Cyberbullying is often defined as one or more acts of intentional aggression carried out online by one or more individuals who repeatedly target other individuals who cannot easily defend themselves (Kowalski et al., 2014). In Spain, 10.3% of the students from 5th grade of primary education to 4th grade of secondary education report having experienced some kind of cyberbullying situation frequently (Díaz-Aguado et al., 2024). A recent meta-analysis conducted on 42 studies, which encompassed data from 266,888 participants from different countries, found a similar average prevalence of cyberbullying victimization (11.1%) (Li et al., 2024). A large number of studies indicate that the prevalence of cyberbullying begins to increase during the last years of primary education (ages 10-12), reaching its peak in secondary education (Álvarez-García et al., 2018; Arnarsson et al., 2020; Yudes et al., 2020). However, some recent studies have found higher prevalences at younger ages (Aizenkot & Kashy-Rosenbaum, 2021). A possible explanation is the greater access to technology by primary school students. Most studies have explored the phenomenon of cyberbullying primarily among secondary school students, with fewer studies addressing the transition between primary and secondary education (Vismara et al., 2022).

In addition, levels of unsupervised Internet access and social-media use among 5th- and 1st-grade students have risen sharply in recent years (e.g., 47.7% access without supervision and 50.8% social-network accounts in 5th grade; 58.1% and 70% in 6th grade; Díaz-Aguado et al., 2024). This underscores the need to re-examine cybervictimization and social support in today’s technologically saturated context, rather than relying on earlier datasets when digital penetration among young students was substantially lower. There is a causal relationship between suffering cyberbullying and the subsequent development of mental health problems, suicide attempts, suicidal ideation, and poor overall health (Moore et al., 2017). Specifically, cyberbullying increases suicidal thoughts by 14.5% and suicide attempts by 8.7% (Nikolaou, 2017). Cybervictimization has also been associated with physical self-harm (Azami & Taremian, 2020; Kwan et al., 2020) and digital self-harm (Patchin & Hinduja, 2017). Furthermore, it is associated with a higher likelihood of developing generalized anxiety (Álvarez-García et al., 2025; Gong et al., 2022), depression (Álvarez-Marín et al., 2022; Balluerka et al., 2023), social stress (Garthe et al., 2023), and posttraumatic stress (Liu et al., 2020). Additionally, it implies a greater likelihood of developing Internet addiction (Aktaş Terzioğlu & Büber, 2024), sleep problems (Mang et al., 2023) and eating disorders (Cheng et al., 2023).

Social support is defined as the perception or experience of being cared for and valued and of being part of a social group that offers and enjoys interdependent relationships of affection and socio-emotional support (Taylor, 2011). There is consistent evidence that social support plays a key role in the prevention of a large number of physical and psychological diseases at any age (Chang et al., 2023; Galindo-Domínguez & Losada, 2023; Jameel et al., 2024; McLean et al., 2022; Wang et al., 2024). Social support enhances immune function, and is associated with greater longevity and a better ability to cope with adverse life events (Vila, 2021).

Despite its profound impact, the transition from childhood to adolescence has received comparatively little attention relative to later adolescence (ages 15–19) (Blum et al., 2014). This period is marked by significant physical, psychological, and social changes: the child with more “childish” traits develops a more adult appearance, enhanced cognitive capacities, and social relationships in which peers assume greater importance (Papalia & Martorell, 2024). Until this point, the family typically serves as the primary source of support, but with the onset of adolescence, peers and close friends gradually exert greater influence on emotional well-being and the development of social skills. This dynamic can be understood through Bronfenbrenner’s bioecological model (Bronfenbrenner, 1979; Bronfenbrenner & Morris, 2006), which situates the individual within interrelated systems—microsystem (family, peers, school), mesosystem (interactions among these), exosystem (broader social contexts) and macrosystem (cultural values)—where perceived social support is a resource continually shaped by the quality of relationships across these levels. At these ages, social support primarily stems from four key sources: family, teachers, classmates, and friends (Harter, 2012). Adopting such a global, ecological perspective helps explain why certain sources of support gain prominence at distinct stages of the school trajectory and how their combined reinforcement can serve to prevent cybervictimization.

