We contrast evidence of urban path dependence with efforts to analyze calibrated models of city sizes. Recent evidence of persistent city sizes following the obsolescence of historical advantages suggests that path dependence cannot be understood as the medium-run effect of legacy capital but instead as the long-run effect of equilibrium selection. In contrast, a different, recent literature uses stylized models in which fundamentals uniquely determine city size. We show that a commonly used model is inconsistent with evidence of long run persistence in city sizes and propose several modifications that might allow for multiplicity and thus historical path dependence.