The research on complexity, although interdisciplinary by nature, stems originally from studies of natural systems. Artificial systems have only recently reached the level of sophistication that requires explicit handling of complexity to achieve appropriate understanding about their properties and behaviour. The necessity to capture sufficiently precise behaviour of a system has lead to building more and more sophisticated structures of primitive (and not so primitive) components. Whereas each of those components can (in principle) be modelled by formal (e.g. mathematical) models that approximate many aspects of the respective subsystems with high precision, the interactions of those components depend on a multitude of incompletely known factors that, in many cases, can not be included into the formal models. Those interactions may occur sporadically, and the properties of interactions may change dynamically depending on the incompletely known factors. Hence, the overall behaviour of many systems can not be described as a conventional mathematical superposition of behaviours from their components. Examples from neuroscience, e.g. modelling of neurons and neural networks (Clewley 2001), condensed matter and statistical physics (Pietronero 2007), non-linear mechanics (Engelbrecht) support the above observation.