We conducted an experiment on a total of three confluence areas (1 with a burned stream (confluence 1, Deadman) and 2 with burned + debris flow streams (confluence 2 & 3, Frog Pond and Low Gradient) throughout August 2011 to help identify mechanisms responsible for confluence selection by fish. This experiment was needed to test hypothesis 2, which predicted that higher subsidies of prey to adjacent ecosystems would increase local abundances of predators. An experiment (modeled after Fausch et al. 1997) that reduced inputs of drifting invertebrate prey from the tributary stream helped determine if increased prey from headwater streams was the primary reason for preferential use of confluences. We selected confluence areas with high degrees of selection by trout and high insect inputs from tributaries, which we documented in August 2010. Further, we selected confluences with different topologies to allow us to evaluate if characteristics of the recipient habitat mediated the influence of cross-ecosystem subsidies on recipient consumers. Confluences 1 and 2 consisted of the tributary entering the main-stem river through a culvert. Compared to confluence 2, confluence 1 was located in shallow water in a low-velocity area. Confluence 3 was more isolated from the main-stem river because of the presence of large boulders, exported from the tributary during the debris flow. We placed drift nets (250 µm mesh size) in the tributary channels just upstream from their confluences with downstream-main-stem channel. Because nets spanned the entire width of tributary channels, we captured most of the tributary flow, subsequently reducing tributary inputs of insects to confluence areas. This allowed us to reduce inputs of drifting invertebrates to confluence areas without affecting other factors associated with these areas (i.e., thermal refugia). After nets were placed, we allowed 1 h to pass before the post-drift depletion surveys were conducted. We examined nets at least twice per 1 h and cleaned them to reduce possible flow alterations. We performed snorkel surveys of abundance and behavior (scan and focal observations Altman 1974) pre-and-post-drift depletion with each survey lasting approximately 1 h. Experiments were repeated three to four times per confluence reversing the order of pre and post drift-depletion surveys to account for varying feeding patterns throughout the day. At the beginning of each survey, we recorded all fish species present and their sizes (to the nearest 5 cm). We observed both juvenile Chinook salmon and rainbow trout focal fish, because these 2 species were the most common in confluence areas. Each focal fish was observed for 5 min per fish, recording all-feeding attempts and classifying them as either benthic, mid-water, or surface to detect any possible shifts in foraging mode. Additionally, we recorded all agonistic behaviors (both chases and fleeing) during focal observations. During each survey, we attempted to record at least three focal fish observations per species. Concurrent focal fish observations were conducted in nearby non-confluence habitats that did not receive direct inputs from a tributary, but exhibited similar water depth and velocity as confluence habitats. These observations were compared to observations taken in confluences to detect differences in feeding and agonistic behavior rates between habitats. Each confluence was treated as its own experimental unit in analysis because of its unique geometry and differences in concentration and yield of insect export between tributaries.