Effects of delay, length, and frequency on onset RTs and word durations: Articulatory planning uses flexible units but cannot be prepared

Abstract

Models of speech production differ about whether articulatory planning occurs online or offline and about whether a separate articulatory buffer is needed. The hypothesis that most articulatory planning is completed before articulation starts (offline) assumes a buffer where articulatory plans are accrued and kept ready before articulation starts. The hypothesis that articulatory planning can occur after the beginning of articulation (online) does not need a proper buffer but only limited memory resources encapsulated within articulatory processing (not independently available). We have tested these alternative hypotheses by comparing onset RTs that measure offline processes, and word durations that measure online processes. We tested young control speakers, an aphasic speaker, CS, and an age and education matched speaker, using repetition, reading and picture naming tasks. Our results support the view that articulatory planning occurs mostly online and that articulatory plans cannot be prepared, buffered, and refreshed like phonological representations can. The possibility to prepare (in delayed naming conditions) strongly reduced onset RTs, but had no benefit for word durations (in fact, we found small effects in the opposite direction). Moreover, we found significant effects of frequency and imageability on word durations, which are not predicted if all processing is completed before articulation starts. Finally, in word durations we found no weakening of frequency or length effects with preparation, again suggesting that these variables affect online processing. The same pattern of results was found in CS and in control participants, strengthening confidence in our results.

Timing in Conversation

“…The fourth claim of Levinson and Torreira’s model is the distinction between response planning and launching: Speakers begin to prepare a response to their partner as soon as possible, but only launch it shortly before the anticipated end of the partner’s turn. This proposal is consistent with a large body of experimental work using delayed naming tasks, which has shown that speakers can indeed generate speech plans internally, retain them in working memory, and produce them upon presentation of a response cue (for recent discussions see Kawamoto, Liu, & Kello, 2015 ; Krause & Kawamoto, 2020 ; Piai, Roelofs, Rommers, Dahlstaett, & Maris, 2015 ; Romani, Silverstein, Ramoo, & Olson, 2022 ). The latencies to produce prepared utterances are much shorter than those observed for utterances not planned ahead of time.…”