T7 RNA polymerase presents a very simple model system for the study of fundamental aspects of transcription. Some time ago it was observed that in the presence of only GTP as a substrate, on a template encoding the initial sequence GGGA., T7 RNA polymerase will synthesize a 'ladder' of poly-G RNA products. At each step, the ratio of elongation to product release is consistently approximately 0.75 until the RNA reaches a length of approximately 13-14 nt, at which point this ratio drops precipitously. One model to explain this drop in complex stability suggests that the nascent RNA may be structurally hindered by the protein; the RNA may be exiting via a pathway not taken by normally synthesized RNA and therefore becomes sterically destabilized. The fact that the length of RNA at which this occurs is close to the length at which the transition to a stably elongating complex occurs might have led to other mechanistic proposals. Here we show instead that elongation falls off due to the cooperative formation of structure in the nascent RNA, most likely an intramolecular G-quartet structure. Replacement of GTP by 7-deaza-GTP completely abolishes this transition and G-ladder synthesis continues with a constant efficiency of elongation beyond the limit of detection. The polymerase-DNA complex creates no barrier to the growth of the nascent (slippage) RNA, rather termination is similar to that which occurs in rho-independent termination.