The development of a specific and sensitive radioimmunoassay for the detection of buprenorphine in urine samples is described. With minor adjustments, the assay was also applied to the analysis for buprenorphine in plasma samples. The 2-diazobenzoic acid derivative of buprenorphine has been prepared as a hapten. The immunization of rabbits with the hapten-bovine serum albumin conjugate resulted in the production of antibodies, which cross-reacted with N-dealkylbuprenorphine up to about the 90% level. The antibodies showed very low cross-reactivities with the 3-O-glucuronides and with the structural analogue etorphine. The assay was mainly used to pre-screen for buprenorphine in urine samples of persons suspected of Temgesic misuse and to determine buprenorphine in plasma samples. A linear calibration graph for buprenorphine was obtained after logit-log regression [Y = 0.383 (s, 0.059) - 0.535 X (s, 0.025); r = 0.997 (s, 0.001)]. The spiking recovery study showed a linear regression of Y(observed) = 0.94 + 0.84 X (expected); r = 0.997. Intra- and inter-assay relative standard deviations were <4.35 and <6.36%, respectively. A comparison study of the high-performance liquid chromatographic determination (X) to the radioimmunoassay (Y) resulted in the following regression equation for the urine samples: Y = 1.44 + 1.64 X(n = 32; r = 0.910) and Y = 0.007 + 1.58 X (n = 10; r = 0.930) for plasma specimens. The minimum detectable dose of the immunoassay was calculated to be 10 pg ml-1 (Student's t-distribution, p = 0.01, degrees of freedom = 8).