Bedrock rivers set the pace of landscape adjustment to tectonic and climatic forcing by transmitting signals of base-level change upstream through the channel network and ultimately to hillslopes. River incision is typically modeled as a monotonic function of bed shear stress or stream power, modulated by sediment tools and cover effects, but these models do not apply in channels with steep or vertical bedrock reaches due to changes in flow dynamics, hydraulic geometry, and bed cover. Here, we investigate how such knickzones (oversteepened channel reaches often containing waterfalls) influence the propagation of slope-break knickpoints that separate relict from adjusting topography, and thus the response times of landscapes to external forcing. We use a conceptual long-profile model to explore the consequences of waterfalls and knickzones on channel response and compare predictions to light detection and ranging (LiDAR) topography, field observations, and cosmogenic radionuclide data from Big Tujunga Creek, a 300 km2 watershed in the San Gabriel Mountains, California. Three prominent knickzones along Big Tujunga Creek, characterized by numerous waterfalls, show contrasting behavior. For the upper knickzone, waterfalls align with bands of harder rock exposed on adjacent hillslopes, and between waterfalls, the channel is mantled by large (>2 m) boulders, indicating knickzone retreat is slow compared to predictions of slope-break knickpoint retreat from stream-power models, enhancing the preservation of an upstream relict landscape. The middle knickzone shows evidence for both fast and slow knickzone retreat, as well as significant deviations from predictions of uniform tributary knickpoint elevations derived from stream-power models. The lower knickzone is characterized by a waterfall and knickzone within an incised inner gorge that provide evidence of rapid retreat relative to background channel incision. Overall, we find a pattern of decreasing knickzone and waterfall retreat rate with distance upstream of the range front, beyond decreases predicted by simple area-dependent celerity models. Our results highlight that waterfalls and knickzones can both enhance and inhibit landscape adjustment, leading to divergent controls on the pace of landscape evolution.