Home · Themes · Metacognition and self-regulation
Metacognition and self-regulation
Generated view: 12 sources whose Kunskapsbank note was filed under this theme, and the claims and candidates they link to.
Claims
C15 weakened
Explaining newly learned material to others (actually teaching it, not just preparing to) improves learners' own understanding on a delayed test more than studying it for themselves when learners studied the material expecting to teach it; a lasting gain is not established for a short explanation added after ordinary study.
Serves: Q1
Ages: grundskola-hog, gymnasium, vuxen-livslangt
Bounds: The learning of the person who explains, compared with studying for themselves; outcome delayed tests of the explained material; not the effect on the tutee (C5, C10) or peer discussion of concept questions (candidate K38). Evidence on file: Kobayashi 2018 (Japanese Psychological Research; meta-analysis of 28 studies; Hedges' g 0.35 for preparing to teach and 0.56 for teaching after preparing; benefits also after a delay; larger for interactive teaching); Fiorella & Mayer 2013 (Contemporary Educational Psychology; actually teaching after preparing helped on a delayed test about one week later). Expectancy and durability bound (added 2026-10-25, nightly attack): In three classroom experiments with 1,074 German grade 7–8 physics students, 15 minutes of explaining a lesson by voice message to a fictitious peer, with the materials at hand, beat 15 minutes of restudying on a test right afterwards but not on a test eight weeks later (difference not significant; a later advantage only for students with low or medium interest in physics) (Lachner et al. 2025). Lachner, Russ, Hübner, Sibley & Scheiter 2025 (Educational Psychology Review 37, 88; open access, CC BY; full text read on the publisher's site): preregistered one-stage individual-participant analysis of three analogous classroom experiments (project GRAIL) with 1,074 German grade 7–8 students in 11 schools; after the same 45-minute curriculum lesson on the converging lens, students were randomly assigned to 15 minutes of non-interactive teaching (recording a voice message to a fictitious peer, Mia, who had stated two common misconceptions; materials allowed, not to be read aloud) or 15 minutes of restudying the materials with notes. Because assignment came after the lesson, the teaching group had not studied expecting to teach. Scientific knowledge (15-item ROC-CI test, α = .72): immediate posttest β = 0.23, p < .001; delayed posttest eight weeks later β = 0.09, p = .110 (not significant); interest in physics moderated the delayed effect (β = −0.18, p = .043; low- and medium-interest students gained, high-interest students did not). Of the three sub-studies, Study 1 found no long-term effect and Study 3 (under review) no effect even immediately. The authors attribute the missing delayed effect to the short, single, open-book session and leave open whether repeated or closed-book teaching would last. Kill withheld: C15's kill-condition is about teaching after preparing to teach, and these students had not studied expecting to teach; the delayed estimate points the same way as the claim and was not shown to be zero (no equivalence test); one topic, one 15-minute dose. The evidence for a delayed benefit on file comes from studies in which learners studied expecting to teach (Fiorella & Mayer 2013; Kobayashi 2018). Consistent with this, Kobayashi 2024 (Educational Psychology Review 36, 26; meta-analysis of 39 studies; abstract only, paywalled) reports that teaching after studying with teaching expectancy beat studying without it, g = 0.48 [0.34, 0.63], while teaching after studying without teaching expectancy did not, g = −0.02 [−0.14, 0.11]; whether those outcomes were delayed is not stated in the abstract.
Attack record: 1 attack: 1 narrowed · Evidence links: 9 linked sources (3 supports, 4 limits, 2 challenges)
1 link(s) from sources in this view.
C19 weakened
Teaching pupils in grades 1–8 (about ages 6–14) how to plan, monitor and check their own learning (metacognitive strategy instruction) raises their school performance compared with ordinary teaching in research studies, and a gain of roughly 0.3 standard deviations or more is still there at a follow-up test weeks or months later. It is not guaranteed at scale: in a large English trial with cascade-trained teachers, two years of a self-regulated writing programme did not raise writing, and pupils scored lower in reading, spelling and maths (Torgerson et al. 2018).
Serves: Q1 (How do humans actually learn, and what helps them learn best?)
