Competence-Affect Separation on the Math and Verbal Scales of the Self-Description Questionnaire II in Chinese Sample

The Multidimensional Construct of Self-Concept

Self-concept is defined as an individual’s perception of self as a result of interactions and experiences with and within the environment (Shavelson, Hubner, & Stanton, 1976). It was considered as a global or unidimensional construct by early theorists. However, recent studies have supported the multidimensionality and domain specificity of self-concept instead (e.g., Leung, Marsh, Craven, Yeung, & Abduljabbar, 2013; Leung, Marsh, Yeung, & Abduljabbar, 2015; Marsh & Craven, 2006). Shavelson et al. (1976) proposed that self-concept is multi-faceted and hierarchical rather than unidimensional in nature. According to Shavelson et al.’s model, there is a general self-concept at the apex of this hierarchy. General self-concept is divided into academic and nonacademic self-concepts. Academic self-concept is further subdivided into different domains in particular subject areas, such as mathematics and verbal, whereas nonacademic self-concept is subdivided into different areas such as social, physical, and emotional domains.

Separation of the Competence-Affect Components of Academic Self-Concept

In Shavelson et al.’s (1976) model, self-concept consists of both self-evaluation and self-description components and they were not differentiated. Competency component of academic self-concept is evaluative in nature as it is concerned on a person’s ability in the academic work whereas affect component is descriptive in nature as it depicts student’s affective and motivational responses such as enjoyment in the schoolwork (Marsh, Craven, & Debus, 1999). Hence, self-concept consists of both competence and affect components.

Self-Description Questionnaire (SDQ) instruments, which were developed on the basis of Shavelson et al.’s (1976) model, contain items measuring both student’s competence and affect components. According to Shavelson et al.’s original work, student’s competence and affect components were not differentiated. However, Marsh et al. (1999) reported that competence and affect components within each specific subject of academic self-concept could be separated. Specifically, it demonstrated that confirmatory factor analysis (CFA) models positing the distinct separation between the competence-affect components for the different academic facets of self-concept (i.e., math competence, math affect, reading competence, and reading affect) had better model fit than models that posited only distinct facets of academic self-concept (i.e., math and reading) without the competence-affect distinction in a sample of elementary Australian students using the SDQ I. Similarly, Arens, Bodkin-Andrews, Craven, and Yeung (2014) revealed that Math and English self-concepts could be separated into competence and affect components in a sample of Australian high school students using the SDQ II. Thus, investigation of separation between the competence-affect components for the different academic domains of self-concept was important.

However, no studies have investigated the distinction of the competence and affect components for the Math and Verbal scales of the SDQ II in Chinese sample. Hence, the present investigation attempts to address the gap by using a CFA approach to test such distinction.

Method

Results and Discussion

Descriptive Statistics for the Chinese Sample

Table 1 shows the descriptive statistics, and results indicate that the coefficient alpha reliability estimates for each subscale in the Chinese version of the SDQ II were high and ranged from .90 to .95 (see Table 1).

OPEN IN VIEWER

Table 1.

Descriptive Statistics for the SDQ II.

Subscale	M	SD	Kurt	Skew	N	α	SEm
Mc	3.19	1.28	−0.70	0.08	480	.92	.08
Ma	3.71	1.39	−0.99	−0.12	480	.93	.06
Vc	3.19	1.19	−0.40	0.24	480	.93	.05
Va	3.52	1.25	−0.69	0.04	480	.90	.06
Math	3.34	1.23	−0.83	0.04	480	.94	.06
Verbal	3.29	1.14	−0.48	0.19	480	.95	.05
Total	3.31	0.83	0.23	0.15	480	.89	.04

Note. SDQ = Self-Description Questionnaire; Kurt = Kurtosis; Skew = Skewness; α = coefficient alpha estimate of reliability; SEm = standard error of the mean; Mc = math competence; Ma = math affect; Vc = verbal competence; Va = verbal affect; total = math and verbal.

