We investigate the influence of casting solvent on the final spatial dispersion of nanoparticles (NPs) in polymer nanocomposites (PNCs). We prepared nanocomposites of bare silica NPs and poly(2-vinylpyridine) (P2VP) by casting from two different solvents-methyl ethyl ketone (MEK) and pyridine-which are theta/good solvents, respectively, for both the polymer and the NPs. In MEK, we show that P2VP strongly adsorbs onto the silica surface to create a temporally stable bound polymer layer. The resulting "hairy" particles are sterically stabilized against agglomeration, and thus good NP dispersion in PNCs is always achieved, independent of P2VP molecular weight, concentration, or NP loading. On the contrary, in pyridine, P2VP does not adsorb on the silica NPs. The phase behavior in this case is thus governed by a subtle balance among electrostatic repulsion, polymer-induced depletion attraction, and the kinetic slowdown of diffusion-limited NP aggregation. While there is little remnant solvent in the dry PNC, and since these dispersion states are hardly altered on annealing, these results serve to emphasize the crucial role played by the casting solvent in the spatial dispersion state of NPs in a polymer matrix.