"Switching" in predators which attack several prey species potentially can stabilize the numbers in prey populations. In switching, the number of attacks upon a species is disproportionately large when the species is abundant relative to other prey, and disproportionately small when the species is relatively rare. The null case for two prey species can be written: P1/P2 = cN1/N2, where P1/P2 is the ratio of the two prey expected in the diet, N1/N2 is the ratio given and c is a proportionality constant. Predators were sea—shore snails and prey were mussels and barnacles. Experiments in the laboratory modelled aspects of various natural situations. When the predator had a strong preference (c) between prey the data and the "null case" model were in good agreement. Preference could not altered by subjecting predators to training regimens. When preference was weak the data did not fit the model replicates were variable. Predators could be trained easily to one or other prey species. From a number of experiments it was concluded that in the weak—preference case no switch would occur in nature except where there is an opportunity for predators to become trained to the abundant species. A patchy distribution of the abundant prey could provide this opportunity. Given one prey species, snails caused a decreasing percentage mortality as prey numbers increased. This occurred also with 2 prey species present when preference was strong. When preference was weak the form of the response was unclear. When switching occurred the percentage prey mortality increased with prey density, giving potentially stabilizing mortality. The consequences of these conclusions for prey population regulation and for diversity are discussed.