The paper proposes to revisit Keynes’s Treatise on Probability . First, it stresses the Leibnizian origin of Keynes thought process. Second, it tries to unveil the real scope of the research program which underlies the book. Finding the logical foundations of probability characterized as the degree of rational belief in the truth of a proposition. It shows how such a program has been badly translated in the inadequate terms of Russell’s system of logic. Reducing the degree of belief to a logical relation opens the way to Ramsey’s criticism and misunderstandings. Finally, the paper suggests that some ideas to be found in the Treatise , as for example the notion of “relevance”, have anticipated several concepts being developed today in a cognitivist perspective for understanding individual beliefs.