Because of their comparative economy, the most commonly used methods for counting the homeless focus on users of shelters, food lines, health clinics for the homeless, and other services for the homeless. This paper argues that surveys restricted to shelter users are of limited usefulness, but that joint surveys of food - line, shelter, and clinic users include very substantial proportions of the homeless in many communities. Such comprehensive surveys can provide an accurate basis for research on the homeless in communities with reasonably capacious service systems. The reliability of such surveys has grown as service systems for the homeless have improved. Groups of the homeless that tend to be missed in service user surveys include homeless youths on their own as well as substantial portions of the rural homeless population and of the doubled-up or institutionalized homeless populations. Surveys of homeless persons sleeping in known places "on the streets" can be used to supplement surveys of service users. Research in Colorado suggests that such supplemental surveys of the street population provide a practical basis for comprehensive estimates of the homeless street population.