A number of DV prevention programs have been developed in the last decades (see for a review De La Rue et al., 2017; Whitaker et al., 2006). Some programs are aimed at at-risk populations, for instance, adolescents who have been victims of child abuse (e.g., the Expect Respect Program; Ball, Kerig, & Rosenbluth, 2009; the Youth Relationships Project; Wolfe et al., 2003). Universal preventive interventions aimed at the general population of adolescents have also been developed and evaluated. Some noteworthy programs in this area are the Fourth R Skills for Youth Relationships (Wolfe et al., 2009) and the Safe Dates Project (see, for example, Foshee et al., 2005). In addition, innovative methodologies, as for instance game-based interventions, have also been proposed for primary DV prevention. For example, Bowen et al. (2014) developed an educational computer game (Green Acres High) aimed at raising awareness of and changing attitudes towards DV in adolescents.

As mentioned, although most DV prevention programs developed to date have shown efficacy in modifying cognitions and attitudes related to partner violence, behavioral changes have been reported in very few cases (De La Rue et al., 2017; Shorey et al., 2012; Whitaker et al., 2006). De La Rue et al.’s (2017) meta-analysis examined 23 studies that had assessed the efficacy of DV prevention programs implemented in middle and high schools. Their results evidenced that school-based programs are effective in increasing DV knowledge and shifting attitudes legitimizing violence in dating relationships, but DV behaviors – both perpetration and victimization – are not affected to a significant extent. Another limitation of existing programs is related to the study design. RCTs and long-term follow-ups are not usual (Shorey et al., 2012). Additionally, the programs developed until now are mostly multi-session, and some of them imply the involvement of various actors (students, schools, and parents). Although multi-session, multi-component programs that include a skill-building component have been highlighted to achieve behavioral changes and long-term effects (De La Rue et al., 2017; Whitaker et al., 2006), some researchers have also emphasized the potential relevance of brief interventions (Schleider & Weisz, 2017a).

A meta-analysis conducted by Schleider and Weisz (2017b) explored the efficacy of single-session interventions for youth psychological problems. The findings from 50 RCTs demonstrated a significant beneficial effect for certain psychological problems, with the largest effect for anxiety and conduct problems although, overall, the effects waned over time (Schleider & Weisz, 2017b) and were slightly smaller than those of larger interventions (Weisz et al., 2017). Despite these limitations, other advantages are relevant for consideration and may justify the choice of single-session interventions under certain circumstances. Specifically, its brevity and flexible format (e.g., via computers, individuals without psychotherapy training, or self-administered) reduce costs, broaden accessibility to populations who might not otherwise access mental health care, and maximize scalability (Schleider & Weisz, 2017a).

Particularly in the area of DV prevention, in recent years, a few studies have brought into focus the effect of single-session interventions for secondary and tertiary prevention. For example, the studies by Cunningham et al. (2013) and Rothman and Wang (2016) examined the effect of brief interventions using the principles of motivational interviewing to reduce perpetration of DV in medical center settings among at-risk adolescents, showing its potential to influence target attitudinal and behavioral outcomes. Beyond the DV prevention area, other types of brief interventions directed at adolescents have emerged. For instance, some researchers have focused on the potential effect of brief interventions based on promoting an incremental theory of personality (ITP) in emotional well-being of adolescents and youth (Schleider & Schroder, 2018; Yeager, Lee, & Jamieson, 2016). ITP interventions specifically aim at changing entity theories of personality (i.e., the belief that personal characteristics are fixed and cannot be changed) for an incremental theory of personality (i.e., the belief that people do have the potential to change). ITP interventions use strategies derived from research on persuasion and attitude change, such as a focus on students’ perspectives and scientific information. Moreover, they are presented to participants with no intention of modifying their characteristics so that the adolescents do not feel manipulated, thereby minimizing resistance to change (for a review, see Yeager & Walton, 2011).

