English (pdf)
Article in xml format
Article references
Automatic translation
Send this article by e-mail
Cited by SciELO 
Cited by Google 
Similars in
SciELO 
Similars in Google 
Permalink
Introduction
Adolescents frequently use digital devices and content (Organisation for Economic Co-operation and Development [OECD], 2019). Adolescents frequently use digital devices and content (OECD et al., 2020). Although differences in Internet access persist among Latin American countries, some studies have reported adolescents' frequent use of this resource for entertainment, socialization, and learning (OECD, 2020; Trucco & Palma, 2020). In Mexico, the National Institute of Statistics and Geography [INEGI], 2020) reports that 90% of adolescents between 12 and 17 years old are frequent internet consumers. Although Internet use has brought substantial benefits to adolescents' cognitive and social development (Ang, 2017; Skryabin et al., 2015), it has also had adverse effects, including online aggression (Chester et al., 2019; Rice et al., 2015).
Cyberbullying involves aggressive, repetitive, and intentional attacks through electronic devices against victims who cannot easily defend themselves (Hamm et al., 2015; Smith et al., 2008). Studies have revealed a substantial prevalence of cyberbullying in Latin American countries such as Ecuador (Calmaestra et al., 2020), Colombia and Uruguay (Yudes-Gómez et al., 2018), Argentina (Pengpid & Peltzer, 2022), and Brazil (Malta et al., 2022). In Mexico, empirical studies show that 20 – 30% of Mexican students have been victims of cyberbullying at some point in their schooling (INEGI, 2021; Madrid-López et al., 2020). Cyberbullying can have long-lasting adverse effects on youth mental health, including depression, anxiety, substance abuse, suicidal ideation, and attempts (Graham & Wood, 2017; Iranzo et al., 2019; Pengpid & Peltzert, 2022; Santos et al., 2021). Moreover, cyberbullying is strongly related to school absenteeism, low academic performance, and negative perceptions of school climate (Hinduja & Patchin, 2008; Ortega & González, 2016).
Throughout literature, cyberbullying is perceived as a social phenomenon that involves an aggressor, a victim, and a bystander. While the roles of aggressors and victims have been broadly explored, the role of bystanders is still understudied despite their potential to inhibit aggression and promote victims' well-being (Machackova & Pfetsch, 2016; Zych et al., 2019). Overall, scholars agree that bystanders should be behavior in cyberbullying events: passive, reinforcement of aggression, or defender of cyber victims (Mallmann et al., 2017; Olenik-Shemesh et al., 2017; Song & Oh, 2018). While passive and reinforcing bystanders have been shown to promote cyberbullying and its harmful effects on victims, defenders are expected to be potential shields that may not only hinder cyberaggression but also attenuate its negative effect on victims (DeSmet et al., 2019; Holfeld, 2014; Torgal et al., 2023).
Bystander-defender behavior involves stopping cyber aggression or comforting cyber victims (Olenik-Shemesh et al., 2017; Sarmiento et al., 2019). Traditionally, bystander defenders' interventions have been conceptualized and measured as one-dimensional constructs (see Reijntjes et al., 2016; Salmivalli et al., 1996). However, current studies suggest that defender intervention is a multidimensional construct involving aggressive and constructive interventions to help victims (Bussey et al., 2020; García-Vargas et al., 2023; Moxey & Bussey, 2020). Defending constructive interventions include prosocial behaviors that may be oriented toward victims (by providing support and orientation) or toward the bully (by trying to stop aggressive behaviors and motivating the aggressor to apologize to the victim) (Cassidy et al., 2013; DeSmet et al., 2016). However, the intervention of aggressive bystander defenders implicates retaliation behavior against the bully (e.g., spreading rumors or posting images or videos denigrating the aggressor) (Bussey et al., 2020; Moxey & Bussey, 2020).
Despite being aimed at defending victims, bystander intervention effectively stops aggression only when it is constructive; otherwise, it contributes to increasing violence (Moxey & Bussey, 2020; Pronk et al., 2019). Nonetheless, adopting a constructive or aggressive stance remains unpredictable, because it relies on individuals' psychological resources and moral development. In this regard, Bussey et al. (2020) reported that self-efficacy and low moral disengagement are positively related to constructive defender intervention and negatively related to aggressive intervention. A recent study (Valdés-Cuervo et al., 2021) shows that moral guilt and sympathy are negatively related to the aggressive defending intervention.
