Characteristics of nonrespondents, respondents who were easy to locate, and respondents who were hard to locate were examined with the use of data from a telephone health survey of male, US Army, Vietnam-era veterans. Of 17,867 eligible men discharged from active military duty in the late 1960s and early 1970s, 15,288 (86%) were successfully located and interviewed during 1985-1986. Veterans who could not be located were more likely than respondents to possess baseline characteristics predictive of increased mortality. In contrast, subjects who were located but refused to be interviewed were similar to respondents. Among veterans who were interviewed, those who were hardest to locate had the highest prevalence of known risk factors for diminished health status and reported many health problems with higher relative frequencies than respondents who were easier to locate. Odds ratios comparing the prevalence of each of 11 health outcomes in men who had served in Vietnam with that in men who had served elsewhere did not vary appreciably by intensity of follow-up. In particular, the subgroup of respondents that was located and interviewed within 2 weeks of initiation of follow-up (comprising 25% of all respondents) produced odds ratios for 10 of the 11 outcomes that were not appreciably different from odds ratios based on all respondents.