Much of the focus on Poincaré’s philosophy of science has been on the notion of convention, a crucial concept that has become distinctive of his position. However, other notions have received much less attention. That is the case of verifiable hypotheses. This kind of hypotheses seems to be constituted from the generalization of several observable facts. So, in order to understand what these hypotheses are, we need to know what a fact to Poincaré is. He divides facts into brute and (...) scientific facts. The characterization of this duality is not trivial at all, and leads us to the following questions that we will discuss in this paper: which the part of construction that exists in a scientific fact and which the part of translation, that is, what remains from the brute fact in the scientific one?; and when we conceive a generalized hypothesis, are we supposed to do it from scientific or from brute facts? The clarification of these questions could lead to distinguish the part of construction and the part of translation in the first steps of science, which is essential to get a better understanding of Poincaré’s conception of science. (shrink)