Mindfulness-based interventions in schools—a systematic review and meta-analysis

This is a meta-analysis, meaning the authors pooled results from 24 separate studies of school-based mindfulness interventions. The interventions varied widely but all involved teaching mindfulness meditation or mindfulness-based practices to students during school hours. Typical programs included:

• Mindfulness exercises: Breathing meditation, body scans, mindful movement (yoga), and attention-to-present-moment practices.
• Formal practice: Sessions ranging from 5–45 minutes, 1–5 times per week, for 4–24 weeks.
• Informal practice: Bringing mindful attention to everyday activities like eating, walking, or listening.

Comparators: Most studies used a control group that received either:

• No intervention (waitlist or business-as-usual)
• An active control (e.g., relaxation training, social-emotional learning programs, or physical education)

Outcome measures were grouped into four domains:

1. Cognitive performance: Attention, concentration, executive function, working memory, academic achievement.
2. Stress: Self-reported stress, anxiety, physiological stress markers (e.g., cortisol).
3. Resilience: Coping skills, emotional regulation, adaptability, positive affect.
4. Emotional problems: Depression, internalising symptoms, behavioural problems.
5. Third-person ratings: Teacher or parent reports of student behaviour, attention, or emotional state.

• Total participants: 2,224 students (1,348 received mindfulness instruction; 876 served as controls).
• Age range: Grade 1 through grade 12 (approximately ages 6–18).
• Settings: Schools in multiple countries (primarily Western, including the US, UK, Canada, Australia, and several European nations).
• Diversity: Studies included mainstream classrooms, special education settings, and schools in low-income or high-stress communities. Some studies focused on students with pre-existing anxiety or attention difficulties.
• Gender: Not consistently reported across studies, but generally balanced where reported.
• Exclusions: Not uniformly reported; some studies excluded students with diagnosed psychiatric disorders or those on medication.

Important caveat: The meta-analysis includes both published (13) and unpublished (11) studies. Unpublished studies tend to have smaller or null effects, so their inclusion makes the overall estimate more conservative and realistic.

The meta-analysis did not use a single instrument but aggregated results across many different measurement tools. Here are the most common instruments used in the included studies:

Cognitive performance:

• Test of Everyday Attention for Children (TEA-Ch)
• d2 Test of Attention (measures processing speed and concentration)
• Continuous Performance Test (CPT)
• Stroop test (inhibitory control)
• Working memory span tasks
• Teacher-rated academic performance

Stress:

• Perceived Stress Scale (PSS, 0–40 scale, higher = more stress)
• State-Trait Anxiety Inventory for Children (STAI-C)
• Cortisol levels (salivary samples)
• Self-report stress questionnaires (various)

Resilience:

• Resilience Scale for Children (various versions)
• Coping strategies questionnaires
• Positive and Negative Affect Schedule (PANAS-C)
• Self-efficacy scales

Emotional problems:

• Children's Depression Inventory (CDI)
• Strengths and Difficulties Questionnaire (SDQ)
• Internalising/Externalising subscales of the Child Behavior Checklist (CBCL)

Third-person ratings:

• Teacher ratings of classroom behaviour (e.g., Conners' Rating Scales)
• Parent reports of emotional and behavioural functioning

Meta-analytic approach: The authors calculated Hedge's g (a standardised effect size corrected for small sample bias) for each study, then pooled them using random-effects models (which assume true effects vary across studies—appropriate given the heterogeneity). They also assessed publication bias using funnel plots and Egger's test.

Study design: Systematic review and meta-analysis of 24 studies (19 controlled, 5 uncontrolled pre-post designs). The authors searched 12 databases in August 2012, plus hand-searched references and contacted experts for unpublished data.

Inclusion criteria:

• Mindfulness-based intervention delivered in a school setting (K–12)
• At least one quantitative outcome measure (psychological, cognitive, or behavioural)
• Published or unpublished (to reduce publication bias)
• Any language (though searches were in English)

Exclusion criteria:

• Studies where mindfulness was only one component of a multi-component program (unless mindfulness was the core)
• Studies with only physiological outcomes (e.g., heart rate without psychological measures)
• Case studies or qualitative-only reports

Data extraction: Two independent reviewers extracted data on study characteristics, intervention details, outcomes, and effect sizes. Discrepancies were resolved by consensus.

Quality assessment: The authors assessed study quality using a custom checklist (not a standardised tool like the Cochrane Risk of Bias), which is a limitation. They rated studies on: randomisation, control group, blinding, attrition, and outcome measurement.

