Bitumen fumes emitted during road paving or roofing contain polycyclic aromatic hydrocarbons (PAHs). Experimental studies have been previously performed to test the carcinogenic potency of bitumen fumes. Some of them have been criticised either on the grounds that the fume condensates were not representative of fumes to which humans are exposed or because the fumes were never characterised in terms of particle size and poorly in terms of composition and concentration in the chambers. For a nose-only inhalation study, we have evaluated the ability of a new fume generation system to deliver stable and reproducible atmospheres of bitumen fumes to an inhalation chamber and investigated the representativity of the fumes generated at a concentration level of 5 mg/m3. The fume generator comprises: (1) an insulated 20 l heated kettle (200 degrees C for bitumen); (2) an insulated inlet pipe with a needle valve to adjust the flow of the test compound from the kettle; (3) a fume generation chamber equipped with a series of interchangeable channels of different width. The fume concentration in the exposure chamber can be controlled by changing the channel width or by restricting the evaporation surface with aluminium foil, and/or by changing the flow rate. Samples of the atmosphere in the chamber were collected and analysed for quantitative determination of total particulate matter (TPM), soluble matter, benzo[a]pyrene (B[a]P) content of the fumes and other PAHs, and evaluation of the particle size distribution. The representativeness of the fumes has been tested by comparison with fumes generated in the Shell small-scale fume rig, which was previously validated against field fumes collected during paving operations. Evaporative losses from the filters during sampling, transport and storage have been also assessed. At 5 mg/m3 TPM, the agreement between laboratories was quite good for the TPM analyses and was good for the soluble matter and B[a]P. Evaporative losses may lead to underestimation of the true exposure level in the inhalation chambers but the use of an XAD-2 cartridge backup is one approach to partially recover losses which occur on the filter. The particle size distributions are somewhat different from those reported for fumes associated with roofing and indoor mastic laying works, in that we found more than 85% of particles to be smaller than 1 micron, compared with 40% particles in the previous analyses. In conclusion, this equipment allows reproducible generation of fumes at the 5 mg/m3 TPM that are fairly representative of those produced in the field with the same bitumen.