ABSTRACT
According to Levelt (1989) and Levelt, Roelofs, and Meyer (1999) (a) selfmonitoring of speech production employs the speech comprehension system, (b) on the phonological level the speech comprehension system has no information about the lemmas and forms chosen in production, and (c) lexical bias in speech errors stems from the same perception-based monitoring that is responsible for detection and overt correction of speech errors. It is predicted from these theoretical considerations that phonological errors accidentally leading to real words should be treated by the monitor as lexical errors, because the monitor has no way of knowing that they are not. It is also predicted that self-corrections of overt speech errors are also sensitive to lexicality of the errors. These predictions are tested against a corpus of speech errors and their corrections in Dutch. It is shown that the monitor treats phonological errors leading to real words in all respects as other phonological, and not as lexical errors and that no criterion is applied of the form “is this a real word?” It is also shown that, whereas there is considerable lexical bias in spontaneous speech errors and this effect is sensitive to phonetic similarity, self-corrections of overt speech errors are not sensitive to lexical status or phonetic similarity. It is argued here that the monitor has access to the intended word forms and that lexical bias and self-corrections of overt speech errors are not caused by the same perception-based self-monitoring system. Possibly fast and hidden self-monitoring of inner speech differs from slower and overt self-monitoring of overt speech.
