The Prevention of Bullying at Schools

From the 40 years following Olweus’ (Reference Olweus1973) pioneering study, with all the evidence and knowledge generated in relation to bullying and its prevention, we can affirm at least two things. The first has been widely confirmed: Bullying is associated with multiple negative consequences for the physical and psychological well-being of the children involved. A large number of studies, both cross-sectional and longitudinal (e.g., Arseneault, Reference Arseneault2018; Baier et al., Reference Baier, Hong, Kliem and Bergmann2019; Bannink et al., Reference Bannink, Broeren, van de Looij-Jansen, de Waart and Raat2014; Hase et al., Reference Hase, Goldberg, Smith, Stuck and Campain2015; Young-Jones et al., Reference Young-Jones, Fursa, Byrket and Sly2015), as well as a multitude of meta-analyses (e.g., Holt et al., Reference Holt, Vivolo-Kantor, Polanin, Holland, DeGue, Matjasko, Wolfe and Reid2015; Koyanagi et al., Reference Koyanagi, Oh, Carvalho, Smith, Haro, Vancampfort, Stubbs and DeVylder2019; Kowalski & Limber, Reference Kowalski and Limber2013; Kowalski et al., Reference Kowalski, Giumetti, Schroeder and Lattanner2014; Moore et al., Reference Moore, Norman, Suetani, Thomas, Sly and Scott2017; Schoeler et al., Reference Schoeler, Duncan, Cecil, Ploubidis and Pingault2018; Tsaousis, Reference Tsaousis2016; van Geel et al., Reference van Geel, Goemans, Zwaanswijk, Gini and Vedder2018), show that children and adolescents who experience bullying tend to become depressed, distressed and lonely. They miss more school and show a decline in their academic performance. Bullying undermines their self-esteem as well as their physical health. In severe cases, bullying can lead to self-harming and suicidal behavior. Moreover, according to recent meta-analyses, it seems that these negative consequences are even worse in the context of cyberbullying (Baier et al., Reference Baier, Hong, Kliem and Bergmann2019; Katsaras et al., Reference Katsaras, Vouloumanou, Kourlaba, Kyritsi, Evagelou and Bakoula2018; Schoeler et al., Reference Schoeler, Duncan, Cecil, Ploubidis and Pingault2018). Some authors attribute this result to the particular characteristics of cyberbullying compared to traditional or face-to-face bullying, which place the victimized student in a situation of greater helplessness (Smith, Reference Smith and Smith2019). Among these characteristics are the fact of not being able to escape the bullying situation (it can occur 24 hours a day, 7 days a week), not knowing who is carrying out the aggression (cyberbullying can be anonymous), and the ease with which it is spread, reaching an unlimited number of viewers with a single click, without the victimized student being able to know who or how many people have been reached.

The second aspect of bullying that boasts robust empirical evidence is that its initiation, maintenance, and consequences depend not only on the personal characteristics of the students involved, but also on the context of their behaviors (Garandeau & Salmivalli, Reference Garandeau and Salmivalli2019). In this regard, Bronfenbrenner’s ecological perspective (Bradshaw, Reference Bradshaw2015; Bronfenbrenner, Reference Bronfenbrenner1979; Hymel & Swearer, Reference Hymel and Swearer2015) has been the most widely used both to study and understand the phenomenon, as well as to prevent it. Following this line, the preventive and educational point of view seeks to understand which contexts are less likely to present bullying. Again, the evidence is conclusive. Classrooms show lower levels of bullying and victimizationFootnote ¹ , as well as more peer defending behaviors (i.e., any action aimed at stopping the bullying, helping and/or comforting the victimized student, or requesting adult intervention), under the following conditions: There is a good relationship between teachers and students, there are clear rules against bullying, both teachers and students express attitudes against bullying, and families, for their part, convey messages to their children against the use of violence as a way of solving conflicts (Hymel & Swearer, Reference Hymel and Swearer2015; Jungert et al., Reference Jungert, Piroddi and Thornberg2016; Lucas-Molina et al., Reference Lucas‐Molina, Williamson, Pulido and Pérez‐Albéniz2015; Lucas-Molina et al., Reference Lucas-Molina, Giménez-Dasí, Fonseca-Pedrero and Pérez-Albéniz2018; Peets et al., Reference Peets, Pöyhönen, Juvonen and Salmivalli2015; Saarento et al., Reference Saarento, Kärnä, Hodges and Salmivalli2013; Sentse et al., Reference Sentse, Veenstra, Kiuru and Salmivalli2015; Thornberg et al., Reference Thornberg, Wänström and Pozzoli2017). It is worth noting that these contexts not only protect against bullying, but also against its negative effects (Morin et al., Reference Morin, Bradshaw and Kush2018). For instance, several studies have shown that in schools where students have a greater sense of belonging, there was no significant relationship between bullying (or cyberbullying) and various mental health indicators, such as suicidal behavior (Lucas-Molina et al., Reference Lucas-Molina, Pérez-Albéniz, Solbes-Canales, Ortuño-Sierra and Fonseca-Pedrero2022).

