[Nota: revista indexada en Philosopher´s Index, Latindex y Dialnet, entre otros índices].In this work, we will try to state the opposition between two approaches to the problem of the overall reliability of human knowing capacities, and a possible solution to that conflict. On the one hand, as we will point out, there exist a number of approaches that fall under the broad term of ?evolutionary reliabilism? and according to which the reasons that we have for believing in the reliability of human cognition are empirical in character. Namely, the adaptive success of our species in a biological environment characterised by the survival of the fittest provides us with a reason to believe that our belief-forming mechanisms are truth-tracking; if they were not, we would have gone extinct. On the other hand, nevertheless, we find analyses?of which we will focus on the case of Ernest Sosa?according to which the tenet that our knowing capacities are reliable cannot be empirically based, and thus contingent; on the contrary, they point out that we need to presuppose?without need of a proof?that such capacities are trustworthy. Given this conflict between an ?empirical? defence and a defence ?in principle? of our knowing capacities, we will consider two possible objections against Sosa?s proposal and try to answer them.