Introduction

This study tries to explore the potential influence of social masculine traits on bullying. Our starting point is the research on the relation between bullying and gender stereotype traits, specifically, masculine and feminine ones, in the tradition of Bem’s (1974) pioneering research on gender stereotypes. More recently, Choi, Fuqua, and Newman (2008) have suggested their own model of gender stereotypical traits for adolescents consisting of three factors: one feminine factor and two masculine ones. We contend that their introduction of two dimensions of masculinity and the accompanying distinction between “social” and “personal” masculine traits may prove useful in better explaining the nature of bullying. More specifically, we hypothesize that if gender plays such an important role in school bullying it is because bullying is an attempt to demonstrate a social dominant position and an unbalance of power as a way to exert social control and establish a certain hierarchy among peers. In this respect, bullying arises as an expression of the “social” masculine traits such as the ones introduced by Choi et al. (2008, 2009). Our research, therefore, focuses on the relation between “social” masculinity and bullying behaviours in a sample of male and female Spanish adolescents.

Coexistence problems at school, including bullying, currently feature among topics that cause much concern within the education framework (e.g., Berger, and Rodkin 2009). Bullying is a phenomenon that occurs worldwide. Harel-Fisch et al. (2011) reported evidence of cross-national importance for adolescent behaviour in all 40 European and North American countries they studied, including Spain. Bullying also has a negative impact on perpetrators. In studies from the U.S. (Crick et al. 2002; Kochenderfer, and Troop-Gordon 2010), U.K. (Arseneault et al. 2006), the Netherlands (Bruyn et al. 2010), Italy (Gini, and Pozzoli 2009), Australia (Fitzpatrick, and Bussey 2011) and Spain (Navarro et al. 2012), victimizing youths have considerable problems including interpersonal relationships and psychological functioning. Being a perpetrator of bullying is associated with increased behaviour risk and a range of social and emotional problems. Perpetrators tend to feel chronically unsafe and to participate in a variety of antisocial behaviors (see Cowie 2013, in U.K.).

The sex/gender paradigm considers that the term sex refers to the physical characteristics that define people as men and women, while gender deals with the social rules that guide the conduct of men and women, thus presenting different ideals for each of them (e.g., Fernández 2010). The stereotypes of gender features have been linked to the concepts of masculinity and femininity. These concepts consist of common traits that strongly discriminated between men and women in the general population. “A personality characteristic was qualified as feminine if it was independently judged by both females and males to be significantly more desirable for a woman than for a man” (Bem 1981, p. 19), and vice versa. The items in the masculine subscale purport to measure socially desirable masculine characteristics, the items in the feminine subscale are intended to measure socially desirable feminine characteristics, with the central characteristic being “affective concern for the welfare of others” (Bem 1974, p. 156). Masculinity is built upon the perception that men have more instrumental features such as an independent and strong personality, while femininity is determined by the possession of expressive characteristics such as understanding of others and being gentle. The Social Role Theory states that men’s and women’s social roles cause differences in aggressive behaviour through the mediation of social and psychological processes (Eagly et al. 2004, in U.S.). One of these processes is learning masculine and feminine traits; for example, males learn that aggression is useful to maintain masculinity (Underwood et al. 2001, in U.S.).

The Three-Factor Model of Gender Stereotypical Traits

Research on adolescents from various countries confirms that the structure of gender stereotypical traits has evolved (England: Wilcox, and Francis 1997; U.K.: Colley et al. 2009; Spain: Fernández, and Coello 2010). Choi and Fuqua (2003) summarised the results of 23 factor analysis studies (from U.S., U.K. and Asian countries) using the Bem Sex Role Inventory (BSRI, Bem 1981) by stating that, ‘In spite of the differences in the number of factors retained in 23 factor-analytic studies, the model was one clean F (femininity) factor and two or more complex M (masculinity) factors’ (p. 883). In a more recent study with 661 students at an urban Midwestern university, Choi et al. (2008) confirmed the three-factor model. The structure consists of two masculine factors and a single feminine factor. These authors hypothesised that the resulting two factors seemed to reflect what they described as social and personal factors of masculinity. The items related more strongly with the personal factor included, for example: defends own beliefs, independent and self-sufficient. These descriptors have an internal, more personal, or self-control locus. The items associated more closely with the social factor of masculinity included: dominant, aggressive and forceful. These traits might all represent some form of social control of others. The three-factor model was invariant across gender groups, as well as the sample of 666 college students and 312 accountants from the U.S. using the short form of the BSRI (Choi et al. 2009). In these studies, women obtained higher scores for the feminine factor. For the masculine factor of social control, men scored significantly higher than women. There were no differences between men and women for the masculinity personal factor. This model is relevant for research because the traits forming the social masculinity factor coincide to a great extent with the traits defined as being a bully in the research.

