There have been few attempts to model subjective symptoms of drug withdrawal using animals as subjects. Two approaches for developing such models are reviewed. First, using drug discrimination methodology, it may be possible to train animals to detect the effects of withdrawal. This method has two difficulties: 1) the only discriminations trained to date involve precipitated withdrawal, and 2) the stimulus controlling behavior is difficult to specify. Second, withdrawal from many drugs of abuse produces the symptom of anxiety, and it seems likely that animal models of anxiety could be useful for studying drug withdrawal. This hypothesis has been explored most fully using subjects trained to detect the discriminative stimulus properties of the putative anxiogenic drug pentylenetetrazole (PTZ). Withdrawal from benzodiazepines or ethanol substitutes fully for PTZ, and withdrawal from cocaine, morphine, and nicotine substitutes partially for PTZ. Emerging data suggest that other animal models of anxiety may also be useful for detecting drug withdrawal. The final portion of this review examines a behavioral test that is very sensitive for detecting physical signs of withdrawal in animals. In subjects maintained on an operant baseline using food as a reinforcer, withdrawal from a drug of dependence frequently is associated with disruption of that operant behavior. For example, tetrahydrocannabinol and cocaine, drugs that are not traditionally seen as having significant withdrawal signs, produce disruption of operant responding when high-dose administration is terminated, and their readministration reverses this behavioral disruption. Based on the observation that withdrawal is associated with anxiogenic stimuli, we suggest a method to determine if disruption of operant behavior may be related to these stimuli.