Home · C31
C31 weakened
For children with special educational needs, being placed in mainstream inclusive classes rather than separate special settings has not been shown to change how fast their academic achievement grows: studies that follow pupils over time find about no difference (d = 0.05 for pupils with general learning difficulties), although at a single point in time pupils with general learning difficulties in inclusive classes score higher (d = 0.35 on average, 0.64 in propensity-matched comparisons; Krämer, Möller & Zimmermann 2021), which may partly reflect which pupils are placed where.
- Status
- weakened
- Last change
- 2026-10-31
- Serves
- Q1 (How do humans actually learn, and what helps them learn best?)
- Ages
- grundskola-lag, grundskola-hog, overgripande
- Area
- Special education and inclusion
- Relates to
- C31 relates-to C32 (special needs: where vs what is taught)
- Bounds
- Placement itself, compared across settings; not the quality of teaching in either setting and not wellbeing. Evidence on file: Dalgaard, Bondebjerg, Viinholt & Filges 2022 (Campbell Systematic Reviews, DOI 10.1002/cl2.1291; abstract only): no consistent positive or negative effect on language, literacy or maths; averages favoured inclusion slightly but none significant; non-randomised OECD studies. Stefanidis et al. 2023 (International Journal of Educational Research, DOI 10.1016/j.ijer.2023.102208; abstract only): co-taught classes gave no significantly higher achievement for students with or without disabilities. All evidence is non-randomised. Weakened 2026-10-31 (nightly attacks): Krämer, Möller & Zimmermann 2021 (Review of Educational Research 91(3), 432–478, DOI 10.3102/0034654321998072; full text): 40 non-randomised studies, N = 11,987; pupils with general learning difficulties (often IQ about 50–90) scored higher in inclusive settings, d = 0.35 (95% CI 0.18 to 0.52; 0.28 without outliers), maths and verbal alike; cross-sectional d = 0.48 vs longitudinal gains d = 0.05 (−0.16 to 0.25), a significant difference; propensity-matched d = 0.64 (4 samples), covariate-controlled 0.21 (ns), unmatched 0.29; peers without difficulties d = −0.14 (ns); psychosocial d = 0.00. Covers general learning difficulties only, not all special needs.
- What would kill this
- A meta-analysis of longitudinal or randomised comparisons finds a reliable academic advantage, or disadvantage, in achievement gains for children with special needs in inclusive rather than segregated placement, large enough to rule out a near-zero average (sharpened 2026-10-31 with the weaken: Krämer et al. 2021 already show a reliable level difference in cross-sectional comparisons). Promoted from K60 on 2026-10-29 in the bold expansion Jonatan approved after four empty attack nights (26–29 Oct): new sharp claims in thin areas, each naming a comparison, a rough size or direction and a population.
- Tutor consequence
- none yet
Attack record
1 attack: 1 narrowed
2026-10-31 narrowed
- Attack
- A meta-analysis of 40 studies (N = 11,987) found that pupils with general learning difficulties in inclusive settings had reliably higher academic results than comparable pupils in segregated settings (d = 0.35, 95% CI 0.18 to 0.52; 0.64 in propensity-matched samples), so inclusive placement does not leave their achievement unchanged "by a reliable amount on average".
- Target
- kill-condition
- Source
- Krämer, Möller & Zimmermann 2021 | Review of Educational Research 91(3), 432–478, DOI 10.3102/0034654321998072 (full text read 2026-10-31; Crossref-checked)
- Steelmans to
- C31's kill line asks for a meta-analysis of controlled comparisons with a reliable academic advantage large enough to rule out a near-zero average; this one gives a confidence interval well clear of zero, the same for maths and verbal outcomes, robust to outliers (0.28), and largest where samples were fully matched by propensity scores, which argues against pure selection.
- Why
- Nightly-attacks 2026-10-31 Europe/Stockholm. Full text: 40 non-randomised studies, 428 effect sizes; cognitive outcomes of pupils with GLD d = 0.35 (SE .09, CI 0.18 to 0.52; 16 samples), 0.28 without one or two outliers (Table 6). Design moderated it strongly: cross-sectional d = 0.48 (0.30 to 0.66) vs longitudinal gains d = 0.05 (−0.16 to 0.25), F(1, 159) = 30.68, p < .001 (Table 2). Matching: full propensity matching 0.64 (0.32 to 0.96; 4 samples), covariate-controlled 0.21 (−0.05 to 0.47; 7), unmatched 0.29 (5), groups not significantly different (p = .12). Peers without GLD −0.14 (ns); psychosocial 0.00. Kill withheld: all studies are non-randomised; the authors discuss selection; and the comparison closest to a causal test — gains over time in longitudinal studies — shows about zero, which is what C31's "on average" can still honestly say. The claim narrows from "neither raises nor lowers achievement" to "no shown difference in achievement growth", with the level difference stated. Population caveat: general learning difficulties only. Not a duplicate (empty ledger before tonight); not a restatement (Bounds rested on two abstracts finding no consistent effect). Not a preprint. Status alive → weakened. Claim field, Bounds and kill line rewritten in the same turn; brief and diagram core rewritten (C31 has no front-page rule).
- Rewrite
- claim-field+brief
Linked sources
4 linked sources (2 supports, 2 limits)
Supports (2)
Limits (2)
Review state: first-pass draft, pending human review.