Aim: The geological and palaeo-climatic forces that produced the unique biodiversity in the Red Sea are a subject of vigorous debate. Here, we review evidence for and against the hypotheses that: (1) Red Sea fauna was extirpated during glacial cycles of the Pleistocene and (2) coral reef fauna found refuge within or just outside the Red Sea during low sea level stands when conditions were inhospitable. Location: Red Sea and Western Indian Ocean. Methods: We review the literature on palaeontological, geological, biological and genetic evidence that allow us to explore competing hypotheses on the origins and maintenance of shallow-water reef fauna in the Red Sea. Results: Palaeontological (microfossil) evidence indicates that some areas of the central Red Sea were devoid of most plankton during low sea level stands due to hypersaline conditions caused by almost complete isolation from the Indian Ocean. However, two areas may have retained conditions adequate for survival: the Gulf of Aqaba and the southern Red Sea. In addition to isolation within the Red Sea, which separated the northern and southern faunas, a strong barrier may also operate in the region: the cold, nutrient-rich water upwelling at the boundary of the Gulf of Aden and the Arabian Sea. Biological data are either inconclusive or support these putative barriers and refugia, but no data set, that we know of rejects them. Genetic evidence suggests that many endemic lineages diverged from their Indian Ocean counterparts long before the most recent glaciations and/or are restricted to narrow areas, especially in the northern Red Sea. Main conclusions: High endemism observed in the Red Sea and Gulf of Aden appears to have multiple origins. A cold, nutrient-rich water barrier separates the Gulf of Aden from the rest of the Arabian Sea, whereas a narrow strait separates the Red Sea from the Gulf of Aden, each providing potential isolating barriers. Additional barriers may arise from environmental gradients, circulation patterns and the constriction at the mouth of the Gulf of Aqaba. Endemics that evolved within the Red Sea basin had to survive glacial cycles in relatively low salinity refugia. It therefore appears that the unique conditions in the Red Sea, in addition to those characteristics of the Arabian Peninsula region as a whole, drive the divergence of populations via a combination of isolation and selection.