It is argued that for theoretical and empirical reasons an entirely Empiricist account for perception and action is not possible. This is related to results on the necessity of hidden layers in connectionist models of cognition. For paradigmatic examples in space perception and motor behavior it is investigated what processes can account for the apparent order in these domains. A general tendency towards stability is identified but, for instance, the phenomena of apparent depth in plane drawings or multistability reveal that different kinds or points of stability may compete. The finally achieved stable percept usually depends on symmetry breaking. Similar phenomena can be found in motor behavior where singularities play a similarly decisive role. Finally, the "cost" of these order producing processes in discussed in terms of veridical representations and learnability.