According to Keyes’s model (2007), positive mental health and well-being can be defined by the presence of subjective well-being, which goes beyond the absence of mental illness. In line with this theoretical framework, emotional well-being is one of the three main dimensions contributing to subjective well-being (along with psychological and social dimensions), referring to the predominance of satisfaction with life and the experience of positive emotions as opposed to negative ones (e.g., anger, sadness, fear; Diener, 1984). Cyberbullying is known to negatively impact this dimension of well-being in adolescents. For instance, students who were aggressors and/or victims of cyberbullying reported less satisfaction with their lives (UNESCO, 2017), and victims reported more emotional adjustment difficulties, such as depressive, anxiety, and psychosomatic symptoms and low self-esteem (e.g., Cowie, 2013; Ortega et al., 2012). On the other hand, a state of positive mental health and well-being can act as a protective factor and significantly change the impact of cyberbullying (Brailovskaia et al., 2018). Therefore, emotional well-being seems to play an important role in cyberbullying, especially in terms of the effect it may have on adolescents.

Anger is a predominant negative emotion shared by adolescents in cyberbullying events and also an important predictor of cyberbullying behavior (Lonigro et al., 2014). Specifically, anger can be expressed through aggressive behavior toward others (i.e., in aggressors), may be suppressed or may turn toward the self (i.e., in victims). In addition, coping negatively with anger seems to result in higher levels of cyberbullying behavior (Den Hamer & Konijn, 2016). Moreover, cyberbullying aggressors seem to reveal lower emotional regulation (Lin, 2017) and greater difficulties regulating negative emotions, such as sadness and anger, which seem to be related to different forms of aggressive behavior (Zeman et al., 2002). Considering the definition of emotional well-being (Diener, 1984), the predominance of negative emotions, such as anger, may lead individuals to experience low emotional well-being.

Anger as a personal factor (i.e., emotion) seems to determine cyberbullying behavior, which goes in line with a socio-cognitive perspective on the factors determining an individual’s behavior (Bandura, 2001). It is usually directed towards others through words in cyberbullying, since aggressive language is often used to communicate with others in this phenomenon (Veiga Simão et al., 2018). Accordingly, low levels of positive emotions might be related to anger and help explain the relation between cyberbullying and emotional well-being. Thus, we argue that an aggressive communication style may also be related to emotional well-being. Therefore, we hypothesize that:

Hypothesis 2: An aggressive communication style will predict adolescents’ emotional well-being, such that those who have this style of communication will report less emotional well-being.

This research considers the social cognitive theory of moral agency (Bandura, 2008) to conceptualize personal moral beliefs as being influenced by an aggressive communication style and as determiners of behavioral intentions. According to this perspective, individuals develop agency through self-regulation, which links thought to action (Bandura, 2001). In moral agency, individuals refrain from behaving in ways which violate their moral standards (Osofsky et al., 2005). Self-regulation of harmful behavior functions at a personal level and encompasses social aspects. Individuals are not autonomous moral agents, but rather behave morally according to cognitive, affective, and social influences (Bandura, 2008). Moreover, morality is socially founded, thus, moral standards function at individual and social level, and may cause substantial harm if these standards are not within social boundaries (Bandura, 2008). An aggressive communication style may be linked to how individuals interpret the world surrounding them and the specific circumstances they experience (Crick & Dodge, 1994). How individuals interpret the world is linked to the attitude they have towards these circumstances (Dill et al., 1997). These attitudes may be determined by other individual factors, such as personal beliefs, which Ajzen and Fishbein (1980) have defined as subjective cognitions about different aspects of life in general (i.e., behavior or attributes). An aggressive communication style is often a sign of someone wanting to protect his/her own ideas and opinions, as well as having them being accepted by others, even if it is at the expense of others (Salmivalli & Nieminen, 2002). That is, individuals with this communication style tend to interpret situations as if they were battles, which they want to win at all cost. In view of this, this study proposes that an aggressive communication style may be related to adolescents’ personal moral beliefs about cyberbullying. In accordance, we argue that:

Hypothesis 3: An aggressive communication style will predict adolescents’ personal moral beliefs about cyberbullying behavior, such that those who have this style of communication will believe that this type of behavior is fair.

