Breakdown of food by early fossil primates, assessed with the aid of a machine that simulates mastication

To infer diet in early primates, I built a machine that simulates mastication. This machine models the actions of masticatory muscles on the thick-tailed bushbaby, Otolemur crassicaudatus. The masticatory forces in this species are well known and mastication is primitive for Primates. Nineteen post-canine dentitions---each representing a separate species---were used separately in this machine to test fourteen kinds of food. Masticatory performance was assessed by determining the size distributions of chewed fragments.; Three commonly used methods for inferring diet (referred to as Seligsohn's method, Kay's method and Evans and Sanson's method) from molar morphology were applied to the test dentitions. To assess their predictive power, results obtained using these methods were compared to masticatory performances.; Methods that measure the sharpness of cusps and cusp tips are accurate, but apply to few foods and dental features. Methods that integrate several molar dimensions apply to several foods and dental features but are somewhat inaccurate.; Several hypotheses about the diets of early primates are questioned. Late Paleocene Plesiadapiformes (?Primates) were probably not primarily insectivorous; they probably concentrated on plant parts, especially nuts, seeds and leaves. Some adapids (Adapidae, Primates), some omomyids (Omomyidae, Primates) and the late-occurring plesiadapiform Phenacolemur praecox probably concentrated on fleshy fruit. Other adapids and omomyids probably were insectivore-graminivores. Relative to plesiadapiforms, Eocene true primates are marked by greater dietary specialization. Performances of extant test species reflected their known diets. Flat dentitions are not specialized for fruit: pointy dentitions are not specialized for insects.; This experiment provides an independent means to infer diet and it tests established methods that rely on dental morphology alone.