In order to determine the mechanism by which testosterone (T) regulates the expression of vasopressin (VP) in the bed nucleus of the stria terminalis (BNST) and medial amygdala (MA), in situ hybridization's were performed to measure the expression of cytoplasmic VP mRNA and nuclear VP RNA in Long-Evans rats that were castrated or T replaced for various lengths of time. One hour after castration, plasma T levels were significantly reduced and remained at this low level until the T was replaced in animals. Three hours after T replacement, plasma T levels were significantly elevated compared to those of control animals and returned to control levels within 9 h of replacement. Modulation of nuclear VP RNA expression by castration and T replacement was rapid. Six hours after castration, levels of nuclear VP RNA in the BNST and MA were reduced below limits of detection. Three hours after T replacement nuclear VP RNA expression was elevated in BNST and MA, achieving levels comparable to controls after 9 h of T replacement. In the BNST, the number of cells positively labeled by the nuclear VP RNA probe closely paralleled the changes observed in plasma T; in that 24 h after T replacement there was a significant increase in the number of cells positively labeled for nuclear VP RNA, with control levels returning 3 days later. Cytoplasmic VP mRNA expression in the BNST and MA, on the other hand, was slower to respond to both castration and T replacement than nuclear VP RNA. A significant change in the number of cells expressing cytoplasmic VP mRNA was observed approximately 3 days after castration in the BNST and 7 days after castration in the MA. The amount of labeling per cell of cytoplasmic VP mRNA responded more rapidly to the reduction of circulating T than the number of positively labeled cells. A significant decrease in grains/cell was observed 24 h after castration in the BNST and 3 days after castration in the MA. The number of cells expressing VP mRNA in both the BNST and MA required 3 days of T replacement to return to control levels; while the amount of labeling per cell required 7 days of replacement. Therefore, these data suggest that T's ability to regulate the expression of the VP mRNA occurs at least in part at the transcriptional level, since the expression of VP primary transcript was altered prior to changes in the hybridization signal for the cytoplasmic VP mRNA.