Psychometric Evaluation of the Chinese Version of Cognitive Style Questionnaire-Short Form Among University Students

Abstract

Objective

The Cognitive Style Questionnaire-Short Form (CSQ-SF) is a self-report measure of negative cognitive style. This study aimed to evaluate the psychometric properties of a Chinese version of the CSQ-SF among Chinese university students.

Methods

Following approval of the Chinese translation by an original author, 3525 university students completed the CSQ-SF along with the Center for Epidemiologic Studies Depression Scale (CES-D), State-Anxiety Inventory (SAI), Dysfunctional Attitude Scale (DAS), and Ruminative Response Scale (RRS). After a 5-week interval, 429 students completed the CSQ-SF twice. Confirmatory factor analysis tested a first-order five-factor model and a second-order single-factor model. Multi-group confirmatory factor analysis examined the measurement invariance across gender and time. Internal consistency and convergent validity were assessed.

Results

Both models of the CSQ-SF fit the data well. The first-order model showed measurement invariance across gender and time. Internal consistency (Cronbach’s α: 0.60–0.92, McDonald’s omega: 0.61–0.92, mean inter-item correlations: 0.15–0.41, split-half reliability: 0.58–0.83) and test-retest reliability (0.59–0.91) were acceptable The scores of CSQ-SF and its factors had significantly positive correlations with scores of CES-D, SAI, DAS and RRS (ps < 0.001).

Conclusion

The Chinese version of CSQ-SF demonstrates acceptable psychometric properties and can be used to assess cognitive vulnerability in Chinese-speaking university individuals.

Introduction

The Hopelessness Theory of Depression posits that a core vulnerability to depression lies in individual differences in attributions for negative life events.¹ Individuals with a negative cognitive style tend to attribute adverse events to stable (enduring) and global (pervasive across life domains) causes, and consequently infer negative implications for their self-worth and future outcomes.² This maladaptive pattern of attribution and inference may interfere with adaptive problem-solving and coping processes, foster a state of hopelessness characterized by pessimistic expectations, and increase the risk of depression.^3–5 Both cross-sectional and prospective studies have shown that individuals with more negative cognitive styles experience poorer psychological outcomes, including higher levels of hopelessness, more severe depressive symptoms, and an elevated risk of developing depression.^6–10

To assess cognitive vulnerability to negative life events, the Cognitive Style Questionnaire (CSQ) was developed based on the Hopelessness Theory of Depression.^6,11,12 The CSQ uses a scenario-based format, presenting 12 positive and 12 negative hypothetical scenarios to elicit spontaneous attributions and evaluations.¹¹ For each scenario, respondents are asked to write down the single most likely cause of the event and then rate the stability and globality of that cause, as well as the perceived implications for consequences and self-worth, using a 7-point Likert scale. The CSQ has been translated into several languages, including Chinese, Spanish, and Romanian, and has shown good reliability and validity in clinical and non-clinical samples (including university students).^13–18 Chinese adaptations of the full CSQ include both a 24-scenario version and a 12-scenario version. Both retain the written-cause procedure and remain time-consuming to complete. The original CSQ typically requires approximately 45 minutes to complete, which limits its feasibility in research and applied settings, particularly when cognitive vulnerability is assessed alongside other measures. To address this practical limitation, the Cognitive Style Questionnaire-Short Form (CSQ-SF) was developed.¹⁹ The CSQ-SF includes eight negative scenarios closely associated with depression (eg, conflict with parents, negative job evaluation), eliminates the written-cause component, and simplifies scoring by reducing the Likert scale from 7 to 5 points.

