Fluorescence correlation spectroscopy (FCS) can provide a wealth of information about biological and chemical systems on a broad range of time scales (<1 micros to >1 s). Numerical modeling of the FCS observation volume combined with measurements has revealed, however, that the standard assumption of a three-dimensional Gaussian FCS observation volume is not a valid approximation under many common measurement conditions. As a result, the FCS autocorrelation will contain significant, systematic artifacts that are most severe with confocal optics when using a large detector aperture and aperture-limited illumination. These optical artifacts manifest themselves in the fluorescence correlation as an apparent additional exponential component or diffusing species with significant (>30%) amplitude that can imply extraneous kinetics, shift the measured diffusion time by as much as approximately 80%, and cause the axial ratio to diverge. Artifacts can be minimized or virtually eliminated by using a small confocal detector aperture, underfilled objective back-aperture, or two-photon excitation. However, using a detector aperture that is smaller or larger than the optimal value (approximately 4.5 optical units) greatly reduces both the count rate per molecule and the signal-to-noise ratio. Thus, there is a tradeoff between optimizing signal-to-noise and reducing experimental artifacts in one-photon FCS.