Connection and Coordination: The Interplay Between Architecture and Dynamics in Evolved Model Pattern Generators Psujek Ames Beer 2006

This paper explores all possible topologies on networks of up to four nodes, with a representative sample of 5 node networks. it is a great illustration of how topology does not necessarily restrict dynamics, there is still plenty of wiggle in the dynamics for all kinds of different kinds of behavior to be selected for in a given topology.

The work eventually functions to show the necessary and sufficient conditions for attaining each of the three plateaus of the fitness landscape: single steps, coordinated steps, well coordinated steps.

it also investigated how the different architectures influence 'evolvability'

I remember being disatisfied with this paper, though i've forgotten why. I think I wanted it to be something besides what it was.