Previous studies have shown that Korean high vowels are highly likely to be devoiced between fricatives and aspirated consonants based on acoustic and EMG data [Hirose (1974); Kim et al. (1993); Jun and Beckman (1993, 1994)]. In this paper, intraoral air pressure during vowel devoicing was examined together with acoustic data. Intraoral pressure was collected by inserting a sense tube Gaeltec S7b through the nostrils to the oropharynx, and the data were measured using the high-speed speech analysis system at RILP, Univ. of Tokyo. Three native Korean speakers repeated test words [siC] in a phrase at a normal and fast rate. The results show that there are four types of intraoral air pressure configurations: (1) completely devoiced vowel---a single peak rising during [s] and falling after C; (2) completely voiced vowel---a small peak during [s] and a fall during [i] and a bigger peak during C; (3) partially devoiced vowel, I---a single peak with a plateau or shoulder during [i]; (4) partially devoiced vowel, II---a high peak from [s] to C with a few quasiperiodic waves during [i]. More aerodynamic measurements based on EGG, airflow, and oral air pressure will be examined to understand the glottal gestures during the partially devoiced vowels.