Bowlby’s (1988) attachment theory highlights that individuals’ emotional experiences with primary caregivers are associated with their capacity to connect with others and cope with affective or stressful situations. Armsden and Greenberg’s (1987) theoretical model of attachment considers secure attachment to be characterized by high levels of trust and communication with others and low levels of alienation. Trust and communication reflect understanding and mutual respect, whereas alienation refers to feelings of isolation. These dimensions are shaped in the relationships with caregivers (Kerns et al., 2015). It is now widely accepted that both parents are attachment figures but may play different roles (Bacro, 2012). Classical attachment theory has traditionally emphasized the role of the mother as the primary attachment figure (Bowlby, 1969, 1984). However, recent research also highlights the relevance of the father as a significant figure in the development of attachment within the father-child relationship (Cabrera et al., 2018).

Parental secure attachment has been linked to better social, emotional, and academic functioning, which may extend beyond adolescence (Holt et al., 2018). Adolescents with secure parental attachment learn to regulate emotions and to understand the need to control them, expressing them more positively (Dane & Marini, 2014).

Secure parental attachment also facilitates the establishment of secure peer relationships (Zou et al., 2020), indicating close ties between parental and peer attachment (McDowell & Parke, 2009). Parental warmth appears to be especially valuable in helping adolescents feel loved and appreciated by the family (Ertema et al., 2025), develop a secure attachment (Cornella-Font et al., 2020), and achieve good adaptation (García et al., 2020; Krauss & Orth, 2024), this being especially relevant in the academic field (Fuentes et al., 2019; Reyes et al., 2023). Conversely, adolescents with insecure parental attachment experiences face greater difficulties in establishing peer relationships (Ward et al., 2018) and may even be at greater risk of affiliating with deviant peers (Kliewer et al., 2018). Thus, during adolescence, the role of peers is added to the predominant role of parental attachment figures, which contributes reinforcement and well-being (Laghi et al., 2016). Regarding sex differences, adolescent girls show higher peer attachment, as peer influence begins earlier for them than for boys (Miljkovitch, 2021).

Emotional regulation refers to the process by which individuals consciously or unconsciously modulate their emotions to respond appropriately to contextual demands; it includes managing both positive and negative emotions, either by decreasing or enhancing their impact (McRae & Gross, 2020). The ability to regulate emotions according to goals and context allows children and adolescents to cope with daily challenges (Oram et al., 2017).

The theoretical model of emotional regulation divides strategies into two phases (McRae & Gross, 2020). The first phase involves antecedent-focused strategies, i.e., those activated before the emotional experience occurs, such as cognitive reappraisal (Gross & John, 2003). The second phase involves response-focused strategies, i.e., strategies implemented after the emotion is triggered, including expressive suppression (Gross & John, 2003).

Research in this field has shown that cognitive reappraisal use is associated with more adaptive outcomes, such as better physical health (Appleton et al., 2014), higher academic performance (Ivcevic & Brackett, 2014), and more adaptive social relationships (English & Eldesouky, 2020). Adjusted behavior requires effective regulation and control of both positive and negative emotions. The greater the ability to manage and sustain positive emotions, the greater the resources to buffer against negative ones and the better the preparation to promote positive behaviors that may result from environmental situations (Caprara et al., 2010). Conversely, expressive suppression is associated with lower well-being and greater difficulties in establishing interpersonal relationships (Cameron & Overall, 2018; English & Eldesouky, 2020).

There is a developmental evolution in the use of regulation strategies: response-focused strategies are more common in childhood, while antecedent-focused ones tend to increase with maturation and better emotional management (Stifter & Augustine, 2019). During adolescence, emotional regulation becomes increasingly complex due to developmental changes (Denham, 2019).

