This study analyzed the phonetic realizations of devoiced vowels in Mandarin Chinese. First, the target vowels were acoustically examined and classified in three categories: completely devoiced, partially devoiced, and completely voiced. Second, the segmental duration and intensity of the target vowels, and the aspiration/noise duration of the preceding consonants were measured between the devoiced and voiced tokens. Our findings indicate that 1) the vowels of Mandarin Chinese were seldom realized as prototypical devoiced vowels like Japanese or the American Indian languages: the frequency of devoiced vowels was less than 5% regardless of the contexts eliciting vowel devoicing; and 2) there were considerable individual variations across speakers. However, it was also found in the acoustic measurements of the target vowels that the devoiced vowels were realized with shorter duration and lower intensity than the voiced counterparts. As for the duration of the aspiration/noise of the preceding consonants, a voiceless fricative, but not a voiceless affricate or stop, was realized with longer duration when it was immediately preceded by the (completely and partially) devoiced vowel as compared to the completely voiced vowel. The present results demonstrate a cross-linguistic variation in the realization of devoiced vowels.(Konkuk University)