The CSQ-SF has been widely used in empirical studies and has demonstrated good internal consistency and test-retest reliability.^20–25 Studies examining its factor structure have yielded mixed findings. Evidence from independent translations has consistently supported a single-factor structure for the CSQ-SF.^19,23 For instance, Meins et al conducted an exploratory factor analysis based on scores from the five dimensions (internality, stability, globality, negative consequences, and self-worth implications) and identified a single-factor structure.¹⁹ Similarly, a comparable dimension-level analytic approach yielded a single-factor structure for the German version of the CSQ-SF.²³ In contrast, confirmatory factor analysis (CFA) has indicated that both single-factor and five-factor models fit the data adequately,²² suggesting that the scale may be used as either a unidimensional or multidimensional measure. Notably, previous studies differ in the inclusion of the internality dimension when computing total scores: some include all five dimensions,^21,23,25 whereas others exclude the internality dimension,^20,24,26 reflecting ongoing ambiguity regarding the conceptual role of this component. Moreover, model fit improved after excluding internality, suggesting that this dimension may contribute less consistently to the overall factor structure.²² Overall, these inconsistencies highlight the need for further clarification of the factorial validity of the CSQ-SF.

Cultural context may shape both the expression of cognitive vulnerability and the interpretation of psychological constructs. In particular, the meanings attached to hypothetical interpersonal and achievement failures, as well as to self-worth and personal responsibility, may differ across cultural settings. Consequently, psychometric findings obtained in one cultural context cannot be assumed to generalize to another.²⁷ These considerations are particularly pertinent to university students, who face academic, interpersonal, and developmental demands that may interact with cognitive vulnerability to depression.^28,29 Therefore, further psychometric evaluation of the CSQ-SF is warranted in the Chinese cultural context to determine whether its factorial and measurement properties are sufficiently adequate for meaningful interpretation and comparison among Chinese populations.

A Chinese version of the CSQ-SF has previously been examined. Chen et al translated the CSQ-SF into Chinese and conducted an initial psychometric evaluation in a sample of 415 university students, providing evidence for its internal consistency, test–retest reliability, and criterion-related validity. Using exploratory factor analysis, they identified a single-factor structure.³⁰ Although this represents an important first step, several critical psychometric issues remain unresolved. First, the factorial structure of the Chinese CSQ-SF has not been comprehensively evaluated using confirmatory factor analysis, particularly regarding whether a unidimensional or multidimensional structure provides a better representation of the scale. Second, measurement invariance across gender and over time has not been systematically tested. Given that cognitive style scores are commonly compared by gender,^29,31 and negative cognitive style is conceptualized as a relatively stable vulnerability assessed longitudinally,^31,32 establishing measurement invariance is a prerequisite for any meaningful mean-level or longitudinal comparisons.³³ Third, the initial findings by Chen et al³⁰ on the Chinese CSQ-SF require replication in a substantially larger sample to generate more robust and precise estimates of reliability and validity, and to permit a more rigorous evaluation of its factorial structure.

Therefore, with formal permission from Professor Charles Fernyhough, one of the original developers,¹⁹ the present study translated the CSQ-SF into the Chinese language and conducted a comprehensive psychometric evaluation in a large sample of Chinese university students. Specifically, we used CFA to compare alternative factorial structures, examined measurement invariance across gender and over time, and assessed internal consistency, test–retest reliability, and associations with external variables to evaluate construct validity. Given the mixed findings regarding the internality dimension, we additionally examined its psychometric performance and contribution to the overall factor structure. Together, these analyses were intended to clarify the factorial and measurement properties of the Chinese CSQ-SF and to provide evidence supporting its use as an efficient measure of cognitive vulnerability in Chinese-speaking populations.

Methods

Participants

From October 2024 to December 2024, a total of 3743 undergraduates were recruited using convenience sampling from two universities in Hunan Province, China, and completed paper-based questionnaires. After excluding questionnaires with more than 10% missing data or exhibiting patterned responses, including invariant strings of ≥ 10 consecutive identical responses or obvious non-random geometric marking patterns (eg, straight-line or diagonal-line patterns), 3525 valid questionnaires were retained, resulting in a valid response rate of 94.18%. Participants ranged in age from 16 to 24 years (18.52 ± 1.29 years). The sample included 1242 males (35.23%) and 2283 females (64.77%). At the first assessment (Time 1, T1), participants completed a set of measures assessing cognitive style, depressive symptoms, anxiety, dysfunctional attitudes, and rumination.