ITP-based interventions have shown efficacy to improve internalizing problems such as depression and anxiety in adolescents (Miu & Yeager, 2015; Schleider & Weisz, 2018). A positive effect has also been proved in other behavioral areas such as academic achievement (Romero, Master, Paunesku, Dweck, & Gross, 2014; Yeager et al., 2014). Particularly, in the area of peer aggression, the only previous study that has explored the effect of a single-session intervention based on ITP in peer conflict situations (Yeager, Trzesniewski, Tirri, Nokelainen, & Dweck, 2011) found that experimentally inducing an incremental theory (a belief in the potential for change) about oneself and about bullies in high school students reduced the desire to take revenge on perpetrators. These results suggest that interventions based on promoting an incremental theory of personality show promise to reduce peer aggression, although behavioral intentions, but not actual aggressive behaviors, were measured. Thus, conclusions about the efficacy of single-session interventions based on ITP to reduce adolescent aggressive behaviors are still preliminary, and no study to date has explored its effect, particularly on aggressive behaviors in the context of a dating relationship.

Why may ITP interventions help to reduce DVP? These types of interventions have been shown to reduce aggression toward peers, but also to promote overall prosocial behaviors such as being more respectful and friendly to others (Yeager et al., 2011; Yeager, Trzesniewski, & Dweck, 2013), and therefore, they could also encourage behaving more prosocially and less aggressively in the context of a dating relationship. Moreover, increasing empathy and the ability to put oneself in other’s place, promoting self-esteem and self-confidence to handle stressful situations, and decreasing hostility and angry feelings are core elements targeted by the ITP intervention and, at the same time, variables that have been identified as relevant DV risk factors (e.g., Pflieger & Vazsonyi, 2006; Vagi et al., 2013).

Considering the above, the main objective of this study was to explore the effect of a brief single-session ITP intervention on DVP and DVV. Based on the results of previous studies exploring the effect of this type of intervention on adolescent aggressiveness in peer conflict situations (Yeager et al., 2013; Yeager et al., 2011), we hypothesized that perpetration of DV behaviors will be significantly reduced in adolescents who received the ITP intervention in comparison to a control group of adolescents who received an educational intervention about the human brain. Regarding DVV, considering the predominant bidirectional pattern of aggression in dating relationships (Chiodo et al., 2012; O’Leary et al., 2008), we expected that a potential impact of the intervention in victimization may also occur. Another objective of this study was to explore whether the effect of the ITP intervention on DVP and DVV was moderated by participants’ sex. As the prevalence of DV is high in both boys and girls and previous studies have not evaluated sex differences in the effects of interventions based on ITP in aggressive behavior, we did not state any specific hypothesis about sex differences.

We conducted a double-blind randomized controlled trial (RCT) with two parallel groups (experimental vs. control). Recruitment was carried out via educational centers. We invited a random sample of 20 school headmasters from all high schools in Bizkaia (Basque Country, Spain) to participate in the study. Of them, the headmasters of ten schools agreed to participate and four schools were randomly selected for this study, which is part of a larger research project aimed at assessing the effect of the ITP intervention on several internalizing and externalizing problems. Eligible participants were those who were enrolled in any of the four high schools and who spoke Spanish or Basque fluently (N = 603). Informed consent was required both from parents and adolescents. Ten adolescents declined to participate in the study, and 147 parents did not provide consent. Moreover, 44 adolescents were excluded from the study because they were not in class on the days of the pretest and/or intervention, and 279 adolescents were excluded because they had not begun dating (see Figure 1). Randomization took place on the day of the intervention and was done at individual level within each classroom, blocked by sex. Allocation was concealed to participants, researchers, and teachers. Participants completed task interventions on paper individually, and assessment measures were administered one week prior to the intervention (pretest), one week after the intervention (post-test), six months after the intervention (6-month follow-up), and one year after the intervention (1-year follow-up). Because of the short time frame of the post-test (see Measures section), the resulting behavioral assessment of dating aggression at the post-test was not comparable to the measures obtained in the other waves, and therefore it was not used for this study’s analyses. Both interventions and assessment measures were administered by research assistants during normal class time (clinicaltrials.gov identifier NCT03583645). The Ethics Committee of the University of Deusto approved this study and confidentiality was guaranteed.

