Pain thresholds are increased following central administration of arginine vasopressin (AVP), an effect which appears not to be mediated through opioid analgesic processes. In addition to magnocellular projections to the posterior lobe of the pituitary gland and parvocellular projections to the zona externa of the median eminence, the paraventricular nucleus (PVN) of the hypothalamus contains VP parvocellular neurons which also project to extrahypothalamic structures involved in pain inhibition. The present study examined whether AVP analgesia as measured by the tail-flick test was altered in animals with lesions placed in the PVN at either 7 or 35 days after surgery. VP levels in the pons-medulla and the lumbo-sacral spinal cord were measured by radioimmunoassay, as well as VP-like immunoreactivity in the PVN and spinal cord with immunocytochemistry. Lesions placed in the PVN eliminated AVP analgesia on the tail-flick test at both 7 and 35 days after surgery, and decreased radioimmunoassayable VP by 59% in the lumbo-sacral spinal cord and 36% in the pons-medulla. The extent of the lesions ranged from complete destruction of the PVN to partial sparing of ventro-medial PVN cells with VP-like immunoreactivity. These data indicate that the PVN is a critical structure for the integrity of AVP analgesia.