We develop a theory to study apertureless scanning near-field optical microscopy which takes into account retardation, higher multipoles of the tip, and the multiple scattering between the tip and the surface. We focus on metallic systems and discuss the implication of the formation of tip-induced surface plasmon modes in the tip-surface system. We discuss the effects associated with the shift in energy of those modes as a function of the tip-surface distance. Both the local field and the scattering cross section are enhanced when the tip approaches the surface, but there is no general correspondence between the two enhancements.