Theory suggests that intraspecific trait variation will alter species interaction strengths through nonlinear averaging when interaction strengths are nonlinear functions of individuals' traits. This effect is expected to be widespread, yet what factors mediate its magnitude in nature and hence its potential effects on ecosystems and communities are unclear. We sought to quantify how nonlinear predator functional responses, variation in prey densities and counteracting variation in attack rates and handling times among predator individuals of similar body size alter their population-level feeding rates through nonlinear averaging in a natural system, and to determine the processes influencing the net magnitude of this effect. We used a field caging experiment in the rocky intertidal of Oregon, USA to quantify attack-rate variation and feeding rates of the whelk Nucella ostrina on its barnacle and mussel prey. We also used empirically parameterized simulations to examine the effects of handling-time variation among individuals on population-level feeding rates. Within cages, individual attack-rate variation reduced population-level whelk feeding rates. However, the magnitude of this reduction differed among prey species and cages depending on cage-specific magnitudes of attack-rate variation and functional-response nonlinearity. The inferred effects of handling-time variation among individuals were of smaller magnitude than those of attack-rate variation, yet counteracted them to cause a net weakening of the effect of individual attack-rate variation on population-level feeding rates. Across cages, attack-rate and prey-density variation had non-additive effects that produced greater feeding-rate reductions at the experiment scale relative to the cage scale. Our results indicate that the effects of trait variation via nonlinear averaging depend critically on the features of systems that determine the magnitudes of nonlinearities and trait variation. Because of counteracting trait variation, nonlinear-averaging effects may be quite complex, involving both the variances and covariances of all traits and environmental variables influencing the ecological process of interest.