This paper evaluates the potential contribution(s) of faunal analysis to hominin palaeoecology at regional and continental scales, through an explicit investigation of the values, methods and conceptual frameworks of palaeoanthropology and their compatibility with real data structures. It employs a problem-framing method developed in policy-relevant science to establish a suitable research design for ?large scale? faunal analysis, before testing the method in a pilot study of 48 faunal assemblages from the African Plio-Pleistocene. Hitherto, taphonomic bias has discouraged attempts to study faunal assemblages on large spatiotemporal scales, and most scientists have restricted their work to the smaller (site or local) scale and/or a subset of the total fauna. Furthermore, palaeoanthropological studies of fauna tend to address pre-determined questions through analysis of statistical outputs (patterns), rather than investigating the limitations and potential of the data through exploratory work. This paper, despite identifying a number of inherent constraints on palaeocommunity analysis at the large scale?including a clear tendency towards the segregation of faunal assemblages along taphonomic and geographic lines?successfully defines palaeocommunities and identifies systematic variation in their distribution in several regional datasets and at the continental scale. It suggests that the potential viability of faunal analyses for a given project could be made empirically testable, and further work on the lines defined here might provide insight into the impacts of taphonomy and ecology at the large scale. Although there are conceptual and methodological problems associated with large-scale faunal analyses, this paper suggests that they could provide some insight into hominin environments, evolutionary ecology and biogeography as part of a holistic, multi-scale approach to our lineages' history.