We analyze flash flooding in small urban watersheds, with special focus on the roles of rainfall variability, antecedent soil moisture, and urban storm water management infrastructure in storm event hydrologic response. Our results are based on empirical analyses of high‐resolution rainfall and discharge observations over Harry's Brook watershed in Princeton, New Jersey, during 2005–2006, as well as numerical experiments with the Gridded Surface Subsurface Hydrologic Analysis (GSSHA) model. We focus on two subwatersheds of Harry's Brook, a 1.1 km2 subwatershed which was developed prior to modern storm water management regulations, and a 0.5 km2 subwatershed with an extensive network of storm water detention ponds. The watershed developed prior to modern storm water regulations is an “end‐member” in urban flood response, exhibiting a frequency of flood peaks (with unit discharge exceeding 1 m3 s−1 km−2) that is comparable to the “flashiest” watersheds in the conterminous U.S. Observational analyses show that variability in storm event water balance is strongly linked to peak rain rates at time intervals of less than 30 min and only weakly linked to antecedent soil moisture conditions. Peak discharge for both the 1.1 and 0.5 km2 subwatersheds are strongly correlated with rainfall rate averaged over 1–30 min. Hydrologic modeling analyses indicate that the sensitivity of storm event hydrologic response to spatial rainfall variability decreases with storm intensity. Temporal rainfall variability is relatively more important than spatial rainfall variability in representing urban flood response, especially for extreme storm events.