Neuronal activity in the primate superior colliculus (SC) is modulated by physical salience, reward value, and attention, but how these signals are integrated remains unknown. One hypothesis is that SC activity reflects a unified “priority map” of the visual field. However, it is unclear how a unified priority map might support distinct roles for SC activity in different contexts: If the same SC neurons are activated by saccade cues and attention cues, how can SC activity evoke a saccade in one context but covert orienting of attention in another context? We hypothesized that distinct sources of modulation cause dissociable rather than unified patterns of SC activation, which would support a context-specific relationship between SC activity and behavior. To test this hypothesis, we recorded 220 SC neurons in a macaque performing two tasks: a covert change detection task manipulating attention, and a saccade task independently varying reward value and salience. All three factors influenced behavior: salience affected saccadic endpoint error, reward modulated reaction time, and attention cueing affected detection rate. Single-neuron, ROC areas for each factor suggested that salience, reward, and attention exert independent influences on SC. At the population level, linear classifiers trained to decode conditions for one factor (e.g., reward) failed to generalize to other factors (e.g., salience), indicating independent population-level encoding. These findings challenge the unified priority map model and suggest SC could support flexible visually guided behaviors by selective routing. This mechanism could explain how SC contributes to overt orienting and covert attention depending on context, advancing our understanding of how the brain efficiently processes visual information to guide behavior. Our results open new avenues for investigating how neuronal modules participate in diverse functions.