Species in community food webs are commonly aggregated or lumped on th
e basis of overlap in predators and prey. This note reports an unexpec
ted result of lumping species in observed food webs and simulated food
webs. The main result is that it is much easier to lump species in ob
served webs than in simulated webs. The reason is that there is much m
ore overlap in predators and prey for species in observed webs than in
simulated webs. This points to a fundamental deficiency in web assemb
ly models that assume a random distribution of links.