The vowel devoicing of (Standard) Japanese is a well-studied phenomenon in which high vowels are pronounced devoiced when flanked by voiceless obstruents: sUkiyaki, kItakaze 'northwind' (devoiced vowels are capitalized). Less studied is the vowel devoicing of Ainu, a genetically isolated language and the geographic neighbor of the Japanese. The current research revealed that the vowel devoicing of Ainu exhibits characteristics that are unknown or rare to vowel devoicing of Japanese, in contrast with previous studies which claimed that the phonological environment of the phenomenon should not differ in both languages. Examining the sound recordings of the Saru Dialect of Ainu, I found that the speakers never devoiced vowels in the syllables tu, pi and pu (yaytupareno 'carefully', akohepututu 'to pout'), while these are devoiceable syllables in Japanese: pIkupIku 'jerkingly', resepUshon 'reception'. The Ainu data suggest that the vowel devoicing in this language cannot be simply due to ease of articulation (saving vocal cord vibration throughout the syllable in question), an explanation often preferred for the vowel devoicing of Japanese. This deviation suggests that vowel devoicing of Ainu might be governed by distinct mechanisms.