The study of alcohol abuse traditionally has placed great emphasis on the development of tolerance and dependence as key factors. However, animal models of ethanol self-administration in dependent rats have been difficult to establish, caused in part by ethanol's aversive taste cues and subsequent aversive effects (i.e., "hangover" malaise) that prevent substantial ethanol consumption. In this study, this problem was addressed in animals trained to self-administer ethanol (10% w/v) in a sweetened-solution fading procedure before induction of dependence and repeated exposure to withdrawal. Once stable rates of responding for ethanol were achieved, a palatable liquid diet containing 8.7% (v/v) ethanol was introduced as the sole source of calories and fluid for one group of rats [ethanol diet (ED) group]. A second group of rats received a control diet with sucrose isocalorically substituted for ethanol (CD group). After 14-17 days of liquid diet exposure, the rats were withdrawn once a week for 4 weeks and 8 hr into each withdrawal session were allowed to self-administer ethanol or water for 60 min. As compared with CD rats, ED rats showed significantly greater intake of ethanol, but not water. No significant differences were found when separate groups of ED/CD rats were allowed to self-administer an alternate reinforcer (0.0075% saccharin solution). Rats who consistently had blood alcohol levels (BALs) above 100 mg% at the time of withdrawal sustained high levels of ethanol self-administration throughout the four withdrawal sessions. In contrast, rats who had an average BAL at withdrawal below 100 mg% showed progressive decreases in ethanol self-administration during repeated withdrawal episodes. The results demonstrated that chronic exposure to ethanol and repeated periods of abstinence are accompanied by elevated rates of ethanol intake in certain animals, and the persistence of elevated self-administration behavior of individual rats is predicted by their BAL at the time of withdrawal.