The electromagnetic theory of surface-enhanced Raman spectroscopy (SERS), despite its simplicity, can account for all major SERS observations, including: the need for a nanostructured material as the SERS-active system; the observation that some metals form good SERS-active systems while others do not; the observation that strongly interacting metal nanoparticles result in very much more effective SERS-active systems; the observed polarization sensitivity shown by nanoparticle aggregates; and the optical behavior of nanostructured metals in the absence of a molecular adsorbate. By extending the ideas inherent in the electromagnetic model one can also understand the seminal features reported for single-molecule SERS, including the puzzling observation that only a few silver ‘particles’ in an ensemble are ‘hot’ (they are appropriately structured nanoparticle clusters) and that for a hot particle, once one is able to observe SERS, adding more adsorbate does not significantly alter the intensity (once the electromagnetic hot spot is occupied, adding adsorbate to other sites on the nanoparticle cluster will not add greatly to the observed intensity). However, the electromagnetic model does not account for all that is learned through SERS. Molecular resonances, charge-transfer transitions and other processes such as ballistic electrons transiently probing the region where the molecule resides and then modulating electronic processes of the metal as a result certainly contribute to the rich information SERS reports; and by virtue of the fact that these contributions will vary from molecule to molecule, they will constitute the most interesting aspects reported by SERS. But, the overall reason why SERS produces such inordinate enhancements is largely an electromagnetic property of nanostructures. Copyright © 2005 John Wiley & Sons, Ltd.