Clearly, if bullying depends on the context in which it occurs, the design of successful prevention programs must have a socio-ecological perspective that seeks to improve the social context; it must go beyond mere stop-gap solutions such as informative talks or sessions with at-risk students only (Salmivalli et al., Reference Salmivalli, Laninga‐Wijnen, Malamut and Garandeau2021; Smith, Reference Smith and Smith2019). A school wide lifestyle change is needed in a way that promotes positive relational climates.

After more than four decades of research, we also have more than enough evidence in this regard. Recent meta-analyses conducted by Hannah Gaffney and her team (e.g., Gaffney, Farrington, et al., Reference Gaffney, Farrington and Ttofi2019; Gaffney, Ttofi, et al., Reference Gaffney, Ttofi and Farrington2019; Gaffney et al., Reference Gaffney, Ttofi and Farrington2021) show that prevention programs can reduce bullying and victimization levels in schools by up to 20%, and that there are several components that appear to be more effective than others. These components include adopting a whole-school approach to prevention that involves the entire school community, providing teacher training, implementing classroom activities based on cooperative learning, encouraging peer support, holding meetings with families, providing increased supervision in high-risk areas of the school (e.g., school yard and corridors), and establishing clear disciplinary measures against bullying.

Nonetheless, despite the numerous meta-analyses from the past five years (see the Gaffney sources cited above), there seems to be no clear relationship between implementing a greater number of components and greater effectiveness (Salmivalli et al., Reference Salmivalli, Laninga‐Wijnen, Malamut and Garandeau2021). Nor has it been possible to determine which components are indispensable, despite the large number of studies included in the meta-analyses (Salmivalli et al., Reference Salmivalli, Laninga‐Wijnen, Malamut and Garandeau2021). In particular, the latest meta-analysis by Gaffney et al. (Reference Gaffney, Ttofi and Farrington2021) evaluated more than 100 anti-bullying interventions. Moreover, results from the different meta-analyses show some discrepancies. For example, Ttofi and Farrington (Reference Ttofi and Farrington2011), in one of the first such meta-analyses, found that the ’work with peers’ component (defined as formal peer involvement in anti-bullying, including mentoring and peer mediation, as well as promoting bystander intervention in bullying prevention) was associated with higher levels of victimization and bullying. Nevertheless, subsequent meta-analyses (e.g., Gaffney et al., Reference Gaffney, Ttofi and Farrington2021; Lee et al., Reference Lee, Kim and Kim2015; Polanin et al., Reference Polanin, Espelage and Pigott2012) have repeatedly shown the opposite. It should be noted that when integrating and interpreting these findings and their contradictions, it is important to consider the limitations of meta-analyses in general (e.g., publication, location and inclusion biases), and of bullying prevention analyses in particular (Menesini & Salmivalli, Reference Menesini and Salmivalli2017; Salmivalli et al., Reference Salmivalli, Laninga‐Wijnen, Malamut and Garandeau2021). For example, most of the studies in these meta-analyses are correlational, so it is inappropriate to make causal inferences. In addition, when identifying the components, researchers tend to rely on program manuals rather than on what was actually implemented. Certain components that appear explicitly in the manual may not have been implemented, or perhaps they were implemented differently from what was recommended, or the implementation may even have included another component not indicated in the program (Johander et al., Reference Johander, Turunen, Garandeau and Salmivalli2021). Similarly, certain components may be effective when combined with others and not when applied independently. Finally, the effectiveness of different components is likely to depend on the age of the students to whom they are applied (e.g., elementary vs. middle school students) (Yeager et al., Reference Yeager, Fong, Lee and Espelage2015). Indeed, some recent meta-analyses have tried to ascertain not only which components are effective, but specifically for whom (Hensums et al., Reference Hensums, de Mooij, Kuijper, Fekkes and Overbeek2022). Even so, it is indisputable that the promotion of bystander intervention is at the core of programs that currently have the greatest effectiveness (e.g., the KiVa program) (Sainio et al., Reference Sainio, Herkama, Turunen, Rönkkö, Kontio, Poskiparta and Salmivalli2020; Salmivalli et al., Reference Salmivalli, Laninga‐Wijnen, Malamut and Garandeau2021).

