UC Davis

Structures of evolving populations are traditionally derived from traits of its members. An alternative approach uses network metrics to define groups that evolve jointly. This supposes that selection acts not only on who members are (i.e., traits) but also on to whom they are connected (i.e., interdependent relationships). This paper presents a method to meaningfully quantify differences in evolutionary forces over multiple levels of population taxonomies and tests almost 1000 multilevel partitions of 8 empirical networked populations evolving over time. It shows that multilevel network metrics as selection criteria identifies stronger evolutionary natural selection than trait based population taxonomies.