The middle of a hillslope hollow in the Oregon Coast Range failed and mobilized as a debris flow during heavy rainfall in November 1996. Automated pressure transducers recorded high spatial variability of pore water pressure within the area that mobilized as a debris flow, which initiated where local upward flow from bedrock developed into overlying colluvium. Postfailure observations of the bedrock surface exposed in the debris flow scar reveal a strong spatial correspondence between elevated piezometric response and water discharging from bedrock fractures. Measurements of apparent root cohesion on the basal (C_b) and lateral (C_l) scarp demonstrate substantial local variability, with areally weighted values of C_b = 0.1 and C_l = 4.6 kPa. Using measured soil properties and basal root strength, the widely used infinite slope model, employed assuming slope parallel groundwater flow, provides a poor prediction of hydrologic conditions at failure. In contrast, a model including lateral root strength (but neglecting lateral frictional strength) gave a predicted critical value of relative soil saturation that fell within the range defined by the arithmetic and geometric mean values at the time of failure. The 3‐D slope stability model CLARA‐W, used with locally observed pore water pressure, predicted small areas with lower factors of safety within the overall slide mass at sites consistent with field observations of where the failure initiated. This highly variable and localized nature of small areas of high pore pressure that can trigger slope failure means, however, that substantial uncertainty appears inevitable for estimating hydrologic conditions within incipient debris flows under natural conditions.