We compared data from clinical surveillance of acute respiratory infections in elderly people reported by a network of sentinel general practitioners, virus isolate data for respiratory syncytial virus (RSV) and influenza A and B viruses, and registered deaths, during the past four winters. Cases of acute respiratory disease (ARD) and deaths (3-week moving averages) showed strong similarity in timing. Periods of high incidence of ARD and numbers of deaths were followed within 2-3 weeks by reports of high numbers of RSV or influenza virus isolates. The epidemic period in December, 1989, was followed by high numbers of both virus isolates, but in all other epidemic periods one virus predominated. The lethality of epidemics due to RSV or influenza A or B varied both within and between years. These analyses suggest that RSV is as important as influenza viruses in causing morbidity and excess deaths among elderly people. RSV was prevalent at the end of each of the past four years in the UK; its importance as a cause of excess deaths would therefore be concealed in any analysis based on several years' experience. By contrast, the inconsistency of epidemics due to influenza viruses makes excess deaths due to these viruses easily recognisable. Integrated clinical and viriological surveillance is essential if epidemics due to RSV, influenza virus, or any other virus are to be correctly interpreted.