The determination of trace elements in powdered solids with direct powder introduction ICP-AES and ICP-MS is subject to errors associated with particle-size if (a) the powder sampler hardware introduces bias in the particle size distribution during the sample transport, or (b) if the efficiency of the processes leading to emission of analytes in the plasma is a function of particle size. The magnitude of particle size-segregation which is caused by a fluidized-bed powder sampler was estimated with two simple fluid dynamic models of powder collection: the first based upon the relative forces of gravity and the viscous drag on particles moving through a gas, and the second based upon the relative effects of particle inertia and the force of viscous drag on the particles. Experimental measurement of size-segregation was based on a series of 13 synthetic binary mixtures of 21 +/- 6 (+/- 1 SD) and 34 +/- 10-mu-m silica particles ranging in composition from 1 to 99% by weight of the smaller particle, the balance made up of the larger particle. The smaller particle was labelled with Mn and Co via complexation with immobilized 8-hydroxyquinoline, and the larger was labelled with Cu and Cd. Mixtures of the particles were analyzed by acid extraction of these labels, followed by conventional liquid nebulization ICP-AES. The composition of the particle mixture after passage through the sampler was related to the original mixture by (y(a)/y(b)) = 1.96 (x(a)/x(b)) where y(a) and y(b) are the fractions by weight of 21 and 34-mu-m particles after, and x(a) and x(b) are the fractions before, transport through the sampler. A mixture which was 1:1 by weight of these particles prior to transport, contained about 65% small and 35% large particles (by weight) after passage through the fluidized-bed sampler. Hardware designs which minimize the particle size-segregation should include mechanical provision for continuous formation of an aerosol of sample in the vicinity of the uptake capillary.