Studies on 'hidden populations', such as homeless people, prostitutes and drug addicts, raise a number of specific methodological questions usually absent from research involving known populations and less sensitive subjects. This paper examines the advantages and limitations of nonrandom methods of data collection such as snowball sampling. It reviews the currently available literature on sampling hard to reach populations and highlights the dearth of material currently available on this subject. The paper also assesses the potential for using these methods in nursing research. The sampling methodology used by Faugier (1996) in her study of prostitutes, HIV and drugs is used as a current example within this context.