Rats were trained to detect the stimulus properties of fentanyl, 0.04mg/kg. Tested acutely, fentanyl generalized to, and morphine substituted for, the fentanyl training stimulus, although morphine was approximately 100-fold less potent than fentanyl in producing this stimulus. Subsequently, these rats were implanted with pumps (0.25ml volume) that delivered osmotically 1µl/h of solution containing either 3µg/µl (n = 16) or 6µg µl (n = 16) of fentanyl. When tested shortly after pump implantation, saline administration resulted in fentanyl-lever selection, indicating that the pumps were releasing fentanyl. Fentanyl-lever selection peaked at approximately 8 h after implantation; by 24 to 72 h after implantation, saline injection produced mainly saline-lever selection. On days 4 and 5 of chronic infusion of fentanyl, the fentanyl and morphine dose-effect curves were redetermined in both groups: both curves had shifted upwards, and the degree of shift occurred as a function of the dose of fentanyl infused continuously. After 6 days of infusion, the pumps were removed, and the detection of the fentanyl training stimulus was tested at 4, 12 and 24h after pump removal. At 12 h after removal, the percentage of subjects selecting the fentanyl lever following 0.04 mg/kg of fentanyl decreased to 53%, and this recovered to 87% by 24 h following pump removal. These results demonstrate tolerance to the stimulus properties of opioids with a few hours from the time of insertion of the pumps; the tolerance was not complete as shown by the results obtained during dose-effect testing with the pumps in place; finally tolerance was observed for a brief period when the pumps infusing fentanyl were removed (withdrawal tolerance). No evidence was obtained for asymmetric tolerance between fentanyl and morphine.