The main aim of this work is to investigate how reconstructing instantaneous carrier-phase velocities from filtered ones through defiltering improves the prediction of particle concentration in large-eddy simulation (LES) of particle-laden turbulent flows. A particle-laden homogeneous turbulent shear flow is simulated by LES employing the approximate deconvolution method (ADM) to approximate the instantaneous velocities of the carrier phase at the location of particles to solve their Lagrangian momentum equations. The carrier phase is simulated using a Fourier pseudo-spectral method with dynamic Smagorinsky model for subgrid-scale closures. The level of particle concentration is measured by the radial distribution function of particles and the probability density function of the particle number density. Particles with various time constants and terminal velocities are considered and it is shown that employing ADM highly improves the prediction of particle concentration by LES as results are compared against the results obtained by the direct numerical simulation performing a posteriori test. (c) 2007 Elsevier Ltd. All rights reserved.