Is it possible to be realistic and say truth in human and social sciences (HS), when examining the way we thought, mean, or behave? Traditionnally, the HS answer these questions in a metaphysical way, founding themselves in the reality and truth of things they examine seen as specifically human. But what we are asking to the HS is no more than examining human comprehension – in this, they are just ordinary activities using the language. So, our question is one of these to be answered within the ordinary language. In an opposite way, when considering that meaning and comprehension are no more but uses, one may incline to consider any knowledge about this as absolutely relative to situations; truth itself may be seen as relative, and HS as non-realistic. These two ways are aporetic: on one side, the HS appear as non-scientific and metaphorical discourses, on the other, quite impossible. Our sciences are worth another way making possible realistic uses of ''true''; I hold that this can be found in a philosophy of ordinary language sharing with HS the same material: meaning, mutual comprehension. From Austin, they can inherit a non-essential conception of what appears as reality in meaning; from Cavell, the analysis of human comprehension in our agreement within the langage; from Diamond, a solution to the question of realism by considering it in our attitudes. To illustrate these uses of ''true'' in HS, I propose a reading of Foucault showing him search a realistic attitude. As a conclusion, I describe how HS must be seen as practices: not in a verificationist hope, but in the acceptation of their depending of our uses of ''true'' and attitudes.