Regarding sex differences, results are inconclusive. In adults, women seem to have more difficulties with emotional regulation (Momeñe et al., 2017). In adolescents, it remains unclear whether sex plays a differential role in the use of regulation strategies (Kokkinos et al., 2019). For cognitive reappraisal, some studies found no significant differences between adolescent boys and girls (Gross & John, 2003; Gullone & Taffe, 2012), whereas others reported lower levels in boys (Gullone et al., 2010). Cognitive reappraisal showed a strong relationship with emotional self-concept, as men generally reported a higher emotional self-concept than women (Alcaide et al., 2025; García et al., 2024; Martin-Blesa et al., 2024). Boys also show greater use of expressive suppression (Gullone & Taffe, 2012), possibly linked to their greater impulsivity in social relationships and tendency toward externalizing behaviors such as aggression (Knyazev, 2004). Men show greater difficulty in adequately expressing emotions compared to women (Reyes et al., 2023; Villarejo et al., 2024). Recent studies in the Spanish adolescent population show that boys obtain higher scores in cognitive reappraisal and expressive suppression, while girls obtain better scores in emotional self-efficacy skills, both positive and negative (Doménech et al., 2025). These findings suggest that educational programs focusing on teaching cognitive reappraisal strategies and adaptive emotional management could be especially beneficial for promoting adolescent well-being and academic adjustment, considering sex differences in emotional regulation.

In recent years, there has been growing interest in analyzing the relationships between attachment and emotional regulation at different stages of the life cycle (Chen et al., 2019; Kokkinos et al., 2019; Morris et al., 2017). Cognitive reappraisal and expressive suppression are also related to emotional self-efficacy (Doménech et al., 2024). Children learn to manage their emotions based on parental attitudinal models that they later apply in other contexts (Denham, 2019; Leerkes & Bailes, 2019; Morris et al., 2017). Parenting is inherently an emotional task (Leerkes & Augustine, 2019). However, beyond the perceptions and attitudes associated with emotional regulation, it is necessary to analyze the influence parents exert on such regulation and on academic functioning (Oram et al., 2017).

Machado and Duarte (2014) found that secure paternal attachment is associated with more effective emotional regulation strategies. Conversely, insecure attachment predicts more maladaptive emotional management, both in adolescence (Machado & Duarte, 2014) and in emerging adulthood (Tani et al., 2018). Positive family relationships during adolescence are associated with greater emotional regulation skills (Herd et al., 2022). Affective attachment bonds with parents have implications for children’s emotional development (Cooke et al., 2019; Karreman & Vingerhoets, 2012). Although research traditionally focused on the mother’s role, recent studies in children indicate that the combined influence of attachment security with both parents predicts better emotional regulation outcomes than the specific contributions of each parent separately (Fernandes et al., 2021). Attachment security to both mother and father was negatively associated with the limited use of emotional regulation strategies (Gambin et al., 2021).

Furthermore, recent studies have concluded that secure attachment to mothers and fathers is positively related to cognitive reappraisal and negatively related to expressive suppression in U.S. preadolescents (Abtahi & Kerns, 2017), Chinese adolescents (Chen et al., 2019), and Greek adolescents (Kokkinos et al., 2019). Peer attachment fosters close social behavior requiring more adaptive regulation strategies, such as cognitive reappraisal, which directs the emotional state toward the common good (Sabatier et al., 2017). Secure attachment is associated with regulation strategies that favor adaptation and effective emotional management, such as cognitive reappraisal (Huang et al., 2022). Nevertheless, culture and socioeconomic circumstances may shape the role of parents as facilitators of self-regulation (Grolnick et al., 2019).

Academic self-efficacy is a crucial component of motivation and learning theories and can be defined as an individual’s beliefs about their own abilities regarding learning or carrying out educational activities (Khine & Nielsen, 2022). It is related to aspects such as school performance, motivation, academic activities, work pace, or school attendance (Tur-Porcar et al., 2019). According to Khine and Nielsen (2022), increasing attention is being paid to the measurement and research on academic self-efficacy.

