We conducted this study at two middle schools in the public school system of a large city in the Southeastern U.S. and a middle school in an adjoining county. The university’s institutional review board approved all procedures. We recruited participants from a sample of sixth grade students in intervention classrooms who completed pretest measures for Sullivan and colleagues’ (2014) study and who were present on at least half of the days the Second Step intervention was taught. The Sullivan et al. sample represented 82% of the students in the two urban schools and 75% of students in the county school who provided both parental consent and student assent.

The urban schools served a predominantly African American student population. The majority of students (i.e., 83%) were eligible for the federally subsidized school lunch program. We recruited 91 students (41 from one school and 50 from the other) from all 102 students that met the study’s selection criteria. The final sample included a slightly higher percentage of boys (54%) than girls. The average age was 11.8 (SD = 0.58). The majority (83%) identified themselves as Black or African American, and 12% identified themselves as representing more than one race. Only one student self-identified as Hispanic or Latino. Most participants described their family structure as two-parent (24%), single mother without other adult (20%), single mother with stepfather (19%), single mother with adult other than grandparent (9%) single mother with grandparent (8%), or grandparent without father or mother (6%).

The county school was in a rural setting, classified as “Rural Fringe” by the Census Bureau, which was in close proximity to the urban area. Twenty-two percent of the students were eligible for the federally subsidized school lunch program. We recruited 50 students from a random of 66 students that met the study’s selection criteria. The final sample included a slightly higher percentage of boys (54%). The average age was 11.9 (SD=0.78). The county sample was more diverse than the urban sample, with 52% identifying themselves as White, 32% as Black or African American, and 12% as more than one race. None identified themselves as Hispanic or Latino. Most students in the county schools described their family structure as two-parent (58%), single mother without other adult (19%), or single mother with stepfather (10%).

Sullivan et al. (2014) implemented Second Step during the 2008-09 school year as part of a randomized trial. This universal school-based violence prevention curriculum was designed for middle and junior high school students, and focuses on decreasing aggression by enhancing adolescents’ social and emotional competencies (Committee for Children, 1997b). The curriculum includes 15 lessons that cover interpersonal conflicts, empathy and perspective-taking, anger management, and problem-solving. Lessons also address applying these skills to make a complaint (i.e., talking to someone when they did something that bothered or hurt the student), address peer and gang pressure and bullying situations, and diffuse a fight. Sullivan and colleagues (2014) included an additional lesson on prosocial behavior and social networks, and incorporated problem situations for role plays derived from prior qualitative work to identify relevant problem situations encountered by urban youth (Farrell et al., 2006, 2007). Implementation training for project staff addressed both the curriculum and best practices for effective classroom instruction. Sullivan and colleagues encouraged teachers to be present to assist with the lessons. Fidelity of implementation was high based on observers ratings of a random sample of each interventionist’s lessons (see Sullivan et al, 2014 for details).

We interviewed students during the same school year in which the curriculum was implemented (N = 42) if they were in classrooms in the urban schools where the intervention was completed during the first half of the school year. We interviewed students during the following school year if they were in classrooms in which the curriculum was completed during the second half of the school year (N = 49 students in the urban schools and 50 in the county school). Research staff conducted the interviews at participants’ schools, typically in the media center or in unoccupied classrooms. Interviews were semi-structured, lasted about an hour, and were organized into four parts (see Table 1). During Phase 1, interviewers asked participants open-ended questions about their general reactions to the intervention. In Phase 2 interviewers asked participants to describe a problem situation they had encountered with a peer and to respond to a series of questions about any experience they had in which they attempted to apply something they learned from the intervention. In Phase 3 interviewers asked participants about their reactions to and experiences attempting to apply each of eight key skills targeted by the intervention. In Phase 4 they asked participants to rate specific aspects of the intervention, and to discuss their general reactions to it. For the protocol used during the spring of 2009, interviewers asked participants to describe experienced or anticipated outcomes for each of the eight intervention skills, and to rate these outcomes on a five-point scale from 1 (terrible) to 5 (great).

