I did a few experiments in Conway Life a long time ago. I tried adding a low-probability of error to the rule -- after the regular rule executed, I modified each cell with a small probability. I've forgotten most of the details, but generally .... An initially 50% random bit field decays to about 2.8% of the cells ON. The timescale is very roughly ~10K generations. The final percentage of ON cells has a weak dependence on the starting %. (If P is near 0 or 1, the field is blank after a few steps.) [Several Lifers have done extensive experiments, measuring the probability of various patterns in the 'final' configurations.] If ~.01% of the cells are randomly forced to OFF after each time step, or, " " " forced to ON, or, " " " complemented, the eventual outcome is a blank pattern. I did one experiment about the speed and distance of 'change propagation': I made two random 8Kx8K fields, with the left halves matching and the right halves random. I ran a number of steps, and looked at how well the left halves matched. The 'change front' propagated about 500 cells into the left half, gradually slowing, and seemed to have roughly stabilized after O(10K) steps. Rich