Using sentence dictation to practise and assess taught spelling and punctuation skills: a Year 2 Explicit Instruction intervention

60 pupils in 3 classes across 2 schools (IntA n=17, IntB n=18, comparison n=25)

Year 2 pupils (aged roughly 7-8) in two New South Wales primary schools

The youngest controlled study of punctuation and capitalisation teaching that exists. Nine weeks of Explicit Instruction in spelling concepts plus sentence punctuation (capital letters and full stops), assessed by researcher-written sentence dictation. Design limits are severe and structural: only TWO clusters, one intervention school and one comparison school, so the school effect is completely confounded with treatment; no randomisation; the first author delivered and designed the programme; and the outcome is punctuation in a DICTATED sentence, where the child does not have to compose - i.e. conventions measured at the lowest possible composition load. Between-teacher variance inside the intervention school was larger than the treatment contrast in one comparison, which is itself informative.

Punctuation errors fell substantially in both intervention classes and barely moved in the comparison class. Dictation 1, percentage of pupils making the error, pre to post: full stop omitted - IntA 71% to 41%, IntB 72% to 0%, comparison 64% to 64%; capital letter omitted - IntA 88% to 59%, IntB 83% to 39%, comparison 80% to 76%. Dictation 2: full stop - IntA 87% to 47%, IntB 94% to 28%, comparison 76% to 72%; capital letter - IntA 80% to 47%, IntB 89% to 22%, comparison 84% to 56%. Composite within-class Cohen's d was 0.2 and 0.4 for IntA, 1.4 and 1.2 for IntB, and 0.1 and 0.7 for the comparison class. The two intervention classes differ from each other by more than IntA differs from the comparison class, so teacher effects are not a footnote here - they are the largest source of variation in the study.

Medium-high in principle: two whole schools compared without randomisation means any intake difference between the schools is inside the treatment estimate. The pre-test error rates were broadly comparable across classes, which is the available mitigation, and the comparison class's own flat punctuation trajectory over nine weeks is the more informative half of the result.