The Healthy Context Paradox

Thus far we have highlighted the characteristics of effective bullying prevention programs and educational contexts that are ’healthy’, or less prone to bullying, according to empirical evidence. Given the profusion of research, one might have the impression that virtually everything is known about bullying prevention. As noted, a multitude of meta-analyses indicate the effect sizes of the different programs, as well as the components that in principle are most effective, despite the need for further research (considering the disadvantages of meta-analyses noted above). Nevertheless, in recent years, with Christina Salmivalli and her team’s extensive implementation and evaluation of the KiVa program in Finnish schools (i.e., Haataja et al., Reference Haataja, Sainio, Turtonen and Salmivalli2016; Herkama et al., Reference Herkama, Kontio, Sainio, Turunen, Poskiparta and Salmivalli2022; Herkama et al., Reference Herkama, Saarento, Salmivalli and Sturmey2017; Sainio et al., Reference Sainio, Herkama, Kontio, Salmivalli and Smith2019), and its implementation in other countries in Europe and Latin America, certain rather alarming results began to be detected. Improved social climates, or ’healthy contexts’, placed the most vulnerable children at a higher risk.

Hence, in classes where victimization levels had been reduced by 75%, students who were still victims of bullying showed higher levels of depression and lower levels of self-esteem, compared to the pre-intervention time and to the control group that had not received any intervention (Garandeau et al., Reference Garandeau, Lee and Salmivalli2018; Huitsing et al., Reference Huitsing, Lodder, Oldenburg, Schacter, Salmivalli, Juvonen and Veenstra2019). That is, victimized students had worse indicators of psychosocial adjustment after a highly effective intervention. In 2018, Salmivalli coined the phrase "the healthy context paradox" to refer to this phenomenon (Salmivalli, Reference Salmivalli2018).

This situation had already been detected previously in cross-sectional and longitudinal studies carried out in different countries (e.g., the United States, Netherlands, Finland, Italy, etc.), with no prior intervention. For instance, several cross-sectional studies found that in classrooms with lower levels of victimization, victimized students presented worse self-esteem (Huitsing et al., Reference Huitsing, Snijders, van Duijn and Veenstra2014), greater somatic problems (Gini et al., Reference Gini, Holt, Pozzoli and Marino2020) and lower levels of peer acceptance (Sentse et al., Reference Sentse, Scholte, Salmivalli and Voeten2007), compared to classrooms with higher levels of victimization. In the same vein, longitudinal studies showed that after one year, victimized students presented worse indicators of depression and self-esteem in low-victimization contexts than in high-victimization contexts (Garandeau et al., Reference Garandeau, Lee and Salmivalli2018). However, what was paradoxical was that these results were also being obtained after the successful implementation of intervention programs (Huitsing et al., Reference Huitsing, Lodder, Oldenburg, Schacter, Salmivalli, Juvonen and Veenstra2019; Kaufman et al., Reference Kaufman, Kretschmer, Huitsing and Veenstra2018), so they could be due to possible iatrogenic effects of school-based prevention programs and, more specifically, of the KiVa program.

These disturbing results caused a great impact among experts in the prevention of bullying in educational contexts, who tried to understand the causes. It could be said that, for a brief period, the scientific community was stunned. Some authors even questioned the usefulness of universal prevention programs that promoted the intervention of bystander students (Healy, Reference Healy2020), finding them possibly detrimental to students who were in a more vulnerable situation.

