Many fields seek to identify steric influences in protein-ligand binding specificity. In some cases, these influences can be found by visually comparing protein structures, but subtler influences, whose significance may only be apparent from the analysis of many structures, are harder to find. To assist this process, we present VASP-S (Volumetric Analysis of Surface Properties with Statistics), an unsupervised volumetric analysis and statistical model for isolating statistically significant structural variations that may influence specificity. We applied these methods to analyze sequentially nonredundant structural representatives of two well-studied protein families: the canonical serine proteases and the enolase super family. We observed that statistically significant structural variations, as identified by VASP-S, reproduced experimentally established determinants of specificity. These results suggest that unsupervised methods, supported by statistical models, may be able to automatically identify variations that sterically influence specific binding in catalytic sites.