Mutualistic networks provide an excellent example of how coevolutionary forces shape the structure of the network (Thompson 2005, Bascompte 2009). Two major properties of mutualistic networks that are believed to develop through coevolution are asymmetry and nestedness (Bascompte et al. 2003, Bascompte et al. 2006, Thompson 2006). Bascompte et al. (2006) described asymmetry in mutualistic networks as a skew in the dependencies of interacting species relative to one another. In other words interacting pairs of species will tend to involve a specialist interacting with a generalist. When a network is described as nested in an ecological context, arranging species from most generalist to specialist should show that the specialist is using a subset of the resources used by those that are more generalist.
The alternative to nestedness in a network is compartmentalization, whereby there are small groups of strongly interacting species connected by weak interactions (Krause et al. 2003). In ecological networks a general pattern has emerged in that networks of antagonistic interactions (e.g. trophic or competition) tend to exhibit compartmentalization (Krause et al. 2003, Stouffer and Bascompte 2011), while networks of mutualistic interactions tend to be nested (Bascompte et al. 2003). Furthermore it is thought that nestedness may result from increasing asymmetry within the network.
Two coevolutionary processes are thought to be responsible for the patterns of asymmetry and therefore nestedness in mutualistic networks, coevolutionary complementarity and convergence (Bascompte 2006). Coevolutionary complementarity refers to the similarity of species in traits involved in the interaction (e.g. fruits or flowers in plants) while convergence is the idea that new species enter and persist in the network by developing these traits. Complementarity of traits generates the foundation of the network. From a fitness point of view, complementarity should be expected when there are fitness gains to increasing individual “attractiveness” for flowering plants or increasing the base number of species from which an animal can eat fruit or nectar.
In mutualistic networks most generalist species interact with one another. These interactions form a core structure in the network and generate strong selective pressure to develop complementary traits through diffuse pair-wise interactions. New species can be added to the network of mutualistic interactions by developing traits that converge upon those of the core of generalists (Bascompte et al. 2003, Bascompte et al. 2006). Mutualistic interactions often favor the incorporation of new species and as the network grows the core set of traits become increasingly more important to the network (Thompson 2005, Thompson 2006). Addition of new species through complementarity and convergence results in the “coevolutionary vortex” creating increased asymmetry in interactions. Newer species will be more specialized relative to the core of generalists (Thompson 2006, Bascompte et al. 2006).
In contrast to mutualistic networks, trophic networks are most likely shaped by different forces. While complementarity and convergence cause a pattern of asymmetry and nestedness in mutualistic networks, coevolutionary alternation may be responsible for developing compartmentalized trophic networks (Thompson 2005). Coevolutionary alternation occurs through prey switching by predators in response to development of stronger predator defenses (Thompson 2005, Thompson 2006). When the predator switches to alternative prey the original prey species is now burdened with a costly defense and therefore individuals with weaker defenses should have higher fitness. This continual cycle allows for tighter coevolution between the predator and their prey options generating a more compartmentalized network structure. Compartmentalization is often found in empirical food webs (Krause et al. 2003, Stouffer and Bascompte 2011), while it is generally found only in more species rich mutualistic networks (Jordano 1987).
Asymmetrical interactions and nestedness may also be found in trophic networks although not as often as in mutualistic networks (Bascompte et al. 2003). Arditi and Ginzburg (2012) demonstrate that asymmetrical interactions can result from donor-controlled predator-prey dynamics. Interference among predators in this case limits the mortality induced by the predators on the prey population (Arditi and Ginzburg 1989). Mathematically this leads to a Jacobian matrix (matrix of interactions between all species) with zeros on one side of the diagonal at equilibrium. Such a triangular matrix is similar to the nested architecture arising from asymmetrical dependencies presented by Bascompte and colleagues (2003). Since asymmetrical interactions are expected to lead to a nested architecture (Bascompte et al. 2003) and donor-control predator-prey interactions generate asymmetries then nestedness should be expected in trophic networks to the extent that donor-control occurs in the web.