The counting of fibres on membrane filters could be facilitated by using size-selective samplers to exclude coarse particulate and fibres that impede fibre counting. Furthermore, the use of thoracic size selection would also remove the present requirement to discriminate fibres by diameter during counting. However, before thoracic samplers become acceptable for sampling fibres, their performance with fibres needs to be determined. This study examines the performance of four thoracic samplers: the GK2.69 cyclone, a Modified SIMPEDS cyclone, the CATHIA sampler (inertial separation) and the IOM thoracic sampler (porous foam pre-selector). The uniformity of sample deposit on the filter samples, which is important when counts are taken on random fields, was examined with two sizes of spherical particles (1 and 10 microm) and a glass fibre aerosol with fibres spanning the aerodynamic size range of the thoracic convention. Counts by optical microscopy examined fields on a set scanning pattern. Hotspots of deposition were detected for one of the thoracic samplers (Modified SIMPEDS with the 10 microm particles and the fibres). These hotspots were attributed to the inertial flow pattern near the port from the cyclone pre-separator. For the other three thoracic samplers, the distribution was similar to that on a cowled sampler, the current standard sampler for fibres. Aerodynamic selection was examined by comparing fibre concentration on thoracic samples with those measured on semi-isokinetic samples, using fibre size (and hence calculated aerodynamic diameter) and number data obtained by scanning electron microscope evaluation in four laboratories. The size-selection characteristics of three thoracic samplers (GK2.69, Modified SIMPEDS and CATHIA) appeared very similar to the thoracic convention; there was a slight oversampling (relative to the convention) for d(ae) < 7 microm, but that would not be disadvantageous for comparability with the cowled sampler. Only the IOM thoracic sampler tended to undersample the fibres relative to the thoracic convention. With the data divided into four classes based on fibre length, the size-selection characteristics appeared to be unaffected by fibre length for GK2.69, Modified SIMPEDS and CATHIA. Only the IOM thoracic sampler (with the foam selector) showed slightly lower selection for longer length classes of fibres. These results indicate that the tested samplers follow the thoracic sampling convention for fibres, and may be used to improve the quality and reliability of samples that are taken when there is likely to be significant background dust.