Replication forks, moving opposite to the direction of transcription, are arrested at the 3' ends of the 35S transcription units in the rDNA locus of S. cerevisiae. Because of its position and polarity, we tested the hypothesis that this replication fork barrier (RFB) results from the act of transcription. Three results contradict this hypothesis. First, the RFB persists in a strain containing a disruption of the gene for the 135 kd subunit of RNA polymerase I. Second, the RFB causes a polar arrest of replication forks when transplanted to a plasmid. Third, transcription by RNA polymerase II of a plasmid copy of the 35S transcription unit lacking the RFB does not generate a barrier. We propose that replication forks are arrested in a directional manner through the binding of one or more proteins to two closely spaced sites in the RFB.