Psychometric properties of the Emotion Regulation Scale for Children and Adolescents (ERQ-CA), an assessment of emotion regulation strategies, were examined among older children. The ERQ–CA was evaluated with 147 participants between 9 and 12 years old. Explorative factor analysis and confirmatory factor analysis were calculated twice within a four-month interval. We meticulously assessed internal consistency, convergent, and concurrent validity by analyzing the relationship between ERQ-CA scores and measurements of resilience and intrinsic motivation. Measurement invariance was tested for the pre- and post-test across gender, age group, intervention group, and longitudinally. A robust two-factor structure of cognitive reappraisal (CR) and expressive suppression (ES) was found. Internal consistency was adequate, with Cronbach’s α = 0.76 for CR and α = 0.77 for ES. Test-retest reliability in four months was r = 0.50 for CR and r = 0.32 for ES. The ERQ-CA showed convergent and concurrent validity with established measures of resilience and motivation. The gender and age-based mean scores were consistent with those of previously reported studies. We found measurement invariance across gender at the pre-test and intervention groups at the pre-and post-test, indicating stability across responses to the ERQ-CA. These findings significantly contribute to the theoretical understanding of emotion regulation in children and this knowledge’s practical application in clinical, educational, and research settings.
Opmerkingen
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Introduction
The Emotion Regulation Questionnaire for Children and Adolescents (ERQ-CA) is derived from the ERQ (Gross & John, 2003) but specifically tailored to younger populations. The ERQ-CA assesses the two primary emotion regulation strategies, cognitive reappraisal, and expressive suppression, which evaluate a child’s or adolescent’s emotional regulation capabilities. Previous research by Gullone and Taffe (2012) show the ERQ-CA, used with n = 827 10- to 18-year-old (M = 13.9, SD = 2.46) children and adolescents, revealed Cronbach’s alpha coefficients (CAs) of α = 0.85 and α = 0.76 for the CR and ES scales, respectively. A 12-month interclass correlation coefficient of 0.37 and 0.40 was reported for the 10–12-year age group for the CR and ES, respectively. A confirmatory factor analysis (CFA) showed a poor fit to a single underlying factor of the ERQ-CA. In contrast, an adequate fit was reported for a two-factor model. A significant positive association was found between the ES subscale and the Children’s Depressive Inventory (CDI). Correlating the CR subscale with the Big-Five Questionnaire for Children (BFQ–C; Barbaranelli et al., 2003) showed a negative correlation with the neuroticism/emotional instability factor and a positive correlation with extraversion.
The ERQ-CA has been validated with Chinese children (N = 1381; 7–12 years; Liu et al., 2017), Portuguese adolescents (N = 809; 12–19 years; Teixeira et al., 2015), and Spanish adolescents (N = 462; 11–17 years; Martín-Albo et al., 2018), replicating the two-factor structure and the intergroup invariance for age, gender, and other demographic variables. More importantly, results showed positive relations between CR and measures of psychological functioning, such as lower neuroticism, fewer depressive symptoms, and higher levels of self-esteem and well-being (Teixeira et al., 2015), with expressive suppression (ES) showing the same relations in the opposite direction. A study by Martín-Albo et al. (2018), also using a Spanish sample of adolescents, demonstrated an adequate CA (0.77) for the CR factor and a low (0.52) for the SE. The predictive validity showed that depression was negatively related to CR and positively related to SE. However, CR and SE were positively related, indicating that the same participants in this study applied both emotional regulation strategies. The finding from Martín-Albo et al. (2018) contradicts the finding of Gullone and Taffe (2012) of a negative relation between these two regulation strategies.
Examining the latent structure of the ERQ for adults across N = 11,288 individuals from 29 countries, Burghart et al. (2023) indicate that emotion regulation strategies may not readily converge across cultures. This underscores the need to assess and test the psychometric properties of such scales in various cross-cultural settings. In addition, no examination of the relationship between emotion regulation strategies and intrinsic motivation, social support, and social self-efficacy using a sample of younger children has been reported. Emotion regulation ability is significantly related to resilience and self-efficacy (Wu et al., 2022). Lande et al. (2023) have shown that cognitive appraisal processes involve emotion regulation, building resilience, and developing self-efficacy. Furthermore, emotion regulation, resilience, and self-efficacy are all significant predictors of mental health outcomes (Lande et al., 2023; Wu et al., 2022). Deci and Ryan’s self-determination theory (SDT) relies on intrinsic motivation (Ryan & Deci, 2017). SDT defines intrinsically motivated behavior as driven by inner joy or personal interest in the activity (Ulstad et al., 2020). The Intrinsic Motivation Inventory (IMI) questionnaire was developed to measure intrinsic motivation. Numerous studies have linked IMI results to the potential for experiencing the joy of movement (Ryan & Deci, 2017; Ulstad et al., 2020). In conclusion, scientific literature strongly supports assessing emotion regulation ability together with intrinsic motivation, resilience, and self-efficacy because of their interconnected nature, shared underlying mechanisms, and comprehensive view of mental health, especially in children and adolescents.