First, the family, parents, or other close members are the primary sources of support and care. They provide emotional security, material sustenance, and moral guidance (Michaelson et al., 2021; Tan & Yasin, 2020). Older children and adolescents progressively seek greater autonomy from their family. However, family support remains fundamental for their proper development (De Meulenaere et al., 2022). A strong attachment to the family is a protective factor against many problems both in childhood and adulthood (Khairunnisa et al., 2024; Rawatlal et al., 2015) and, additionally, a significant buffering factor against the negative effects of cyberbullying (Hellfeldt et al., 2020). The absence of adequate family support is associated with difficulties in socio-emotional and academic development (Chan et al., 2022).

Children and young people spend a significant part of their lives at school. Teachers not only facilitate the acquisition of knowledge, skills, and academic abilities, but they also act as role models and sources of emotional support (García-Rodríguez et al., 2023). They can help students develop a healthy and resilient personality that aids them in facing personal, social, and academic challenges at all educational stages (Pakarinen et al., 2020; Roorda et al., 2021; Wentzel, 2022). The role of teachers can be even more critical in cases where students lack adequate family support (Burdick & Corr, 2024). The relationship of children and adolescents with parents and teachers is asymmetrical. In contrast, peer relationships allow for the natural learning of social skills for conflict resolution, cooperation, respect, setting boundaries, and developing empathy (Eccles, 1999; Ogden & Hagen, 2018). Moreover, it fosters the construction of a social identity that is gradually differentiated from that of their family (Reitz et al., 2014). Feeling supported or rejected by peers has important implications for the healthy development of children and adolescents (Wentzel, 2022).

Finally, relationships with close friends are essential for proper socio-emotional development (Bukowski & Sippola, 2005). They provide a safe environment where young people can freely express their most personal emotions and thoughts and receive close support in a context of symmetry (Krammer et al., 2023). Not having close friends at any stage of life is associated with poorer physical and emotional health (Allen et al., 2022; Howick et al., 2019).

Social support is a key factor in the prevention of traditional school bullying. Many studies have found that students with high levels of social support are victimized less (Holt & Espelage, 2007; Murphy et al., 2017; Rothon et al., 2011; Shaheen et al., 2019). Furthermore, social support appears to have a buffering effect on the psychological consequences of bullying (Lee et al., 2022; Rothon et al., 2011). However, few studies have examined the relationship of social support with cyberbullying. Additionally, some of the valuable efforts made regarding this have not considered the evaluation of relevant sources of support at these ages, such as peers (Arató et al., 2022; Hellfeldt et al., 2020), focusing solely on secondary school students (Jaskulska et al., 2022) or presenting limitations that may be significant.

Although prior research has documented the decline in perceived family support and the rise of perceived peer support during adolescence, very few studies have focused specially on the transition from primary to secondary school or have simultaneously examined multiple sources of support (family, teachers, classmates, and close friends) in relation to cybervictimization. Our study fills this gap by assessing perceived social support across a wide age span (5th grade of primary to 4th grade of secondary education), comparing in a single model the protective effects of these four distinct sources, exploring both direct and indirect pathways as well as interactions with school grade, and focusing explicity on cybervictimization during this critical transition. Moreover, to ensure methodological rigor, we employ standardized, validated instruments for both social support and cybervictimization.