Five weeks after the initial assessment (Time 2, T2), based on convenience sampling, a subset of participants enrolled in the Foundations of Education course were recruited to complete the CSQ-SF again to assess test–retest reliability. Of the 463 questionnaires collected, 429 valid responses were retained, yielding a valid response rate of 92.66%. The final sample consisted of 144 males (33.57%) and 285 females (66.43%), aged 16 to 23 years (18.13 ± 1.39 years).

Instruments

Cognitive Style Questionnaire-Short Form (CSQ-SF)

With permission from the original authors,¹⁹ the CSQ-SF was translated and culturally adapted into Chinese following a standard translation–back-translation procedure. Specifically, the original scale was independently translated into Chinese by two doctoral-level psychologists, and discrepancies were resolved through discussion to produce a preliminary version. This version was subsequently back-translated by a bilingual doctoral-level psychologist to ensure translation accuracy. A bilingual professor of psychology then reviewed and refined the translated version. To further improve clarity and readability, the revised version was evaluated by four undergraduate students, and their feedback was incorporated to produce a pilot version. The pilot version was administered to a sample of 60 undergraduate students to assess clarity and comprehensibility. The pilot administration took approximately 6 minutes to complete. Based on the results, minor modifications were made to wording and formatting, resulting in the final Chinese version.

The CSQ-SF consists of eight hypothetical scenarios in which participants are asked to imagine the causes of specific negative events occurring to themselves. Based on expert review and pilot testing, all scenarios were considered culturally appropriate and relevant for Chinese participants in the present study, with no substantial difficulties identified in translation or adaptation. Each scenario includes nine items assessing five dimensions of negative cognitive style: internality, globality, stability, negative consequences, and self-worth implications. All items are rated on a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree). The scale yields dimension scores by summing items within each dimension, and a total score is calculated by summing all 72 items, with higher scores indicating a more negative cognitive style.¹⁹

Center for Epidemiologic Studies Depression Scale (CES-D)

The CES-D was used to assess individuals’ depressive symptoms over the past week.³⁴ It consists of 20 items rated on a 4-point Likert scale (1 = none or almost none to 4 = almost always), yielding a total score ranging from 20 to 80. Higher scores indicate more severe depressive symptoms.³⁵ In the present study, the CES-D demonstrated excellent internal consistency, with a Cronbach’s α of 0.91.

State-Anxiety Inventory (SAI)

The SAI was performed to estimate the severity of anxiety over the past week.³⁶ The SAI comprises 20 items rated on a 4-point Likert scale (1 = none to 4 = severe), producing total scores ranging from 20 to 80, with higher scores reflecting greater anxiety.³⁷ In the current sample, the scale demonstrated excellent internal consistency, with a Cronbach’s α of 0.92.

Dysfunctional Attitude Scale (DAS)

The DAS was used to assess individuals’ maladaptive beliefs and negative cognitive schema.³⁸ It comprises 40 items rated on a 7-point Likert scale (1 = totally disagree to 7 = totally agree), yielding total scores ranging from 40 to 280, with higher scores indicating more severe dysfunctional attitudes. In this study, the DAS demonstrated excellent internal consistency, with a Cronbach’s α of 0.92.

Ruminative Response Scale (RRS)

The RRS consists of 22 items rated on a 4-point Likert scale (1 = almost never to 4 = almost always), yielding total scores ranging from 22 to 88. Higher scores indicate a greater tendency toward rumination.³⁹ In the current study, the RRS demonstrated excellent internal consistency, with a Cronbach’s α of 0.91.

Data Analysis

Data analyses were performed using IBM SPSS Statistics 26.0 and Mplus 8.3.

The T1 sample (N = 3525) was used for confirmatory factor analysis (CFA), internal consistency reliability, and construct validity analyses. The matched sample that completed both assessments (N = 429) was used for longitudinal measurement invariance testing and test–retest reliability analysis.