There is a pressing need for more comprehensive research on the psychometric properties of the ERQ-CA in younger children. Furthermore, there is a crucial need for testing and reporting measurement invariance to examine whether differences in scores indicate whether the measure is variant (unstable) or invariant (stable) across groups that differ in characteristics such as gender and age. Moreover, the literature lacks ERQ-CA assessments of samples from a Scandinavian country, which underscores the importance of our study in filling these gaps.
The current study has several objectives. First, using both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA), we aim to examine the factor structure of the official Norwegian version of the ERQ-CA. Second, we report on the scale’s internal stability, the reliability of a four-month test–retest trial, and measurement invariance. Third, we examine the correlations between the ES and CR Scales of the ERQ-CA, the Social Support and Social Self-efficacy subscales from the Resilience Scale for Adolescents (READ), and the subscales of Enjoyment, Competence, and Value/Usefulness from the IMI.
Materials and Methods
The data were derived from a comprehensive experimental intervention study involving a tuition-free football [soccer] school (Pedersen et al., 2024). The study comprised an intervention and control group and used questionnaires and skill tests. The pre-test data collected include information from the Emotion Regulation Questionnaire (ERQ-CA; Gullone & Taffe, 2012), the Social Support and Social Self-efficacy subscales from the READ (Hjemdal et al., 2006), and three of the six IMI subscales – Interest/Enjoyment, Perceived Competence, and Value/Usefulness (McAuley et al., 1989).
Participants
Participants were older children aged 9–12 who had applied to join a football school free of charge. A total of 175 children applied, and 147 (mean age = 10.3, SD = 1.2) were accepted (boys n = 101, M = 10.3, SD = 1.2; girls n = 46, M = 10.5, SD = 1.3) into the school and completed the questionnaire. The number of boys and girls by age is as follows: age 9 boys (n = 38), age 9 girls (n = 16); age 10 boys (n = 26), age 10 girls (n = 7); age 11 boys (n = 14), age 11 girls (n = 6); age 12 boys (n = 24), age 12 girls (n = 16).
Measures
The ERQ-CA is a self-report instrument that measures children’s emotion regulation strategies. The ERQ-CA consists of ten items assessed on a five-point Likert scale, with responses ranging from 1 (strongly disagree) to 5 (strongly agree). The scale assesses two dimensions of emotion regulation: CR and ES (Gullone & Taffe, 2012). CA coefficients range from 0.73 to 0.82 for the CR subscale and 0.64 to 0.81 for the ES subscale (Gullone & Taffe, 2012). The ERQ-CA has demonstrated limited but adequate convergent and discriminant validity in studies with children and adolescents (Liu et al., 2017; Martín-Albo et al., 2018; Teixeira et al., 2015). The scale is positively correlated with measures of emotion regulation and negatively correlated with measures of anxiety and depression (Gullone & Taffe, 2012). The ERQ-CA was translated from English to Norwegian by two translators—the first author and Kyrre Svarva, who took turns evaluating each other’s suggestions until they agreed on a translation. Three girls aged 9 and 10 volunteered to help select competing item formulations, focusing on readability and understandability. The final version was back-translated by a licensed translator. Eleonora Gullone approved the back-translation.
READ (Hjemdal et al., 2006). The original READ addresses five factors; the two factors included in this study were titled social self-efficacy (S-SE) (e.g., “I easily make others feel comfortable around me”) and social support family (SSF) (e.g., “In my family we support each other”). The five response options for each range from “I totally disagree” (1) to “I totally agree” (5). READ shows adequate psychometric properties and promising validity compared with measures of mental difficulties (Askeland & Reedtz, 2015; Hjemdal et al., 2006). See Aune et al. (2023) for a more detailed description of these two scales. In this study, CAs for the Social Self-Efficacy and the Social Support scales were 0.74 and 0.78, respectively.
