We present a quantitative discussion of the processes active in bedrockfloored river channels, drawn from field observations, erosion rate measurements, and simple scaling rules. Quantitative documentation of process is needed to improve our understanding of bedrock river channels and aid in the formulation of erosion rules to be used in landscape evolution simulations. Our observations in a channel with “hard” rock (Indus River, Pakistan) suggest quarrying and abrasion are the primary erosion processes. It appears that block quarrying is the most efficient process when joints and bedding planes are sufficiently close. The block thickness a river is capable of quarrying goes as the square of the local flow velocity, v. Quarrying requires block “preparation”, during which subaerial weathering, bedload bashing, and/or hydraulic wedging, a previously undocumented process, act to free a block for quarrying. The Indus River is capable of quarrying blocks of up to ~0.7 m during annual peak flows. Rock abrasion should go as ~v5. Abrasion is most effective in regions of separated flow, generating a suite of sculpted rock bedforms that includes flutes, and this suggests abrasion occurs primarily by suspended sediment. Cavitation is unlikely to be a major process, as it requires unusually high velocity, and is suppressed by flow aeration. Abrasion measured on the Indus over 1 year using drill holes is lt; mm, with maximum rates within flutes, and in locally steep, narrow channel segments. Cosmogenic radionuclides from the same bed locations reveal average erosion rates over -1.5-2.0 ka that are an order of magnitude lower than the maximum 1 year rates. We reconcile these measurements by appealing to the passage of bedforms such as flutes. Our Indus River rate measurements are many times lower than longer-term rates, possibly implying substantial hydrologic variation induced by climate change. Incision rates in bedrock channels are controlled by very local hydraulic conditions well below the resolution of reach-based erosion rules. Incorporation of this geometric complexity represents a significant challenge to the landscape evolution modeling community.