Spherical cap gold nanocavity arrays with internal diameters of 240, 430, 600 and 820 nm were fabricated on smooth gold films using nanosphere lithography with electrochemical metal deposition. Each array was prepared to the same normalized film thickness to diameter ratios, t(N), of 0.8 +/- 0.04. Selective modification of the top surface and interior walls of the gold nanocavity arrays with [Ru(bpy)(2)(Qbpy)](2+), where bpy is 2,2'-bipyridyl and Qbpy is 2,2':4,4 '':4,4 ''-quarterpyridyl, was accomplished using a two step adsorption process exploiting the assembled polystyrene spheres as masks. This selective modification approach permitted direct quantitative comparison, for the first time, of plasmonic enhancement of Raman signal and luminescence signal from a monolayer adsorbed at the top surface versus interior walls of all-gold nanocavity arrays. For all cavity sizes, significantly greater Raman and luminescence signal enhancement was observed from [Ru(bpy)(2)(Qbpy)](2+) monolayer adsorbed at the top surface of the array compared with the cavity walls. This disparity in Raman intensity from top versus cavity interior increased as the cavity dimensions decreased. For example, the Raman signal intensity from [Ru(bpy)(2)(Qbpy)](2+) adsorbed at the top surface of 240 nm gold arrays was 170 times greater than SERS signal for this material adsorbed at the interior walls of this array, whereas the relative Raman signal enhancement was 6 from top versus interior for the 820 nm internal radius arrays under 785 nm excitation. The origin of the relatively greater signal at the top surface is discussed in the context of plasmonic distribution at each surface.