A series of deletion mutants of the wild-type Escherichia coli lactose promoter, with endpoints at +25, +19, +14, +1 and -6 (relative to the start of transcription at +1), was constructed and the deleted DNA replaced with non-lac DNA. These mutants were used to show that no specific DNA sequences downstream from -6 are required for efficient promoter utilization in vitro. In all cases transcription is dependent on the presence of the catabolite activator protein (CAP) and cAMP, and begins at +1 at a level indistinguishable from that at the wild-type promoter. A set of lac DNA fragments deleted to -6 was constructed, having an A, C, G or T residue at +1 and heterologous DNA downstream. These synthetic promoters allow systematic testing of the effect of the initiating nucleotide on the transcription process. Again, transcription occurs mainly from +1, at a level similar to the normal wild-type level. No substantial differences between these promoters are observed in the rates of formation of stable complexes, in the degree of complex formation, in the rate at which polymerase "escapes" from the complex or in abortive transcription products. Equivalent results are seen with a related set of constructs based on the CAP-insensitive lac UV5 promoter. Thus, lac promoter sequences including consensus hexamers at -10 and -35, plus the spacer region between them, provide specificity and efficiency both in initiation of transcription by RNA polymerase and in CAP-polymerase interactions. A question as to whether there is a third RNA polymerase binding site at lac, in addition to the known overlapping P1 and P2 regions, was not unambiguously answered. However, if a "P3" site does exist, it must lie between P1 and P2. Alternatively, the variety of polymerase interactions at wild-type lac may reflect different structural states of the enzyme. The results presented here indicate that DNA downstream from -6 plays little part in determining the conformation of the enzyme at the lactose promoter.