In a previous study of the Andersen 8-stage non-viable cascade impactor (ACI), we observed maldistribution of monodisperse particles of known aerodynamic diameter between the four concentric rings of nozzles associated with stages 0 and 1 of an ACI (operated at 28.3 L/min without pre-separator). In the current study, we investigated this behavior by recovering the calibration particles from each ring of deposits separately. The performance of stages 0 and 1 were close to expectations when deposits from all four rings were assessed as a single entity. However, the fraction of the total deposits in the individual rings was not only different than the ideal value of 25% but changed with the size of the calibration particles. For example, the fraction of the total deposit in the outermost ring of both stages increased as the calibrant particles decreased in size. This outcome can happen only if the smaller particles are preferentially dispersed to the outermost ring, indicating an airborne pre-classification of large from small particles before the incoming aerosol enters into the nozzles. This undesirable airborne size classification is consistent with the focused recirculating dispersion of a laminar jet in the Andersen cone predicted by its large expansion angle (110º). We conclude that when a polydisperse aerosol enters the ACI, the particle masses collected on stages 0 and 1 bear no known relationship to the aerodynamic particle size of this aerosol in the size range where these two stages are operational.