Gender Stereotypical Traits and Bullying

The investigation has focused on the gender standards that can influence aggressive behaviour, placing emphasis basically on masculinity (e.g., Cohn, and Zeichner 2006; Dohi et al. 2001). Several studies have found that the most aggressive subjects are those that possess a strong and traditional masculine orientation, regardless of the gender of the one who attributes himself/herself these masculine traits (e.g., Kinney et al. 2001). When examining bullying among adolescents, gender is also a crucial aspect to consider (e.g., in US: Berger, and Rodkin 2009; Peeters et al. 2010).

Studies have suggested that for males and females bullying is also associated with masculinity. Young, and Sweeting (2004) analysed the association between scores in masculinity as well as the implication of bullying in a sample of 15-year-old secondary school students living in and around Glasgow, Scotland. The results showed that, without taking into account participants’ gender, masculinity was associated positively with the aggressor role in school bullying dynamics. In two schools in Stockholm, Sweden, Eliasson et al. (2007) used observations which included interviews of children aged 14–15 years and their class teachers, analysed verbal abuse and identified that masculinity contributes to bullying. They affirmed that, “boys who aspire to hegemonic masculinity can gain from the exercise of verbal abuse” (p. 601). Likewise, the studies done by Leach (2003) in Africa (Zimbabwe, Malawi, and Ghana) with adolescents examined males’ abusive behaviour toward females revealing that intimidation for males relates strongly to proving one’s masculinity. Going well beyond mere association between bullying and masculinity, Gini and Pozzoli (2006) performed a hierarchical regression analysis to explore to what degree “bullying can be predicted by masculine and feminine traits”. They found that masculinity “independently contributes to predict bullying” (p. 587).

In Spain, based on a qualitative methodology using focus groups of 198 adolescent students and 22 teachers whose average teaching experience was 13.90 years, research confirmed that masculinity can favour bullying (Yubero, and Navarro 2006). In previous research carried out by Navarro et al. (2011), employing two gender dimensions of Bem and aggressive tendencies, with 1654 Spanish high school students aged 12–18 years, from Castilla-La Mancha, confirmed the relationship between masculinity and high levels of bully perpetration. Males and females with a high adherence to masculine stereotype traits were more frequently involved in the perpetration of bullying.

A step ahead was clearly needed in order to align this line of research with the Three-Factor Model of gender stereotype developed by Choi et al. (2008). The question now becomes how the specific traits of masculinity not only relate, but contribute, to bullying. The analysis of bullying as the product of social dynamics arising in a group of peers has been provided by Rigby (2004) and as systematic abuse of power by other authors (Leach 2003; Smith, and Sharp 1994; Valls et al. 2008). A similar point of view is adopted by Pellegrini (2001) in his explanation of bullying as an instrumental form of aggression associated with social control in adolescents aged 13–17 years in the U.S. In the U. K., Lahelma (2002) used natural observations, conflicts between females and males aged 13–14 and 17–19, group interviews, and autobiographical memories in the U.S. This author concluded that individuals employ aggression as a form of social control to maintain hierarchical boundaries between genders. Salmivalli and Peets (2009), in the Netherlands, reviewed findings concerning the characteristics of individual bullies and suggested that the search for a dominant position within a group of peers is the basis for the motivation to bully. Bergeron and Schneider (2005) performed cross-national studies (derived from the studies of Hofstede, Bond, & Schwartz) on peer-directed aggression and associated bullying with individualism and dominance. In Finland, Björkqvist et al. (1982) used semantic differentials with 401 adolescents aged 14–16 years old. They also found that adolescent harassers considered themselves as being dominant. Phillips (2007), based on interviews and discussion groups with 32 adolescent boys in U.S., suggested that males used bullying to affirm their strength and dominance. These results are very important for our research.

