We evaluated the effects of brown trout on boreal stream food webs using field surveys and enclosure/exclosure experiments. Experimental results were related to prey preference of uncaged trout in the same stream, as well as to a survey of macroinvertebrate densities in streams with vs. without trout. Finally, we assessed the generality of our findings by examining salmonid predation on three groups of macroinvertebrate prey (chironomid midges, epibenthic grazers, invertebrate predators) in a meta-analysis. 2. In a preliminary experiment, invertebrate predators showed a strong negative response to trout, whereas chironomids benefited from trout presence. In the main experiment, trout impact increased with prey size. Trout had the strongest effect on invertebrate predators and cased caddis larvae, whereas Baetis mayfly and chironomid larvae were unaffected. Trout impact on the largest prey seemed mainly consumptive, because prey emigration rates were low and independent of fish presence. Despite strong effects on macroinvertebrates, trout did not induce a trophic cascade on periphyton. Uncaged trout showed a strong preference for the largest prey items (predatory invertebrates and aerial prey), whereas Baetis mayflies and chironomids were avoided by trout. 3. Densities of invertebrate predators were significantly higher in troutless streams. Baetis mayflies also were less abundant in trout streams, whereas densities of chironomids were positively, although non-significantly, related to trout presence. Meta-analysis showed a strong negative impact of trout on invertebrate predators, a negative but variable impact on mobile grazers (mainly mayfly larvae) and a slightly positive impact on chironomid larvae. 4. Being size-selective predators, salmonid fishes have a strong impact on the largest prey types available, and this effect spans several domains of scale. Discrepancies between our experimental findings and those from the field survey and meta-analysis show, however, that for most lotic prey, small-scale experiments do not reflect fish impact reliably at stream-wide scales. 5. Our findings suggest that small-scale experiments will be useful only if the experimental results are evaluated carefully against natural history information about the experimental system and interacting species across a wide array of spatial scales.