The IMI (McAuley et al., 1989). This is a self-report multidimensional measurement device intended to assess participants’ subjective experience related to a target activity in laboratory experiments. It has been used in several experiments on intrinsic motivation and self-regulation (e.g., Cocca et al., 2022; Deci et al., 1994; Ryan, 1982; Ryan & Deci, 2017; Ulstad et al., 2020). The instrument assesses participants’ (1) Interest/Enjoyment, (2) Perceived competence, (3) Value/Usefulness, (4) Effort/Importance, (5) Pressure/Tension, and (6) Perceived choice. This study only assesses the first three areas. The Interest/Enjoyment subscale is considered a self-report measure of intrinsic motivation. The Perceived competence concepts are theorized to be positive predictors of both self-report and behavioral measures of intrinsic motivation. The Value/Usefulness subscale is used in internalization studies (e.g., Deci et al., 1994), the idea being that people internalize and become self-regulating concerning activities that they experience as functional or valuable for themselves. The IMI consists of varied items from these subscales, all analytically coherent and stable across various tasks, conditions, and settings (Cocca et al., 2022; Deci et al., 1994; Ryan & Deci, 2017). The general criteria for including items on subscales have been a factor loading of at least 0.6 on the appropriate subscale and no cross-loadings above 0.4. The questionnaire applied in this study consists of 20 items assessed on a seven-point Likert scale, with responses ranging from 1 (not at all true) to 7 (very true). Interest/Enjoyment included seven questions (e.g., “I enjoyed doing this activity very much”). Perceived competence included six questions (e.g., “I think I am pretty good at this activity”). Value/Usefulness included seven questions (e.g., “I believe this activity could be of some value to me”).
One question had a very low factor loading in the principal component analysis (PCA) of the items covering Interest/Enjoyment. It belonged to another factor with an eigenvalue higher than one and thus was excluded. The final PCA indicated that all items related to one another were included in each of the three factors, with high eigenvalues and satisfactory factor loadings for each question. The factor loadings for each item were as follows: Interest/Enjoyment = 2.6 (factor loadings 0.51–0.87), Perceived competence = 3.4 (factor loadings 0.38–0.83), and Value/Usefulness = 3.5 (factor loadings 0.47–0.81). The reliability analyses revealed CAs of 0.51 for the Interest/Enjoyment subscale, 0.83 for the Perceived competence subscale, and 0.79 for the Value/Usefulness subscale.
Procedure and Data Collection
The questionnaire had ID numbers linked to the respondents’ names in a password-encrypted data file on OneDrive.
A pilot study examined the participants’ understanding of the 20 IMI questions. The questionnaire was completed by 17 participants (15 boys and two girls) from two football teams in a medium-sized city in Norway. The participants were 9–12 years old (SD = 10.3) and completed the questionnaire in 4 to 15 minutes (SD = 11 minutes). Seven participants noted that some of the 10 ERQ-CA questions were difficult to understand. Specifically, questions 5 and 7–10 were mentioned as challenging. Some parents who observed their children completing the questionnaire also pointed out that these questions were difficult for children aged 9–12 to grasp. However, the questions were not rewritten during this validation phase, allowing the findings to be compared with other studies using the ERQ-CA.
The data were collected at the football school on November 5 th, 2023 (Pedersen et al., 2024). The children were given the questionnaire upon arrival, with boys receiving it at noon and girls receiving it at 4:00 PM. When necessary, the children’s parents assisted the children during this process.
Data Analysis
The reported means (M) and standard deviations (SDs) for the EAQ-CA scale for each gender and age group were calculated using the IBM SPSS Statistics package 29. Additionally, internal stability (CA) and four months of test-retest reliability are examined for both genders and reported for the overall ERQ-CA scale and the CR and ES subscales. Means and SDs from the two factors according to gender and age group are reported. Bivariate correlations were calculated for the ERQ-CA items and measurements exploring resilience and intrinsic motivation to assess convergent validity. An EFA was conducted, followed by a CFA using AMOS (Version 26.0; Arbuckle, 2019). Model fit was assessed using maximum likelihood estimation with robust standard errors (MLR; Yuan & Bentler, 2000). The following fit indices were applied to determine acceptable model fit: The comparative fit index (CFI), Tucker–Lewis’s index (TLI), the adjusted goodness-of-fit index (AGFI), the root mean square error of approximation (RMSEA), the x^2 df (normed fit index), and the incremental fit index (IFI), which adjusts for sample size and degrees of freedom. According to Garver and Mentzer (1999), the TLI and CFI should be greater than 0.90 to ensure an acceptable fit. It is also recommended that the IFI be 0.90 or higher (Tabachnick & Fidell, 2007). The recommended RMSEA score is below 0.08 (Hu & Bentler, 1999). Hair et al. (2014) commented that the recommendation of a nonsignificant chi-square is a very stringent criterion, as the chi-square statistics are susceptible to sample size and models with more observed variables. With 10 items on the ERQ-CA, our sample (N = 147) falls within the methodological guidelines for factor analyses (Kyriazos, 2018).
PCA was used in conjunction with reliability analyses to examine the IMI’s factor structure and report on the scale’s internal stability and reliability.
Measurement invariance was assessed using Mplus 8.6 (Muthén & Muthén, 2017). Measurement invariance was examined based on gender (boys and girls), age group (9–10 and 11–12 years), treatment group (intervention vs. control), and longitudinally. Guidelines for measurement invariance can be found in supplemental material (Hu & Bentler, 1999).
