Relative contributions of feedback and editing in language production: Behavioral & articulatory evidence

Abstract

Psychologists normally attribute the surfacing of phonological speech errors to one of two factors: editing of the speech plan [Levelt (1989)] or feedback between word and phoneme levels [Dell (1986)]. This paper assesses the relative contributions of each factor, focusing on the perception and articulation of elicited speech errors. Experiments one and two measure the likelihood of phonological exchange errors as a function of phonetic similarity [Frisch (1996)], using the SLIP paradigm and a tongue-twister task. Both experiments show that error likelihood increases with phonetic similarity between intended and actual utterance, an effect easy to account for in terms of feedback but not in terms of editing. Experiment three uses EPG to analyze the tongue-twister utterances: many errors occur at the articulatory level but are not easily perceived in the speech signal. Preliminary analysis suggests three patterns of error: (1) substitution of segments, which may be the result of editing; (2) simultaneous double articulation, hypothesized to be the result of residual activation due to feedback; and (3) overlapping double articulation, representing partial execution of one articulation before substitution with another. Taking these findings together, we hope to evaluate the relative contributions of editing and feedback to phonological speech errors.