Functional responses describe the per capita consumption rate of consumers and have been central to ecological theory since Holling’s seminal work. Holling observed that feeding rates generally saturate with increasing resource density and may decline at high consumer densities due to interference among individuals. Although functional responses underpin much of ecological modeling, their derivation from first principles has remained unclear. Here we show that this depends critically on whether feeding rates are averaged over time for individuals or across populations—a distinction absent from existing literature. We provide a unified framework that clarifies the common foundations and assumptions of different functional response forms, which offers a new perspective on how consumer behavior and stochasticity shape emergent feeding rates and, as a consequence, consumer–resource dynamics. We do so by revisiting a classic approach that models feeding interactions using individual-based reaction schemes, which highlights the shared assumptions underlying various functional response models. We show that the classic Beddington–DeAngelis functional response can be derived by time-averaging the consumption rate of a single consumer. In contrast, averaging across a population of consumers yields a distinct predator-dependent functional response—a natural extension of the population-averaged Holling Type II response that incorporates different forms of consumer interference. We further extend this framework to derive Holling Type III and multispecies versions of the Holling Type II functional response. Using synthetic data, we illustrate how the associated probability distributions can be used to estimate model parameters from feeding experiments. In contrast, averaging across a population of consumers yields a distinct predator-dependent functional response.