This study investigated the devoicing rate of Japanese high vowels, focusing on regional and generational differences by acoustically analyzing vowels from two large speech databases. The first speech database used in this study was collected between 1986 and 1988 from 41 areas (prefectures) which included 607 participants (299 high school students and 308 their grandparents). The second was taken from a 2006-2007 collection from seven areas as a follow-up investigation to the first database consisting of 463 participants ranging in age from 8-90 year olds. The results revealed there is a generational as well as regional difference in the devoicing rate in almost all areas. Based on those results, a new distribution map reflecting a current devoicing rate of the younger generation was presented. Furthermore, by comparing the two data sets, this study confirmed age difference in the devoicing rate is not age-grading but a sound change in progress. This study discusses the social factors for changes in the devoicing rate of some areas and then applies the devoicing rate of five areas to an S-curve model to predict the future devoicing rate.