We used a repeated-measures factorial field experiment to ask whether the arrangement and stability of leafy debris affect the density of chironomids in leaf packs in dynamic, streambed landscapes. We replaced natural debris in 24 streambed landscapes with leaf packs suspended from wires that snagged organic debris as it moved downstream. We arranged 64 wires (each with 3 leaf packs) in each landscape so that they were distributed among 8 large (aggregated) or 64 small (fragmented) patches. We manipulated landscape stability by removing accumulated debris from 0% (stable), 50% (semi-stable), or 100% (unstable) of the patches in each landscape 3 d after a flood. We determined chironomid density in leaf packs 2 d before the flood, 3 d after the flood, and 5 d after manipulating stability (8 d after the flood). We mapped the debris in each landscape and used multiple regression to evaluate the effect of landscape characteristics on density. Mean patch area was significantly smaller, and the number of patches, landscape perimeter, and landscape perimeter to area (P:A) ratio were significantly greater in fragmented than in aggregated landscapes. Chironomid density was significantly higher in fragmented than in aggregated landscapes and significantly lower in all landscapes after the flood. Density was ∼30% higher in unstable landscapes than in stable or semi-stable landscapes after debris removal. Density was negatively correlated with patch area, and debris accumulation may have caused patches to lose their value as refugia by making patches too large for chironomids to enter. Thus, disturbances that maintain fragmented patch arrangements with small patches and high landscape P:A ratios may be critically important for retention of individuals and to population persistence in dynamic landscapes.