Respondent-driven sampling has been used to survey rare or hard-to-reach populations (Salganik & Heckathorn, 2004; Thompson, 1997). The District of Columbia (DC) Mayor’s Office of GLBT Affairs, the DC Department of Health, and the American Institutes for Research (AIR) conducted a health survey of GLBT individuals in DC, a relatively rare and often hard to reach population. Respondent driven sampling was used to construct the sample. This paper discusses details of the sampling process, the issues that had to be addressed, the characteristics of the effective sample, and lessons learned. DC government agencies identified a number of GLBT organizations in DC; many of these organizations agreed to contact a sample of their members and request that they complete the survey. Each organization drew a sample of members using a sampling interval provided by AIR. These samples represented the first stage in the sampling process. The organizations were provided a letter that was sent to each sampled member explaining the survey’s purpose, requesting participation, and giving a link to the web -based questionnaire. In addition to asking the respondents to complete the questionnaire, the letter asked them to provide the names and contact information of three friends who were part of the DC GLBT community but did not belong to their organization. These individuals comprised the second-stage sample. AIR sent letters similar to the ones sent in the first stage to these potential respondents requesting both their participation and three more names of GLBT residents of DC along with the contact information. This process continued until a sample of sufficient size was achieved. The goal of the process was to represent eligible individuals not belonging to participating organizations.