Three thousand, four hundred and eighty-nine (N= 3,489) students of the second stage of Compulsory Secondary Education (E.S.O.) and the Spanish Baccalaureate (Bachillerato) attending secondary schools in the eight provinces of Andalusia, Spain participated in the study (girls 51.1% and boys 48.9%). The age range was between 12 and 18 years (average age =16.85; SD=1.24). The sample was determined based on simple random sampling, using the levels corresponding to Compulsory Secondary Education and the Baccalaureate as independent samples and by calculating the number of participants according to these levels (Moreno, Martínez, & Chacón, 2000). The data contained in the “Education Statistics of Andalusia” of the Ministry of Education's Statistical Unit, Regional Government of Andalusia, were consulted. Using a table of random numbers, a total of 48 schools were selected: 28 for Compulsory Secondary Education and 20 for Baccalaureate (state and state-subsidised schools).

Sexual Harassment Survey (AAUW, 1993). The adapted version by Ortega and colleagues (2010) was used. Numerous studies have used this instrument to measure sexual harassment in adolescence (McMaster et al., 2002; Ortega et al., 2010). The questionnaire asked students to report how often they had perpetrated or received a variety of sexual harassment behaviour during the last three months. The questionnaire instructions explicitly asked to report only unwanted a sexual behaviour. Thirteen items measured aggression and thirteen items measured victimization with a format scale on five points (from 0= never; 4= daily).

The instrument was administered with the prior consent of family members and the participating schools, during school time in a 60-minute session, by trained researchers. The students were informed that their responses to the questionnaires would remain anonymous. The rejection rate of students to complete the questionnaires was 1%. A survey implementation protocol was used to avoid any potential biases in data collection (Moreno et al., 2000).

Prior to conducting the CFAs, the occurrence of each item for sexual aggression and victimization in boys and girls was analyzed. In all cases, item frequency was at least 1% except for female aggression, where four items presented lower percentages (items 7, 8, 9 and 12), according to previous studies (McMaster et al., 2002; Ortega et al., 2010) these items were removed from the analysis (see Table 1). The statistical software LISREL 9.1. (Jöreskog & Sorbom, 2012) was used to perform confirmatory factor analyses. Two models for aggression and another two for victimization were tested among boys and girls separately: a one-dimensional model comprising all the behaviours contained in the instrument, and a two-dimensional model encompassing the visual/verbal and physical. Because of the non-normal distribution of the study population, the large sample size, the ordinal response format (Flora & Curran, 2004) and high correlation factors (Holgado-Tello, Chacón-Moscoso, Barbero-García, & Vila-Abad, 2010), diagonally weighted least squares (DWLS) estimation was used, whereby estimations are based on a asymptotic covariance matrix of the data (Morata-Ramírez & Holgado-Tello, 2013). In order to compare the suitability of the chosen models, this study considered the indexes proposed by Hu and Bentler (1999): CFI, GFI and NNFI values greater than .95; an RMSEA less than .05; and SRMR values below .08 as acceptable and less than .05 as adequate. The Parsimony Goodness-of-Fit Index (PGFI) takes into account the complexity of the hypothesized model in the assessment of overall model fit. PGFI value equal to .67 is acceptable (Byrne, 2013). The Expected Cross-Validation Index (ECVI;Browne, 2000) was measured to obtain a comparison of both models, where the model which yielded a lower ECVI value would have the most potential for replication (Morata-Ramírez & Holgado-Tello, 2013). To calculate involvement in visual/verbal and physical sexual aggression and victimization, the scores were dichotomized, with zero indicating that adolescents reported not having been involved in any scenarios either as aggressor or victim in relation to the behaviours subject to analysis. In contrast, they were considered to be involved if they answered affirmatively to any of the items. The percentages of involvement are shown in Table 3.

Models were tested for female and male victimization and aggression, obtaining a total of eight models: four one-dimensional and four two-dimensional types. The one-dimensional models contained 13 items to measure male aggression and male and female victimization. Nine of the models’ items tested female aggression. The two-dimensional models differentiated between visual/verbal aggression and victimization (items 1, 3, 4, 5, 11 and 12) and physical aggression and victimization (2, 6, 7, 8, 9, 10 y 13). In terms of female aggression, the physical scale was represented by three items (2, 6 and 13). The mean scores and standard deviations of the scales were: visual/verbal aggression -girls M=0.09; SD=0.22; boys M=0.35; SD=0.56-; visual/verbal victimization -girls M=0.21; SD=0.33; boys M=0.41; SD=0.57-; physical aggression -girls M=0.03; SD=0.19; boys M=0.15; SD=0.46-; physical victimization -girls M=0.05; SD=0.22; boys M=0.18; SD=0.44-.

