This paper rejects Matsumoto's (1984, 1995) arguments that o1 and o2 in Old Japanese (OJ) are allophones of the phoneme /o/. Matsumoto claims that a restricted distribution of the phonetically unmarked o1, its low frequency, and the anomalous direction of its merger with o2 should be regarded as denoting their status as allophones, rather than two different phonemes. The phonological distinction of vowel quantity in OJ and pre-OJ, and Short-mid-vowel-raising in pre-OJ (Hattori 1976, 1979a, b) and Vowel-shortening, which shortens the vowel of the first syllable in a disyllabic morpheme containing two long vowels in pre-OJ, can explain all the alleged anomalies and serve to invalidate Matsumoto's arguments.
