Foraging profitability can be strongly affected by the size structure of different prey, so that predator distributions are not a simple function of total prey biomass. For a bottom-feeding avian predator, the surf scoter Melanitta perspicillata, we assessed effects of prey size and other prey attributes on seasonal shifts in scoter use of 2 major foraging habitats in Puget Sound, Washington, USA. During early winter, many thousands of scoters fed at an unvegetated site where profitable prey appeared limited to mussels Mytilus trossulus of smaller sizes (2 to 30 mm) despite their much lower biomass relative to larger mussels and several other prey types. Accordingly, scoter numbers decreased at that site as small mussels declined over winter. During pre-migratory fattening in spring and feather molt in summer, >8000 surf scoters aggregated at a seagrass site where they fed mainly on epifaunal crustaceans (50 to 73%) and gastropods (12 to 27%). Body sizes of most crustacean prey had increased substantially since winter. Thus, prey size had opposite effects on the profitability of unvegetated habitats that provide mainly mussels (smaller items likely reduce shell processing costs) versus seagrass crustaceans (larger items are likely more visible and yield greater energy per prey item, although relative mobility of prey can alter their value). Total prey biomass, and prey distributions relative to water and sediment depths, appeared less important than prey size to shifts in scoter diets and numbers. Our synthesis of past studies indicates that biomass and production of mussel beds are typically an order of magnitude greater than for entire assemblages of seagrass macroinvertebrates. However, because of seasonal shifts in prey size structure, seagrass sites can be an important complement to mussel beds when the narrow size fraction of mussels that are profitable to scoters declines.