C34 weakened
For language-minority children, school programmes that also teach in the home language give majority-language reading results about as good as teaching only in the majority language once the children have moved on to reading in that language, and have not been shown to lower it by a reliable amount from then on; while reading is still being taught in the home language, their majority-language reading lags behind (in the one long randomised trial, by about 0.4–0.5 standard deviations in kindergarten and grade 1, shrinking to mostly non-significant gaps of about 0.2 in grades 2–4; Slavin et al. 2010/2011).
Bounds: Language-minority children in school programmes; reading in the majority language as the main outcome. Evidence on file: Reljić, Ferring & Martin 2015 (Review of Educational Research, DOI 10.3102/0034654314548514; abstract only): European studies, bilingual over submersion on reading g = 0.23 (95% CI 0.10 to 0.36), but only five studies. Thayer 2021 (University of Nebraska dissertation; abstract, flagged as non-journal): mostly US Spanish–English; bilingual vs mainstream English g = 0.028 (95% CI −0.144 to 0.201). The claim says 'no harm', not 'better': a small gain is possible but not established. Weakened 2026-11-01 (Slavin, Madden, Calderón, Chamberlain & Hennessy 2011, Educational Evaluation and Policy Analysis 33(1), 47–58, DOI 10.3102/0162373711398127; numbers from the January 2010 report version of the same trial, full text, Best Evidence Encyclopedia / ERIC ED539723): Spanish-dominant children randomised within six US schools from kindergarten to transitional bilingual education or structured English immersion, same Success for All programme in both; English reading behind by an average effect size of −0.54 in kindergarten and −0.42 in grade 1 (significant on all English measures), then −0.20, −0.16 and −0.25 in grades 2–4 (one of nine comparisons significant; grade-4 sample 115); English vocabulary (PPVT) stayed lower in grades 2–4 (−0.17, −0.26, −0.29; the report's Table 3 marks only grade 3 as significant); Spanish reading higher in K–3. The early lag is a reliable reading disadvantage during the home-language phase, so 'do not lower reading by any reliable amount' and 'at least as good' no longer hold without a time bound; at the end point the trial's point estimates still lean towards English-only, so 'about as good' replaces 'at least as good'. Cross-cohort note (corrected 2026-11-07): grade rows are separate cohorts, not one panel (report p. 8), so the grade-4 row is the first cohort only (60 vs 55); pretest TVIP favoured the bilingual group in every row. The trial cannot rule out a small later English-reading lag (grade 3–4 point estimates −0.16/−0.25, ns). The published 2011 article is paywalled (papers-needed). Kill-condition test survived 2026-11-14: Valentino & Reardon 2015 (Educational Evaluation and Policy Analysis 37(4), 612–637, DOI 10.3102/0162373715573310; accepted manuscript read, ERIC ED566267): 13,750 English learners in one large US district who entered kindergarten 2001–02 to 2009–10, followed on the state English Language Arts test from grade 2 to grade 7–8, not randomised but with school fixed effects and controls for parents' programme preferences. Relative to English immersion, English-reading scores grew faster in transitional bilingual (+0.023 SD per grade) and dual immersion (+0.064; dual immersion started 0.19 lower in grade 2 and passed English immersion by grade 6), and not significantly differently in developmental (home-language maintenance) bilingual (−0.022, p < .10). Subgroup limit: for Chinese-speaking learners, developmental bilingual grew significantly more slowly (−0.068 SD per grade), while Latino learners in the same programme type grew faster (+0.056).
Attack record: 2 attacks: 1 survived, 1 narrowed · Evidence links: 7 linked sources (5 supports, 2 limits)