This study presents a probability density function propagation approach to dispersion modeling. The model calculates the instantaneous spatial spread about the ensemble-mean trajectory for a group of particles as they move in the Lagrangian reference frame, and thus it precludes the need to generate large numbers of individual particle trajectories to represent the particle phase. This method for finite-inertia particle dispersion is based on Taylor＇s approach and it approximates the normalized particle velocity correlation functions with Frenkiel functions. The required particle time scales are based on published analytical studies and some independent analysis. All the turbulent scales needed in this approach can be obtained from practical turbulence models. A new procedure to estimate the particle fluctuating velocity statistics along the ensemble mean particle trajectory is also developed. This procedure is based on the particle momentum equation and does not involve any empirical constants. The present model is evaluated by use of the experimental data of Snyder and Lumley and those of Wells and Stock. The ability of the model to predict particle dispersion and particle velocity decay is quite satisfactory for the cases studied. The study also demonstrates the computational advantage of the present model in comparison with the Lagrangian stochastic models that rely on the Monte Carlo procedure to represent the particle phase. The Frenkiel functions seem adequate to model the normalized particle velocity correlations. The validation studies indicate that the crossing trajectory effects induce a negative loop in the normalized particle transverse (relative to particle drift) velocity correlations. These negative loops apparently do not exist in the normalized particle longitudinal (relative to particle drift) velocity correlations.