The study reported here is a substantial first step towards using granular flow modelling to elucidate the conditions for unbiased sampling of particulate materials. Three mechanisms that can lead to sampling bias are identified and discussed at the particle level. They all tend to lead to undersampling of large particles compared to small particles. One of these mechanisms was investigated using two-dimensional modelling of sample cutters, including at least 46 simulation runs for each of 23 different sets of conditions. The configurations satisfying standard sampling criteria demonstrated biases ranging from 2.4% to 1.6% in a size score, indicating that large particles are under-sampled. They also showed extraction ratios substantially below the desirable levels because of particle hold up in the region above the cutter blades. The sensitivity of the sampling to the cutter aperture, cutter blade thickness and tip shape, cutter speed and the coefficients of restitution and friction of the particles were evaluated The most interesting result was that for particles with nominal topsize of 20 mm, a cutter with aperture 100 mm moving at 1200 mm/s (which does not satisfy the conditions of Gy) showed less bias than a cutter with aperture 70 mm moving at 600 mm/s (that does satisfy these requirements). Similarly a trimaran cutter showed little bias but does not currently meet these requirements.