The presence of beliefs that justify violence in romantic relationships, such as gender stereotypes, negative attitudes toward women, and favorable attitudes toward violence, has been identified as risk factors for TDV (Rubio-Garay et al., 2015; Reidy et al., 2016; Hunt et al., 2022). Adolescents with maladaptive schemas may play a relevant role in victimization in dating relationships (Calvete et al., 2018). This is particularly observed among adolescents who hold accepting attitudes toward the use of violence in dating relationships (Jennings et al., 2017; Smith-Darden et al., 2017).

Sexist beliefs are prevalent among Spanish adolescents (Ferragut et al., 2017; Ramiro-Sánchez et al., 2018), encompassing both hostile and benevolent sexism. Both forms of sexism are associated with the justification and perpetration of various types of violence, including TDV (Carrascosa et al., 2018; Vives-Cases et al., 2021).

Regarding gender differences, Ayala et al. (2021) found that males scored higher than females on sexism, and females who had experienced TDV reported higher levels of hostile sexism than those who had never been in an intimate relationship. Conversely, Pazos-Gómez et al. (2014) found that high levels of sexism were associated with increased threats of sexual, relational, verbal-emotional, and physical violence in males, whereas in females, it was only associated with higher levels of relational violence. Rey-Anacona et al. (2017) also found significantly higher scores on hostile sexism among males than females. Furthermore, hostile sexism was linked to severe physical aggression perpetrated by females, while benevolent sexism was associated with mild aggression perpetrated by females.

In terms of the relationship between victimization and sexist beliefs, Marcos et al.’s (2020) study supports the association between victimization in relationships and sexist beliefs and romantic love myths, suggesting that these attitudes and myths can facilitate the initiation and perpetuation of TDV. This aspect has also been demonstrated in other studies (Fernández-Fuertes et al., 2018; Rey-Anacona and Martínez-Gómez, 2022).

Among the psychological factors that can act as protective factors against TDV, high empathy is highlighted (Vagi et al., 2013; Davis et al., 2019). According to Wolfe et al. (2004), empathy can moderate the relationship between attitudes of violence justification, self-efficacy, and TDV. This finding is consistent with multiple studies that indicate empathic capacity as an inhibitory factor of aggression (Tur-Porcar et al., 2016; Deschamps et al., 2018; Song et al., 2018), although authors like Vachon et al. (2013) suggest that this relationship is moderated. On the other hand, Espelage et al. (2020) observed through a longitudinal study that male perpetrators of TDV had significantly higher levels of empathy compared to non-perpetrators, but over time, the perpetrators’ empathy decreased while non-perpetrators’ empathy increased.

The study of empathy in this field seems more appropriate from Davis's (1983) multidimensional model. This model differentiates between cognitive empathy (Fantasy and Perspective Taking), which is the ability to understand how others feel, and affective empathy (Empathic Concern and Personal Distress), which corresponds to the vicarious experience of feelings expressed by others. Several studies report that aggressive individuals exhibit lower cognitive empathy than non-aggressive individuals (Gantiva et al., 2018). Additionally, Berger et al. (2015) indicate that affective empathy inhibits aggression but not cognitive empathy.

Regarding specific dimensions of empathy, Van Heerebeek (2015) found that perspective-taking was negatively related to indirect aggression in both males and females. However, personal distress was positively related to indirect aggression in males and direct aggression in females. Authors such as Guzmán-González et al. (2014) observed that university students who engaged in psychological TDV had low levels of perspective-taking and empathic concern. Similarly, Valdivia-Salas et al. (2021) showed a positive relationship between high personal distress and TDV in physical and relational forms. Regarding the relationship between victimization and empathy, the results of Dodaj et al.’s (2020) study indicate that empathic concern and perspective-taking dimensions are significant predictors of victimization in sexual coercion and psychological aggression. In summary, not all dimensions of empathy have the same protective capacity against TDV perpetration.

Assertiveness is recognized as a valuable interpersonal skill for expressing needs, fostering effective romantic relationships, obtaining support, and facilitating successful conflict resolution (Xia et al., 2018). The cognitive model of assertiveness proposed by Vagos and Pereira (2010) encompasses four interpersonal schemas or fundamental beliefs, reflecting individuals’ beliefs about their ability to express themselves and respond appropriately in different social contexts. These schemas include external emotional support, interpersonal management, practical personal ability, and affective personal ability. The first schema refers to a positive representation of others as suppliers of support, acceptance, and affection (e.g., “When I am sad, angry, or upset, I have someone to support me and help me feel better”). The second pertains to a representation of the self as possessing the abilities needed to manage daily life (e.g., “I am capable of performing tasks at work (or school) as well as most people”). The third relates to the ability to solve problems as part of interpersonal encounters (e.g., “When someone I like pulls away from me, I try to understand why and solve the situation”). The fourth conveys beliefs about the self being lovable (e.g., “I feel I am special to some people”).

