It is well established that the a "diameter' of a powder and its explosivity are strongly correlated. But, in fact, what diameter should be considered? The mass median diameter d50 is often used as an indicator of the particle size distribution (PSD) of a powder, but this choice is not always consistent with the mechanisms involved in a dust explosion. Moreover, authors mainly refer to the initial diameter of the powder, whereas it is more relevant to consider its characteristics at the exact time of the ignition. Many other metrics can be of interest concerning dust explosion: Sauter diameter, d10, solid surface concentration. To highlight these points, tests were performed on raw starch powders of 24 microm d50 and 0.49 m2.g-1 specific surface area. They were mechanically agglomerated with an Instron press and then coarsely ground by a blade mill. Different size classes up to 283 microm d50 and 0.16 m2.g-1 specific surface area were selected by sieving. The PSD of the samples were compared before and during their injection in the explosion vessels. The influence of the actual PSD on their explosion severity and minimum ignition energy (MIE) was determined. It has been demonstrated that the evolution of the MIE with regard to PSD is subject to a threshold phenomenon. Moreover, if the mass median diameter can be interesting for kinetics controlled by homogeneous chemical reaction and directly related to the fuel equivalent ratio, the reactive surface is more relevant when the combustion is controlled by surface reaction or diffusion to the surface as it is the case for the various starch samples. Chromatography measurements of the combustion gases confirm these results. Finally, when the agglomerates are partially fragmented by the injection, both the agglomerate and primary diameters must be considered.