For beginning readers at risk of reading difficulties, systematic phonics instruction (explicitly teaching the links between letters and sounds in a planned order) gives better word decoding than teaching with little or no phonics (such as whole-language or whole-word teaching); how much a planned sequence adds over "balanced" teaching that already covers some phonics now and then is an open question (open question 16), and a better reading comprehension is not established in randomised trials: in a UK government systematic review, four randomised trials comparing systematic phonics with whole-language or whole-word teaching gave a comprehension effect of about 0.24 standard deviations that was not statistically significant, while word-reading accuracy improved by about 0.27 (Torgerson, Brooks & Hall 2006), and a Cochrane review of phonics training for poor readers found only a small, uncertain comprehension gain (Torgerson et al. 2006; McArthur et al. 2018).
Status
weakened
Last change
2026-10-24
Serves
Q1 (How do humans actually learn, and what helps them learn best?)
Ages
grundskola-lag (early reading, roughly grades K–3; the Swedish trial is grade 2; Torgerson et al. 2006 trials mostly ages 5–7), overgripande (the National Reading Panel covered K–6)
Area
Reading and comprehension
Bounds
This claim is about how the alphabetic code is taught to beginning readers at risk (systematic phonics versus teaching with little or no phonics; the edge over "balanced" or incidental phonics is open question 16), on decoding and comprehension, not about later reading strategies, vocabulary or background knowledge, which skilled comprehension also needs (Scarborough 2001; Castles, Rastle & Nation 2018; Elwér 2014). Evidence in the project: National Reading Panel 2000 (meta-analyses: systematic phonics significantly more effective than little or no phonics in K–6 and for pupils with difficulties; strongest for decoding and spelling, comprehension gains clearest in the early years); Castles, Rastle & Nation 2018 (review: systematic phonics is needed to crack the alphabetic code but is not enough for expert reading); Lindström-Sandahl, Elwér, Samuelsson & Danielsson 2023 (randomised trial in Swedish grade 2, 68 pupils at risk: nine weeks of intensive one-to-one systematic phonics improved word recognition d = 1.04 [0.54, 1.53], decoding d = 0.54 [0.10, 0.97] and reading comprehension d = 0.50 [−0.02, 1.02] compared with teaching-as-usual controls, in semi-transparent Swedish spelling; comprehension CI includes ~0; control supplemental dose not measured; posttest only, no longer delayed follow-up; small, one-to-one). Context: Herkner 2022 (Swedish grade 3 national test missed pupils with decoding problems, who still struggled in grade 9; small follow-up); IFAU 2024:6 (Läslyftet, a teacher professional-development programme without a phonics focus, gave about 1.3 % of an SD on Swedish national tests, not significant in grade 3; not a phonics trial). Most of the evidence is English-language; Swedish spelling is more transparent, so dosage may differ. Comprehension bound (added 2026-10-16, extra run 3): Torgerson, Brooks & Hall 2006 (DfES Research Report RR711; systematic review of 20 randomised trials, children mostly aged 5–7, some up to 11): systematic phonics versus whole-language or whole-word teaching, word-reading accuracy d = 0.27 (fixed effects, 95 % CI 0.10 to 0.45; 12 trials; random effects 0.38, 0.02 to 0.73; without one outlier trial 0.20 fixed and 0.21 random, the latter not significant), no difference between normally developing children and children at risk; reading comprehension d = 0.24 (fixed, approximate CI −0.03 to 0.51, p = .08; 4 trials; random effects 0.35, not significant); weight of evidence judged weak for comprehension. McArthur et al. 2018 (Cochrane review, 14 randomised or quasi-randomised studies, 923 English-speaking poor readers, phonics training versus no training or non-literacy training): word-reading accuracy SMD 0.51 (0.13 to 0.90; 11 studies), reading comprehension SMD 0.28 (−0.07 to 0.62; 5 studies, 343 participants; low-quality evidence). Head-to-head test against explicit but unplanned phonics (kill-condition survived 2026-10-20; since the 2026-10-24 rewording this trial bears on open question 16 rather than the claim's core): in a randomised trial with at-risk first graders in six Texas schools, a systematic small-group programme with a fixed scope and sequence (Proactive Reading) gave better decoding of new words than an equally intensive programme that taught letter–sound links explicitly but without a planned order, using leveled books (Responsive Reading): WJ-III Word Attack difference 6.88 W-score points (SE 2.80, p < .05); word identification and passage comprehension did not differ (Mathes, Denton, Fletcher, Anthony, Francis & Schatschneider 2005; about 80 pupils per arm, 40 minutes a day from October to May).
