In a longitudinal study, subject attrition constitutes a potential threat to the usefulness of data collected from subjects remaining in later waves of the study. If attrition is selective, then the sample at later waves of the study may be representative of a different population than that represented at the outset of the study. This study examined selective attrition between the first three waves of the Medicine, Health, and Aging study, a study which explored the health and psychosocial characteristics of persons enrolled in a state-wide pharmaceutical assistance program for senior citizens in Pennsylvania and compared these characteristics to persons not enrolled in the program. Subjects were 1,803 older adults who were participating in the 3-year study. Every year subjects completed a survey concerning physical and psychological health, functional well-being, and prescription and non-prescription drug use. Analyses focused on comparisons of the characteristics of returning, non-returning, and deceased subjects from wave to wave of the study. The results indicated that attrition and mortality were related to a set of health status, drug use, and functional status characteristics; in general, attrited subjects had experienced lower levels of health and functioning than had returning subjects, and before their death, deceased subjects appeared to have been functioning at levels lower than the non-returning subjects. The findings suggest different processes which may underlie attrition and mortality. Thirteen references, three tables and four figures are included. (Author/NB)