Abstract

In cognitive neuroscience, there is great consensus that the brain interprets sensory input based on expectations and that it continuously generates predictions about what happens next. However, sensory input is often ambiguous, i.e. allows several interpretations, and there is often more than just one possible future scenario. How does the brain deal with those multiple options? In our natural environment, objects never occur in isolation (e.g., object ‘A’ is followed by object ‘B’; ‘C’ appears together with ‘D’). Such statistical regularities can serve as a constraint for interpretation and prediction. I will present experimental work related to the question how such associative learning is used to form accurate predictions about uncertain upcoming visual events. Furthermore, the centrality of prediction for our nervous system raises the question of whether associative information, as a proxy for predictability, is linked to reward signals. Such an affective tag could enforce an active search for predictable information, thereby warranting the creation of an ecological niche with a high degree of predictability and stability. Here I will present results of experiments that provide evidence that such associative information is preferred over non-associative information.