Russell's philosophy is rightly described as a programme of reduction of mathematics to logic. Now, the theory of geometry developed in 1903 does not fit this picture well, since it is deeply rooted in the purely synthetic projective approach, which conflicts with all the endeavours to reduce geometry to analytical geometry. The first goal of this paper is to present an overview of this conception. The second aim is more far-reaching. The fact that such a theory of geometry was sustained by Russell compels us to question the meaning of logicism: how it is possible to reconcile Russell's global reductionism standpoint with his local defense of the specifities of geometry?