There seems to be a strong relationship between adolescents’ attitudes towards cyberbullying and their behavioral intention (Heirman & Walrave, 2012). Specifically, attitudes towards cyberbullying seem to be a strong predictor of adolescents’ behavioral intention to engage in cyberbullying. To fully understand adolescents’ behavioral intentions, and consequently behavior, it is crucial to understand how personal and social factors influence psychological processes to yield behavioral effects (Bandura, 2005). This study focuses specifically on adolescents’ personal moral beliefs and how these may be related to their intentions to engage in cyberbullying. Social experiences affect behavior in the different contexts in which they unfold and how individuals interact within these situations influences how their moral standards develop throughout time and, therefore, guide their behavior (Bandura, 2005). This theoretical approach may help explain how these processes occur in cyberbullying, which is a social phenomenon (Allison & Bussey, 2017). In line with this, we argue that moral beliefs may determine adolescents’ behavioral intentions and consequently, their behavior. For instance, lower moral standards have been linked to higher levels of cyberbullying behavior (Perren & Gutzwiller-Helfenfinger, 2012). Thus, we focus on adolescents’ personal moral beliefs about cyberbullying behavior, since these beliefs may influence their actions (Barchia & Bussey, 2011). To specify, adolescents may not intervene or may even interfere in the situation in an aggressive manner if they believe the actions of the aggressor are justifiable, which constitutes a moral disengagement mechanism with regards to the situation (Allison & Bussey, 2017; Barchia & Bussey, 2011). Research has shown how moral disengagement could predict adolescents’ aggressive behavior in cyberbullying (Pornari & Wood, 2010). Therefore, we hypothesize that:

Hypothesis 4: Adolescents’ personal moral beliefs about cyberbullying behavior will predict their intentions to engage in this type of behavior, such that those who find cyberbullying behavior unfair will have lower intentions to engage in this type of behavior.

How individuals interact in face-to-face contexts is different from online environments, especially since pseudonymity and anonymity enable them to be more disinhibited (Bandura, 2004). Furthermore, online contexts enable individuals to disguise their identity and detach themselves from the physical world due to the lack of personal and social sanctions for injurious behavior. Therefore, individuals may feel a moral disconnection more easily when engaging in hurtful behavior and find it more difficult to regulate their moral conduct. In line with this, we argue that adolescents’ personal moral beliefs help explain the nature of the relationship between an aggressive communication style and their intentions to engage in cyberbullying behavior. Accordingly, we hypothesize that:

Hypothesis 5: Adolescents’ personal moral beliefs about cyberbullying behavior will mediate the relationship between an aggressive communication style and their intentions to engage in this type of behavior, such that the indirect effect will be lower than the direct effect.

In conclusion, this paper intends to achieve a better understanding of the role of a verbal aggressive communication style in cyberbullying. Thus, we studied the relationship with different individual variables, which may be associated with this phenomenon. Specifically, the aims of this study are to assess the relationship between an aggressive communication style and intentions to engage in cyberbullying behavior. We also tried to understand the role of an aggressive communication style on adolescents’ emotional well-being. Furthermore, we intended to understand how this communication style influences adolescents’ personal moral beliefs about cyberbullying, since they enable individuals to gain control over their thoughts, feelings, motivation, and actions (Bandura, 2006). Finally, to understand the specific role of personal moral beliefs as potential determiners of individuals’ actions (Bandura, 2006), another aim of this study was to determine the mediating effects of personal moral beliefs in the relationship between an aggressive communication style and adolescents’ intentions to engage in cyberbullying. Figure 1 presents our conceptual model.

A convenience sample of 218 9th graders (M_age = 14.67, SD = 0.84, 53% girls) from three public schools in Lisbon participated in this study. Student participation in data collection depended on students’ own volunteerism and parental consent. Specifically, all students from the schools were contacted. However, only those who had parental consent and gave their own consent participated in the study. The final sample was therefore not chosen by the research team, as this could create bias in the participation, but rather, only based on parental and students’ own consent.

During the initial development of the measures used in the present study within our research projects on cyberbullying (SFRH/BPD/110695/2015; PTDC/MHC-PED/3297/2014), facial validity was tested with three adolescents and by a panel of seven experts to understand whether the items of assessment instruments were appropriate for the specific constructs and assessment objectives (Hardesty & Bearden, 2004).

Aggressiveness in Interpersonal Communication (AIC). This instrument is a 10-item task and was developed based on Jakubowski and Lange’s (1978) theory of communication styles, specifically with regards to aggressiveness. As this resource is a performance task, and not a questionnaire, we decided to assess its internal structure and reliability with the Item Response Theory (IRT) approach by computing Rasch analysis with the Winsteps program (Linacre, 2013), which measured its unidimensionality, as well as participants’ scores of aggressiveness in hypothetical situations. Participants were asked to respond to daily life situations of adolescents (e.g., “A friend says to you [Your shirt is really horrible!]. How likely do you respond? [What’s that got to do with you?]”). In doing so, they should position themselves in each situation and respond in the manner presented in terms of likelihood from 1 (not likely at all) to 5 (very likely). Winsteps enabled us to estimate participants’ scores on a one-dimensional logit scale and evaluate the properties of the AIC.

