The RNA-dependent RNA polymerase (replicase) mediating the replication of tobacco mosaic virus (TMV) has been investigated in a number of laboratories over a period of 20 years. Cell-free enzyme preparations have been prepared which can continue the synthesis of nascent complementary RNA, initiated in vivo; however, the enzyme does not require, nor does it respond to, exogenous viral RNA as a template. The presence in plants of a virus-stimulated, host-encoded RNA-dependent RNA polymerase (RdRp) has added confusion to this field; it is now generally conceded, however, that this enzyme is not the TMV replicase. Our recent studies have emphasized several aspects of TMV RNA replication. We have examined the nature of TMV replicative structures synthesized in vitro by utilizing a partially purified enzyme preparation isolated from TMV-infected tobacco tissue. Radiolabelled products of the reaction were analysed on agarose gels and fractions with the predicted electrophoretic migration and nuclease sensitivities of replicative form (RF) and replicative intermediate (RI) were isolated. These fractions were hybridized to a collection of bacteriophage M13 clones containing portions of the TMV genome of both plus and minus polarity. The nascent synthesis in the RI-like molecules was restricted to the plus viral strand, while the new synthesis in the RF-like molecules was of both plus and minus polarity. Solubilization of the membrane-bound replicase with the non-ionic detergent CHAPS has yielded complexes which remain in solution after high-speed centrifugation. The solubilized replication complexes have been utilized as starting material for enzyme purification by Sepharose 4B gel filtration chromatography. The intracellular site of synthesis of TMV RNA has been reinvestigated in the light of reports suggesting a nuclear site of replication. The conclusion for nuclear synthesis has been based on fractionation of subcellular homogenates of virus-infected leaves or mesophyll protoplasts and identification of virus-related proteins associated with these fractions. In our studies, however, we conclude that these procedures can be misleading in that the 126,000 Mr TMV protein (and replicase activity) were found in all fractions of the homogenate analysed. Double-stranded TMV RNA, on the other hand, was barely detectable in preparations of purified nuclei; instead it was concentrated in the post-nuclear supernatant, suggesting that the nucleus is not the site of TMV RNA synthesis.