Recent theoretical and experimental work on linear exciton-light coupling in single and coupled semiconductor microcavities is reviewed: emphasis is given to angular dispersion and polarization effects in the strong-coupling regime, where cavity-polariton states are formed. The theoretical formulation is based on semiclassical theory. The energy of single-cavity modes is determined by the Fabry-Perot frequency ω c as well as by the center of the stop band ω s of the dielectric mirrors; the phase delay in the dielectric mirrors carries a nontrivial angle-and polarization dependence. The polarization splitting of cavity modes depends on the mismatch between ω c and ω s, and increases with internal angle as sin2 θ eff. Interaction between the cavity mode and quantum-well (QW) excitons is described at each angle by a two-oscillator model, whose parameters are expressed in terms of microscopic quantities. Weak and strong coupling regimes and the formation of cavity polaritons are described. Comparison with experimental results on a GaAs-based cavity with In0.13Ga0.87As QWs shows that a quantitative understanding of polariton dispersion and polarization splitting has been achieved. Coupling of two identical cavities through a central dielectric mirror induces an optical splitting between symmetric and antisymmetric modes. When QW excitons are embedded in both cavities at antinode positions, the system behaves as four coupled oscillators, leading to a splitting of otherwise degenerate exciton states and to separate anticrossing of symmetric and antisymmetric modes. These features are confirmed by experimental results on coupled GaAs cavities with In0.06Ga0.94As QWs. An analysis of reflectivity lineshapes requires the inclusion of the effect of resonance narrowing of cavity polaritons. Finally, the polarization splitting in a coupled cavity depends both on the single-cavity factors and on the angle-and polarization dependence of the optical coupling between the cavities. Inclusion of all these effects provides a good description of the experimental findings.