Skip to content

Bullying Prevention and Intervention

Principle · All principles
Evidence · 31 claims (23 for, 7 mixed, 1 against) · 18 studies (6 causal, 5 review, 4 associational, 2 design, 1 theoretical), q1–q3 · 4 of 18 report an effect size · 31 claims rest on one study

Conditional relationship

For school-age learners (child, adolescent, pre-k-12) in a shared school setting (classroom and the unstructured spaces around it: recess, corridors, buses, group chats), bullying is expected to fall, and fear of it to ease, when the whole school acts on it over a term-plus horizon: every adult uses one definition (a power imbalance plus repetition, in verbal, relational, physical and online forms), the adults set and keep a positive, supervised climate in every setting, classes rehearse what bystanders can do (the indirect roles of assistant, reinforcer, outsider and defender, not only the bully–target pair), learners have several ways to report that do not depend on being seen to report, and incidents get a planned response that protects the target. That is expected to do more than a response aimed only at the pair after an incident, or a single assembly or poster campaign. The outcomes are behavioural (bullying perpetration and victimisation, as rated by teachers or reported by students), affective (fear of bullying, feeling safe) and, further downstream, attendance and graduation. The target for learners is attitude-motivation and a complex-skill: recognising bullying when it happens and choosing a safe move as a bystander.

The relationship is conditional on breadth and duration (the controlled evidence comes from programmes run schoolwide for two to four years), on the learners' age (bullying and peer rejection rose across elementary school even in schools that slowed the rise), on who reports the outcome (teacher ratings, student self-reports and incident records count different things, and many bullied students tell no adult), and on whether the setting is in school or online. No claim in the wiki tests a programme built around the participant roles against one aimed only at the bully–target pair, and no claim in the wiki tests a cyberbullying-specific response. What the wiki has is (a) a group-randomised trial of a schoolwide positive behaviour framework that slowed the rise in teacher-rated bullying and peer rejection over four years, (b) a randomised experiment of an adult-organised recess programme in which teachers reported less recess bullying, (c) a quasi-experiment, reported in a review, of a restorative whole-school approach with less self-reported bullying after two years, (d) a small play-test of a role-play game that raised children's confidence to defend a peer, and (e) a Chicago study linking schools that reduce fear of bullying to higher graduation. So the role-focused design below is a default, labelled step by step.

How converted siblings sit beside this: the Positive Behavioral Interventions and Supports pattern holds the tiered schoolwide behaviour sequence (taught expectations, acknowledgement, data review, small-group and individual support); this page uses it as the climate layer and does not repeat it. The Social-Emotional Learning pattern holds the explicit teaching of named social and emotional skills; defending, empathy for a target and asking for help are taught here with that sequence. Character Education holds the schoolwide, non-punitive, conflict-resolution framing of discipline that bullying-specific work builds on. Effective Classroom Management Plan holds the classroom-wide plan for disruption. Building Empathy holds structured perspective-taking.

Default design, while the relationship is untested

The earlier page's guidance, kept as a concrete default a designer can act on and revise. Each step is a proposal unless a claim is linked beside it; quantities are untested proposals.

