The relationship between identification and discrimination in cross-language perception

The effects of stimulus presentation contexts on cross-language consonant category identification and category goodness rating were examined in two experiments. In Experiment 1, native Korean listeners’ identification and goodness ratings of Thai stop consonants were obtained under two conditions: ‘single’ and ‘triadic’ stimulus presentations. In the ‘single’ stimulus presentation, each target Thai stop consonant was presented in isolation for categorical identification and goodness rating, while the target stimulus (X) was presented between two other stimuli (A and B) in the ‘triadic’ stimulus presentation. Korean listeners’ identification data obtained under both presentation contexts were then used to generate ‘predicted’ discrimination scores. However, Korean listeners’ ‘actual’ (AXB) discrimination scores of Thai stop consonant contrasts were also obtained. The results indicated that the two stimulus presentation conditions (i.e., ‘triadic’ and ‘single’) did not affect the choice of modal categories with which the Thai consonants were identified and that native Korean listeners had no perceptual difficulty discriminating among Thai stop consonants. However, a slightly better fit between ‘actual’ and ‘predicted’ discrimination scores derived from the ‘triadic’ identification data was observed. On the other hand, the identification of Korean stop consonants obtained from the Thai listeners in Experiment 2 showed a strong effect of stimulus presentation contexts. Specifically, the identification of Korean lax /p/, /t/, and /k/ varied depending on whether they were presented in the context of an aspirated or a tense stop in the ‘triadic’ stimulus presentation format. Nonetheless, like the Korean listeners in Experiment 1, Thai listeners had no difficulty discriminating among Korean stops. In contrast with Experiment 1, however, a much stronger correlation between ‘actual’ and ‘predicted’ discrimination scores derived from Thai listeners’ ‘triadic’ identification data was observed.