This paper reports the successful completion of Japanese POLYPHONE project, Voice Across Japan (VAJ) data collection project. The database has the following characteristic, 1) large speakers database (8,866 spk.) through telephone line, 2) to gather participant's personal information such as gender, age, growing place, and so on, and 3) to put data segmented by phone or word boundary. This paper describes several aspects of Japanese dialects and also, reports the results of experiments. How much percents do dialects make influence on speech recognition. In our result, dialects makes 2-4% influence on speech recognition rate. The results are useful information for building practical speech recognition system as well as data collection.