The key to developing cost-effective measures to reduce public exposure to PM is knowledge of the contribution that different sources make to the concentrations measured in the air, a scientific task known as source apportionment. This has been a very active area of air pollution research internationally, with the University of Birmingham being a major player. This task is more difficult than might appear at first sight for a number of reasons. If high-quality data were available on all the emission sources, then combining this with meteorological data and use of sophisticated dispersion models would give a good estimate of the concentration in air from each source. Unfortunately, though, this method is of limited value, as knowledge of emissions of particles from some sources is very poor (eg, woodsmoke) or non-existent (eg, road dust), and a large proportion of the mass of particles is not directly emitted but forms in the atmosphere from the oxidation of trace gases such as sulphur dioxide and nitrogen dioxide by processes which are challenging to simulate reliably. Consequently, the best knowledge of source apportionment is derived from using extensive measurements of the chemical composition of airborne particles (which are source-dependent) in statistical models using methods generically referred to as receptor modelling.