Three proboscidean species inhabited Africa in the middle Pleistocene: Loxodonta atlantica, Loxodonta africana, and Elephas recki. L. africana can easily be distinguished from L. atlantica and E. recki by the broad, lozenge-shaped wear surfaces of its molars. L. atlantica and E. recki are more difficult to distinguish on the basis of molar characteristics. These two species rarely co-occur in the African fossil record, with L. atlantica in northern and southern Africa and E. recki in eastern equatorial Africa. Consequently, taxonomic assignment of isolated molars from these species may have been based, in some cases, on geography rather than morphology. Elandsfontein (EFT), a middle Pleistocene locality on the west coast of South Africa, has produced hundreds of skeletal elements referred to L. atlantica. Among these are 15 complete or nearly complete permanent molars, one of which is newly recovered and has not yet been referred to a species. The last taxonomic revision of EFT proboscidean material was done in the 1970s, and additional material has been recovered since then, so an updated assessment could yield new insights into the biogeography and evolutionary history of middle Pleistocene African elephants. With this goal in mind, standard molar characteristics (crown height, width, length, enamel thickness, lamellar frequency, number of enamel plates, and hypsodonty index) were recorded for EFT molars, and characteristics of M3s (n=5) were evaluated against diagnostic characteristics for L. atlantica and E. recki. EFT molars were in general more similar to L. atlantica than to E. recki, but they exhibited features of both species and could not definitively be assigned to one. Principal components analysis (PCA) was conducted to compare EFT M3s to those of E. recki and L. atlanticadescribed in the literature. Variables used in the PCA were those not substantially affected by tooth incompleteness: height, width, enamel thickness, and average lamellar thickness (a measurable reflection of lamellar frequency). The first principal component sorted individuals by species and showed that the newly recovered M3 was more similar to L. atlantica than to E. recki, but species assessment based on PCA was unreliable because the distinction between species was unclear on the first or any axis. Qualitative criteria may be more useful for identifying species based on molars; for example, E. recki exhibits irregular enamel folding, usually not present on L. atlantica and not present on EFT molars. Overall, there is little reason to revise taxonomic assignment of EFT specimens or to refer the new molar to E. recki. However, if not for geographic differences and based on molar criteria, E. recki and L. atlantica might be described as members of the same genus.