The Factor Structure for Competence-Affect Separation on the Math and Verbal Scales

The four-factor model had better goodness-of-fit indices than the two-factor and one-factor models (see Table 2). The TLI, RNI, and CFI values indicate that the model had a good fit because these three scores were greater than .95. Also, the RMSEA was equal to .05, and the SRMR was smaller than 0.10, which indicates good model fit (Browne & Cudeck, 1993; Hu & Bentler, 1999; Kline, 2005).

OPEN IN VIEWER

Table 2.

Goodness-of-Fit Summary for Alternative Models of the SDQ II Subscales (Math and Verbal).

Model	χ2	df	TLI	RNI	CFI	RMSEA	SRMR
a.  Four-factor model	374.60	164	.985	.987	.990	.052 90% CI = [.045, .059]	.031
b.  Two-factor model	1,424.58	169	.912	.922	.940	.125 90% CI = [.120, .130]	.056
c.  One-factor model	8,962.56	170	.386	.450	.740	.329 90% CI = [.320, .330]	.30

Note. SDQ = Self-Description Questionnaire; TLI = Tucker–Lewis Index; RNI = Relative Noncentrality Index; CFI = Comparative Fit Index; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual; CI = confidence interval.

These results suggest that the four-factor model provides the best fit to the data. The target factor loadings were reasonable (most > .75) (see Table 3). The correlation between the competence and affect components were high but not perfectly correlated (Math competence and Math affect: r = .77; Verbal competence and Verbal affect: r = .82) (see Table 4). This suggests that the competence and affect components of the Math and Verbal scales formed four distinct factors. Hence, it provided clear support for the reliability and validity of the competence-affect separation of the Math and Verbal scales of the Chinese version of the SDQ II.

OPEN IN VIEWER

Table 3.

CFA Completely Standardized Solution for Four-Factor Model.

Indicator	Factor loadings for each subscale a
	Mc	Ma	Vc	Va
Item 1	.84	.88	.76	.87
Item 2	.80	.89	.86	.86
Item 3	.81	.89	.78	.87
Item 4	.80	.86	.87	—
Item 5	.84	—	.87	—
Item 6	.81	—	.77	—
Item 7	—	—	.78	—
Uniquenesses
Item 1	.29	.23	.42	.24
Item 2	.37	.21	.27	.25
Item 3	.34	.21	.38	.24
Item 4	.36	.26	.24	—
Item 5	.30	—	.24	—
Item 6	.35	—	.41	—
Item 7	—	—	.38	—

Note. CFA = confirmatory factor analysis; Mc = math competence; Ma = math affect; Vc = verbal competence; Va = verbal affect.

a

All factor loadings are statistically significant at p < .001.

OPEN IN VIEWER

Table 4.

CFA Correlations Among Math and Verbal Subscales of the SDQ II.

Subscale	Math competence	Math affect	Verbal competence	Verbal affect
Math competence	—
Math affect	.77***	—
Verbal competence	−.08	−.04	—
Verbal affect	.02	.01	.82***	—

Note. CFA = confirmatory factor analysis; SDQ = Self-Description Questionnaire.

***

p < .001.

The above results have important implications for self-concept research, theory, and practice. The present investigation advances our knowledge in self-concept research in Chinese culture as it is the first to examine the separation of competence-affect components for the Math and Verbal scales in a Chinese population. And these results imply that researchers can confidently adopt the Chinese versions of the SDQ II for Chinese samples regarding the competence-affect separation of Math and Verbal scales. Regarding the self-concept theory, it indicated that competence and affect components are not inseparable in Chinese culture as opposed to the assumption in Shavelson et al.’s (1976) original work.

From practical perspective, as mentioned in Marsh et al. (1999), the present investigation provides further evidence on the direction for self-enhancement interventions. Interventions that can target to promote both the competence and affect components (such as enjoying, liking) of Math and Verbal self-concept would be more effective than interventions targeting only one component.

Limitations and Future Research

A major limitation of the present study is that external criteria, such as academic achievement, were not adopted, which prevented the examination of the differential relationship of competence and affect on achievement as done in Arens et al.’s (2014) study. Hence, further research can use objective measures of achievement as external criteria to provide support for between-network construct validity as well as to evaluate whether a similar differential relation between competence and affect on achievement can be found in the Chinese population.