Abstract Partitioning of Cu, Pb, Zn, U, As and Mo between the minus 70 μm and minus 200 μm fractions of stream sediments from arid and semi-arid terrains is examined in the light of published case histories supplemented by new data. The advantages of selecting a particular fraction for routine sampling in such arid environments are assessed in terms of five criteria: (1) homogeneity of background population; (2) definition of threshold; (3) absolute element abundance levels; (4) contrast between anomalous and background populations; and (5) length of dispersion train. The most homogeneous background population distributions and improved definition of the threshold between background and anomaly occur in the very fine, minus 70 μm fraction of stream sediments for the majority of elements, in particular for Zn, Cu, U and As. Data for Pb and Mo do not consistently favour either size fraction in the case histories studied. Increased abundance levels of elements which are normally close to the analytical detection limit (U, Mo, As) occur most frequently in the minus 70 μm fraction, although Cu, Pb and Zn levels are commonly higher in the coarser fraction. In addition the finer size fraction better defines the anomalous population and provides the longer dispersion trains for Cu, Pb, Zn, U and As in the majority of case histories. The data examined indicate that the minus 70 μm fraction provides more useful information, in many instances, than the minus 200 μm fraction. The evidence suggests that problems expected with the use of the fine fraction — dilution through the abundance of wind blown material, and insufficient fine sediment — do not restrict the use of this fraction in stream sediment surveys in arid terrains.