When Is a Verb? Using Functional Grammar to Teach Writing

3 intact tenth-grade classes; 18 + 21 treatment and 18 control pupils analysed

Tenth graders (aged ~15-16) in one urban US high school in the bottom 10% of the state, ~65% classified limited English proficient, nearly 100% eligible for free or reduced lunch

Rare and valuable because it measured conventions KNOWLEDGE and conventions USE in the same pupils at the same time, which almost nothing in this literature does. Two intact classes were taught "functional" grammar in writing by the investigators (10-12 min twice weekly plus 8-10 min teacher follow-up twice weekly, ~200 minutes over five weeks); a third class on the other side of the same campus was taught traditional identification-definition-description grammar by its own English teacher, using the whole 47-minute period daily on identification worksheets, cloze exercises, editing prepared sentences for subject-verb agreement, and daily process writing. So the CONTROL arm is the conventions-instruction-heavy arm, and it received several times more instructional time. Outcomes were five-minute first-draft writing samples scored blind for fluency (words) and mechanical control (mean errors per sentence: punctuation, capitalisation, spelling, tense agreement; inter-rater 97%) and a six-point holistic general-impression rubric (three trained raters, inter-rater 96%), plus a grammar test. Serious limitations: no randomisation; n ~57; every measure researcher-designed; the investigators both delivered the treatment and operationalise their own published writing programme; and the pre- and post-test grammar instruments were DIFFERENT (18 recognition items before, 7 "write a sentence that contains..." production items after), so pre-post grammar gains are not interpretable and only the between-group post-test comparison carries information.

Mechanical control did not move in any arm. Errors per sentence went 1.3 to 1.3 in the first treatment class, 1.6 to 1.3 in the second, and 1.3 to 1.2 in the traditional-grammar control - a control class that had spent a full 47-minute period every day for five weeks on grammar identification, cloze work, subject-verb agreement editing and writing. The authors state it plainly: "neither mode of instruction seemed to influence students' use of mechanics and the conventions of written text", and "the more traditional grammar instruction had no greater influence on students' error rate than did the grammar-driven writing instruction that was not directed at reducing error rate." On the seven-item grammar production test the arms were statistically indistinguishable (treatment 4.00 and 4.00 vs control 4.72; p < .324 and p < .330) - the class taught grammar the traditional way did not end up knowing more grammar. On holistic writing quality the treatment classes gained (2.94 to 3.61 and 2.95 to 3.48) while the traditional-grammar control DECLINED (2.78 to 2.61), both between-group differences significant (p < .002, p < .003). Fluency rose comparably everywhere (roughly 63-76 to 88-93 words in five minutes), so all three classes were writing more by the end.

Medium. Assignment was by intact class within one school rather than by randomisation, so class-level differences in prior attainment could drive the holistic-quality contrast; the authors report pre-test equivalence on writing and grammar, which is the available mitigation. The mechanics result is a within-arm pre-post comparison in every arm simultaneously, which is the part least exposed to selection: no group improved, so no selection story is needed to explain a difference that does not exist.