Measures of Bystander Defender Behavior in Cyberbullying
Salmivalli et al. (1996) define bystander defender intervention as a unidimensional construct that combines indicators of constructive and aggressive interventions. However, the current scales (DeSmet et al., 2016; Ferreira et al., 2020; Pozzoli & Gini, 2020; Sarmiento et al., 2019) exclude items aimed at assessing aggressive interventions. The Styles of Bystander Defender Intervention scale (SBDI; Moxey & Bussey, 2020) is the only scale found by the authors that measure bystander defender intervention as a multidimensional construct that comprises both constructive and aggressive interventions.
Moxey and Bussey (2020) used a sample of Australian high school students to report the validity of interpretation based on the SBDI. However, this measure presents at least five issues that must be addressed. First, no known study by the authors has compared the adjustment of the two-dimensional model proposed in the SBDI scale with the one-dimensional model, as traditionally proposed (see Salmivalli et al., 1996). Second, cross-validation was required to ensure the stability of the measurement model. Third, no study has examined the discriminant validity of each subscale, a condition needed to verify the construct's uniqueness (Shiu et al., 2011). Fourth, few studies have examined the measurement invariance of SBDI in critical variables such as gender and educational level. Fifth, the analysis of evidence based on relationships with other variables remains limited. Furthermore, no studies have explored the validity and reliability of the scale score to measure constructs among Mexican adolescents.
Gender Differences in Bystander Defender Intervention
Although limited, previous findings of gender effects in bystander aggressive and constructive intervention indicate that constructive defending intervention was more frequent in females than males, whereas aggressive intervention was more prevalent in males (Bussey et al., 2020; Moxey & Bussey, 2020). Regarding the length of enrollment, whereas Moxey and Bussey (2020) found fewer constructive interventions for 9th-grade students than for 7th-grade students, Bussey et al. (2020) reported significant differences. Nonetheless, these studies did not examine the scale's measurement invariance by gender. It is essential to mention that measurement invariance ensures that differences between groups result from variances in the expression of the construct, rather than from measurement bias (Byrne, 2016; Van de Schoot et al., 2015). Therefore, it is necessary to establish measurement invariance for meaningful group comparisons.
Relationships With Related Constructs
Based on theoretical considerations, we expected that both types of bystander interventions would be differently associated with cyberaggression and moral identity. In line with previous literature (Chan & Wong, 2019; Marín-López et al., 2019), we posit that constructive bystander interventions are negatively associated with cyberbullying, and aggressive interventions are positively associated.
Constructive bystander interventions are associated with higher moral emotions and lower moral disengagement (Bussey et al., 2020; Moxey & Bussey, 2020; Valdés-Cuervo et al., 2021). We expected that moral identity, the importance of being a moral person based on the person's identity (Gibbs, 2014; Hardy & Carlo, 2011), would encourage constructive interventions and hinder aggressive interventions.
The Present Study
The present study proposed examined the validity of the interpretation of scores of Styles of Bystander Intervention in Cyberbullying (SBDI) to describe the frequency of constructive and aggressive styles of intervention in cyberbullying events in Mexican adolescents. The study assumed that validity is a unitary concept, so various sources of evidence should be integrated into the judgment of the validity of the interpretation (American Educational Research Association et al., 2014; Zumbo et al., 2014). To examine evidence based on internal structure: (1) calculate several confirmatory factorial analyses (CFA) to compare the goodness of fit of the one-dimensional and two-dimensional measurement models of bystander defender intervention (see Figure 1). (2) Verify whether the differences between the factors are empirically grounded. (3) Cross-validation was performed to examine the stability of the internal structure in an independent sample. (4) Use robust measures to ensure scale reliability (McDonald's Omega and variance extracted). (5) Explore scale measurement invariance by gender and educational level (secondary vs. high school). (6) Comparison of latent means by gender and educational level. Additionally, (7) we examined evidence validity based on the scale dimensions' relations with external variables such as cyber aggression and moral identity.
Method
We used a non-probabilistic sample of Mexican adolescents. The sample came from urban secondary schools (N = 68) and high schools (N = 68) in Sinaloa (Sample 1) and Sonora (Sample 2), Mexico. These schools are located in the middle and low-class areas of these cities. In both samples, nine participants were recruited from each school (three from each grade). We ensured that the ratio of male to female students was similar to that in the schools. Sample 1 (calibration sample) included 612 (nine from each school) adolescents (42.3% males and 57.7% females), with 306 (50%) secondary students aged from 12 to 15 years old (M years = 13.2, SD = 1.04), and 306 high school students (50%) aged from 15 to 19 years old (M years = 16.2, SD = 1.01). Sample 2 (cross-validation sample) contained 612 (nine from each school) adolescents (43% male and 57% female), with 306 (50%) secondary students aged from 12 to 15 years old (M years = 12.9, SD = 1.24), and 306 (50%) high schools students aged from 15 to 19 years old (M years = 15.8, SD = 1.11).