The 13 interviewers had bachelors or masters degrees, with the exception of one undergraduate senior in the honors program. We trained interviewers with a combination of group meetings and individual supervision. Supervisors listened to audio recordings and provided written and verbal feedback for at least two practice interviews for each interviewer. Interviewers practiced conducting interviews with their supervisors and did not conduct interviews in the schools until their supervisors had confidence in their proficiency. Supervisors provided written feedback on each interviewer’s first three interviews. In cases where supervisors noted any significant issues, we required interviewers to practice the interview again. African American women conducted 58% of the interviews, European American women 26% of interviews, European American men 13% of interviews, and a Latino American woman 3% of the interviews. Female interviewers interviewed both boys and girls, but male interviewers only interviewed boys. Research assistants transcribed the interviews. For the first wave of data collection, research assistants reviewed 36% of transcriptions randomly selected for accuracy. For the second wave, they reviewed all transcriptions for accuracy. Although 141 students completed interviews, three could not be transcribed due to faulty recording equipment.

We conducted a qualitative analysis of interview transcripts to identify themes related to participants’ general reactions to the intervention, descriptions of any intervention skills or changes in behavior they made following the intervention, and their perceived outcomes from use of the skills. Coding was an iterative process in which we generated and refined themes based on multiple passes through the transcripts. We began with open coding wherein two investigators independently read approximately 25% of the transcripts from the first wave of interviews and identified themes. This led to an initial set of codes and preliminary definitions with associated examples from the transcripts. We further refined themes and definitions based on consensus coding by the same two investigators, who reviewed the full set of 42 interviews collected during the first year. We conducted consensus coding using NVivo 7 qualitative coding software to generate documents that included all of the text we coded for each theme. During a third stage of coding, three different investigators read through these documents and identified subthemes within each category. We used the resulting themes with definitions and examples to create a preliminary coding manual. For one question in which we asked how students handled a problem situation, we established codes a priori. We assumed that students would describe one of the intervention skills, a prosocial response consistent with the intervention, or an antisocial or aggressive response.

During the fourth stage of coding, a team of five investigators coded all 138 interview transcripts based on the preliminary coding manual. They organized themes into five groups representing higher order categories (e.g., general reactions to the intervention, changes in behavior, outcomes) and we assigned a pair of coders to each group. As part of the training process, each coder independently coded ten randomly selected transcripts. We then compared their codes and discussed any discrepancies until the coders reached consensus. In some cases this led to further revisions to the definitions that we incorporated into the coding manual. During this process it became evident that the a priori codes for behavior change did not effectively capture the participants’ responses. We therefore created new codes using the iterative process described previously. We compared codes for three transcripts halfway through the coding to address potential coder drift. Each pair of coders reached consensus on these transcripts. We assessed intercoder reliability by overlapping the two coders on a randomly selected subset of 28 interviews (i.e., 20%). We also identified and defined subthemes for outcomes that were initially coded into broad heterogeneous categories (e.g., “Positive Outcomes of Using Skill).

Although inter-coder reliability for coding of themes was fairly high, there were several exceptions. Several categories that included themes that did not fit into more specific categories (e.g., Other Positive Reactions) had acceptable percentages of agreement (i.e., 88% to 95%), but low base rates that resulted in low kappa coefficients. Agreement across coders as to whether students misused or misunderstood a skill was also acceptable in terms of percentage agreement (i.e., 89% to 93%), but low base rates again resulted in low kappa coefficients. We addressed this issue by having two investigators independently code the transcripts, and then discuss and resolve any discrepancies. For all other themes kappa coefficients ranged from .43 to 1.00, and percentages of agreement ranged from 76% to 100% (see Tables 2 and 3).

Participants’ overall reactions to the intervention were generally quite favorable evidenced by ratings of 4 or 5 on each of the five-point scales assessing their perceptions of different aspects of the intervention (see Table 4). The majority (76%) rated the intervention as “pretty good” or “great,” indicated that what they learned in the intervention would work “pretty well” or “great” in solving or avoiding conflict (69%), that they found the intervention “pretty useful” or “really very useful” for addressing problems (71%), and that they would recommend the intervention to other students (87%). The great majority of participants (87%) also reported that they had used something they learned in the intervention at least once in the past month, and 76% indicated that using the skills helped them get what they wanted. Ratings on the six scales did not differ across sites (i.e., urban versus county schools) at p < .05.

