1995A&A...294..107V

Using the most recent evolutionary computations of massive single stars and of massive close binaries, we discuss the theoretically predicted distribution of WR stars in the Galaxy and in the MC's. We conclude that when standard massive single star evolutionary computations are used, correspondence between theory and observation is very poor. We then propose an alternative model for single star evolution. Using this model, we rediscuss the predicted WR distribution. We conclude that correspondence with observations is obtained provided that a. the stellar wind mass loss rate during the RSG phase for single stars with initial mass larger than 20M_☉ and during the LBV phase for single stars and binary components with initial mass larger than 50M_☉ is significantly larger (up to a factor 10) than predicted by semi-empirical M_dot relations used in present day stellar evolutionary codes, b. the stellar wind mass loss rate during the WR phase is slightly metallicity dependent and c. the O-type binary frequency is a decreasing function of metallicity.