1. In vertebrates, bite force is a measure of whole organism performance that is associated with both cranial morphology and dietary ecology. Mechanistic studies of bite force production have identified morphological features associated with bite force, and linked bite force with diet, but this approach has rarely been used in mammals.

2. Mammals are a good system with which to investigate the function of the feeding apparatus because of the relative simplicity of their skulls and their high dietary diversity. Phyllostomid bats are one of the most trophically and morphologically diverse groups of mammals, but we know little about the relative importance of biomechanical variables in producing bite force or how these variables vary with diet.

3. We combined in vivo measurements of bite force with assessments of muscular and bony morphology to build and validate a model describing the mechanics of bite force production in 25 species of bats. We used this model to investigate how bats with different diets vary in biomechanical parameters that contribute to bite force. In addition to traditional dietary categories, we used a functional definition of diet that reflects the mechanical demands (hardness) of the food items in the natural diet.

4. Our model provided good predictions of in vivo bite forces and highlighted behavioural variation that is inherent in the in vivo data. The temporalis generates the highest moment about the temporomandibular joint (TMJ) axis, but the moment generated by the masseter is the most important variable in explaining variation among species. The dietary classification based on the hardness of the diet was more effective than traditional dietary categories in describing biomechanical differences among groups. The temporalis generated the highest proportion of the moment about the TMJ axis in species with very hard and hard diets, the masseter was most important for species with soft diets, and the medial pterygoid was most important for species with liquid diets.

5. Our results highlight the utility of combining a modelling approach with in vivo data when conducting ecomorphological studies, and the importance of ecological classifications that reflect functional importance of performance traits.