We have measured the transcription of Xenopus laevis oocyte and somatic-type 5 S RNA genes in S-150 extracts prepared from ovaries of mature X. laevis females (Glikin, G. C., Ruberti, I., and Worcel, A. (1984) Cell 37, 33-41). We find that somatic-type 5 S genes are transcribed at least 40-fold more efficiently than oocyte-type 5 S genes. Since adenovirus VA, Xenopus OAX, and Xenopus tRNAMet genes are all transcribed at levels similar to that of the somatic-type 5 S gene, this difference reflects a low level of oocyte-type 5 S gene transcription. Somatic-type 5 S transcription is a linear function of somatic-type 5 S DNA concentration and this, together with the efficient transcription of other class III genes, suggests that RNA polymerase III and the general class III transcription factors are not limiting under the conditions used here. Moreover, the 5 S gene-specific transcription factor TFIIIA does not limit transcription under these conditions as preincubation with purified TFIIIA does not alter transcription of either gene, and both genes exhibit characteristic TFIIIA footprints under transcription conditions in the S-150. Somatic-type 5 S DNA incubated in the S-150 and then injected into oocyte nuclei is actively transcribed whereas oocyte-type 5 S DNA treated in the same manner is inactive. We conclude that factors in the S-150 distinguish between somatic and oocyte-type 5 S genes, assembling active complexes preferentially on the former and inactive complexes preferentially on the latter. This process is not explained by binding properties of transcription factors for these genes as currently understood and represents a previously unrecognized mechanism of transcriptional regulation of the Xenopus 5 S genes.