Measures
Styles of Bystander Intervention in Cyberbullying Incidents
The Styles of Bystander Defender Intervention Scale (SBDI; Moxey & Bussey, 2020) was used. The scale includes 15 items grouped into two dimensions: constructive defending intervention (10 items, e.g., encouraging the child to report being picked on) and aggressive defending intervention (5 items, e.g., sharing humiliating images or videos of the bully). Likert scale response with five points (0 = never to 4 = always) was used to answer questions such as, "During the last term, how often did YOU respond to defend a cyber-victimized kid?"
A back-translation procedure was adopted to translate the items from English to Spanish. Three professional bicultural bilingual translators participated in the scale translation. One translator, whose mother language was Spanish, translated the English version of the scale into Spanish. A second translator, whose mother language was English, translated the Spanish version back to English. A third translator certified the equivalence of these versions.
Cyberbullying
The Adolescent Cyber-Aggressor Scale (CYB-AGS; Buelga et al., 2020) was used. This scale comprises 18 items to measure the frequency of harassment and intimidation suffered by classmates using the Internet and social media (e.g., average variance extracted AVE = .58, McDonald's Omega ω = .83). The scale uses a Likert response format (0 = never to 4 = always). The CFA evidence the model fit to the data (SBX 2 = 163.43, df = 132, p = .033; SRMR = .04, CFI = .98, TLI = .97, RMSEA = .05, 90% CI [.03, .06]).
Moral Identity
The Moral Identity Scale (MIE; Aquino & Reed, 2002) was used to measure the construct. A back-translation procedure was used to translate the scale from English to Spanish. The scale comprises two dimensions: (1) internalization, with four items to assess the level of importance that individuals assign to moral traits (e.g., I feel good to be a person who holds features such as compassion, kind, fairness, generosity, and honesty; AVE = .62, ω = .74), and (2) symbolization, with five items to assess the level of moral traits reflected in individuals' behavior (e.g., the things I do during my free time portrays me as a person that holds: compassion, kindness, fairness, generosity, and honesty; AVE = .51, ω = .88). The items were rated on a Likert scale (0 = strongly disagree to 4 = strongly agree). The results of the CFA supported the model's goodness of fit (SBX 2 = 38.20, df = 20, p = .008; SRMR = .019, TLI = .99, CFI = .99, RMSEA = .02, 90% CI [.01, .04]).
Procedure
The Ethical Commission of the Institute Technology of Sonora approved the study. Secondary and high schools were invited to participate in the study. Schools that agreed to participate in this study were also included. Data were collected in the classroom during school hours. Parents were then asked for permission from their children to respond to the questionnaires. Only 4% of the parents rejected authorization for their children to participate. Next, the students were informed about the study's objective and that their participation would be voluntary. Finally, the participating students and parents were informed about the confidentiality of the participants.
Data analysis
The dataset contains no missing data. Means, standard deviation, symmetry, and kurtosis were calculated. We assumed that skewness and kurtosis values between -2 and 2 indicated univariate normality (George & Mallery, 2001).
Analysis of Evidence Based on Internal Structure
The goodness of fit of the one-dimensional (Model 1) and two-dimensional (Model 2) measurement models was compared (see Figure 1). The CFA uses the Diagonal Weighted Least Squares Robust (DWLS) estimation method with robust X 2 (SBX) correction using the LISREL 12 software. The goodness of fit of the one-dimensional (Model 1) and two-dimensional (Model 2) measurement models was compared (see Figure 1). Because the SBX statistic is sensitive to large sample sizes (Byrne, 2016; Powell & Schafer, 2001), we additionally used fit indices, such as the standardized root mean square (SRMR), comparative fit index (CFI), Tucker Lewis index (TLI), and root mean square error of approximation (RMSEA). The structural equation literature (SEM) suggests that the values of SRMR and RMSEA ≤ .08 and CFI and TLI are ≥ .95 indicating a good model fit (Brown, 2015; Byrne, 2016). The goodness of fit of the model was compared using differences in SBX (ΔSBX), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). When the difference in ΔSBX2 was significant, a model with greater SBX had a worse fit (Brown, 2015; Byrne, 2016). Furthermore, differences in AIC and BIC > 10 indicate distinctions in the model's fit, and a model with greater AIC and BIC has poor fit (Byrne, 2016; Vrieze, 2012).
