It is undisputed that there is a link between statistical (in)dependencies and causal relations between events, but the exact nature of this link is still controversial. We will only consider the epistemological question: how does statistical information (in the crude sense of dependency vs. independency) guide the discovery of causal relations? Our starting point is Reichenbach's screening off criterion, and its generalization in the Markov condition. This will allow us to present an adaptive mechanism that enables one to infer causal relations from (in)dependency information, conditional on some constraints. It is clear that this makes possible the development of an adaptive logic for causal reasoning, about which we will give some clues.