Suspected adverse drug reactions first reported in 1963 in the "British Medical Journal," the "Lancet," the "Journal of the American Medical Association," and the "New England Journal of Medicine" were reviewed 18 years later to assess their initial validity and subsequent verification. Of 52 first reports, five were deliberate investigations into potential or predictable reactions, and in each case causality was reasonably established; the other 47 reports were essentially anecdotal. Of these 47 reports, 14 related to categories of adverse reaction where false-positive reports were unlikely: immediate reactions, local reactions, and known reactions caused by a different mode of administration or a brand previously thought or claimed to be safe. The problem of false alarms rose in the remaining types of reactions: general reactions that did not occur immediately after administration and arose for the first time with a new chemical entity. Of 33 reports of such suspected adverse reactions, validity was satisfactorily established in 14 cases on the basis of rechallenge, predictability from known pharmacology, or the unique nature of the reaction. Of the remaining 19 reports, further verification still has not been satisfactorily established in 12. Seven of these possible false alarms were haematological reactions.Although 35 of the 47 anecdotal reports were clearly correct, of the 19 reports that were not reasonably validated at the time of the report, only seven were subsequently verified. This suggests that agencies monitoring adverse drug reactions should adopt criteria for assessing the validity of first reports of suspected adverse reactions. Such criteria should include: reactions on rechallenge, a pharmacological basis for the adverse reaction, immediate acute reactions, local reactions at the site of administration, reactions with a new route of administration of a drug known to provoke such reactions by another route, and the repeated occurrence of very rare events.