There is less evidence for the relation of the role of feminine traits in bullying aggression processes. Conversely, the results of feminine traits are more inconsistent. Young and Sweeting’s (2004) study showed that femininity relates negatively with being a bully. Tapper and Boulton (2000) found a link between females’ greater expressive representations and poor aggression expression among 130 primary school children in the U. K. Crothers et al. (2005) studied the relationship between feminine traits and relational aggression in a sample of 52 girls from U.S., mean age 15 years old, with quantitative methodology and focus group interviews. These authors found that relational aggression was positively associated with feminine traits. However, they did not replicate this finding in a 697 female college student sample from U.S.; results indicated that femininity is unrelated to relational and social aggression (Kolbert et al. 2010). In Spain, Navarro et al. (2011) confirmed that traditional feminine traits relate negatively with bullying; adolescents who obtained a higher score in feminine gendered traits reported less implication in bullying behaviours. There were significant effects for feminine traits on physical bullying displayed by females, but no significant effects for males.

What is clearly lacking in this recent line of research is an analysis of how masculinity contributes to predict bullying, especially once it has been shown that masculinity, far from being a single factor, is an internally complex one. Taking the Three-Factor Model by Choi et al. (2008) as our starting point, we try to clarify in the present study the relationship between gender stereotypical traits and bullying and, more specifically, using regression analysis, to show that it is the social masculinity factor the one that contributes to bullying behaviour.

Current Research Overview

All reviewed research analysed the influence of the gender stereotypical traits from the two classical masculinity and femininity dimensions of Bem. To our knowledge, there are no studies available that analyse the relationship of bullying with the three-factor model of gender stereotype traits. The objective of the present study was to analyse the gender traits of the structure proposed by Choi et al. (2008, 2009), that is, feminine traits, social masculine traits and personal masculine traits, in native Spanish adolescents and to test the implications of this model for bullying. The two dimensions of masculinity may prove useful to better explain the nature of the influences of gender stereotypical traits on bullying.

Based on previous data, we can hypothesize that gender influences bullying as a way to affirm social masculine identity. Bullying would thus fulfill the function of showing off a social dominant position and an imbalance of power in the forms of social control and the maintenance of hierarchical boundaries.

Given that previous research focused on associations of bullying with social masculine traits, we expected that, from the two masculinity factors, personal and social, only the second would be associated with bullying aggression, while femininity was expected to be associated negatively with bullying aggression both for males and females (Hypothesis 1). Adolescents involved in bullying behaviour, males as well as females, would attribute to themselves more social masculine traits and less feminine traits (Hypothesis 2). We also expected social masculinity will predict active bullying behaviour and femininity factor will predict lower intervention bullying behaviour by males and females (Hypothesis 3).

Method

Participants

Our sample was composed of high school students (Compulsory Secondary Education in Spain), aged 12–18 years, from 27 public schools located in central Spain (Castilla-La Mancha and Castilla-León). Although the total sample consisted of 2761 students, only data about the Spanish participants, representing 93 % of the original sample, are presented here (n = 2560). The small subsample of immigrants (n = 201) was excluded from the analysis due to its internal heterogeneity. Table 1 provides information on the total number of participants in the final sample, as well as their distribution by age and gender. It is worthwhile noting that age distribution fits the extant situation in Spanish classrooms for this level of education.

Table 1 Demographic information
Full size table

Measures

Demographics

The questionnaire included age, gender and ethnicity.

Gender Stereotypical Traits

The Personality Traits Questionnaire by López-Sáez et al. (2008) was the instrument we used to measure gender stereotypical traits. Previously validated and used in representative samples of the Spanish population over the years (López-Sáez and Morales 1995; Morales and López-Sáez 1993), it showed adequate reliability (.72 on the global scale in López-Sáez et al. 2008) and was composed, as in the case of Bem’s Sex-Role Inventory, of positive and negative traits of masculinity (instrumental) and femininity (expressive). The nine instrumental masculine traits were “athletic”, “adventurous”, “egotistic”, “dominant”, “individualist”, “acts like a leader”, “aggressive”, “strong personality”, and “hard-hearted”, while the nine expressive/feminine ones were “submissive”, “loves children”, “cries easily”, “understanding”, “compassionate”, “sensitive”, “warm”, “affectionate”, and “soft-hearted” (see López-Sáez et al. 2008, p. 6, and Table 2, this paper, for the translation).

