In this note we discuss the following problem: Whether or not there exists a nonlinear mean-square consistent estimate of a linear function of regression coefficients, in case that a linear one does not exist? Examples and counter-examples are presented, together with some conjectures related to this problem.