You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

 
Poster D55, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

Declarative and procedural memory substrates of the categorical perception of speech

F. Sayako Earle1, Emily B. Myers2, Jarrad A.G. Lum3, Michael T. Ullman4;1University of Delaware, 2University of Connecticut, 3Deakin University, 4Georgetown University

Background. For speech sounds to be considered ‘categorically’ perceived, two behavioral phenomena have traditionally been required: (1) categorical (rather than linear) identification (e.g. “/ka/ or /ga/?”) of sounds that vary across an acoustic-phonetic continuum, and (2) greater accuracy at discriminating sounds that fall across vs. within a category boundary. It is often assumed that performance on these two tasks indexes a common underlying representation. A growing literature challenges this view, suggesting instead that the two tasks may reflect qualitatively different types of speech knowledge (Schouten, Gerrits, & van Hessen, 2003; Antoniou, Best, & Tyler, 2013; Earle & Myers, 2015). However, the precise nature of such potential differences remains unclear. Hypothesis. We propose that learning speech sounds relies on two general-purpose learning and memory systems in the brain, declarative and procedural memory. Under this view, observed dissociations between identification and discrimination (and other) tasks may in part be due to task-specific preferential recruitment of declarative or procedural knowledge (Earle & Myers, 2014). In particular, we suggest that categorical identification, which requires the explicit recall of a category label, generally recruits phonetic features learned by declarative memory, whereas discrimination may be a skill-based task that is learned mainly in procedural memory. Methods. To test this hypothesis, adult participants (18-24) completed a two-session experiment designed to assess the learning (day 1) and retention (day 2) of (i) a set of (non-native) speech sounds and (ii) knowledge in both declarative and procedural memory. To control for circadian effects on learning, and to limit the amount of between-session linguistic exposure, all participants completed the training/assessment session at 7PM, and returned on the next day at 8AM for re-assessment. Speech-sound learning and retention was measured by identification and discrimination tasks on a Hindi dental-retroflex contrast. Learning and retention in declarative memory were measured in a non-verbal recognition memory task (incidental encoding, with a real/novel object decision task). Learning and retention in procedural memory were measured with a serial reaction time task. Between-session sleep was tracked via wrist actigraphy. Results. Regression models on the 18 participants tested thus far revealed that the overnight changes in declarative knowledge (=.604, t=3.686, p=.002), but not procedural knowledge (=-.132, t=-.871, p=.398), significantly predicted overnight gains in speech-sound identification. In contrast, the overnight changes in procedural knowledge (=.303, t=3.043, p=.008) but not declarative knowledge (=.116, t=1.081, p=.297) significantly predicted changes in speech-sound discrimination. These relationships were maintained even after controlling for sleep duration and sleep efficiency. Conclusion. Thus, it appears that the overnight memory processes that act upon declarative memory also act on speech-sound identification. In contrast, the memory processes that act on procedural memory also act on speech-sound discrimination. These findings support the interpretation that, at least in nonnative contrast learning, the identification of speech sounds relies heavily on declarative memory, whereas the discrimination of speech sounds depends importantly on procedural memory. In sum, speech-perceptual task performance may rely on knowledge acquired by two general-purpose learning systems, and moreover, different aspects of speech processing may depend on different types of memory.

Topic Area: Perception: Speech Perception and Audiovisual Integration

Back to Poster Schedule