We confirmed that the distinctions between the scale dimensions were empirically supported. Based on the guidelines proposed in the literature, we expected that the square of the correlation (R 2) between the SBDI factors would be less than the AVE of each factor (Hair et al., 2010).
Cross-Validation Analysis
Structural cross-validation examines the replicability of the measurement model in an independent sample (Byrne, 2016). Multigroup analysis was used to assess the replicability of the factor structure in an independent sample (sample 2). Configural, metric, and scalar invariances were also examined. Measurement invariance was supported when ΔSBX2 was not significant (p > .001), ΔCFI ≤ .01, and ΔRMSEA ≤ .015.
Reliability Analysis
The scale's reliability was tested using McDonald Omega (ω) and the average variance extracted (AVE). The results of ω > .70 and AVE > .50 suggest adequate reliability of the scale scores (Dunn et al., 2014; Hair et al., 2017).
Measurement Invariance Analysis by Gender and Educational Level
Using a multigroup procedure, we tested configural (constrained the number of factors and factor-loading structure to be the same across groups), metric (fixed factor loadings across groups), scalar invariance (constrained intercept across groups), and residual invariance (constrained item residuals variance across groups). The nested factor model was used to examine measurement invariance in the groups by gender and educational level. The difference in SBX (ΔSBX) was not statistically significant (p > .001), suggesting that the constraints imposed were equal between the groups (Brown, 2015; Byrne, 2016). We also used the differences in CFI and RMSEA to assess the invariance. The structural modeling literature proposed that differences in CFI (ΔCFI) less than .01 and differences in RMSEA (ΔRMSEA) less than .015 confirmed model invariance (Sass & Schmitt, 2013). When the results were contradictory, we assumed that the differences in CFI and RMSEA were due to the large sample size.
Latent Means Differences
Latent means by gender and education level were compared. The reference groups (girls and high school students) were fixed at zero, while the factor means of the other groups were estimated freely. A Z-statistic was used to examine the differences between latent means (Brown, 2015; Byrne, 2016).
Analysis of Evidence Based on the Scale Relations With External Variables
Correlations between the SBDI with external variables were calculated to examine concurrent validity. Then, the correlation between SBDI and cybervictimization was calculated. The effect size was assessed based on guidelines from the literature (see Funder & Ozer, 2019). An effect size r of .10 is small, .20 indicates a medium effect, and .30 suggests a large effect.
Results
Descriptive Analysis
Table 1 provides the descriptive statistics of the items. For ten items, the means center in the "sometimes" category, with the remaining five in the "never" category. The results of skew and kurtosis indicated a normal univariate for 11 items. However, statistics indicate departures of univariate normality for items 11, 12, 13, and 14. These results indicate that adolescents rarely intervened in defending themselves against cyberbullying victims.
Assessing One-Dimensional and Two-Dimensional Measurements Models
The goodness-of-fit statistics for the one-dimensional (Model 1) and two-dimensional (Model 2) measurement models in Sample 1 (calibration sample) were assessed. Confirmatory factor analyses showed that Model 1 did not have a reasonable adjustment to the data (SBX2 = 680.14, df = 76, p < .001, SRMR = .12, CFI = .88, TLI = .77, RMSEA = .13, 90% CI [.12, .14]), whereas Model 2 had an acceptable goodness of fit (SBX2 = 105.61, df = 78, p = .014, SRMR = .04, CFI = .98, TLI = .97, RMSEA = .04, 90% CI [.03, .05]). The fit of Model 2 showed better adjustment than Model 1 (ΔSBX2 = 384.53, df = 2, p < .001, ΔAIC = 515.73, ΔBIC = 515.62; see Table 2).
Table 2 Goodness-of-fit Statistics of the One-Dimensional and Two-Dimensional Measurement Models (N = 612)
In Model 2, the factor loadings were from .54 to .88 (p < .001); the constructive defending and aggressive defending intervention factors were positively correlated with each other (see Figure 2). The remaining analyses were based on model 2. These results indicate that bystander defender intervention is a two-dimensional model.
Figure 2 Results of the Two-Dimensional Model of Styles of Bystander Defender InterventionNote. Standardized factor loadings are also reported.
The square of the correlation between the SBDI factors in both samples was less than the AVE of each factor (Sample 1 AVE = .52, R 2 = .01; Sample 2 AVE = .53, R 2 = .02). We concluded that these scores confirm the uniqueness of each SBDI dimension.
