Psychometric Properties and Factor Structure of the Ruminative Responses Scale-Short Form in Colombia

Psychometric Properties and Factor Structure of the Ruminative Responses Scale-Short Form in Colombia

Full reference: Ruiz, F. J., Suárez Falcón, J. C., Sierra, M. A., Barreto Montero, K., García Martín, M. B., Bernal, P. A., & Ramírez, E. S. (2017). Psychometric Properties and Factor Structure of the Ruminative Responses Scale-Short Form in Colombia. International Journal of Psychology & Psychological Therapy, 17(2), 199-208.

Study type: Cross-cultural validation and adaptation / Factor structure and measurement invariance study

Background and objectives

Rumination, defined as a style of coping with sadness and depression characterized by repetitive and passive thinking concerning the causes, consequences, and meaning of negative emotional states, has been extensively investigated within Response Style Theory (RST). Empirical evidence has demonstrated that rumination is a particularly pernicious coping strategy, as it traps ruminators in cycles of repetitive thinking that do not lead to effective problem-solving. Numerous longitudinal studies have confirmed that the tendency to ruminate predicts more severe depressive symptoms and the onset of depressive disorders, and is associated with other psychological problems including anxiety, binge eating, addictions, and self-harm.

The Ruminative Response Scale (RRS; Nolen-Hoeksema & Morrow, 1991) is the most widely used instrument for measuring rumination in response to depressive mood. However, subsequent research identified that several items of the original 22-item RRS overlapped with items from depression measures, which could artificially inflate correlations between the two constructs. In response to this limitation, Treynor, González, and Nolen-Hoeksema (2003) developed a short form (RRS-SF) consisting of 10 items that eliminated overlapping content, organized into two subscales: Brooding (the most maladaptive form of rumination) and Reflection (which may have both maladaptive and potentially adaptive aspects). The short form demonstrated acceptable psychometric properties (Cronbach's alphas of .77 for Brooding and .72 for Reflection) in the original English version.

Subsequently, the RRS-SF was translated into Spanish in samples from Spain and Chile (Hervás, 2008; Cova Solar, Rincón, & Melipillán, 2007). Hervás (2008) reported acceptable psychometric properties in Spanish nonclinical samples (alphas of .80 for Brooding and .74 for Reflection), with confirmatory factor analysis revealing the presence of the two proposed factors. However, no previous study had evaluated the psychometric properties of the RRS-SF in clinical samples in Spanish-speaking countries, nor specifically in Colombian populations. This represented an important gap, as testing measures in culturally diverse samples not only increases confidence in the instrument but also enhances the cross-cultural relevance of the underlying theory.

The objective of the present study was to investigate the psychometric properties and factor structure of the Spanish version of the RRS-SF (Hervás, 2008 adaptation) in both clinical and nonclinical Colombian samples, and to examine measurement invariance across these groups and between genders.

Method

Participants

The study included two samples. Sample 1 consisted of 368 nonclinical participants (62% female) with ages ranging from 18 to 89 years (M = 26.65, SD = 9.81). Educational level was heterogeneous: 49.2% had completed primary education or mid-level studies (high school or vocational training), 33.4% were undergraduate or college graduates, and 16.4% were currently studying or held a postgraduate degree. Participants completed an anonymous online survey distributed through social media. Thirty-nine percent reported having received psychological or psychiatric treatment at some time, but only 7.3% were currently in treatment. Additionally, 4.3% reported consumption of psychotropic medication.

Sample 2 consisted of 236 patients (74% female) with ages ranging from 18 to 63 years (M = 27.02, SD = 10.23), suffering from emotional symptoms. All participants were being evaluated at private psychological consultation centers. Only 7.1% reported consuming psychotropic medication. In both samples, the percentage of women was significantly higher than men, especially in the clinical sample.

Instrument under study

Ruminative Responses Scale-Short Form (RRS-SF; Treynor et al., 2003; Spanish version by Hervás, 2008). The RRS-SF is a 10-item self-report instrument rated on a 4-point Likert scale (4 = almost always; 1 = almost never), designed to measure the tendency to ruminate in response to feelings of sadness and depression. The instrument contains two subscales: Brooding (items 1, 3, 6, 7, 8) and Reflection (items 2, 4, 5, 9, 10). According to Treynor et al. (2003), brooding represents the most maladaptive form of rumination, while reflection may have both maladaptive and potentially adaptive aspects.

Other outcome measures

Several instruments were used to assess convergent validity evidence:

Acceptance and Action Questionnaire-II (AAQ-II; Bond et al., 2011; Spanish version by Ruiz, Langer, Luciano, Cangas, & Beltrán, 2013). A 7-item instrument on a 7-point Likert scale (7 = always true; 1 = never true) measuring experiential avoidance. The Spanish version demonstrated excellent psychometric properties (overall alpha of .90) and one-factor structure in Colombia.

