Late Pleistocene records of island settlement can shed light on how modern humans (Homo sapiens)adapted their behaviour to live on ecologically marginal landscapes. When people reached Sahul(Pleistocene New Guinea-Australia), between 65 and 50 ka, the only islands they would have encoun-tered were in the tropical north. This unique geographic situation therefore offers the only possibility ofmodelling human adaptive behaviour to islands in Australasia during the Late Pleistocene. Cave exca-vation on the uplifted limestone island of Panaeati in the Massim region of Southeastern New Guinearevealed a cultural sequence commencing from 17,300e16,800 cal. BP, suggesting habitation of highercoastlines occurred as low-lying shorelines destabilised during the initial stages of deglacial sea-levelrise. No cave use was evident between 12,400 and 4780 cal. BP when the continental shelf was fullyinundated, and Panaeati reduced in size by 90%. It is likely that diminished coastlines and the reducedresources of low-lying islands could no longer support pre-agricultural populations during this time.Cultural groups that were better adapted to living on small islands returned to Panaeati by 4780e4490 cal. BP when sea levels had stabilised, lagoons formed, and coastal ecosystems had diversified.Investigations demonstrate the role of larger islands as refugia during deglacial sea-level rise and theeffects on human dispersals and cultural diversity.