Prey age-at-death is often used to assess hunting strategies, but the age profiles of animals represented in Middle Palaeolithic ungulate assemblages vary through time and across space. Understanding this diversity is important for reconstructing human behavioural evolution. To explore this issue, I studied one species, red deer (Cervus elaphus), in late Acheulean and Middle Palaeolithic assemblages from France (Combe-Grenal, Lazaret, and Piégu) and northern Spain (Gabasa 1 and Labeko Koba). I reconstructed age distributions using three methods: histograms, boxplots of median crown heights, and modified triangular plots, and two tooth combinations: dp4/p4 and dp4/m1. Two modern samples served as comparative baselines: a sample of C. elaphus killed by modern hunters in western Montana, USA and a sample of C. elaphus hunted by wolves reintroduced into Yellowstone National Park, Wyoming, USA. The mortality distributions of the modern samples differ dramatically. The wolf-hunted sample contains more juvenile and old individuals, while the human-hunted sample has more young adults. The fossil age structures fall into two clusters: 1) Gabasa 1 and Labeko Koba with a high proportion of juvenile red deer and few prime-aged adults, and 2) Combe-Grenal, Lazaret, and Piégu with equal abundances of juvenile and prime-aged red deer. I consider possible explanations for this pattern. More sites and more detailed analyses are needed to determine conclusively that the difference is related ancient human behaviour and not other pre- and post-depositional processes. However, the assemblage from Gabasa Cave is important for demonstrating that deciduous teeth can be well preserved in the Late Pleistocene fossil record. Copyright © 2004 John Wiley & Sons, Ltd.