Whether an ecological network is nested, compartmentalized, or asymmetric is clearly influenced by the developing coevolutionary interactions between species as the network grows. Through feedback, however, network structure determines the interactions that can and do occur as well as the strength of those interactions. Constraints on the interactions that can occur develop as a property of the ecology and phylogenetic history of the current species and connections in the network (Jordano et al. 2003). These so called “forbidden links” result primarily from phylogenetic constraints on characters preventing trait convergence and thus assimilation into the network.
The structure of mutualistic networks reduces the ability of species to become involved in tight reciprocal interactions. A diffuse selection regime is at least in part responsible for the lack of tight interactions. Regardless of whether or not two species are more efficient mutualists with one another, having a diverse community of potential partners through complementary and convergent traits will result in selection that is too diffuse to allow for tight coevolution (Hoeksema and Bruna 2000). Tight reciprocal selection between two species is rarely found in mutualistic networks (Joppa et al. 2009). Most mutualistic networks either display a pattern that is opposite of reciprocal specialization, or at least no different than chance (Joppa et al. 2009), possibly due to a lack of compartmentalization (Bascompte et al. 2003).
Mutualistic networks such as those involving seed-dispersing frugivores and plants, however, were found by Jordano (1987) to compartmentalize around specific major food sources. These major food sources provide a core set of species in the network upon which further interactions can develop. Multiple core sets of species are likely to occur in these networks due to increasing species richness and phylogenetic constraints. As species richness increases the particular set of species with which any other species is interacting will become a smaller proportion of the total number of species. Thus compartments should develop around several core sets of generalist species based on the traits that are most similar to their own through convergence. Furthermore, phylogenetic constraints may limit which set of species can interact with each other (Jordano’s forbidden links). The influence of phylogeny may also lead to more related species interacting with similar species. For example, a species with a small gape width will be less likely to interact strongly with large-fruit bearing plants.
Relatively recent research has lead to the development of a new understanding of how evolutionary change can occur and be influenced by species interaction. Evolutionary responses can occur through the adaptation of one species to one other species or by adaptation to environmental change. An adaptation may produce either reciprocal adaptation (coevolution), or it could invoke a response in another interacting species or multiple other interacting species. Thus within a network evolutionary responses may take the form of one-on-one adaption, reciprocal evolution, cascading adaptation which can be either pair-wise or diffuse, and cascades that include reciprocal evolution (Guimarães Jr. et al. 2011). Cascades can be especially influential in the degree of evolutionary response within the network.
Not all species are equally important in the evolutionary dynamics within the network (Guimarães Jr. et al. 2011). Because mutualistic networks are built upon a core of generalist species whose interactions encompass most of the network they can be especially influential in determining the course of evolution for the system (Bascompte et al. 2003). Thompson has called these influential species “keystone mutualists” while Guimarães Jr. et al. (2011) uses the term “supergeneralist” to denote these well-connected generalists. This core is responsible for the maintenance of convergence through complementary traits in diverse assemblages (Guimarães Jr. et al. 2011). Furthermore, interwoven effects of both coevolutionary and evolutionary change cascade through the network and lead to faster trait evolution in models of plant-animal mutualistic networks (Guimarães Jr. et al. 2011).
In trophic networks that are compartmentalized, such as those studied by Krause et al. (2003), interaction strength is weaker among compartments than within. Therefore, reciprocal adaptation should be more likely to occur within a compartment than between. The results found by Joppa et al. (2009) indicate that it is possible for compartments or modules to arise from somewhat tighter reciprocal specialization of taxonomically related individuals on similar resources. Their results were not significant but suggest that more research should be done in this area.