Secure maternal attachment has been associated with better academic performance and adjustment (Furrer & Marchand, 2020). In addition to maternal attachment, paternal, and peer attachment have also been linked to improved academic, social, and emotional functioning (Holt et al., 2018), partly because parents’ emotional expression can help regulate and manage negative emotions to foster academic success (Moed et al., 2017). Thus, alongside the role of parents, peer groups act as mechanisms of personal protection and school connectedness (Oldfield et al., 2018). Other authors have directly linked paternal attachment to academic self-efficacy in children (Bacro, 2012). During adolescence, attachment is connected with other dimensions that affect academic self-efficacy, such as motivation and school participation (Duchesne & Larose, 2007). Secure attachment has been positively associated with school motivation, from which academic self-efficacy derives (Yang et al., 2024). Emotional competence learned during childhood and adolescence supports the resolution of tasks that facilitate academic and social mastery (Denham, 2019). By contrast, emotional dysregulation may act as a risk factor for academic success (Oram et al., 2017). Students reporting secure attachment show greater confidence, stronger belief in their academic decisions, and in the correctness of those decisions (Wright et al., 2014). Peer support creates a favorable climate for participation and learning in school (Villar et al., 2025).

As can be observed, most studies link attachment with academic outcomes, school participation, motivation, or academic adjustment. All these factors influence academic self-efficacy, which, according to Khine and Nielsen (2022), is associated with student engagement, study habits, learning styles, and personality, and predicts and explains a wide range of student activities and achievements. Attachment—whether parental, peer, or romantic—is closely linked to the sense of academic self-efficacy (Macakova & Wood, 2022).

Regarding academic self-efficacy and sex differences, adolescent girls appear to show higher levels of academic efficacy and performance than boys, as well as a more positive evaluation of their studies (Instituto Nacional de Evaluación Educativa [INEE, 2015]). Sex moderates the association between attachment security and academic motivation (Yang et al., 2024).

This study integrates parental (mother/father) and peer attachment with emotion regulation and perceived academic self-efficacy in middle to late adolescence (15-18 years), a period characterized by marked maturational changes. Our aim is to clarify sex specific patterns of association consistent with prior evidence on the distinct parental roles and the rising salience of peers in adolescence.

This research has two objectives:

Since fathers and mothers may act and influence their children differently, and sons and daughters may also respond and influence their parents differently, as explained by Bornstein (2013), the study examines the four types of dyads (mother-son, father-son, mother-daughter, and father-daughter).

Two hypotheses are formulated:

Of the 750 students invited, 703 participated, yielding a response rate of 93.7%. No participants withdrew during data collection. Finally, a total of 703 adolescents participated: 351 boys (49.9%) and 352 girls (50.1%), aged 15 to 18 years (M = 15.86, SD = 0.30). The age distribution was as follows: 15 years: 251 (35.7%), 16 years: 325 (46.2%), 17 years: 98 (13.9%), and 18 years: 29 (4.1%). Sex was recorded based on biological criteria (male/female) as reported by participants. Although gender identity was not assessed in this study, all procedures were conducted with respect for participants’ self-perception and diversity, in accordance with ethical standards for research with minors. SAGER guidelines were followed, differentiating between sex (biological attributes) and gender (social roles and identities) in the design, analysis, and interpretation of data. Participants were enrolled in the final year of compulsory secondary education 349 (49.6%) and in high school 354 (50.4%). Participants were recruited using stratified convenience sampling across eight secondary schools (public and private-subsidized) in the province of Valencia. Of these, 348 (49.5%) attended public schools and 355 (50.5%) private-subsidized schools. Regarding family background, 652 (92.7%) were Spanish, 10 (1.4%) from other European countries, 36 (5.1%) from Latin America, 2 (0.3%) from North Africa, and 3 (0.4%) from Southeast Asia.

The sample was selected based on the classification of secondary education centers in the province of Valencia. Participation was preceded by authorization from the educational centers, families, and students, who provided informed consent. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki and national regulations on research with minors. It was carried out within the official Doctoral Program in Psychological Research at the University of Valencia (Faculty of Psychology), under academic supervision and institutional oversight. All procedures ensured voluntary participation, anonymity, confidentiality, and informed consent from legal guardians and adolescents. Students could withdraw at any time, but none chose to do so. No missing data were recorded. Data collection sessions lasted 40–45 minutes.

