The effect of excitation frequency and state of aggregation on the sensitivity obtained in ultratrace analysis using colloidal suspensions of silver nanoparticles and surface-enhanced resonance Raman scattering (SERRS) detection is explored to define suitable conditions for quantitative analysis. Two structurally similar dyes, only one of which causes aggregation, were used as analytes without the use of external aggregating agents, thus simplifying the surface chemistry and removing a major source of error. Addition of the nonaggregating dye caused no change in particle charge or size and no time-dependent aggregation as measured by zeta potential and particle size analysis. The most intense single-particle scattering was obtained using excitation at the wavelength of the plasmon resonance. Molecular resonance added approximately 2 orders of magnitude in sensitivity. Addition of the aggregating dye caused a reduction in surface charge of the particles and initiated a time-dependent aggregation process. However, constant SERRS with time is obtained at some excitation wavelengths probably because a constant number of clusters active at these wavelengths is maintained in the dynamic aggregation process. The additional enhancement caused by aggregation and molecular resonance is spread over a range of excitation frequencies. However, electronic spectra suggested that plasmon resonance enhancement would be effective at the longest wavelength of excitation used (785 nm), but there was a significant drop in intensity this far away from the absorbance maximum of the dye (429 nm). Thus, sensitive analysis using suspensions of single nanoparticles is feasible provided the excitation frequency used is close to that of the plasmon resonance frequency. Aggregation adds only an enhancement of approximately 6 in the experiments performed since only some particles in aggregates will have an active plasmon at any one wavelength, but the range of excitation wavelengths at which good enhancement is obtained is wider giving more flexibility if more complexity.