Reviewing archaeological research in the Willandra Lakes area over the past forty years shows a number of methodological and conceptual problems. One of these is the use of ethnographic models of Aboriginal behaviour to organise the archaeological data. Such models privilege spatial over temporal relationships and minimise the extent of archaeological changes over time. Of research to date, it is Bowler's correlations of archaeological and hydrographic changes since the Late Pleistocene which best match the limited archaeological data available. A comparison of the Pleistocene archaeology from the Willandra Lakes and Tasmania reveals time-averaged records that are amenable to the analysis of patterns of mobility and material use, but not of functionally defined sites nor of typologically defined artefact assemblages. The periodicity and intensity of human occupation of the Willandra Lakes can be studied using large samples of stone artefacts drawn from wide areas, allied with a strategy of dating of land units based on multiple C14 and Optical dates. Such a strategy requires the reconsideration of the relationships between behaviour, function, space and time and a reformulation of much of our existing knowledge of the archaeology of Pleistocene Australia.