Normality was assessed using skewness and kurtosis values, with a normal distribution assumed when skewness values fell within the range of −2 to +2 and kurtosis values within −7 to +7.^40,41 CFA was conducted using the T1 sample to examine the latent structure of the CSQ-SF. Two theoretically relevant models were tested: a first-order five-factor model with five correlated factors (internality, globality, stability, negative consequences, and self-worth implications), and a second-order single-factor model in which these five first-order dimensions loaded onto a second-order general cognitive style factor. A unidimensional single-factor model was not tested because the theoretical structure of the CSQ-SF specifies five distinct dimensions. To reduce model complexity and maintain stable parameter estimates, item parcels were created following established recommendations.^42,43 Within each of the five dimensions, items were randomly assigned to three parcels, with each parcel including items from two to three different scenarios (eg, I11, I12, and I13 in Figures 1 and 2). Thus, a total of 15 parcels were created, with three parcels representing each dimension. The same 15 parcels were used as observed indicators in both factor models. This parceling approach is justified by the unidimensionality of each dimension and serves to reduce the influence of item-level idiosyncrasies. Model fit was evaluated using the Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), Standardized Root Mean Square Residual (SRMR), and Root Mean Square Error of Approximation (RMSEA) with its 90% confidence interval (CI). Values of CFI and TLI greater than 0.90, and RMSEA and SRMR less than 0.08 were considered indicative of acceptable model fit.⁴⁴

Figure 1.

First-order five-factor model of the Cognitive Style Questionnaire-Short Form.

Figure 2.

Second-order single-factor model of the Cognitive Style Questionnaire-Short Form.

Multi-group confirmatory factor analysis (MGCFA) was performed to examine measurement invariance across gender (based on the T1 sample; N = 3525) and across time (based on the matched sample; N = 429). Measurement invariance was tested sequentially using four increasingly restrictive models, including configural, metric, scalar, and strict invariance.⁴⁵ First, configural invariance was established by allowing all parameters to be freely estimated across groups (or time points). Metric invariance was then evaluated by constraining factor loadings to be equal. Building upon this, scalar invariance was tested by further imposing equality constraints on item intercepts. Finally, strict invariance was examined by constraining error variances to be equal across groups (or time points). Changes in model fit were evaluated using ΔCFI, with ΔCFI ≤ 0.01 indicating that invariance was supported.^46,47

Sample-size adequacy was evaluated using the N:q ratio (sample size to estimated parameters), a widely used approach in SEM-based psychometric research, with 10:1 as the minimum acceptable ratio.⁴⁸ The models tested in the present study were relatively parsimonious, containing no cross-loadings or correlated error terms, which supports the appropriateness of this criterion. For the baseline CFA (N = 3525), the first-order five-factor model (40 estimated parameters) yielded an N:q ratio of 88.13:1, and the second-order factor model (35 estimated parameters) yielded a ratio of approximately 100.71:1. Both substantially exceeded the 10:1 minimum. For the longitudinal measurement invariance analyses, the matched sample comprised 429 participants who completed both assessments. The first-order model in this subsample yielded an N:q ratio of 10.73:1, exceeding the minimum guideline and indicating that the sample size was adequate for these analyses.

Internal consistency reliability was assessed using Cronbach’s α, McDonald’s omega (ω), and mean inter-item correlations (M_IC). Cronbach’s α values of 0.70 or higher are generally recommended, but values as low as 0.60 are also considered acceptable.⁴⁹ The M_IC is considered acceptable at between 0.10 and 0.50.⁵⁰ Test–retest reliability was evaluated using Intraclass Correlation Coefficient (ICC), with values below 0.50 considered poor, 0.50–0.75 moderate, 0.75–0.90 good, and above 0.90 excellent.⁵¹ Convergent validity was examined by calculating Pearson correlations between the CSQ-SF and related variables, including depressive symptoms (CES-D), anxiety symptoms (SAI), dysfunctional attitudes (DAS), and rumination (RRS). Those variables were selected based on prior evidence linking cognitive vulnerability to both depressive and anxiety symptoms.¹² In addition, dysfunctional attitudes and rumination represent distinct cognitive vulnerability constructs.⁵²

