The objectives of this study were to describe follow-up dynamics in a longitudinal study on aging conducted in Spain between 1993 and 1997, and to identify the demographic, behavioral and health characteristics of persons who would later refuse to continue participating, move out of the area or be hard to locate (i.e., become lost to follow-up subjects). Data from the 1993 baseline survey were used to predict the probabilities of being lost to follow-up in the 1995 and 1997 waves. Structural multiple logistic regressions were fitted and mean probabilities were estimated to identify patterns of loss to follow-up. After 4 years, 52% of baseline participants remained in the study, 24% had died, 17% refused to continue participating, and 8.7% were impossible to locate. In the multivariate analysis, advanced age and living alone were independent predictors of loss to follow-up, and none of the health status variables remained significant. However, participation status in previous waves and the number of nonresponse items were strong independent predictors of further non-participation. Our results suggest that an attitude against participation in surveys may be an independent predictor of losses to follow-up and efforts should be made to retain this subgroup of the population in the study. Further research on reasons why people are unwilling to participate in surveys and strategies to retain people in longitudinal studies is needed.