The opioid epidemic has stimulated efforts to discover new treatments for opioid use disorder (OUD); however, since the approval of buprenorphine in 2002, no new OUD medications have attained approval by the U. S. Food and Drug Administration.  One obstacle to OUD medication development has been an excessive reliance on preclinical testing procedures that yield a high rate of false positive effects.  Preclinical-to-clinical predictive validity can be improved with the use of drug-choice procedures, in which subjects choose between a drug reinforcer (e.g. an opioid like fentanyl or heroin) and a concurrently available non-drug reinforcer (e.g. food in laboratory animals or money in humans).  In these procedures, promising medications not only produce a decrease in drug choice, but also promote a reallocation of behavior to choice of the alternative non-drug reinforcer.  Less promising candidates fail to produce this outcome, and instead often have little effect on drug choice while depressing overall rates of responding and reinforcement.  This talk will describe drug choice procedures developed for use in nonhuman primates and rats, describe effects of both effective and ineffective treatments, and highlight the importance of opioid dependence/withdrawal state on medication effects.