  1. Commit the whole school, for at least two years. Choose the scope as a school or a grade band, not one class; agree the plan with every adult who supervises learners, including lunch, recess, bus and support staff. Schoolwide frameworks are where the controlled evidence is: in a group-randomised trial in 37 elementary schools, a schoolwide positive behaviour framework slowed the rise in teacher-rated bullying over four years (SWPBIS and bullying [+M]); a restorative whole-school approach in four Hong Kong secondary schools was followed by less self-reported bullying after two years where it was fully implemented (RWSA [+M]). Whether the bullying-specific steps below add to a general climate framework is untested.
  2. Give every adult one definition and one set of cases. Bullying is a power imbalance plus repetition; a single conflict between equals is a behaviour problem but not bullying. In a 45–60 minute staff session before the year starts (a proposal), sort short cases (verbal, relational exclusion and rumour, physical, online) into bullying or not, and agree the first response to each. Include relational cases on purpose: teachers in one survey rated relational aggression as less serious than verbal aggression, and the severity a teacher sees predicted how likely they were to react (perceived severity predicts reaction [+M], associational; verbal rated more serious than relational [+M], descriptive). Treat a learner as possibly in more than one role (bullying in one place, targeted in another), not as fixed "bully" or "victim"; untested, from the earlier page.
  3. Supervise and organise the unstructured times. Map where incidents happen (recess, corridors, toilets, buses, group chats), put an adult where they cluster, and give recess an organised, inclusive structure with games an adult or trained older learner runs. In a randomised experiment, teachers in schools with an adult-organised recess programme reported less bullying and exclusion at recess than control-school teachers (Playworks, recess bullying [+M]).
  4. Teach and rehearse the bystander roles in class. A proposed dose: one 30–40 minute session a week for 8–10 weeks in the first term, then a 20-minute booster each month. In each session, show a short scenario, name the roles present (who started it, who joined in, who laughed, who watched, who helped; laughing and watching reinforce the behaviour by example, Social Learning Theory), and have learners choose and rehearse a safe defending move: comforting the target afterwards, getting an adult, telling the person to stop, or bringing the target away. In a play-test with 18 children aged 9–10, a robot-mediated role-play game raised perceived self-efficacy for defending, Cohen's d = 0.74 (REMind, self-efficacy [+M]), and children explained a bystander's inaction mainly as low confidence or fear of retaliation (REMind, reflections [+M]); it did not significantly change their expectation that defending would help the target feel better (REMind, victim expectations [~M]). Rehearse comforting as well as confronting: confronting is the riskier move for a child who is afraid of the bully. Whether rehearsal changes actual bystander behaviour is untested in the wiki.
  5. Open several reporting routes, including ones nobody else sees. A named adult in each grade, an anonymous written or online route, and a way for a bystander to report for someone else. Tell learners what happens after a report, and report back to the person who reported. Untested proposal; from the earlier page, which reported that about 40% of bullied students do not tell an adult (2012 national survey data, not on a claim page here), so incident counts undercount.
  6. Respond to an incident with a fixed sequence that protects the target. Stop it; meet the target and the learner who bullied separately; record what happened and where; agree a safety plan with the target (where they sit, who they can go to); give the learner who bullied a consequence and a chance to repair; tell both families; check in with the target after a week and after a month. Do not use a joint mediation as the default. A restorative meeting between the two is possible only with a trained adult facilitator and the target's free agreement, and the target may still be uncomfortable facing the bully (restorative responses need trained facilitators [~W]). Do not put learners identified as bullying others together in one group for intervention (peer contagion, Adolescent Peer Group Structure); untested proposal from the earlier page.
  7. Handle cyberbullying as its own case. It can follow a learner home at any hour, can be copied and kept, may be anonymous, and happens on platforms adults do not see. Add to step 6: help the target save evidence (screenshots with dates) before blocking; report to the platform; agree with families who contacts whom; act when it affects school even if it happened off site, within the school's policy. Untested proposal; from the earlier page. For learners who may miss social cues online, see step 9.
  8. Keep the target's wider risk in view. Being bullied is linked to depression, anxiety, school avoidance and, with other risk factors, suicide. Ask about mood and attendance when a learner is targeted and refer to counselling where needed; do not present bullying as the single cause of a suicide in messages to learners, which overstates causality and hides other risks. Untested proposal; from the earlier page. Anxiety and fear from bullying are argued to take attention away from learning (toxic stress impedes learning [+W], narrative synthesis).
  9. Teach learners at higher risk explicitly. For learners with autism or other learning differences, teach what bullying looks like (including online), when to tell, and whom, with simple, frequent conversations and clear online rules agreed with the family (Frequent conversations and clear limits). Youth with an IEP report bullying at higher rates than peers (IEP youth, bullying [+M], associational), and adolescents with ASD are described as frequent targets of cyberbullying because they may miss social cues and lack self-advocacy skills (ASD and cyberbullying [+W], an expert brief without cited data). Transgender students report bullying and fear of violence at more than twice the rate of peers (transgender students [+M], associational).
  10. Plan the arc across sessions and years, and measure separately. Early (first 6–8 weeks): adults lead, scenarios are fictional or played by puppets, robots or characters, and the adult names the roles. Middle (rest of the first term): learners bring anonymised scenarios from their own school life and the class decides the safe move; the adult only adds what is missed. Late (from the second term): learners lead short scenarios for younger classes or new arrivals, and boosters run monthly. Move a class on when most learners can name the roles in a new scenario and give a safe move without prompting (an untested threshold). Each year, record student-reported bullying and fear of bullying, teacher-reported incidents and where they happened, separately and by group, at the start and end; the trials here counted teacher ratings or student reports, not both. Expect rates to rise with age even where the programme works: in the four-year trial, bullying and rejection rose in both conditions (rise across elementary school [~M]), so compare with where they would otherwise be, not with last year.