One hundred and twenty-three adolescents participated in this study, with 62 participants allocated to the experimental condition and 61 to the control condition (see Figure 1). Of these, 28 participants did not complete measures at the 6-month follow-up (attrition rate = 22.76%), and another 16 participants did not complete measures at the 1-year follow-up (attrition rate = 35.77%), mainly because they were not in class on the day of data collection. Thus, 79 participants (41 in the experimental group and 38 in the control group) completed the three assessment waves. Attrition analyses revealed that there were no significant differences between completers (n = 79) and non-completers (n = 44) in condition or sex, although non-completers were significantly older (M_age = 15.57) than completers (M_age = 15.00), t(121) = 3.23, p < .01. Mean scores of DV were higher for non-completers compared to completers, mainly in the case of victimization, although the differences were not statistically significant. In addition, comparisons between conditions showed that there were no differences in sex or age between the experimental and control groups. With regard to DV pretest scores, this preliminary analyses revealed that participants in the experimental condition scored significantly higher in DVP, total score: t(121) = 2.24, p = .027; traditional DVP: t(121) = 2.36, p = .020; online DVP: t(121) = 1.72, p = .088, than participants in the control condition, without differences in DVV.

Sixty-six of the participants were female (53.7%) and 57 were male (46.3%), with a mean age of 15.20 years at the pretest (SD = 0.99, range = 13.06-17.77). Their parents’ socioeconomic status was determined by using the recommendations of the Work Group of the Spanish Society of Epidemiology and the Spanish Society of Family and Community Medicine (2000), which consider parents’ last job. According to this criterion, the distribution was as follows: 11.1% low, 23.2% low-medium, 12.1% medium, 32.3% high-medium, and 21.2% high socioeconomic class.

Perpetration of DV was measured by asking adolescents to report if they had carried out any of the 25 listed behaviors on her/his partner or ex-partner in the last six months (last week for the post-test measure). Fourteen of the items refer to traditional (face-to-face) dating aggressive acts and 11 items assess acts of online dating abuse. The measured aggressive behaviors were based on previous validated scales for the assessment of DV. However, as there was no previous measure that assessed the entire range of target aggressive behaviors of this study (for instance, face-to-face and online aggressions), a new ad hoc scale was elaborated which covers different types of traditional aggressions (physical, psychological, and sexual) and online aggressions. Specifically, the Conflict in Adolescent Dating Relationships Inventory (CADRI; Wolfe et al., 2001; Spanish version by Fernández-Fuertes, Fuertes, & Pulido, 2006), the Psychological Abuse in Partner Relationships Inventory (Calvete, Corral, & Estévez, 2005), and the Spanish validation (Buesa & Calvete, 2011) of the Subtle and Overt Psychological Abuse of Women Scale - SOPAS (Marshall, 1992) were considered for the construction of the items examining traditional DV. Items assessing online dating abuse were taken from the Cyber Dating Abuse Questionnaire (Borrajo, Gámez-Guadix, Pereda, & Calvete, 2015). All the items are shown in the Appendix. Next, to assess victimization, participants were asked to report if they had been victims of the same offline and online aggressive behaviors by their partner or ex-partner in the last six months. The response choices for each item were defined with a 4-point Likert-type scale ranging from 0 (never) to 3 (often).

The DV scale showed adequate psychometric properties. Specifically, the factor structure was examined through confirmatory factor analysis (CFA) with LISREL 8.8 (Jöreskog & Sörbom, 2006). We used the robust maximum likelihood (RML) method, which requires an estimate of the asymptotic covariance matrix of the sample variances and covariances and includes the Satorra-Bentler scaled χ² index (S-Bχ²). A correlated two-factor (traditional and online aggression) model was tested both for DVP and DVV. Each item loaded exclusively on one factor, and the measurement error terms associated with each item were uncorrelated. The fit indices obtained for the two-factor solution were acceptable both for DVP and DVV (see Table 1). The covariance between the two factors was .91 (p < .001) for DVP and .87 (p < .001) for DVV. We also tested a unidimensional model in which all the items were explained by only one factor. This model also showed adequate fit indices both for DVP and DVV (see Table 1), although worse than those obtained for the two-factor models. The unidimensional model increased chi-squared significantly both for DVP, Δχ²(1, N = 123) = 21.43, p < .001; and DVV, Δχ²(1, N = 123) = 57.87, p < .001. Therefore, the two-factor model, which includes the differentiation between the two types of aggression (i.e., traditional and cyber dating abuse) was considered preferable. Finally, reliability was examined. Cronbach’s alphas for this study’s sample at pretest, 6-month follow-up, and 1-year follow-up, respectively, were .91, .97, and .90 for DVP total score; .83, .93, and .88 for traditional DVP; .90, .94, and .78 for online DVP; .93, .94, and .93 for DVV total score; .88, .92, and .89 for traditional DVV; and .88, .93, and .85 for online DVV.

