Measured rates of river incision into bedrock are commonly interpreted as proxies for rates of rock uplift (see refs 1 and 2, for example) and indices of the strength of climatic forcing of erosion over time (see refs 3 and 4, for example). This approach implicitly assumes that river incision rates are in equilibrium with external forcings over a wide range of timescales. Here we directly test this assumption by examining the temporal scaling of bedrock river incision from 155 independent measurements of river incision compiled from 14 sites. Of these sites, 11 exhibit a negative power-law dependence of bedrock river incision rate on measurement interval, a relationship that is apparent over timescales of 10(4)-10(7) years and is independent of tectonic and geomorphic setting. Thus, like rates of sediment accumulation(5), rates of river incision into bedrock exhibit non-steady-state behaviour even over very long measurement intervals. Non-steady-state behaviour can be explained by episodic hiatuses in river incision triggered by alluvial deposition, if such hiatuses have a heavy-tailed length distribution(6). Regardless of its cause, the dependence of incision rate on measurement interval complicates efforts to infer tectonic or climatic forcing from changes in rates of river incision over time or from comparison of rates computed over different timescales.