When a condition is not met: if the school will not act schoolwide, run steps 4–6 in the classes you control, say the reach is smaller, and supervise your own unstructured times; if there is no time for a weekly series, keep step 6 (the response) and step 5 (reporting) before anything else.

Fitting the design to a situation

The three facts that most change the decision:

  • How far the plan reaches and for how long. One class, one school or a district; one session, a term or several years. If unstated, ask: which adults supervise these learners across the day, will they all respond in the same way, and for how many years can the plan run?
  • Who the learners are and where the bullying happens. Primary or secondary age; in class, at recess, on the way home or online; whether some groups (learners with an IEP or autism, transgender or LGBTQ learners) are targeted more. If unstated, ask: where and when do incidents happen, and who reports them?
  • Which outcome counts, and who reports it. Fewer incidents recorded, fewer students reporting being bullied, less fear, more defending, better attendance or graduation. If unstated, ask: what would tell you it worked in a year, and from whose report?
If the brief says… Then change… Basis
Whole school, primary age (child), goal is less bullying over several years Adopt a schoolwide positive behaviour framework as the climate layer, keep it for four years, and track teacher-rated bullying and peer rejection each year against the expected rise SWPBIS and bullying [+M]; SWPBIS and peer rejection [+M] (group-randomised trial, 37 Maryland elementary schools, teacher ratings)
Secondary age (adolescent), goal is less bullying and more empathy Implement a restorative whole-school approach in full (not partly), for two years, and measure student-reported bullying, empathy and self-esteem RWSA [+M] (quasi-experiment, grades 7–9, four Hong Kong schools, reported in a review)
Incidents cluster at recess or lunch Organise recess with adult- or older-learner-led inclusive games and more adults outside; record incidents by place and time Playworks, recess bullying [+M] (randomised experiment, teacher reports, no effect size printed)
Learners watch bullying but do not act, saying they are not confident or are afraid Rehearse defending in role-play with comforting and getting help as options beside confronting; begin with fictional or robot characters REMind, self-efficacy [+M] and REMind, reflections [+M] (18 children aged 9–10, one pre/post play-test, no control); untested for actual defending
The school wants to bring the two learners together to resolve it Meet them separately first; use a restorative meeting only with a trained facilitator and the target's free agreement, and offer another route restorative responses need trained facilitators [~W] (narrative review)
Staff disagree on what counts, or ignore exclusion and rumours Run the staff case-sorting session with relational cases; agree one first response per case perceived severity predicts reaction [+M] and verbal rated more serious than relational [~M] (Romanian teachers, survey; associational and descriptive)
Learners with an IEP or autism Teach what bullying is, when and whom to tell, with frequent short conversations and agreed online rules with the family; check in with them directly rather than waiting for reports IEP youth, bullying [+M] (national survey, associational); ASD and cyberbullying [+W] (expert brief)
Transgender, LGBTQ or racially minoritised learners report more bullying Break every count down by group; add staff training and positive and restorative discipline schoolwide; check whether the gap closes transgender students [+M] (California survey, associational); gaps converged after SWPRD [+M] (one school, reported in a review, no statistics)
Bullying is online or on phones (online, phone-only) Add the cyberbullying steps: save evidence, report to the platform, agree family contact, act when it affects school; teach these to learners as a short routine untested proposal, from the earlier page
