The effects of parent-implemented language interventions on child linguistic outcomes: A meta-analysis

25 randomised or quasi-randomised controlled trials, 1,734 children, mean age 3.7 years

Young children with or at risk for language impairment (mean age 3.7, mostly below this archive's 4-18 range), whose parents were trained to deliver a language protocol

Recorded here for the DELIVERY-AGENT question rather than for an age-relevant effect size: this is the largest body of trials in which parents were trained to deliver a specified instructional protocol, so it is the best available evidence on what a trained parent can and cannot achieve as an instructor. Effect sizes below are Hedges g as summarised in ASHA's structured review of the paper; the full text is closed and the business-as-usual versus clinician-delivered comparison moderator could not be obtained, so this source does NOT establish parent-delivered/clinician-delivered equivalence.

The training works on the parent and the parent works on the trained target. Parent use of language-facilitating behaviours moved g=1.20 — the largest effect in the meta-analysis is on the ADULT, not the child. Child effects: expressive vocabulary g=0.42 (significant), expressive language g=0.27 (small-to-moderate), receptive vocabulary g=0.18 (not significant), receptive language g=0.07 (not significant). By context, expressive vocabulary was g=0.37 for shared book reading and g=0.50 for play/routine-based interventions. Effects were moderate to large for children with or at risk for primary language impairment and smaller and non-significant for children with autism spectrum disorder. The shape is the archive's recurring pattern in a new setting: a trained parent moves the specific trained target and does not move the broader capacity.

Low — randomised trials, parents in both arms equally selected.

Updates and partly deflates Roberts & Kaiser 2011 on an enlarged and RCT-restricted study set; the direction agrees, the receptive-language claim does not survive.