C20 Interactive book reading
narrowed 4 attacks: 3 survived, 1 narrowed · serves Q1 · last change 27 October 2026
Reading picture books with young children in an interactive way (asking them questions and getting them to talk about the story) grows their spoken vocabulary more than no shared reading or a passive control in several meta-analyses, but the edge over ordinary reading aloud is not clearly shown when studies are pooled (Noble et al. 2019), and shared-reading language gains are near zero against active non-language controls.
In practice
When sharing picture books with young children, ask them questions and get them to talk about the story. Interactive / shared reading raised spoken vocabulary more than no shared reading or a passive control in several metas (Mol, Bus & de Jong 2009 oral language d = .54; Dowdall et al. 2020 RCT-only caregiver meta expressive d = 0.41, receptive d = 0.26), and parent Dialogic Reading beat reading-as-usual on vocabulary in a dedicated meta (Mol, Bus, de Jong & Smeets 2008 overall d = .42, expressive d = .59; smaller/ns for kindergarten and at-risk families). The pooled edge over ordinary reading aloud is still not clearly shown across broader shared-reading interventions (Noble et al. 2019 dialogic moderator ns), and gains are near zero against active non-language controls.
Go deeper
Where it holds
Main limits
- About oral vocabulary in early childhood, not decoding (C12) or topic knowledge for comprehension (C17).
- Mol 2009 full text: oral language d = .54; Dialogic Reading by teachers d = −.08 (ns) vs experimenters d = .58.
- Mol 2008 parent Dialogic Reading vs reading-as-usual: overall d = .42, expressive d = .59; kindergarten and at-risk ns.
- Dowdall et al. 2020 caregiver RCT meta: expressive d = 0.41, receptive d = 0.26; low-intensity (<60 min) ns.
- Noble et al. 2019: overall g = 0.194; active controls g = 0.028; follow-up ns.
All limits (7)
Inside the circle: the core of the claim. Outside: every bound and carve-out in the claim file, numbered, with the attack that forced or added it where the record shows one.
forced by a narrowing attackin the original wordingwritten at creation, or added without a narrowing
Warm cards in the list were added by later attacks; plain cards were in the original draft or written at creation.
Interactive shared reading (dialogic / interactive book reading) versus ordinary reading aloud, passive control, or no shared reading, with spoken vocabulary / oral language as the outcome, for young children.
Written when the claim was created (27 October 2026); not forced by an attack
Kill-condition test survived 2026-10-27 (Mol, Bus & de Jong 2009, Review of Educational Research 79(2), 979–1007, DOI 10.3102/0034654309332561; full text from Leiden thesis ch.4): 31 (quasi-)experiments, oral language d = .54 (95% CI [.33, .74]); expressive vocabulary d = .62; receptive d = .45; control with no intervention d = .34, with non-language intervention d = .95; teacher-delivery limit from the same full text — Dialogic Reading by teachers d = −.08 (ns), by experimenters d = .58, teacher–small group d = .15 (ns).
Added with attack #1 + #3 (survived; it did not narrow the claim) · Mol, Bus & de Jong 2009; Mol, Bus, de Jong & Smeets 2008 · 27 October 2026
Kill-condition test survived 2026-10-27 extra8 on the ordinary-reading arm (Mol, Bus, de Jong & Smeets 2008, Early Education and Development 19(1), 7–26, DOI 10.1080/10409280701838603; Leiden thesis ch.3 OA): parent Dialogic Reading vs reading-as-usual; overall vocabulary d = .42 [.30, .53] (k = 16, N = 626); expressive d = .59 [.44, .75] (k = 9); receptive d = .22 [.05, .39]; preschool d = .50 vs kindergarten d = .14 (ns); not-at-risk d = .53 vs at-risk d = .13 (ns).
Added with attack #1 + #3 (survived; it did not narrow the claim) · Mol, Bus & de Jong 2009; Mol, Bus, de Jong & Smeets 2008 · 27 October 2026
Kill-condition test survived 2026-10-27 extra9 on the no-shared-reading / control arm (Dowdall, Melendez-Torres, Murray, Gardner, Hartford & Cooper 2020, Child Development 91(2), e383–e399, DOI 10.1111/cdev.13225; CentAUR OA accepted MS): RCT-only meta of caregiver-directed interactive / dialogic picture-book sharing, 19 RCTs, N = 2,594; expressive language d = 0.41 [0.20, 0.61] (16 studies, 35 ES, N = 1,664); receptive d = 0.26 [0.12, 0.40] (16 studies, 24 ES, N = 1,871); caregiver competence d = 1.01; dosage limit — medium–high intensity expressive d = 0.54 vs low intensity (<60 min) d = 0.19 (ns), receptive low-intensity d = 0.09 (ns).