Participants were 654 students from 5th grade of primary education to 4th grade of secondary education, aged 9 to 17 (Mage=12.80, SD=1.64). According to the International Standard Classification of Education (ISCED) (UNESCO Institute for Statistics, 2012), the Spanish 5th and 6th grades of primary education correspond to ISCED Level 1 (primary education, typically covering ages 6 to 12), while 1st to 4th grades of secondary education correspond to ISCED Level 2 (lower secondary education, typically covering ages 12 to 16). Participants were recruited via convenience sampling from six schools located in different provinces of Galicia (Spain). Schools were selected to ensure representation of both rural and urban settings. Concerning gender, 48.8% identified as girls, 50% as boys, and 1.2% as non-binary. Of the students, 60.7% were enrolled in public schools (government-funded and managed) and 39.3% in government-dependent private schools (schools under the administration of private entities with public/government funding). Additionally, 43.2% attended rural schools, and 56.8% attended urban schools. Socio-demographic information (gender, age, and school attended) was obtained through a multiple-choice item for gender and two open-ended items in which students specified their age and the school they attended.

An adapted version of the Social Support Scale for Children (SSSC; Harter, 1985) was used, based on a previous adaptation by Pastor et al. (2012) and taking into account the observations of Lipski et al. (2014) and Malecki and Demaray (2002). This 24-item scale evaluates children’s and adolescents' perceptions of social support through four subscales corresponding to four sources of support: parents, classmates, teachers, and close friends. Each of the four SSSC subscales consists of 6 items, with a four-point response format ranging from 1=very false to 4=very true. Thus, scores could range between 6 and 24, with higher values indicating the reporting student’s greater perception of social support in the specific target area.

The validation of this adapted version (Rodríguez-Enríquez et al., in press), used in the present study, showed that the factorial structure that best fit the data from the tests conducted is composed of four subscales or types of perceived social support, which are interrelated. The items of the Parent(s)’ Social Support (PSS) subscale specifically measure the degree to which respondents perceive that their parents understand them, show interest in them, and value their importance (e.g., “I feel that my parents really understand me”). The classmates’ social support (CSS) subscale measures the extent to which they feel accepted, valued, and heard by their peers (e.g., “I have classmates who pay attention to what I say”). The teacher(s)’ social support (TSS) subscale measures the degree to which teachers show interest, treat students equitably, and offer help for their development (e.g., “If I feel unwell, there is a teacher who cares about me”). Lastly, the close friends’ social support (FSS) subscale measures the extent to which they perceive they have a close friend they can trust and share their experiences and concerns with (e.g., “I have a close friend who truly understands me”). Although the model comprising these four interrelated types of perceived social support demonstrated the best fit to the data, a hierarchical model-consisting of these four factors as first-order dimensions and a general second-order factor-also provided an adequate fit (Rodríguez-Enríquez et al., in press). In the present work, the general score of total social support (SS) based on the scores of the 24 items will also be used, due to its summarizing nature. This general score provides information about the extent to which the child or adolescent feels supported by all sources of support. In this case, the minimum score is 24 points, and the maximum is 96. The reliability of the scores obtained on the four subscales in the present study, in terms of internal consistency, was high or very high (PSS: α=.820, ω=.850, composite reliability=.845; CSS: α=.804, ω=.810, Composite reliability=.858; TSS: α=.855, ω=.858, composite reliability=.924; FSS: α=.922, ω=.924, composite reliability=.808). The internal consistency of the scores on the overall scale was moderate (α=.877, ω=.886, composite reliability=.674). Convergent validity, analyzed in terms of average variance extracted, is adequate (AVETSS=.671, AVECSS=.504) or slightly below the threshold typically considered for this purpose (AVEPSS=.479, AVEFSS=.425) in its subscales. In contrast, the AVE is lower in the second-order factor (AVESS=.345), which is nevertheless retained for analyses in this study due to its sufficient internal consistency and informative value.

An adapted version (Rodríguez-Enríquez et al., 2025) of the Cyberbullying Test by Garaigordobil (2013) was applied.. Specifically, this work will analyze the results of the cybervictimization scale from the cyberbullying section. This scale consists of 15 items that measure the level of cyberbullying experienced by the informant in the last three months (e.g., Has anyone pretended to be you, making defamatory comments, lies, or sharing your secrets on social media or the Internet?). The items are rated on a 4-point response format ranging from 0=never to 3=always. The score ranges from 0 to 45, with higher scores indicating elevated levels of cybervictimization perceived by the informant. The internal consistency of the scores obtained with the complete scale was high (α=.880, ω=.886, composite reliability=.876). However, its convergent validity, assessed through the Average Variance Extracted (AVE), was modest (.330).

