self-concept and native language background: A study ofmeasurement invariance and cross-group comparisons in third grade. Journal ofEducational Psychology, 105(1), 226–240. http://dx.doi.org/10.1037/a0030556.

Organisation for Economic Co-operation and Development (OECD) (2009). PISA2006 technical report. Paris: OECD Publications. http://dx.doi.org/10.1787/9789264048096-en.

Pajares, F., & Miller, M. (1994). Role of self-efficacy and self-concept beliefs in mathemat-ical problem solving: A path analysis. Journal of Educational Psychology, 86(2),193–203. http://dx.doi.org/10.1037/0022-0663.86.2.193.

Peters,M. (2013). Examining the relationships among classroomclimate, self-efficacy, andachievement in undergraduate mathematics: A multi-level analysis. InternationalJournal of Science and Mathematics Education, 11(2), 459–480. http://dx.doi.org/10.1007/s10763-012-9347-y.

Peterson, E., &Whiteman, M. (2007). “I think I can, I think I can…”: The interrelationshipsamong self-assessed intelligence, self-concept, self-efficacy and the personality traitintellect in university students in Scotland and New Zealand. Personality andIndividual Differences, 43(4), 959–968. http://dx.doi.org/10.1016/j.paid.2007.02.019.

Pietsch, J., Walker, R., & Chapman, E. (2003). The relationship among self-concept,self-efficacy, and performance in mathematics during secondary school. Journal ofEducational Psychology, 95(3), 589–603. http://dx.doi.org/10.1037/0022-0663.95.3.589.

Sawtelle, V., Brewe, E., & Kramer, L. (2012). Exploring the relationship betweenself-efficacy and retention in introductory physics. Journal of Research in ScienceTeaching, 49(9), 1096–1121. http://dx.doi.org/10.1002/tea.21050.

Schwartz, R., & Lederman, N. (2008).What scientists say: Scientists' views of nature of sci-ence and relation to science context. International Journal of Science Education, 30(6),727–771. http://dx.doi.org/10.1080/09500690701225801.

Spinath, B., & Steinmayr, R. (2012). The roles of competence beliefs and goal orientationsfor change in intrinsic motivation. Journal of Educational Psychology, 104(4),1135–1148. http://dx.doi.org/10.1037/a0028115.

Tsai, C. -C., Ho, H., Liang, J. -C., & Lin, H. -M. (2011). Scientific epistemic beliefs, conceptions oflearning science and self-efficacy of learning science among high school students.Learning and Instruction, 21(6), 757–769. http://dx.doi.org/10.1016/j.learninstruc.2011.05.002.

Urhahne, D., Kremer, K., & Mayer, J. (2011). Conceptions of the nature of science — Arethey general or context specific? International Journal of Science and MathematicsEducation, 9, 707–730. http://dx.doi.org/10.1007/s10763-010-9233-4.

Usher, E. L., & Pajares, F. (2009). Sources of self-efficacy inmathematics: A validation study.Contemporary Educational Psychology, 34(1), 89–101. http://dx.doi.org/10.1016/j.cedpsych.2008.09.002.

Van Dinther, M., Dochy, F., & Segers, M. (2011). Factors affecting students' self-efficacy inhigher education. Educational Research Review, 6(2), 95–108. http://dx.doi.org/10.1016/j.edurev.2010.10.003.

Velayutham, S., Aldrige, J., & Fraser, B. (2012). Gender differences in student motivationand self-regulation in science learning: A multi-group structural equation modelinganalysis. International Journal of Science and Mathematics Education, 10(6),1347–1368. http://dx.doi.org/10.1007/s10763-012-9339-y.

Wagner, W., Göllner, R., Helmke, A., Trautwein, U., & Lüdtke, O. (2013). Construct validityof student perceptions of instructional quality is high, but not perfect: Dimensionalityand generalizability of domain-independent assessments. Learning and Instruction,28, 1–11. http://dx.doi.org/10.1016/j.learninstruc.2013.03.003.