Traditional approaches to diet analysis in fish established that niche breadth increases as the fish grows. However, this appears to be artefactual, influenced by sampling, measurement, and statistical difficulties. On a ratio scale, niche breadths appear to be characteristic of a species and (generally) constant as the fish grows. In general there seems to be no equivalent to Roughgarden's 'between-phenotype-component' in ratio-based niche widths of fish. From the data available, there also appears to be no consistent trend of niche breadth among fish predator species of different mean sizes. Similarities to and differences from Ursin's log weight-ratio analysis are discussed. Assuming equal or predictably declining prey biomass in equal logarithmic length intervals (a Sheldon spectrum), a constant logarithmic (ratio) niche breadth means a constant or declining effective prey biomass pool as predators become larger. This contrasts with the generally expanding prey biomass pool postulated in the size-efficiency hypothesis. When prey biomass equilibrium can be assumed, declining productivity with size will further reduce resources available to increasingly large predators. A declining resource base would act to offset any size-related gains in metabolic or foraging efficiency.