Information on the physical health of homeless adults is potentially biased either by sampling strategy or by measurement of physical health. Studies that used comprehensive health measures (self-reported and objective measures) relied on samples from shelters or hotels. However, more representative community-based studies relied on self-reports or ratings. We conducted the first study to use both a community-based sample (N = 529) and comprehensive measures of health (an interview, a limited physical examination, and blood testing). Shelter dwellers compared with homeless persons sampled elsewhere were less likely to have used illegal drugs, to have been victimized, to have injured skin, and to have elevated aspartate aminotransferase levels and mean corpuscular volumes. Sixty-two percent of persons observed to have high blood pressure were unaware of their condition. Sampling only shelter dwellers, or relying only on reports of illness by homeless adults, may mask or underestimate existent health problems that are revealed by community-based sampling techniques and more objective measures.