whether faunal region entered their models. Starting
with the individual variables, they find constraining
(triangular) relationships between richness as the
dependent variable and area, temperature, precipita-
tion, and (inversely) distance from the nearest conti-
nent, indicating that these variables may each set
upper limits to richness, but that richness frequently
falls below these limits because of the effects of the
other limiting variables. Using correlation and gen-
eral linear modelling, they show first, that the inter-
action terms between area and precipitation, and
between area and temperature, are crucial to push-
ing the explained variance above 50%, and second,
that the slope of the ISAR increases with rise in mean
annual temperature, and with precipitation.
Although their simpler models are easiest to inter-
pret and readily reach R^2 values in the range of
0.64–0.70, their final multiple regression model
based on quite complex manipulations of area, pre-
cipitation, isolation, temperature, and elevation
attained a cumulative R^2 of 0.877. Halting the model
after the inclusion, in this order, of ‘the area–precipi-
tation interaction’, ‘distance to continent’, and ‘tem-
perature’, provides a model explain 82% of the
variation, which given that the data encompass
global variation in island environments, and a large
range of island areas, is a remarkably good fit.
Faunal region had only a minor signal in the data,
accounted for by lower than expected diversity in
the Sub-Antarctic/Antarctic.
One of the most intriguing findings of their
analyses relates to the role of isolation. Although
they find that isolation has similar explanatory
power to climate and area, the interactions of the
explanatory variables tell a different story.
According to MacArthur and Wilson (1967), the
slope (zvalues) of ISARs should vary in relation to
the difficulty of species dispersing to islands, i.e.
more isolated sets of islands should produce
steeper ISAR slopes. By contrast, in these global
analyses, Kalmar and Currie (2006) report strong
area–climate interactions, which are consistent with
Wright’s species-energy ideas, but they do not find
area–isolation interactions. This is particularly
intriguing given that their isolation metrics alone
are capable of explaining just under 50% of the vari-
ation in their data set, and together with area 54%.
TURNOVER 99These seemingly contradictory findings may indi-
cate that by selecting a globally representative set of
islands, they have assembled a data set in which the
variation in ISAR slope due to climate has largely
swamped the intraregion influence expected as a
function of island isolation. So, it is not that isola-
tion has no correlation with richness of island birds,
but it is hard to disentangle the independent effect
of isolation from variation in area (with which dis-
tance is negatively correlated), topography, and
climate and the further complications of stepping-
stone effects that effectively alter isolation from
continents. As Kalmar and Currie are quick to point
out, these findings sit uneasily with the prominent
role asserted for isolation in MacArthur and
Wilson’s dynamic island model.
David Lack (1969, 1976) argued that given the abil-
ity of birds to colonize remote islands comparatively
early in their existence, and then to speciate on them,
the apparent impoverishment of remote islands
might be less a reflection of immigration–extinction
dynamics than of a relative paucity of resources
within their food chains. Kalmar and Currie’s analy-
ses seem to lend some support to this argument.
The issue remains, to what extent do the patterns
discussed above reflect dynamic equilibria of the
form proposed in equilibrium theory? To address
this, it is essential, as Gilbert (1980) advocated, to
give close consideration to species turnover.4.4 Turnover
According to equilibrium thinking, it is a reason-
able assumption that most islands are at, or close
to, their equilibrium species number most of the
time. The EMIB postulates that this should be a
dynamic equilibrium, approached by means of
monotonic alterations in rates of immigration and
extinction. Each of these postulates requires
examination. Before doing so, it is important to
consider the nature of the evidence involved.
Whether or not these postulates are well founded,
there is another important issue that must be con-
sidered, which is: when turnover does occur at
measurable rates on ecological timescales, is it
homogeneous (across all species) or heteroge-
neous (involving particular subsets of species)?