"Vowel devoicing" occurs more frequently in the Tokyo dialect (standard Japanese) than in the Osaka dialect. Our previous study showed that devoicing was a) highly frequent when a stop or an affricate was at least on one side, b) less frequent when fricatives were on both sides, and c) infrequent when there was a following /h/. In order to clarify why asymmetric effect of consonantal kind on vowel devoicing arises especially in C2 position, characteristics of glottal opening were examined using photoelectric glottography (PGG). The results showed that glottal opening degree in C2 position, in coordination with vocal tract constriction, can account for such difference of devoicing rate. For the devoiced tokens, glottal opening degree was comparable to that of a single consonant for the Osaka speaker, while it was larger than that of a single consonant for the Tokyo speaker. This fact suggests that the mechanism of devoicing differs between Tokyo and Osaka dialects.