Fortunately, there was a prompt response from the scientific community, led especially by Christina Salmivalli, and the crisis was seized as an opportunity to advance the field of bullying prevention. The first conclusion reached was immediate: The evaluation of bullying prevention programs must include other effectiveness indicators beyond the levels of victimization and bullying (Garandeau & Salmivalli, Reference Garandeau and Salmivalli2019), such as levels of self-esteem, depression, life satisfaction, and so on. This conclusion, which had previously been proposed by Juvonen et al. (Reference Juvonen, Schacter, Sainio and Salmivalli2016), was merely practical in nature and somewhat superficial, as it did not involve rethinking the way bullying was being studied and prevented.

From a deeper and more conceptual point of view, possible explanatory hypotheses for this paradoxical outcome began to emerge, with possible impact on the way bullying had been addressed thus far. Several explanatory hypotheses have been proposed since 2019 (e.g., Garandeau & Salmivalli, Reference Garandeau and Salmivalli2019; Huitsing et al., Reference Huitsing, Lodder, Oldenburg, Schacter, Salmivalli, Juvonen and Veenstra2019), with three of them attracting particular attention. These three hypotheses attempt to explain the causes of the poorer psychological adjustment of victimized students in contexts with low levels of victimization. The first hypothesis argues that the greater maladjustment of victims after intervention is simply a reflection of a high baseline distress that, over time, may become chronic. The second attributes the worse adjustment to the result of a cognitive mechanism related to social comparison and causal attributions. The third emphasizes that victimization leads to a series of interpersonal problems for the student victim, which then have an effect on their psychological adjustment.

These three hypotheses are presented in more detail below. To give them empirical validity, we rely on some recent work, particularly, the two-year longitudinal study conducted in China by Pan et al. (Reference Pan, Li, Ji, Malamut, Zhang and Salmivalli2021). To our knowledge, this is the only study that has empirically examined some of these explanatory hypotheses using a longitudinal design. Specifically, this study involved 2,643 children from 51 primary school classrooms in China. The participants were assessed at three different times, at the beginning of the study (T1), at one year (T2) and at two years (T3). In order to explore the explanatory hypotheses, the authors combined multilevel and mediated moderation modeling. In this way, they were able to examine the simultaneous and interactive effects of different individual variables (i.e., depressive symptomatology, victimization, number of reciprocal friendships, and social self-concept) and the classroom variable (level of victimization). These complex analyses controlled for different variables pertaining to the individual (students’ baseline depression levels, gender, and grade) and to the classroom (classroom size). The victimization levels of each classroom were obtained by averaging the victimization levels of the students in that classroom. These individual victimization levels were in turn provided by peers through peer nomination procedures.

As previously stated, the first hypothesis postulates that victimized students with greater psychological maladjustment are more likely to become chronic victims, and therefore would also present worse levels of depression and self-esteem over time. When this hypothesis was tested in the study by Pan et al. (Reference Pan, Li, Ji, Malamut, Zhang and Salmivalli2021), the results again evidenced “the healthy context paradox”, namely, classrooms with low levels of victimization showed increased depression in students who remained victims. In other words, “healthier contexts” would increase the psychological difficulties of chronically victimized students (Garandeau & Salmivalli, Reference Garandeau and Salmivalli2019). What was most interesting, however, was that this occurred regardless of the students’ initial level of depression, whether high or low. These findings, therefore, seem to dismiss the hypothesis that “the healthy context paradox” is due solely to the initial adjustment levels of the student victim (Garandeau & Salmivalli, Reference Garandeau and Salmivalli2019).

The second explanatory hypothesis proposes that classrooms with lower levels of victimization lead to worse levels of psychological adjustment in the victimized students through two possible cognitive mechanisms. One of them would be explained by Weiner’s Causal Attribution Theory (Reference Weiner1986). In contexts with low victimization, students who continue to be victims when their other victimized peers are not, are more likely to blame themselves and attribute their situation to internal, stable, and uncontrollable causes. What is more, this attribution would not only be made by the victimized students, but also by the rest of their peers who see how other peers have managed to change their situation. To our knowledge, this explanatory mechanism has not yet been empirically tested.

