The nature of error in adolescent student writing

178 essays from 67 students (33 in Grade 10, 34 in Grade 12; 49 first-language, 18 second-language) across five US states

US adolescents in Grades 10 and 12

Descriptive error analysis of student essays written for English language arts and for social studies, coded against an error taxonomy. Not an intervention study and not randomised - it describes what errors adolescents actually make and where. Recorded for two facts a conventions topic needs and cannot get anywhere else: which conventions errors dominate, and whether the same student makes more of them when the composing task is harder. Sample is small and the essays are naturally occurring rather than standardised, so treat magnitudes as indicative.

Spelling, capitalisation and some punctuation errors are the most frequent error types, and the eight commonest types account for a little over half of all errors observed; most other error types are rare. That matters practically: the conventions problem in real writing is concentrated in a short list, not spread evenly across the grammar curriculum. The second finding is the more interesting one for this archive - error rates were significantly HIGHER in social studies essays than in English language arts essays, in the same students. Holding conventions knowledge constant, raising the content and composing load raises the surface error rate. That is direct observational support for treating "knowing a convention" and "applying it under composition load" as different outcomes rather than as one skill measured twice.

Not applicable to the within-student subject-area contrast, which compares each student to himself across two writing tasks and therefore nets out stable individual differences. The descriptive error frequencies are a corpus description, not a claim about causes of individual differences.