In particle-laden turbulent flows, the statistics that are key to quantifying the turbulence-driven particle dispersion include the particle spatial number density, and the moments of particle velocity such as the mean velocity and the velocity covariance. These particle statistics could be obtained through rigorous computational approaches such as the direct numerical simulations (DNS) and large-eddy simulations (LES) of turbulent flows carrying disperse particles. However, DNS, and to some extent LES, of high Reynolds number, particle-laden turbulent flows are computationally too taxing to be of practical use. In the current study, a computationally efficient stochastic method is presented for tracking the velocities and positions of discrete particles in homogeneous, isotropic turbulent flow. The principal advantages of this approach when compared to DNS/LES are: (i) it provides reasonably accurate predictions of the particle statistics at a significantly reduced computational cost, and (ii) for a given particle Stokes number, there is no significant increase in computational expense with the flow Reynolds number. The stochastic method adopted in this study is rooted in the transport equation for the probability density function (PDF) of particle position and velocity, P(x,v;t). The PDF equation consists of an unclosed term representing particle-turbulence interactions. A rigorous closure is derived for this term for high Stokes number particles in isotropic turbulence. The closure leads to a diffusion-like term, thereby transforming the PDF equation into a Fokker-Planck equation. Using the closure theory, stochastic simulations were performed by solving the Langevin equations which are the Lagrangian equivalent of the Fokker-Planck equation. Particle velocity, acceleration and dispersion statistics obtained from these simulations show good agreement with the DNS results from three prior studies [1-3]. (C) 2013 Elsevier B.V. All rights reserved.