Added with attack #4 (survived; it did not narrow the claim) · Dowdall, Melendez-Torres, Murray, Gardner, Hartford & Cooper 2020 · 27 October 2026
Weakened 2026-10-27 (Noble, Sala, Peter, Lingwood, Rowland, Gobet & Pine 2019, Educational Research Review 28, 100290, DOI 10.1016/j.edurev.2019.100290; OA full text stored): shared-reading interventions overall g = 0.194; dialogic-reading moderator ns (p = .579 / .370); active controls g = 0.028 (p = .703); follow-up g = 0.139 (ns).
Forced by attack #2 · Noble, Sala, Peter, Lingwood, Rowland, Gobet & Pine 2019 · 27 October 2026
Whitehurst et al. 1988 (record verified; numbers not yet read).
Written when the claim was created (27 October 2026); not forced by an attack
Not decoding (C12) and not topic knowledge for comprehension (C17).
Written when the claim was created (27 October 2026); not forced by an attack
1 of 7 bounds and carve-outs were forced by a specific attack. 3 attacks hit the boundary and did not move it: #1, #3, #4.
How it changed
The ledger records 4 attacks, but no earlier wording of C20 is on record, so the text cannot be replayed step by step. Only the current wording is shown.
Reading picture books with young children in an interactive way (asking them questions and getting them to talk about the story) grows their spoken vocabulary more than no shared reading or a passive control in several meta-analyses, but the edge over ordinary reading aloud is not clearly shown when studies are pooled (Noble et al. 2019), and shared-reading language gains are near zero against active non-language controls.
What would change this
What would kill it: A large randomised trial or meta-analysis finds that interactive book reading gives no more vocabulary growth than ordinary reading aloud or no shared reading, including when children's own teachers or parents deliver it. Added as a claim on 27 October 2026, promoted from a candidate belief in the evidence library. Uncovered area: Early childhood and language.
Attack record
4 attacks: 3 survived, 1 narrowed. Counts of attacks, not a score.
Full attack list (4)
A meta-analysis of interactive book-reading experiments in early education would meet C20's kill-condition if interactive reading gave no more oral-language / vocabulary growth than control (ordinary programme or no shared reading); Mol, Bus & de Jong 2009 found oral language d = .54 (95% CI [.33, .74]) across 31 studies.
Strongest reading and verdict reasoning
Steelmans to: C20 dies if a large randomised trial or meta-analysis finds interactive book reading gives no more vocabulary growth than ordinary reading aloud or no shared reading; this is the main classroom interactive-reading meta named in the claim's own Bounds and kill sentence.
Why: Extra7 attack pass 2026-10-27 Europe/Stockholm (parent: never-targeted C20 first honest OA kill-condition test). Full text read and stored (Leiden Scholarly Publications thesis chapter based on the RER article). 31 (quasi-)experiments; oral language d = .54 [.33, .74]; expressive vocabulary d = .62; receptive d = .45; fail-safe N for oral language 1,724. Kill fails: interactive reading beat control on oral language. Limits logged in Bounds same turn (not a separate weaken): DR by teachers d = −.08 (ns) vs experimenters d = .58; teacher–small group d = .15 (ns); authors state experimenter findings were not simply replicable in natural classrooms. Control-type split: no-intervention control d = .34; non-language intervention control d = .95. Not a preprint. Claim sentence unchanged on this entry; Status still alive until the Noble weaken below. First survival (1 of 3).
- #2 narrowed 27 October 2026 · Noble, Sala, Peter, Lingwood, Rowland, Gobet & Pine 2019 · target: claim-text
A later meta-analysis of shared book-reading interventions would undercut C20's claim that interactive reading grows vocabulary more than ordinary reading aloud if dialogic techniques were not a significant moderator and if shared-reading language gains vanished against active controls; Noble et al. 2019 found dialogic moderator ns (p = .579 / .370) and active-control g = 0.028 (p = .703) against overall g = 0.194.
Strongest reading and verdict reasoning
Steelmans to: C20 asserts interactive reading beats ordinary reading aloud or no shared reading; this OA meta pools shared-reading interventions, tests dialogic as a moderator of language outcomes, and splits passive vs active controls — the strongest available OA challenge to the ordinary-reading conjunct and to a general no-shared-reading advantage.
Why: Extra7 attack pass 2026-10-27 Europe/Stockholm (second applied attack; strongest honest challenge after the Mol survival). Full text read and stored (Hope HIRA OA PDF). After excluding influential cases: overall g = 0.194 [0.088, 0.299], p = .002; passive controls g = 0.263; active controls g = 0.028, p = .703; dialogic reading moderator not significant; follow-up g = 0.139, p = .200. Kill withheld: Mol 2009 still shows a moderate oral-language advantage vs control, and Noble's active-control null is about shared reading generally, not a pure interactive-vs-ordinary head-to-head. Narrowed instead: keep a passive/no-shared-reading advantage from metas including Mol; drop a clear pooled edge of interactive over ordinary reading aloud; note active-control near-zero. Not a restatement of Mol's teacher-delivery limit. Not a preprint. Status alive → weakened. Claim and Bounds rewritten same turn; brief and the practical rule updated.