First, descriptive statistics were explored, followed by t-tests to examine possible sex differences in the analyzed variables and correlational analyses among them. Subsequently, a structural equation model was estimated using confirmatory techniques. Predictor variables included attachment to peers, mothers, and fathers, to observe their effects on outcome variables: cognitive reappraisal, expressive suppression, and perceived academic self-efficacy.

This model was estimated using raw scores for the variables and a robust estimator for quantitative variables (MLR). Given deviations from multivariate normality, the robust MLR estimator (maximum likelihood robust) was applied in Mplus. This estimator adjusts standard errors and fit indices for non-normal data, ensuring greater accuracy in parameter estimation and model interpretation. Statistical analyses were conducted using SPSS 24.0 and Mplus 8 (Muthén & Muthén, 1998-2017). Prior to the analyses, the absence of missing data was verified through visual inspection of the database and confirmation in SPSS (complete frequencies for all variables). Therefore, no imputation procedures or MCAR tests were required A cross-sectional design was employed due to feasibility constraints and the aim of exploring associations among attachment, emotional regulation strategies, and academic self-efficacy at a specific developmental stage. While this design does not allow for causal inference or temporal sequencing, it provides valuable insights into concurrent relationships and informs hypotheses for future longitudinal research. Multivariate normality was assessed using Mardia’s coefficient, and given deviations from normality, the robust MLR estimator (Maximum Likelihood Robust) was applied. The model was estimated using observed scores (rather than latent variables) for reasons of parsimony and statistical power. Considering that the aim of the study was to analyze structural relationships for applied purposes rather than to validate the internal structure (already supported by previous research), this approach reduced the number of parameters and improved the stability of estimates. Given the high correlation between maternal and paternal attachment (r ≈ .70), multicollinearity was assessed by calculating the variance inflation factor (VIF) and tolerance. The values were acceptable (maternal attachment: VIF = 1.97, tolerance = .51; paternal attachment: VIF = 2.00, tolerance = .50; peer attachment: VIF = 1.07, tolerance = .94), indicating no problematic collinearity. A two-factor MANOVA (Sex × Age group: 15-16 vs. 17-18) was conducted on secure attachment (mother, father, peers), cognitive reappraisal, expressive suppression, and perceived academic self-efficacy. Multivariate and univariate tests were examined, and effect sizes are reported as partial η². Assumptions were checked (Box’s M for homogeneity of covariance matrices; Levene’s tests for homogeneity of variances). Given multivariate non-normality, the SEM analyses used MLR; MANOVA results are reported with standard GLM assumptions.

A leptokurtic distribution is observed, more pronounced in the attachment variables (peers, mothers, and fathers) and less accentuated in cognitive reappraisal and perceived academic efficacy, which indicates a concentration around the mean values. These same variables show a longer left tail (negative skewness). In expressive suppression, skewness is positive; therefore, there is less clustering around the mean values, with more scores above the mean (Table 1).

The independent-samples Student’s t-test, accompanied by effect size using Cohen’s d (1992), was conducted to examine the strength of the differences as a function of sex.

The results show a significant difference in peer attachment and perceived academic efficacy, favoring females. Significant differences also emerged in cognitive reappraisal and expressive suppression, favoring males. The effect size, according to Cohen’s (1992) d, falls within the small to medium range (Table 2). Attachment to mothers and fathers did not show significant differences, indicating that adolescent boys and girls perceive their attachment relationships with both parents in a similar manner.

With regard to the dimensions of emotion regulation, boys reported higher scores in both cognitive reappraisal and expressive suppression. The effect size was medium for cognitive reappraisal and small for expressive suppression (Table 2).

Given the significant differences found in the variables analysed by sex, correlational analyses were conducted separately for boys and girls (Table 3).

