In this paper we argue that landscape spatial structure is of central
importance in understanding the effects of fragmentation on population
survival. Landscape spatial structure is the spatial relationships am
ong habitat patches and the matrix in which they are embedded Many gen
eral models of subdivided populations make the assumptions that (I) al
l habitat patches are equivalent in size and quality and (2) all local
populations (in the patches) are equally accessible by dispersers. Mo
dels that gloss over spatial details of landscape structure can be use
ful for theoretical developments but will almost always be misleading
when applied to real-world conservation problems. We show that local e
xtinctions of fragmented populations are common. From this it follows
that recolonization of local extinctions is critical for regional surv
ival of fragmented populations. The probability of recolonization depe
nds on (1) spatial relationships among landscape elements used by the
population, including habitat patches for breeding and elements of the
inter-patch matrix through which dispersers move (2) dispersal charac
teristics of the organism of interest, and (3) temporal changes in the
landscape structure For endangered species, which are typically restr
icted in their dispersal range and in the kinds of habitat through whi
ch they can disperse, these factors are of primary importance and must
be explicitly considered in management decisions.