Timing and tempo in spontaneous phonological error repair

• We model the temporal organisation of spontaneous phonological error repairs.
• We quantify the relevance of repair timing for repair tempo.
• We find a predominance of temporal compression after repair initiation.
• We find evidence of a significant effect of repair timing on repair tempo.
• We compare methods of quantifying repair timing.

This paper reports on a study of the temporal characteristics of phonological error repair in spontaneous Dutch speech, with a focus on how the articulation rate of the correct target word production — the repair — compares to that of the preceding erroneous target word attempt — the reparandum. The study is motivated by two findings from recent independent studies: first, that self-repair is generally associated with relative temporal compression — that is, a local increase in articulation rate — following the repair initiation; second, that the timing of the repair initiation relative to the error is consequential for the prosody of the repair component. This study investigates to what extent these findings generalise to a collection of spontaneous phonological error repairs sampled from the Spoken Dutch Corpus. The study also asks how best to quantify repair timing, considers whether timing is consequential for the duration of the ‘offset-to-repair interval’, and tests for effects of lexical frequency. The results confirm that temporal compression following the repair initiation is more common than temporal expansion, and that repair timing has a significant effect on both offset-to-repair duration and repair tempo — at least in a subset of the data. A frequency effect is also observed. The results suggest that proportional measures of target word completeness provide the most informative quantifications of repair timing in modelling the overall temporal organisation of phonological error repairs.