Based on the Bell theorem, it has been believed that a theoretical computation of the Bell correlation requires explicit use of an entangled state. Such a physical superposition of light waves occurs in the down-converter sources used in Bell experiments. However, this physical superposition is eliminated by wave propagation to spatially separated detectors. Bell correlations must therefore result from local waves, and the source boundary conditions of their previously entangled state. In the present model, Bell correlations are computed from disentangled separated waves, boundary conditions of nonlinear optics, and properties of single-photon and vacuum states specified by quantum electrodynamics. Transient interference is assumed between photon-excited waves and photon-empty waves based on the possibility of such interference found to be necessary by the designers of Bell-experiment sources. The present model employs local random variables without specifying underlying causality.