We investigated web relocation and habitat selection in Latrodectus re
vivensis (Theridiidae), a widow spider that constructs webs on scatter
ed shrubs in the Negev desert. We used repeated observations of indivi
dual spiders' habitat preferences, movements to new web sites, growth,
fecundity, and survival under natural and manipulated habitat conditi
ons to examine relationships between movement, habitat, and components
of fitness. Our main objectives were to determine the spatial and tem
poral patterns of movement, and the causes, costs, and fitness consequ
ences of shifting habitat. Repeated censuses of individually marked fe
males showed that web relocation in L. revivensis is associated with s
pider growth. Spiders tended to move to larger shrubs after one or two
molts, and the size of web components and web-site features scaled to
body size. Moves occurred mainly in Spring (March-May), and most move
s (74%) were by juveniles. The main cost of web relocation was increas
ed mortality: there was a 40% chance of death during a move. A compari
son of occupied shrubs with those available in the habitat indicated s
trong site selection involving seasonally varying and age-dependent pr
eferences for particular shrub species, and a general preference for t
aller shrubs. Prey availability at shrubs (as measured by pitfall trap
s, and by prey remains in nests) varied spatially and was affected by
shrub species and size. However these effects were small compared to h
abitat-wide temporal variation in prey availability, suggesting that w
eb-site relocations would not result in substantially higher prey capt
ure. Nonetheless, movement to larger shrubs resulted in improved spide
r body condition and, ultimately, greater reproductive success. The ad
vantages of web-site relocation in this species may relate to architec
tural features of shrubs that provide support for the web, and minimiz
e exposure to thermal extremes and to predators. However, experimental
manipulation of exposure by trimming shrubs did not decrease spider f
itness. We suggest that web-site selection in this species is less pre
cise than expected because of the risk of dying during a move.