This paper experimentally and theoretically investigates the influence of an underlying metallic substrate (i.e., gold and silver) on the surface plasmon resonance (SPR) of labeled gold nanoparticles and the concomitant impact on the surface-enhanced Raman scattering (SERS) signal from the labels. These experiments employ nanoparticles of varied sizes (30-100 nm) that are coated with a bifunctional Raman scatterer composed of (1) a disulfide for chemisorption to the nanoparticle surface, (2) a succinimidyl ester for formation of a covalent linkage to an amine-terminated self-assembled monolayer on the underlying substrate, and (3) an aryl nitro group with an intrinsically strong Raman active vibrational mode. This approach allows facile systematic assessments of how variations in nanoparticle size, substrate composition, and the gap between the nanoparticle and substrate affect the SPR of the bound particles. Both UV-vis transmission and reflection absorption (incident angle of 58 degrees) spectroscopy are used to characterize the effect of each of these parameters on SPR. These results are then correlated with SERS enhancement factors (EFs) that were determined by accounting for particle surface concentrations, which were measured by atomic force microscopy, and the absolute number of labels, which were calculated on the basis of the surface area of each of the different-sized particles. All SERS spectra were collected at an incident angle of 58 degrees with respect to the surface normal. As expected, the SPR for particles in solution red-shifts with increasing particle size. More importantly, the SPR moves to even longer wavelengths as the size of immobilized particles increases and as the gap between the immobilized particle and substrate decreases. The red shift is also greater for a gold nanoparticle tethered to a gold substrate compared to a silver substrate. A theoretical model for the extinction of a particle above a flat substrate, corrected for surface scattering, radiation damping, and dynamic depolarization, is also briefly detailed. SPR results calculated with the model are consistent with the shifts observed in the SPR position for each of the manipulated experimental variables. The largest EFs are found for samples with an SPR maximum (lambda(max)) between the wavelengths for laser excitation (633 nm) and the Raman band for the symmetric nitro stretch of the particle coating (690 nm). As an example, an order of magnitude in the SERS enhancement factor is gained for a 60-nm particle immobilized 1.2 nm above a gold substrate (SPR lambda(max) = 657 nm) compared to that for a 30-nm particle (SPR lambda(max) = 596 nm).