What would kill this
Large-scale evidence (a meta-analysis or large randomised trial) in which teaching with little or no phonics matches or beats systematic phonics for at-risk beginning readers on word decoding (reworded 2026-10-24; the earlier wording also covered "balanced" teaching, which now sits in open question 16). Original bank wording (Swedish, verbatim, before the 2026-10-24 rewording): “Storskalig evidens att icke-systematisk/‘balanced’ avkodning utan phonics matchar systematisk phonics för at-risk early readers på decoding+comprehension.” Promoted from K05 on 2026-10-16 under the project's promotion rule.
Tutor consequence
none yet
Attack record
2 attacks: 1 survived, 1 narrowed
Reworded on 24 October 2026 (a change of wording, not an attack, so it is not counted). The evidence behind the core (National Reading Panel 2000; Torgerson, Brooks & Hall 2006; McArthur et al. 2018) compares systematic phonics with little or no phonics. The edge over "balanced" or incidental phonics is smaller and contested (Camilli et al. 2003/2006; Stuebing et al. 2008; Mathes et al. 2005), so it is held as open question 16 rather than as the claim's core. Kill-condition narrowed to match. No new numbers.
2026-10-16 narrowed
Attack
In a UK government systematic review of randomised trials, systematic phonics compared with whole-language or whole-word teaching improved word-reading accuracy (d = 0.27) but its effect on reading comprehension (d = 0.24, four trials) was not statistically significant, so C12's claim that systematic phonics gives better reading comprehension is not established by randomised evidence.
Target
claim-text
Source
Torgerson, Brooks & Hall 2006 | Department for Education and Skills Research Report RR711 (ISBN 1 84478 659 6)
Steelmans to
C12 asserts better decoding and comprehension; the review restricted to randomised trials of systematic phonics against whole-language or whole-word approaches — exactly C12's comparison — and found the comprehension pooled effect non-significant under both fixed (0.24, approximate CI −0.03 to 0.51, p = .08) and random (0.35, ns) effects, with the weight of evidence judged weak; the Cochrane review of phonics for poor readers (McArthur et al. 2018) points the same way (comprehension SMD 0.28, −0.07 to 0.62, low quality).
Why
Extra run 2026-10-16 (extra 3) Europe/Stockholm, ordered by Jonatan. C12's first turn (never targeted, the promotion rule). Full RR711 report read (PDF, not stored): 20 RCTs identified, all on initial reading teaching, children mostly 5–7; 14 compared systematic phonics with whole-language or whole-word approaches, 12 individually randomised and pooled: accuracy fixed d = 0.27 (0.10 to 0.45, p = .002), random d = 0.38 (0.02 to 0.73); sensitivity analysis without Umbach et al. 1989: fixed 0.20 (0.02 to 0.37, p = .03), random 0.21 (−0.03 to 0.44, p = .09); no significant difference between normally developing and at-risk children for accuracy. Comprehension: four trials (Leach & Siddall 1990; Lovett et al. 1989; Torgesen et al. 2001; Umbach et al. 1989), fixed 0.24, approximate 95 % CI −0.03 to 0.51, p = .08; random 0.35, ns. McArthur et al. 2018 (Cochrane CD009115.pub3, full text read via PMC6517252): poor readers, phonics vs no or non-literacy training; word-reading accuracy SMD 0.51 (0.13 to 0.90); comprehension SMD 0.28 (−0.07 to 0.62; 5 studies, 343). Kill withheld: decoding/accuracy still favours systematic phonics in both reviews, and the kill-condition requires a match on both. Limits: few, small and old comprehension trials; the reviews do not show phonics harms comprehension, only that a benefit is not established; the National Reading Panel (2000, on file) found comprehension gains clearest in the early years, from a broader set of studies including non-randomised ones. Not a preprint. First ledger entry on C12. Claim, Bounds, Ages and status rewritten same turn; brief and the practical rule updated.
Rewrite
claim-field+brief
2026-10-20 survived
Attack
In a randomised trial with at-risk first graders, a small-group programme that taught letter–sound links explicitly but without a planned order (Responsive Reading, guided-reading style with leveled, non-decodable books) was judged by its authors "essentially equally effective" as a systematic programme with a fixed scope and sequence (Proactive Reading), which would meet C12's kill-condition if it matched on both decoding and comprehension.