Ethics
The Norwegian Agency for Shared Services in Education and Research (SIKT) approved the study on October 24, 2023 (#449551). The study was conducted strictly in accordance with local legislation and institutional requirements to ensure the ethical conduct of research. Prior to the commencement of the study, parents provided their written informed consent for themselves and their children to participate. This consent process ensured all participants were fully aware of the study’s purpose, procedures, potential risks, and benefits.
Results
Gender and Age Differences
Table 1 shows the means M and SDs for two emotion regulation strategies: ES and CR. The CR scale exhibited no gender difference, with t = −1.58, p > 0.05. However, a significant difference was found between genders on the ES scale, with t = 3.42, p < 0.001. The boys reported using ES significantly more than the girls. This finding is consistent with previous studies by Gullone and Taffe (2012), Teixeira et al. (2015), and Martín-Albo et al. (2018) but not with the study by Pastor et al. (2019). The original five-point Likert scale of Gullone and Taffe (2012) was used by Teixeira et al. (2015), Pastor et al. (2019), and in our study. Martín-Albo et al. (2018), Gong et al. (2022), and Liu et al. (2017) used a seven-point Likert scale. Converting the seven-point Likert scale to a five-point scale score shows similar results across the various studies, with Gong et al. (2022; CR = 18.55; ES = 11.70), Liu et al. (2017: CR = 20.64; ES = 10.92), Martín-Albo (2018; CR = 19.43; ES = 10.64), Teixeria et al. (Teixeira et al., 2015: CR = 21.39; ES = 11.31), and Gullone and Taffe (2012: CR = 21.55; ES = 10.54). Our CR score of 20.82 falls in the middle of the previously reported scores, while our ES score of 10.44 falls on the lower end. Pastor et al. (2019) and Martín-Albo (2018) assessed two Spanish samples, with the latter having a more comprehensive age range (11–17) than the former (13–14 years old). The two Chinese samples were similar to our sample age ranges: Gong et al. (2022), 7–13 years old, and Liu et al. (2017), 7–12 years old. Teixeira et al. (2015) used a Portuguese sample of adolescents aged 14–18, while Gullone and Taffe (2012) assessed a sample of Australian children and adolescents aged 10–18. However, the data from the previously reported studies provides strong evidence that ERQ-CA mean scores are consistent across cultures and age ranges, indicating that the measurement yields similar results in various settings. The finding that boys more commonly use ES as a regulation strategy than girls could have long-term consequences.
Table 1
Means (M) and standard deviations (SD) for the two emotion regulation strategies (expressive suppression and cognitive reappraisal) for the overall sample, by age group and sex (N = 147)
Samples
Emotion regulation strategies
Expressive suppression
Cognitive reappraisal
M
SD
M
SD
Overall sample (N = 147)
10.44
2.69
20.82
3.42
Boys (n = 101)
10.93
2.75
20.52
3.41
Girls (n = 46)
9.34
2.21
21.48
3.38
9 to 10 years
10.39
2.74
21.71
3.86
11 to 12 years
10.60
2.97
21.36
3.75
Reliability
For the total sample, the alpha reliability coefficient for the six-item CR scale was 0.76 and 0.77 for the four-item SE scale. For boys, the CAs for CR and ES were 0.76 and 0.78, respectively. For girls, the CAs were 0.77 for CR and 0.67 for the ES scale.
The internal consistency values in our study are higher than those reported by Pastor et al. (2019) for the CR (0.67) and ES (0.64) and higher than those reported by Teixeira et al. (2015), assessing a sample of Portuguese adolescents. In addition, the four ES scale items indicate that this ERQ-CA version also has sound internal consistency among girls.
Pearson’s product-moment coefficients for the two subscales remained moderately high over four months, ranging from 0.32 (ES) to 0.50 (CR). The four-month test–retest correlations were aligned with Gullone and Taffe’s (2012) 12-month test–retest period.
There were no significant correlations between CR and ES at either assessment point. Table 2 presents the correlations between the ERQ-CA total scale and the two subscales.
Table 2
Pearson correlations between the ERQ-CA total score and the two subscales: cognitive reappraisal (CR) and expressive suppression (ES) at assessment points I (T1) and II (T2) four months later. Skewness and kurtosis of all variables
Scales
Skewness
Kurtosis
ERQ-CA
T1
ERQ-CA
CR T1
ERQ-CA
ES T1
ERQ-CA
T2
ERQ-CA
CR T2
ERQ-CA T1
0.509
0.471
ERQ-CA CR T1
0.190
0.086
0.782**
ERQ-CA ES T1
−0.027
0.141
0.606**
−0.023
ERQ-CA T2
0.394
0.727
0.421**
0.387**
0.121
ERQ-CA CR T2
0.107
0.181
0.363**
0.499**
−0.104
0.778**
ERQ-CA ES T2
0.219
−0.207
0.254**
0.041
0.319**
0.690**
0.082
** Correlation is significant at the 0.001 level (2-tailed)
The values reported are like those reported by Gullone and Taffe (2012) but differ from those of Martín-Albo et al. (2018). While Martín-Albo et al. (2018) reported a significantly positive correlation between the CR and SE factors, both Gullone and Taffe (2012) and ours indicate no significant correlation and, for some, correlation values in a negative direction.
