The previous simulation model of motor synergies is a very applied model, whose results immediately relate to empirical science. This very tangible way of using ER simulation models gives an example of the potential of simulation models to generate proofs of concepts and to illustrate logical and mathematical states of affair beyond our cognitive grasp. By contrast, this chapter presents a simulation model whose results are of a more general and theoretical nature. It investigates the conceptual soundness of arguments proposing a certain type of control architecture for life-time adaptation. The architecture modelled is very wide-spread and features a "value system' for self-supervised behavioural learning. The term "value system' is borrowed from Edelman et al.'s work (e.g., Sporns and Edelman, 1993), but the idea is much more generally applied. The simulation model illustrates some of the implicit premises that underlie this kind of architecture and demonstrates that the adaptive capacity of such circuits can break down in closed-loop agent environment interaction, if no additional mechanisms to secure intact functioning are in place. The results from the model presented in this chapter have been partially published in (Di Paolo et al., forthcoming; Rohde and Di Paolo, 2006).