Most of the violence against women is intimate partner violence. Worldwide, almost one third (27%) of women aged 15–49 who have been in a relationship report that they have been subjected to some form of physical and/or sexual violence by their intimate partner, and this figure is practically reached by the age of 15–19 (WHO, 2013). A study of 42,000 women in Europe that included the most frequent manifestations of this violence (such as psychological violence and control) estimated that around 43% of women aged 15–74 had suffered such violence in their lives (FRA, 2014).

Studies of adolescents show that male adolescent dating violence against women (DVAW) can appear in the earliest relationships they establish (Díaz-Aguado et al., 2021), can be a predictor of intimate partner violence in adulthood (Manchikanti Gómez, 2011), and relates to other forms of violence as bullying, with serious consequences for the subsequent development of both victims and perpetrators (Taquette and Monteiro, 2019).

Studies on male adolescent DVAW from a gendered perspective (Anderson, 2005; Santoro et al., 2018; Ferrer-Pérez and Bosch-Fiol, 2019) are scarce, and the few that exist are usually based on what the victims report to have suffered (Bradbury-Jones et al., 2019). Investigating DVAW based on self-reporting by adolescent boys is important, to know how different types of violence against women coalesce within adolescence, and their conditions of risk and protection, as well as helping to improve the efficacy of measures to prevent DVAW.

As the majority of researchers on violence acknowledge, in order to eradicate violence and alleviate the damage it causes, it is important to consider the multiple contexts within which the risk and protection factors exist (Krug et al., 2002; Fulu and Miedema, 2015; Boonzaier and van Niekerk, 2021; Zinzow et al., 2021). The analysis of such violence carried out from this ecological perspective (Bronfenbrenner, 1979; Heise, 2011) considers characteristics from: (1) individuals; (2) interpersonal relationships or micro-system; (3) the community or meso-system; and (4) society or macro-system. As the ecological approach of human development recognizes, the influence of playing a social role (such as the role of traditional masculinity in a relationship with a women) is seen at different levels, including at the individual and micro-system levels with the interpersonal relationships that occur therein, but also at the community and the macro-system levels in which the role has its roots (Bronfenbrenner, 1979). This underlines the importance of representative research of the population to study violence against women in terms of the cultural context in which it occurs (Webster et al., 2019).

One of the main risk conditions for male DVAW from adolescence, in which the social macro-system and individual levels interact, is masculinity based on control and the use of violence to exercise it. In support of this argument is the fact that sexist attitudes and acceptance of violence are among the main predictors of the violence that some adolescent boys inflict on women (Shen et al., 2012; Díaz-Aguado and Martínez-Arias, 2015; Reyes et al., 2016; Bradbury-Jones et al., 2019; Taquette and Monteiro, 2019; Wesche and Dickson-Gomez, 2019), and that this violence diminishes with interventions designed to override the mentality the underlies it (Heise, 2011). The results from studies of adolescents in Spain also show that the mentality based on male dominance against women is the main risk condition for behaviors of maltreatment that boys report to have shown in their relationships with girls, evaluated from: identification with male dominance and violence against women, and the use of violence to resolve conflicts (Díaz-Aguado et al., 2011, 2014, 2020, 2021).

Studies on the emotional characteristics of male abusers show that men who exercise violence against women have lower self-esteem than men who do not (Gallagher and Parrott, 2011), which is supported by some studies on adolescents (Díaz-Aguado et al., 2011, Díaz-Aguado et al. 2013, 2021) but not by others (Foshee et al., 2001; Merino et al., 2021). This contradiction could be explained by the fact that the relation between self-esteem and aggression can vary according to age and type of study of adolescents, as well as the instrument applied to evaluate aggression (Teng et al., 2015).

One of the emotional problems that has received most attention in studies of male adolescent DVAW from the gendered perspective is the gender role stress caused by situations in which masculinity based on dominance is impeded from being exercised (Gallagher and Parrott, 2011; Baugher and Gazmararian, 2015; Reidy et al., 2015; Merino et al., 2021). It has even been confirmed that such masculine gender role stress (MGRS) predicts violence against women to a greater extent than other more cognitive components of sexism, both in adults (Jakupcak et al., 2002) and in adolescents aged 13–16 (Merino et al., 2021).

The increasing use of digital technologies by adolescents has transformed all the environments in which we live and act, representing both an opportunity and a risk (Stonard, 2018), especially for those who tend to spend more time on their digital devices and act more impulsively when using them (Kim et al., 2020). A study across 11 countries in Europe by Durkee et al. (2016) found that 90% of adolescents with problematic internet use (PIU) presented multiple risk behaviors. As an example of the relation between PIU and male adolescent DVAW, it was found that boys who acknowledged perpetrating such violence against girls, and girls who reported having suffered this violence, both presented higher levels of PIU (Díaz-Aguado et al., 2021).