Table 2 Structure coefficients, communalities, means, and standard deviations for items of Personality Traits Questionnaire
Full size table

According to Choi and Fuqua (2003), “M factors (…) tended to be more complex” than M single factor in Bem’s Inventory and “the majority of the studies reported two to three factors derived from masculine items” (pp. 879–882). This fact led us to test Choi et al.’s Three-Factor Model in our sample of Spanish native adolescents via EFA. Commonalities lower than .40 were found in four items that were excluded from the analysis: athletic (deportivo/a) .30, individualist (individualista) .33, loves children (amante de los niños) .35 and submissive (sumiso/a) .34.The value of the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was .85, indicating that the fit of the data to the factorial model was acceptable. Bartlett’s test of sphericity yielded statistically significant results, χ 2(91) = 8619.56, p < .000, suggesting that the data are adequate for a factor analysis. Trait stereotyping structure was explained through a factor analysis of oblique rotation, the same one used by Choi et al. (2008, 2009). Three factors emerged accounting for 53.59 % of the variance. The first factor, femininity, includes seven items referring to expressive traits (e.g., warm, understanding, compassionate, affectionate) and explained 24.45 % of the variance. The second factor, personal masculinity, comprised three items referring to personality traits (e.g., strong personality, adventurous) and explained 21.18 % of the variance. The third factor, social masculinity, consisted of four items relating to social relationship characteristics (e.g., dominant, aggressive, leader) and explained 7.96 % of the variance. The multivariate F was significant, F(14,2545) = 7706.11, p < .000, indicating that measures of traits differed between males and females. The results of the analysis are shown in Table 2. An additional CFA verified not only this factorial structure but also its good fit. In this research, scale reliability (Cronbach’s α) was .79 for femininity, .60 for personal masculinity, and .73 for social masculinity.

Our data, then, show that masculinity is not a single factor, but an internally complex one. As in the case of the Three-Factor Model by Choi et al.’s (2008, 2009), there is a difference between a personal masculine factor and a social masculine one. As already established, bullying, as an instrumental form of aggression (see Pellegrini 2001), is expected to be predicted by this social masculinity factor. The alpha of .60 for the personal masculine factor could be questionable. Clearly, this factor, consisting only of three items, due probably to the fact that two masculine traits (“athletic” and “individualistic”) had low commonalities and had to be dismissed, demands further elaboration. However, the social masculine factor was the crucial in our predictions and its reliability of .73 was acceptable.

Participants were asked to apply each of the eighteen original items to describe themselves using a 7-point Likert type scale ranging from 1 (never or almost never true) to 7 (always or almost always true).

Previous studies using this questionnaire (López-Sáez et al. 2008) showed adequate reliability (.72 on the global scale). In this research, global scale reliability (Cronbach’s α) was .70, .79 for femininity, .60 for personal masculinity, and .73 for social masculinity. When all traits were included in EFA, four variables obtained commonality values below .40, which led us to eliminate them. After testing sample adequacy (KMO) and performing Bartlett’s sphericity test, EFA was recalculated (Izquierdo et al. 2014) obtaining a three factor structure. According to George & Mallery (2003), an alpha of .60 is questionable, although the number of variables defining each factor is something that has to be considered (see Conway and Huffcut 2003). Our Personal Masculinity factor was composed of three variables and, as suggested by Mundford et al. (2005), with less than four variables lower reliability values are usually found. In the absence of reliability values in the Choi et al.’s study, no comparison could be established. More important in our case is the indication by Nunnally (1978) that the demand level of reliability is contingent on how the measures are used and, as already noted, our crucial variable in the study is Social Maculinity and it does have an adequate reliability value.

Bullying

We used the Instrument to Assess the Incidence of Involvement in Bully/Victim Interactions at School (CAME; Rigby, and Bagshaw 2003) to measure bullying. It was translated and adapted to Spanish samples and had been previously used with Spanish adolescents (Navarro et al. 2011; Yubero et al. 2010). Participants were asked if they had participated in repetitive aggressive behaviours at school during the previous year. Four levels of intervention were considered: 0: never, 1: sometimes, 2: weekly, 3: daily. The five items of the scale are the following: physically direct bully “push, hit” (golpear, empujar); physically bully indirectly, “broke or hid things” (romper, esconder o robar cosas); verbally bully directly, “call names or insult” (insultar, poner motes); verbally bully indirectly, “said mean things behind him/her back or spread rumours about him/her” (hablar mal de otro a sus espaldas, rumorear); and bullying through exclusion, “ignore him/her or didn’t let him/her participate in games and other activities” (ignorar a alguien, no dejar participar de una actividad). In order to examine the participation in bullying aggression, the items were grouped for the analysis. Several authors agree that students with scores higher than a standard deviation above the mean comply with the frequency and intensity criteria that define the behaviour of bullying (in U.S., Espelage and Holt 2001; in Canada, Marini et al. 2006; in Germany, Schäfer et al. 2002; Scheithauer et al. 2006; in Australia, Yoneyama and Rigby 2006). Our bully category included only students with scored above 1 SD on the CAME scale. In previous studies carried out with Spanish adolescents (Navarro et al. 2011; Yubero et al. 2010), the bullying scale ranged from .65 to .76. In the present study, reliability was .76 using Cronbach’s α.

