smaller islets), with predominantly desert-like con-
ditions. Here they have radiated into 13 different
species (Sato et al. 1999; but see Zink 2002). The
family is found also on Cocos Island, a lone,
forested island of about 47 km^2 , which is, just, the
nearest land to the Galápagos. Here there has been
no possibility of the island hopping that appears to
characterize archipelago speciation, and the group
is represented by a single endemic species
Pinaroloxias inornata, one of just four species of land
bird (Williamson 1981; Werner and Sherry 1987).
Analyses of mtDNA data indicate that the Cocos
finch is probably a descendent of a Galápagos tree
finch (Sato et al. 1999).
Yet, as Sauer (1990) cautions, within an archipel-
ago, in addition to single species being present on
different islands, quite commonly several sibling
species occur on a single island, where they are
mainly segregated by habitat. In many instances,
isolation may have been only indirectly relevant as,
by screening out competitors, isolation offers
opportunities for adaptive radiation to those
species that do arrive. Some, at least, of this radia-
tion may take the form of the next class of mecha-
nism, competitive speciation.
Competitive speciation
This is the term that Rosenzweig (1978, 1995) gives
to cover a variety of related modes of sympatric
speciation. The idea here is that a species expands
its niche to occupy an unexploited ecological
opportunity, followed by that species’ sympatric
break-up into two daughters, one in essence occu-
pying the original niche, and the other the newly
exploited one. The expansion happens because of a
lack of competition with other species, a feature
common to remote island faunas and floras. But
the break-up into separate species happens
because of increased competitive pressure between
those portions of the population best able to
exploit the two different niches. The operation of
these processes must take place over the course
of many generations, and observation of the com-
plete process in the field is thus not an option.
Rosenzweig illustrates the idea by means of a
‘thought experiment’, the essence of which, placed
in the island context, can be rendered as follows.
A colonizing species of bird will have a particular
feeding niche, determined morphologically and
behaviourally, and perhaps most clearly expressed
by features such as bill size and shape (Lack 1947a;
Grant and Grant 1989). According to theory, such
characters would be expected to have a unimodal
distribution of values. The environment of the
island contains empty niches, unexploited ecologi-
cal opportunities, which for the purpose of our
experiment are taken to be distributed bimodally,
i.e. there are two resource peaks. The species,
unconstrained by competition with the full range of
mainland forms, expands from its original modal
position somewhere along this resource space so
that it occupies each of these niches. Individuals
with genotypes that match best to one or the other
of these niches become numerically dominant
within the population, which has by this stage
grown close to its carrying capacity. At this point,
disruptive selection kicks in, such that the mid-
range phenotypes lose fitness relative to those that
are better suited to one or the other of the main
feeding niches. Those individuals adapted prima-
rily to a peak in the resource curve may also exploit
the valleys either side of that peak. If they do so at
all successfully, no opportunity remains for valley
phenotypes in between the two peaks. They will be
few in number in the population, and their off-
spring will have fewer resources to tap. Although
they may also reach up to the peaks, they will not
be effective competitors in those portions of the
resource continuum. Valley genes will thus have lit-
tle success and will be bred out, as in time members
of each ‘peak’ population that can recognize others
of their type (and breed with them alone) will be at
a selective advantage. Therefore, isolating behav-
ioural mechanisms develop and at this point the
lineage can be viewed as having split.
If neither resource peak matches closely the orig-
inal niche of the species, the divergence may result
in two novel forms and the effective disappearance
of the original species (termed anagenesis: see
below). There may also be more than two resource
peaks, thus accommodating a larger radiation.204 SPECIATION AND THE ISLAND CONDITION