Metric characters closely connected with fitness have little additive genetic variability, presumably because it is quickly exhausted under continuous directional selection on fitness. Other metric characters have substantial additive genetic variability with a typical heritability of about 0.5. A popular model is that the second class of characters is subject to weak stabilizing selection for an optimal value, which depletes genetic variability, while recurrent mutation tends to restore it. Can this model account for the variability observed, given the evidence available about the strength of selection and mutation rates? Much theoretical work has been done on this complex problem. This work is reviewed, with the intention of simplifying it as much as possible.Key words: mutation–selection balance, genetic variability, continuum-of-alleles model, house-of-cards approximation.