Prosodic characteristics of interrupted versus completed speech error repairs in spontaneous Dutch speech

We present an analysis of speech error repairs aimed at establishing whether prosodic differences between ‘interrupted’ and ‘completed’ error repairs found in elicited data [8] are observed in a corpus of Dutch spontaneous speech. It has been assumed that the interrupted–completed distinction is associated with the coordination of pre - and post-articulatory self-monitoring processes; we explore the use of more fine-grained measures of reparandum completeness than considered so far. We find no evidence in intensity and f0 measures to support a distinction between ‘interrupted’ and ‘completed’. Results of regression analyses differ in detail from the findings in [8] and provide only weak support for prosodic marking of early repairs. We suggest that analysis of interactional factors absent from experimental data may prove fruitful.