The morphological and ecological diversity of lemurs and lorisiformes once rivaled that of the rest of the primate order. Here, we assemble a dataset of 3D models representing the second mandibular molars of a wide range of extant and fossil strepsirrhines encompassing this diversity. We use these models to distill quantitative descriptors of tooth form and then analyze these data using new analytical methods. We employ a recently developed dental topography metric (ariaDNE), which is less sensitive to details of random error in 3D model quality than previously used metrics (e.g., DNE); Bayesian multinomial modeling with metrics designed to measure overfitting risk; and a tooth segmentation algorithm that allows the shapes of disaggregated tooth surface features to be quantified using dental topography metrics. This approach is successful at reclassifying extant strepsirrhine primates to known dietary ecology and indicates that the averaging of morphological information across the tooth surface does not interfere with the ability of dental topography metrics to predict dietary adaptation. When the most informative combination of dental topography metrics is applied to extinct species, many subfossil lemurs and the most basal fossil strepsirrhines are predicted to have been primarily frugivorous or gummivorous. This supports an ecological contraction among the extant lemurs and the importance of frugivory in the origins of crown Strepsirrhini, potentially to avoid competition with more insectivorous and folivorous members of Paleogene Afro-Arabian primate faunas.