ITP intervention. The experimental intervention was developed in the United States by David Yeager and colleagues (see, for example, Yeager et al., 2014). The intervention had three main parts that are presented to adolescents as a writing assignment to be completed in about 50-60 minutes. These parts come from now-standard methods for “wise interventions” in social psychology (Walton, 2014; Walton & Wilson, 2018; Yeager & Walton, 2011). First, participants are asked to read scientific information that provides evidence that individuals have the potential to change. They read about neurological and behavioral studies showing that behaviors are controlled by “thoughts and feelings in brains,” and that pathways in the brain have the potential to be changed under the right circumstances. After reading this information, participants are asked to write three sentences to explain in their own words why scientific evidence shows that it is true that people have the potential to change. Second, participants read several normative quotes purportedly written by upperclassmen that previously read the same scientific information and endorsed its conclusions (i.e., “descriptive norms”; Cialdini, 2003). These testimonials are provided to give credibility to the incremental theory of personality. They were obtained from previous interventions in the United States and edited by the research team. The third and last part consists of a self-persuasive writing exercise (Aronson, 1999). In this final task, participants are asked to write their own version of such a narrative to share with future students (see Aronson, Fried, & Good, 2002). Specifically, adolescents have to describe a time when they felt withdrawn, rejected, or disappointed by another person at school. Then, they are asked to imagine that the same event he or she has described has happened to another student and write one to three paragraphs describing what he or she can say to help the other student to understand that people can change and that the things that are happening to him or her can also change. This activity has been shown to facilitate the internalization of the ITP intervention message, building on a long line of research on cognitive dissonance (Walton & Cohen, 2011).

Educational intervention. The control intervention involved scientific information and education about the different areas and functionalities of the human brain. This educational intervention was designed to run parallel to the experimental one; hence, it also has three main parts that are completed in around 50-60 minutes. First, participants are asked to read scientific information about the different areas of the brain and their specialties. After, participants are asked to write three sentences to explain in their own words why this information is interesting. Second, participants read several testimonials written by upperclassmen about their transition to high school and how their brains help them to adapt to the new space. All the testimonials refer to physical aspects of the school building. As for the experimental intervention, these testimonials were obtained from previous interventions in the United States and edited by the research team. The third part consists of a writing exercise. In this final task of the control intervention, participants are asked to write a letter (3-5 sentences) to another student explaining the main things he or she has learned about the human brain and what he or she thinks are important for adapting to the new physical environment in high school. Both the experimental and control interventions were adapted to our cultural specificities and translated into Spanish and Basque languages.

Hierarchical linear models with HLM 6.0 (Raudenbush, Bryk, & Congdon, 2004) were used to test whether the ITP intervention reduced DV over time. We estimated separate models for each outcome (DV perpetration total score, traditional DV perpetration, online DV perpetration, DV victimization total score, traditional DV victimization, and online DV victimization) with measurement occasions (level 1) nested within individuals (level 2). We estimated the intraclass correlation coefficient (ICC) for all the outcome variables of the study over time to determine whether school centers and classes should be included as a third level in the analyses. The ICC values ranged between .00 and .08 (< .10 in all cases), suggesting that it was not necessary to include a third level in the analyses. Full information maximum likelihood (FIML) was used as the estimation method for all the models tested. FIML estimates parameters using all the available data, including cases without data (Little, Jorgensen, Lang, & Moore, 2014). For level 1, regression equations modeled variation in the repeated measures as a function of time, which was coded as 0 (pretest), 1 (6-month follow-up), and 2 (1-year follow-up). For level 2, the equations modeled individual differences in the level 1 parameters (i.e., intercepts and slopes) as a function of between-subject variables. Level 2 predictors of the intercept and the slope included condition (0 = control, 1 = experimental), sex (0 = female, 1 = male), and the interaction term between condition and sex. The inclusion of these parameters at level 2 allowed the effects of condition and sex on both the intercept and the change in the outcome variables over time to be tested. However, to increase the sample’s power to detect significant changes and following the parsimony principle, sex was excluded from the models when its interaction with condition was nonsignificant. Random effects at level 2 for intercept and slope were included, thereby allowing variability between individuals in the initial levels and changes over time. Where the random effects were nonsignificant, they were removed from the final models. Finally, effect sizes were calculated so that differences between conditions in mean score changes in the outcome variables from pretest to follow-ups could be compared, using the estimated marginal means obtained in the mixed models. Where the differences between conditions in mean score changes were statistically significant, positive values of Cohen’s d indicated greater decreases in DV for adolescents in the ITP condition compared to adolescents in the control condition. Negative values of Cohen’s d indicated greater decreases in the control condition. Effect sizes were defined as small for Cohen’s d values of 0.20, medium for values of 0.50, and large for values of 0.80 (Cohen, 1992).

