Covello and Gray proposed a model for the evolution of RNA editing wherein genetic drift results in fixation of genetic changes that must be transcriptionally corrected for production of a functional protein product. In this model, the enzymatic activity that alters the transcript is already present, and it is the chance fixation of an editable sequence that leads editing to be necessary for continued protein production. We previously observed that RNA polymerase slippage is required for cotranscriptional correction of frameshifted genes in Buchnera, an endosymbiont of aphids. Slippage is an intrinsic property of RNA polymerase, in Buchnera and in its relative, E. coli. However, E. coli lacks frameshifted open reading frames, whereas these are widespread among Buchnera, which is prone to Muller’s Ratchet. We reasoned that, under conditions where E. coli is subjected to bottlenecks, frameshift mutations may become fixed through drift. Following repeated single-cell bottleneck events, we observed emergence of dozens of frameshifts that require slippage for production of protein. We show that slippage has a real fitness cost, and that frameshifting impacts protein production. Our work is consistent with Covello & Gray’s evolutionary model. We see no evidence of adaptive cooption of slippage in our bottlenecked lines, consistent with the expectation that the majority of slippage events would yield non-functional or deleterious alternative products. Our results demonstrate the nonadaptive origins of editing, that we conclude selection normally keeps in check, and reveal the initial evolutionary path taken by genes that have evolved additional functions through editing or slippage.