A meta-analysis that pits Dialogic Reading against reading-as-usual would meet C20's kill-condition on the ordinary-reading-aloud arm if interactive reading gave no more vocabulary growth than ordinary shared reading; Mol, Bus, de Jong & Smeets 2008 found overall vocabulary d = .42 (95% CI [.30, .53]) and expressive d = .59 [.44, .75] across 16 parent–child studies.
Strongest reading and verdict reasoning
Steelmans to: C20 dies if a large randomised trial or meta-analysis finds interactive book reading gives no more vocabulary growth than ordinary reading aloud or no shared reading; this is a dedicated parent Dialogic Reading vs reading-as-usual vocabulary meta named in the kill sentence's ordinary-reading arm, distinct from Mol, Bus & de Jong 2009 (classroom interactive vs control) and from Noble et al. 2019 (shared-reading pool with ns dialogic moderator).
Why: Extra8 attack pass 2026-10-27 Europe/Stockholm (parent: C20 at 1/3 after Mol 2009 survive + Noble 2019 weaken; find ONE more honest OA kill-condition survival). Full text read and stored (Leiden Scholarly Publications thesis chapter based on the EED article). 16 studies, N = 626; overall d = .42 [.30, .53]; expressive d = .59 [.44, .75]; receptive d = .22 [.05, .39]. Kill fails on the ordinary-reading arm: Dialogic Reading beat reading-as-usual. Limits logged in Bounds same turn (not a separate weaken): kindergarten d = .14 ns; at-risk d = .13 ns; parents only (teachers/strangers excluded). Does not undo Noble 2019's ns dialogic moderator in a broader shared-reading intervention pool — claim sentence and Status (weakened) unchanged. Not a preprint. Not a restatement of Mol 2009 (different contrast, different sample). Second survival (2 of 3).
- #4 survived 27 October 2026 · Dowdall, Melendez-Torres, Murray, Gardner, Hartford & Cooper 2020 · target: kill-condition
An RCT-only meta-analysis of caregiver-directed interactive picture-book sharing interventions would meet C20's kill-condition on the no-shared-reading / control arm if interactive book-sharing gave no more language growth than control; Dowdall, Melendez-Torres, Murray, Gardner, Hartford & Cooper 2020 found expressive language d = 0.41 (95% CI [0.20, 0.61]) and receptive d = 0.26 [0.12, 0.40] across 19 RCTs (N = 2,594).
Strongest reading and verdict reasoning
Steelmans to: C20 dies if a large randomised trial or meta-analysis finds interactive book reading gives no more vocabulary growth than ordinary reading aloud or no shared reading; this is the first RCT-only meta of caregiver interactive / dialogic book-sharing vs control for child language, distinct from Mol, Bus & de Jong 2009 (classroom interactive vs control, mixed designs) and from Mol, Bus, de Jong & Smeets 2008 (parent Dialogic Reading vs reading-as-usual).
Why: Extra9 attack pass 2026-10-27 Europe/Stockholm (parent: C20 at 2/3 after Mol 2009 + Mol 2008 survivals and Noble 2019 weaken; find ONE more honest OA kill-condition survival that is not Flack 2018 same-force). Full text read and stored (CentAUR accepted manuscript PDF). 19 RCTs, N = 2,594; expressive d = 0.41 [0.20, 0.61], p < .001 (16 studies, 35 ES, N = 1,664); receptive d = 0.26 [0.12, 0.40], p = .001 (16 studies, 24 ES, N = 1,871); caregiver competence d = 1.01. Kill fails on the control arm: interactive caregiver book-sharing beat control on expressive and receptive language. Limits logged in Bounds same turn (not a separate weaken): low-intensity interventions (<60 min contact) expressive d = 0.19 ns, receptive d = 0.09 ns; few long-term follow-ups; controls mix passive and active non-book-sharing. Does not undo Noble 2019's ns dialogic moderator in a broader shared-reading pool — claim sentence and Status (weakened) unchanged. Not a preprint. Not a restatement of Mol 2009 (classroom, mixed designs) or Mol 2008 (vs reading-as-usual). Third survival (3 of 3). Floor reached → the promotion rule promotion same turn.
Evidence
5 linked sources: 4 supports, 1 limits (first-pass draft, pending human review). supports: Mol et al. 2009, Whitehurst et al. 1988, Mol et al. 2008, Dowdall et al. 2020 · limits: Noble et al. 2019