Understanding river response to sediment pulses is a fundamental problem in geomorphic process studies, with myriad implications for river management. However, because large sediment pulses are rare and usually unanticipated, they are seldom studied at field scale. We examine fluvial response to a massive (~20 Mt) sediment pulse released by the largest dam removal globally, on the Elwha River, Washington, United States, in an 11‐year before‐after/control‐impact study of channel morphology and grain size. We test the hypothesis that for a given flow magnitude, greater geomorphic change occurs under sediment‐rich conditions than under sediment‐starved conditions. Channel response to flow forcing was significantly different during the sediment‐pulse peak, 1–2 years after dam removal began, than earlier or later. During peak sediment supply our hypothesis was supported; major geomorphic change occurred under low flows and unit stream power ≤60 W/m2. However, by 4–6 years after dam removal began, rates of geomorphic change and sensitivity to stream power had decreased substantially such that our hypothesis was no longer unequivocally supported. These findings are consistent with a two‐phase conceptual model of dam‐removal response, involving a transport‐limited state followed by a more supply‐limited state. From comparisons with other dam removals and natural sediment pulses, we infer that the longevity of sediment‐pulse signals in gravel‐bed rivers depends upon gradient, river discharge, valley morphology, and sediment grain size. Stream power associated with substantial geomorphic change varies with sediment supply, such that assigning a general threshold stream power to gravel‐bed rivers may be untenable.