Discussion of: Statistical Methods in Psychiatry: Virtual Models

Abstract

Victor, whom Prof. Barra cited, took a stand against the tendency to believe that there is “automatically good science” if results are based on tests of statistical significance. In effect, these presuppose a univocal hypothesis and specific conditions. Now in psychiatry, on the one hand, the state of knowledge requires above all that research proceed step by step as one would explore a poorly known continent and, on the other hand, the prerequisites of strictly decisional statistics are not respected in the investigations owing to the practical difficulties that they raise. This is why the first “Meeting of Methodology of Research in Psychiatry” in Marseille in 1971 took these “new” statistics represented by data analysis as its theme. They were just becoming known then and have since enjoyed an extraordinarily wide diffusion penetrating all those disciplines which CANGUILHEM qualified as “formless” in opposition to the methodologically constituted sciences. They are in effect “statistical instruments”, the application of which is particularity seductive because of the few constraints they place on the collection of “phenomena” and at the same time because of all they enable to reveal concerning structures or inapparent regularities. But in our workshop, which had decided to carry on its reflective effort, the same questions cropped up time and time again asking whether precisely this facility was not dangerous and this fecundity illusory.