The Development of Population Genetics 3254 GENES AND SELECTION: THE SYNTHESIS ESTABLISHEDAlthough Fisher’s synthesis of Mendelian genetics with biometrical Darwinism
marked a turning point in the debate with Pearson disagreements of a somewhat
different kind began to emerge with the biologist Sewall Wright, another of the
founders of population genetics. Wright [1921] also showed how selection operated
in Mendelian populations and did so not only by using a different quantitative
method but also by incorporating an entirely different qualitative theory about
the way selection operated. Wright was convinced by his experimental work with
guinea pigs that it was the interaction systems of genes, rather than single genes,
that were the important elements of evolution. He also believed, contrary to Fisher,
that natural selection operated most effectively in smaller populations where in-
breeding was sufficiently intense to create new interaction systems as a result of
random drift, but not intense enough to cause random non-adaptive fixation of
genes. In those populations natural selection could act on the newly produced
interaction systems, resulting in a more rapidly changing population than that
produced by mass selection of single genes (as described by Fisher’s approach).
Wright felt that there was sufficient evidence from animal breeders that mass se-
lection was slow and somewhat unsure; hence he assumed that natural populations
tended to become subdivided into partially isolated groups small enough to cause
random drifting of genes. Thus, one of the key differences between Fisher and
Wright in attempting to isolate the way selection worked concerned population
structure. Fisher was convinced that it worked more efficiently in large groups in
which there was mass selection and random breeding, whereas Wright’s account
favoured small inbred groups in which obvious variation would then serve as the
basis for selection.
Wright’s evolutionary models and mathematical techniques were intended to
provide a way of focusing on a quantitative analysis of inbreeding. His use of the
method of path coefficients (a system he devised) enabled him to determine the
degree to which a given effect was produced by each of a number of different causes.
In that analysis the causes of factors that affected a trait (e.g.the general size of an
organism) and factors that affected its separate parts could be treated as indepen-
dent. A path coefficient could then be defined as the ratio of the variability of the
effect that is found when all causes are kept constant except the one in question
to the total variability. The path coefficient is identical with a partial-regression
coefficient provided that it can be measured in standard-deviation units. Wright,
however, wanted to stress the difference between dealing with multiple regressions
and dealing with his complex network of relations that involved hypothetical fac-
tors. As far as he was concerned, correlation and regression coefficients provided
descriptive/predictive statistics from which scientists unjustifiably inferred the ex-
istence of causal relationships. By contrast, the method of path analysis was not
used toinferacausalschemeordeduceparticular causal relationships; instead, it
required a specific causal hypothesis to be stated at the outset and was best ap-
plied in cases where causal pathways were known or could reasonably be assumed,