Three clinical pharmacologists independently evaluated 500 untoward clinical events reported by physicians as adverse drug reactions (ADRs). They often disagreed with the reporting physicians and with each other. They judged 14.4-28.2% of the events to be definite, 26.4-38.0% probable, 21.4-31.0% possible and 15.8-28.2% of the events to be definite, 26.4-38.0% probable, 21.4-31.0% possible and 15.8-28.2% unlikely ADRs. In their opinion 19.1-32.4% of the drugs blamed were definitely, 29.8-34.4% probably, 24.9-36.7% possibly and 8.4-14.5% not responsible for the adverse reactions. Evaluators disagreed among themselves about the drug most likely to have been responsible in 36.4% of the events, about ADRs causing hospital admission in 56.8%, about severe ADR morbidity in 55.8%, about ADR prolongation of hospitalization in 67.3% and ADR prolongation of hospitalization in 67.3% and about ADR contribution to death in 71.0%. The divergence of judgements suggests that suspected ADRs are usually ambiguous clinical events, and that incidence, severity, medical consequences and cost of ADRs can only be estimated.