Evaluating medications in animal laboratory paradigms can reveal whether the compound is effective in an established alcoholism model, at clinically relevant doses and exposure conditions, when administered orally (or transdermally) and without serious limiting side effects. Positive outcomes constitute a possible discovery for relevance to alcoholism and, under favorable marketing conditions, encourage further development. Medication testing using animal models of alcoholism might also guide clinical testing by discriminating clinically effective from clinically ineffective compounds. This ability rests on whether there are tests or, more reasonably, batteries of tests having this discriminative ability. The present paper examines this possibility. Effects of naltrexone and acamprosate in animal paradigms which model behavioral aspects of alcoholism are reviewed and compared with the effects of compounds which have limited effects in alcoholics. It is not clear at present whether any single paradigm or combination of paradigms differentiates clinically effective from clinically limited compounds. Steps are suggested to improve the use of preclinical laboratory tests to predict which compounds are likely to be effective medications for reducing drinking and sustaining abstinence in human alcoholics.