The history of statistics, a laboratory for social theory.

La politique des grands nombres (The politics of large numbers) by Alain Desrosieres and La mesure de I'Etat (The measure of State) by Eric Brian are devoted to the history of statistics or mathematics and their links with social sciences and the administration. By comparing the theoretical approach developed in these two works, it is possible to examine in what manner recent developments in history and sociology of sciences have contributed to sociology as a whole. if the authors of these works take a stand in relation to the problems put forward by constructivism and relativism, it is not in the same way. This can be seen by the questions they ask, in the methods adopted and in the type of reply they are seeking, as well as in the period covered : short and detailed for E. Brian, longer and more extensive for A. Desrosieres. For the latter, constructivism is found in the concern for political consequences leading to the construction of statistical objects. His position, basically relativist, is also accompanied by a classificatory approach aiming to highlight what he considers to be the intrinsic (and trans-historic) characteristics of statistics. On the contrary, E. Brian puts forward a methodological constructivism which takes into account both the constructed nature of historical sources and that of his own analysis. This is the approach of a realist who needs a base in order to hold an epistemological control over the construction of this object.