We present and illustrate methods of analyzing food web structure base
d on generalized linear models. Applied to the published food webs of
Schoenly et al., these methods reveal strong associations of trophic f
ractions and connectance with the number of species in the web. Furthe
rmore, the nature of these relationships depends on the habitat type o
f the web. Measures of goodness-of-fit of the statistical models revea
l pronounced overdispersion, i.e., web-to-web variation in trophic fra
ctions and connectance that greatly exceeds the variation attributable
to the number of species and habitat type of the web. These results s
uggest it may be difficult to identify laws governing food web propert
ies over a wide range of organisms and habitats.