One of the primary particle-retention mechanisms in a filtration process is straining, which happens when the size of a particle is larger than the pore throat trapping it. The occurrence of straining depends on both the particle size distribution (PSD) of the suspension and the pore throat size distribution (PoSD) of the filter, with measurement of the latter being especially difficult. This paper presents a new approach for evaluating the PoSD of a filter medium using a discrete element method (DEM). The approach computes the filter efficiency (FE), defined as the fraction of particles of a certain size retained by the filter, for incremental particle sizes in a filtration process. Special care was taken to turn off the interactions between injected particles in the simulation so that the computed filter efficiencies reflect an incremental sampling of pore throat sizes, and thereby can be used to infer the PoSD. Good agreement was observed between the simulated and theoretical PoSDs in an ideal sieve and a simple cubic packing case. The study showed that for a monodisperse random close packing (RCP) of spherical grains, the largest pore throat is about 1/4.5 of the grain size, and the smallest pore throat is about 1/6.6 of the grain size. For a poly disperse RCP formed by 20-40 mesh grains, the largest pore throat is about 1/5.3 of the effective grain size, and the smallest pore throat is about 1/10 of the effective grain size. The approach can be easily extended to evaluate the PoSD of any type of filter medium, including RCPs with any PSD. The PoSD computed provides essential information for filter layer design and selection. (C) 2017 Elsevier B.V. All rights reserved.