Factorial Structure Cross-Validation
A multigroup procedure was used to assess the stability of the two-dimensional measurement model in an independent sample of adolescents. The configural model (SBX2 = 192.65, df = 152, p = .014, SRMR = .04, CFI = .97, TLI = .96, RMSEA = .04, 90% CI [.03, .07]) had an adequate fit for the data. Furthermore, the results confirmed the metric and scalar invariance of the model (see Table 3). Additionally, the reliability for each factor in both samples was acceptable: constructive defending (Sample 1, ω = .77 and AVE = .52; Sample 2, ω = .79 and AVE = .53), and aggressive defending (Sample 1, ω = .81 and AVE = .55; Sample 2, ω = .78; AVE = .51). Overall, these results confirmed that Sample 1 factor structures were replicated in Sample 2, which confirms the stability of the two-dimensional first-order structure.
Assessing Measurement Invariance by Gender
The configural model had a goodness-of-fit on both samples (Sample 1: SBX2 = 205.45, df = 152, p = .002, SRMR = .051, CFI = .98, RMSEA = .03, 90% CI [.02, .03]; Sample 2: SBX2 = 196.25, df = 152, p < .001, SRMR = .06, CFI = .98, RMSEA = .03, 90% CI [.02, .04]). When factor loadings were fixed to be equal across genders (metric invariance), the difference in the configural model was not statistically significant in either sample, and changes in the CFI and RMSEA were small (ΔCFI < .01 and ΔRMSEA < .015). When intercepts of the observed variables were forced to be equal by gender, the differences were not statistically significant, and differences in CFI and RMSEA were small (ΔCFI < .01 and ΔRMSEA < .015), which supports scalar invariance in Samples 1 and 2. Finally, the results support residual invariance (see Table 4).
Assessing Measurement Invariance by Educational Level
The fit indices indicated that the configural model fit the data by educational level in both samples (Sample 1: SBX2 = 205.67, df = 152, p = .002; SRMR = .04, CFI = .97, TLI = .96, RMSEA = .04, 90% CI [.03, .06]; Sample 2: SBX2 = 194.38, df = 152, p = .011; SRMR = .05, CFI = .96, TLI = .95, RMSEA = .05, 90% CI [.03, .07]). The factor loadings were constrained to be equal between secondary and high school students (metric invariance), the comparison with the configural model was not statistically significant in either sample, and the changes in CFI and RMSEA were not significant (ΔCFI < .01, ΔRMSEA < .015). We then added constraints on the intercepts of the model (scalar invariance). These results show that the difference in SBX2 between models was not significant, and the changes in CFI and RMSEA were smaller than those suggested in the literature (ΔCFI = .003, ΔRMSEA = .002). Also, the results confirmed residual invariance by educational level (secondary school vs. high school) (see Table 4).
Latent Means Differences
Females were chosen as a reference group to compare factors by gender. Then, the male means informed the difference in constructs across the groups. The results show a statistically significant difference in gender in the model dimensions. Notably, girls had higher levels of constructive defense and lower levels of aggressive defense than did boys.
Concerning possible differences by educational level, the high school group was chosen as the reference and the secondary students' group was informed of the difference in factor means. The test results revealed a statistically significant difference that was unique to the aggressive defense dimension. High school students had higher levels of aggressive intervention than secondary school students when they observed cyberbullying incidents (see Table 5).
Relations With External Variables
As expected, constructive defending had a negative relationship with cyberbullying and a positive relationship with moral identity (see Table 6). Furthermore, as expected, aggressive defending interventions were positively correlated with cyberbullying and negatively correlated with moral identity. The effect size of the correlation ranged between low and medium, suggesting explicative and practical consequences (Funder & Ozer, 2019). Overall, these correlations confirmed the concurrent validity of the SBDI.
Discussion
Prior studies have investigated bystander-defender intervention in cyberbullying events; most conceptualized bystander-defender intervention in cyberbullying as prosocial. However, this study confirms that defender intervention should be aggressive or constructive. Thus, unlike other scales, the Styles of the Bystander Defender Intervention (SBDI) considers two possible stances for defenders: constructive and aggressive. Given its relevance to advancing the current understanding of multidimensional defenders' interventions in cyberbullying, this study sought to examine the psychometric properties of the SBDI in a Mexican sample. Our findings confirm what other studies report, bystander defender intervention is a multidimensional construct that comprises aggressive and constructive intervention. Furthermore, the results indicate that the scale may be suitable (valid and reliable) for use in Mexican populations, facilitating research on variables associated with constructive interventions in cyberbullying.
