Abstract
Espinosa-Soto and Wagner (2010) introduced a domain with weak assumptions on biology and environment, where modular structures emerge under simple evolutionary processes.
We found a number of anomalous behaviours: modularity emerged in this domain, but could not dominate populations as observed in biology. Highly fit, modular solutions exist in the search space, can be readily found by a simple deterministic procedure (and presumably could dominate populations if found), but evolutionary search never found them, despite mutation biases that appear to favour those solutions. Moreover, emergence of modularity was promoted by stochastic dynamicity in the fitness function: a stochastic but fixed fitness function generated much less modular solutions.