Corey McMillan(1), Martin Corley(1), and Robert Hartsuiker(2)
(1)University of Edinburgh (2)Ghent University

The lexical bias effect (LBE) is a well-established finding that speech errors result in real words more often than chance would predict. Two competing models of speech production attempt to account for the LBE that mainly differ in the way information is passed in the speech planning process; either feedforward or feedback. Support for the feedforward model comes from a speech error eliciting study, the SLIP task, showing that the LBE is modulated by lexical context (Baars et al., 1975). Recently, however, Baars et al.'s findings have come into question. Using variants of the Baars et al.'s (1975) SLIP task, Humphreys (2002) finds support for a feedback model with evidence for a LBE regardless of context, whereas Hartsuiker et al.(in press) find support for a combined model with an increase in real-word exchange errors in contexts which include lexical items. This talk will discuss in detail the LBE in the context of the dominant theories of speech production. Evidence for the LBE in corpus analyses will be reviewed and three experiments using the SLIP task, designed to investigate and account for the differences in recent findings will be reported. Two experiments manipulated response-time constraints to investigate whether faster responses lead to a larger (suggested by Humphreys, 2002) or smaller (found by Dell, 1986) lexical bias across conditions. Additionally, in Experiment 2, participants were required to listen to white noise throughout the task (as was the case for Baars et al. (1975) and Hartsuiker et al. (in press) but not Humphreys (2002)) in an attempt to establish whether differential access to internal/external monitoring could account for the differences between studies. Lastly, there will be a discussion about future directions of research in this area including measurements of articulation and the development of new speech-error eliciting experimental paradigms.