Experimental observations comparing two steep unchanneled valleys in the Oregon Coast Range, one intensively instrumented (CB1) and the other monitored for runoff but which produced a debris flow (CB2), shed light on the mechanisms of shallow flow in bedrock, its interaction with the vadose zone, and its role in generating landslides. Previous work at CB1 led to the proposal that during storms pulses of rainfall transmit pressure waves through the vadose zone and down to the saturated zone to create rapid pore pressure response and runoff <Torres et al., 1998>. Here, we document the associated rapid pore pressure response in the shallow fractured bedrock that underlies these colluvium-mantled sites and examine its influence on the generation of storm flow, seasonal variations in base flow, and slope stability in the overlying colluvial soil. Our observations document rapid piezometric response in the shallow bedrock and a substantial contribution of shallow fracture flow to both storm flow and seasonal variations in base flow. Saturated hydraulic conductivity in the colluvial soil decreases with depth below the ground surface, but the conductivity of the near-surface bedrock displays no depth dependence and varies over five orders of magnitude. Analysis of runoff intensity and duration in a series of storms that did and did not trigger debris flows in the surrounding area shows that the landslide inducing storms had the greatest intensity over durations similar to those predicted by a simple model of piezometric response. During a monitored storm in February 1992, the channel head at the base of the neighboring CB2 site failed as a debris flow. Automated piezometric measurements document that the CB2 debris flow initiated several hours after peak discharge, coincident with localized development of upward spikes of pressure head from near-surface bedrock into the overlying colluvial soil in CB1. Artesian flow observed exfiltrating from bedrock fractures on the failure surfaces at CB2 further implicates bedrock fracture flow in both runoff generation by subsurface storm flow and suggests a connection to landslide initiation. Our observations show that the timing of shallow landslide initiation can be delayed relative to both peak rainfall and peak runoff and support the argument that the influence of fracture flow on shallow landsliding helps explain the wide variability in the occurrence of slope instability in topographically analogous locations.