In January 1984 the River Dee in north Wales was contaminated with phenol, with subsequent contamination of the tap water received by about two million consumers. A retrospective postal survey of 594 households was undertaken to determine whether consumption of this contaminated water was associated with illness. Subjects in areas that received contaminated water reported significantly more gastrointestinal illness than those in a nearby unexposed area (32.6% v 8.7%, p less than 0.00001) as well as reporting a higher incidence of any symptoms (43.6% v 18.4%, p less than 0.00001). Symptoms were consistent with phenol poisoning and bore a strong temporal relation to the pollution of the supply, but they developed at concentrations of phenols previously considered to be safe by the water authorities concerned. Chlorophenols produced during the treatment of water may have aggravated the problem.