Attrition, or the loss of study subjects, confounds the subject identification process. Design attrition occurs when subjects initially involved in the study are excluded because of changes in study focus or rationale. Data collection attrition occurs when subjects are excluded because they were not contacted by the data gatherer, or because the contact did not yield the information required for analysis. Data processing attrition occurs when subjects included in the design fail to be included in the study data records. Analysis attrition occurs when subjects included in the design are not deemed valid subjects for the analytic design. Using the evaluation of Project Follow Through as an example, attrition in the data collection process is discussed. The point is made that the data collection staff can play an important role in minimizing attrition in longitudinal studies by gaining the cooperation of subjects, by correctly locating and identifying subjects, and by the use of proper management and administration procedures. (BW)