SUMMARY The origin of the tridactyl hand of crown birds from the pentadactyl hand of those early theropod dinosaurs lying along the avian stem has become a classic, but at times seemingly intractable, historical problem. The point in question is whether the fingers of crown birds represent digits 1–3 as predicted by generalized trends in the fossil record; or digits 2–4, as evidenced by the topology of the embryonic mesenchymal condensations from which the digits develop. The frame shift hypothesis attempted to resolve this paradox by making these signals congruent by means of a homeotic transformation in digital identity, but recently the paleontological support for this hypothesis was questioned. Here, we reassess the frame shift from a paleontological perspective by addressing what predictions a frame shift makes for skeletal morphology, whether the frame shift remains a viable explanation of the known fossil data, and where on the theropod tree the frame shift most likely occurred. Our results indicate that the frame shift remains viable, and based on the inferred pattern of digit loss, the frame shift likely occurred at a deeper position in theropod phylogeny than previously proposed. A new evolutionary model of the frame shift is described in which the early history of the frame-shifted hand is marked by an extended zone of developmental polymorphism. This model provides a new conceptual framework for the role of developmental variability in communicating broad evolutionary patterns on a taxonomically inclusive phylogenetic tree.