The other potential cognitive mechanism is explained by the Social Comparison Theory (Festinger, Reference Festinger1954). According to this theory, people have a natural tendency to compare themselves with others and generate, from this, an evaluation of them. When we compare ourselves with others, two evaluative outcomes can emerge: One is positive (descending or downward) when we benefit from the comparison, and one is negative (ascending or upward), when we are disadvantaged. The type of comparisons made by the person (ascending or descending) will therefore have an impact on his or her self-concept and well-being. Thus, in a context with low levels of victimization or with few victimized students, those students who are victims would tend to make ascending comparisons with negative results, leading them to negatively self-evaluate their social competence. This would affect their social self-concept leading to higher levels of depression. This maladaptive upward comparison and its impact on social self-concept are less likely to occur in a context with high levels of victimization or with a high prevalence of victimized students. This hypothesis was also examined in the longitudinal study by Pan cited above. The results showed that, in those classrooms with lower levels of victimization, victimization at baseline (T1) predicted lower social self-concept the following year (T2), and this in turn predicted greater depressive symptomatology at two years (T3). That is, social self-concept at T2 had a mediating role between victimization at T1 and depression at T3. Therefore, healthier classroom contexts led to worse levels of depression over time, in students who remained victimized, through the negative impact of victimization on social self-concept.

The third explanatory hypothesis proposes that contexts of less victimization or decreasing victimization could worsen the psychological adjustment of victimized students by negatively influencing their social relationships, especially their opportunities to make friends. A multitude of studies indicate that victimization predicts lower peer acceptance and reduces the likelihood of forming friendships (Sentse et al., Reference Sentse, Dijkstra, Salmivalli and Cillessen2013; Sijtsema et al., Reference Sijtsema, Rambaran and Ojanen2013). Indeed, the bidirectional and transactional relationship between bullying and peer rejection has been extensively studied (Pouwels et al., Reference Pouwels, Lansu and Cillessen2016; Pouwels et al., Reference Pouwels, Salmivalli, Saarento, van den Berg, Lansu and Cillessen2018). This hypothesis is based on the Peer-group Dissimilarity Theory (Wright et al., Reference Wright, Giammarino and Parad1986). According to this theory, the attitudes of the group members towards their peers depend on what is considered normative, and therefore valued and accepted, in the peer group (Wright et al., Reference Wright, Giammarino and Parad1986). Consequently, victimized students would be considered “misfits” in classrooms with low levels of victimization and would be more likely to be peer rejected in these classrooms (Kaufman et al., Reference Kaufman, Laninga‐Wijnen and Lodder2022; Sentse et al., Reference Sentse, Dijkstra, Salmivalli and Cillessen2013). This leads to their non-victimized peers avoiding them as friends (Sentse et al., Reference Sentse, Dijkstra, Salmivalli and Cillessen2013; Sijtsema et al., Reference Sijtsema, Rambaran and Ojanen2013), as well as them tending to associate with other victimized students who share their plight (Huitsing et al., Reference Huitsing, Snijders, van Duijn and Veenstra2014; Lodder et al., Reference Lodder, Scholte, Cillessen and Giletta2016). The latter would be even more difficult in classrooms with a low presence of victimized students. If we add to this the fact that having friends not only protects against victimization, but also buffers the negative impact of victimization on internalization problems (Hodges et al., Reference Hodges and Perry1999; Schmidt & Bagwell, Reference Schmidt and Bagwell2007), victims in a low victimization context find themselves in a highly vulnerable position. Pan et al. (Reference Pan, Li, Ji, Malamut, Zhang and Salmivalli2021) also explored this third explanatory hypothesis in their longitudinal study. The authors found that, in those classes with lower levels of victimization, victimized students (T1) had fewer reciprocal friendships one year later (T2), resulting in greater depressive symptomatology the following year (T3).

It is important to note that in the study by Pan et al. (Reference Pan, Li, Ji, Malamut, Zhang and Salmivalli2021), only the number of reciprocal friendships was taken into account, without considering their quality. On the other hand, it should be mentioned that 96% of the participants in the study remained in the same classrooms throughout the two years; therefore, the stability of the context and of the roles played (e.g., victim) may have had an impact on their results. It is likely that the impact of victimization on psychological adjustment is greater in more stable contexts than in more dynamic contexts (where it is potentially easier to change roles). Therefore, these results should be replicated and investigated in other contexts where the social structure of the classroom, its stability, and its effect on “the healthy context paradox” are analyzed. In addition, it would be interesting to examine the classroom social dynamics by adopting a social networking approach (Kaufman et al., Reference Kaufman, Huitsing, Bloemberg and Veenstra2021; Rambaran et al., Reference Rambaran, Dijkstra and Veenstra2020).

