The population of the nationwide study comprises Muslim adolescents in juvenile rehabilitation residential schools, ages between 13 and 21 years old. The target sample of the current study is Muslim adolescents between 14 and 18 years old undergoing rehabilitation programs in all eight juvenile rehabilitation residential schools in Malaysia. A total of 440 Muslim adolescents were selected through stratified random sampling; the population was stratified according to their gender. The present sample includes 268 male (60.9%) and 172 female (39.1%) participants, most between 16 and 17 year old (52%; mean age = 16.71 years; Sd. = 1.14). Participants reported that their parents were in their forties (Father: Mean = 47.8, Sd. = 9.5; Mother: Mean = 43.04, Sd. = 7.6) and almost all of them (93.9%) came from a family with low socio-economic status or low parent monthly income (< USD 1052).

Ethical clearance from the ethics committee of University Putra Malaysia (UPM) was obtained as the first step in the data collection process, followed by submission of the research proposal and questionnaires to the Department of Social Welfare for data collection permission. Next, all juvenile rehabilitation residential schools in Malaysia were informed about the list of Muslim juvenile residents and written consent. A pilot test was first conducted among randomly selected Muslim respondents in the rehabilitation schools who were later excluded from the actual study. Then, the questionnaires went through several modifications, such as item deletion and restructured of items, and were distributed to all Muslim respondents in the actual study. During the actual data collection, respondents were split into smaller groups and they completed the questionnaire in roughly 30 minutes, with a 100% return rate.

Data analysis was conducted using the Statistical Package for the Social Sciences (SPSS) version 21.0 and Analysis of Moment Structure (AMOS) 20.0. Preliminary analyses were applied in order to examine the descriptive statistics and correlations among variables of interest. For multivariate analysis, structural equation modeling (SEM) with the maximum likelihood method was utilized. The model goodness of fit was tested using several fit indices (Kline, 2005). The recommended indices to estimate goodness of fit in the path analyses are the normed chi-square (χ²/df ) < 5, the Root Mean Square Error of Approximation (RMSEA) < 0.05, Goodness of Fit Indices (GFI) > 0.90, and Comparative Fit Index (CFI) > 0.90 (Kline, 2005). A full structural model encompassing parceled variables of secure and insecure maternal attachment subscales, self-regulation, and cognitive distortion was assessed. Mediation analysis consuming bootstrapping estimation procedure (Preacher & Hayes, 2008) was applied to examine the mediating effect of adolescent self-regulation through which maternal attachment (secure and insecure) was linked with cognitive distortion.

The results of the descriptive and correlational analysis for all variables of interest are presented in Table 1. Significant correlations were observed for both secure and insecure maternal attachment with self-regulation and with cognitive distortion. Secure maternal attachment and self-regulation were negatively related with cognitive distortion (secure: r = −.11, p < .05; self-regulation: r = −.33, p < .01). Insecure maternal attachment was also found to be negatively correlated with self-regulation (r = −.23, p < .01). There was a significantly positive relationship between insecure maternal attachment and cognitive distortion (r = .10, p < .05) and a significant positive relationship between secure maternal attachment and self-regulation (r = .22, p < .01). In general, both independent variables (secure and insecure maternal attachment) were significantly associated with self-regulation and cognitive distortion, in which the directions of associations were as predicted with acceptable magnitude ranged from 0.10 to 0.33. These results also suggest that Malaysian Muslim juvenile offenders with poor maternal attachment and lower level of self-regulation were more likely to have higher degree of cognitive distortion.

The first step in conducting structural equation modeling (SEM) to examine the structural relations among the study variables is through the validation of respective measurement model of all latent variables. The results of Confirmatory Factor Analysis (CFA) suggested that all measurement model fit indices reflected satisfactory model fit. Next, parceling technique was applied as some of the study variables possessed large number of items and structural equation modeling was found to be more convenient if indicators were lesser (Little, Cunningham, Shahar, & Widaman, 2002).

The random assignment technique was utilized to parcel the latent variables into a few indicators. Particularly, items were randomly selected to form the parcels in each construct. All in all, the fit indices of the parceled model of secure maternal attachment (χ^² = 2.63; df = 2; χ^²/df = 1.32; RMSEA = .03; CFI = .99; GFI = .99), insecure maternal attachment (χ^² = 4.50; df = 2; χ^²/df = 2.25; RMSEA = .05; CFI = .99; GFI = .99), self-regulation (χ^² = 4.58; df = 2; χ^²/df = 2.29; RMSEA = .05; CFI = .99; GFI = .99), and cognitive distortion (χ^² = 2.88; df = 2; χ^²/df = 1.44; RMSEA = .03; CFI = .99; GFI = .99) proposed sufficient in overall goodness of fit statistics. Finally, the full structural model (as illustrated in Figure 1) was evaluated based on the recommended goodness-of-fit.