Factor Analyses
Applying EFA with alpha factoring and an oblique rotation revealed a − 0.036 correlation between the two factors. Because of the low correlation between the two extracted factors, a second EFA was conducted using maximum likelihood extraction and a varimax rotation method. This EFA revealed two factors with eigenvalues above 1.0. Inspection of the scree plot also clearly demonstrated two factors explaining 52.5% of the total variance. The first factor consisted of four items, identical to those constituting the ES factor proposed by Gullone and Taffe (2012). This factor explained 28.3% of the variance and had an eigenvalue of 2.83. The second factor comprises six items and explains 24.2% of the variance with an eigenvalue of 2.42. These six items are the same as those Gullone and Taffe (2012) included in the CR factor. Table 3 shows the factor matrix scores and eigenvalues for each ERQ-CA item.
Table 3
Emotion Regulation Questionnaire for Children and Adolescents (ERQ-CA) (N = 146). Means, standard deviation, skewness, and kurtosis for each item. Exploratory factor analysis (EFA): Factor loadings and communalities (extraction) for the two factors, CR and ES
ERQ-CA items
Mean (SD)
Skewness
Kurtosis
Factor loading
CR/ES
Communality
1. When I want to feel happier‚ I think about something different
3.32. (0.95)
−0.234
0.455
0.576/0.102
0.343
3. When I want to feel less bad (e.g., sad‚ angry, or worried). I think about something different.
3.43 (0.96)
−0.252
−0.416
0.733/–0.060
0.541
5. When I’m worried about something‚ I make myself think about it in a way that helps me feel better.
3.61 (0.79)
−0.391
0.231
0.608/–0.161
0.396
7. When I want to feel happier about something‚ I change the way I’m thinking about it.
3.46 (0.72)
0.203
−0.182
0.527/0.065
0.278
8. I control my feelings about things by changing the way I think about them.
3.50 (0.83)
0–.145
0.186
0.554/–0.068
0.312
10. When I want to feel less bad (e.g., sad‚ angry‚ or worried) about something‚ I change the way I’m thinking about it
3.53 (0.73)
0.249
0.724
0.587/0.107
0.356
2. I keep my feelings to myself.
3.11 (0.91)
0.274
0.668
0.081/0.716
0.519
4. When I am feeling happy‚ I am careful not to show it.
1.71 (0.85)
1.121
1.365
−0.143/0.534
0.306
6. I control my feelings by not showing them.
2.80 (0.90)
−0.115
−0.408
0.065/0.761
0.583
9. When I’m feeling bad (e.g., sad‚ angry‚ or worried). I’m careful not to show it.
2.78 (0.83)
−0.145
0.186
−.028/0.698
0.488
After four months, the same but reduced (n = 104) sample was subjected to CFA, which confirmed the two-factor structure obtained from the EFA. Fig. 1 illustrates the two-factor CFA model of the ERQ-CA. This model exhibited good model fit and parsimony: χ2 = 1.62, TLI = 0.90, CFI =.94, AGFI = 0.90, IFI = 0.94, RMSEA = 0.06. This is consistent with the two-factor model revealed by the EFA and the original two-factor structure proposed by Gullone and Taffe (2012). Furthermore, the factor loadings for the CR scale are equivalent to those reported by Gong et al. (2022), Liu et al. (2017), Teixeira et al. (2015), and Pastor et al. (2019). In contrast, the factor loadings of the ES scale are different and higher than those previously reported. This study supports the soundness of the ERQ-CA two-factor structure.
Fig. 1
CFA model of the ERQ-CA with two latent factors (CR and ES). The paths indicate the standardized regression coefficients
Validity
Our study revealed a significant positive correlation between the READ Social Self-Efficacy and Social Support Scale and the CR scale for boys. However, we found no such significant association for girls. Additionally, there were no significant correlations between the ES and the two resilience scales for both boys and girls. Our findings are consistent with those of Gullone and Taffe (2012), who used different scales, such as the CDI. It is important to note that the READ Social Support and Social Self-efficacy scales are reliable predictors of depression among children and adolescents (Hjemdal et al., 2006). Table 4 summarizes the correlations between the measurements.