Table 2 shows the results obtained. The fit indexes were correct for both the one-dimensional and two-dimensional models. The reliability indexes of the scales for the one-dimensional models were good. As for the two-dimensional models, the reliability indexes yielded better outcomes for the physical scale than for the visual/verbal scale in both male and female victimization and aggression, especially among girls. Reliability was acceptable in all cases except for female visual-verbal sexual victimization and aggression, where slightly lower values were found (.66 and .61). The correlation between factors reached moderate to high values across the board, reaching .90 for male sexual aggression. This led us to consider whether we should opt for a global or a two-dimensional model. Finally, and bearing in mind that the one-dimensional model could also fit the data for boys, we opted for the two-dimensional model. This was in spite of the high correlation between latent factors and which did not produce a substantially better fit than the one-factor model, namely for the following reasons: because a) the global fit indexes (RMSEA, NNFI, GFI and SRMR); b) the X2 test (chi-square/degrees of freedom); c) the replicability index (ECVI); and d) factorial loads and item errors yielded better results in the two-dimensional models. Furthermore, according to the theory furnished by Holgado-Tello et al. (2010), in the field of psychology and interpersonal relationships, and more so when talking about phenomena that involve violence, it is highly likely that the behaviours being measured are closely related, although this does not mean that they are measured exactly the same way. Further support for the decision to adopt the two-factor model resides in the recommendations made by some authors who have studied the dimensional structure of adolescent sexual harassment. Two-dimensional structures were found to better fit the data than the one-factor kind, although the differences between the one- and two-dimensional models were not substantial.

The graphical solutions of the aggression and victimization models (see Figures 1, 2, 3 and 4) found all saturations to be above .50 and below 0.94.

Figure 1.

Two-dimensional model of female sexual victimization.

(0.26MB).

Figure 2.

Two-dimensional model of male sexual victimization.

(0.27MB).

Figure 3.

Two-dimensional model of female sexual aggression.

(0.21MB).

Figure 4.

Two-dimensional model of male sexual aggression.

(0.26MB).

As the results show, the highest percentage of involvement was found in visual/verbal rather than physical forms, and in victimization over aggression. Whereas visual/verbal victimization exceeded 50% across both sexes, visual/verbal aggression maintained the same levels in boys but fell to 25% in girls. In terms of the physical forms, victimization reached around 30% involvement, whereas aggression did not exceed 25%. Boys showed more involvement than girls in visual/verbal forms and, fundamentally, in physical forms.

The first aim of this research was to test a one-dimensional and two-dimensional model of sexual victimization and aggression among adolescent peers by studying boys and girls separately, thus contributing to the small body of research available on the dimensions and factors related to adolescent sexual harassment. The results obtained for sexual victimization in girls and boys confirmed that the two-factor models (visual/verbal and physical contact) reported some fit and replicability indexes slightly more suitable than the one-factor kind. Thus, we chose to adhere to the approach taken in those studies that defend the two-dimensional structure as the most adequate when it comes to defining the adolescent sexual harassment construct (McMaster et al., 2002; Ortega et al., 2010; Wei & Chen, 2012; Witkowska & Kjellberg, 2005).

If we compare our results with those obtained in research using similar statistical analyses (CFAs), we will find that the model developed in the present study shares the two-factor structure approved by Witkowska and Kjellberg (2005), as well as instrument validation for girls and boys separately. However, these authors did not find a satisfactory victimization model that fits both sexes. Therefore, the findings of this study are seen as a step forward in identifying the dimensions for female and male victimization and aggression, highlighting the importance of aggression models and, most importantly, male ones, given that this is the first Spanish study that has successfully tested models whereby the data pertaining to sexual aggression fit well for both sexes, albeit slightly better among girls.

Victimization models. The CFAs have shown both the one-dimensional and two-dimensional models to correctly fit the data in the female and male samples. However, the two-dimensional model (visual/verbal and physical) was deemed most suitable for both boys and girls as indicated by the fit indexes. This seems to indicate that irrespective of sex, adolescents perceive two forms of sexual victimization in peer sexual harassment: one characterized by insults and behaviours with an important visual component of a sexual nature, and another characterized by including physical contact. This constitutes a relevant contribution to the scientific field of adolescent sexual harassment given that, to date, there were no victimization models that fit both male and female samples.

Aggression models. The models for female and male aggression presented the same structure as those for victimization, although with differences in the female model, where the physical sexual aggression scale fell from 7 to 3 items. Hence, it appears that the factorial structure is similar to that observed for female and male victimization and for male aggression. However, the results seem to highlight the importance attached to the expression of behaviours that boys and girls perpetrate and receive; while they seem to be the same in nature, both sexes express them differently. It is for this reason that girls failed to acknowledge performing certain behaviours that involved physical contact. Specifically, boys were found to express themselves by displaying different behaviours when it comes to sexual aggression, whereas girls primarily did so via visual/verbal behaviours, meaning that the severity of peer sexual harassment may be greater in the male population.