First, a meeting was arranged with the school managers to inform them about the objectives and characteristics of the study, as well as the voluntary nature, anonymity, and confidentiality of the data to be collected. After obtaining authorization from the management teams, the same information was sent to the families in writing, offering them contact information of the research team and encouraging them to ask any questions they deemed necessary to decide whether to authorize the student's participation. Prior to each application in the classroom, about 5-10minutes were dedicated to providing the students with information about the characteristics of the study (anonymity, confidentiality, and voluntariness), emphasizing the importance of answering sincerely. They were also encouraged to raise any concerns or questions they might have, allowing enough time to clarify them. The average time to complete the tests was 30minutes, although flexibility was prioritized, especially considering the specific age of the group, the questions raised, or the presence of students with cognitive difficulties.

In the 5th- and 6th-grade classrooms (ages 9-12), the characteristics of cyberbullying were briefly enumerated and explained to help them distinguish what is cyberbullying and what is not. This decision was made based on some comprehension difficulties encountered during the pilot application of the test on younger students. This enhanced the clarification of doubts and increased the chances of obtaining reliable results. The researchers administered the tests to all groups in all the centers during school hours.

All phases of the study followed the International Ethical Code in Humanities and Social Sciences of the Center for Research Ethics & Bioethics, and the protocol of the larger research project of which it is part was approved by the Institutional Ethics and Bioethics Committee (Ref. CE-DCEC-UVIGO-2020-12-02-8129).

First, the reliability of the questionnaires used was analyzed by examining the internal consistency of the scores obtained for each factor. For this purpose, Cronbach’s alpha, McDonald’s omega, and composite reliability were calculated for each factor. To obtain the composite reliability, the factor loadings were used. These loadings were derived from a confirmatory factor analysis conducted for each of the questionnaires. Given the ordinal nature of the data, the WLSMV estimation method was employed. Internal consistency indices above .70 are generally considered adequate, above .80 high, and above .90 very high. The analysis of the internal consistency of the scores for each factor was complemented by the analysis of the average variance extracted (AVE) for each factor, as a measure of the factor’s convergent validity. An AVE value of .50 is commonly used as the cutoff point to consider the average variance extracted adequate.

Next, descriptive and correlational analyses were conducted to determine the distribution and simple correlations among the scores of the target variables. Skewness and kurtosis indices were analyzed to assess the normality of the distribution of scores for each item. Subsequently, to examine the degree of perceived social support and possible grade-related differences (the first objective of the study), one-way ANOVAs were performed for each variable of perceived social support. Welch's statistic was used for variables with different variances among the groups, and Fisher's statistic for variables with equal variances. Post-hoc analyses were conducted using the Games-Howell test in the case of different variances or Tukey's test in the case of equal variances.

Afterwards, to analyze the effect of the four sources of social support on cybervictimization and the possible moderating effect of the grade (the second objective of the study), a three-step multiple regression analysis was conducted. In the first step, the four types of social support were introduced; in the second step, the course was added to the four types of social support; and in the third step, interaction effects of the course on each of the four types of social support were included. Given that some sources of social support showed a statistically significant simple correlation with cybervictimization but no statistically significant effect in the regression analyses, mediation analyses were subsequently performed to check whether the effect might be indirect rather than direct, mediated through other forms of social support that were analyzed with a clearer direct effect. Finally, moderation analyses were conducted to study in greater detail the interaction effects found in the multiple regression analyses. All analyses were conducted using the statistical software Jamovi 2.3.28.

