Several types of narrative errors are discussed that were found in the course of an analysis of stories collected in casual settings from a number of American speakers in undirected conversation. The approach to the question of error correction is sociocultural; the emphasis is on the motivation for the error correction. This paper explores the reason why some sorts of lexicalizations, ordering of sentences and "false starts" are viewed as errors. The thesis here is that the correction of errors is actually used strategically in talk, so the correction of errors in talk is not random but must be learned, like other aspects of language use. Errors in event structuring and descriptive structuring are examined first, then errors of evaluation are investigated. Examples of dialogue and stories accompany the discussion. (Author/AMH)