OBJECTIVE To review the literature on ambient air pollution and respiratory disease, and to consider the criteria for defining causation. METHODS Medical and scientific journals indexed by Medline, conferences, proceedings and monographs. METHODS Two kinds of study were selected--(i) controlled clinical trials which have exposed normal or asthmatic subjects and/or patients with chronic obstructive airways disease to sulphur dioxide, nitrogen dioxide or ozone; and (ii) epidemiological studies which have investigated the chronic toxicity of these pollutants, acid aerosols and polycyclic aromatic hydrocarbons. METHODS Experimental studies were tabulated under the headings "Design", "Subjects", "Pollutant concentration", "Duration of exposure", "Outcome measures" and "Conclusions". Epidemiological studies were summarised and compared in an attempt to reconcile conflicting results. (The experimental and epidemiological evidence has been used by regulatory bodies to develop ambient air quality guidelines.) CONCLUSIONS At the present state of knowledge, it is not possible to conclude that air pollution can cause respiratory disease de novo, but levels marginally above current guidelines certainly have adverse effects on individuals with pre-existing chronic lung disease.