Sonority constraints on perceptual errors

Abstract

This study explored phonological constraints upon word misperceptions elicited from normal listeners by degrading the speech signal with noise and various high- and low-pass filters. Pre-vocalic sound sequences of error words were compared to American English targets. Syllable onset substitutions were evaluated with respect to the sonority sequencing principle (SSP), which holds that the least sonorous sounds of languages (those produced with a relatively constricted vocal tract) possess little perceptual prominence and are preferred in onset position to more sonorous sounds of languages (those produced with a more open vocal tract). Results suggest that errors pattern in accordance with predictions from sonority theory and suggest that several different phonological mechanisms may account collectively for findings. The primary mechanisms include partial signal analysis and/or activation of a default lexical search mechanism which accesses words possessing the least-marked of possible pre-vocalic segmental sequences (e.g., obstruent-vowel strings). Confusion matrices indicate that misperceptions are remarkably resistant to filter type.