For many years, measuring crime on the sole basis of computations of the operations of the public agencies in charge of its control implied great implicit confidence in the adequacy of their action to expectations for safety-maintenance. Over the last thirty years, all comparable countries have gradually come to resort at least partially to surveys, revealing their increasing doubts about this adequacy. In 1996 the Insee [the French national institute for statistics} inserted a module on victimisation in its survey of the living conditions of households, thus yielding an opportunity to measure medium-term variations from one decade to another, for the first time in France. Comparison of these findings with the only national survey available so far (Cesdip, 1986) shows how different the picture is for violence and property offences, in terms of numbers and of trends. It is also possible to evaluate the extent -variable, depending on the kinds of victimisation- to which official statistics still reliably account for trends as well as for orders of magnitude. Aside from its methodological teaching, this analysis modifies and clarifies present controversies over crime.