Tolerance develops to the antinociceptive effects of morphine with repeated microinjections into the ventrolateral periaqueductal gray (PAG). This tolerance could be caused by adaptations within the PAG or anywhere along the descending pathway (rostral ventromedial medulla to spinal cord). If tolerance is caused by a change along the descending pathway, then tolerance should develop to direct activation of PAG output neurons. However, if tolerance is caused by a change to neurons within the PAG, then tolerance will not occur with repeated direct activation of PAG output neurons. This hypothesis was tested by assessing antinociception following repeated microinjections of the GABA antagonist bicuculline and the excitatory amino acid kainate into the ventrolateral PAG. Microinjection of bicuculline and kainate produces antinociception by disinhibition and direct excitation of ventrolateral PAG output neurons, respectively. Repeated administration of these drugs into the ventrolateral PAG produced antinociception with no evidence of tolerance. That is, the hot-plate latency and responsiveness to intraplantar formalin administration was comparable whether rats received the drug for the first or fifth time. Moreover, microinjection of bicuculline or kainate produced comparable antinociception in rats pretreated with these drugs and saline-treated control rats. These data demonstrate that repeated activation of ventrolateral PAG output neurons is not sufficient to produce tolerance. Thus, tolerance must be caused by a change in neurons preceding output neurons in this circuit, presumably opioid-sensitive GABAergic neurons.