In culture, hippocampal neurons develop a polarized form, with a single axon and several dendrites. Transecting the axons of hippocampal neurons early in development can cause an alteration of polarity; a process that would have become a dendrite instead becomes the axon (Dotti, C. G., and G. A. Banker. 1987. Nature (Lond.). 330:254-256). To investigate this phenomenon more systematically, we transected axons at varying lengths. The greater the distance of the transection from the soma, the greater the probability for regrowth of the original axon. However, it was not the absolute length of the axonal stump that determined the response to transection, but rather its length relative to the lengths of the cell's other processes. If one process was greater than 10 microns longer than the others, it invariably became the axon regardless of its identity before transection. Conversely, when a cell's processes were nearly equal in length, it was impossible to predict which would become the axon. In these cases, axonal outgrowth began only after a long latency. During this interval, the processes appeared to be in dynamic equilibrium, some growing for short distances while others retracted. When one process exceeded the others by a critical length, it rapidly elongated to become the axon. The establishment of neuronal polarity during normal development may similarly involve an interaction among processes whose identities have not yet been determined. When, by chance, one exceeds the others by a critical length, it becomes specified as the axon.