Abstract: Pollen and spore analysis is the most successfully used palaeobotanical discipline for reconstructing Holocene vegetation and climate history throughout the world. Subantarctic islands are very specific areas. They are located in the circum-Antarctic Southern Ocean in latitudes that are under strong influence of the southern westerly winds, and are characterised by a treeless, phanerogam-poor flora. Palynological research on many of these islands has resulted in diverging conclusions about how to infer climate history from pollen data. In this study we compare pollen data with macrofossil data on the one hand, and the palaeoenvironmental history based on a multiproxy record on the other hand, of two peat sequences from two different subantarctic islands, South Georgia and Île de la Possession (Îles Crozet). We conclude that palynology must be used with caution as a proxy for climate change on these islands, especially when no other proxy data are available. The uplandlowland principle, as it has been applied in pollen studies in the South Indian Ocean islands, results in erroneous conclusions about climate change on Île de la Possession. More palaeoclimatic multiproxy and pollen studies, in combination with pollenvegetation relationship studies, can, however, contribute to a more reliable model of how to interpret pollen data in the Subantarctic. We want to stress that our conclusions are only based on Holocene records. Consequently, the question remains if palynology can be used as a palaeoclimatic proxy when climatic changes were more pronounced such as during the last glacialinterglacial transition.