The performance of a modified Harvard high-volume cascade impactor (HVCI) was evaluated in six field campaigns with size-segregated particulate samplings for chemical and toxicological characterization. The 7-week sampling campaigns in 2002-2003 in Duisburg (autumn), Prague (winter), Amsterdam (winter), Helsinki (spring), Barcelona (spring), and Athens (summer) were selected to represent contrasting urban environments and seasons of public health interest due to high particulate concentrations or previous findings in epidemiological studies. Particulate samples were collected in parallel with the HVCI (PM(10-2.5), PM(2.5-1), PM(1-0.2), PM(0.2)), a virtual impactor (VI; PM(10-2.5), PM(2.5)), and a Berner low-pressure impactor (BLPI; 10 stages between 0.035 and 10 mum in particle diameter) using a 3- or 4-day sampling duration. The campaigns exhibited different profiles with regard to particulate mass concentration, size distribution, chemical composition and meteorological conditions, thus providing a demanding setup for an overall field comparison of the HVCI with the VI and BLPI reference samplers. Size-segregated particulate mass concentration could be reasonably well measured with the present HVCI configuration. The coarse (PM(10-2.5)) and fine (PM(2.5)) particulate mass agreed within 10% with the low-volume reference samplers, and the four-stage size distribution of the HVCI followed the modal pattern of urban aerosol. The concentrations of chemical constituents measured and integrated especially for the HVCI-PM(2.5) differed to some extent from those measured from the corresponding VI-PM(2.5) samples. This implies that when investigating the association of toxicological responses with the chemical constituents of particulate matter, it is necessary to use the chemical composition data of the same samples as used in toxicological experiments.