Allele Fixation in Island Populations
Reduction of Heterozygosity in Small Demes
Consider an archipelago consisting of 1000 similar islands. Each island contains an isolated population of Species X, which has a gene locus segregating a dominant (A) and recessive (a) allele that is not under selective pressure.
All 1000 populations start with equal freqencies of dominant (A) and recessive (a) alleles at a particular locus (i.e., p and q both = 0.5)
As the number of generations passed increases, the 1000 populations
begin to diverge from the initial 0.5 frequencies due to SAMPLING ERROR.
As some populations go to q = 1.0, they will also go to p = 0.0
Approximately the same number are predicted to go to p = 1.0
Hence, the distribution of allele frequencies found in the 1000 populations
spreads symmetrically.
When the allelic frequency in a population reaches 1.0, the allele is
the only one left in the population, and it becomes fixed for that allele. The other
allele is permanently lost.
In populations in which an allele has become either fixed or lost, the
process of random genetic drift stops at that locus.
Without further input (mutation), the populations that have allelic
frequencies of 1.0 or 0 for either allele will maintain those allelic
frequencies forever.
Somewhere between 100 (N) and 200 (2N) generations, the distribution
of allele frequencies flattens out as some of the 1000 initial
populations are lost to
fixation/loss of alleles. (Think of the sides of the graph as "sinks" into
which these fixed/lost allele populations disappear.)
With the population size we chose for our example, there's a
loss of about 0.5% of the total 1000 populations to fixation/loss with each
generation.
Subsequent loss to fixation/loss becomes vanishingly small after the
generations reach about 300. It would take a HUGE amount of time for the
rest of the populations to be absorbed--practically infinite.
NOTE that if the initial allelic frequency is not 0.5, the curve will be
shifted to the right or left (depending on which allele is more frequent),
but the processes of movement towards homozygosity will be the same.

In this example, we started out with a very low frequency of a. The
probability of its being LOST from the population is far greater than its
being fixed. In this graph, most of the populations are fixed with AA--not
aa.