We also opted for two-dimensional models to measure aggression; although the one-dimensional model fit the data correctly, the two-factor model yielded better outcomes. However, this was not an easy decision to make for the male aggression model, given that the high correlation between latent factors (visual/verbal and physical) posed difficulties when choosing the model. Finally, we also decided on the two-dimensional model to measure male aggression; we found theoretical support for selecting the two-dimensional model as a better option than the one-dimensional model (Ortega et al., 2010; Wei & Chen, 2012; Witkowska & Kjellberg, 2005). Furthermore, the chi-square, goodness-of-fit, and replicability indexes also proved to be slightly better in the two-dimensional model.

Thus, it can be said that peer sexual harassment seems to clearly distinguish two types of behaviours in both female victimization and aggression and in male victimization (visual/verbal behaviours and behaviours involving physical contact), and somewhat less conclusive but acceptable in male aggression. It is thereby fair to say that this study contributes to the small body of research that has examined the dimensions of peer sexual harassment.

The second aim of this research was to analyze the prevalence of sexual harassment among peers. From this perspective, the results revealed higher indexes of victimization than aggression, as well as greater involvement by boys than by girls in both phenomena (AAUW, 2011; Wei & Chen, 2012). The prevalence of visual/verbal victimization involvement yielded the highest rates in the male sample. Physical victimization was less pronounced in both the male and female populations, although boys were two times more likely than girls to be physically victimized (33.8% boys and 14.2% girls). Our findings confirmed prevalence levels in visual/verbal victimization similar to studies that reported the highest indexes (AAUW, 2011; Mumford et al., 2013; Ortega et al., 2008), whereas the results corresponding to physical victimization coincided more with those studies where the percentages ended up being lower (Bucchianeri et al., 2013; Ortega et al., 2010; Wei & Chen, 2012).

The aggression prevalence results yielded higher percentages than most of the previous international studies that used the Sexual Harassment Survey (Espelage et al., 2012; McMaster et al., 2002; Mumford et al., 2013), except for the study by the AAUW (1993) which found that more than half of boys and girls had assaulted their peers (66% boys and 52% girls). What is more, the majority of authors did not differentiate between one form of violence and another, which poses a problem when it comes to us making comparisons. What is clear from the results of this study, and those that have identified less and more severe displays of aggression, is that adolescents use milder forms of aggression, mainly verbal (Mumford et al., 2013; Ortega et al., 2010).

In the same vein, the results have shown that both sexes share the same most frequent and less frequent behaviours. Most boys and girls were recipients and initiators of “sexual looks, comments, jokes and movements”. This observation finds support in most international research (AAUW, 2011; Attar-Schwartz, 2013; Tillyer et al., 2010) as well as on a national level (Ortega et al., 2010; Vicario-Molina et al., 2010). The behaviour that fewer participants reported to have experienced and perpetrated was “forcing someone to do something other than kissing”. In this case, previous research has not come to a conclusive argument regarding this point, primarily because data regarding the less common behaviour are usually not provided. However, the scientific community is generally in agreement that the most frequent behaviours are those that entail some form of forced physical contact (Espelage et al., 2012; McMaster et al., 2002; Ortega et al., 2010).

The two-dimensional approach chosen to analyze and explore adolescent sexual harassment is arguably one of the strengths of this study. This approach has allowed us to more accurately gain a sense of how these types of violent sexual relationships work over the course of adolescence, to enable us to respond more appropriately to the needs of both sexes. Specifically, this research has served to further develop understanding of the male victimization model as well as that of male victimization and aggression, given that we have successfully validated aggression models in both girls and boys.

Another strength of this study resides in the fact that it constitutes one of the first prevalence studies carried out in Spain, which gives us a more realistic picture of peer-to-peer sexual harassment.

In terms of the study's limitations, a high correlation between factors was found in the two-dimensional model of male aggression; as a result, the data should be interpreted carefully. A replication of the model in a sample with similar characteristics to those addressed in this study is also warranted. That said, the replication data yielded slightly better outcomes in the two-dimensional model, which was one of the reasons why we opted for this particular model.

The instruments used in this study also came with certain limitations, one of them would be inherent to use of self-reports, because of social desirability that could be associated with, and another one could be the lack data supporting the criterion validity of this questionnaire. Also the inclusion of qualitative measures to this set of questionnaires would have helped us to evaluate other aspects that we were unable to assess. Examples may include focus groups or interviews.

Following on from this study, we consider it important to continue exploring the evolution of sexual harassment over the course of adolescence, the effects that sexual harassment has on male and female populations, and the coping strategies that both sexes employ when faced with these aggressive behaviours. Furthermore, comparing this and other phenomena that occur during this stage of the life cycle is also a topic that would be interesting to pursue. It would also be advantageous to examine which variables predict and influence involvement in these types of phenomena.