Table 4
Pearson product-moment correlation coefficients between the ES and CR scales of the Emotion Regulation Questionnaire for Children and Adolescents (ERQ-CA) and the Enjoyment, Competence, and Value/Usefulness subscales from the Intrinsic Motivation Inventory Questionnaire (IMI), both assessed at pre-test, and the two Resilience Scale for Adolescents (READ) subscales: Social self-efficacy and social support, assessed at post-test
Samples/Scales
Social self-efficacy
Social support
Enjoyment
Competence
Value/Usefulness
Overall sample
ES
−0.130
0.021
−0.100
0.041
0.083
CR
0.269**
0.253**
0.250**
0.302**
0.228**
Boys
ES
−0.193
−0.010
−0.089
0.105
0.060
CR
0.340**
0.349**
0.250*
0.293**
0.126
Girls
ES
0.037
−0.055
−0.072
−0.063
0.074
CR
0.136
0.011
0.240
0.299**
0.490**
** Correlation is significant at the 0.01 level (2-tailed)
* Correlation is significant at the 0.05 level (2-tailed)
The three IMI subscales and the CR scale showed an overall statistically significant correlation. However, no significant correlations were found for the ES scale with the same IMI subscales. In contrast, girls showed relatively high and statistically significant correlations between the CR regulation scale, Value/Usefulness, and Competence, while boys showed a statistically significant correlation between CR and Competence. Although a significant correlation was shown between CR and Enjoyment for boys, the correlation coefficients for boys and girls were similar.
Measurement Invariance
Invariance testing metrics are shown in Table 5, with guidelines for invariance outlined in supplemental material. A three-step approach was used to test for measurement invariance. First, the highest factor loading on each factor was set to one, with factor means in each group set to zero (configural invariance). A model has configural invariance if the fit is adequate. Next, all factor loadings were constrained to be equal across the three groups (gender, time, and group) to test for metric invariance; changes in fit indices were compared with the configural model. Lastly, factor loadings and intercepts were constrained to be equivalent across the three groups, and scalar invariance was tested by comparing this model with the metric invariance model. Chen (2007) notes that χ2 tests for invariance are sensitive to sample size and violations of normality, so small discrepancies may result in model rejection. Thus, Chen outlined guidelines for determining invariance relying on other model fit indices for larger samples with unequal group size and/or non-normal data (Chen, 2007). Specifically, (1) a change of ≤ −0.005 in CFI and a change of ≥0.010 in RMSEA OR (2) a change of ≥0.025 in Standardized Root Mean Square Residual (SRMR), in a comparison of configural with metric models would indicate metric noninvariance. For scalar invariance (comparing the scalar with the metric model) [1], a change in Comparative Fit Index (CFI) ≥ −0.005, and a change in Root Mean Square Error of Approximation (RMSEA) of ≥0.010 OR [2] a change in SRMR ≥0.005 indicates scalar noninvariance. The configural invariance model examines whether the factor structure is similar across the three groups.
Table 5
Measurement invariance and Longitudinal measurement invariance by pre- and post-test
1Gender: Pre-test
RMSEA
CFI
TLI
SRMR
Configural invariance
0.089
0.883
0.845
0.090
Metric invariance
0.096
0.849
0.821
0.117
Δ from configural
+0.007
+0.034
−0.024
+0.027
Scalar invariance
0.101
0.812
0.799
0.132
Δ from metric
+0.005
−0.037
−0.022
+0.015
Gender: Post-test
Configural invariance
0.133
0.813
0.753
0.105
Metric invariance
0.124
0.818
0.785
0.119
Δ from configural
−0.009
+0.005
+0.032
+0.014*
Scalar invariance
0.119
0.817
0.804
0.120
Δ from metric
−0.005
0.001
+0.019
+0.001*
2Age Group: Pre-test
Configural invariance
0.083
0.891
0.855
0.090
Metric invariance
0.092
0.850
0.823
0.111
Δ from configural
+0.009
−0.041
−0.032
+0.021*
Scalar invariance
0.084
0.862
0.852
0.116
Δ from metric
−0.008
+0.012
+0.029
+0.005
Age Group: Post-test
Configural invariance
0.095
0.861
0.816
0.097
Metric invariance
0.085
0.877
0.854
0.110
Δ from configural
−0.01
+0.016
+0.038
+0.013*
Scalar invariance
0.082
0.872
0.863
0.116
Δ from metric
−0.003
−0.005
−0.009
+0.006
3Treatment Group: Pre-test
Configural invariance
0.077
0.906
0.875
0.077
Metric invariance
0.081
0.881
0.860
0.117
Δ from configural
+0.004
−0.025
−0.015
+0.04
Scalar invariance
0.079
0.878
0.869
0.124
Δ from metric
−0.002
−0.003
+0.009
+0.007
Treatment Group: Post-test
Configural invariance
0.132
0.748
0.667
0.108
Metric invariance
0.116
0.784
0.745
0.123
Δ from configural
−0.1
+0.036
+0.078
+0.015
Scalar invariance
0.111
0.783
0.767
0.132
Δ from metric
−0.005
−0.001
+0.022
+0.009
Longitudinal Measurement Invariance by Pre- and Post-test
Configural invariance
0.129
0.466
0.400
0.166
Metric invariance
0.128
0.455
0.401
0.177
Δ from configural
−0.001
−0.011
+0.001
+0.011*
Scalar invariance
0.123
0.459
0.447
0.181
Δ from metric
−0.005*
+0.004
+0.046
+0.004*
Note. SRMR Standardized Root Mean Square Residual, RMSEA Root Mean Square Error of Approximation, CFI Comparative Fit Index. 1Gender: 1 = Boys; 2 = Girls. 2Age group: 1 = 9–10 years; 2 = 11–12 years. 3Treatment group: 1 = Intervention; 2 = Control. *Variance found
For pre-test measurement invariance by gender, the model fit was adequate, χ2(68) = 109.023, p = 0.0012, CFI = 0.883, RMSEA = 0.089 (90% CI = 0.056, 0.119), SRMR = 0.090. No change in fit occurred relative to the configural model, indicating metric invariance for pre-test by gender. Additionally, there were no substantial decreases in model fit relative to the metric model, after constraining factor loadings and item intercepts to be identical and allowing factor means to vary between genders at pre-test, indicating no scalar variance. However, in the post-test, constraining factor loadings to be equivalent across genders resulted in metric and scalar variance per standard cutoff criteria for SRMR (≥ 0.025 configural to metric and ≥ 0.005 metric to scalar). Therefore, variance by gender is seen in post-test participants.