We used Rasch polytomous methodology to examine the instrument and participants’ scores. All items were assessed to understand whether they had excessive infit and outfit mean square residuals. None of the items showed infit/outfit higher than 1.5, as well as z statistic > 2.00, as suggested in the literature (Bond & Fox, 2007). We considered other reliability indicators from Rasch measures for AIC, including Cronbach’s alpha, Person separation reliability (PSR), and item separation reliability (ISR) (Smith, 2001). AIC revealed a Cronbach’s α of .72, a PSR of .69, and an ISR of .99. These scores indicate good internal consistency reliability (Fox & Jones, 1998), even though the PSR revealed difficulty on participants’ behalf. After removing subjects with excessive infit/outfit, PSR was .70, ISR remained stable (.99), and Cronbach’s α increased to .73.

Personal moral beliefs about cyberbullying behavior. This instrument is a 9-item one-dimensional questionnaire that asks adolescents (on a Likert-type scale from 1 = fair to 6 = unfair) to think about whether the specific behavior of cyberbullying is fair or unfair (“I think seeing someone being threatened online is fair/unfair”; α = .81). After an exploratory factor analysis explaining 59% of the variance, the values of a confirmatory factorial analysis were good according to the literature in a previous study (Hooper et al., 2008), namely, χ²(22) = 26.55, p > .05, χ²/df = 1.21, CFI = .98, GFI = .93, IFI = .98, RMSEA = .02 [.00, .04], SRMR = .04, AIC = 72.55 (see Veiga Simão et al., 2018).

Behavioral intentions in cyberbullying is a two-dimensional questionnaire. This instrument asks adolescents (on a Likert scale of 1 = not probable to 5 = very probable) to indicate their behavioral intentions regarding cyberbullying behavior. After an exploratory analysis explaining 85% of the variance, confirmatory factorial analysis values were good according to the literature in a previous study (Hooper et al., 2008), namely χ²(134) = 296.06, p < .001, χ²/df = 2.20, CFI = .91, IFI = .91, RMSEA = .04 [.03, .05], SRMR = .09, AIC = 370.06. We only used “intentions to engage in cyberbullying behavior” (α = 0.91), which asked adolescents whether they would engage in cyberbullying behavior in the next six months (i.e., “If in the next six months I see someone threatening someone on the Internet, there is a possibility that I will do the same.”).

Emotional well-being sub-scale. This instrument is a 3-item sub-scale of the Portuguese version of the Mental Health Continuum-Short Form for adolescents (Matos et al. 2010), which evaluates the predominance of positive emotions and quality of life (e.g., ‘‘How often have you felt happy?’’; α = .79) on a 6-point Likert type scale (varying from 0 = never to 5 = every day). A confirmatory factorial analysis revealed good values according to the literature (Hooper et al., 2008), namely χ²(72) = 194.15, p <, χ²/df = 2.697, CFI = .93, IFI = .93, RMSEA = .06 [.05, .08], SRMR = .05, AIC = 260.15.

In a first phase, we requested and were granted authorization to conduct this study by the Ministry of Education of Portugal, the Portuguese National Commission of Data Protection, the Deontology Committee of the Faculty of Psychology of the University of Lisbon, schools’ boards of directors, teachers, parents, and adolescent participants. The instruments were administered to adolescents in a classroom context with computers with Internet access in their own schools by researchers of this study. Prior to filling out the instruments and performing the AIC task, all students were informed that their participation was based on confidentiality and their data would remain anonymous, and that they could have psychological support (i.e., with a professional psychologist) if they needed to talk to someone during or after participating. Moreover, we informed all students that they could quit at any time they wished to. All the students in the final sample chose to participate.

Database was previously treated for missing values by transforming raw data into z scores and by removing outliers. Before performing structural equation modeling, we computed Pearson correlations of the variables included in our structural equation model. We evaluated the significance of regression coefficients with AMOS (v. 23, SPSS Inc., Chicago, IL). Parameters were estimated through the asymptotically distribution-free method due to the non-normal distribution of data, which is characteristic of cyberbullying data. Normality of variables was evaluated with the univariate and multivariate skewness and kurtosis. We used χ² tests to assess the significance of the total, direct and indirect effects (Marôco, 2010). Moreover, effects p < .05 were considered significant. Also, we used the bootstrapping method (2,000 samples, CI 90%) to test for mediation effects (Preacher & Hayes, 2008).