There is general agreement that the diet of early hominins underwent dramatic changes shortly after the appearance of stone tools in the archaeological record. It is often assumed that this change is associated with dietary expansion to incorporate large mammal resources. Although other aspects of the hominin diet, such as aquatic or vegetal resources, are assumed to be a part of hominin subsistence, identifying evidence of these adaptations has proved difficult. Here we present a series of analyses that provide methodological support for the inclusion of aquatic resources in hominin dietary reconstructions. We suggest that bone surface modifications in aquatic species are morphologically distinguishable from bone surface modifications on terrestrial taxa. We relate these findings to differences that we document in the surface mechanical properties of the two types of bone, as reflected by significant differences in bone surface microhardness values between aquatic and terrestrial species. We hypothesize that the characteristics of bone surface modifications on aquatic taxa inhibit the ability of zooarchaeologists to consistently diagnose them correctly. Contingently, this difficulty influences correspondence levels between zooarchaeologists, and may therefore result in misinterpretation of the taphonomic history of early Pleistocene aquatic faunal assemblages. A blind test using aquatic specimens and a select group of 9 experienced zooarchaeologists as participants was designed to test this hypothesis. Investigation of 4 different possible explanations for blind test results suggest the dominant factors explaining patterning relate to (1) the specific methodologies employed to diagnose modifications on aquatic specimens and (2) the relative experience of participants with modifications on aquatic bone surfaces. Consequently we argue that an important component of early hominin diets may have hitherto been overlooked as a result of (a) the paucity of referential frameworks within which to identify such a component and (b) the inability of applied identification methodologies to consistently do so.