Implications for the Study and Prevention of School Bullying

As can be seen, the conceptual and empirical background of the three explanatory hypotheses is relatively short, as they have been proposed and examined in the last three years. As is often the case in the field of psychology, it is likely that these hypotheses explain “the healthy context paradox” in a complementary way, and only in part. In other words, it is unlikely that any one of them alone can comprehensively and exclusively explain this phenomenon, especially considering that the socio-cognitive mechanisms involved are closely related, as we have seen.

Although “the healthy context paradox” initially generated some alarm in the area of bullying prevention, it has undeniably led to a better understanding of the mechanisms underlying bullying. Certain implications for the study and prevention of bullying have recently been derived from this paradoxical effect. One of them has already been discussed above and has to do with the need to incorporate other indicators of psychological adjustment in evaluating the effectiveness of prevention programs, beyond the levels of victimization and bullying.

Claire Garandeau (Reference Garandeau2021), one of the first, along with Christina Salmivalli, to identify and label this phenomenon resulting from reduced levels of victimization, presents a series of suggestions for bullying prevention that seek to avoid the appearance of “the healthy context paradox”. These proposals consist of: (a) Helping the victimized students to modify their causal attributions, making them see that they are not to blame for their situation, (b) changing their perceptions to see that they are not the only ones who have been bullied, thus helping them protect their self-concept by modifying their frame of reference when making possible social comparisons, and (c) developing coping strategies among these students, so that they learn to ask for help from school staff and to respond assertively to bullying.

As can be seen, these actions would focus mainly on victimized students. Undoubtedly, these more individualized interventions are necessary, but not sufficient. We cannot again lose sight of the context when it comes to understanding and preventing bullying, and focus only on the victim. As we stated at the beginning of this paper, bullying is initiated and maintained in a specific context and not in another. It goes beyond the characteristics of the student who suffers or perpetrates it. In this regard, while it is important to help victimized students make healthier and more adaptive causal attributions and social comparisons, as well as develop their assertiveness and coping strategies in bullying situations, it is also essential to do so with the rest of the students. It is unlikely (or at least more difficult) for a student to change the causal attributions of their victimization situation if their immediate context is reinforcing the opposite. It will be easier to change these attributions and perceptions in a context that does not tolerate this type of behavior and, therefore, neither blames the victims for the bullying they experience, nor leaves them alone. The victim is not responsible for learning to see their experience differently, but rather this responsibility must fall on the school, by activating all its resources to modify the victim’s reality. In order to achieve mutually beneficial coexistence at school, there must be shared responsibility among all the components of the educational community, where the goal becomes the wellbeing of each and every person. On the other hand, it is necessary to reflect on the real effectiveness of an intervention that, after several years of implementation and having managed to significantly reduce the levels of victimization, has failed to improve the relational climate of the context in which the bullying took place.

Thus, one of the greatest risks derived from the phenomenon of “the healthy context paradox” may be to lose the contextual perspective in understanding, explaining, and preventing bullying, and once again leave the victim alone in the face of a group and contextual problem. Let us recall here the definition of bullying: that type of violence that is perpetrated intentionally and repeatedly by a student or group of students against another student who cannot defend himself/herself due to an imbalance of power (Hymel & Swearer, Reference Hymel and Swearer2015; Salmivalli, Reference Salmivalli2010). Consequently, the student victim is in a situation of helplessness from which they cannot escape by their own means, it is, therefore, necessary for the context to act in their defense and protection (Rambaran et al., Reference Rambaran, Dijkstra and Veenstra2020; Yun & Juvonen, Reference Yun and Juvonen2020).

