The concept of size-differential sampling of aerosol particles and the methods of data evaluation and presentation are reconsidered. Starting with a careful definition of logarithmically equidistant sampling intervals, the differences between mean interval positions and mean particle sizes are outlined. It is shown that for reason of particle-number conservation, log-differential concentration distributions, dN/d log D, must be presented as a function of log D, not as a function of the particle size D. Procedures for evaluating low-resolution (impactor-type) data are discussed. The properties of lognormal model distributions are analysed with the aim of evaluating the relation between concentration distributions and the derived distributions of joint length, surface area and volume. One important result is that for relatively narrow distributions the spacing between the distributions of the surface area and the volume (or mass) may be too small to be detectable experimentally. The findings derived from the basic investigation are used to evaluate recently reported particle-concentration distributions of urban aerosols which exhibit a very rapid decrease at particle sizes between 0.4 and 0.8 mum. As a result, the peak positions of the (calculated) distributions of the surface area and mass as well as large fractions of the cumulative distributions around the median differ by only 20%, a difference that cannot be identified by currently available techniques. Methods to derive information on particle sizes from combined time-dependent measurements of total particle concentrations and mass are discussed. Also addressed are the problems associated with attempts to determine the mass density of aerosol particles. (C) 2002 Elsevier Science Ltd. All rights reserved.