The topographic evolution of many active orogens is thought to be dominated by fluvial incision of bedrock rivers, that is rivers that have channels that are actively cutting into bedrock. Several incision models are based on the concept that incision of a bedrock channel is limited by basal shear stress or stream power. Collectively, these models can be represented by a power-law relationship, E=k S^n A^m, where A is upstream area, S is channel slope, E is incision rate, and k, m, and n are parameters. Specific incision models are distinguished by different predicted values or ranges for the parameters m and n, or by expansion of the equation to account for other processes. We evaluate the current crop of incision models by comparing their predictions with observations obtained from the Clearwater River in the western Olympic Mountain of northwest Washington State. We use long-term incision rates determined for the Clearwater by Pazzaglia and Brandon (2001) from their study of terrace incision over the last ~140 k.y. Those rates, together with long-term erosion rates determined by low-T thermochronometric studies (Brandon et al., 1998; Batt et al., 2001), show that over long time scales (> ~20 kyr.), the drainage has maintained a steady-state form, defined by a local balance between river incision and rock uplift. The incision-rate data allows us, through the use of inversion methods and statistical tests, to directly evaluate the applicability of the different incision models for the Clearwater. We were surprised to find that all models fail to fit the data. Our analysis is particularly striking for the commonly used stream-power model, which gives physically implausible best-fit values for m and n. Attempts to use realistic m and n values results in large systematic misfits. Our conclusion is that current bedrock incision models have not been adequately tested and many of the models may be inappropriate for representing the incision process over long time scales. This is an important issue in that it undermines quantitative modeling of tectonically active landscapes, where realistic representation of the bedrock incision process is critical.