A model for visual feeding by aquatic predators is derived. The predator's visual range, which depends on its visual capability, surface light, water clarity, and size and contrast of the prey, is emphasised. Central to the model is the assumption that a prey may be recognized only if the difference in retinal flux, with and without the prey image, exceeds a threshold. This assumption is equivalent to requiring that the product of apparent contrast at retina, retinal background irradiance and area of prey image must exceed a threshold. Visual range (r) is found from the equation r2exp (cr + Kz) = rhoE0 \ C0 \ pibeta2DELTAS(e)-1, where c is beam attenuation coefficient, z is depth, K is diffuse attenuation coefficient, rho is light loss through the surface, E0 is surface light intensity, C0 is inherent contrast of prey, beta is prey radius and DELTAS(e) is sensitivity threshold of the eye for detection of changes in irradiance. The model predicts that visual range increases non-linearly with increasing predator size and ambient light. Visual range also increases almost linearly with increasing prey size and decreases non-linearly with increasing turbidity. These predictions are compared with experimental data. It is shown that characteristic fluctuations in light regime may be more important to feeding than characteristic variations in prey abundance in aquatic environments. Due to the direct impact of light on the feeding process of several predators (and thereby on the mortality process of prey), we conclude that light should be considered an important top-down control in aquatic ecosystems in addition to the bottom-up control exerted through primary production. Finally, the model is testable, and should stimulate a stronger interaction between theory and experiments in aquatic feeding ecology of visual predators.