Paleontologists use fossil teeth to reconstruct the diets of early hominins and other extinct species. Some evidence is adaptive: nature selects for tooth size, shape, and structure best suited to specific food types. Other evidence includes traces left by actual foods eaten, such as microscopic tooth wear. This critical review considers how molars work, how they are used, and how occlusal topography and dental microwear can be used to infer diet and food preferences in the past, particularly for hominins of the Pliocene and early Pleistocene. Understanding that cheek teeth function as guides for chewing and tools for fracturing allows us to characterize aspects of occlusal form that reflect mechanical properties of foods to which a species is adapted. Living primates that often eat leaves, for example, have longer crests and more sloping occlusal surfaces than those that prefer hard foods. Studies of feeding ecology have shown, however, that tooth shape does not always correspond to preferred food items. It often follows mechanically challenging foods whether eaten often or rarely. Other lines of evidence that reflect actual tooth use are required to work out food preferences. Microwear textures, for example, reflect foods eaten by individuals in the past such that hard seeds and bone tend to leave complex, pitted surface textures, whereas tough leaves and meat more often leave anisotropic ones covered in long, parallel scratches. The study of fossil hominin molars shows how these various attributes are combined to infer diet and food preference in the past. A trend in occlusal morphology suggests decreased dietary specialization from Australopithecus to early Homo, and increasing dispersion in microwear complexity values is consistent with this. On the other hand, occlusal morphology may suggest dietary specialization in Paranthropus, although different species of this genus have different microwear texture patterns despite similar craniodental adaptations.