Learners join remotely while others are in the room Give remote learners a private written reporting route and a named adult; run bystander scenarios with chat and polls so remote learners choose moves too; watch the class chat as a supervised space untested proposal
One adult for a large group, or scarce facilitator time (volunteers, a counsellor with a short slot) Keep the response sequence (step 6) and an anonymous reporting route first; use trained older learners to lead recess games and short scenarios; review reports weekly in 15 minutes untested proposal
A single session or a fixed short slot (for example 45 minutes) Do not call it a programme or expect less bullying: 10 minutes on the definition and roles, 20 minutes rehearsing two defending moves in one scenario, 10 minutes on how and where to report, 5 minutes for anonymous questions untested proposal; the controlled evidence comes from two- to four-year schoolwide programmes
Learners have low confidence or low literacy Rehearse moves by acting them out, not by writing; use pictures or puppets for scenarios; accept a spoken or drawn report untested proposal; the REMind reflections point to low confidence as a reason for inaction (REMind, reflections [~M]), carried from 9–10-year-olds
Adults in a work setting (staff, apprentices, a workplace) Use the same definition and roles with cases from their own workplace; the organisation's own policy is the response sequence; reporting routes outside the line manager untested proposal; every claim here is about school-age learners
The brief names its own programme, survey, reporting form or behaviour policy Keep it as given; change only what the evidence here says matters: schoolwide and multi-year reach, supervised unstructured times, rehearsed defending, and counts by group untested proposal (wave-5 rule), with the claims in the rows above
Reports are private or sensitive (named learners, sexual or identity-based bullying) The target decides who is told and whether it is discussed; scenarios in class are anonymised or fictional; records are kept by the named adult only untested proposal
A student makes a threat of violence, or a target shows signs of depression or suicidal thinking Refer to the school's threat-assessment or mental-health route at once; prefer counselling and a parent conference to exclusion; do not treat bullying as the whole explanation CSTAG [+M] (randomised studies of a threat-assessment model, reported in a review; its entries report counselling and suspension, not bullying); shooters not bullied loners [~M] (review)
Goal is attendance or graduation in high school Track students' fear of bullying and the school's supportive environment as leading measures; give most attention to students who were afraid of bullying on arrival fear reduction and graduation [+M]; larger for previously victimised students [+M]; supportive environment [+M] (one Chicago study; the last is associational)
Learners make a product for a real audience (an anti-bullying campaign, a performance for younger classes) Keep the product whole and real, and pair it with rehearsed defending and the reporting routes; do not count a campaign alone as prevention untested proposal

Eleven of the nineteen rows cite a claim. Every claim rests on one study; four (SWPBIS twice, Playworks, CSTAG) come from randomised designs, one from a quasi-experiment reported second-hand, one from an 18-child play-test without a control, and the rest are associational or descriptive surveys, reviews or an expert brief. Rows on online bullying, mixed modes, scarce time, a single session, adults at work, private reports and public products are untested proposals.

Observation, state and explanation

What can be observed: incidents recorded by staff (what, where, when, who was present), students' answers to a survey about being bullied, bullying others and fear of bullying, teacher ratings, what learners choose and say in a rehearsed scenario, who uses each reporting route, attendance, and what happens after a report. "Is a bully", "is a defender", "the school is safe" are inferred states. A learner who names the right defending move in class has shown what they say in that room, not what they do on the bus.

