Observing the contribution of both underlying and surface representations: Evidence from priming and event-related potentials

Abstract

This dissertation aims to uncover the role of the acoustic input (the surface representation) and the abstract linguistic representation (the underlying representation) as listeners map the signal during spoken word recognition. To examine these issues, tone sandhi, a tonal alternation phenomenon in which a tone changes to a different tone in certain phonological environments, is investigated. This dissertation first examined how productive Mandarin tone 3 sandhi words (T3 → T2/___T3) are processed and represented. An auditory priming lexical decision experiment was conducted in which each disyllabic tone 3 sandhi target was preceded by a tone 2 monosyllable (surface-tone overlap), a tone 3 monosyllable (underlying-tone overlap), or an unrelated monosyllable (unrelated control). Lexical decision RTs showed a tone 3 (underlying-tone overlap) facilitation effect for both high and low frequency words. A second priming study investigated the processing and representation of the more complex and less productive Taiwanese tone sandhi. Lexical decision RTs, examining sandhi 24 → 33 and 51 → 55, showed that while both sandhi types exhibited facilitatory priming effects, underlying tone primes showed significantly more facilitation than surface primes for sandhi 24 → 33, while surface tone primes showed significantly more facilitation than underlying primes for sandhi 51 → 55, with both effects modulated by frequency. A third study used event-related potentials (ERPs) to examine Mandarin tone 3 sandhi. Using an oddball paradigm, participants passively listened to either Tone 2 standards ([tʂu2 je4] /tʂu2 je4/), Tone 3 standards ([tʂu3 je4] /tʂu3 je4/), Tone Sandhi standards ([tʂu2 jen3] /tʂu3 jen3/), or Mix standards (i.e., both tone 3 sandhi and tone 3 words), occasionally interspersed with a tone 2 word [tʂu2] (i.e., the deviant). Results showed a mismatch negativity (MMN) in the Tone 2 condition but not in the Sandhi condition, suggesting different neural processing mechanisms for Tone 2 and Sandhi words. Together, the current data suggest that the underlying tone contributes more to the processing of productive tone sandhi and the surface tone contributes more to the processing of less productive tone sandhi. In general, this dissertation provides evidence for the representation and processing of words that involve phonological alternation, both within the same language and across different languages.