Tracing the causes of writing errors : a revision of error analysis procedures and assumptions

Typescript (photocopy).Error analysis (EA) in composition studies attempts to determine the causes of writers' uses of nonstandard language. Typified by the work of Kroll and Schafer, Bartholomae, and Shaughnessy, traditional EA draws extensively upon second language acquisition research from the 1970s and generally relies on researchers' textual interpretations of sample errors. The present study attempts to refine EA assumptions and methodology by proposing a model that utilizes a greater variety of procedures and sources of data. The model entails two types of descriptions necessary for understanding error. The three primary sources of data are sample errors (including naturally occurring and elicited samples), writers' interpretations of error, and writers' evaluations of error; all three sources involve direct and indirect means for obtaining writers' responses. For complementary data, EA can draw upon non-errors as well. The model suggests various techniques and data needed to form grammatical and rhetorical descriptions of error, descriptions that allow an analyst to define the constraints on error and its causes. For errors related to linguistic competence, grammatical descriptions involve the error's linguistic function, transitional/stable nature, and relationship to SE, dialect, and/or interlanguage grammars. For errors dealing with linguistic performance, grammatical descriptions involve various syntactic environments conducive to oversights or retention problems. Rhetorical descriptions, based on the communications triangle, explore the relationship of error with both textual and extra-textual factors of the writing situation. The components of the triangle are broadly defined so that they include cognitive, affective, and social factors. The study applies the suggested assumptions and procedures through an analysis of nonstandard uses of in which. The analysis draws upon elicitation, evaluation, revision, and rewrite tasks to form grammatical and rhetorical descriptions. Results indicate the various nonstandard functions of the variable and the constraints imposed by an interaction of the writer's gender and the perceived formality of the writing situation