A theoretical construct for examining the use of reading in composing, developed from the literature review and pilot study, was tested using audio-videotaped protocols of students composing two drafts of an essay, followed by retrospective interviews.

Fifteen skilled and fifteen unskilled college writers, so designated by tests and a writing sample given by the English department, participated in the study at a rural, two-year technical college in a Sunbelt state during the 1984-85 school year.

The construct for examining reading in composing included the focus of reading: word, multi-word, sentence, multi-sentence, paragraph, multi-paragraph, and draft levels; the purpose of reading: to verify, clarify, provide direction, edit, or refresh memory; the effect (or outcome) of reading: no/change or change, using Faigley and Witte's revision classification scheme to describe changes; and the amount of reading, which was a count of all occurrences of reading. The construct was useful in identifying the aspects of reading in composing. Thus, a major benefit of the study is empirical data on reading-in-composing for both skilled and unskilled writers.

A profile of each group's use of reading was developed. Then a comparison of reading by the two groups was made using chi-square and percents.

The findings revealed that (1) 29 of the 30 students were readers of their texts; (2) the skilled writers wrote and read twice as much as the unskilled writers did, but the proportion in both drafts was the same; (3) reading occurred within and between drafts for both groups; (4) both groups read most often at the multi-word level; (5) both groups read for all five purposes and when ranked by frequency, the order was the same for both groups; (6) the effect of reading differed significantly in the no-change/change categories with the skilled writers making more changes.

Findings which were statistically significant included: the size of the corpus; the focus of reading; the difference in no-change/change decisions; and the categories of change at surface, meaning-preserving and meaning-changing levels. The study confirmed that writing is a recursive process with reading as a major component and that both skilled and unskilled writers are readers of their texts.

The study revealed that protocol analysis and the Faigley and Witte classification scheme for revision can work well together. Second, the amount, focus, purpose, and effect of reading can be examined through thinking-aloud composing protocols. Third, reading is a more appropriate term than re-reading to describe the reading which occurs during the composing process.