Ages: grundskola-lag, grundskola-hog, gymnasium
Bounds: Teaching learners strategies for planning, monitoring and evaluating their own learning, with school performance as the outcome and a later follow-up; not retrieval practice itself (C0), spacing/interleaving schedules (C11), or feedback type (C4). Evidence on file: de Boer, Donker, Kostons & van der Werf 2018 (Educational Research Review 24, 98–115, DOI 10.1016/j.edurev.2018.03.002; full text stored from University of Groningen Pure): meta-analysis of 48 metacognitive strategy-instruction interventions in school contexts with follow-up at least three weeks after the programme (mean 21.6 weeks, range 3–108); posttest Hedges' g = 0.50 (SE 0.07), follow-up g = 0.63 (SE 0.07); post-to-follow-up difference g = 0.12 (SE 0.04, 95% CI 0.05–0.19). Kill-condition test survived 2026-10-27 (de Boer et al. 2018): the follow-up gain was not gone and was slightly larger than the posttest effect. Scope note (mining 2026-10-27): the meta's school sample is grades 1–8 only (higher secondary not represented); trim-and-fill for publication bias lowers the post–follow-up difference to g = 0.09 (SE 0.04) but it stays significant. Kill-condition test survived 2026-10-27 extra5 (Eberhart, Ingendahl & Bryce 2025, Metacognition and Learning 20:7, DOI 10.1007/s11409-024-09405-x; OA full text read): multi-level meta of 67 metacognition interventions in typically developing pre- and elementary children (≤12 years; 349 effect sizes); immediate g = 0.48 (95% CI 0.35–0.61); follow-up (19 studies, mean 15 weeks, range 3–48) g = 0.29 (SE 0.06, 95% CI 0.17–0.40); Language at follow-up g = 0.45; Metacognition-alone follow-up ns (g = 0.12); age band younger than gymnasium; funnel asymmetry at follow-up. Kill-condition test survived 2026-10-27 extra11 (Tuero, Núñez, Vallejo, Fernández, Añón, Moreira, Martins & Rosário 2022, Frontiers in Psychology 13:889201, DOI 10.3389/fpsyg.2022.889201; Frontiers OA full text): classroom-randomised school-based SRL (Rainbow Program; planning–execution–evaluation) vs standard curriculum; 761 grade 3–4 pupils in 50 classes / 14 schools; teacher-rated academic scores (science, Spanish, English, maths) at post-test and 3-month follow-up; combined subjects post-test d = 0.25, follow-up d = 0.33; per-subject follow-up significant for science (d = 0.31), Spanish (d = 0.35) and English (d = 0.28), maths ns (d = 0.11). Dignath & Büttner 2008 (full text stored 2026-10-27 extra11): posttest academic ES primary 0.61 / secondary 0.54; no follow-up durability pool — Bounds support only, not a kill-condition survival. Scale bound (added 2026-11-15, nightly attack): Torgerson et al. 2018 (IPEELL: Evaluation report and executive summary, Education Endowment Foundation; independent evaluation by Durham University and the University of York; full report read): two pragmatic school-randomised trials of IPEELL, an English adaptation of Self-Regulated Strategy Development (planning, drafting, self-monitoring, goal setting and evaluating one's writing) bundled with 'memorable experiences', in 84 primary schools in Leeds and Lincolnshire (2015–2017), teachers trained for one day by newly recruited trainers (train-the-trainers), control = usual teaching. One year in Year 6 (2,465 pupils analysed): KS2 writing g = −0.09 (95% CI −0.34 to 0.15); KS2 reading −0.23 (−0.43 to −0.04). Two years in Years 5–6 (2,196 analysed; about 20% missing): independently administered, blind-marked writing test g = 0.11 (−0.13 to 0.34), low prior attainers 0.26 (−0.01 to 0.53); KS2 reading −0.17 (−0.33 to −0.01), maths −0.30 (−0.53 to −0.07), spelling, punctuation and grammar −0.28 (−0.49 to −0.06). Outcomes taken at the end of the programme, not at a later follow-up, so the kill-condition was not tested. The earlier efficacy trial of the same programme, delivered by its developers to low-attaining pupils at the primary–secondary transition, had found an effect size of about 0.75 (Torgerson et al. 2014, as cited in the 2018 report; +0.74 in the EEF summary). The evaluators point to weak training (inexperienced trainers, little modelling) and possible loss of curriculum time from other subjects; 59.5% of intervention schools were rated as implementing well; exploratory complier estimate for two years 0.19 (−0.23 to 0.60).
Attack record: 4 attacks: 3 survived, 1 narrowed · Evidence links: 6 linked sources (4 supports, 2 limits)
6 link(s) from sources in this view.
C29 weakened
Training preschool children's executive functions (ages 3–6) raises their scores on executive-function tests by about a third of a standard deviation, but has not been shown to raise learning-related or academic outcomes reliably compared with control groups: one meta-analysis found no significant extra effect on learning and behaviour outcomes pooled together (g = 0.17, not significant; Scionti et al. 2020), while another found literacy and numeracy gains of about 0.4 standard deviations in 4- to 6-year-olds right after training, resting on very few studies per outcome and including non-randomised ones (Birtwistle et al. 2025).
Serves: Q1 (How do humans actually learn, and what helps them learn best?)
