At the posterior pole of the Drosophila oocyte, oskar induces a tightly localized assembly of pole plasm. This spatial restriction of oskar activity has been thought to be achieved by the localization of oskar mRNA, since mislocalization of the RNA to the anterior induces anterior pole plasm. However, ectopic pole plasm does not form in mutant ovaries where oskar mRNA is not localized, suggesting that the unlocalized mRNA is inactive. As a first step towards understanding how oskar activity is restricted to the posterior pole, we analyzed oskar translation in wild type and mutants. We show that the targeting of oskar activity to the posterior pole involves two steps of spatial restriction, cytoskeleton-dependent localization of the mRNA and localization-dependent translation. Furthermore, our experiments demonstrate that two isoforms of Oskar protein are produced by alternative start codon usage. The short isoform, which is translated from the second in-frame AUG of the mRNA, has full oskar activity. Finally, we show that when oskar RNA is localized, accumulation of Oskar protein requires the functions of vasa and tudor, as well as oskar itself, suggesting a positive feedback mechanism in the induction of pole plasm by oskar.