In addition, among the main risk conditions for male adolescent DVAW online include offline DVAW (Hinduja and Patchin, 2021) and stereotypical gender and dating beliefs (Reed et al., 2021).

One of the most widespread problems among adolescents who have grown up with digital devices is the reception of unwanted sexual solicitation, or sexual harassment online. One of the most significant studies on the prevalence of these behaviors in the European Union found that 11% of women older than 15 had been exposed to such problem; this was reported by double the figure of those aged 15–29 than in the 40–49 age group, and three times more than those aged 50–59 (FRA, 2014). The results of the study carried out in Spain on the generational differences in child sexual abuse remembered by adults aged 18–74 (Ferragut et al., 2021) reflect that while most forms of such abuse have decreased in the younger generation (18–24 years old), exposure to pornographic material and solicitations of personal photographs or videos of sexual content have increased, probably because they are exerted through the various digital devices that this generation uses much more frequently than the previous ones.

Based on these results, it is no surprise that studies on sexual harassment online in adolescents have multiplied in recent times, with findings that suggest it is one of most commonly perpetuated abuses, and that girls are the victims in the vast majority of cases (Baumgartner et al., 2010; Calvete et al., 2021; Díaz-Aguado et al., 2021; Mishna et al., 2021). This over-representation of girls among the victims of sexual harassment online leads us to consider it as a new form of violence against women (SHOAW), according to the definition of violence against women adopted by the Council of Europe (2011). In support of research on such violence from a gendered perspective, it should also be noted that the situations in which sexual harassment online occurs are perceived very differently by boys and girls; it has a far more negative emotional impact on girls, whereas some boys misconceive the severity of the impact of sexual harassment online on women (Brown et al., 2020).

Sexual harassment against women (SHAW) seems to be closely related to other forms of violence (as bullying) and risk behaviors that include browsing high risk websites and interacting with someone they have met online (Ybarra et al., 2007; Chang et al., 2016; Calvete et al., 2021).

Defining the typologies of adolescent boys in relation to the perpetration of different types of violence against women could be very useful for prevention. We found only two studies with this focus, both based exclusively on male adolescent DVAW. In the first, Lindhorst and Beadnell (2011) identified three patterns of exposure to violence in adolescent mothers: battered, moderate exposure, and low exposure. They concluded that typology is useful for illustrating an underlying “level of risk” in male adolescent DVAW, showing that psychological abuse might exist where physical abuse is absent, but that serious physical abuse exists in conjunction with psychological maltreatment. These findings were supported by research by Díaz-Aguado and Martínez-Arias (2015) whose probabilistic study of 4,147 boys in Spain aged 14–18 identified four discrete, identifiable groups. The first (76%) consisted of non-violent adolescent boys. The second (17%) comprised boys who isolated and controlled their partners. The third group (5%) was formed of boys who exerted only medium-level emotional abuse, and the fourth (2%) was represented by teenage boys who frequently engaged in all types of violence. In a logistic regression that compared the non-violent adolescents to the rest, the other groups presented lower self-esteem and greater justification of male dominance and intimate partner violence against women, greater justification of aggression in conflict resolution, as well as receiving more dominance and violence messages from adults in their family environment; they perceived violence against women as an abuse of lesser importance.

As these studies show, there has been little investigation of male adolescent violence against women as acknowledged by boys themselves, and less on such violence within different contexts, which include sexual harassment online outside the relationship with a partner, and comparative studies of behavior between those who perpetrate this violence and the population at large, as is the aim of this current research.

To understand the context of this research, the rapid changes toward equality between men and women that have occurred in Spain in the last decades should be taken into consideration. These changes are especially noticeable in the awareness and rejection of violence that men inflict on women within an intimate relationship; to counter this phenomenon, a law was approved in 2004 (Ministry of Equality, 2004) that became an international reference in this field, and a cross-party parliamentary pact was established in 2017 to fight gendered violence. International recognition of the passing of this law came from bodies such as UN Women, the World Future Council and the Inter-Parliamentary Union, as one of the most effective regulatory frameworks worldwide for eradicating this type of violence.

The study on adolescent violence against women in Spain (Díaz-Aguado et al., 2021) showed that the most common situations girls reported experiencing were: emotional abuse (insulting or ridiculing, 17.3%), general abusive control (deciding for me, down to the smallest detail, 17.1%), and control by mobile phone (14.9%). The percentage of boys who acknowledge perpetrating all types of DVAW is much less than that of girls who report suffering them. For example, 2.3% of boys recognized hitting their partner, while 3.6% of girls acknowledged having received physical abuse; 3.1% of boys acknowledged pressuring girls to have sex against their will, against 11.1% of girls who reported this situation; 2.8% of boys reported sending messages via Internet or mobile phone in which they insulted, threatened, caused offense or attempted to scare their partner, whereas 6.3% of girls acknowledged receiving such messages.

