Particle size distribution, obtained under standard sieving method, is usually given as a function of a single characteristic length of the particle. It is fully characterized for spherical particles, but it presents uncertainties when the particle has more complex morphologies as it is the case of biomass powders. The aim of this work is to characterize the standard sieving method in order to determine which particle dimension is being measured, as well as the repercussions of non-spherical shapes on a correct size classification. For this purpose, samples of milled poplar and corn stover have been classified in six size ranges between 0 and 5 mm. Each group of particles has been studied by means of scanned imaging to characterize their real 2D dimensions (width and length) and their shapes according to six different categories: circle, square, rectangle, rectangle fibrous, hook and hook fibrous. Results from image analysis show that sieve size corresponds mostly with particle width (shorter dimension), finding a sieving efficiency around 70%. Most wrongly classified particles showed a high aspect ratio, a hook shape or silhouette irregularities at fracture section, thus proving the importance of particle shape in the classifying process. (C) 2013 Elsevier Ltd. All rights reserved.