Karin R. Humphreys & Angela Swendsen
University of Illinois at Urbana-Champaign
Speech errors, while disordered speech, are nonetheless remarkable in their orderliness; an example of this is the lexical bias effect. Baars, Motley and MacKay (1975), replicated by Dell (1986), found experimentally elicited phoneme exchange errors ("barn door" spoken as "darn bore") are more likely when the errors form words than when they form nonwords ("dart board" spoken as "bart doard"). This effect is initially puzzling. Generally, errors respect only the constraints that apply to the processing level at which they occur, e.g., phonological errors respect phonotactic, but not syntactic category constraints.
Lexical bias could arise from feedback along backward links from phonological to morpheme nodes in the production system (Dell 1986; Dell & Reich, 1981). Activated phonemes feed back to connected morphemes, which then feed forward to other phonemes. This resonance of activation makes word errors more likely, as nonwords have no morpheme node to support the accumulation of activation. However, these backward links are prohibited by the model of Levelt et al. (1999). They instead account for the effect via an editor, which is more likely to prevent nonwords from being articulated than words.
This experiment investigates further the lexical bias effect. We elicited phoneme exchanges from 100 speakers using stimuli where the error outcomes had either the first word a real word, and the second a nonword, or vice versa ("dean beak" slips to "bean deak", word-nonword; "deal bead" slips to "beal deed", nonword-word). This follows up Dell and Reich's (1981) corpus finding that only the first words in exchanges had a lexical bias.
We found that complete phoneme exchanges were far more likely for word-nonword than nonword-word outcomes (47 vs. 19). This asymmetry can be accounted for naturally by a feedback account, plus an incremental model of exchange errors where they are precipitated by an anticipation (e.g., "dean beak" starts as "bean ") displacing the intended phoneme ("d"), leaving it nowhere to go but the vacated onset of the second word, thus completing the exchange ("deak"). Feedback explains the automatic lexical bias on the first word, and then the second part of the error happens by default. The lexical editor however, does not predict that word-nonword errors are more likely than nonword-word errors, unless it specially adds a mechanism by which the articulation of an initial nonword is more likely to be prevented. Implications for feedback and incrementality in models of language production will be discussed.
References
Baars, B. J., Motley, M. T., & MacKay, D. G. (1975). Output editing for lexical status in artificially elicited slips of the tongue. Journal of Verbal Learning and Verbal Behavior, 14, 382-391.
Dell, G. S., & Reich, P. A. (1981). Stages in sentence production: An analysis of speech error data. Journal of Verbal Learning & Verbal Behavior, 20, 611-629.
Dell, G. S. (1986). A spreading-activation theory of retrieval in sentence production. Psychological Review, 93, 283-321.
Levelt, W. J. M. (1989). Speaking: From Intention to Articulation. Cambridge, MA: MIT Press.
Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production. Behavioral & Brain Sciences, 22, 1-75.