The importance of genetic interactions in the evolutionary process has
been debated for more than half a century. Genetic interactions such
as under dominance and epistasis (the interaction among genetic loci i
n their effects on phenotypes or fitness) can play a special role in t
he evolutionary process because they can create multiple fitness optim
a (adaptive peaks) separated by fitness minima (adaptive valleys). The
valleys prevent deterministic evolution from one peak to another. We
review the evidence that genetic interaction is a common phenomenon in
natural populations. Some studies give strong circumstantial evidence
for multiple fitness peaks, although the mapping of epistatic interac
tions onto fitness surfaces remains incompletely explored, and absolut
e proof that multiple peaks exist can be shown to be empirically impos
sible. We show that there are many reasons that epistatic polymorphism
is very difficult to find, even when interactions are an extremely im
portant part of the genetic system. When polymorphism results in the p
resence of multiple fitness peaks within a group of interbreeding popu
lations, one fitness peak will quickly be nearly fixed within all inte
rbreeding populations, but when epistatic or under dominant loci are n
early fixed, there will be no direct evidence of genetic interaction.
Thus when complex landscapes are evolutionarily most important, eviden
ce for alternative high fitness genetic combinations will be most ephe
meral. Genetic interactions have been most clearly demonstrated in wid
e crosses within species and among closely related species. This evide
nce suggests that genetic interactions may play an important role in t
axonomic diversification and species-level constraints. Population gen
etic analyses linked with new approaches in metabolic and molecular ge
netic research are likely to provide exciting new insights into the ro
le of gene interactions in the evolutionary process.