While gathering the 1980 census, the U.S. Census Bureau tried several methods of improving its training of census interviewers. This paper covers three general areas of that experiment. First, it reviews bureau efforts to improve verbatim training and presents selected results from an alternative training experiment conducted in the 1980 census. In this experiment, a method of instruction called job-aided training was developed, was used experimentally, and was compared to census verbatim training. The job-aided training was an enhancement of verbatim training that changed it from a lecture-oriented approach to one that required active learning by the trainee. It included simplification of training materials, especially the training manual, and the addition of task analysis to the design of the materials. As a result, this training approach was more successful than others, and the bureau is planning to use it more widely for the 1990 census. Second, the paper presents an overview of training approaches being developed to support the bureau's increasing activities in centralized telephone interviewing. These approaches include self-paced instruction and classroom-housed performance-based training, especially practice interviewing exercises. Finally, the paper reviews some of the evaluation strategies being implemented to determine training effectiveness and to improve the quality of interviewing. Specifically, these strategies include selection, training, and monitoring techniques for use in a centralized telephone operation. Because of the expense of their use and the questionable results, however, such techniques may remain experimental. (KC)