Interactive networks are binary or quantitative representations of interspecific interactions in ecological communities. Complex networks and other theoretical approaches offer metrics and descriptors to investigate the organization of various kinds of interactive networks. Large sets of mutualistic networks, notably of plant/pollinator and plant/disperser webs, have been analyzed for general patterns and features. By contrast, antagonistic bipartite networks (sometimes called ‘trophic’ webs, such as hosts/parasitoids), are only beginning to be explored. Here we present a cross-sectional analysis of a large set of plant-insect herbivore community webs obtained in various latitudes, continents and ecosystems. We examine descriptors of host specialization of insects and of its converse, herbivore richness on plants, which are key organizing features of trophic webs, in two ways: first, the distribution of numbers of associated hosts or herbivores (‘degree distribution’), fitted to exponential, power-law and truncated power law models; second, the asymmetry in numbers of links of each species and of its direct associates. On theoretical grounds we expect few generalists (closer to an exponential degree distribution) and a noticeable proportion of reciprocal specialists, which should reduce average link asymmetry in herbivory networks.
Herbivory networks ranging from 22 to over 1,600 species show very different specialization patterns. Degree distributions of plants and herbivores are often distinct. In most webs, herbivore richness on plants is best fit by exponential or truncated-power models, whereas host richness of herbivores fits either power or truncated-power models. Linkage asymmetry is high when specialists have mostly generalist associates; as expected in networks with pronounced nestedness. In many herbivory webs, species’ degrees are weakly correlated with the average linkage of their associates. However, the actual degree distribution of species associated with specialists shows other trends. Specialized plants, those with low herbivore richness, often are consumed by insects with either very few or many hosts. In turn, host plants of specialized insects often have highly variable numbers of herbivores, but, even so, clusters of specialist-to-specialist links are often recognizable, indicating that herbivory networks are at least partly modularized or compartmented. These very different patterns suggest that herbivory networks are heterogeneous not only among ecosystems, but also according to feeding modes, taxonomic span of herbivores and plants, and methodology. No general model or pattern is likely to fit or describe all of these.