I develop an explanation of Zipf's law that is consistent with the observed marked heterogeneity in the growth of US cities. The explanation is that heterogeneous growth results in heterogeneous size distributions across cities, with the heaviest tailed distributions being Zipf and dominating...