Animals can be important in nutrient cycling through a variety of direct and indirect pathways. A high biomass of animals often represents a large pool of nutrients, leading some ecologists to argue that animal assemblages can represent nutrient sinks within ecosystems. The role of animals as sources vs. sinks of nutrients has been debated particularly extensively for freshwater fishes. We argue that a large pool size does not equate to a nutrient sink; rather, animals can be nutrient sinks when their biomass increases, when emigration rates are high, and/or when nutrients in animal carcasses are not remineralized. To further explore these ideas, we use a simple model to evaluate the conditions under which fish are phosphorus (P) sources or sinks at the ecosystem (lake) level, and at the habitat level (benthic and water column habitats). Our simulations suggest that, under most conditions, fish are sinks for benthic P but are net P sources to the water column. However, P source and sink strengths depend on fish feeding habits (proportion of P consumed from the benthos and water column), migration patterns, and especially the fate of carcass P. Of particular importance is the rate at which carcasses are mineralized and the relative importance of benthic vs. pelagic primary producers in taking up mineralized P (and excreted P). Higher proportional uptake of P by benthic primary producers increases the likelihood that fish are sinks for water column P. Carcass bones and scales are relatively recalcitrant and can represent a P sink even if fish biomass does not change over time. Thus, there is a need for better documentation of the fraction of carcass P that is remineralized, and the fate of this P, under natural conditions. We urge a more holistic perspective regarding the role of animals in nutrient cycling, with a focus on quantifying the rates at which animals consume, store, release, and transport nutrients under various conditions.