Well I was actually thinking about mutations :) Let's consider a simplified model, where there are only two different bases (A and G instead of A, T, G, C) and

Well I was actually thinking about mutations :) Let's consider a simplified model, where there are only two different bases (A and G instead of A, T, G, C) and assume that the probabilities of mutations A->G and G->A are the same. Let's consider a single base mutation[1]. No matter what the mutation probability is, and no matter what the initial distribution is, I expect that the population will converge to about 50% A and 50% G. I guess you would agree. Now if we make this very wrong model a bit less wrong by including the four base types A, T, G, C, we can expect that each allele will be found in about 25% of the population. And that is for single base mutations only. A mutation that depends on the precise value of 5 different bases would be displayed by a much smaller share of the population (1/4^5 = 0.1% I guess). Of course people don't get born with fully random genomes like that, the vast majority of these mutations would not yield a living embryo. But still, the intuition we can get from this is that without evolutionary pressure, a particular mutation will eventually be found in a fraction of the population that is inversely proportional to the complexity of the mutation (in most cases that would be far from the 50% of my initial intuition). No idea how close these layman's speculations are from the reality... [1] Such as the gene that determines the type of ear wax: http://udel.edu/~mcdonald/mythearwax.html http://udel.edu/~mcdonald/mythearwax.html