In any flux-limited sample a tight correlation of luminosity (P) and redshift (z) is inevitable. It is therefore necessary to obtain complete samples at lower and lower flux-limits in order to have adequate coverage of the P–z plane, essential if we are to decouple the trends in epoch from trends in luminosity. This we have done for a number of flux-limits — giving coverage of the P–z plane seen in Fig. 1. Our redshift information is spectroscopic; the results of Eales et al (in prep.), namely the increased scatter in the

K-z plot for samples lower in luminosity than 3C, strongly warn us against using redshifts estimated from K-magnitudes.