We report a novel combinatorial methodology for characterizing the effects of polymer surface features on cell function. Libraries containing hundreds to thousands of distinct chemistries, microstructures, and roughnesses are prepared using composition spread and temperature gradient techniques. The method enables orders of magnitude increases in discovery rate, decreases variance, and allows for the first time high-throughput assays of cell response to physical and chemical surface features. The technique overcomes complex variable spaces that limit development of biomaterial surfaces for control of cell function. This report demonstrates these advantages by investigating the sensitivity of osteoblasts to the chemistry, microstructure, and roughness of poly(D,L-lactide) and poly(epsilon-caprolactone) blends. In particular, we use the phenomenon of heat-induced phase separation in these polymer mixtures to generate libraries with diverse surface features, followed by culture of UMR-106 and MC3T3-E1 osteoblasts on the libraries. Surface features produced at a specific composition and process temperature range were discovered to enhance dramatically alkaline phosphatase expression in both cell lines, not previously observed for osteoblasts on polymer blends.