In relation to the latter, some authors have asked whether other contextual variables may have a protective effect against the negative impact of low-victimization classrooms on the adjustment of victimized students. For instance, Yun and Juvonen (Reference Yun and Juvonen2020) wondered whether in contexts with low levels of victimization, peer defending behaviors might moderate the effect of victimization on depression. To this end, they conducted a study with 1,373 adolescents in 54 classrooms in South Korea and found that, as expected by “the healthy context paradox”, the association between victimization and depression was greater in those classrooms with lower levels of victimization. However, they also found that the levels of peer defending behavior moderated this relationship. In this way, they observed that classrooms with higher levels of defending behaviors presented smaller effects of victimization on depression than did classrooms with lower levels of defending behaviors. Nevertheless, in order to have a buffering effect on the relationship, the defending behaviors had to be directed at the bully (bully-oriented or aimed at confronting and stopping the bully) and not at the victim (victim-oriented or aimed primarily at offering comfort to the victimized student). This moderating effect occurred independently of the levels of victimization in the classroom. Therefore, the victim’s adjustment was not only dependent on the levels of victimization in the classroom, but also on the levels of defending behavior. It is likely that in classrooms where peers confront the bully students, the victim students have less tendency to blame themselves and, consequently, to experience depressive symptomatology. This result opens another line of research aimed at further exploring the protective effect of other contextual variables on the distress experienced by the victimized student. In the case of the impact of defending behaviors, further investigation must differentiate whether these are directed at the bully or the victim. It should be noted that this typology of defending behaviors has become established in recent years (Gini et al., Reference Gini, Pozzoli, Angelini, Thornberg and Demaray2022; Reijntjes et al., Reference Reijntjes, Vermande, Olthof, Goossens, Aleva and van der Meulen2016; Yun, Reference Yun2020; Yun & Graham, Reference Yun and Graham2018), and is providing valuable insights into the mechanisms underlying bullying, as well as essential clues for its prevention (Yun & Juvonen, Reference Yun and Juvonen2020).

In summary, as contributions to the prevention of bullying that can be drawn from “the healthy context paradox”, we can highlight:

1. 1. When evaluating the effectiveness of bullying prevention programs, other effectiveness indicators beyond levels of victimization and/or bullying (e.g., indicators of psychosocial adjustment) should be considered, and interventions should be monitored, incorporating these other indicators as well. Students with higher levels of victimization and maladjustment need to be identified at an early stage for more detailed follow-up.

2. 2. Within the whole-school approach to prevention, apart from incorporating and combining universal preventive actions (for the whole educational community and students) with specifically-indicated actions (for students who show behaviors and symptomatology related to victimization or bullying), more individualized interventions must also be aimed at persistent cases of bullying (e.g., individual sessions outside school hours aimed at developing adaptive coping strategies for bullying or encouraging re-attributional training). These interventions should be provided by professionals specialized in bullying prevention, and coordinated with the actions implemented by educational staff with the rest of the students in the classroom and school settings. Through this combination of multilevel actions, it would be possible, on the one hand, to change the relational climate of the classroom, by promoting friendship and defending behaviors, and on the other hand, to offer individualized and group support to the victimized students.

3. 3. Finally, the protective role of other contextual variables in “the healthy context paradox” will need to be further considered and investigated. For instance, considering the role of teacher-student relationships in bullying and defending behaviors (e.g., Lucas-Molina et al., Reference Lucas‐Molina, Williamson, Pulido and Pérez‐Albéniz2015; Sjögren et al., Reference Sjögren, Thornberg, Wänström and Gini2021; Thornberg et al., Reference Thornberg, Wegmann, Wänström, Bjereld and Hong2022), it would be desirable to include the effect of other variables related to teachers’ attitudes and behaviors on “the healthy context paradox”.

In conclusion, much remains to be done to understand the causes of “the healthy context paradox”. While there is no doubt that schools must continue to make efforts to reduce victimization and bullying in the classrooms, making these contexts truly healthy for all students requires additional measures to effectively protect students who continue to be victims of bullying after intervention. And these measures must involve the entire educational community. In addition, further research into the possible iatrogenic effects of anti-bullying interventions is essential (Healy, Reference Healy2020; Salmivalli et al., Reference Salmivalli, Laninga‐Wijnen, Malamut and Garandeau2021). In that regard, as previously stated, future research should examine other contextual variables apart from the levels of victimization in the classroom (e.g., defending behaviors, teacher-student relationships, etc.). It should also analyze the social structure of classroom relationships and how they change with the intervention (studying social networks and the stability/dynamism of the context), and consider the quality of friendships, not merely their quantity. If the ultimate goal is to improve the relational climate of the classroom, it is crucial to establish a baseline assessment. The approach based on social network analysis can be very helpful in this regard (Martín Babarro, Reference Martín Babarro2014; Kaufman et al., Reference Kaufman, Huitsing, Bloemberg and Veenstra2021). We cannot leave the victimized student alone; it is imperative that future work identifies the factors that enable classrooms and schools to be safe places for all students.