Surveys and registers are the sources of statistics. Their merits lie chiefly in their cost and their contribution to knowledge. They give form to the social world, but they depend on conventions regarding registration. Making surveys and keeping registers imply the cognitive and political construction of an area of equivalence affected by institutional contexts. They are used to assess public policies through indicators, by distinguishing between ends and means. Yet registers reflect means more than the state of society. This critical examination has been carried out by analysing the joint construction of ways of thinking: society, public policy, and descriptive tools.