Supported lipid bilayers are model membranes formed at solid substrate surfaces. This architecture renders the membrane experimentally accessible to surface-sensitive techniques used to study their properties, including atomic force microscopy, optical fluorescence microscopy, quartz crystal microbalance, and X-ray/neutron reflectometry, and allows integration with technology for potential biotechnological applications such as drug screening devices. The experimental technique often dictates substrate choice or treatment, and it is anecdotally recognized that certain substrates are suitable for a particular experiment, but the exact influence of the substrate has not been comprehensively investigated. Here, we study the behavior of a simple model bilayer, phase-separating on a variety of commonly used substrates, including glass, mica, silicon, and quartz, with drastically different results. The distinct micron-scale domains observed on mica, identical to those seen in free-floating giant unilamellar vesicles, are reduced to nanometer-scale domains on glass and quartz. The mechanism for the arrest of domain formation is investigated, and the most likely candidate is nanoscale surface roughness, acting as a drag on the hydrodynamic motion of small domains during phase separation. Evidence was found that the physicochemical properties of the surface have a mediating effect, most likely because of the changes in the lubricating interstitial water layer between the surface and bilayer.