ISCA Workshop on Plasticity in Speech Perception (PSP2005)

Senate House, London, UK
June 15-17, 2005

Lexically-Driven Perceptual Adjustments of Vowel Categories

James McQueen, Holger Mitterer

MPI for Psycholinguistics, Nijmegen, The Netherlands

Previous research has examined lexically-driven perceptual learning about consonants. Norris, McQueen & Cutler (2003) used an exposure-test paradigm. Listeners exposed to an ambiguous fricative in lexical contexts which biased its interpretation towards [f] identified more sounds on an [Ef]-Es] test continuum as [f] than listeners exposed to the same ambiguous sound in [s]-biased lexical contexts. Listeners can thus use lexical knowledge to adjust consonant categories while listening to a speaker who -- perhaps because of dialect differences -- produces unusual tokens of those sounds.

Dialect differences are often carried by vowels, however. We therefore tested here whether this type of lexically-driven perceptual learning occurs for vowels. We also tested whether learning generalises to the perception of other vowels. We used the same exposure-test paradigm, with critical stimuli based on the [i]-[e] height distinction. In the lexical decision exposure phase in two experiments, one group of listeners heard ambiguous vowels, midway between [i] and [e], in sequences which were Dutch words if the final syllable contained [i] but not if it contained [e] (e.g., "satell?t", from "satelliet", satellite), plus unambiguous words with [e] in the final syllable (e.g., "atleet", athlete). A second group heard the reverse (e.g., "atl?t" and "satelliet").

In Experiment 1, listeners were tested on three continua, in this order: 1. Exposed: [ift]-[eft]; 2. (or 3.) Near: [Ift]-[Eft] (also a front height contrast); 3. (or 2.) Far: [oft]-[aft] (a low back height contrast); and 4. Exposed. There was a vowel identification shift only in Block 4: Listeners exposed to [?] in [i]- biased contexts identified more sounds on the [ift]- [eft] continuum as [i] than those who heard [?] in [e]-biased contexts. In Experiment 2, these test continua were presented in a different order: 1. Near (or Far); 2. Far (or Near); 3. Exposed; 4. Exposed; 5. Near. There was a reliable lexically-driven shift on the Exposed continuum only in Block 3. Although there was again no effect for the Near continuum, there was for the Far continuum, consistent with the direction of the vowel height adjustment encouraged by the exposure conditions (i.e., more [o] responses from the listeners exposed to [?] in [i]-biased contexts than from those who heard [?] in [e]- biased contexts). These results show that lexicallydriven perceptual adjustments can be made to vowel as well as consonant categories, and can sometimes generalise from the vowel contrast heard during exposure to perception of another vowel contrast.

Full Paper

Bibliographic reference.  McQueen, James / Mitterer, Holger (2005): "Lexically-driven perceptual adjustments of vowel categories", In PSP2005, 233-236.