Animals scan their visual environment to respond to threats and locate food sources. The neural computations underlying the selection of a particular behavior, such as escape or approach, require flexibility to balance potential cost and benefit to the animal’s survival. For example, avoidance of a visual object reduces predation risk, but negatively affects foraging success. Zebrafish larvae approach small, moving dots (‘prey') and avoid large, looming dots ('predator'). We found that this binary classification of objects by size is strongly influenced by feeding state. Changes in behavior correlate with shifts in neuronal responses to prey cues in the optic tectum. Both behavior and tectal function are modulated by signals from the hypothalamic-pituitary-interrenal axis and the serotonergic system. Our study has revealed a neuroendocrine mechanism by which the motivation to eat influences the perception of food in a vertebrate visual system.