Results

Descriptive Statistics of the CSQ-SF

Table 1 shows descriptive statistics and correlations among the CSQ-SF subscales and total scale. Subscales were positively correlated with each other (r = 0.29–0.70, ps < 0.001), with higher correlations with the total scale (r = 0.67–0.90, ps < 0.001). The mean corrected item-total correlations (CITCs) and corrected dimension-total correlations (CDTCs) were calculated. Although the mean CITC for the internality dimension was slightly lower than those of other subscales, it exceeded the recommended lower bound of 0.30.⁵³

Table 1.

Descriptive Statistics and Correlations Among CSQ-SF Subscales and Total Score (N = 3525)

	M (SD)	r					CITCs	CDTCs
		1	2	3	4	5
1. Internality	42.83 (4.98)						0.33	0.53
2. Globality	41.17 (5.97)	0.45***					0.42	0.76
3. Stability	40.99 (6.77)	0.34***	0.70***				0.45	0.70
4. Negative consequences	21.14 (3.65)	0.29***	0.58***	0.58***			0.41	0.62
5. Self-worth implications	37.47 (8.50)	0.61***	0.67***	0.62***	0.54***		0.58	0.77
CSQ-SF total score	183.60 (24.21)	0.67***	0.86***	0.83***	0.71***	0.90***	-	-

Note: ***p < 0.001.

Abbreviations: M, means; SD, standard deviations; CITCs, mean corrected item-total correlations; CDTCs, corrected dimension-total correlations.

The Goodness-of-Fit Indexes of CFA

All variables met the predefined criteria for normality, with skewness values ranging from −0.63 to 1.05 and kurtosis values ranging from −0.77 to 1.88. These results supported the use of maximum likelihood estimation.

The first-order five-factor model (Figure 1) also demonstrated acceptable fit to the data, χ² (80) = 1416.24, CFI = 0.941, TLI = 0.920, SRMR = 0.034, RMSEA = 0.061 (90% CI [0.058, 0.064]). The standardized factor loadings of parcels on the five first-order factors ranged from 0.51 to 0.81. The second-order model (Figure 2) showed an acceptable fit to the data, χ² (85) = 2417.67, CFI = 0.924, TLI = 0.904, SRMR = 0.045, RMSEA = 0.074 (90% CI [0.071, 0.077]). The standardized factor loadings of parcels on first-order factors ranged from 0.53 to 0.81, and the loadings of first-order factors on the second-order factor ranged from 0.70 to 0.96. Model comparison indicated that the first-order five-factor model provided a significantly better fit than the second-order model (Δχ²(5) = −1001.43, p < 0.001; ΔCFI = 0.017, ΔRMSEA = −0.013). Although the second-order model is more parsimonious, the first-order five-factor model was retained for subsequent measurement invariance analyses because it demonstrated superior fit and directly represented the five hypothesized dimensions of the CSQ-SF.

Measurement Invariance of the Five-Factor Model of the CSQ-SF Across Gender and Time

As shown in Table 2, the five-factor model of the CSQ-SF demonstrated acceptable model fit in each subgroup separately, across gender (male vs female) and time (T1 vs T2). Measurement invariance analyses supported configural, metric, scalar, and strict invariance across both gender and time, with satisfactory fit indices (CFI, TLI, RMSEA, and SRMR) and ΔCFI values ≤ 0.01.

Table 2.

