narrow or broad groups of species. It might thus be
possible for chequerboard distributions and the like
to be detectable within a particular guild of birds,
while the avifauna analysed as a whole exhibits a
tendency towards nestedness.
Tendencies towards nestedness were recognized
by early island biogeographers, such as Darlington
(1957), but it was not until the 1980s that statistical
tests were developed for formal detection of the
pattern. When the first indices were developed by
Patterson and Atmar (1986), they found that insular
systems are commonly nested (Patterson 1990;
Blake 1991; Simberloff and Martin 1991). Patterson
and Atmar (1986) took the view that nestedness
was most likely to occur in extinction-dominated
systems and developed their first metric on this
presumption. However, other authors argued that
immigration might also produce nestedness pat-
terns, requiring different metrics for quantifying
nestedness. This led to much debate about how best
to measure nestedness, a matter that is most likely
still unresolved (e.g. Wright and Reeves 1992; Cook
and Quinn 1995; Lomolino 1996; Brualdi and
Sanderson, 1999; Wright et al. 1998; Rodríguez-
Gironés and Santamaría 2006). Although the choice
of metric is important at a detailed level, the
general proposition that nestedness is a common
feature of ecological assemblages (both insular and
non-insular) is robust to the choice of metric
(Wright et al. 1998).
The nestedness concept, as applied by Patterson
and Atmar and as generally followed by others, is
based on ordering the data matrix by the size of
fauna or flora, i.e. it is richness-ordered nestedness.
Some authors, however, have used the same term
‘nestedness’ when ordering the data matrix not by
species richness but by island area, which we might
termarea-ordered nestedness, or even by island
isolation, i.e. distance-ordered nestedness(e.g. see
Roughgarden 1989; Lomolino and Davis 1997). The
extent to which island assemblages are found to be
nested when ordered in these different ways may
be used to infer processes responsible for structur-
ing them. However, whichever metric is used, and
whichever variable is used for ordering the data
matrix, inferences of causation ideally require
independent lines of verification.
As Wright et al. (1998, p. 16) state ‘nestedness is
fundamentallyordered composition... Any factor
that favours the ‘assembly’, or disassembly...of
species communities from a common poolin a consis-
tent orderwill produce nested structure’ [italics in
original]. They identify four factors or filters that
may be important contributors in generating nested
structure: passive sampling, habitat nestedness,
distance, and area.
●So called passive samplingof a pool of species
by a recipient island may produce nested structure
if species are drawn randomly from the pool with
the constraint that the availability of propagules is
itself strongly non-random, assuming for instance a
log-normal species abundance distribution for the
pool community. Simulation models on this basis
generate highly nested model communities, gener-
ally in fact more strongly nested than observed.
Where passive sampling simulations do compare
well with observed data sets, this may lead to the
inference that differential immigration has struc-
tured the assemblages. However, Wright et al.
(1998) argue that the simulation is not dependent
on a particular mechanism of community assembly,
and that other scenarios are conceivable. One might
be that abundant species from within the pool are
regionally ‘successful’ species with a low probabil-
ity of going extinct from isolates.
●Forhabitat nestednessto produce biotic nested-
ness requires, first, a strong fidelity of species to
particular micro-habitats, and second, for islands to
acquire additional habitats in a fairly fixed order.
Given which, we would anticipate that it will be
more readily detectable in plant data sets (as plants
often show high habitat affinity), as exemplified by
Honnayet al.’s (1999) study of plants in forest
fragments in Belgium.
●If nestedness is clearly related to distancefrom
mainland sources, this suggests that there are
predictable limits to species’ dispersal abilities,
such that the system is made up by islands
‘sampling’ a series of species’ isolation-incidence
functions (as Fig. 5.4). In many case studies, where
available data are limited to a single time-point, it
may be safest simply to label this pattern a
‘distance’ effect (Wright et al. 1998), but there areNESTEDNESS 127