When asked to explain their ratings, nearly all participants (99%) made at least one positive comment (see Table 2). These were typically general comments indicating that the intervention helped them stay out of trouble, prevented or reduced fights and other problem behavior, or would help them with future decisions and problems. For example, a male student at one of the urban schools indicated that the lessons “kept us out of problem instead of always goin’ around fightin’ and doin’ different things like that.” Some participants (55%) discussed learning a specific skill they found useful. For example, one female student in an urban school stated that “it helped me think it out. When I get angry, think it out before I like make a wrong decision.” Some participants (24%) reported that they found the intervention interesting or enjoyable. The percentage of interviews in which each positive theme was coded did not differ across sites at p < .05.

The great majority of participants (79%) also identified at least one thing they did not like about the intervention. Almost two-thirds (64%) said that some aspects were not helpful or useful. Comments focused on the relevance of the intervention (“I really don’t need ‘em”) and on the perceived ineffectiveness of certain aspects of it. One female student at an urban school pointed out that one aspect of the intervention was ignoring provocative behavior and said, “If you ignore them, all they do is want to mess with you.” Some students (16%) said that they found the intervention boring. One male student in a county school explained, “They kept saying the same thing like over and over again.” There was one significant site difference in the percentage of codes for negative reactions. A higher percentage of students in the county school made negative comments that did not fit into a specific theme (50% versus 18%; χ²(1) = 15.79, p < .001). These comments generally indicated that the lessons were too long, that they cut into health class or gym time, or that other students didn’t listen to or like the lessons. Less frequent comments were that the students did not like to have to talk about “stuff” or that the students didn’t understand aspects of the intervention.

Interviewers also asked participants about their friends’ reactions to the intervention. Thirty-eight percent identified positive reactions, such as finding the intervention useful or fun. A male student at one of the urban schools reported that one of his friends “really liked it because it calms his anger down.” Forty-two percent of participants identified negative reactions by friends. These included statements that their friends found the intervention boring, were frustrated that the intervention cut into their gym time, did not pay attention during class, or continued to fight regardless of the intervention. A female student in an urban school provided the following insight into her friends’ perspectives: “Some people don’t like sitting down and talking to some adults about things, about problems and issues because they think it’s kind of punkish.” There were no significant site differences in the percentage of students who reported positive or negative friends’ reactions at p < .05.

Participants offered several suggestions for ways to improve the intervention. Twenty percent suggested that the intervention could be improved by making it more fun or adding activities. For example, a female student in an urban school said, “Like make it fun and not just, you know, sitting in the seats all the time, like do different games and stuff.” Others (14%) emphasized the need to discuss a wider variety of problems and include more relevant situations. For example, a female student in one of the urban schools stated that “a lot of the people there, they don’t have problems like what they had on they, um, scenario.” Another female student at one of the urban schools suggested that interventionists ask students to identify problem situations before teaching skills: “Ask children, like before you start, asking kids what they going through and see what they going through and then you make a list of things that they going through and go, and then you see how they will handle it.” Others responded that they would like the lessons to go into more depth. As one male student at an urban school explained:

There were no site differences across categories in the percentage of students who made suggestions for improving the intervention at p < .05.

During the initial stage of the interview, interviewers asked students if they had done anything differently because of what they learned in the intervention. The majority of participants (59%) identified one or more changes they made following the intervention (see Table 3). The most frequent (22%) concerned controlling anger. For example, one female student at an urban school reported that she “calmed down a little bit when something made me mad.” Participants also described efforts to improve relations with others (19%). One male student at an urban school explained, “Instead of getting ready to get into a fight, we talked it out.” A female students at an urban school said, “I’ve started respecting more people.” Participants also indicated decreasing their involvement in problem behaviors (12%), such as giving into peer pressure, talking back to teachers, or (more vaguely) getting into trouble. One male student at an urban school reported, “Last year I was getting a lot of referrals, but this year I ain’t got not one referral.” Participants also reported decreasing their aggressive behavior (11%). For example, one female student at the county school said, “I’ve stopped gossiping about people, I’ve stopped being a bully.” Students in the urban schools were more likely than those in the county school (i.e., 13% versus 0%) to report that they decreased other negative behaviors (χ²[1] = 7.01, p < .008). There were no other site differences in these codes at p < .05.

During the second phase of the interview, interviewers asked participants to identify a problem they recently had with a peer or at school to provide a more specific context for discussing their use of the skills taught in Second Step. The majority of participants (80%) described at least one way in which they used something they learned in the lessons to help them deal with the situation (see Table 3). The most frequently mentioned theme (43%) was improving communication or relations with others. Examples included efforts to talk it out with the other person, asking the other person to stop the provocation, and taking responsibility for their own actions. One male student at an urban school said, “I just apologized to him…and told him I didn’t really mean it.” When a male student at an urban school had an electronic item broken by another student, he explained, “I talked to him and told him that he would have to give me another one, but usually…I would’ve gotten mad.”

