Dietrich et al. (1992, 1993) proposed a digital terrain model for predicting the location of channel heads on the basis of the assumption that they occur where saturation overland flow exerts a boundary shear stress in excess of a critical value. Flume experiments were conducted in the modeled field site to evaluate the threshold hypothesis and to constrain critical shear stress and flow resistance parameters. Under complete grass cover, microtopography and grass stems were found to prevent significant sediment transport at all but the highest flows. When the grass stems were cut close to the ground, flow resistance and critical shear stress for significant sediment transport were reduced by up to an order of magnitude, but the remaining dense root mat prevented deep scour. These field experiments support the threshold assumption and the model estimations of the critical shear stress if local topographic convergence of flow is taken into account. The experiments also support the interpretation that significant degradation of vegetation cover is required for channel incision.