Syntactic Maturity, Mechanics, and Vocabulary as Predictors of Quality Ratings.

An analysis of student expository writing by Stewart and Grobe (1979), showed that holistic quality scores awarded by teacher-raters were influenced to a greater extent by essay length and freedom from spelling errors than by syntactic maturity and mechanics. A replication of their investigation using a sample of students* narrative writing produced the same findings. HolisticaUy derived writing scores were regressed in a step-wise fashion on 14 syntax, usage and mechanics variables. Separate analyses of grades 5, 8 and 11 pupil writing yielded regression solutions which accounted for 59.8, 42.0 and 31.8% of holistic score variance. Ten variables, containing vocabulary information, were added to the existing 14 variable prediction system, and the step-wise regressions repeated. At each grade level, the inclusion of vocabulary information greatly improved the accuracy of the predictions. The amounts of explained variance were 81.4, 64.9 and 59.8% respectively for grades 5, 8 and 11. Vocabulary characteristics had a stronger relationship to holistic scores than any other classes of information included in the prediction systems. A recent paper by Stewart and Grobe (1979) detailed the relationship between teachers' quality ratings of grades 5, 8 and 11 students and various measures of syntatic maturity and mechanics of writing. Their study suggested that teacher-markers, in general, were "more influenced by the length of the compositions and their freedom from simple mechanical errors," than by students' control of the syntactic resources of their language. The Stewart and Grobe investigation concluded that, as predictors of quality, composition length and freedom from spelling errors accounted for about 25% of the variance in teacher quality ratings. In grade 5, syntactic maturity variables explained about 10% of the accountable variance, while at grades 8 and 11, they contributed virtually nothing to the prediction of quality rating. The purpose of this paper is two-fold: (1) to replicate the original StewartGrobe prediction study using a similar data base, and (2) to explore the relationship between specific vocabulary characteristics and teacher quality ratings.