Another frequently described change (i.e., 39% of participants) involved ignoring the provocation or walking away from problem situations. For example, one female student at an urban school said, “I ain’t say nothing back. I just walked away.” A male student at an urban school explained that he was able to ignore the provocation: “[he] just come up to me and like say something. I ain’t even worried about it.” Other students (21%) reported that they handled the situation by controlling their anger. For example, a female student in an urban school reported, “I, um, calmed down and then I took them deep breaths and I count back and then I just moved away from her.” Still others (12%) explained that they avoided making aggressive responses. One student at an urban school said that she handled the situation differently by “learning to use my mind instead of my fist.” A male student in an urban school said, “I didn’t fight him because if I would have it could have caused trouble.” Some participants (11%) stated that they sought support from parents, teachers, or friends. For example, one female student in an urban school said, “I went to go talk to my teacher and worked it out with the other person.” A student in the county school noted that she was able to “just breathe and go talk to…friends about it.” A few students reported that they changed their behavior by thinking before acting. Another female student in the county school said, “I was probably thinking before I acted, or thinking before I said anything.” There were no site differences in the percentage of students identifying themes within these categories at p < .05.

Interviewers asked those students who were interviewed the year following the intervention to rate how well using what they learned in the lessons worked out for them. The mean of these ratings was 3.7 (SD = 0.75) which falls between ok and pretty good on the scale of 1 to 5. Mean ratings did not differ across sites (t[83] = 0.28, ns). Very few participants (i.e., 2) assigned ratings of 2 or lower, corresponding to “terrible” or “pretty bad.”

During the third phase of the interview, interviewers asked participants about their use of the eight skills that constituted the focus of the intervention. Interviewers asked participants: if they had used each skill, to describe how they had used it, and to rate how well things worked out. On average, participants reported using 4.6 skills (SD = 1.8). The mean number of skills used did not differ across sites (t[139] = 1.01, ns). The percentage of participants who reported using each skill ranged from 13% for resisting gang pressure to 82% for anger management. With the exception of resisting gang pressure, at least half of the participants reported using each skill (see Figure 1). Results of a repeated measures logistic regression analysis indicated that there were significant differences in rates of use across skills, (χ²[7] = 138.72, p < .001). Contrasts comparing each skill to the average indicated that a higher percentage of participants reported using anger management, empathy, making a complaint, and diffusing a fight, and a lower percentage reported resisting gang pressure (all ps < .05). There were significant differences in patterns of skills used across sites, (χ²[7] = 24.39, p < .001). Follow-up analyses revealed that compared to students at urban schools, those in the county schools were less likely to report using anger management (70% versus 88%; χ²[1] = 6.88, p = .012), and diffusing a fight (52% versus 70%; χ²[1] = 4.70, p = .043), but were more likely to report making a complaint (76% versus 57%; χ²[1] = 4.97, p = .029).

For each skill participants used, interviewers asked them to describe how they used it. Although their descriptions generally did not provide a sufficient basis for determining the degree to which they made appropriate use of skills, there were some cases in which it was quite evident that they did not use a skill correctly. Coders coded these as a misuse of a skill when they reached a consensus that the students’ descriptions made it clear that they used it inappropriately. Inappropriate use of skills included descriptions of antisocial, aggressive, or clearly ineffective responses, and when participants described use of a different skill (e.g., explained how they ignored someone when asked to describe how they used problem solving). Close to half the sample (45%) gave a description of using at least one skill that clearly suggested they misused it. This percentage did not differ across sites (χ²[1] = 1.12, ns). Across the skills, the percentage of the total sample whose descriptions suggested they used a skill incorrectly ranged from 1% for resisting gang pressure to 18% for making a complaint (see Figure 1). These percentages ranged from 4% to 23% when based on the percentage of students who reported using each skill rather the percentage of the total sample. Misusing a skill was most common with making a complaint. We defined making a complaint as talking to someone when they did something that bothered or hurt the student, but many students seemed to understand the skill as simply complaining. For example, one female student in the county school said, “Well, I always complain in gym about when we have to go outside.” A male student in an urban school explained how he used dealing with peer pressure when his mother told him to do his chores: “When I about to lay down, I have to do my chores and stuff like, I be like, oh, Lord, help me.” When discussing empathy, one student in an urban school described how she noticed her friend looking worried, and discovered that her friend was going to be banked (i.e., physically attacked) by a group of children after school. She told her friend, “If they bank you, I’ll help you [fight].” Compared to students in the county school, students in the urban schools were more likely to misuse problem solving (27.3% vs. 4.0%; χ²[1] = 5.82, p = .016).

