Abstract
The aim of this paper is to show that the concept of Fréchet differentiability of von Mises statistical functionals appears naturally in statistical inference if we impose, on “approximately" correct parametric models, regularity conditions similar to those typically used for the parametric inference. It is shown that a functional which is regular at a rich family of smooth parametric models containing F and for which the asymptotic limits satisfy a natural continuity property over shrinking Cramér von Mises neighbourhoods of F, is Fréchet differentiable at F for the Cramér von Mises norm.