Schedule of Events | Symposia

Do listeners access abstract morphological representations while processing speech? A representational similarity analysis of event-related potentials

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Chiung-Yu Chang¹ (chiungyuchan@umass.edu), Alexandra Jesse¹, Lisa D. Sanders¹; ¹University of Massachusetts Amherst

The English plural suffix is pronounced as /z/, /s/, or /ɪz/ depending on the phonological context. Classical linguistic analyses posit an abstract representation of the suffix, while episodic theories do not. Neither of these approaches specify the timing of abstraction during speech processing. This study explored the representational geometries predicted by these theories. We recorded event-related potentials (ERPs) elicited by four types of sounds: /z/ as the canonical plural suffix (e.g., hill-/z/), /s/ as an illegal substitute for the plural suffix (e.g., hill-/s/), /z/ as the coda of a noun (e.g., maze), and /s/ as the coda of a noun (e.g., moose). Additionally, we manipulated the suffix predictability by adding a number (e.g., two hill-/z/) or a definite article (e.g., the hill and the hill-/z/). The acoustic-phonetic model supposes no abstract representation of the suffix, predicting that ERPs for /z/ will always differ from those for /s/. In contrast, abstractionist models predict that ERPs for /z/ and /s/ will be more similar when they serve the same syntactic function. Representational dissimilarity matrices (RDMs) based on these predictions were compared to RDMs derived from the observed ERP data in early and late time windows (i.e., 100-200 and 400-500 ms after /s/ or /z/ onset). Results from 25 participants suggest that the acoustic-phonetic model best predicts the observed data, supporting episodic theories. However, an abstractionist model performs equally well for the late time window when the plural suffix is predictable, suggesting an interplay between predictability and allomorph representations during speech processing.

Topic Area: PERCEPTION & ACTION: Audition