Procedure

For selection of the sample, we used a stratified multi-stage procedure (cities and educational cycles) by conglomerates with a selection of the primary sample units (educational centers) at random, and of the secondary units (school classroom) by proportional affiliation according to the stage of education and the final units (students) by quotas.

To select the centers, the Educational Council of the provinces covered by the study used a list of high schools in the study area. Geographical and population criteria were used. The selection included three centers in each province, one centre in the capital of the three provinces, another one for those areas with a population over 5000 inhabitants, and one for areas with a population under 5000 inhabitants. In each centre, the selection of the classroom was random. On the agreed date, a researcher went to the school and administered the questionnaire to the selected classroom. Previously, we had obtained passive consent from parents for their children to participate in the study. The tutors of each classroom that participated in this research were in charge of handing out the documents to students so it would reach their parents. A letter was sent to all parents to let them know about the nature of the study. In the letter parents were also asked to contact the tutors in case they did not want their son/daughter to participate in the study. No parent refused the participation of his/her child.

Questionnaires were completed by students during class time. Before the questionnaires were handed out to them, students were informed of the voluntary nature of their participation and that the anonymity of their responses was guaranteed. They were asked to respond as truthfully as possible. At the beginning, the researcher was reading aloud each item while the participants were following along on their own. They were given the opportunity to pose any question about the statements just read by the researcher. After the provided explanations, if needed, they proceeded to answer the questionnaire.

Results

Descriptive Statistic and Gender Comparisons

We tested for gender differences in gender stereotypical traits in a three gender stereotypic dimensions. Gender differences in the level of adherence to the three gender stereotype factors and were analysed through multivariate analysis. The multivariate F was significant, F(3, 2556) = 329,67, p < .000, η2 = .28, indicating that measures of gender stereotypic dimensions differed between males and females (see Table 3). Multivariate analysis indicated differences between males and females in self-adherence to feminine stereotype traits, personal masculine traits and social masculine traits. Males showed higher levels of personal and social masculine stereotype traits than females. On the other hand, females displayed significantly higher levels of feminine stereotype traits than males.

Table 3 Effects of gender on gender stereotypic dimensions
Full size table

Gender and Bullying

As noted previously, the bully category included those students who scored above 1 SD on the CAME scale. With these criterion, 367 bullies were identified, 179 female (13.7 %) and 188 male (15 %). Gender differences in the participants’ classification as bullies or non-bullies were analysed by a chi-square test. No gender differences were found in bully frequency, χ 2(1) = 0.92, p = .337.

Hypothesis 1 was tested with Pearson correlation. We had hypothesised that, from the two masculinity factors, personal and social, only the second would be associated with bullying aggression, while femininity was expected to be associated negatively with bullying aggression both for males and females. Partial correlation analyses were initially performed for males and females separately to examine the associations among personal masculinity, social masculinity, femininity, and bullying. The results, reported in Table 4, demonstrated a strong association between bullying, and personal and social masculinity for males and females. The association between bullying and femininity was not significant.

Table 4 Intercorrelations for females (over the diagonal) and males (under the diagonal)
Full size table

