Environmental interpretation of fossil assemblages requires an accurate reconstruction of the community from which the assemblage was derived, which in turn depends on the quality of a comparative model usually based on the study of modern equivalents. The degree of inaccuracy introduced by taphonomic and other types of bias is often difficult to assess and the suitability of comparative models has rarely been addressed in this light. Here we apply a recently developed method to assess the bias present in a range of key hominin bearing localities from the Neogene of East and South Africa. The ecological structure of several of the investigated faunas can be shown to depart substantially from that of a comprehensive range of modern comparative faunas. Bias, where present, affects primarily the small mammals, which tend to be under-represented, and the large primary consumers, which tend to be over-represented. This has potentially significant implications for past and future palaeoecological reconstruction of these localities as numerous methods that are currently in use rely extensively on either the small mammals or the large primary consumers, and in particular the bovids. Understanding the nature of the bias, when present, will go some way towards improving the quality of environmental reconstructions.