Research on the role of assertiveness in TDV suggests that this skill can serve as a protective factor in escalating TDV conflicts. For example, Simpson-Rowe et al. (2012) found that assertive skills were associated with a reduced risk of TDV, particularly for young females, including sexual victimization and coercion. Similarly, Fortin et al. (2021) reported a negative association between assertiveness and TDV perpetration. However, Xia et al. (2018) did not find significant relationships between assertiveness and physical or psychological TDV. On the other hand, Valdivia-Salas et al. (2023), drawing on Vagos and Pereira’s (2010) cognitive model of assertiveness, indicated that practical personal ability was explicitly associated with high TDV perpetration or victimization among adolescents.

Gender differences in interpersonal skills during adolescence have been documented, with females generally exhibiting higher levels of general interpersonal skills than males (Salavera et al., 2019; Persich et al., 2020). Notably, Ayala et al. (2021) found that high assertiveness was associated with higher levels of benevolent sexism in both sexes and higher levels of hostile sexism in males.

Despite the extensive literature linking these variables to TDV, no theoretical model has been found that comprehensively considers all of them. Although the well-established link between sexist beliefs and TDV, limited attention has been given to understanding the role of empathy and assertiveness as mediators in this relationship. A model in this sense can contribute to developing preventive and treatment programs that cultivate interpersonal skills capable of moderating the impact of sexist beliefs on TDV perpetration and victimization. It is important to note that the relationships among these variables may be sex-dependent and can vary depending on the specific constructs examined.

Considering the previous literature that establishes the relationship of sexism, empathy, and assertiveness with TDV, are there gender differences regarding all these variables, and consequently, is it justified to develop a predictive model of TDV differentiated by gender? Assuming gender differences are found according to previous literature, does sexism predict both perpetration and victimization in TDV? Are empathy and assertiveness mediating variables between sexism and TDV in both directions, perpetration, and victimization? The main objective of the present study is to propose predictive models of sexism differentiated for perpetration and victimization of TDV, with empathy and assertiveness as mediating variables for both genders. To achieve this, several intermediate objectives are established aimed at confirming the conditions that support the formulation of such a model. Firstly, it examines whether significant gender differences exist in TDV perpetration and victimization, including different typologies (relational, verbal-emotional, and physical) and roles (non-perpetrator and non-victim, perpetrator, victim, and perpetrator-victim). Secondly, explore potential gender-based differences in sexism, empathy, and assertiveness and their potential associations with TDV perpetration and victimization. Lastly, develop separate models for each gender to examine the predictive role of sexism on TDV, with empathy and assertiveness as mediating variables.

The study sample consisted of 1,650 students attending public secondary schools in various regions of Spain, namely Asturias, Murcia, Teruel, and Valencia. Out of the total sample, 862 students (49.80% males and 50.20% females) who reported having been in an intimate relationship within the past 12 months were included in the study. The mean age of the students was 14.13 years (range = 11–17, SD = 1.35), and they were enrolled in the first (15%), second (28.10%), third (28.10%), or fourth year (28.90%) of compulsory secondary education, which corresponds to grades 7, 8, 9, and 10 in the North American educational system.

The Conflict in Adolescent Dating Relationships Inventory (CADRI, Wolfe et al., 2001), adapted to Spanish by Fernández-Fuertes et al. (2006), was used in this study. It comprises 25 items designed to assess conflictive actions within the past 12 months among adolescent dating partners who either engage in abusive behavior or are victimized. Respondents rated the frequency of experiencing these actions on a 4-point scale, ranging from 1 (never) to 4 (often). The inventory follows a five-factor structure representing five types of abuse: physical, sexual, threatening, relational, and emotional or verbal. Previous Spanish validation demonstrated that three factors showed optimal reliability: emotional violence (10 items, Cronbach’s alpha = 0.79), physical violence (4 items, Cronbach’s alpha = 0.76), and relational violence (3 items, Cronbach’s alpha = 0.73). In this study, all scales exhibited high reliability for both perpetration of TDV (Cronbach’s alpha = 0.83) and victimization of TDV (Cronbach’s alpha = 0.90).

