In ABX discrimination and similarity ranking, the perceptual status of phonological and phonetic categories was assessed. The results are summarized as follows: first, phoneme pairs were fully discriminated in ABX discrimination and similarity ranking ...
In ABX discrimination and similarity ranking, the perceptual status of phonological and phonetic categories was assessed. The results are summarized as follows: first, phoneme pairs were fully discriminated in ABX discrimination and similarity ranking tasks, while allophonic pairs and phonetic variant pairs were less saliently discriminated than phonemic pairs. However, more importantly, Korean listeners distinguished the phonetic differences of allophones, even though they discriminated such allophonic pairs below the chance level. There were significant differences in the scores between allophone pairs and control pairs in the similarity ranking task. Similarly, the phonetic variant pair with 130 ms of VOT was rated as more different than that with 100 ms of VOT. These differences were not likely to be due to the raw phonetic differences between the stimuli as proposed in previous studies such as Jaeger (1980, 1986), Pegg & Werker (1997) and Whalen et al. (1997) where two allophones were argued to be distinguished only at a much lower level. The similarity ranking task is usually used to capture the phonological processing, not surface phonetic processing (Boomershine et al. in press), and thus those significant differences between the allophonic pair and the identical pair on the one hand, and between the phonetic variant with a relatively large amount of acoustic cues and that with a small amount of acoustic cues on the other suggest that they are not simply a result of the raw overall auditory qualities of the sounds, but are strongly tied to the phonological system of the native language of the listeners. However, it is interesting to note that the results for discrimination and similarity ranking were not consistent. In similarity ranking, Korean listeners’ perception of the two types of phonetic variant pairs was sensitive to the amount of aspiration: the pair with 130 ms of VOT was rated as more different than that with 100 ms of VOT. However, in discrimination, there were no significant differences between the pair with 130 ms of VOT and that with 100 ms of VOT.
Second, the results of the present study showed that overall real word pairs showed higher mean correct scores in discrimination than non-word pairs. Phonemic pair types, whose categories were already robustly established, were not overridden by lexicality effect, but in allophonic and phonetic variant pair types, lexicality played a crucial role in discrimination: real words were more successfully discriminated than non-words.
Finally, discrimination seemed to be only partly affected by the episodic information such as frequency and recency. It was expected in Exemplar Theory that more recently given words and high frequency words were easier in discrimination, while in SPE, such episodic information was expected to be trivial in discrimination. However, it was shown that there was no better performance in the case where the target token was presented more recently; only in allophone pair type words the high frequency word was easier in discrimination than the low frequency word. Thus the expectation from Exemplar Theory was not borne out in the case of episodic information of recency, and information from frequency only influenced the discrimination of allophonic pairs.
Based on these results, it was shown that both models are successful in predicting one of another part of our findings relating to the phonological and phonetic categories such as phonemes, allophones, and phonetic variants. However, Exemplar Theory seems to be better in accounting for a wider range of results obtained in the present study.