isolated, the land surrounding others was
abandoned during the study and became infilled
by forest regrowth dominated by Cecropia.
Stouffer and Bierregaard (1996) examined the use
of these forest fragments by understorey humming-
birds. The three species found before fragmentation
persisted at similar or higher numbers in the frag-
ments through the 9 years of the study. Use of the
fragments did not differ between 1 and 10 ha frag-
ments. Furthermore, it mattered little whether the
surrounding matrix included cattle pasture, aban-
doned pasture, or Cecropia-dominated second
growth. This particular guild of birds thus appears
to be able to persist in a matrix of fragments, sec-
ondary growth, and large forest patches.
By contrast, the guild of insectivores, which
dominate the understorey bird community, was
much more responsive to fragmentation: both abun-
dance and richness declined dramatically (Stouffer
and Bierregaard 1995). Three species of obligate
army-ant followers disappeared within 2 years.
Mixed-species flocks drawn from 13 species disag-
gregated over a similar timescale, although three of
the species persisted in the fragments. The 10 ha
fragments were less affected than the 1 ha frag-
ments, and in time the species flocks reassembled in
those 10 ha fragments that were surrounded by
Cecropiaregrowth. The communities in these recon-
necting fragments converged compositionally on
pre-isolation communities, while communities in
the smaller or more isolated fragments continued to
diverge. Particular terrestrial insectivores (e.g.
Sclerurusleafscrapers and various antbirds) failed
to return to any of the fragments during the study.
To sum up: (1) the impact of fragmentation on birds
is strongly ecologically structured, (2) different
‘guilds’ respond in divergent fashion, and (3) the
types of species most vulnerable to fragmentation
are fairly predictable, with insectivores that partici-
pate in mixed-species flocks being the most vulner-
able guild (e.g. Lovejoy et al. 1986; Stouffer and
Bierregaard 1995, 1996; Christiansen and Pitter 1997;
Ribonet al. 2003).
We may thus conclude that habitat fragmentation
is associated with compositional change and
species losses. Theory predicts significant losses
where fragmentation is extreme, but over an uncer-
tain time frame. However, it has yet to be satisfac-
torily demonstrated that the stochastic ‘island’
relaxation effect is generally a strong one, and it
seems likely that there will be a degree of scale
dependency in relation both to area and isolation:
factors that warrant further theoretical and empiri-
cal attention. Recent work suggests that there may
be taxon dependent thresholds of habitat loss,
above which few species losses occur, but beyond
which losses accelerate (Ewers and Didham 2005).
Given the significance of the assumptions of relax-
ation to the debate on species extinction rates, it is
surprising that more attention has not been given to
demonstrations of these effects (see Simberloff and
Levin 1985; Simberloff and Martin 1991; Whitmore
and Sayer 1992; Brooks et al. 1999; Laurance 2002).10.8 Succession in fragmented landscapes
The processes of land-use change which create frag-
mented systems typically initiate successional
changes in the habitat island remnants. This was
apparent in the Barro Colorado island study.
Another example is provided by Weaver and
Kellman’s (1981) study of newly created woodlots
in southern Ontario, Canada. The species losses that
occurred did not take the form of stochastic
turnover, with different species ‘winning’ in differ-
ent woodlots. Instead, the ‘relaxation’ involved the
successional loss of a particular subset of species.
With appropriate management, the observed losses
could be avoided. If occurring at all in the absence
of disturbance and succession, turnover appeared to
be very slow-paced among the vascular flora. In this
system, the EMIB effects were clearly weak, and
subordinate to other ‘normal’ ecological processes.
In practice, nature-reserve management often
pays considerable attention to ecological succession,
particularly in small reserves. Failure to do so often
leads to the loss of desirable habitats. This applies to
many of the lowland heath reserves of southern
England, which have long been anthropogenically
maintained. Without appropriate management,
most areas suffer woodland encroachment. The
concepts of minimum dynamic areasandpatch
dynamics(Pickett and Thompson 1978) are thus274 ISLAND THEORY AND CONSERVATION