This paper describes how decisions were made on procedures for counting the homeless population for the 1990 census. These topics are covered: (1) reasons the Census Bureau did not specifically label anyone as "homeless" while still including homeless persons in the census; (2) how the lists of sites where homeless persons might be found were compiled; (3) enumeration procedures, including procedures for shelters, streets, abandoned buildings, places of commerce, and open public locations; (4) what happened on "Shelter and Street Night" ("S-Night"), including media activity, hiring census takers, and training and payment; (5) impressions of data quality, including the count in shelters, the count of persons visible in the streets, and coverage of "doubled-up" families; and (6) publication of data. It is concluded that the Census Bureau had an effective method that was practical within the constraints of a decennial census and that many persons were added to the census who otherwise would have been missed; however, concern is expressed about the count of persons visible in the street. It is noted that information about limitations to the data are available to all users, and certain "lessons" for future improvements learned from S-Night are outlined. In general, the report finds that selected major components of the homeless population have been included and that the 1990 Census was an opportunity to get a clearer idea of the demographic diversity of homeless persons as well as of differences among areas of the country. A glossary of terms is included. (ABL)