Knowledge of the dispersal status of group members is important to understanding how sociality may have evolved within a species. I assessed the effectiveness of four techniques for elucidating dispersal behaviour in a rock-dwelling rodent (Ctenodactylus gundi) with small group sizes (2-10 animals): genetic parentage assignment, haplotype data and kinship analyses, assignment testing, and F-statistics. The first two methods provided the greatest insight into gundi dispersal behaviour. Assignment testing and F-statistics proved of limited use for elucidating fine-scale dispersal, but could detect large-scale patterns despite low sex-biased dispersal intensity (1.9 : 1) because of moderate genetic differentiation among groups (F ST = 0.10). Findings are discussed in light of current dispersal theory. In general, gundi dispersal is plastic, and seems to be dependent on body weight (for males), group composition, and scale of analysis (total dispersal events recorded within the population were almost twice the immigration rate into the population). Most groups were comprised of a single matriline and one immigrant male. Immigrant rather than philopatric males bred with group females. Dispersal among groups was male-biased, but dispersal or philopatry could occur by either sex. During a drought, both sexes delayed dispersal and cooperative social units formed. Whether such behaviour resulted directly from the drought or not remains unclear, however, since comparative information was not available from nondrought years. Combining fine-scale analyses with information on large-scale patterns provided substantial insight into gundi dispersal behaviour despite the limited movement of animals during a drought, and may prove useful for elucidating dispersal behaviour in other social animals.