In a follow-up of 85 alcoholic men, 93% were interviewed or confirmed deceased 8 years after discharge from hospital. The sample was also followed up at 1 and 3 years postdischarge. Follow-up rates of other published studies are discussed. Scales to assess difficulty of interview and difficulty of location were developed. Factor analysis revealed the differential characteristics of subjects defined as difficult to locate, difficult to interview or missing. Subjects who are difficult to locate, or who in the extreme case will go missing, tended to have poorer social functioning prior to intake and to be residentially unstable during the follow-up period, characteristics that tend to correlate with worse drinking outcome. Subjects who are unwilling to be interviewed tended to be residentially stable and show better interpersonal adjustment at follow-up. Less intensive location procedures would have resulted in data loss from those classified as difficult to locate. Less persuasive interview techniques would have resulted in data loss from those classified as difficult to interview. Early termination of the follow-up would also have resulted in data loss from the latter group. The distinct types of data biases that would be introduced by loss of information from each of the above subgroups are examined.