Aggregation of data collected in 28 controlled experiments reveals reproducible debris-flow behavior that provides a clear target for model tests. In each experiment ∼10 m3 of unsorted, water-saturated sediment composed mostly of sand and gravel discharged from behind a gate, descended a steep, 95-m flume, and formed a deposit on a nearly horizontal runout surface. Experiment subsets were distinguished by differing basal boundary conditions (1 versus 16 mm roughness heights) and sediment mud contents (1 versus 7 percent dry weight). Sensor measurements of evolving flow thicknesses, basal normal stresses, and basal pore fluid pressures demonstrate that debris flows in all subsets developed dilated, coarse-grained, high-friction snouts, followed by bodies of nearly liquefied, finer-grained debris. Mud enhanced flow mobility by maintaining high pore pressures in flow bodies, and bed roughness reduced flow speeds but not distances of flow runout. Roughness had these effects because it promoted debris agitation and grain-size segregation, and thereby aided growth of lateral levees that channelized flow. Grain-size segregation also contributed to development of ubiquitous roll waves, which had diverse amplitudes exhibiting fractal number-size distributions. Despite the influence of these waves and other sources of dispersion, the aggregated data have well-defined patterns that help constrain individual terms in a depth-averaged debris-flow model. The patterns imply that local flow resistance evolved together with global flow dynamics, contradicting the hypothesis that any consistent rheology applied. We infer that new evolution equations, not new rheologies, are needed to explain how characteristic debris-flow behavior emerges from the interactions of debris constituents.