Fit Indexes for the Measurement Invariance Tests of CSQ-SF Across Gender and Across Time

	χ2	df	CFI	TLI	SRMR	RMSEA (90% CI)	ΔCFI
Across gender (N = 3525)
Model male (N = 1242)	523.562	80	0.946	0.928	0.031	0.062 (0.057, 0.068)	–
Model female (N = 2283)	1288.082	80	0.926	0.901	0.038	0.074 (0.070, 0.078)	–
Configural invariance	1811.643	160	0.933	0.911	0.036	0.070 (0.067, 0.073)	–
Metric invariance	1851.121	170	0.931	0.914	0.042	0.069 (0.066, 0.072)	−0.002
Scalar invariance	2068.461	180	0.921	0.907	0.045	0.071 (0.068, 0.074)	−0.010
Strict invariance	2094.323	195	0.921	0.914	0.050	0.069 (0.066, 0.072)	0
Across time (N = 429)
Model T1	332.547	80	0.928	0.905	0.046	0.073 (0.064, 0.083)	–
Model T2	275.395	80	0.945	0.928	0.033	0.075 (0.066, 0.085)	–
Configural invariance	820.905	345	0.935	0.918	0.039	0.057 (0.052, 0.062)	–
Metric invariance	833.246	355	0.934	0.920	0.044	0.056 (0.051, 0.061)	−0.001
Scalar invariance	948.566	365	0.928	0.914	0.049	0.058 (0.053, 0.063)	−0.006
Strict invariance	1016.812	380	0.921	0.909	0.058	0.060 (0.055, 0.064)	−0.007

Abbreviations: df, degrees of freedom; CFI, Comparative Fit Index; TLI, Tucker– Lewis Index; SRMR, Standardized Root Mean Squared Residual; RMSEA, Root-Mean-Square Error of Approximation; ΔCFI, Change in CFI.

Reliability

As shown in Table 3, the Cronbach’s α coefficients for the total scale and subscales of the CSQ-SF ranged from 0.60 to 0.92, and McDonald’s ω coefficients ranged from 0.61 to 0.92. The mean inter-item correlations (M_IC) ranged from 0.15 to 0.41, and the split-half reliability coefficients ranged from 0.58 to 0.83. The test–retest reliability of the total scale was 0.91, while that of the subscales ranged from 0.59 to 0.85.

Table 3.

Reliability of the Chinese Version of the CSQ-SF and Its Subscales

	Total Score	Internality	Globality	Stability	Negative Consequences	Self-Worth Implications
Cronbach’s α	0.92	0.62	0.72	0.75	0.60	0.85
McDonald’s ω	0.92	0.64	0.73	0.76	0.61	0.85
MIC	0.20	0.18	0.25	0.27	0.15	0.41
Split-half reliability	0.83	0.58	0.68	0.73	0.58	0.82
ICC (N = 429)	0.91	0.59	0.74	0.76	0.65	0.85

Abbreviations: M_IC, mean inter-item correlations; ICC, intraclass correlation coefficient.

Validity

As shown in Table 4, all CSQ-SF subscales were significantly and positively intercorrelated and correlated with the total score (r = 0.29–0.90, ps < 0.001). CSQ-SF subscale and total scores were also significantly positively correlated with CES-D, SAI, DAS, and RRS scores (ps < 0.001). Notably, correlations involving the internality subscale (r = 0.15–0.30) were generally lower than those of the other dimensions.

Table 4.

Descriptive Statistics and Pearson Correlations Among the CES-D, SAI, DAS, RRS, and CSQ-SF (N = 3525)

	M (SD)	CSQ-SF	Internality	Globality	Stability	Negative Consequences	Self-Worth Implications
CES-D	38.65 (9.34)	0.46***	0.16***	0.41***	0.52***	0.37***	0.36***
SAI	40.91 (9.74)	0.44 ***	0.15***	0.39***	0.48***	0.36***	0.34***
DAS	137.90 (24.94)	0.62 ***	0.30***	0.50***	0.62***	0.47***	0.53***
RRS	45.41 (9.83)	0.42***	0.16***	0.35***	0.43***	0.33***	0.37***

Note: ***p < 0.001.

Abbreviations: CSQ-SF, Cognitive Style Questionnaire-Short Form; CES-D, Center for Epidemiologic Studies Depression Scale; SAI, State-Anxiety Inventory; DAS, Dysfunctional Attitude Scale; RRS, Ruminative Response Scale.