According to the goodness-of-fit indices, the initial structural model was found to fit well to the data (χ^² = 183.12; df = 98; χ^²/df = 1.87; p < 0.001; RMSEA = 0.05; CFI = 0.96; GFI = 0.94). Even supposing that the structural model fitted the data well, a potential better-fit statistics for the data could be achieved by examining the modification indices. Therefore, error term 1 and 5 of maternal attachment was suggested to be correlated due to the possibility of the similar item content or wording in each parcel. Consequently, the goodness-of-fit indices revealed an improvement with χ^² = 164.46; df = 97; χ^²/df = 1.70; p < 0.001; RMSEA = 0.04; CFI = 0.97; GFI = 0.95, confirming an excellent model fit to the data. Hence, a final (revised) structural model as depicted by Figure 1 was attained. Besides, squared multiple correlation coefficient exposed that the final (revised) model accounted 24% of the variance in cognitive distortion and 2% of the variance for self-regulation.

Significant correlations were revealed through the examination of the final (revised) model. Independent variables of insecure maternal attachment (B = .12, p = 0.040) and self-regulation (B = −.45, p = .001) were indicated to possess direct and small influence on cognitive distortion. However, the path between secure maternal attachment and cognitive distortion was found to be insignificant. On the other hand, secure maternal attachment had a small (B = .15, p = 0.030) influence on self-regulation, while insecure maternal attachment was reported to possess insignificant relationship with self-regulation.

Bootstrapping method was applied for mediation analysis to investigate the significance of self-regulation as a mediating effect in the final structural model. Indirect effect can be claimed as significant when zero is not in the lower and upper limit of the 95% confidence intervals (Preacher & Hayes, 2008). The result of bias-corrected confident interval bootstrapping of the indirect model is presented in Table 2. However, only indirect effect of secure maternal attachment on cognitive distortion via self-regulation (SIE = −.07, SE = .04, p < 0.05, bootstrap 95% CI = −.15 to −.01) was statistically different from zero, which suggested that there was a significant mediating effect via self-regulation. All in all, the findings of mediation analysis revealed that secure maternal attachment with the mediating effect of self-regulation were considered as significant factors that influenced cognitive distortion among Malaysian Muslim juvenile offenders. Moreover, the findings of this study highlighted the roles of secure maternal attachment and self-regulation in controlling cognitive distortion among the institutionalized study population.

The results of the current study revealed a significant negative relationship between insecure maternal attachment and cognitive distortion, and is consistent with past studies (Dykas & Cassidy, 2011; Mason, Platts, & Tyson, 2005). This finding indicates insecure maternal attachment resulted in a greater degree of cognitive distortion among Malaysian Muslim juvenile offenders. In other words, the level of cognitive distortion among Malaysian Muslim juvenile offenders was determined by the quality of mother-adolescents attachment. Attachment relationships built from early childhood created internal working model that comprised positive and negative attachment-relevant memories or mental schemas (Pietromonaco & Barrett, 2000; Thompson, 2008). Insecure attachment relationships with mothers contribute to negative internal working models or weak attachment schemas which may lead to a negative perception of self, others, and the world (Bowlby, 1963). Weak attachment schemas may strengthen the Malaysian Muslim juvenile offenders’ negative schemas, which study results reflect in their distorted perception and later demonstrated through their misbehavior (Amani, Majzoobi, & Azadi Fard, 2017).

In contrast, secure maternal attachment was significantly positively related to self-regulation among research participants. The finding is consistent with earlier studies (Calkins & Leerkes, 2004; Meriac, Slifka, & LaBat, 2015) emphasizing that better self-regulation may be explained by a strong mother-adolescent attachment. The current finding supports the opinion by Kopp (1982) regarding attachment and self-regulation. Self-regulation in adolescents was found to begin formation in attachment with their caregivers during childhood and also known as a developmental capacity that was formed via interactions with caregivers (Calkins & Leerkes, 2004). For that reason, the feeling of security from mothers as a secure base among study participants was found to be essential in the development of better self-regulation in them. In other words, securely attached Malaysian Muslim juvenile offenders were able to develop advanced self-regulatory capacities due to availability and emotional support from their caregivers during distressing situations (Karreman, Van Tuijl, van Aken, & Deković, 2006).

Findings from the current study reveal that self-regulation was significantly inversely related to cognitive distortion among Malaysian Muslim juvenile offenders. Results from the current study extend prior literature (Volling et al., 2002; Muraven & Baumeister, 2000) by suggesting that the presence of good self-regulation could lead to lesser degree of cognitive distortion as self-regulation acts as a controller of emotion, cognition, and behaviors. Other findings support self-regulation being related to cognitive outcomes. Van der Merwe and Dawes (2000) described children involved in the violence being stimulated by the failure in controlling negative cognitions and possessing self-regulatory inconsistencies. Raffaelli and colleagues (2005) suggested that risks related to cognitive and behavior problems can be reduced if the adolescents are capable to control their behavior, attention, and emotion. Thus, self-regulation is important as a protective factor against cognitive-based psychopathology and problem behaviors (Gestsdottir & Lerner, 2008).

Although the study has achieved its aims, there were some unavoidable limitations. Results of the current study were based on self-reports provided by respondents about their past experiences with their mothers, which can be assumed to possess potential biases. With this in mind, future investigation should refer to the information collected from multiple informants. Both caregivers and adolescents can be recruited and the information may be more accurate as compared to single informant.