Irruptive population dynamics are characteristic of a wide range of fauna in the world's arid (dryland) regions. Recent evidence indicates that regional persistence of irruptive species, particularly small mammals, during the extensive dry periods of unpredictable length that occur between resource pulses in drylands occurs as a result of the presence of refuge habitats or refuge patches into which populations contract during dry (bust) periods. These small dry-period populations act as a source of animals when recolonisation of the surrounding habitat occurs during and after subsequent resource pulses (booms). The refuges used by irruptive dryland fauna differ in temporal and spatial scale from the refugia to which species contract in response to changing climate. Refuges of dryland fauna operate over timescales of months and years, whereas refugia operate on timescales of millennia over which evolutionary divergence may occur. Protection and management of refuge patches and refuge habitats should be a priority for the conservation of dryland-dwelling fauna. This urgency is driven by recognition that disturbance to refuges can lead to the extinction of local populations and, if disturbance is widespread, entire species. Despite the apparent significance of dryland refuges for conservation management, these sites remain poorly understood ecologically. Here, we synthesise available information on the refuges of dryland-dwelling fauna, using Australian mammals as a case study to provide focus, and document a research agenda for increasing this knowledge base. We develop a typology of refuges that recognises two main types of refuge: fixed and shifting. We outline a suite of models of fixed refuges on the basis of stability in occupancy between and within successive bust phases of population cycles. To illustrate the breadth of refuge types we provide case studies of refuge use in three species of dryland mammal: plains mouse (Pseudomys australis), central rock-rat (Zyzomys pedunculatus), and spinifex hopping-mouse (Notomys alexis). We suggest that future research should focus on understanding the species-specific nature of refuge use and the spatial ecology of refuges with a focus on connectivity and potential metapopulation dynamics. Assessing refuge quality and understanding the threats to high-quality refuge patches and habitat should also be a priority. To facilitate this understanding we develop a three-step methodology for determining species-specific refuge location and habitat attributes. This review is necessarily focussed on dryland mammals in continental Australia where most refuge-based research has been undertaken. The applicability of the refuge concept and the importance of refuges for dryland fauna conservation elsewhere in the world should be investigated. We predict that refuge-using mammals will be widespread particularly among dryland areas with unpredictable rainfall patterns.