[en] The authors try to solve the following problems: provided that variation freedom is not restricted and intragroup correlation is taken into account in a given approximation, is it possible to develop for group functions any analogue of the well known (in the case of orbitals) Adams-Gilbert-Kunz method that could give one a possibility to go from delocalized group functions to localized ones seeking for minimization of the intergroup correlation? 20 refs, 1 tab