Electron microscopy was used to study the formation of random complexes between Escherichia coli RNA polymerase (nucleosidetriphosphate:RNA nucleotidyltransferase, EC 2.7.7.6) and a promoterless fragment (Mbo I-C) of bacteriophage T7 DNA, and to determine the location of the polymerase molecules bound at 3 degrees to the promoter-containing (Hinf)1100 fragment of the same DNA. The value of the Ka of random binding is about 3 times 10(4)M-1 when the enzyme is slowly diluted from its storage condition and is incubated with DNA for up to 2 min at 37 degrees. If dilution is rapid and occurs in a single step, or if incubation extends beyond 5 min, a substantial portion of RNA polymerase is converted to a form that binds randomly with a much greater affinity (about 10(8)M-1). Hence true random binding by RNA polymerase holoenzyme is much weaker than previously thought. However, great caution is required in assessing the extent of random binding where damage to the enzyme may occur. When RNApolymerase holoenzyme is incubated at 0 degrees with promoter-containing fragment (Hinf)1100, complexes form at the same promoter sites utilized at 37 degrees, although the highly stable "open" promoter complex is not formed under these conditions. However, the extent of binding is reduced as compared to promoter complexes formed at 37 degrees. This gives direct evidence for formation of complexes with promoter sites that have properties of the hypothetical "closed"complexes formed between RNA polymerase and duplex DNA.