For the age grouping, variance was found in both pre-and post-test measures. At pre-test, the model fit was adequate, χ2(68) = 104.175, p = 0.0031, CFI = 0.891, RMSEA = 0.083 (90% CI =0.049, 0.114), SRMR = 0.090 for age group. Constraining factor loadings to be equivalent across age groups resulted in metric and scalar variance per standard cutoff criteria for SRMR (≥ 0.025 configural to metric and ≥ 0.005 metric to scalar) at both pre- and post-test.
For pre-test measurement invariance by treatment group, the model fit was adequate: χ2(68) = 98.556, p = 0.0091, CFI = 0.906, RMSEA = 0.077 (90% CI = 0.039, 0.108), SRMR = 0.077. No change in fit occurred relative to the configural model, indicating metric invariance. There were no substantial decreases in model fit relative to the metric model after factor loadings and item intercepts were constrained to be identical. Factor means varied across treatment groups in the post-test, indicating scalar invariance.
After factor loadings were constrained to be equivalent across the intervention group at the post-test, no change in fit occurred relative to the configural model, indicating metric invariance. Finally, no substantial decreases in model fit relative to the metric model, indicating no scalar variance.
Longitudinal Invariance
Longitudinal invariance testing metrics are shown in Table 5. The model revealed variance when tested for longitudinal invariance across time (Table 5). Constraining factor loadings to be equivalent across time resulted in changes to fit relative to the configural model, indicating metric variance. Finally, there was a substantial decrease in model fit relative to the metric model after factor loadings, item intercepts were constrained to be identical, and factor means were allowed to vary across time, indicating scalar variance.
Discussion
Our study’s findings have significant implications for the validity of the ERQ-CA. The ten items of the ERQ-CA are divided into two distinct factors: CR and ES. A CFA confirmed that this factor structure aligns well with the theoretically and empirically proposed two-factor model. The internal stability (CA) for the two subscales (CR and ES) demonstrated satisfactory values for both girls and boys. The relatively low four-month test-retest correlation (0.32) for the ES scale may be attributed to the rapid pace of development among children in this age group. CR has been associated with lower levels of negative affect and psychopathology, while ES has been linked to increased physiological stress and social impairment (Kneidinger et al., 2024). The results of our study support these suggestions, showing that the CR factor is significantly positively associated with well-recognized resilience features, such as experience of familial social support and social self-efficacy. In contrast, the ES factor was negatively associated with these resilience factors. These findings bolster the validity of the ERQ-CA. Furthermore, our examination of the overall sample revealed a statistically significant positive correlation (P < 0.01) between CR and the experience of Enjoyment, Competence, and Value/Usefulness assessed by the IMI. Specifically, the Value/Usefulness subfactor has been associated with the ability to regulate (Deci et al., 1994). The Value/Usefulness subscale is used in internalization studies (e.g., Deci et al., 1994), which propose that people internalize and become self-regulating concerning activities that they perceive as functional or valuable for themselves. More intriguingly, private self-consciousness (self-reflection and insight) was found to be a central antecedent for the use of CR. Our results demonstrate a relatively high correlation between CR and the IMI Usefulness/Value subfactor, especially among girls, indicating that self-reflection regarding the values of an activity should be targeted.
The findings from this study and previous research strongly suggest that mean ERQ-CA scores are consistent across different cultures and age groups, indicating that the measurement is experienced similarly across various settings. This global applicability ensures that researchers and professionals from diverse cultural and geographical backgrounds can benefit from the insights provided. Furthermore, the discovery that boys use significantly more ES as a regulatory strategy compared with girls could have long-term consequences and needs further exploration. Baars et al. (2017) have shown that problem-solving and learning are linked to self-regulation and motivation. The way behaviors related to these perceptions work in tandem must be investigated in the future.
