Abstract
The authors analyse and extend the real-space renormalisation group proposed by Reynolds et al. (1977) to treat site percolation. The best of three possible definitions of 'percolating' configurations and of five possible methods of weighing these configurations, are established for percolation in square lattices. The use of n*n square clusters leads, for n=2 (n=3 and n=4), to nu p approximately=1.635, nu p approximately=1.533 and yp approximately=1.498, and also to pc approximately=0.382, pc approximately=0.388 and pc approximately=0.398, respectively exhibiting in this way the correct (but slow) tendency towards the best available values.