Three synthetic promoters, P(S1), P(S2) and P(S3), which differ in their core promoter elements, were studied in vivo and in vitro. Whereas an increased homology score correlates with higher rates of RNA polymerase binding, it does not correlate with activity in vivo. Permanganate probing in vivo reveals that P(S1), which exhibits the lowest homology score, is rate- limited during the early phase of promoter-RNA polymerase interactions. By contrast, P(S2) and P(S3), with higher homology scores, are limited at a late step involving an open DNA region spanning from +6 to +12, indicating a stalling of RNA polymerase. These complexes disappear upon treatment of cells with rifampicin and are replaced by open complexes covering the start site. Because initiated complexes are selectively insensitive to rifampicin action, this confirms that RNA polymerase stalled at +6 to +12 has initiated RNA synthesis. Kinetic studies indicate that the enzyme is released slowly from this position and that this slow release appears to be responsible for the low promoter activity. For P(S3), which exhibits the highest homology score and which binds RNA polymerase most efficiently, the release of the stalled complex is particularly slow. P(S3) is found to be the weakest of the three promoters in vivo. These results support models in which promoter activity can be determined by various rate limiting steps, including those following the formation of open complexes and even the initiation of RNA synthesis. © 1994 Academic Press Limited.