Читать книгу The Handbook of Speech Perception - Группа авторов - Страница 65

In speech perception, selective adaptation for both consonants and vowels results in changes to category boundaries after exposure to a repetitive adapting stimulus. Cooper (1974; Cooper & Lauritsen, 1974) reported production changes in produced VOT following repeated presentation of a voiceless adapting stimulus. In a manner similar to the effects of selective adaptation on perceptual category boundaries, talkers produced shorter VOTs after adaptation. The effect was attributed to a perceptuomotor mechanism that mediates both the perception and the production of speech. More recently, Shiller et al. (2009) found that, when subjects produced fricatives with frequency‐altered feedback, they produced fricatives that compensated for the perturbation. Most interestingly, the subjects shifted their perceptual boundary for /s/–/sh/ identification following this production perturbation. However, as Perkell (2012) cautions, the segment durations of the fricatives were far beyond the natural range, raising the possibility that the effect was acoustic rather than phonetic. Lametti et al. (2014) showed the opposite direction of influence. A perceptual training task designed to alter perceptual boundaries between vowels preceded a production task. No shift was observed in baseline vowel formant values but a difference was observed in the magnitude of compensation to F1 perturbations. Oddly, this difference was observed in a follow‐up days later. The persistence is surprising for a number of reasons. First, the speech adaptation effects produced by formant shifts themselves drift away relatively quickly within an experimental session following return to normal feedback. Second, the perceptual training didn’t influence baseline vowel production immediately after training nor days later. The influence of perceptual change on production is shown only in the magnitude of compensation (i.e. in the behavior of the auditory feedback processing system). Finally, the length of effect is noteworthy. While it is not unheard of for perceptual effects to persist across many days, it is not common; the McCollough effect in vision has been shown to last for months after a 15‐minute training period (Jones & Holding, 1975). However, the reinforcement learning paradigm used by Lametti et al. (2014) is considerably different from the adaptation approach used in other studies and suggests a more selective influence on the perception–production linkage.

The published data suggest modest effects from speech‐perception training on speech production and vice versa. As Kittredge and Dell (2016) suggest, the pathway for exchange between the input and output systems may be restricted to a small set of special conditions. Kittredge and Dell suggest that one possibility is that perceptual behavior that involves prediction invokes the motor system and this directly influences production.

A separate line of research has suggested this influence may exist but has shown similar, small effect sizes in experiments. In the study of face‐to‐face conversations, considerable theoretical proposals support the idea that interlocutors align their language at many levels (Garrod & Pickering, 2004). At the phonetic level, the findings have been weak but consistent. Few acoustic findings support alignment but small perceptual effects have been frequently reported (Pardo et al., 2012; Kim, Horton, & Bradlow, 2011. The surprising aspect of these findings is the small effect size. Given the proposed importance of alignment in communication (and the proposed linkage between perception and production; Pickering & Garrod, 2013), the small influence is problematic.