STRATEGIES FOR EDITING OUT SPEECH ERRORS IN INNER SPEECH Sieb Nooteboom & Hugo Quené Utrecht institute of Linguistics OTS, Utrecht University Sieb.Nooteboom@let.uu.nl ABSTRACT In a classical SLIP task spoonerisms are elicited with either a lexical or a nonlexical outcome. If the frequency of a particular class of responses is affected by the lexicality of the expected spoonerisms, this indicates that many such responses have replaced elicited spoonerisms in inner speech. This is shown in early interrupted speech errors and in completed speech errors that deviate from the elicited spoonerisms. Keywords: Speech errors, lexical bias, feedback, self-monitoring, inner speech. 1. INTRODUCTION In a classical SLIP (Spoonerisms of Laboratory- Induced Predisposition) task [1], consonant exchanges are elicited that have either a lexical outcome, as in BARN DOOR > DARN BORE, or a nonlexical outcome, as in BAD GAME > GAD BAME. (Recently, it has been made plausible that such ‘consonant exchanges’ really are full or partial exchanges of ‘articulatory gestures’ [4]. In this paper we stick to the notion ‘spoonerisms’ without committing us to either ‘phonemes’ or ‘articulatory gestures’ as the units involved). Spoonerisms with a lexical outcome were shown to be more frequent than those with a nonlexical outcome [1]. This so-called ‘lexical bias’ effect was explained from pre-articulatory editing of inner speech: Nonwords are supposed to be more frequently detected and repaired before speech is initiated than real words. This explanation is supported in [7, 8]. Others have explained lexical bias in phonological speech errors from feedback of activation between phonemes and words in the mental production of speech [2, 11]. Such feedback increases the activation of lexical but not of nonlexical errors, because the latter have no lexical representations. The two mechanisms do not exclude each other, and there is increasing evidence that the relative frequencies of real-word errors and nonword errors are affected both by feedback and by self-monitoring [5, 6]. An important source of evidence stems from errors in a SLIP task that are not identical with the elicited spoonerisms, for example ‘early interrupted’ speech errors, such as G..BAD GAME. If such interruptions are more frequent in the condition with expected lexical outcomes (i.e., positive lexical bias), this might reflect an underlying effect of feedback on the frequency of covert spoonerisms. If interruptions are less frequent in the condition with expected lexical outcomes (i.e., negative lexical bias), this might reflect that self-monitoring detects nonlexical errors more frequently than lexical ones. Note that interrupted errors are detected in inner speech, not overt speech, because speech fragments before interruption are generally shorter than humanly possible reaction times and offset-to-repair times are often close to 0 ms. If in a SLIP task the frequency of speech errors, that are neither completed spoonerisms nor interrupted spoonerisms, is also significantly affected by the lexicality of elicited spoonerisms, then, we argue, that this would suggest that self- monitoring inner speech has replaced elicited spoonerisms with other speech errors. When these other errors are more frequent in the condition with expected lexical outcomes, this would possibly reflect an underlying effect of feedback on the frequency of the covert spoonerisms. When these other errors are less frequent in the condition with expected lexical outcomes, however, this would possibly reflect that self-monitoring more frequently detects and replaces nonlexical than lexical spoonerisms. Note that the standard argument in [1, 7, 8] would be that self-monitoring replaces the spoonerisms with the correct targets. Obviously, this can not be observed in the error counts. Below we describe two experiments employing the SLIP technique for eliciting spoonerisms [1]. In the first experiment the participants are under ICPhS XVI ID 1119 Saarbrücken, 6-10 August 2007 www.icphs2007.de 1945