This study revisits the measurement of vowel contrast by evaluating the Pillai score with a sample-size-adjusted threshold (P!") as a categorical criterion. We examined the assumption that vowels are contrastive when the Pillai score exceeds the P!" threshold as derived from the Stanley and Sneller (2023) ’s contrapositive. Six adjacent English vowel pairs (/i–ɪ/, /ɛ–æ/, /u–ʊ/, /ɑ–ɔ/, /ɑ–ʌ/, /ɔ–ʌ/) produced by 36 native speakers in two speech corpora were analyzed through MANOVA on normalized F1, F2, and duration. Pillai scores of six vowel pairs were calculated for each speaker and then converted to binary values (P#%), coded as 1 if the score exceeded P!" and 0 otherwise. Results showed that /i–ɪ/ and /ɑ–ʌ/ consistently exceeded the threshold across all speakers, while the /u–ʊ/ pair exhibited the lowest rate of threshold achievement as found in 27 out of 36. This indicates that 27 native speakers did not produce these vowels contrastively, reflecting unstable phonemic status of the /u–ʊ/ pair. Logistic regression confirmed that all vowel pairs were significantly more likely to yield P#% =1. We concluded that native speakers tend to produce vowel pairs contrastively, although some speakers exhibited merged productions for certain pairs. The application of the Pillai score with its P!" threshold to vowel contrast analysis provides a unified criterion for assessing vowel contrasts in English. This approach could be further applied in L2 contexts to eliminate the need for direct native-speaker comparison.
Han et al. (Sat,) studied this question.