Ability to vary one's style is an important skill of mature writing, and it would be useful to have tests of this skill. We developed a cloze test to measure writing flexibility, then asked college students (all good writers) to replace sentences that had been deleted from two short stories. The style of the cloze sentences, for students with experience in creative writing, more closely resembled the original story than the cloze sentences of less experienced students. Style differences, between experienced and inexperienced students, appeared in average sentence lengths, sentence types, and verb-adjective ratios. In another experiment, less experienced students were given explicit instructions to imitate story style; they showed virtually the same adaptability to style as the creative writing group in the first experiment. Thus we have evidence that the cloze test measures style differences between experienced and less experienced writers, and also that responsiveness to style features, distinct from the skill needed to change those features, is a significant component of experienced writing.

Computers make it easy to tabulate text style variables [1]. In our laboratory we have gone farther, creating an automated system that uses those tabulations to make detailed editorial comments [2]. Relying on advice from writing experts and psychological research, our “Writers' Workbench” calculates several readability measures, comments on misspelled words and awkward phrases and sentences, measures text abstract-ness, and compares these properties of a text to other texts. Writers find these aids useful, and there are obvious management and educational implications. Before putting these tools together, however, we considered how such aids might be used and misused. Below I review these considerations in light of readability formulas.