The Adolescent Sexism Detection Scale (Recio et al., 2007) was utilized in the study. It comprises 26 items assessing Hostile Sexism (16 items) and Benevolent Sexism (10 items). Participants rated each item on a 6-point scale, ranging from 1 (strongly disagree) to 6 (strongly agree). Previous research conducted with Spanish adolescents reported good internal consistency with Cronbach’s alpha values ranging between 0.80 and 0.86 for Benevolent Sexism and between 0.92 and 0.94 for Hostile Sexism (Recio et al., 2007; Ramiro-Sánchez et al., 2018). In the present sample, Cronbach’s alpha coefficients were 0.93 for Hostile Sexism, 0.89 for Benevolent Sexism, and 0.89 for Trait Sexism.

The Interpersonal Reactivity Index (Davis, 1983), adapted to Spanish by Mestre et al. (2004), was administered in this study. It consists of 28 items that assess cognitive and emotional dimensions of empathy, including Perspective Taking (7 items), Fantasy (7 items), Empathic Concern (7 items), and Personal Distress (7 items). Respondents rated the extent to which each statement described themselves on a 5-point scale, ranging from 1 (does not describe me well) to 5 (describes me very well). Previous studies with Spanish adolescent samples reported Cronbach’s alpha values between 0.56 and 0.76 for the four subscales (Mestre et al., 2004). In our sample, the two reversed items of the Personal Distress subscale were removed due to reliability issues (Józsa and Morgan, 2017). The Cronbach’s alpha coefficients were 0.65 for Perspective Taking, 0.70 for Fantasy, 0.72 for Empathic Concern, and 0.72 for Personal Distress.

The Assertive Interpersonal Schema Questionnaire (Vagos and Pereira, 2010) was employed in this study to assess core beliefs about oneself, others, and social events or interactions. The questionnaire comprises 21 items, which measure four subscales: Outer Emotional Support (5 items), Practical Personal Ability (4 items), Interpersonal Management (8 items), and Affective Personal Ability (4 items). Participants were asked to rate the extent to which the statements described themselves on a 5-point scale, ranging from 1 (completely false to me) to 5 (completely true to me). Previous studies with Portuguese samples reported Cronbach’s alpha values between 0.75 and 0.83 for the subscales (Vagos and Pereira, 2010). In our sample, Cronbach’s alpha coefficients were 0.86 for Outer Emotional Support, 0.86 for Practical Personal Ability, 0.74 for Interpersonal Management, and 0.72 for Affective Personal Ability.

To ensure a geographically diverse sample, authorization was obtained from the Directorates-General for Education of the Autonomous Communities of Comunidad Valenciana, Region of Murcia, Aragón, and Asturias. A total of 10 schools voluntarily participated in the study through convenience sampling. These schools represented both public ownership (82.73%; n = 1,365) and private-concerted ownership (17.27%; n = 285). Out of the total 1,650 participants, 45.63% (n = 753) attended schools in the province of Teruel (Aragón), spread across four schools; 32.9% (n = 543) attended schools in the Region of Murcia, also spread across four schools; 7.88% (n = 130) attended a school in Asturias; and 13.57% (n = 224) attended a school in the province of Valencia (Comunidad Valenciana).

A detailed letter summarizing the research project was sent to the selected schools before data collection. The researchers personally contacted the principals of these schools to provide a comprehensive explanation of the study’s purpose and to obtain permission to conduct the research. Informed consent was obtained from parents or guardians for their children’s participation in the study. The students themselves were fully informed about the study objectives and assured that participation was voluntary and anonymous. Data collection occurred during regular classroom sessions, specifically during a homeroom class, and completing the questionnaire took approximately 20 min. A researcher was present during the administration of the instruments to offer necessary support to the students. Participants were informed of their voluntary participation rights following the Spanish Organic Law 3/2018 on Personal Data Protection and Digital Rights Guarantee. The selection criteria for participation were as follows: (a) age between 11 and 17 years old; (b) explicit agreement to participate; and (c) proper completion of the survey.

The study adhered to The Code of Ethics of the World Medical Association (Declaration of Helsinki) and was approved by the Committee of Evaluation and Follow-up of Research with Human Beings (CEISH) from Valencian International University (protocol code CEID2020_03).

First, specific analyses were conducted to describe the results obtained by the male and female participants in the dependent variables of interest. Central tendency (mean) and dispersion (standard deviation) were calculated for continuous variables, while frequencies were determined for categorical variables. Group comparisons were performed using Student’s t-test, as the sample size allowed for parametric statistics. Effect sizes (Cohen’s d) were calculated to supplement the comparative analysis. Chi-square tests (χ²) were employed to compare categorical variables, such as gender and types of teen dating violence (TDV), with Cramer’s V selected as the effect size measure. The classification of participants into each TDV category was presented, along with the expected values for each case.

