Empirical descriptions of cosmological evolution rely on an assumed relation between luminosity distance and redshift in order to derive the luminosity function and its epoch dependence. This relation is usually taken to be that of an idealised Friedmann cosmology, despite the obvious abundance of small-scale structure in the Universe. However, the assumption that the probability of a lensing event is negligible is now made less tenable with the discovery of the double and triple QSO's, both of which appear to have been magnified by factors of 10–15. With at least two events of this amplification in 1500 known quasars, it may be that a large fraction of observed QSO's have been magnified by significant amounts. Turner (1980) suggests that all evolutionary statistics (e.g. V/Vmax) may thus be misleading, although he considers only an illustrative model of the effect. The problem is that, even if the intrinsic probability of a lensing event is very low, the effect becomes important if the background density of faint sources rises sufficiently quickly. We must therefore calculate not only the intrinsic probability of a given amplification, but incorporate this with the luminosity function to see if lensing could be made dominant by selection effects.