We have used data from three analyses of occupational mortality for England and Wales to investigate a suspected hazard of pneumonia in welders. Mortality from the disease was consistently raised in welders aged 15-64, with standardised mortality ratios of 184 (95% CI 150-224) in 1959-63 and 157 (121-200) in 1970-72. Analysis of data for 1979-80 and 1982-90 showed that the increased risk is attributable mainly to an excess of pneumococcal and unspecified lobar pneumonia (proportional mortality ratio 255, 95% CI 192-332). No excess occurred in men above retirement age (65). A possible explanation of these findings is that welding fume reversibly increases the susceptibility of the lung to pneumonic infection. The observation of a similar mortality pattern in moulders and coremakers points to the metallic component of the fume as a possible culprit, but ozone or oxides of nitrogen could also be implicated. There are grounds for lobar pneumonia to be considered an occupational disease in welders.