Abstract
Semantic errors produced by normal and aphasic speakers on a picture naming test were examined for their phonological similarity to the targets they replaced. In both groups of subjects, semantic errors preserved phonological characteristics of the target word at rates greater than would be expected by chance. These results are consistent with previous findings in speech error corpora (e.g. Dell amp; Reich, 1981) and error elicitation studies (Martin, Weisberg, & Saffran, 1989) and lend support to the assumption that semantic and phonological representations of a word are retrieved in two overlapping stages. In addition to extending this finding to the aphasic population, the use of a picture naming task in the present study eliminated some of the potential confounds that arise in connected speech. In a second study, we examined phonological relationships within sets of semantically related words and showed that there is no tendency for these words to share phonological characteristics. This finding substantiates the argument that the tendency for semantic errors to preserve phonological characteristics of the target word does not arise by chance. We discuss these results in relation to alternative models of lexical retrieval.