Effects of Bullying Intervention on Gender Stereotypic Dimensions

Gender traits and bullying effects were analysed through MANOVAs. It had been hypothesised that adolescents involved in bullying behaviour, males as well as females, would attribute to themselves more social masculine and less feminine traits (Hypothesis 2). We conducted a 2 (target bully) × 2 (participant gender) Manova with an alpha level of .05. As shown in Table 5, significant results were obtained for gender, F(3, 2554) = 174,63, p < .000, η2 = .17 and bullying behaviours, F(3, 2554) = 20.92, p < .000, η2 = .03, meaning that measures of gender stereotypic dimensions differed between participants involved and not involved in bullying behaviours. Bullies displayed higher levels of social masculinity, F(1, 2556) = 60.25, p < .000, η2 = .02, and personal masculinity, F(1, 2556) = 10.21, p < .001, η2 = .01. No differences were found for femininity, F(1, 2556) = 1.17, p = .279, η2 = .00. The sex x bullying interaction was not significant, F(3, 2554) = 1.24, p = .282, η2 = .00, but multivariate analysis revealed differences in personal masculinity, F(1, 2556) = 3.71, p < .054, η2 = .00, among females, F(1, 1306) = 0.84, p = .359, η2 = .00; among the males, F(1, 1250) = 12.56, p < .000, η2 = .01. Males and females participating in bullying behaviours had greater adherence to social masculine stereotype traits. In femininity traits, no differences were found between participants involved and not involved in bullying behaviours. Differences in personal masculinity did appear in boys, though not in girls.

Table 5 Means and standard deviations in relation of gender and bullying intervention
Full size table

We expected that social masculinity would predict active bullying behaviour and femininity factor will predict lower participation in bullying behaviour by males and females (Hypothesis 3). We tested hypothesis 3 using logistic regression. As predicted, the effect of social masculinity factor was significant in the role of bully by males, B = .37, SE = .07, p < .000 and by females, B = .32, SE = .06, p < .000. Additionally, personal masculinity in both males, B = .09, SE = .07, p = .15 and females, B = .00, SE = .06, p = .95 were not significant predictors. The femininity factor displayed an opposite relationship with bullying for males, B = −.06, SE = .07, p = .38 and for females, B = −.12, SE = .07, p = .07, but not significantly.

Next, we tested a Structural Equation Model with LISREL8.54 to confirm the model and the relationships existing among the variables. We followed the Maximum Verisimilitude procedure for estimations. We calculated several indices of the various measure types. We assessed the goodness of fit of the model proposed with a series of indices: Goodness-of-Fit Index (GFI), Comparative Fit Index (CFI), Normed Fit Index (NFI), Relative Fit Index (RFI) and the Root Mean Square Error Approximation (RMSEA). After typifying the variables in the model, we analysed the structural equation model to test the relationships between three gender factors stereotyping and bullying separately for males and females (see Fig. 1). The results for the females indicate a direct negative relationship between femininity and bullying, a positive relationship between bullying and social masculinity, no relation between personal masculinity and bullying. The calculated model provides acceptable data adjustment: GFI = .99, CFI = .98, NFI = .98, RFI = .90, RMSEA = .03. Also, adequate model adjustment in bullying for the males shows: GFI = .97, CFI = .96, NFI = .96, RFI = .90, RMSEA = .08. The data indicate a positive relationship between bullying and personal masculinity and social masculinity, less relation between femininity and bullying.

Fig. 1
figure 1

Regression analysis for a females, b males. *p < .05, **p < .01, ***p < .001

Full size image

In short, data from a regression computed separately for the boys and the girls on bullying showed that, regardless of gender, social masculine traits predicted active bullying behavior.

Discussion

Insofar as characteristics attributed to men and women determine what is acceptable in society for members of each gender group, they become important means for exercising power in relationships. Therefore, a significant influence of gender stereotype on intra- and inter-group relations should be expected (Vázquez, and Martínez 2011).

The three factors of Choi et al. (2008, 2009) gender stereotypical traits, that is, feminine, personal masculine and social masculine, reappeared in our sample of Spanish adolescents. Males assign to themselves more instrumental traits and females more feminine ones. Our findings revealed that girls increased their self-attribution of masculine traits and boys did the same with feminine traits, albeit at a significantly lower scale.

Gender stereotypes are dynamic and sensitive to societal changes (Diekman and Eagly 2000). Our findings are consistent with Twenge’s (1997) meta-analysis of samples of college students in the U.S., which indicated that women’s masculinity traits have been increasing and gender differences on masculinity have been decreasing over time. Also, in Turkey, Özkan, and Lajunen (2005), using 536 university students, found that some instrumental characteristics (i.e., independent, strong personality, defends own beliefs) were desirable for both genders. The socio-demographic and political changes that have taken place in the last 25 years have involved modifications in the perception of gender. López-Zafra et al. (2008) carried out a transversal study to analyse the evolution of the gender stereotypes in the U.S., Spain and Germany. Their results show that the evolution was greater in Spain, thus reducing the differences with other countries. In a recent study with a sample of 277 Spanish men and women aged 15–87 years, Lopéz-Zafra and García-Retamero (2012) confirm that young Spanish people present a more equalitarian vision of gender: young women are perceived with masculinity traits but young men are not perceived as having more expressive traits.

