Predator-prey interactions are one of the main ecological factors influencing the structure of fish communities. The impact of wading and diving semi-aquatic predators on riverine fish populations is poorly known. We studied the effect of feral American mink predation on brown trout juveniles during winter in two experiments conducted in semi-natural streams (length 26 m, width 1.5 m). In the first experiment, we compared the vulnerability of age-1+ hatchery (length 142 +/- 16 mm, average +/- SD) and wild (112 +/- 8 mm) brown trout of similar genetic origin in sympatry and allopatry. In the second experiment we used age-0+ brown trout (79 +/- 5 mm), increased habitat heterogeneity by addition of fine wood, and compared those to treatments without fine wood addition. Hatchery fish were more vulnerable to mink predation than their wild counterpart, and the predation rate increased with increasing body size among the hatchery trout. Predation by mink on wild trout was higher in sympatric than in allopatric treatments suggesting that stocking of hatchery fish may increase predation on wild conspecifics. Increased habitat heterogeneity resulted in reduced predation rate. The results show that a large size of hatchery fish in small streams was a negative trait, which was opposite to the mainstream observations of salmonid stockings made directly of feeding areas in lakes and oceans. Adding habitat heterogeneity was found important for habitat enhancements in streams with mammalian predators. We highlight the importance of taking all the habitats during the life cycle of migratory fish into account in management decisions and carefully considering whether using hatchery fish to support wild populations in streams.