Table 1 shows a generally moderate to high level of perceived social support and a low level of cybervictimization, a positive correlation between the different forms of perceived social support, a negative correlation of all the analyzed forms of social support with reporting being a victim of cybervictimization, and a negative correlation of the grade with some forms of social support.

Regarding perceived social support, the average and total levels found in each form of social support tended to be high, above the scale’s mean value. Students perceived greater social support from close friends and family than from teachers and classmates. Despite the scores’ tendency to be distributed above the mean, their distribution did not deviate severely from normality. The correlations between the four forms of social support were positive and statistically significant. The magnitude of the correlations between perceived social support from close friends and classmates, from family and teachers, and from family and classmates was moderate. In the other cases, the correlations were low.

Regarding cybervictimization, the distribution of scores deviated significantly from normality, with a significant tendency towards low scores. The correlations of all forms of perceived social support with reporting being a victim of peer cyberaggression were negative and statistically significant. The magnitude of the correlation tended to be moderate or low.

Finally, a negative and statistically significant correlation was observed between some forms of perceived social support and grade. Specifically, student-perceived social support from adults (parents and teachers) tended to decrease throughout their school years. In contrast, no significant relationship was observed between grade and social support from peers (close friends and classmates) or with cybervictimization.

As both the general measure of perceived social support and several specific forms of support were significantly related to the grade (Table 1), we compared perceived social support in the different grades (from 5th grade of primary education to 4th grade of secondary education). Table 2 shows statistically significant differences as a function of grade in all kinds of perceived social support, as well as in the total measure of social support. In general, primary students tended to perceive greater social support than secondary students. This is more evident in the social support from adults (parents and teachers) than from peers (classmates and close friends), where the level of support was more stable, indicating fewer differences as a function of grade. As shown in Table 2, regarding the score of total social support, students in 5th and 6th grade of primary school perceived greater social support than those in 1st, 2nd, 3rd, and 4th grade of secondary school. More specifically, concerning parent(s)’ social support, 5th- and 6th-graders of primary school and 1st-graders of secondary school perceived more parental support than 2nd-, 3rd-, and 4th-graders of secondary school. Students from 5th grade of primary school perceived more parental support than 1st- graders of secondary school. Regarding teacher(s)’ social support, 6th-graders of primary school perceived more teacher support than those in 1st, 2nd, 3rd, and 4th grade of secondary school. Students from 5th grade of primary school perceived more teacher support than 3rd-graders of secondary school. Regarding classmates’ social support and close friends’ social support, in both cases, 5th-graders of primary school perceived greater social support than 2nd-graders of secondary school. In both grades (5th grade of primary school and 2nd grade of secondary school), students reported the highest and lowest levels, respectively, of social support from peers and close friends.

Given that the four forms of perceived social support analyzed had negative and statistically significant correlations with cybervictimization, but at the same time, they correlated with each other (Table 1), a multiple regression analysis was conducted to more precisely determine the specific effect of each form of perceived social support on cybervictimization.

The analysis was conducted in three steps (Table 3). In the first step, a model was tested in which the four forms of perceived social support were included as independent variables. Two of the variables presented negative and statistically significant regression coefficients (social support from family and peers), whereas the other two forms of support (teachers and close friends) no longer presented statistically significant coefficients once the effect of the other types of support was statistically controlled. The resulting model was statistically significant and explained 16% of the variance in cybervictimization scores. In the second step, a model was tested that added the students’ grade to the variables already included in Step 1. The grade did not show a statistically significant effect, nor was there a significant variation in the effect of the other variables included in the model. The increase in the coefficient of determination (R²) was not statistically significant. Finally, in the third step, a model was tested in which the interactions of the grade with each of the forms of perceived social support analyzed were added to the independent variables from Step 2 to explore the possible moderating effect of the grade on the relationship between social support and cybervictimization. Two of the interaction effects were statistically significant: the grade moderated the effect of both parental and peer social support on cybervictimization. The statistically significant effect of social support both from parents and peers was maintained. The introduction of the interactions between the grade and the different forms of social support in the model resulted in a small but statistically significant increase in the model's explanatory power (R²), increasing to 17% of the variance in cybervictimization scores.