As Eaton and Rose (2011) indicated in their review of dating practices over the past 35 years in Sex Roles, there is still a prevalence of gender stereotypes in the cultural areas of adolescents that present stereotyped images of how men and women should behave serving many cognitive and social functions for success. The representations of gender may be influenced by accessible information about men and women through mass media. Television exerts a powerful socialising influence on viewers regarding attitudes toward gender. Soulliere (2006) examined 118 episodes of World Wrestling Entertainment, revealing that the messages emphasise hegemonic masculinity. In Spain, TV commercials also show gender stereotypes (Valls-Fernández, and Martínez-Vicente 2007) in a similar way to other countries such as Italy (Furnham, and Voli 1989), Portugal (Neto, and Pinto 1998), and Turkey (Uray, and Burnaz 2003). In this way, the exposition to models in other media: books, videogames and others show similar effects to the natural exposition of models exerting a modelling action on the stereotypes and gender behaviour. Studies carried out in several countries show qualitative and quantitative gender bias in textbooks (e.g., in Hong Kong, Lee, and Collins 2008; in U.S., Low, and Sherrard 1999; in Spain, Táboas-Pais, and Rey-Cao 2012). A recent study on the exhibition of women as a sexual object in video games, with 74 male participants aged 18–47 from U.S., showed that men exhibit sexist behaviours in social situations with women as a result of this (Yao et al. 2010). This masculinity model based on strength, dominance and aggression can have implications in the way social interactions among adolescents are produced (Soulliere 2006).

Gender and Bullying

The novel aspect of this research which also extended the literature on bullying involves the importance of social masculinity in relation to bullying aggression. The effect of gender stereotypical traits on bullying corroborates the role of gender. With respect to Hypothesis 1, our correlation results reveal a relation of two masculinity factors with bullying. Personal masculine traits present moderate association between bullying for males and females. Feminine traits, in both males and females, are negatively associated with bullying behaviour but this relation is not significant. These results confirm other studies which have revealed that adherence to masculinity gender traits are positively associated with being a bully for males and females (Eliasson et al. 2007; Gini, and Pozzoli 2006; Leach 2003; Navarro et al. 2011; Young, and Sweeting 2004; Yubero, and Navarro 2006).

As expected (Hypothesis 2), females and males participating in bullying behaviours displayed higher levels of social masculinity. Also, male bullies displayed more personal masculinity than non bullies, no differences were found for personal masculinity in females. However, no differences were found in femininity, in both males and females, between bully and non bully adolescents in bullying behaviours. Those adolescents who identified with traits of dominance and power were more prone to be aggressors of their peers. This relationship is present in both males and females. The association found for males and females is that they can legitimate the use of violence to gain recognition of others by exerting power (Leach 2003; Smith, and Sharp 1994; Valls et al. 2008) and group control (Pellegrini 2001; Phillips 2007). In fact, social masculinity represents an unequal distribution of power based on the control and dominance of others (Lahelma 2002). In this way, our results coincide with those of other authors (Björkqvist et al. 1982; Salmivalli, and Peets 2009; Valls et al. 2008) who described adolescents involved in bullying as dominant individualists who exert power over a group. The desire for leadership and power leads to bullying, which provides adolescents with an opportunity to construct the social reputation they desire.

We detected significant effects for social masculine traits on bully behaviours by males and females. But for feminine traits we detected a significant effect only for female participants in bullying (Hypothesis 3). This result is consistent with previous research showing that femininity reduces bullying situations for girls (Navarro et al. 2011; Tapper, and Boulton 2000; Young, and Sweeting 2004).

As Ovejero (2013, p. 9) stated, “school should be an efficient tool of social cohesion and of democratization integration of its citizens”. The observed results enable us to justify the need to continue undertaking educational actions that aim to eliminate gender stereotypes and favour equality. For school interactions, it is important to link interventions with those actions whose aim is to reduce traditional gender stereotyping. It is important to continue researching the deconstruction of traditional gender stereotypes. Some studies have shown that exhibiting various gender models, it is possible to modify or reduce stereotypes (Gartzia et al. 2012, with 301 Spanish employees and managers; Leszczynski 2009, in U.S.).