When asked why they did not use a specific skill, 32% of the sample gave at least one description that indicated that they did not understand it. In some cases, students’ responses suggested that they were confused, but it was not clear whether they misunderstood the skill or the question we asked. The percentage misunderstanding specific skills ranged from 3% for dealing with bullying to 24% for empathy (see Table 5). As with misuse of skills, some students confused making a complaint with complaining. Others misinterpreted resisting gang pressure to mean joining a gang. Similarly, some students misunderstood the skills for dealing with bullying to mean bullying other students. Students interpreted empathy in several different ways. For example, when discussing empathy, one student in a county school explained that he would not use empathy because “I actually do think about people but I don’t really feel like I should like try to be like them.” A higher percentage of students at the urban schools misunderstood at least one of the skills (38.9% versus 18.8%, χ²[1] = 5.85, p = .021). This was largely due to a higher percentage of students in the urban schools not understanding the use of making a complaint (13.5% vs. 2.1%; χ²[1] = 4.72, p = .033), and resisting gang pressure (10.1% vs. 0.0%; χ²[1] = 5.20, p = .027).

Some students (i.e., 12%) did not remember one or more of the skills or could not recall learning about a specific skill even when shown handouts from the intervention. The percentage of students not remembering a skill ranged from 2% for dealing with bullying to 10% for empathy (see Table 5). One student in an urban school explained that he did not use empathy because “we learned about it, but I forgot about it.” Other students explained that they did not remember a skill because they were probably not in class that day, or that the teacher never covered it. One student in the county school indicated that he had not remembered a skill because of the time lag between the intervention and the interview: “I might have, like, did it, but it may have been like last year and lot of stuff’s happened since last year.”

Participants who used a specific skill rated how well it worked for them and how confident they were that they could use it. Those who did not use the skill rated how well they thought it would work and their confidence that they could use it if they wanted to. Comparisons of effectiveness and self-confidence ratings across skills revealed several significant differences between those who reported using and not using skills. Compared to participants who reported not using, those who reported using problem solving and making a complaint rated the effectiveness of these skills higher (d = .47, p = . 018; and d = .62, p < .001, respectively). They also rated their self-efficacy higher for problem-solving (d = .43, p = .01) and for empathy (d = .38, p = .049). We pooled data from users and non-users to allow comparison across skills. Mean ratings of individual skills showed little variation, ranging from 3.73 to 3.90 and did not differ across skills (F[7, 616] < 1, or sites, F[1, 88] < 1), and there was no Skill x Site interaction, (F[7, 616] < 1). Ratings of self-efficacy ranged from 3.69 to 3.94, did not differ across skills (F[7, 917] < 1, or sites, F[1, 131] < 1), and there was again no Skill x Site interaction (F[7, 917] = 1.22, ns). This suggests that most participants felt the skills would produce positive outcomes and were fairly confident that they could use each of them.

In addition to rating how well the skill worked for them, we asked students who used a specific skill to explain why they gave that rating. The majority of students (59% to 86%) who used a specific skill identified at least one desirable outcome. These included experiencing a positive outcome (29% to 76% across skills; see Figure 2). For example, when describing how she used anger management, one female student in an urban school said, “I was calm for the rest of the day.” When discussing empathy, most students said that using the skill helped them to maintain existing relationships or create new relationships. One male student in an urban school explained, “Seein’ how mad he was, was like, kinda like makin’ me feel like I did somethin’ wrong so it was like, I needed to calm down for what I did and apologize.” One student in a county school explained how problem-solving helped her: “Me and her like actually solved the problem together, and we apologized, and I told her that I was like really sorry for what happened.” Participants also described how using a specific skill helped them avoid negative outcomes, especially getting in trouble (7% to 36% across skills; see Figure 2). This theme emerged most frequently in regards to anger management, diffusing a fight, and dealing with peer pressure. For example, one male student in an urban school explained that using anger management “kept me from sayin’ stuff that would get me in even more trouble.” In discussing their use of diffusing a fight, most students mentioned that using the skill prevented them or their friends from fighting or getting into trouble. One student in an urban school described negative outcomes that he avoided: “if I would’ve fought him I would’ve probably still got, like, in trouble and…he like, you know how when you fight somebody and say if they lose or something they’ll just keep on wanting to fight you and fight you until they win.” Negative outcomes that students avoided by using skills for dealing with peer pressure primarily included getting into trouble and bodily harm. As one male student in an urban school so aptly put it, “I’m not [going to] steal from no store and go to Juvy [i.e., juvenile detention] with all them bad children.”