To explore why social support from teachers and close friends did not show a statistically significant relationship when statistically controlling for the other variables and considering that the four types of social support correlated with each other (Table 1), mediation analyses were conducted. As shown in Table 4, we observed that the effect of both social support from teachers and close friends on cybervictimization was not direct but indirect, mediated in both cases by the social support of parents and classmates.

Finally, moderation analyses were conducted to study in more detail the interaction effect of the grade on the social support of parents and classmates, as observed in Table 3. As shown in Table 5, the effect of both types of support on cybervictimization was negative and statistically significant in both lower and medium or higher grades. However, the effect of parental social support was greater in the earlier grades than in the later ones, while the effect of peer support was greater in the later grades than in the earlier ones (Table 5; Figures 1 and 2).

Figure 1.

Graphic representation of the effect of parent(s)’ social support (PSS) on cybervictimization (CBV) at different levels of the moderator (grade).

Figure 2.

Graphic representation of the effect of classmates’ social support (CSS) on cyber-victimization (CBV) at different levels of the moderator (grade).

The results emphasize that, although it decreases during adolescence, perceived family support remains crucial. Thus, it is essential for preventive strategies against cyberbullying to involve families —especially during school transition periods— strengthening their role as agents of emotional and practical support (Holt & Espelage, 2007). Equally important is the cultivation of positive peer networks: programs should empower classmates as defenders and foster genuine, offline social bonds to counteract the isolating effects of excessive online interaction. In the school context, interventions should focus on strengthening peer bonds, given their key role during adolescence. Some educational models dedicate the first weeks of secondary education to strengthening positive relationships among students, promoting a positive classroom climate, and mitigating risk factors associated with victimization, successfully reducing its prevalence (Virtanen et al., 2019). Therefore, training teachers to act as facilitators of these relationships and promoters of anti-aggression norms is fundamental. Outside of school, it is important to promote and nurture friendships, as they complement family and school support and reinforce social skills and self-esteem (Hong et al., 2016). Finally, preventive measures must be adapted to the educational level. In childhood, they should focus on strengthening family support, while in adolescence, healthy peer relationships should be prioritized, without neglecting the ongoing role of family and teachers. This approach will allow for the design of strategies that address both risk and protective factors, optimizing the prevention of cyberbullying and improving school well-being.

Although this study enriches the understanding of cybervictimization and social support, several limitations must also be considered. Firstly, the explanatory power of the social support variables analyzed in relation to cybervictimization tends to be low. This may be because these variables represent only a small part of the set of factors, and their interactions, that explain the likelihood of being a cybervictim. Future studies should expand the explanatory model by incorporating other relevant factors to improve its predictive capacity. Secondly, the use of a cross-sectional methodology is not the most suitable for analyzing the evolution of the phenomenon over time. Although this work is exploratory, future research should employ longitudinal designs to examine how the relationships between various sources of social support and cybervictimization change at different educational stages. Thirdly, the student groups by grade do not have equal sizes. This lack of homogeneity in group sizes may affect the robustness of the statistical analyses. However, precautions were taken to minimize this potential bias, such as checking for homogeneity of variances and using robust tests when necessary. In future replications of this study, it would be advisable to ensure that group sizes are similar. Forthly, the measures used were based on self-reports to assess both social support and cybervictimization. Although this method is particularly suitable for detecting phenomena like bullying that often occur privately, it is subject to possible memory biases or social desirability. Future research could combine self-reports with other complementary methodologies to enrich the understanding of the phenomenon. Finally, although some subscales presented AVE values marginally below the recommended cut-off, this limitation does not compromise the validity of the results or the use of the questionnaires, given their strong internal consistency and reliability. Nonetheless, this aspect should be considered in future studies.