Peer, mother, and father attachment correlated negatively with expressive suppression and positively with cognitive reappraisal in both sexes (Table 3). The internal relationship among the attachment variables was positive and significant for both sexes. A particularly strong association was observed between mother attachment and father attachment (r = .701; p < .01 for girls, and r = .679; p < .01 for boys). Finally, perceived academic efficacy correlated positively with mother attachment, father attachment, and cognitive reappraisal in both boys and girls, whereas the associations between expressive suppression and perceived academic efficacy were weak and only present among boys (r = -.127, p < .05). Besides Fisher’s r-to-z tests were applied to examine whether correlations differed between boys and girls. None of the comparisons was significant (all p > .05), indicating that the correlational patterns are comparable across sex. Full results are provided in Appendix.

Table 4 shows the Multivariate Tests. The Box’s M test was non-significant (M = 46.993, F = 0.723, df₁ = 63, df₂ = 81926.424, p = .952), and Levene’s tests indicated homogeneous variances for all dependent variables. The multivariate effect of Sex was significant, Wilks’ Λ = .949, F(6, 665) = 5.952, p < .001, η²_p = .051; the effect of Age group was non-significant, Wilks’ Λ = .988, F(6, 665) = 1.320, p = .246, η²_p = .012; and the Sex × Age group interaction approached significance, Wilks’ = .982, F(6, 665) = 1.999, p = .064, η²_p = .018 (see Table 4).

Table 4 shows the univariate tests. Significant Sex effects emerged for secure attachment to peers, F(1, 670) = 11.678, p = .001, η²_p = .017; girls > boys, cognitive reappraisal, F(1, 670) = 11.224, p = .001, η²_p = .016; boys > girls, and expressive suppression, F(1, 670) = 4.260, p = .039, η²_p = .006; boys > girls, whereas secure attachment to mothers and fathers were nonsignificant. Age group effects were nonsignificant across outcomes. A small Sex × Age group interaction was found for perceived academic selfefficacy, F(1, 670) = 6.217, p = .013, η²_p = .009, with girls reporting higher selfefficacy at 15-16 than at 17-18, while boys remained stable or slightly higher at 17-18 (see Table 5).

A structural equation model was specified and estimated in which peer attachment, mother attachment, and father attachment, as antecedent variables, had an effect on cognitive reappraisal, expressive suppression, and academic self-efficacy, with participant sex as the grouping variable. Before estimating the structural equation model, multivariate normality was assessed using Mardia’s coefficient, which analyzes multivariate skewness and kurtosis. The results indicated deviations from normality (skewness = 12.84, kurtosis = 146.27, p < .001), confirming that the data did not strictly meet this assumption. This lack of normality is consistent with the observed distributions of the variables, where positive kurtosis values were detected in the attachment measures (e.g., attachment to peers = 2.45). The model was estimated using the robust MLR estimator in Mplus. It is a path analysis with three antecedent variables and three consequent variables (Figures 1 and 2). The model was saturated; therefore, the weakest relationship was removed to allow identification (the relationship between expressive suppression and academic self-efficacy). The model showed satisfactory fit, χ²(2) = 0.275, p > .05, CFI = .99, RMSEA = .000, 90% CI [.000, .055], SRMR = .004.

In the models, cognitive reappraisal was positively affected by father attachment but not significantly influenced by peer or mother attachment. With respect to expressive suppression, it was negatively affected by peer attachment. Regarding the effect of mother attachment, a different pattern emerged for boys and girls: this variable had a negative and significant effect for boys, whereas the effect was nonsignificant for girls. As for the effect of father attachment on expressive suppression, this variable did not show a significant effect in either group.

Finally, concerning the effects on academic self-efficacy, this variable was not significantly influenced by peer attachment. Academic self-efficacy was affected by both father and mother attachment for boys, and by father attachment only for girls.

As for the correlations, these were positive and significant between cognitive reappraisal and expressive suppression, and between cognitive reappraisal and academic self-efficacy in the case of boys, whereas these relationships were not significant for girls.

The model explained a total of 7.1% of the variance in cognitive reappraisal, 14.9% of the variance in expressive suppression, and 12.2% of the variance in self-efficacy for boys, and 6%, 9.2%, and 9%, respectively, for girls.