NOTE 23A: The population is assumed to be so large that Muller’s ratchet does not operate; that is, the fittest class is common enough not to be lost by chance.

NOTE 23B: A single mutator allele is most likely to be lost, but it may increase; the outcome is highly random. However, as long as we concentrate on the *expected* number of offspring, we can ignore this randomness.

NOTE 23C: This is known as **soft selection** (see p. 673 and Web Notes).

NOTE 23D: This model is explored in detail by Burger (1999) and by Waxman and Peck (1999).

NOTE 23E: Recall that if the distribution of offspring number follows a Poisson distribution with mean 1, then the variance is also 1, and we have the standard **Wright–Fisher** formula, *pq*/*2N*. Also note that it is simplest to focus on the fitness of individual genes; Box 15.2 gives a formula for the rate of random drift in terms of the variance in fitness of diploid *individuals*, which is more complex.