A study entitled the Alternative Training Experiment (ATE) evaluated the effectiveness of a method for training census enumerators that involved new processes for developing instructional materials (called job performance aids) and new methods of instruction. During the study three pairs of census decentralized offices were matched on variables related to the difficulty of enumeration. Half the trainees from each office were given the experimental or Job-Performance-Aided (JPA) Training, while the remaining trainees were trained via conventional training methods. Based on enumerator reactions collected immediately after the training and again after two to three days of job experience, researchers found a slight but consistent preference for JPA training. A second study compared the job performance of the two groups of participants on a variety of performance measures. After analyzing data from end-of-training knowledge tests, self-report measures, and quality control records obtained as part of normal field operations, researchers decided that no conclusions about the effects of training on job performance were possible. However, they did make several recommendations concerning training development and research, including designing training packages with some flex time, and making the design of training and field manuals an integrated effort. (MN)