Dartmoor, in SW England, is a classic periglaciated granite upland adorned with a population of over 150 tors. The origin of the tors has been controversial, but their emergence by differentiation after stripping of regolith during Pleistocene cold phases is accepted. However, their actual age has been unknown, with possible scenarios ranging from preservation since the early Middle Pleistocene to relatively short-lived landforms in a maritime climate with high denudation rates. The latter is now supported by 32 cosmogenic surface exposure dates from 28 tors across the whole upland. The distribution of apparent 10Be ages peaks strongly in the Middle Devensian (36-50 ka), which with corrections for weathering and limited ice shielding could be interpreted as Early Devensian. These ages are much younger than those found for three glacially unmodified Cairngorms tors, and somewhat younger even than glacially modified Cairngorms tors. The dates show little spatial variation. Although an ice cap has now been modelled in the heart of northern Dartmoor, tors here are of median age, suggesting that ice cover sufficient to shield tors from incoming radiation was of short duration. The few younger tor ages support the idea of continuing landform instability across the landscape, with weathering flakes redeveloping soon after inferred loss of top pillows by gelifraction or gravitational toppling. The few older tor ages have no systematic explanation, and may indicate inheritance from an earlier cycle of bedrock near-exposure. Since most tors are modest in height (typically 2-5 m), volumetrically insignificant, and often in advanced stages of disintegration, the general impression is that they are evanescent features, which emerge and quickly disappear during every Pleistocene climatic downturn. Tor populations may thus flicker across the landscape rather randomly over the Quaternary. The remarkably consistent age of the present tor population could be associated with a stripping event at the start of the Devensian, but fuller analysis must await closer controls on tor denudation rates in different climatic phases, and on ice cover extent and duration. These results only date extant tor surfaces, not the landscape, but as the best available erosion pins they have evident value in exploring theories of the evolution of Dartmoor during the Quaternary.