A modified and improved technique for the detection of hepatitis B virus-specific DNA polymerase activity is described. DNA polymerase is released from Dane particles by mixing samples with the detergent Nonidet P-40 and beta-mercaptoethanol. After incubation of pretreated samples with a reaction mixture containing tritiated thymidine-methyl-5'-triphosphate (3H-TTP), DNA is precipitated onto a trichloroacetic acid (TCA)-treated paper. Unincorporated 3H-TTP is then chromatographically eluted with a 5% TCA solution and precipitated counts are determined. A sample is considered positive for DNA polymerase if the incorporated counts are significantly higher than the counts of a group of negative control samples. The modifications include pretreatment of the paper with TCA, chromatographic elution of unincorporated 3H-TTP with TCA solution, prefiltration of the sample through bacteriological filters, and use of sound statistical methods for evaluation of data. These changes have led to a highly reproducible, reliable and sensitive technique. The coefficient of variation of negative control samples from various test runs was in the range of 2.7-8.5%. A linear relationship between incorporated counts and DNA polymerase concentration was shown. A total of 419 serum samples from asymptomatic HBsAg-carrying blood donors were tested. Twenty-three (5.5%) of these were found to contain detectable DNA polymerase activity. All 23 samples also contained HBeAg.