Bystander Defender Intervention as a Multidimensional Construct
Findings support the multidimensional conceptualization of the bystander defender intervention proposed by the SBDI. Discriminant validity was confirmed, suggesting that constructive and aggressive factors measure a unique construct. Our results align with those of previous research (Bussey et al., 2020; Luo & Bussey, 2019; Moxey & Bussey, 2020) that distinguishes constructive and aggressive interventions for cyberbullying bystanders. Further studies are needed to explore the roots and effects of different bystander intervention styles on cyberbullying. Similarly, it is necessary to explain the effects of bystander interventions (constructive and aggressive) on the prevalence and prevention of cyberbullying.
Measurement Invariance by Gender and Educational Level
The results provide empirical evidence supporting the measurement invariance of the SBDI scale by gender and education level (secondary and high school). In other words, the SBDI measures the same metrics between the groups. Once measurement invariance was confirmed, mean latent differences were examined in these groups. Consistent with other studies regarding gender differences (Bussey et al., 2020; Moxey & Bussey, 2020), these results confirm the statistical differences in both model dimensions. Girls have higher levels of constructive and lower levels of aggressive intervention in cyberbullying.
Statistically significant differences in aggressive interventions were also found among the education level groups. High school students had higher levels of aggressive intervention in cyberbullying incidents than secondary school students. Further research is needed to test the roots of these differences by analyzing contextual, family, and personal factors as input variables leading to constructive or aggressive defender interventions.
Relations With External Variables
The expected correlation between the SBDI and cyberbullying and moral identity was found, confirming concurrent validity. These results are in line with those of previous studies (see Moxey & Bussey, 2020). Although further studies are needed, we suggest that the violence escalation cycle framework (Anderson & Carnagey, 2004) helps to explain the effect of aggressive interventions in maintaining cyberbullying.
On the other hand, the results are consistent with past research (see Bussey et al., 2020; Valdés-Cuervo et al., 2021), showing that aggressive intervention by bystanders is negatively associated with moral identity, whereas constructive intervention is positively related. Although additional research is needed to understand these relationships, we posit that moral identity is a potential buffer for adolescent aggression (see Colasante et al., 2015).
Theoretical and Practical Implications
This research improves our knowledge of the role of bystanders in cyberbullying. Identifying specific roles and associated factors has contributed to the development of effective targeted bystander cyberbullying interventions. The intervention of defender bystanders is crucial for education, evaluation, and research purposes. The distinction between aggressive and constructive bystander defender interventions facilitates the study of psychosocial factors associated with these behaviors and the development of targeted and helpful anti-cyberbullying programs. Furthermore, an intervention that promotes bystander defender intervention could teach students how to intervene constructively in cyberbullying events. Although further studies are necessary, the findings suggest that strategies to encourage cyber bystanders' defensive interventions must be accompanied by interventions that encourage students' moral development.
The measurement invariance of the SBDI facilitates a meaningful comparison of defender interventions by gender and educational level (secondary vs. high school). Then, practitioners can identify the differences in factors associated with types of defender intervention in both gender and educational level (secondary and high school). In addition, they can evaluate whether programs to promote bystander construction and hinder aggressive defending interventions have different results according to gender and education level.
Limitations
This study has several limitations. First, the results were based on self-reported measures, which may have inherent response biases. Therefore, further studies should include different informants (e.g., peers and teachers) and measurement strategies (e.g., interviews) to provide a more robust scale to measure the styles of bystander defender interventions in cyberbullying events. Second, the findings are based on a non-probabilistic sample from a specific region of Mexico. It is desirable to use random and cross-national samples from diverse cultures (e.g., indigenous students) to examine the psychometric properties of the SBDI.
Third, given the cross-sectional design of this study, it remains challenging to assess whether aggressive or constructive intervention exists. Therefore, further studies should examine the prevalence of the bystander defender type (constructive or aggressive) in cyberbullying events over time. Finally, we need to examine the validity equivalence of the English and Spanish versions of the scale. Further research should examine the validity of the interpretation of this two-scale version.
Future Research
Future studies must further examine the precursors and consequences for victims of both styles of bystander-defender intervention in cyberbullying. Additionally, differences in gender and education levels in defender intervention styles should be studied. Finally, additional studies exploring the consequences of aggressive and constructive interventions in prevention programs are required.