Within the potential limitations of this study, we must consider the relatively small sample size, which prevented the examination of the potential moderator role of other variables (for example, grade) in the effect of the ITP intervention and warrants caution in the interpretation of effect sizes of the intervention. Although the initial number of potential participants was higher, the final sample was reduced because not all the participants had begun to date. The need of early prevention of DV has been strongly emphasized by researchers in the area (e.g., Foshee & Reyes, 2009; O’Leary & Slep, 2012). However, conducting studies to examine the effect of interventions on early adolescents have to deal with the difficulty that a number of adolescents have not begun dating relationships, which hinders obtaining large samples. On the other hand, adolescents who are already involved in dating relationships may differ from those who have not yet begun to date, so that the effect of the intervention on the latter may be not the same.

Despite that, the present study has a relevant number of strengths. First, this is the first study testing the effect of a brief intervention based on ITP on adolescent dating aggression. Second, several strengths related to the study design have to be highlighted. A double-blind RCT was conducted, which has reduced internal validity threats and thus reinforced the conclusions of the study regarding the effectiveness of the intervention. In addition, the fact that participants came from different high schools increased the external validity of the study. Finally, given that the effects of brief interventions may wane over time (Schleider & Weisz, 2017b), long-term follow-ups (until one year after the implementation of the intervention) were collected.

The present study was the first to explore the effectiveness of a brief intervention based on changing entity theories of personality for preventing adolescent dating aggression. Our findings suggest that the ITP intervention decreased the perpetration of aggressive acts toward a dating partner. Although further research with larger samples is needed in order to establish solid conclusions about the effect of the intervention on DV, these study findings show promise about the potential effectiveness of brief interventions for DV prevention. Schools are the natural setting to recruit children and adolescents. However, they do not always have enough time available or personal resources for the implementation of multi-session programs. Therefore, increasing our empirical evidence about the effectiveness of a one-hour, self-applied intervention is of great relevance for moving forward in the prevention of DV.

Another relevant implication of the study findings is related to the fact that the intervention was not specifically designed for the prevention of DV, but rather for adolescents’ other psychological problems (such as depression and peer conflict relationships). On one hand, some authors have claimed that DV prevention may be done indirectly through the prevention of its behavioral precursors (Foshee & Reyes, 2009). In this sense, the previous authors highlight the role of bullying, considering the results of some longitudinal studies showing evidence of the predictive role of peer aggression on DV and partner abuse. The results of our study give some support to the idea that strategies aimed at preventing peer conflict may also prevent dating aggression. On the other hand, Schleider and Weisz (2017a) found in their review that single-session interventions seem to be more effective when addressing specific intervention targets and well-defined behaviors. Thus, adapting the ITP intervention specifically to dating conflict situations may increase its effectiveness. In this sense, a relevant challenge for future research would be to compare the current ITP intervention with a new version adapted to dating conflict situations in order to explore whether the last one has a bigger effect on the prevention of DV. This will allow researchers to make decisions based on empirical evidence. Even though more general interventions may be of great relevance for universal prevention, more specific interventions could be useful for selective prevention.