We have identified a mutation (E148A) in T7 RNA polymerase (RNAP) that results in an enzyme which aborts transcription primarily when the nascent RNA achieves a length of 5 nt. This phenomenon is observed at a consensus promoter, but is even more strongly observed at promoters that are altered in the initiation region. Although the abortive product is of a fixed length (5 nt), the positions of the base substitutions in the initiation region that enhance this effect do not appear to be fixed, and we have observed the effect with a variety of initiation-region promoter variants. The phenomenon is also observed during promoter-independent transcription when transcribing a homopolymeric template such as poly(dC). Under conditions where the active site of the RNAP cannot extend beyond the third nucleotide in the template strand and the maximum length of the RNA:DNA hybrid cannot exceed three base-pairs (i.e. when synthesizing oligoG products due to transcript slippage at a promoter that initiates with the sequence +1 GGG...) the mutant RNAP gives rise to a normal spectrum of products 2 to 14 nt in length with no evidence of a block at 5 nt. Neither promoter binding nor promoter melting appears to be involved in this phenotype, as the mutant RNAP binds normally to promoter sequences and the behavior of the enzyme is unaffected by removal of the non-template strand in the initiation region of the promoter or on a supercoiled template. Importantly, the mutant RNAP is defective in binding single strand oligomers of RNA. These results suggest that the affected region of the RNAP may form part of the RNA product binding site and may be involved in the transition from an unstable initiation complex to a stable elongation complex, perhaps by sensing the presence of a nascent RNA and/or RNA:DNA hybrid.