Studies of hominin dental morphology frequently consider accessory cusps on the lower molars, in particular those on the distal margin of the tooth (C6 or distal accessory cusp) and the lingual margin of the tooth (C7 or lingual accessory cusp). They are often utilized in studies of hominin systematics, where their presence or absence is assessed at the outer enamel surface (OES). However, studies of the enamel-dentine junction (EDJ) suggest these traits may be more variable in development, morphology and position than previously thought. Building on these studies, we outline a scoring procedure for the EDJ expression of these accessory cusps that considers the relationship between these accessory cusps and the surrounding primary cusps. We apply this scoring system to a sample of Plio-Pleistocene hominin mandibular molars of Paranthropus robustus, Paranthropus boisei, Australopithecus afarensis, Australopithecus africanus, Homo sp., Homo habilis and Homo erectus from Africa and Asia (n = 132). We find that there are taxon-specific patterns in accessory cusp expression at the EDJ that are consistent with previous findings at the OES. For example, P. robustus M1s and M2s very often have a distal accessory cusp but no lingual accessory cusp, while H. habilis M1s and M2s show the opposite pattern. The EDJ also reveals a number of complicating factors; some apparent accessory cusps at the enamel surface are represented at the EDJ only by shouldering on the ridges associated with the main cusps, while other accessory cusps appear to have little or no EDJ expression at all. We also discuss the presence of double and triple accessory cusps, including the presence of a double lingual accessory cusp on the distal ridge of the metaconid in the type specimen of H. habilis (OH 7–M1) that is not clear at the OES due to occlusal wear. Overall, our observations, as well as our understanding of the developmental underpinnings of cusp patterning, suggest that we should be cautious in our comparisons of accessory cusps for taxonomic interpretations.