Cognitive Fusion Questionnaire (CFQ; Gillanders et al., 2014; Spanish version by Ruiz, Suárez Falcón, Riaño Hernández, & Gillanders, 2017). A 7-item instrument on a 7-point Likert scale (7 = always; 1 = never true) measuring cognitive fusion. The Spanish version showed psychometric properties similar to the original version (overall alpha of .92) in Colombia.

Depression, Anxiety, and Stress Scales-21 (DASS-21; Antony, Bieling, Cox, Enns, & Swinson, 1998; Spanish version by Daza, Novy, Stanley, & Averill, 2002). A 21-item instrument on a 4-point Likert scale assessing negative emotional symptoms. It contains three subscales: Depression, Anxiety, and Stress. The Colombian version demonstrated adequate psychometric properties with hierarchical structure.

General Health Questionnaire-12 (GHQ-12; Goldberg & Williams, 1988; Spanish version by Rocha, Pérez, Rodríguez Sanz, Borrell, & Obiols, 2011). A 12-item instrument on a 4-point Likert scale widely used as a screening tool for psychological disorders. In the present study, it showed Cronbach's alphas of .91 in both samples.

Procedure

Before administering the RRS-SF, two preliminary studies were conducted to assess item comprehensibility in Colombia and content validity. First, the RRS-SF was administered to 64 clinical psychology trainees to verify comprehensibility, with no understanding problems reported. Subsequently, items were evaluated by 3 experts in rumination who rated their representativeness, comprehensibility, interpretation, and clarity. Aiken's V was above the standard threshold of .50 for all items.

Sample 1 participants completed an anonymous online survey distributed through social media. Sample 2 participants completed questionnaires during initial clinical assessment interviews, in the presence of their therapist. In both samples, following informed consent, participants received a questionnaire package including all mentioned instruments.

Data analysis

Confirmatory Factor Analysis (CFA) was conducted using LISREL 8.71 with weighted least squares estimation method. The fit of a one-factor model was compared versus a two-factor model. Modification indices were explored to identify correlations between error terms of items. Goodness-of-fit indices evaluated were: Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI), Non-Normed Fit Index (NNFI), and Expected Cross-Validation Index (ECVI). RMSEA ≤ .10 indicated acceptable fit and ≤ .05 indicated very good fit; for CFI and NNFI, values > .90 indicated acceptable fit and > .95 indicated very good fit.

Measurement invariance analyses were conducted using multi-group CFA to examine whether factor loadings were equivalent across clinical and nonclinical samples, and between genders. Chen's (2007) criteria were applied: (a) no significant decrement in fit according to chi-square test; (b) difference in RMSEA < .01; (c) difference in CFI > -.01; (d) difference in NNFI > -.01.

Finally, Cronbach's alphas with 95% confidence intervals were calculated for reliability, corrected item-total correlations were obtained, descriptive statistics computed, gender differences examined via Student's t, and Pearson correlations calculated to assess convergent validity.

Results

Factor structure

CFA in the total sample (N = 604) compared one-factor and two-factor models. The two-factor model showed slightly better fit, and modification indices suggested allowing error terms of items 4 and 10 to correlate. The two-factor model with correlated errors showed acceptable fit: RMSEA = .092 [90% CI: .080, .10], CFI = .95, NNFI = .94, ECVI [90% CI: .41, .34, .49]. This model was clearly superior to the one-factor model (RMSEA = .11, CFI = .93, NNFI = .91) and to the two-factor model without error correlation (RMSEA = .11, CFI = .93, NNFI = .91). Figure 1 in the article presents the standardized solution of the final model, showing factor loadings for Brooding ranging from .73 to .84, and for Reflection from .34 to .75.

Measurement invariance

Multi-group baseline models showed good fit to the data for both comparisons (clinical vs. nonclinical samples, and gender). When equality constraints were applied to factor loadings, there was no significant decrement in goodness of fit, suggesting measurement invariance. For invariance across samples, three of four Chen (2007) criteria were fully satisfied; the chi-square difference test was significant [χ²(9) = 28.46, p < .01], but given the sensitivity of this test to sample size, measurement invariance was concluded to be broadly established. For gender invariance, all criteria were fully satisfied: chi-square test not significant [χ²(9) = 18.74, p > .01], differences in RMSEA < .01, differences in CFI and NNFI > -.01.

Reliability

In Sample 1 (nonclinical), Cronbach's alpha for Brooding was .77 [95% CI: .74, .81] and for Reflection .77 [95% CI: .73, .81]. In Sample 2 (clinical), alphas were .71 [95% CI: .65, .77] for Brooding and .70 [95% CI: .63, .75] for Reflection. In the total sample (N = 604), alphas were .79 [95% CI: .76, .81] for Brooding and .75 [95% CI: .72, .78] for Reflection. Corrected item-total correlations for Brooding ranged from .35 (Item 3) to .68 (Item 8), and for Reflection from .32 (Item 5) to .65 (Item 4), indicating acceptable discrimination for all items.

