What a teacher can and must not do in a kindergarten with no speech therapist

The starting point: trained non-specialist staff can deliver effective oral language intervention, and this is the strongest evidence in the field. Fricke, Bowyer-Crane, Haley, Hulme and Snowling (2013) randomly allocated 180 children from 15 UK nursery schools (12 per setting, mean age 4;0) to a 30-week oral language intervention or a waiting control. The programme was delivered by nursery staff and teaching assistants, trained and supported by the research team. Result: oral language and spoken narrative improved significantly both immediately and after six months; phonological awareness improved clearly; the gains generalised to a standardised measure of reading comprehension. The 2017 follow-up with 394 children: the 30-week programme d = 0.30, the 20-week version d = 0.21, both maintained at six-month follow-up — but neither produced statistically significant improvements in early word reading or reading comprehension.

Duration is decisive. The same research group published a cautionary tale: Haley, Hulme, Bowyer-Crane, Snowling and Fricke (2017) took the eight children with the poorest standardised language scores in each of 13 nursery schools (N = 104, mean age 3;11) and ran a 15-week oral language programme. The result: significant differences on taught vocabulary only. There were no group differences on any standardised language measure. Fifteen weeks was not enough; 20–30 weeks was. And a programme can move the words you taught without moving general language. Does it work at scale? In England the Nuffield Early Language Intervention was rolled out to over 10,000 schools; a regression-discontinuity analysis of 19,936 children, comparing those just above and just below the allocation threshold, found a significant language advantage — Hedges g = 0.40. Two-year follow-up: oral language d = 0.22 (d = 0.33 for children with lower initial language ability), reading comprehension d = 0.16.

What a teacher legitimately can do. First, run a structured, manualised 20–30 week oral language programme in small groups, delivered by trained staff with ongoing support and fidelity monitoring. Second, use a validated parent-completed screen (ASQ-3) at defined ages — but record the result as a screen result, never as a diagnosis. Third, conduct and document structured observation: does the child follow a two-step instruction; is their speech intelligible to an unfamiliar adult; do they initiate with peers; have previously used words been lost. Fourth, coach parents — the Cochrane review found no significant difference between clinician-delivered and trained-parent-delivered intervention, and Roberts and Kaiser (2011), across 18 studies, found significant effects of parent-implemented interventions (g from −0.15 to 0.82). Fifth — and this is the cheapest and strongest single move — deliberately pair children with weaker language with typically developing peers in structured activities: this was the largest effect in the Cochrane review, SMD 2.29 (95% CI 1.11–3.48).

What a teacher must not do. Must not diagnose. No teacher should write, say or record "developmental language disorder", "autism", "dyslalia" or the local equivalents. The precision of the instruments does not support it: in a meta-analysis of M-CHAT-R/F covering 15 studies and 49,841 children, the pooled positive predictive value was 57.7%, and 51.2% in low-risk samples, with pooled specificity of 45.7%. In a general kindergarten population, roughly half of positive results will not turn out to be autism. In the Denver II study where 104 children received a full diagnostic battery: sensitivity 83%, specificity 43% — "more than half the children with normal development also received abnormal, questionable, or untestable scores". Must not attempt articulation therapy, oral-motor exercises or targeted speech-sound correction without a qualified specialist's assessment and plan. Must not treat receptive language difficulty as an in-house teaching problem: the Cochrane review found no evidence of effect for interventions targeting receptive difficulties (SMD −0.04, CI −0.64 to 0.56).

The right message to a family is neither "let us wait, they all catch up" nor "your child has a disorder". On Rescorla's review, most late talkers reach the average range by age 5, 6 or 7 — but as a group they score significantly lower on most language measures into adolescence. The age-17 data (26 late talkers and 23 comparison children — a small, socioeconomically advantaged sample) show the same: average range on all language and reading tasks, but significantly lower on Vocabulary/Grammar and Verbal Memory factors. The primary purpose of early screening is to find the minority whose expressive delay is secondary to hearing loss, autism spectrum disorder, intellectual disability or receptive language disorder. Immediate referral for specialist assessment is warranted for: loss of previously acquired words; no words by 18 months; no two-word combinations by 24 months; unintelligible speech at four; receptive difficulty; concerns about social communication. And one organisational rule: do not screen without a referral pathway. Screening that cannot lead to assessment converts anxiety into nothing. If no specialist is reachable, the honest institutional response is to document, monitor with repeated measures, run the language programme, and tell families exactly what is and is not available.