Transfer of statistical learning from passive speech perception to speech production

Abstract

Communicating with a speaker with a different accent can affect one’s own speech. Despite the strength of evidence for perception-production transfer in speech, the nature of transfer has remained elusive, with variable results regarding the acoustic properties that transfer between speakers and the characteristics of the speakers who exhibit transfer. The current study investigates perception-production transfer through the lens of statistical learning across passive exposure to speech. Participants experienced a short sequence of acoustically variable minimal pair (beer/pier) utterances conveying either an accent or typical American English acoustics, categorized a perceptually ambiguous test stimulus, and then repeated the test stimulus aloud. In the canonical condition, /b/–/p/ fundamental frequency (F0) and voice onset time (VOT) covaried according to typical English patterns. In the reverse condition, the F0xVOT relationship reversed to create an “accent” with speech input regularities atypical of American English. Replicating prior studies, F0 played less of a role in perceptual speech categorization in reverse compared with canonical statistical contexts. Critically, this down-weighting transferred to production, with systematic down-weighting of F0 in listeners’ own speech productions in reverse compared with canonical contexts that was robust across male and female participants. Thus, the mapping of acoustics to speech categories is rapidly adjusted by short-term statistical learning across passive listening and these adjustments transfer to influence listeners’ own speech productions.

Bears Don’t Always Mess With Beers: Limits on Generalization of Statistical Learning in Speech

“…The present study also lays groundwork for understanding the currency of representation shared between speech perception and production. We replicated the transfer of statistical learning from perception to production reported in Murphy et al (2023) twice (Experiment 1 and 2, beer-pierbeerpier). Additionally, the present work extends evidence of transfer to a novel context and word pair (Experiment 2, bear-pearbear-pear).…”

“…We modeled the influence of statistical learning on perceptual categorization of test stimuli using mixed effects models (lme4, Bates, Mochler, Bolker, and Walker, 2015) Speech Production. Transfer of statistical learning in listening to speaking was modeled across by-participant z-score normalized utterance F0 (as in Murphy et al, 2023). In brief, the F0 (computed across the first 40 ms) was measured for each utterance.…”

“…A Canonical sampling mirrored the typical American English F0xVOT correlation whereas a Reverse sampling created a subtle accent with an opposing correlation. In a passive exposure version of the paradigm (Hodson et al, 2023;Murphy et al 2023), listeners hear a sequence of beer and pier utterances conveying one of these distributional regularities followed by one of two F0-differentiated test stimuli with ambiguous VOT. With only F0 available to convey category identity, test stimulus categorization indexes listeners' reliance on F0 in speech categorization.…”

“…As a secondary measure in each experiment, we elicit speech productions to attempt to replicate recently reported transfer of statistical learning from perception to production (Murphy et al, 2023) and to examine generalization in production.…”

Is working memory domain-general or domain-specific? Implications for theory and practice

“…However, if the same principles operate across domains, continuous resource models should be able to account for the data in verbal WM. This proposition is not unreasonable, given that despite categorical perception, listeners are sensitive to subtle changes in the distribution of statistical information in acoustic dimensions even when the information does not flip the category [37,38], and such sensitivity is powerful enough to change production [39].…”