In the process of spoken language comprehension, the listener must map acoustic-phonetic information in the waveform onto stored representations of lexical form in the mental lexicon. A major question concerns the nature of the representations which make contact with the lexicon. Are small acoustic details within a phoneme present in this representation, or are they lost at a lower level? In this research project the Voice Onset Time (VOT) of the two Dutch voiced plosives was varied to find out whether lexical access is affected by such subcategorical variation.
In Dutch there are two voiced plosives, namely /b/ and /d/. They are said to have a negative VOT, this means that the vocal cords start vibrating during the closure phase. A small production experiment was carried out to establish the natural variation in VOT for both /b/ and /d/. The results showed that VOT of Dutch voiced plosives varies considerably in natural speech and is therefore a suitable type of subcategorical variation to be studied.
To find out how variation in VOT is perceived, listeners were asked to rate the initial phoneme of several high frequency (HF) words, low frequency (LF) words and nonwords (NW) on a scale from 1 (poor exemplar) to 7 (good exemplar). The items started with a /b/ or /d/, varying from zero periods to 18 periods of prevoicing. The outcomes were used to guide the construction of three variations of both voiced plosives: normal amount of prevoicing, too much prevoicing and too little prevoicing.
In an associative priming experiment subjects were asked to perform a lexical decision task on visual targets that were preceded by an auditory prime. The primes were either HF words or LW words. The results showed a significant priming effect: reaction times to targets preceded by an unrelated prime were significantly slower than to targets preceded by a related prime. There was no difference between the first 3 conditions and there was no interaction with frequency. These results suggest that natural variation within a phoneme such as differences in VOT of voiced plosives does not influence lexical access.