Conclusions

The main contribution of this study is the scientific certification of the weight of social aspects of masculinity in connection to bullying. In short, social masculine stereotypical traits seem to be a significant risk factor for involvement in bully perpetration. Findings of this study suggest the importance of providing an equal model for students to bring about behavioural change. These results emphasise the need of further designs of intervention programmers focusing on multi-dimensional aspects of personal traits. Spanish women have adopted roles and characteristics of masculine personality, but men still must eliminate social masculine traits in order to reduce of dominance and power in favour of gender equality.

Limitations and Suggestions for the Future

It is important to indicate that the tool-administering process is anonymous and that the accountability indices obtained in the questionnaires are acceptable. Adolescents’ responses in such tools can be subject to effects of social desirability and ranks. Moreover, a study such as this one, in which the data analysis is co-relational, requires careful data interpretation. Another limitation to be considered is the model employed. The next step would be to study the validity of the three-factor model with other samples and in different contexts. This is a new approach that needs to be tested and confirmed in future research.

The confirmation of the existence of two masculine factors, and the fact that it is the social masculine factor the one that predicts bullying, should not lead to overlook the somewhat low reliability of the personal masculine factor. Further research is needed in order to produce new evidence in support of a Three-Factor Model in Spanish samples with three highly reliable factors. For the time being, we have to accept that, even though our predictions hold, this is a limitation of our results.

Several reasons, both empirical and theoretical, may be invoked to account for the low reliability of the personal masculine factor. As already stated, the number of traits composing this factor is unusually low. Only three traits are included in it, probably due to very low commonality of some masculine traits of the original questionnaire that resulted in their dismissal. Another related explanation is the nature of our adolescent sample, so different in many respects from the original representative samples of the Spanish population employed in the studies by López-Sáez et al. (2008).

But, at the same time, theoretical explanations should be borne in mind. According to Messick (1989): “(W)hen presumed indicators of a construct split off onto another factor, some new construct distinctions may be called for” (p. 52). This applies, in our opinion, to our case here. An original single factor splits and two factors appear. One of them, the social masculine factor, shows an acceptable reliability and is a predictor of bullying. Its theoretical meaning has to do with the control motive, as captured by Fiske (2004): “people want to be effective, to have some sense of control and competence, and a lack of control provokes (…) an effort to restore control” (p. 20). Now, lingering doubts concerning the personal masculine factor arise not only for its low reliability, that has already been pointed out, but for the lack of clarity of its specific theoretical meaning. It needs to be studied on its own, and in a context different from bullying, in order to establish more clearly its basic social motive.

As far as bullying is concerned, it would be interesting to analyse the influence of gender traits with a three-factor model in males and females on the impact of bullying and the different forms of intimidation: will females with high social masculinity participate more in bullying? Will they employ bullying as a mechanism to maintain power? Will males with high femininity address indirect forms of bullying?

It would also be interesting for future research to carry out the analysis together with sexism in order to know the joint influence of gender socialization on bullying and explore this kind of relation with sexism. Young Spanish people have grown up in a democratic society alongside gender equality social policies where women have entered the labor world (Gartzia, and Lopez-Zafra 2014). The question is, have the same changes taken place in the gender stereotypes of Spanish adolescents regarding sexism? Some research (Lameiras et al. 2007; Spence and Bucker 2000) has confirmed the relationship between both gender variables. The connection between sexism and violence in adult couples is well-studied (e.g., Chapleau et al. 2008; Yamawaki et al. 2009). Studies about bullying are more restricted, but they have shown the role of sexism as the precursor of school harassment (DeSouza and Ribeiro 2005; Valls-Fernández and Martínez-Vicente 2007). It is true that learning about the joint influence of both variables might help us know the repercussions of gender on bullying.

The finding and conclusions of this study reveal various aspects that may be considered when designing intervention practices for bullying schools. The usefulness of adopting a gender perspective in prevention programmers of bullying behaviours may be more effective by focusing on gender stereotypical traits. Boys and girls who engage by bullying depend on the strong adherence to social masculine traits.

We think that our findings can help schools to promote a view fostering non-traditional gender if they wish to create positive interventions that improve, reduce or eradicate bullying in schools. Educators and other professionals in educational settings can remove the social masculine gender traits which support bullying in schools. While we carried out this study in Spain and exclusively with Spanish adolescents, it is just one piece of the puzzle. It is necessary to continue gaining more knowledge of gender stereotypical traits in adolescents and on the influence of femininity, personal masculinity, and social masculinity on bullying in other countries and cultures.