[en] Straight-line Gaussian models adequately describe atmospheric diffusion for many applications. They have been modified for use in estimating diffusion in building wakes by adding terms that include projected building area and by redefining the diffusion coefficients so that the coefficients have minimum values that are related to building dimensions. In a recent study, Ramsdell reviewed the building-wake dispersion models used by the Nuclear Regulatory Commission (NRC) in its control room habitability assessments. The review included comparison of model estimates of centerline concentrations with concentrations observed in experiments at seven nuclear reactors. In general, the models are conservative in that they tend to predict concentrations that are greater than those actually observed. However, the models show little skill in accounting for variations in the observed concentrations. Subsequently, the experimental data and multiples linear regression techniques have been used to develop a new building wake diffusion model. This paper describes the new building wake model and compares it with other models. 8 refs., 2 figs