Target
kill-condition
Source
Mathes, Denton, Fletcher, Anthony, Francis & Schatschneider 2005 | Reading Research Quarterly 40(2), 148–182, DOI 10.1598/RRQ.40.2.2
Steelmans to
C12 dies where an approach without systematic phonics matches systematic phonics for at-risk beginners on decoding and comprehension; this trial ran exactly that head-to-head, with the same amount of extra teaching (40 minutes a day in groups of three, October to May), the same teachers' training hours and the same schools, against an alternative that still taught phonics explicitly — the strongest version of a non-systematic rival — and the authors' own summary was that the two were equally effective (overall average effect sizes .84 and .78 against enhanced classroom teaching).
Why
Nightly-attacks 2026-10-20 Europe/Stockholm. Full text read (published article, PDF hosted by Southern Methodist University; stored privately). Six schools in a large urban Texas district, two first-grade cohorts; at-risk pupils randomised within schools to Proactive (n = 92), Responsive (n = 92) or enhanced classroom teaching only (n = 114); 78, 83 and 91 assessed at posttest. End of year (Table 5): Proactive above Responsive on WJ-III Word Attack, the decoding measure (difference 6.88 W-score points, SE 2.80, t = 2.45, p < .05; effect sizes against enhanced classroom 0.76 and 0.28; only Proactive beat the classroom group on Word Attack); no significant difference on Word Identification (3.23, SE 3.33), WJ-III Passage Comprehension (−1.31, SE 2.40) or CRAB-R Comprehension (−0.46, SE 0.58); neither programme beat the classroom group on passage comprehension. So the rival matched on comprehension but not on decoding, and the kill-condition (match on both) failed. Limits: one trial, about 80 pupils per arm, first grade only, English; both programmes were intensive small-group extras on top of good classroom teaching in high-performing schools; Responsive was explicit about letter–sound links, so this tests the planned order, not phonics versus none. Not a preprint. C12's first survival (now 1 of 3); resets its fruitless streak.
Rewrite
none
Corrections to our notes
These fix a number or a fact in our notes. They are not attacks and do not count toward the floor of 3 survivals.
— From a closer reading of the study: Lindström-Sandahl et al. 2023 full text: comprehension d = 0.50 with 95% CI [−0.02, 1.02]; control was teaching-as-usual with unmeasured supplemental special education; posttest only.
Randomised, at-risk first graders, six Texas schools: a systematic programme with a fixed scope and sequence beat an equally intensive programme that taught phonics explicitly but without a planned order on decoding (WJ-III Word Attack, difference 6.88 W-score points, p < .05); word identification and comprehension did not differ. Kill-condition test survived 2026-10-20.
Re-analysis of the NRP studies: systematic phonics beat less systematic ("some") phonics by d = .24 or .19 depending on weighting, about half the NRP figure, on pooled reading outcomes; decoding and at-risk readers not separated. Held as open question 16 (2026-10-24), not a narrowing.
Multilevel re-analysis of the NRP studies: systematic versus "some" phonics d = 0.123, not significant (p = .183), on pooled reading outcomes (numbers from the authors' 2008 follow-up; rival re-estimate 0.183, p = .042). Held as open question 16 (2026-10-24), not a narrowing.
Cochrane review of phonics for poor readers: word-reading accuracy SMD 0.51; comprehension SMD 0.28 (−0.07 to 0.62, low quality). Cited with Torgerson et al. 2006 in the C12 narrowing 2026-10-16 (extra run 3).
Reading interventions with follow-ups (mean 11 months): phonics interventions' weighted effect fell from 0.29 at posttest to 0.07 at follow-up (reading skills 0.26 → 0.07), more than other intervention types, but the weighted figure leans on two large studies of typical readers, controls had ordinary teaching, and most at-risk phonics groups kept positive effects. Held as open question 25 (2026-11-09), not a narrowing.
Systematic review of RCTs: systematic phonics vs whole-language/whole-word: word-reading accuracy d = 0.27 (significant), comprehension d = 0.24 (4 trials, p = .08, not significant). Narrowed C12 2026-10-16 (extra run 3).
Review state: first-pass draft, pending human review.