Abstract: “Hidden populations” are characterized by their relative small sizes and the unwillingness of the members to be identified. Therefore， it is very difficult to study them by the regular sampling methods. A new sampling method known as respondentdriven sampling （RDS） has emerged in recent years to study hidden populations. Based on the traditional “snowball” sampling method and with the help of social network analysis theories and methods， RDS enables researchers to properly estimate the characteristics of a hidden population from its sample（s）. This paper briefly describes the theoretical and methodological backgrounds， basic ideas， major principles， and operational procedures of RDS. It also discusses the problem areas when RDS is actually practiced.

Key words: respondent driven sampling, hidden population, research methodology