``Selection" is the differential survival and reproduction (aka ``fitness") of individuals. From an evolutionary perspective, we are most often interested in the fitness associated with particular genotypes, and a simple parametrization of single-locus genotypic fitnesses might be:
Different values of s (usually called the selection coefficient) and h (the dominance coefficient) produce the familiar dominance, recessive, additive, overdominant, and underdominant relationships. Kimura[8] used the following parametrization (equivalent to the additive case above)
and showed that the probability of fixation of an allele under this type of genic selection is approximately
in an otherwise ideal population. As 
this
probability approaches 
, as expected from our discrete-process
analysis. For small (positive) s and large N, we can approximate
. There are a few interesting consequences of this
result. The first is that even an advantageous mutation is not guaranteed
to increase in frequency, and in fact, the probability that it does may be
quite small. Notice also that now, the rate of substitution of
advantageous alleles is 
. Many people who are willing to assume
a constant mutation rate in order to be able to construct a phylogenetic
tree balk (with good reason) at the prospect of the number of
simultaneous, delicately balanced, suppositions that would have to be made
to construct a tree of selected substitutions along the same lines. On
the other hand, it has become increasingly clear that the statement that
most evolution at the molecular level is neutral is, at best, an
oversimplification.