Discussion

As a widely used instrument for assessing cognitive vulnerability in recent years, the CSQ-SF has demonstrated good reliability and validity across community, student, and high suicide-risk populations.^19–25 However, controversies remain regarding its factor structure, as well as the inclusion of the internality dimension in the total score. In addition, its measurement invariance across gender and time has yet to be established. Therefore, using a large sample of university students, the present study aimed to systematically evaluate the Chinese version of the CSQ-SF, examine its optimal factor structure, and test its measurement invariance across gender and time. Furthermore, reliability and validity analyses were conducted to re-evaluate its psychometric properties, thereby providing additional evidence to support its application in the Chinese-speaking population.

Regarding the internality dimension, several studies have reported relatively weak correlations between the internality dimension and the other dimensions of the CSQ-SF (r ≤ 0.36),^19,22,23 prompting some researchers to exclude this dimension from total scores under the assumption that internality may not represent a core component of cognitive vulnerability.^20,24,26 In the present study, however, the internality subscale exhibited moderate correlations with the remaining dimensions (r = 0.29–0.61). Furthermore, both the mean CITC (0.33) and CDTC (0.53) exceeded the conventional threshold of 0.30 typically used to identify items or dimensions that contribute meaningfully to the overall construct.^22,54,55 These results indicate that, although the internality dimension contributes somewhat less strongly than other subscales, it still provides meaningful information regarding individual differences in negative cognitive style. Thus, we retained the internality dimension in the total score, consistent with the original scoring approach.¹⁹ Nevertheless, it is important to note that the internal consistency of the internality subscale was relatively low compared to the other dimensions, suggesting that scores on this subscale should be interpreted with caution when used in isolation. This finding echoes the ambiguity noted in the literature and calls for further refinement of internality items in future Chinese adaptations.

Concerning the factor structure, confirmatory factor analyses indicated that both the first-order five-factor model and the second-order single-factor model of the CSQ-SF demonstrated acceptable fit to the data. The CSQ-SF was originally conceptualized as a unidimensional construct, and our findings support this theoretical assumption, as the second-order single-factor model achieved an adequate fit.¹⁹ At the same time, evidence suggests that the CSQ-SF can also be meaningfully represented as a multidimensional construct comprising five factors.²² Consistent with this perspective, our results further confirmed the statistical acceptability of the five-factor model. These findings indicate that the CSQ-SF can be legitimately interpreted at two levels, the second-order model supports the use of a total score as a global index of cognitive vulnerability, whereas the first-order model justifies the examination of individual dimensions (eg, stability, globality). Importantly, the multidimensional structure provides practical value by enabling researchers to examine cognitive vulnerability through subscale scores rather than a single composite score. The results support the stability of the five-factor structure in the present sample. However, its generalizability to other language versions requires further cross-cultural validation. To further substantiate the robustness of the multidimensional structure in the Chinese version, measurement invariance was tested using the five-factor model as the baseline.

Measurement invariance is a prerequisite for meaningful group comparisons and longitudinal analyses.⁴⁵ MGCFA results demonstrated that all four levels of measurement invariance (configural, metric, scalar, and strict invariance) were supported across gender and time. To our knowledge, this is the first study to establish full measurement invariance of the CSQ-SF across gender and over time. The establishment of configural invariance indicates that the factor structure of the CSQ-SF is consistent across groups and time points, with the same number of factors and pattern of factor loadings. The support for metric and scalar invariance further suggests that the observed items and latent constructs have equivalent meanings across gender and time.⁴⁵ Additionally, strict invariance was also established, indicating equivalence of residual variances. Taken together, these findings indicate that full measurement invariance of the CSQ-SF has been established across gender and over time, supporting its robustness for assessing group and longitudinal differences in cognitive vulnerability.

