The verification processes were assessed for 18 important adverse drug reactions. Verification when achieved by formal studies was not obtained through cohort studies such as postmarketing surveillance or other follow up of drug users but through case-control studies or a similar disease orientated approach. When not achieved in this way the most useful aspects of uncontrolled studies were rechallenge and dose-response data. Analysis of the data and of the characteristics of different methods of verification suggested that there were four approaches to earlier discovery. (1) The best system seemed to be some form of record linkage, capable of providing at the same time data on incidence of adverse reactions and on prevalence of drug usage in patients with disease suspected of being drug induced. Until such a system can be established the relative efficacy of alternative approaches appears to be (2) review of all published first alerts, with prompt initiation of case-control or disease orientated studies for verification, (3) postmarketing surveillance of cohorts of drug users, and (4) voluntary reporting systems.