736 THE STRUCTURE OF EVOLUTIONARY THEORY
alleles (even if favorable) will be lost by random non-inclusion in all founding
organisms.
Although both genetic drift and founder effects obviously occur at the
organismal level, our traditions have tended to downplay the role of random
processes vs. selection as sources of sorting—so the phenomena generally receive
short shrift. Some conventional arguments for genuine rarity at the organismal
level may be valid, particularly given the requirement for either small populations
or effective neutrality of drifting sites. (The initiating criterion of low N may,
however, be quite generally met if Mayr's theory of peripatric speciation holds,
hence his emphasis on the "founder principle." Similarly, if bottlenecking to very
small numbers typically occurs during the history of many species, then genetic
drift also becomes important in anagenesis. The argument for effective neutrality,
as discussed previously (pp. 684-689), works best at the genie level, where drift
may predominate by Kimura's neutral theory of molecular evolution.)
However, at the species level, these traditional objections to high frequency
for drift become invalid, and we should anticipate a major role for this second
cause of sorting. Low population size (number of species in a clade) provides the
enabling criterion for important drift in both categories at the species level. The
analog of genetic drift—which I shall call "species drift"— must act both
frequently and powerfully in macroevolution. Most clades do not contain large
numbers of species. Therefore, trends may often originate for effectively random
reasons. Consider a trend produced by random deaths (a comparable argument can
easily be made for random birth differentials), based on Raup's "field of bullets"
model (1991 and Chapter 12). Suppose, for example, that each of the ten species of
a clade lives in a small area, with each species allopatric to all others. Over a
certain period of time, a bolide (or some gentler environmental change with power
to drive a local species to extinction) strikes half the areas at random and
eliminates the resident species of the clade while each of the species in the five
safe areas branches off a daughter, thus restoring the cladal population of 10
species. At an N this low, some trends (and perhaps a substantial number) will
inevitably arise by this mode of random removal. Perhaps, for example, four of the
five species with mean body size below the cladal average will happen to die. A
substantial random trend to increased body size then occurs within the clade.
When we move from the homogenous "field of bullets" model to a scaling of
effects in the real world, and consider the consequences of infrequent, but severe,
mass extinction on a global scale, the potential role of random trends by
elimination only increases—for random effects based on small numbers will be
greatly intensified. (The reduction of species number in mass extinction may be
conventionally causal, but the final death of the clade, after reduction to less than a
handful of species, may then be effectively random. For example, so few trilobite
species still lived when the great Permian extinction occurred that I'm not sure we
need to seek a "trilobite specific" cause for the final elimination of this previously
dominant group.)