Crystallization of 5-methyl-2-[(2-nitrophenyl)amino]-3-thiophenecarbonitrile (1), previously found to produce six conformational polymorphs from solution, on single-crystal pimelic acid (PA) substrates results in selective and oriented growth of the metastable "YN" (yellow needle) polymorph on the (101)(PA) faces of the substrate. Though the freshly cleaved substrate crystals expose (101)(PA) and (111)(PA) faces, which are both decorated with [101](PA) ledges that could serve as nucleation sites, crystal growth of YN occurs on only (101)(PA). Goniometry measurements performed with an atomic force microscope reveal that the (001)(YN) plane contacts (101)(PA) with a crystal orientation [100](YN)//[010](PA) and [010](YN)//[101](PA). A geometric lattice analysis using a newly developed program dubbed GRACE (geometric real-space analysis of crystal epitaxy) indicates that this interfacial configuration arises from optimal two-dimensional epitaxy and that among the six polymorphs of 1, only the YN polymorph, in the observed orientation, achieves reasonable epitaxial match to (101)(PA). The geometric analysis also reveals that none of the polymorphs, including YN, can achieve comparable epitaxial match with (111)(PA), consistent with the absence of nucleation on this crystal face. In contrast, sublimation of 1 on cleaved succinic acid (SA) substrates, which expose large (010)(SA) faces decorated with steps along [101](SA), affords growth of several polymorphs, each with multiple orientations, as well as oriented crystals of a new metastable polymorph on the (010)(SA) surfaces. The lack of polymorphic selectivity on (010)(SA) can be explained by the geometric lattice analysis, which reveals low-grade epitaxial matches between (010)(SA) and several polymorphs of 1 but no inherent selectivity toward a single polymorph. These observations demonstrate the sensitivity of crystal nucleation to substrate surface structure, the potential of crystalline substrates for selective nucleation and discovery of polymorphs, and the utility of geometric lattice modeling for screening of substrate libraries for controlling polymorphism.