Respondent-driven sampling (RDS) is a methodology for sampling and analysing survey data from rare and elusive populations that has gained increasing attention in migration research in recent years. Since its introduction in 1997, it has been applied in numerous surveys in the field of epidemiology (mainly for the estimation of HIV/AIDS prevalence in high-risk populations) and increasingly in migration research. This article summarises some of the experiences of applying RDS to study migrant populations, in light of some of the key assumptions that the methodology builds on, pointing to the advantages, but also the limitations of this approach. If sample frames of reasonable quality are available or can be produced, this will normally be preferable to RDS. But if no sample frame is available, the RDS methodology does create opportunities for a more systematic approach to the study of marginalized migrant populations, where qualitative studies, and convenience samples thus far have dominated knowledge production.