Here is a paper of pivotal importance to all prehistorians attempting to reconstruct societies from assemblages of shells or stone artefacts in dispersed sites deposited over tens of thousands of years. The authors demonstrate the perilous connections between the distribution and content of sites, their geomorphic formation process and the models used to analyse them. In particular they warn against extrapolating the enticing evidence from Pleistocene Willandra into behavioural patterns by drawing on the models presented by nineteenth-century anthropologists. They propose new strategies at once more revealing and more ethical.