Abstract
Contemporary Japanese language surveys using the random sampling method face the following six problems: (1) the difficulty of accessing the basic resident register, (2) an increasing rejection rate of interviews by the respondents, (3) an increasing necessity for time, cost, and a sampling specialist, (4) the linguistic reliability of surveys conducted by external research companies, (5) the difficulty of sampling the native speakers of the dialect in the target area, (6) the paucity of data for linguistically-oriented research. Detailed explanations were given, and possible solutions were suggested for each problem. In particular, it was proposed that by conducting surveys of different kinds (different in scales and in methodology) for the same linguistic phenomenon, one could approach the true picture of the linguistic variation by comparing the results. Examples from domestic and overseas surveys were given and discussed.