Participants also identified some negative outcomes they experienced from using specific skills. The percentage of students using a skill who identified negative outcomes ranged from 34% to 53% across skills (see Figure 2). Some participants reported experiencing general negative outcomes, including finding the skill ineffective, experiencing negative reactions from others involved in the interaction, and other negative outcomes, such as continued or increased provocation from other students. Skills for which participants most frequently identified negative outcomes involved dealing with peer pressure (41%), resisting gang pressure (53%), and dealing with bullying (48%). Some participants who used skills for dealing with bullying or dealing with peer pressure described how their peers’ provocation continued or reported that the other person in the interaction became angrier. For example, one female student in an urban school said, “he just kept trying to beat me up and stuff.” When one student in an urban school resisted peer pressure, he said the students who wanted him to skip class “was like, calling me a little punk.”

Interviewers also asked participants about the reactions of peers not directly involved in the problem situation. Some participants (6% to 14% across skills) stated that their peers had positive reactions (see Table 5) such as helping or encouraging participants to use the skills or approving participants’ use of the skills after the fact. One student in an urban school who used anger management said: “Some of my friend was happy because they ain’t want to see me get in trouble.” Another student in an urban school described how his friends encouraged him to use anger management. “They just said, ‘Calm down, sit in the chair, let’s don’t…just ignore her,’ stuff like that.” Some participants noted that their friends would have done the same thing in their situation. Participants also identified negative reactions from students not directly involved in the situation (9% to 22% across skills; see Table 5). This theme occurred most often when students used skills that involved diffusing a fight (22%), dealing with peer pressure (11%), or anger management (9%). In general, participants who identified negative peer reactions stated that using the skills disappointed other students who wanted to see a fight, or damaged participants’ reputation or social status. For example, when discussing diffusing a fight, one female participant in an urban school said that bystanders “wanted to boost [the fight] up more.” A female student in the county school noted that other students “were a little mad because they wanted to see a fight that day.” When describing her use of anger management, one female student in an urban school said, “the other ones looked at me like I was, like, scared of her or something. They looked like they was getting ready to laugh at me.”

This study has several key limitations. We asked students to provide retrospective accounts of their efforts to apply the skills addressed by the intervention. Observing their behavior in-vivo would provide a better test of their use of skills, but this would be difficult to do unobtrusively and raises ethical concerns related to privacy. More generally, the study focused primarily on participants’ reactions to the intervention, and their reports may have been biased by efforts to tell interviewers what they assumed they wanted to hear. We attempted to minimize this potential bias by means of a thorough interviewer training process that emphasized neutral verbal and nonverbal responses to comments and open-ended questions. This appears to have been at least partially successful, in that many participants freely identified negative reactions to various aspects of the intervention, described negative consequences of their efforts to apply the skills taught in the intervention, and admitted that they did not remember specific skills. Nonetheless, it is not clear that all participants were totally candid in their discussion of the intervention. Another study limitation is that we interviewed participants from three specific schools in two different settings. It is not known how well these findings might generalize to students in other settings. The study also focused on a specific implementation of the Second Step curriculum. It therefore cannot be assumed that the findings would generalize to other violence prevention curricula, to other implementations of Second Step, or to the version of Second Step (Committee for Children, 2008) that became available after we initiated this study. However, as previously noted, many of the skills targeted by the intervention are also addressed by other interventions, which suggests some generalizability. Finally, students interviewed in this study participated in the intervention as part of a larger study that randomized classrooms within each school to intervention and control conditions. The fact that some classrooms of sixth grade students and students in other grades did not receive the intervention may have reduced its effectiveness by not fully addressing peer influences and other factors operating within the school’s climate.