Statistical approach:

• Random-effects meta-analysis (DerSimonian-Laird method)
• Hedge's g as the effect size metric
• Heterogeneity assessed with I² statistic (percentage of variation due to true differences rather than chance)
• Subgroup analyses by age group, intervention duration, and outcome domain
• Sensitivity analyses removing one study at a time

What this design can and cannot prove:

Can prove:

• The overall direction and magnitude of effects across multiple studies
• Whether effects are consistent or vary by outcome domain
• Whether certain study features (e.g., longer duration, older students) are associated with larger effects
• The presence or absence of publication bias

Cannot prove:

• Causality (the included studies were mostly RCTs, but the meta-analysis itself is observational—it summarises existing evidence)
• Which specific mindfulness program is best (too much heterogeneity)
• Long-term effects beyond the study endpoints (most studies were short-term)
• Mechanisms of action (why mindfulness works, if it does)
• Safety or adverse effects (rarely reported in the included studies)

Major methodological weaknesses flagged by the authors:

• High heterogeneity (I² values not fully reported but described as "great")
• Many studies underpowered (small sample sizes)
• Wide variety of interventions, durations, and outcome measures
• Lack of active control groups in many studies (waitlist controls inflate effects)
• Poor reporting of attrition and fidelity (whether teachers actually delivered the program as intended)
• No blinding of participants or teachers (impossible in mindfulness studies, but still a source of bias)
• Most studies short-term (4–12 weeks), with no follow-up beyond intervention end

Overall effect (primary outcome):

• Between-group effect size: Hedge's g = 0.40 (p < 0.0001)
• Within-group effect size: Hedge's g = 0.41 (p < 0.0001)
• Interpretation: Mindfulness groups improved by about 0.4 standard deviations more than controls—a small-to-moderate effect (Cohen's convention: 0.2 = small, 0.5 = medium, 0.8 = large)

Domain-specific effects (between-group):

• Cognitive performance: g = 0.80 (p < 0.05) — large effect

  • This was the strongest finding. Mindfulness training improved attention, concentration, and executive function substantially.

• Stress: g = 0.39 (p < 0.05) — small-to-moderate effect

  • Students reported less perceived stress and anxiety after mindfulness training.

• Resilience: g = 0.36 (p < 0.05) — small-to-moderate effect

  • Improvements in coping, emotional regulation, and positive affect.

• Emotional problems: g = 0.19 (p = not significant) — small, non-significant effect

  • Depression and internalising symptoms did not improve reliably.

• Third-person ratings: g = 0.25 (p = not significant) — small, non-significant effect

  • Teachers and parents did not consistently notice changes in student behaviour.

Moderator analyses (what made effects larger or smaller):

• Age: Effects tended to be larger for older students (adolescents) compared to younger children, but this was not formally tested.
• Intervention duration: No clear relationship between program length and effect size—shorter programs (4–8 weeks) sometimes worked as well as longer ones.
• Control group type: Studies with waitlist controls showed larger effects than those with active controls (e.g., relaxation training), suggesting some effects may be due to expectation or attention rather than mindfulness specifically.
• Publication status: Published studies showed larger effects than unpublished studies, confirming publication bias (studies with null results are less likely to be published).

Heterogeneity:

• The authors report "great heterogeneity" across studies, meaning the true effects vary considerably. Some studies showed large positive effects, others showed no effect or even negative trends. This limits confidence in the overall estimate.

Publication bias:

• Funnel plot inspection suggested asymmetry, indicating that small studies with negative results may be missing from the literature. The authors note this as a concern.

Let's translate these numbers into plain English:

Cognitive performance (g = 0.80): This is a large effect. To put it in perspective, if the average student in the control group was at the 50th percentile on an attention test, the average mindfulness-trained student would be at about the 79th percentile. That's roughly the difference between a student who can focus for 10 minutes versus 18 minutes on a sustained attention task. In classroom terms, this might mean fewer errors on a d2 attention test, better working memory for instructions, or improved ability to resist distractions.

Stress reduction (g = 0.39): A small-to-moderate effect. If the control group's average stress score was 20/40 on the Perceived Stress Scale, the mindfulness group would average about 17–18/40. That's a meaningful but not transformative reduction—like going from "often stressed" to "sometimes stressed." In real-world terms, a student might report feeling less anxious before a test or fewer physical stress symptoms (headaches, stomach aches).

Resilience (g = 0.36): Similar magnitude to stress. A student might report using more active coping strategies (e.g., problem-solving, seeking support) and fewer avoidant strategies (e.g., withdrawal, denial). This could translate to bouncing back faster from a bad grade or a social conflict.