The comparison of prevalence in the last decade (Díaz-Aguado et al., 2021) showed an evident increase in male adolescent DVAW in Spain between 2010 and 2013, which related to the emerging use of digital devices to perpetuate it, and the changes in the nature of relationships as a result of these technologies (Díaz-Aguado et al., 2014). In contrast, 2013–2020 saw a clear decrease in male adolescent DVAW in Spain, as in the previous period based on male violence reported by girls and on boys recognizing such actions. Such violence mainly took the form of psychological abuse and control, which diminished in the latter period thanks to awareness campaigns aimed at adolescents disseminated across the media from 2013 onward. These changes are closely linked to others detected in the same study between 2013 and 2020: among adolescents (less identification with the underlying mentality of gendered violence and greater identification with equality); in the family (fewer messages that support male domination of women, and more conversations between adults and adolescents on DVAW); at school (more activities to build equality and prevent gendered violence). The changes detected in this study (Díaz-Aguado et al., 2021) relate to other important advances toward equality between men and women, and the struggle against gendered violence presented between 2013 and 2020 in the macro-system in Spain, such as the greater scrutiny of this type of violence by the media, the 2017 cross-party parliamentary pact against gendered violence or the mass demonstrations in support of equality that took place on 8 March from 2018, which drew attention from the international press for their huge turnout.

The questionnaires used in 2020 (Díaz-Aguado et al., 2021) included for the first time questions on sexual harassment online perpetrated by a boy against a girl outside a relationship. The results showed that a high proportion of girls aged 14–20 reported being victims of this new type of abuse against women. The situations most frequently reported related to receiving (48%) or requesting sexually explicit photographs (43.9%) from boys; 23.4% of the girls surveyed said they had received requests for online cybersex. As in other forms of male adolescent DVAW, the percentage of boys who recognize having indulged in this type of behaviors is much less than that of girls who report suffering such aggression. The most frequently reported situation, recognized by 17.1% of boys, is requesting sexually explicit photographs online; 10.6% reported having sent girls sexually explicit photographs of themselves, and 7.4% had asked for online cybersex with a girl outside a relationship they were involved in.

The study found a significant relation between having experienced male adolescent DVAW and all the sexual harassment online situations outside the relationship that the participants were asked about in the survey (girls, whether they had experienced such situations, boys, if they had carried them out).

The main aims of this study were to define a typology of adolescent males in relation to violence against women that they reported having perpetrated in three contexts (dating violence offline, dating violence online, and sexual harassment online outside a relationship). Taking a general population sample enables us to understand how different contexts of violent behaviors combine, and their relation to conditions that can be modified through preventive interventions: justification of male dominance and violence against women, justification of violence for conflict resolution, stress resulting from difficulty to comply the traditional male role, problematic Internet use, risk behaviors online and self-esteem. As the survey was aimed at the general population, it was expected that there would be a substantial group of teenagers who showed no kind of violent behavior, and other groups that differed in the type and intensity of the violent behaviors they reported. The non-violent group that we expected to find would be used to compare to the groups that displayed violent behavior. Secondly, it was predicted that one or two groups would be identified as intermediate in terms of the types of violence perpetrated and their risk conditions. Finally, it was predicted that the survey would detect a small minority of adolescents that exercised violence against women in three contexts, with the highest scores for the various risk conditions and the lowest self-esteem.

The study was designed around a probabilistic sample survey with stratified two-stage cluster sampling. The primary sampling unit was the school, and the secondary unit was the classroom. Depending on the school size, one, two or three full classrooms of students were randomly selected. The sample framework was the list of schools in Spain’s autonomous regions provided by the regions’ educational authorities. The sampling design was stratified by region and type of secondary education center (compulsory, academic, and vocational) with sizes proportional to the population sizes. In the Spanish education system, secondary education is divided into compulsory (12–16 years old) and non-compulsory (17–18 years old), the latter divided between academic and vocational. To determine the effective sample size controlling the possible effects of intra-school resemblance, an intra-class correlation of 0.10 was considered. In practice, the effect of school on the main variables (adolescent dating violence against women offline, adolescent dating violence against women online, and adolescent sexual harassment online) was less than 0.04 in all the relevant study variables (0.03 in harassment online, 0.02 in dating violence online, and 0.01 in dating violence offline). Consequently, the design effect was not corrected in the statistical analysis.

