in contrast, were absent from many islands in the
2–10 ha range. Temporary food shortages, or
‘energy crises’ might be more problematic for the
smaller species than for the larger S. araneus, thus
potentially explaining their lower frequency of
occurrence as a function of environmental stochas-
ticity. Arguably, interspecific competition might
also have a role to play, but Peltonen and Hanski
(1991) subsequently reported no evidence for a sig-
nificant effect of competition on extinction rates.
Both studies conclude that the body size of species
has a bearing on their incidence. As the larger
species do not appear to be affected by interactions
with the smaller species within the guild, they pre-
dict that they should occur consistently on islands
greater than a minimum size set by demographic
stochasticity. Small species should have a more
erratic distribution, because of their increased sus-
ceptibility to environmental stochasticity. This
interpretation combines elements of the incidence
function and small-island effect hypotheses intro-
duced in Chapter 4. Indeed, Hanski (1992) shows
how, given knowledge of minimum population
sizes, a simple model of incidence functions can
generate realistic estimates of colonization and
extinction rates for this system.
Incidence functions as originally devised are rel-
atively crude univariate tools. Adler and Wilson
(1985) provided a more sophisticated multivariate
approach, using multiple logistic regression to esti-
mate the probabilities of individual species occur-
rences. Their system consisted of records of 9
species of small terrestrial mammals for 33 coastal
islands off Massachusetts, and their 11 explanatory
variables provided various measures of island area,
length, isolation, habitat, and source pools.
Principal components analysis was used to gener-
ate two composite environmental variables, termed
‘size’ and ‘isolation’. These composite variables
were then used together with a categorical variable
representing the dominant habitat type of each
island (‘domhab’), in multiple logistic regressions.
For all but one species, statistically significant func-
tions were obtained. In four cases (species of
Scalopus, Peromyscus, Microtus, and Zapus), occur-
rence on islands was positively related to increas-
ing island size, and in four cases (Sorex,Blarina,
Tamias, and Clethrionomys), to decreasing island iso-
lation. The habitat variable, domhab, was also
included in the functions of two species
(Peromyscus and Microtus). Their study thus
showed another means of quantifying relationships
between species occurrence and a suite of environ-
mental variables: their probability functions allow-
ing the relative significance of different variables to
be tested within a common framework. One other
feature of interest was that they also used live-trap-
ping over a 2–4 year period to show that the most
widely distributed species across the islands are
also those which reach the highest population den-
sities within them.
The importance of both isolation and of con-
stancy of resource availability were shown by Adler
and Seamon’s (1991) study of the population fluc-
tuations in just a single species, the spiny rat
Proechimys semispinosus, on islands in Gatun Lake,
Panama. No other species of mouse- or rat-like
rodents were captured on the islands. Their data
revealed that colonization and extinction of the rat
occurs regularly to and from the islands. Larger
islands with year-round fruit production, regard-
less of isolation, have persistent populations of rats.
Small, isolated islands only rarely have rats,
because they lack a year-round food source and
because immigration is rare. Small, near islands fre-
quently have rats when fruits are present. These
observations thus reveal how in some circum-
stances (of size, isolation, etc.) fluctuations in island
carrying capacities influence species composition
and turnover.
A particular land or aquatic system may provide
dispersal filters which differentiate between differ-
ent members of a source biota, and indeed different
species within a single taxon (Watson et al. 2005).
For instance, as shown by Lomolino (1986), partic-
ular functional groups of small mammals may be
differentially affected by the same immigration fil-
ters. In regions with season ice cover, small coastal,
inshore, or riverine islands may have their popula-
tions supplemented by mammals moving across
the snow-covered ice in winter. In some cases, such
movements have been recorded by tracking stud-
ies. In his studies of mammal movements across the
snow-covered St Lawrence River, Lomolino foundISLAND ASSEMBLY THEORY 119