Emotional problems (g = 0.19, non-significant): This is a very small effect that could easily be due to chance. If you're hoping mindfulness will cure depression or reduce serious behavioural problems, this meta-analysis suggests it's unlikely to do so on its own. The confidence interval likely includes zero.

Third-person ratings (g = 0.25, non-significant): Teachers and parents didn't reliably notice changes. This could mean the effects are subtle (not visible to outside observers) or that the measures used (e.g., behaviour checklists) are not sensitive enough to capture real changes.

What the authors acknowledge:

• High heterogeneity across studies—different populations, interventions, durations, and outcome measures make it hard to draw firm conclusions
• Many studies underpowered (small sample sizes), increasing the risk of false positives and inflated effect sizes
• Publication bias—unpublished studies with null results were included, but the funnel plot still suggested missing studies
• Lack of standardised mindfulness curricula—programs varied so much that "mindfulness" may mean different things in different studies
• Short follow-up periods—most studies measured outcomes immediately post-intervention, with no data on whether effects lasted weeks or months later
• No assessment of adverse effects—none of the included studies reported whether any students had negative reactions (e.g., increased anxiety, dissociation, boredom)

Critical reader additions:

• No blinding: Participants and teachers knew they were in the mindfulness group. This creates expectation effects—students may report less stress simply because they think they should. Active control groups (e.g., relaxation training) partially address this, but only a few studies used them.
• Waitlist controls inflate effects: Many studies compared mindfulness to "business as usual" or a waitlist. Students in the control group knew they weren't getting the "special" program, which can create resentment or demoralisation, making the mindfulness group look better by comparison.
• Teacher effects: In many studies, classroom teachers delivered the mindfulness program. Teachers who volunteer to teach mindfulness may be more enthusiastic, skilled, or attentive than average, and this enthusiasm (not mindfulness itself) could drive effects.
• Outcome reporting bias: Studies measured many outcomes but often only reported those that were significant. The meta-analysis tried to address this by contacting authors for unpublished data, but it's still a concern.
• Age range too broad: Combining 6-year-olds and 18-year-olds is problematic—their cognitive abilities, attention spans, and understanding of mindfulness are vastly different. The overall g = 0.40 may not apply equally to all ages.
• No long-term data: Even if effects are real, we don't know if they persist. A student might show improved attention immediately after an 8-week program, but does that last 6 months later? The data don't answer this.
• Dose-response not examined: Some studies had 5 minutes of daily practice, others had 45 minutes weekly. The meta-analysis didn't systematically test whether more practice produced larger effects.
• Cultural and socioeconomic factors: Most studies were in Western, relatively affluent settings. Effects may differ in other cultural contexts or in schools with high poverty, trauma exposure, or limited resources.

For someone running their own n=1 experiment (e.g., a student, teacher, or parent trying mindfulness):

What to test

• Specific intervention: A structured mindfulness program with daily practice. Based on the studies in this meta-analysis, a good starting point is:

  • Core practices: 5–15 minutes of breath-focused meditation, body scan (paying attention to sensations in each body part), and mindful listening (e.g., focusing on a bell sound until it fades).
  • Frequency: Daily practice, 5–15 minutes per session.
  • Duration: At least 4 weeks, ideally 8 weeks (most studies in the meta-analysis were 4–12 weeks).
  • Format: Guided audio recordings or in-person instruction (many studies used apps or CDs for home practice).

• What NOT to test: Don't try to treat depression or serious emotional problems with mindfulness alone—the evidence for these outcomes was weak. Focus on attention, focus, and stress reduction.

Minimum meaningful duration

• 4 weeks is the shortest duration that showed effects in the meta-analysis.
• 8 weeks is more common and likely more robust (many studies used 8-week programs).
• Daily practice matters more than total program length—a 4-week program with daily practice may outperform an 8-week program with weekly practice.

What to measure (specific metrics)

Primary outcomes (most likely to show change):

1. Cognitive performance:

   • Sustained attention: Use a free online test like the d2 Test of Attention (or a simplified version: count how many times you can spot a target letter in a grid of distractors over 2 minutes). Measure errors and speed.
   • Working memory: Use a digit span test (repeat increasingly long sequences of numbers forward and backward). Free apps like "Brain Workshop" or "Dual N-Back" can measure this.
   • Executive function: The Stroop test (name the colour of ink, not the word—e.g., the word "RED" printed in blue ink). Measure reaction time and errors.

2. Stress:

   • **Per