Verb retrieval in brain-damaged subjects: 2. Analysis of errors.

Verb retrieval for action naming was assessed in 53 brain-damaged subjects by administering a standardized test with 100 items. In a companion paper (Kemmerer & Tranel, 2000), it was shown that impaired and unimpaired subjects did not differ as groups in their sensitivity to a variety of stimulus, lexical, and conceptual factors relevant to the test. For this reason, the main goal of the present study was to determine whether the two groups of subjects manifested theoretically interesting differences in the kinds of errors that they made. All of the subjects' errors were classified according to an error coding system that contains 27 distinct types of errors belonging to five broad categories-verbs, phrases, nouns, adpositional words, and "other" responses. Errors involving the production of verbs that are semantically related to the target were especially prevalent for the unimpaired group, which is similar to the performance of normal control subjects. By contrast, the impaired group had a significantly smaller proportion of errors in the verb category and a significantly larger proportion of errors in each of the nonverb categories. This relationship between error rate and error type is consistent with previous research on both object and action naming errors, and it suggests that subjects with only mild damage to putative lexical systems retain an appreciation of most of the semantic, phonological, and grammatical category features of words, whereas subjects with more severe damage retain a much smaller set of features. At the level of individual subjects, a wide range of "predominant error types" were found, especially among the impaired subjects, which may reflect either different action naming strategies or perhaps different patterns of preservation and impairment of various lexical components. Overall, this study provides a novel addition to the existing literature on the analysis of naming errors made by brain-damaged subjects. Not only does the study advance our knowledge of the relatively under investigated topic of action naming errors, but it also approaches the analysis from the point of view of a detailed, theoretically motivated, and reliable error coding system.