The ovary and the fat body of Drosophila melanogaster both synthesise vitellogenins in vivo. The ovary contributes nearly as much vitellogenin to the yolk of an oocyte as does the fat body. Densitometry of fluorographs and gels has been used to compare the amount of the smallest vitellogenin polypeptide, yolk protein 3, synthesised by each tissue. Cell-free translations indicate that the ovary, in contrast to the fat body, contains a much reduced level of the mRNA for yolk protein 3 compared with the mRNAs for the other vitellogenin polypeptides. However, if tissues are cultured in vitro, the underproduction of this protein by the ovary is not significant. Because young embryos have levels of this polypeptide which are expected if the ovary has a low level of its corresponding mRNA, we argue that the ovary genuinely underproduces this protein in vivo and that the relative levels synthesised by the ovary in vitro are an artefact. Egg chambers of previtellogenic stages can synthesise vitellogenins, but the maximum level of vitellogenin synthesis occurs in egg chambers of the early vitellogenic stages. We conclude that the expression of the vitellogenin genes is subject to different controls at each site of synthesis. The possible cell types responsible for ovarian vitellogenin synthesis are discussed; the follicle epithelial cells are tentatively nominated for this role. We also suggest that a specific repression mechanism for vitellogenin gene expression exists in the ovary.