Discriminant validity

Participants in the clinical sample obtained significantly higher scores on both Brooding [t(602) = -10.81, p < .001] and Reflection [t(602) = -5.75, p < .001] compared to the nonclinical sample. Mean scores were: Brooding nonclinical M = 10.84 (SD = 3.46) vs. clinical M = 13.90 (SD = 3.48); Reflection nonclinical M = 10.91 (SD = 3.55) vs. clinical M = 12.51 (SD = 3.18). No significant gender differences were found in the nonclinical sample (Brooding: t = .37, p = .71; Reflection: t = 1.85, p = .07) or in the clinical sample (Brooding: t = -.38, p = .70; Reflection: t = .91, p = .37).

Convergent validity

Both RRS-SF subscales showed significant correlations with all assessed constructs in theoretically coherent directions. Brooding correlations were:

• With pathological worry (PSWQ): .65 (Sample 1) and .49 (Sample 2)
• With experiential avoidance (AAQ-II): .70 (Sample 1) and .63 (Sample 2)
• With cognitive fusion (CFQ): .72 (Sample 1) and .62 (Sample 2)
• With depressive symptoms (DASS-21): .66 (Sample 1) and .48 (Sample 2)
• With anxiety symptoms (DASS-21): .62 (Sample 1) and .53 (Sample 2)
• With stress symptoms (DASS-21): .67 (Sample 1) and .53 (Sample 2)
• With psychological distress (GHQ-12): .60 (Sample 1) and .48 (Sample 2)

Reflection correlations were consistently lower, ranging from .24 to .41, confirming that brooding represents the most counterproductive form. All correlations were significant at p < .001.

Discussion and conclusions

This study contributes to the field of rumination measurement by confirming that the Spanish version of the RRS-SF (Hervás, 2008) possesses solid psychometric properties in Colombian samples. Results demonstrated that the two-factor model proposed by Treynor et al. (2003) achieved acceptable fit in the Colombian population, consistent with previous findings in Spanish samples.

A particularly important contribution of this study is the establishment of measurement invariance across clinical and nonclinical samples. Until the time of this study, only Whitmer and Gotlib (2011) had analyzed differences in RRS-SF factor structure between nonclinical and clinical samples. Those authors found that the two-factor structure was inadequate in currently depressed participants but adequate in never-depressed and formerly depressed participants. The present study, however, found measurement invariance between nonclinical and a clinically heterogeneous sample. This difference may be attributed to the fact that the clinical sample was not composed exclusively of currently depressed participants but included patients with diverse emotional symptoms.

The observed psychometric properties were similar and comparable to those of the original English version. Cronbach's alphas (.77 to .79 for Brooding and .70 to .77 for Reflection) were very similar to originally reported values (.77 and .72). Convergent validity was established through significant correlations with measures of pathological worry, emotional symptoms, psychological distress, experiential avoidance, and cognitive fusion—all theoretically related constructs. As expected, correlations were stronger for the Brooding subscale than Reflection, confirming previous literature characterizing brooding as the most counterproductive form of ruminative response.

Furthermore, a previous study by Ruiz, Riaño Hernández, Suárez-Falcón, and Luciano (2016) demonstrated that this version of the RRS-SF is sensitive to change in a Colombian sample, further strengthening its utility for research and clinical practice.

The study has some limitations worth mentioning. First, systematic information on specific diagnoses in clinical participants was not obtained, limiting understanding of the exact composition of that sample. Second, some validity aspects of the RRS-SF were not evaluated in this study, such as discriminant validity and sensitivity to treatment effects. Third, the percentage of women was significantly higher than men, especially in the clinical sample, which may limit generalization of findings.

In conclusion, the Spanish version of the RRS-SF is a psychometrically reliable instrument for measuring rumination in Colombia. The reliability and validity data provided support its use in research and clinical evaluation. Future research is recommended to explore the psychometric properties of this version in other Spanish-speaking countries to strengthen confidence in the measure and its cross-cultural relevance.

Significance and contribution

This study is important because it extends the validation of a widely used instrument—the RRS-SF—to a culturally distinct population (Colombia) that had not been previously evaluated. The RRS-SF is the most widely employed instrument for measuring rumination in response to depressive states and carries significant implications for both research and clinical practice. By demonstrating that the instrument maintains its two-factor structure, acceptable reliability, and convergent validity in Colombian samples—including both nonclinical and clinical participants—the study provides confidence for instrument use in this cultural context. The establishment of measurement invariance between clinical and nonclinical groups is particularly valuable, as it enables valid score comparisons between these populations in future research. This work contributes to closing an important gap in Spanish-language psychometric literature and facilitates future research on rumination and emotional disorders in Colombian populations using a psychometrically sound instrument.