Animals, wild or civilized, are permanently in interaction with their environment through a perception-action loop. Their brains are experts in transforming the continuous flow of perceived information into action – continuously deciding the next action and seamlessly executing it. The present work addresses the mechanisms forming the stream from perception to action, which embody the interaction between signals driven by the environment and signals driven by the goals and expectations of the animal. In the case of the visuo-oculomotor system, which we take as a biological model, those signals converge in the intermediate layers of the Superior Colliculus (SCi), which serves as an interface spatially representing the possible eye movements. Interestingly, action execution can start while the selection is not completed, allowing us to infer the signals present in the SCi from the eye movements. In Chapter 2, using a computational model, we addressed the spatial interactions possibly occurring upstream the SCi and discuss their effects on behaviour. In Chapter 3, we inferred the presence of a spatiotopic signal in the SCi and refute current models of the visuo-oculomotor system. In Chapter 4, we introduced a new way to infer activity of the SCi, and we used it to distinguish the effects of goal-driven and expectation-related signals on the SC map. In Chapter 5, modelling separately the superficial and the intermediate layers of the SC based on recent neurophysiological recordings, we explored how neural properties and connectivity affect signal interactions. Finally, we discussed how we could implement the theories developed in this thesis, how our view of the visuo-oculomotor system could be refined, and whether this system could become a general model of decision making.