[maro]

I don't see why that would be true. Consider the nonfunctional mutation that Nebullis is so fond of. Wouldn't you expect that the percentage of individuals in a population having this mutation (the "Empirical Percentage") would be constant, at least for relatively short periods of time after the first instance shows up? And wouldn't the probability of the mutation occurring at all be precisely the Empirical Percentage? (How could it be anything else?) (When populations come close to extinction, they go through a bottleneck, which randomly causes certain mutations to become prolific. But the a priori probability that any particular nonfunctional mutation would survive and proliferate is the same as its probability of occurring.)

[Nebullis]

But the a priori probability that any particular nonfunctional mutation would survive and proliferate is the same as its probability of occurring.

It depends on neighboring selection. Let's take a protein as an example. The survival of the whole protein and all the neutral mutations in it, depend on the selection of non-neutral mutation. Thus, the rate of neutral mutations are calculated from less sequence dependent portions of the genome.

[edsheppa]

Apologies, I should have explained. Mutations (and I'm thinking snp here - change this G to an A during this replication) occur by a process totally unlike "snuffing out." The former leads to an essentially constant likelihood (but I bet it varies by locus) based on the chemistry of genome replication. The latter is dependent on population size. For example, in a fixed sized asexual population given mild assumptions on reproduction, it is statistically certain that some descendent generation will have a single common ancestor. So, for example, if the population size is 100 and the neutral mutation occurs oin one of them, it has a 1% chance of becoming fixed which of course means it won't be snuffed out.