Regarding measurement invariance, at the post-test, gender had metric and scalar variance according to the standard cutoff criteria for SRMR (≥ 0.025 configural to metric and ≥ 0.005 metric to scalar). Similar patterns were observed between age groups in both the pre- and post-test. Additional variance was observed longitudinally, indicating a change from pre- and post-study time points. These mixed findings could be due to the small sample size, short time frame between pre- and post-tests, and the possibility that the intervention affected boys and girls differently based on age. Findings highlight the importance of measuring emotion regulation in children across these groups.
Strengths and Limitations
Our sample (N = 147) of children enrolled in a tuition-free football [soccer] program is both a strength and a limitation. The sample offers a unique opportunity to examine emotion regulation in a real-world, group-based, physical activity context, which is relevant for interventions targeting children’s mental health and well-being. Although the football school was open for every young kid in a specific municipality to attend tuition-free, there is always a question of the generalizability of the findings to a broader population. However, Juul et al. (2020) have shown that 93% og young kids in Norway do participate in sports activities outside schools. Football is by far the most popular sports activity both among girls and boys. Limitations also include having a relatively small number of girls compared with boys, and the data relied solely on children’s self-reports. Nonetheless, a sample of the size used in this study, using a 10-item inventory, is sufficient for factor analyses (Field, 2013). Testing measurement invariance indicates that the ERQ-CA emotion regulation construct is interpreted and measured similarly across gender and intervention groups.
It would be beneficial to include another measurement that also assesses self-regulation. On the positive side, this study is based on pre-validated questionnaires, and a pre-test was conducted to ensure their validity (Thomas et al., 2023). However, a weakness is that although the questionnaires have been validated, the IMI has rarely been used on such young children. Additionally, the pre-testing of the questionnaire revealed that some children found it difficult to answer some of the questions in the ERQ-CA. Nonetheless, the experience from the test days indicates that the participants were familiar with the wording of the items included in the questionnaire.
Theoretical and Practical Implications
Strengthening the theoretical framework of emotion regulation proposed by Gross and Thompson (2007), this study confirms the two-factor structure (cognitive reappraisal and expressive suppression) in a Norwegian sample, adding to the growing body of evidence supporting the universality of these two fundamental emotion regulation strategies across different cultures. Further, applying the ERQ-CA to children aged nine to 12 provides valuable insights into the development of emotion regulation strategies in younger age groups. This act extends our theoretical understanding of when and how these strategies emerge and evolve during childhood, potentially informing developmental models of emotion regulation. Moreover, accurately assessing emotion regulation strategies in younger children enables earlier identification of potential emotion regulation difficulties. This assessment informs the creation of targeted interventions to enhance emotion regulation skills at a crucial developmental stage, potentially preventing more serious emotional disorders later. Understanding the relationship between emotion regulation strategies and other psychological factors like resilience, self-efficacy, and intrinsic motivation can guide the development of more comprehensive educational programs and clinical interventions. Lastly, the study examines measurement invariance across different groups (e.g., gender, age), giving practitioners confidence in the measure’s stability across these demographics and allowing for more accurate comparisons and interpretations of ERQ-CA scores across diverse children.
In conclusion, the findings significantly contribute to the theoretical understanding of emotion regulation in children and its practical application in clinical, educational, and research settings. By providing a validated tool for assessing emotion regulation in younger children and demonstrating its links to other important psychological constructs, this research paves the way for more effective interventions and a deeper understanding of children’s emotional development.
Acknowledgments
We thank all the participants in this study, especially the children and their parents. We also thank Gunnhild Hammer and Tove Dyrstad for their generous support and effort.
Declarations
Conflict of Interest
None of the participants.
Approval
All authors have read and approved the submitted manuscript.
Ethical Considerations
This research was performed by the code of ethics of the World Medical Association (Declaration of Helsinki) for experiments involving humans. The Norwegian Agency for Shared Services in Education and Research (SIKT) approved the study on October 24, 2023 (#449551). The study was conducted strictly with local legislation and institutional requirements to ensure the ethical conduct of research. Before the commencement of the study, parents provided their written informed consent, permitting both themselves and their children to participate. This consent process ensured all participants were fully aware of the study’s purpose, procedures, potential risks, and benefits.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Met BSL Psychologie Totaal blijf je als professional steeds op de hoogte van de nieuwste ontwikkelingen binnen jouw vak. Met het online abonnement heb je toegang tot een groot aantal boeken, protocollen, vaktijdschriften en e-learnings op het gebied van psychologie en psychiatrie. Zo kun je op je gemak en wanneer het jou het beste uitkomt verdiepen in jouw vakgebied.