The initial sample consisted of 5,150 male adolescents, representative of all Spanish adolescents attending state or private schools of which those with no dating experience with girls (determined by an explicit question in the questionnaire) were not selected, as the objective was to study the dating relationships of boys with girls. The final sample included 3,132 Spanish teenage boys aged 14–18 (M age = 16.03, SD = 1.22). Concerning educational level, 57.4% were enrolled in compulsory secondary education, 29.5% in academic education, and 13.1% in vocational education. Regarding educational level of the parents, reported by the adolescents, 12.6% of the fathers and 9.6% of the mothers had completed an elementary level of education, 45.6% of the fathers and 44.6% of the mothers had completed secondary education, 25.8 of the fathers and 32% of the mothers had completed university education; 16% of the participants did not know the educational level of the father, and 13% of the mother. Considering the family situation, 77.9% lived with the father and mother.

The participants were enrolled at 251 secondary schools, 1,799 (57.1%) in compulsory education, and 1,353 (42.9%) in non-compulsory secondary education (918, or 29.1%, in academic, and 435, or 13.8%, in vocational). The mean number of participants per school was 20.2, ranging from 8 to 90, with a median of 18. The number of participants attending state schools was 1,984 (63.5%), and the participants from private schools numbered 1,138 (36.5%). A total of 2,875 students (91.8%) reported that they were native-born.

The principals of the schools selected were informed of the study and their participation requested.

Passive consent procedures were employed in accordance with recommended ethical guidelines. Parents of students under 18 had the opportunity to refuse consent for their child’s participation by returning a written form to the school office. Informed consent was requested from the 18-year-old students. All students were instructed that the survey was voluntary, they could withdraw at any time, and that their responses were anonymous.

Data collection at the school was carried out via Internet. A teacher remained in the room as the survey was administered to answer questions and resolve potential computer problems. The average time required to complete the questionnaire was about 50 min.

All measures were used and validated in previous research (Díaz-Aguado et al., 2011, 2021), and the technical aspects and detailed psychometric properties of the measures are available for inspection.

Data preparation was carried out using IBM SPSS v.27 software. All analyses were made with SPSS v.27, except for the exploratory factor analysis of some scales that were done with FACTOR v.10.10 software (Lorenzo-Seva and Ferrando, 2020).

Summary scores for some factors were obtained by adding up the scores for the items of the corresponding factors and dividing them by the number of items to maintain the scores within the original scale (self-esteem: 1–4, MGRS-II and MGRS-SW: 0–4, JMDVAW and JVCR: 1–4).

Before computing the summary scores, the items’ missing values were imputed using IBM SPSS v.27 software with the Expectation-Maximization (E-M) algorithm. The procedure is iterative and other variables are used to attribute a value (Expectation) followed by checks to see whether that is the most likely value (Maximization). If not, it re-imputes a more likely value. This continues until the most likely value is reached (Enders, 2010).

The main procedure used for the data analysis was two-step clustering that classifies individuals into groups based on their patterns of responses to sets of observed variables. It is an exploratory multivariate data analysis that identifies similar groups of objects within data sets with both continuous and categorical variables (Norusis, 2009). This study used the likelihood distance measure as the similarity criterion, and the Schwarz Bayesian Information Criterion (BIC) to determine the optimal number of clusters. Two-step cluster analysis also calculates the silhouette coefficient, which measures the quality of the clustering solution. Validity is established based on a combination of two different measures: cohesion, which is the closeness between the members of the same cluster, and separation, the closeness between the members and the centroids of the different clusters. The procedure is performed in two distinct phases. In the first step, a sequential clustering procedure is used to create pre-clusters by scanning the data case by case and assigning them to a previously formed cluster or a new one, based on the distance criterion. When all cases have been assigned to a pre-cluster, all objects in the same pre-cluster are treated as a single entity to reduce the size of the matrix that contains distances between all possible pairs of cases (Norusis, 2009). The second step consists of applying a hierarchical clustering algorithm to the pre-clusters and defining the best number of clusters based on a BIC criterion. Clusters were formed from three standardized quantitative variables: DVAW offline, DVAW online, and SHAW online.

Due to the lack of compliance with the assumptions of the parametric tests, the relationship between cluster membership and quantitative variables was examined by the Kruskal–Wallis non-parametric test with all pairwise comparisons; significance values were adjusted by the Bonferroni correction for multiple tests. Statistical significance for the comparisons was set at 0.01 due to the large sample size.

Rosenthal’s r statistic, with values between 0 and 1 (Rosenthal, 1991), was used to compute the effect sizes of the pairwise comparisons.

Finally, the variables that showed statistically significant differences among groups were entered as predictors in a multinomial logistic regression analysis.