River beds are often arranged into patches of similar grain size and sorting. Patches can be distinguished into "free patches," which are zones of sorted material that move freely, such as bed load sheets; "forced patches," which are areas of sorting forced by topographic controls; and "fixed patches" of bed material rendered immobile through localized coarsening that remain fairly persistent through time. Two sets of flume experiments (one using bimodal, sand-rich sediment and the other using unimodal, sand-free sediment) are used to explore how fixed and free patches respond to stepwise reductions in sediment supply. At high sediment supply, migrating bed load sheets formed even in unimodal, sand-free sediment, yet grain interactions visibly played a central role in their formation. In both sets of experiments, reductions in supply led to the development of fixed coarse patches, which expanded at the expense of finer, more mobile patches, narrowing the zone of active bed load transport and leading to the eventual disappearance of migrating bed load sheets. Reductions in sediment supply decreased the migration rate of bed load sheets and increased the spacing between successive sheets. One-dimensional morphodynamic models of river channel beds generally are not designed to capture the observed variability, but should be capable of capturing the time-averaged character of the channel. When applied to our experiments, a 1-D morphodynamic model (RTe-bookAgDegNormGravMixPW.xls) predicted the bed load flux well, but overpredicted slope changes and was unable to predict the substantial variability in bed load flux (and load grain size) because of the migration of mobile patches. Our results suggest that (1) the distribution of free and fixed patches is primarily a function of sediment supply, (2) the dynamics of bed load sheets are primarily scaled by sediment supply, (3) channels with reduced sediment supply may inherently be unable to transport sediment uniformly across their width, and (4) cross-stream variability in shear stress and grain size can produce potentially large errors in width-averaged sediment flux calculations.