This work is an attempt to evaluate different experimental methods, ABX vs. AXB, and the use of reaction time (RT) measurement in assessing perceptual sensitivity to phonemic similarity based on perceptual representation of Thai initial consonants. Thirty phoneme pairs are selected to represent varying degrees of similarity: highly similar, moderately similar, and clearly distinct. All the phoneme pairs are presented in noise in ABX and AXB tasks to twenty-two normal hearing Thai listeners. Order of the two tasks is counter-balanced across listener groups. Percent correct responses (p(C)), RTs, and preference rating are collected. The findings show that, p(C) is significantly higher in AXB than ABX despite no significant difference in RT values. In both ABX and AXB, listeners' p(C) across 3 levels of similarity varies significantly with the highest score in the clearly distinct group, and lowest score in the highly similar group. RT values across the 3 levels follow similar patterns but are not always statistically significant. ABX and AXB tasks could systematically be used to assess perceptual representation of speech sounds, with AXB eliciting higher p(C) and preference rating. It is suggested that some irregular patterns found in one part of the RT data may reflect some perceptual sensitivity pertaining to perceptual phoneme-cluster boundary.

Index Terms: Thai, initial consonant, perceptual similarity/distance, AXB, ABX, reaction time