Interventions must ensure that adolescents master the skills they are taught, so that they can use them correctly and effectively in the real world. Research suggests that multiple-choice responses and knowledge tests may not capture adolescents’ abilities to use the skills in real-life situations (e.g., Howard et al., 1999). Other efforts may be needed to provide students with additional opportunities to apply the skills they learn within the intervention. This could include capitalizing on “teachable moments” in the classroom when interpersonal problems occur to apply emotion regulation, empathy, and problem-solving strategies. The challenges to using this strategy include teachers’ varying degrees of comfort and competency in observing these issues and skillfully and effectively helping adolescents to walk through a prosocial alternative to conflict. Despite the difficulty, this may be the best way for adolescents to master skills and understand their application to real-life problems.

Another important goal for future violence prevention programs is to optimize their relevance to adolescents. In the case of this implementation of Second Step, Sullivan and colleagues (2014) used prior qualitative studies of urban adolescents to identify difficult problem situations, and created a menu of potential role-plays and discussions based on those situations. Despite these efforts, some adolescents (14%) questioned the relevance of the intervention. Researchers should consider intervention development an iterative process in which adolescents’ concerns are incorporated as the intervention progresses. During the discussion of specific skills, adolescents are likely to raise objections to the content if it doesn’t appear relevant. For example, adolescents may feel that walking away from provocation is not effective (e.g., if the person follows them, or if that response hurts their reputation or social status). For change to occur, adolescents must buy into the idea that using the skills taught is better than their current behaviors. Interventionists should be prepared to identify current problem situations at the school or among particular adolescents, and discuss how intervention skills may relate to those situations. Interventionists should be prepared and able to involve adolescents in a discussion of the pros and cons of using a skill in a particular situation and whether it leads to more desirable outcomes than alternatives. This might include discussing situational factors that could impact how it is applied and influence its likelihood of success. This also highlights the need to better understand potential adaptations to prevention programs, for example, by using a problem situation generated by a class of students on the spot in lieu of following specific problem situations outlined in the lessons. There is some evidence that local adaptations to prevention programs that do not alter the integrity of the intervention may enhance youth outcomes (Berkel, Mauricio, Schoenfelder, & Sandler, 2011).

Mastering the skills taught by an intervention will not ensure that they are used unless adolescents are motivated to change their behavior. Interventions that teach socio-emotional skills might benefit by incorporating techniques that directly address adolescents’ underlying beliefs and values. For example, techniques such as motivational interviewing might help adolescents prioritize their goals and think through their conflicting values. Interventions such as Second Step, which focus primarily on individual-level factors, may also need to incorporate strategies that attempt to alter the overall school environment (e.g., Smith, Schneider, Smith, & Ananiadou, 2004). Peer factors, in particular, may discourage the use of skills taught by an intervention. Adolescents are unlikely to change their behavior if they believe it will result in negative peer reactions or damage their social standing (Farrell et al. 2010). Broader contextual factors such as bystander influences may also make it difficult for adolescents to use nonviolent skills in situations where they are confronted by peers. A key focus will be to alter not only the peer climate, but perceptions of the peer climate based on prior research that has found that adolescents tend to overestimate peer support for aggressive behavior and underestimate support for nonviolent responses (Henry et al., 2013). Changing the school culture will likely increase adolescents’ use of intervention skills if it supports prosocial conflict resolution, includes negative consequences for aggression and if peers provide models of prosocial behavior. Prior work has also supported the influence of parents, particularly the messages they may convey that actively support aggression (Farrell et al., 2010). In short, interventions that focus on promoting individual level skills may best be viewed as one part of a more comprehensive effort to address multiple factors that influence adolescents’ aggressive behavior (Farrell & Camou, 2006).

Although considerable progress has been made in developing interventions to reduce adolescents’ involvement in aggression, further efforts are needed to improve their effectiveness. Universal school-based interventions that focus on enhancing social and cognitive skills appear to have particular promise (Boxer & Dubow, 2002), but it cannot simply be assumed that participants in these interventions will successfully master these skills, use them, or find them effective when they do so. Adolescents encounter many challenges in their efforts to apply these skills outside of the classroom (Farrell et al., 2010). As such, they represent an important source of information that can help developers of interventions determine how best to address these challenges. This underscores the potential value of qualitative efforts such as the current study as part of the iterative process of improving the efficacy of violence prevention programs and enhancing their relevance for specific groups of adolescents.