Bivariate correlations were examined using Pearson’s correlation coefficient (r), which served as the basis for mediation models. Model 4 of the SPSS Macro PROCESS (Hayes) was used for the mediation analysis, with sexism as the predictor variable and violence (perpetration or victimization) as the outcome variable. The factors from the empathy and assertiveness questionnaires were considered mediator variables after confirming their independent predictive capacity for other dimensions in the model. The statistical significance of these analyses was assessed using the bootstrapping method.

All statistical procedures were conducted using the Statistical Package for the Social Sciences (SPSS), version 25. A significance level of 0.05 was set for testing the null hypothesis.

A comparative analysis was conducted between males and females on the variables of TDV (Table 1), revealing statistically significant differences primarily in perpetration. Specifically, females obtained higher scores in verbal-emotional TDV perpetration (t₈₆₂ = −4.60; p < 0.001) and the overall factor of TDV perpetration (t₈₆₂ = −3.53; p < 0.001). Males scored higher in relational TDV perpetration (t₈₆₂ = 3.04; p = 0.002). In all these cases, the effect size was small. Regarding TDV victimization, females obtained higher scores in verbal-emotional TDV (t₈₆₂ = −2.86; p = 0.004).

When comparing males and females in terms of TDV types (Table 2), significant differences were observed between the two groups (χ² = 13.024, p = 0.005). Specifically, males scored higher than expected in the category of “neither perpetrator nor victim,” while females scored higher in “perpetrator and victim.” Despite the observed statistical significance, the effect size was negligible.

Furthermore, when examining the relationship between gender and different modalities of sexism (Table 3), it was found that males had higher scores in hostile sexism (t₈₆₂ = 6.16; p < 0.001) and benevolent sexism (t₈₆₂ = 2.60; p = 0.009). The effect sizes were small for hostile sexism (d = 0.420) and negligible for benevolent sexism.

Regarding empathy (Table 4), it was observed that in all cases, females obtained higher scores than males: perspective taking (t₈₆₂ = −2.83; p = 0.005), fantasy (t₈₆₂ = −5.68; p < 0.001), empathic concern (t₈₆₂ = −3.64; p < 0.001), and personal distress (t₈₆₂ = −3.55; p < 0.001). The effect sizes were small for fantasy (d = 0.385), empathic concern (d = 0.246), and personal distress (d = 0.242). The effect size for perspective-taking was negligible (d = 0.194).

Finally, assertiveness dimensions were compared between both genders (Table 5), and it was found that females obtained higher scores in external emotional support (t₈₆₂ = −4.76; p < 0.001), interpersonal management (t₈₆₂ = −2.06; p = 0.040), and affective personal ability (t₈₆₂ = −2.15; p = 0.032). Only external emotional support showed an appropriate effect size but of a small magnitude (d = 0.324).

Separate analyses were conducted for males and females. In the male group (Table 6), it was observed that both perpetration and victimization of TDV positively correlated with all forms of sexism, assertiveness, and certain empathy factors (fantasy and personal distress). Significant negative correlations were found between all dimensions of sexism and assertiveness. Regarding empathy, negative associations were specifically observed between hostile sexism and the empathic concern subscale. Additionally, empathy showed positive correlations with all dimensions of assertiveness, except for personal distress, which did not exhibit any association with personal abilities (practical and affective).

In the female group (Table 7), positive associations were observed between TDV (perpetration and victimization) and all types of sexism, along with negative correlations between TDV dimensions and assertiveness. Sexism was directly related to personal distress and inversely related to assertiveness, except for benevolent sexism, which showed no association with interpersonal management. Empathy, perspective-taking, and empathic concern were positively associated with different forms of assertiveness (external emotional support, practical personal ability, and interpersonal management).

The significant correlations described allowed the development of mediational models, with sexism as the predictor variable and TDV perpetration and victimization as outcome variables. Assertiveness and empathy served as mediating variables. In the male sample, no relevant mediation effects were observed for assertiveness (Figures 1, 2). Nevertheless, a partial mediation effect was established through personal distress (Figures 3, 4), indicating a statistically significant mediating role between sexism and TDV, encompassing both perpetration and victimization.

In the sample of women, a partial mediation effect for assertiveness (practical personal ability) was observed on the established relationship between sexism and TDV perpetration (Figure 5), although this was not replicated in the same way when TDV (victimization) was used as the outcome variable (Figure 6). Regarding the mediating role of personal distress, it could only be applied in the case of the female sample on TDV victimization, as the variables involved did not present a relevant association with TDV perpetration. In any case, empathy did not generate a relevant mediation effect in this case (Figure 7).