Finite element analysis (FEA) is a powerful quantitative tool that models mechanical performance in virtual reconstructions of complex structures, such as animal skeletons. The unique potential of FEA to elucidate the function, performance, and ecological roles of extinct taxa is an alluring prospect to paleontologists, and the technique has gained significant attention over recent years. However, as with all modeling approaches, FE models are highly sensitive to the information that is used to construct them. Given the imperfect quality of the fossil record, paleontologists are unlikely to ever know precisely which numbers to feed into their models, and it is therefore imperative that we understand how variation in FEA inputs directly affects FEA results. This is achieved through sensitivity and validation studies, which assess how inputs influence outputs, and compare these outputs to experimental data obtained from extant species. Although these studies are restricted largely to primates at present, they highlight both the power and the limitations of FEA. Reassuringly, FE models seem capable of reliably reproducing patterns of stresses and strains even with limited input data, but the magnitudes of these outputs are often in error. Paleontologists are therefore cautioned not to over-interpret their results. Crucially, validations show that without knowledge of skeletal material properties, which are unknowable from fossilized tissues, absolute performance values such as breaking stresses cannot be accurately determined. The true power of paleontological FEA therefore lies in the ability to manipulate virtual representations of morphology, to make relative comparisons between models, and to quantitatively assess how evolutionary changes of shape result in functional adaptations.