Self-monitoring of speech errors: effects of phonetic contrast

Abstract

Speech sound errors are known to occur more often as the interfering speech sounds are phonetically more similar. This paper aims to test the general hypothesis that the phonetic contrast between interfering speech sounds also mediates the odds of detection and repair of such a speech error. This was investigated by re-analyzing responses from four published experiments in which speech errors had been elicited, here using Bayesian modeling with random effects of participants and stimuli. Results show that with increasing phonetic contrast among the segment involved, speech errors occur less frequently (as expected); those errors tend to be detected and repaired more frequently, with early repairs being more prevalent than late repairs, as predicted. These patterns suggest that repair is not triggered by conflict during production, but by error detection during self-monitoring.