Ages: foralder-tidig-barndom
Bounds: Stand-alone cognitive training for ages 3–6 (computer or game-based tasks); not whole curricula such as Tools of the Mind and not physical activity. Evidence on file: Scionti, Cavallero, Zogmaister & Marzocchi 2020 (Frontiers in Psychology 10, 2812, DOI 10.3389/fpsyg.2019.02812; full text read 2026-11-05, open access, PMC6965160): 32 studies (randomised or pre-post with an active or passive control, at least 10 sessions), 123 effect sizes; executive functions g = 0.352; larger for at-risk children. Non-EF outcomes (learning-related, behaviour and problem-behaviour measures pooled together): 39 outcomes from 15 studies, g = 0.169 (SE 0.106, 95% CI −0.047 to 0.383, p = .122); the discussion gives g = 0.10 for the same analysis, an inconsistency in the paper; academic outcomes were not pooled on their own. Melby-Lervåg & Hulme 2013 (Developmental Psychology; already in the library): working-memory training gave near but not far transfer (adjacent support). Weakened 2026-10-31 (nightly attacks): Birtwistle, Chernikova, Wünsch & Niklas 2025 (SAGE Open 15(1), DOI 10.1177/21582440241311060; full text, open access): 57 published RCTs and quasi-experiments of EF training (at least 80 minutes) in typically developing children aged 3–14, immediate post-tests only; ages 4–6 (29 studies): near transfer g = 0.25, far transfer g = 0.34 (p = .05), literacy g = 0.40 (df = 4), numeracy g = 0.39 (df = 5), reasoning ns; I² = 84%. Melby-Lervåg & Hulme 2013 (full text, corrected 2026-10-31): working-memory training at all ages, mean 12 hours; immediate word decoding d = 0.13 (95% CI −0.07 to 0.35) and arithmetic d = 0.07 (−0.13 to 0.27), both non-significant; follow-ups 3–8 months later all non-significant (Table 4). The 'no academic gain' side rests on Scionti et al.'s pooled non-EF estimate, which mixes learning-related and behaviour outcomes (full text read 2026-11-05).
Attack record: 1 attack: 1 narrowed · Evidence links: 4 linked sources (3 supports, 1 limits)
3 link(s) from sources in this view.
C30 killed
Music lessons have not been shown to make children reliably better at thinking skills or school subjects outside music: compared with an active control group the gain is at most small — about zero in the largest meta-analysis's main analysis (g = 0.056; Sala & Gobet 2020), about 0.2 standard deviations when the comparisons in which the control group trained the very skill being tested are dropped (pooled with passive-control studies; Bigand & Tillmann 2022), and 0.28 against active controls for instrumental programmes, with a confidence interval that includes zero (Román-Caballero et al. 2022).
Serves: Q1 (How do humans actually learn, and what helps them learn best?)
Ages: foralder-tidig-barndom, grundskola-lag
Bounds: Music training for children, non-music outcomes (verbal, non-verbal, memory, literacy, maths). Evidence on file: Sala & Gobet 2020 (Memory & Cognition 48, DOI 10.3758/s13421-020-01060-2; full text): 54 studies, 254 effects; active controls g = 0.056 (95% CI −0.069 to 0.182), −0.021 in sensitivity analysis; randomised studies with non-active controls g = 0.064; no difference between cognitive and academic outcomes. Challenge already on file: Bigand & Tillmann 2022 (Memory & Cognition, DOI 10.3758/s13421-021-01226-6; accepted manuscript): reanalysing the same data and dropping 21 comparisons where the control trained the very skill tested gave g = 0.208 — the first attack to weigh. Weakened 2026-10-31 (nightly attacks): Bigand & Tillmann 2022 re-read in the published version (Memory & Cognition 50(2), 339–347): the 21 removed comparisons were 18 phonological-training controls tested on phonology, 2 reading-intervention controls tested on reading and 1 visual-art control tested on visual form; the remaining 51 studies (233 effects) gave g = 0.208 (RVE, 95% CI 0.121 to 0.295; RE 0.195), pooled over active and passive controls, with type of control no longer a significant moderator (p = .193); without post-test-only studies g = 0.243. Román-Caballero, Vadillo, Trainor & Lupiáñez 2022 (Educational Research Review 35, 100436, DOI 10.1016/j.edurev.2022.100436; full text): instrumental programmes, 34 samples, 5,998 children, gΔ = 0.26 (0.12 to 0.39); randomised 0.26; active controls 0.28 (−0.06 to 0.61), randomised with active controls 0.23 (−0.13 to 0.58); one bias correction (RVE PET) cut the overall estimate to 0.01. Weakened 2026-11-05 (nightly attacks): Jamey, Foster, Hyde & Dalla Bella 2024 (Cognition 252, 105913, DOI 10.1016/j.cognition.2024.105913; full text, authors' copy): 22 longitudinal studies with controls, 1,734 children aged 3–11 (mean about 5 years 8 months); the 8 trials randomised at participant level with active non-music control training gave g = 0.60 (SE 0.11, 95% CI 0.39 to 0.82; I² = 0%; robust to rho, 0.55–0.60) on inhibition (flanker, go/no-go, Stroop or preschool versions); all 22 studies g = 0.31 (0.19 to 0.44); Egger p > .33. Limits: one inhibition measure per study, picked by the authors as the most suitable; post-tests right after short programmes; no blinding, every study rated 'moderate' quality; the authors note that playing music itself uses inhibition, so near and far transfer blur. Killed 2026-11-05 (Jonatan's decision): inhibitory control is a non-music cognitive result, and the 8 active-controlled randomised trials in Jamey et al. 2024 put it at g = 0.60 with an interval excluding zero, so the kill line below was met as written. The night run's rewrite of the claim and kill line (naming inhibitory control as an exception) was withdrawn; this is the wording that was attacked. The narrower idea that survives is now its own claim, C35.
