The linear functional and structural relationships are of great importance in the physical, biological, economical and agricultural »cinences but the estimation of these relation is not always saticfactory. We have shown, by an empirical examination, that when the errors are uncorrelated, the likelihood surface contains a saddle point and that the maximum likelihood method leads to this saddle point and not to a real maximum. We have thus confirmed the work done hy Solari (1969) and shown that MAXÎMUM LÎKELIHOOD METHOD fails to solve for this problem.