Latent patterns in pitch-shift responses: association with working memory and temporal changes in response magnitude

This study investigated how working memory capacity is associated with individual differences in vocal responses to auditory perturbations during speech production. Fifty-two native Mandarin speakers participated in an altered auditory feedback (AAF) experiment, sustaining the vowel /ɛ/ while receiving brief pitch shifts (±100 cents) under three predictability conditions: down-only, up-only, and random. Each participant also completed four memory tasks, including digit span forward (DSF), backward (DSB), sequencing (DSS), and auditory memory (AM). Latent class analysis (LCA) of trial-by-trial responses identified three distinct response patterns: Switchers (alternating between opposing and following responses), opposers (consistently compensating for perturbations), and followers (consistently following the perturbations). Separate multinomial mixed-effects models revealed modest associations between memory performance and latent-class membership. Specifically, higher DSF and DSS scores were associated with a greater likelihood of belonging to the opposer class, whereas lower auditory memory scores were associated with a greater likelihood of belonging to the follower class. Generalized additive mixed-effects models (GAMMs) further demonstrated that opposers exhibited reductions in response magnitude over time across all predictability conditions, whereas switchers showed similar reductions only in the random condition and followers showed no such pattern. These findings reveal that memory-related individual differences are associated with variations in vocal feedback control and highlight the importance of considering latent response profiles when examining the interaction between cognitive and sensorimotor processes in speech production.