ABSTRACT
Models of spoken word production have been developed using evidence from two primary sources-speech errors and experimental techniques. Probably the most influential in the early development of these models was speech error data. Slips of the tongue made by speakers were examined in terms of types of error that did and did not occur, and the kinds of factors that influenced their occurrence. From these data inferences were drawn regarding the component processes of word production (although the same data could lead to different inferences; contrast, for example, Dell, 1986, and Shattuck-Hufnagel, 1979). The second main approach that has been used is experimental, where the factors that affect subjects’ performance (error rates, error types, or reaction times) are studied. Meyer (1992) suggests that experimental procedures have played a relatively minor role, at least in the development of models of phonological encoding. She argues that “detailed and comprehensive models . . . cannot be developed solely on the basis of error analyses [and that] future research should directly investigate the normal process of phonological encoding” (p. 181). Since Meyer made this point, experimental investigations of speech production with normal subjects have become a primary source of data in developing theoretical models (e.g., Levelt, Roelofs, & Meyer, 1999a). Beyond this, however, we would assert
