Neuroimaging techniques enable us to establish associations between
brain activity and cognitive functions. Associations do not inform us,
however, about the causal mechanisms that govern how brain functions
give rise to cognitive processes. To advance our understanding of the
brain, it is essential to move beyond associational studies and identify
the causal mechanisms that form the basis of cognition. In this talk, I
investigate which causal statements are warranted and which ones are not
supported by neuroimaging experiments. I argue that in order to decide
this question it is necessary to distinguish between encoding- and
decoding models as well as between stimulus- and response-based
experimental paradigms. I then present a set of statistical conditions
that is, under some technical assumptions, sufficient (but not
necessary) to establish that a brain process is a cause of a behavioural
response, even in the presence of latent confounders. I discuss how to
test these conditions on empirical data and demonstrate the viability of
this approach on EEG data recorded during a neurofeedback study.