Our collaborative work focused on understanding the system of mechanisms influencing the mortality of juvenile pink salmon (Oncorhynchus gorbuscha) in Prince William Sound, Alaska. Coordinated field studies, data analysis and numerical modelling projects were used to identify and explain the mechanisms and their roles in juvenile mortality. In particular, project studies addressed the identification of major fish and bird predators consuming juvenile salmon and the evaluation of three hypotheses linking these losses to (i) alternative prey for predators (prey-switching hypothesis); (ii) salmon foraging behaviour (refuge-dispersion hypothesis); and (iii) salmon size and growth (size-refuge hypothesis). Two facultative planktivorous fishes, Pacific herring (Clupea pallasi) and walleye pollock (Theragra chalcogramma), probably consumed the most juvenile pink salmon each year, although other gadids were also important. Our prey-switching hypothesis was supported by data indicating that herring and pollock switched to alternative nekton prey, including juvenile salmon, when the biomass of large copepods declined below about 0.2 g m−3. Model simulations were consistent with these findings, but simulations suggested that a June pteropod bloom also sheltered juvenile salmon from predation. Our refuge-dispersion hypothesis was supported by data indicating a five-fold increase in predation losses of juvenile salmon when salmon dispersed from nearshore habitats as the biomass of large copepods declined. Our size-refuge hypothesis was supported by data indicating that size- and growth-dependent vulnerabilities of salmon to predators were a function of predator and prey sizes and the timing of predation events. Our model simulations offered support for the efficacy of representing ecological processes affecting juvenile fishes as systems of coupled evolution equations representing both spatial distribution and physiological status. Simulations wherein model dimensionality was limited through construction of composite trophic groups reproduced the dominant patterns in salmon survival data. In our study, these composite trophic groups were six key zooplankton taxonomic groups, two categories of adult pelagic fishes, and from six to 12 groups for tagged hatchery-reared juvenile salmon. Model simulations also suggested the importance of salmon density and predator size as important factors modifying the predation process.