We carried out a study in five London paediatric intensive care units (PICUs), with the objectives of describing a cohort of infants with respiratory syncytial virus (RSV) infection, comparing hospital diagnosis with PCR results and investigating the spread of RSV in families. Eligible infants were under 5 months old and admitted betweem November 1998 and October 1999 with respiratory failure, apnoea and/or bradycardia or acute life threatening episodes (ALTE). We diagnosed RSV by PCR analysis of nasopharyngeal aspirate, and in contacts by PCR of pernasal swabs. Of the 137 eligible infants, 66% (91/137) were recruited; of these, 82% (75/91) had RSV, with 47% (35/75) diagnosed by hospital laboratory tests and 93% (70/75) by PCR. The median duration of ventilation was 4.4 days, the length of stay on PICU, 8.6 days, and the length of stay in hospital, 15.9 days. In most families (62%), the parents and siblings developed symptoms of RSV infection at the same time as the infant. When the index infant was a secondary case, primary cases occurred in both older siblings (16 families) and adults (11 families). Silent RSV infection occurred frequently amongst children and adults. RSV is under-diagnosed in PICUs. PCR increases the rate of diagnosis of RSV compared to routine hospital diagnostic methods. Young infants are most often infected at the same time as or before their parents and siblings, indicating that the source may be outside the household; vaccinating family members may not prevent RSV infection in infants.