Reliability and validity findings further support the psychometric quality of the Chinese CSQ-SF. In the present study, the CSQ-SF demonstrated excellent internal consistency, with Cronbach’s α = 0.92, which is comparable to values reported in previous studies (α = 0.85–0.94).^{19,20,23–25} Additionally, McDonald’s ω for the scale was 0.92, and the M_IC was 0.20, further supporting the good internal consistency of the Chinese version of the CSQ-SF. The scale further exhibited excellent test–retest reliability for the total score (ICC = 0.91), suggesting that the overall measure of negative cognitive style is highly stable over time. Pearson correlation analyses showed that cognitive vulnerability was significantly positively associated with depression, anxiety, dysfunctional attitudes, and ruminative tendencies, consistent with previous research.^{12,19,21,25,52} These results suggest that higher levels of cognitive vulnerability may contribute to increased depressive and anxiety symptoms, as well as greater maladaptive cognitive patterns, providing further support for the construct validity of the CSQ-SF. Taken together, these results align with the theoretical framework of the Hopelessness Theory of Depression, supporting the view that the CSQ-SF captures a cognitive style that is meaningfully associated with depression-relevant constructs.

Overall, the present study extends the existing psychometric evidence for the CSQ-SF by comprehensively evaluating a Chinese translation in a large sample of university students. The confirmed factor structure and strict measurement invariance across gender and time indicate that the core dimensions of negative cognitive style identified in previous research are also applicable to Chinese university students. These findings provide a stronger empirical foundation for future research on cognitive vulnerability and its associations with depression, anxiety, rumination, and dysfunctional attitudes in Chinese populations. Moreover, the brief format of the CSQ-SF reduces respondent burden relative to the full-length CSQ, facilitating its use in large-scale surveys, longitudinal studies, and broader batteries of psychological measures. Thus, the present Chinese CSQ-SF provides a psychometrically sound and practical tool for assessing cognitive vulnerability in university settings.

There are several limitations for this study. First, the sample consisted exclusively of university students, limiting the generalizability of the findings to other age groups and to populations with more diverse educational and socioeconomic backgrounds. Using the CSQ-SF with participants of different ages may require modification of some item scenarios, which could affect the applicability of the scale across age groups. Future studies should include a broader age range and more representative multi-age samples, potentially adapting item scenarios to better capture cognitive vulnerability across different age groups. Second, participants were not screened for psychiatric disorders, which restricts the applicability of the findings to clinical populations. Examining the psychometric properties of the CSQ-SF in clinical populations is therefore essential to establish its applicability among individuals with mental health conditions. Third, the present study examined only convergent validity, without testing discriminant validity (eg, against measures of social desirability). Incorporating discriminant validity tests in future research would provide a more comprehensive assessment of construct validity. Fourth, the internality subscale showed relatively low internal consistency, raising questions about its reliability. Further investigation is needed to determine whether this reflects the nature of the internality construct or limitations of the current item set, and whether refinements are warranted.

Conclusion

In conclusion, the Chinese version of the CSQ-SF demonstrated acceptable reliability and validity, with a supported five-factor structure and strict measurement invariance across gender and time. Compared with the full-length CSQ, the CSQ-SF substantially reduces respondent burden and is better suited for large-scale surveys, longitudinal studies, and multi-measure research batteries. These features advance the measurement of negative cognitive style and cognitive vulnerability in Chinese university students. However, caution is warranted when interpreting the internality subscale in isolation due to its relatively lower internal consistency.

Funding Statement

This work was supported by the National Natural Science Foundation of China (Grant number: 82271573), the Hunan Provincial Innovation Foundation for Postgraduate (Grant number: CX20230358), and the Fundamental Research Funds for Central Universities of the Central South University (Grant number: 2023ZZTS0288).

Data Sharing Statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Ethics Approval and Informed Consent

The study was approved by the Ethics Committee of the Second Xiangya Hospital, Central South University, under registration number [2022-059]. The procedures performed were in accordance with the Declaration of Helsinki. All participants participated voluntarily and provided written informed consent at enrollment.

Disclosure

All authors declare no conflict of interest.