Response Plausible explanations Observation that could change the interpretation
Recorded incidents rise after the programme starts More bullying; more reporting because routes are open and trusted; staff now count relational cases they used to ignore Compare anonymous student-reported victimisation with incident records over the same period; check which reporting routes the new reports came through.
Incidents fall, students' fear of bullying does not Bullying moved where adults do not see (online, off site); learners stopped reporting; fear lags behaviour Ask about online and off-site bullying in the survey; ask learners whether they would report now and why.
Learners rehearse defending well but bystanders still do nothing Fear of retaliation; the bully has high social status; the move rehearsed does not fit the real setting Ask, privately, what stopped them; rehearse comforting and getting help, which carry less risk, and record which moves learners later use.
One group (for example learners with an IEP, transgender learners) does not improve Bullying of that group is identity-based and the programme did not name it; that group reports less; different places Break counts down by group and place; ask learners in that group, in private, where and how it happens.
A targeted learner's attendance and mood fall even after the bullying stops Lasting effects of what happened; other risk factors (depression, family stress); the bullying continues online Check in with the learner and family; refer to counselling; ask specifically about online contact.

These rows are proposals, untested with learners as diagnostics. Keep the explanations that remain open rather than choosing the one whose name fits; a survey or probe can itself change what learners report.

Evidence and qualifications

No claim in the wiki tests the role-focused programme against one aimed only at the bully–target pair, a cyberbullying-specific response, or a reporting system. The claims below test schoolwide programmes against comparison schools, or describe who is bullied.

Learners, outcomes and reporters differ across these claims (teacher ratings in Maryland elementary schools, teacher reports of recess, student reports in Hong Kong secondary schools, a single-item self-efficacy scale, graduation records in Chicago), so do not rank them by effect label. No claim measures bystander behaviour, cyberbullying or reporting rates as an outcome.

The evidence by question

Further evidence, not yet read against this model

Objective and learner-valued goal

The designer's objective is usually fewer learners bullied, less fear, and, downstream, attendance and graduation. Elicit separately what learners want: to be left alone, to have a friend at recess, not to be seen as a "snitch", for the adults to act when told, to stop something happening to a friend. Agreement and divergence are both observations. A learner who values not being seen to report will judge the programme by whether an anonymous route exists and is acted on; a learner who wants the adults to act will judge it by what happened after their last report.

For example, a school may aim at fewer recorded incidents, while a target wants the group chat that excludes them to stop. Recorded incidents may rise as relational and online cases begin to be counted; record both what the school counted and whether the learner says the exclusion stopped.

What would revise this model?

The expectation should weaken if schoolwide programmes that add role-focused bystander rehearsal and reporting routes show no less bullying than the same climate framework without them, or if rehearsal raises confidence to defend without changing what bystanders do. If joint restorative meetings with trained facilitators are found to leave targets as safe as separate meetings, step 6's default should change. If cyberbullying falls with the same in-school programme and no specific steps, the separate cyberbullying routine is a preference, not a requirement. If teacher ratings fall while students' own reports do not, the model is changing what adults see, not what learners experience. Do not protect the model by calling every null "not schoolwide enough" or "not long enough" after the fact.

Bullying rated by teachers, bullying reported by students, fear of bullying, bystander behaviour, reporting, online bullying, and attendance or graduation are separate claims. The present evidence establishes no dose for bystander sessions, no reporting design and no criterion for moving a class on.

Claims

Examples

Key Sources

  • U.S. Centers for Disease Control and Prevention. StopBullying.gov.
  • Arduini-Van Hoose, N. (2020). Peer relationships. In Educational psychology. Retrieved from https://edpsych.pressbooks.sunycreate.cloud. CC BY-NC-SA 4.0.
Page metadata
Field Value
type principle
id bullying-prevention-and-intervention
title Bullying Prevention and Intervention
description For school-age learners, a schoolwide plan kept for two or more years (one definition shared by all adults, supervised and organised unstructured times, rehearsed bystander and defender moves, several reporting routes including anonymous ones, and an incident response that protects the target, with separate steps for cyberbullying) is expected to reduce bullying and fear of it more than a response aimed only at the bully and target after incidents; trials of schoolwide behaviour, recess and restorative programmes test schoolwide action against comparison schools, but no claim tests the role-focused design or a cyberbullying response.
status review
canonical True
generated by claude/unspecified on 2026-10-15
trust unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.