can jeopardize local populations. The effects of inter-
actions and disturbance can also be indirect, For ex-
ample, an introduced species can reduce the availabil-
ity of food or nest sites for resident species, making it
more difficult for them to reproduce or survive envi-
ronmental stresses.
Habitat fragmentation (the isolation of habitats into
islandlike "fragments," remnants, or patches sepa-
rated by physical barriers to dispersal) plays a major
role in rarity and extinction (Saunders et al. 1991). If
the habitat patch in which a population occurs is sepa-
rated from other populations by inhospitable habitat
(e.g., large agricultural areas for forest-dwelling spe-
cies, or lowland areas for highland endemics), the
exchange of individuals and genes among populations
will be reduced. Even if isolation acts as a filter rather
than a strict barrier, allowing only certain species or
genotypes to disperse, the composition of isolated
communities and the structure of individual popula-
tions (e.g., age distribution, sex ratio, genetic diver-
sity) may be altered. Moreover, isolation reduces the
probability that a patch in which a population has
gone locally extinct will be recolonized. Edge effects
can also arise from habitat fragmentation (Angelstam
1992, Murcia 1995). For example, forests that are
penetrated by roads, reduced in size by logging, or
surrounded by fields experience an altered micro-
climate along their edges; conditions may be hotter,
more desiccating, and windier at the edge and as
much as 100 m into the forest. Brood parasites and
predators, many of which prefer open habitats, are
more likely to penetrate core areas within forests by
entering along newly created edges, water lines, or
roads. Although habitat fragmentation and edge ef-
fects are usually thought of on the scale of forest
patches, they also occur at the scale of vast geographi-
cal regions (Edwards et al. 1994). Changes in the land-
scape can disrupt gene flow among populations, pre-
vent recolonization following local extinctions, alter
regional climates, and facilitate the spread of preda-
tors and parasites (Dunning et al. 1992).
As populations decline, the probability of their
going extinct increases for several reasons. Individu-
als in small populations are more likely to be related
than individuals in large populations, which means
the risk of inbreeding rises (Soule 1987). Inbreeding
leads to greater homozygosity (the state of having two
copies of the same allele, or variant of a particular
gene) and, consequently, the increased expression of
deleterious recessive alleles (harmful traits whose
effects are not ordinarily apparent except when ho-
mozygous). The results of excessive inbreeding in-
clude reduced fecundity, developmental defects, low-
ered life expectancy, and loss of genetic variability
within individuals and populations (Pusey and Wolf
1996). Over the long term, populations that are geneti-
cally homogeneous may be more susceptible to dis-
ease or physical stresses, and they may not be able to
evolve adaptations that could enable them to survive
in a dynamic or hostile environment. We do not know
how large populations must be before they are safe
over the long term from the effects of inbreeding or
the random loss of genes through genetic drift in small
populations. Effective population sizes (Ne) might
have to be as large as 500 for species prone to inbreed-
ing depression, or far less for species that have passed
through population "bottlenecks" and eliminated
most deleterious alleles (Lande and Barrowclough
1987).
In small populations, the risk of random demo-
graphic events such as skewed population sex ratios
or unbalanced age distributions quickly becomes a
major factor in population declines. Small localized
populations no longer have a margin of safety in the
face of random population fluctuations and environ-
mental disturbances. Below a certain population size,
such "demographic stochasticity" becomes increas-
ingly perilous. Rarity begets rarity until small popu-
lations are drawn into "extinction vortices" in a posi-
tive feedback cycle (Soule and Wilcox 1980). For
example, small populations can by chance become
dominated by older age classes or pre-reproductive
individuals, by a single sex, or by relatives, which
makes it more likely that the populations will con-
tinue to shrink (Soule and Wilcox 1980, Nunney and
Campbell 1993).
The conservation of biodiversity must be considered
not only at the scale of populations but also in the con-
text of landscapes (Dunning et al. 1992). The assem-
blage of plants and animals can be viewed as the net
outcome of a dynamic process of colonizations, local
extinctions, and recolonizations (MacArthur and Wil-
son 1967). Under natural conditions, habitats are het-
erogeneous or patchy at every scale, broken up by
landscape features (e.g., rivers, valleys) and distur-
bances (e.g., landslides, fires, treefalls). A habitat such
as a swamp forest may be effectively disconnected
from other habitats of the same type, and become an
ecological "island" in a matrix of an unlike habitat.
To the degree that the surrounding habitat matrix is
inhospitable or that patches are isolated from one an-
other, the matrix acts as a barrier. The forest fragmen-
tation that results from clearing for agriculture is the
most familiar example of the creation of ecological
islands by humans. The model of "island biogeogra-
phy" and the concept of habitat heterogeneity can be
applied at scales from individual plants (or parts of
plants) to ecosystems and broad geographical regions
(MacArthur and Wilson 1967, Williamson 1989). As
deforestation moves up Costa Rica's mountain slopes,422 Conservation Biology