Attack record: 2 attacks: 1 narrowed, 1 killed · Evidence links: 4 linked sources (1 supports, 2 limits, 1 challenges)
1 link(s) from sources in this view.
Music lessons do not reliably raise children's school results or broad thinking-skill measures more than active non-music activities do; the one shown exception is inhibitory control (g = 0.60; Jamey et al. 2024).
Serves: Q1 (How do humans actually learn, and what helps them learn best?)
Ages: foralder-tidig-barndom, grundskola-lag
Bounds: Children, music training compared with active non-music control activities; outcomes are school results (reading, maths, literacy, grades, standardised tests) and broad cognitive measures (IQ, general or composite reasoning, verbal or non-verbal ability, memory composites). Not about a single executive function picked per study, and not about music skills themselves. Created 2026-11-05 on Jonatan's decision after C30 was killed on its old kill line; every size below is carried over from C30's record. Sala & Gobet 2020 (Memory & Cognition 48, DOI 10.3758/s13421-020-01060-2; full text): 54 studies, 254 effects; active controls g = 0.056 (95% CI −0.069 to 0.182), −0.021 in sensitivity analysis; no difference between cognitive and academic outcomes. Bigand & Tillmann 2022 (Memory & Cognition 50(2), 339–347, DOI 10.3758/s13421-021-01226-6; published version): dropping 21 comparisons where the control trained the very skill tested left 51 studies (233 effects), g = 0.208 (95% CI 0.121 to 0.295), pooled over active and passive controls; without post-test-only studies 0.243 — the strongest challenge on file, but not an active-control-only estimate. Román-Caballero, Vadillo, Trainor & Lupiáñez 2022 (Educational Research Review 35, 100436, DOI 10.1016/j.edurev.2022.100436; full text): instrumental programmes, 34 samples, 5,998 children, gΔ = 0.26 (0.12 to 0.39) overall; active controls 0.28 (−0.06 to 0.61) and randomised with active controls 0.23 (−0.13 to 0.58), both pooled over all outcome types (Table 2). The paper gives no estimate for school results or broad cognitive measures against active controls alone: its literacy (0.20, 0.06 to 0.34), mathematics (0.23, −0.41 to 0.87) and intelligence (0.29, −0.00 to 0.58) figures pool active and passive controls (Table 3). Bias corrections: over randomised and non-randomised studies one correction (RVE PET) cut the estimate to 0.01; in randomised studies only, PET gave −0.04 (p = .881) and the selection model 0.15 (p = .207), while PEESE (0.18) and trim-and-fill (0.19 and 0.25) stayed near 0.2, and no test for publication bias was significant (Tables 4–5). Of the randomised studies with active controls, the authors cite the two with school-result outcomes (D'Souza & Wiseheart 2018; Haywood et al. 2015, 269 vs 279 pupils) as null results and two with intelligence outcomes (Schellenberg 2004; James et al. 2020) as positive (p. 2; Table 1). Outcomes measured at programme end only; no follow-up (corrected 2026-11-11). Exception: Jamey, Foster, Hyde & Dalla Bella 2024 (Cognition 252, 105913, DOI 10.1016/j.cognition.2024.105913; full text): 8 participant-randomised trials with active non-music controls, inhibition g = 0.60 (95% CI 0.39 to 0.82; I² = 0%); all 22 studies (1,734 children aged 3–11) g = 0.31; one measure per study picked by the authors, immediate post-tests, no blinding, and no academic or broad cognitive outcome.
Attack record: No attacks yet · Evidence links: 6 linked sources (2 supports, 2 limits, 2 challenges)
1 link(s) from sources in this view.
Candidates (not approved)
- K52: When a learning aid or lesson addition makes practice easier or more engaging, learners' own judgement of how much they learned or how well they did does not follow their unaided test results: in randomised comparisons they rated themselves the same or higher while their measured performance was unchanged or lower. (1 link(s))
Sources