The objective of this paper is to analyse some of the conceptual and methodological contributions that complexity theory can make to the study of socio-technical cooperative systems. The theory of complex systems has developed along two complementary, but nevertheless distinct, axes. Chronologically, the first unifying concepts of the complexity paradigm resulted from the study of non-linear systems. Later, the study of distributed self organising systems made it possible to widen this initial approach to the analysis and modelling of social cognitive systems. The first school (non-linear systems) brought many conceptual and methodological contributions, however, these contributions are not directly applicable to the study of complex socio-technical systems, which are precisely the systems of interest to ergonomists and sociologists. On the contrary, the distributed approach, being interested in local interactions rather than structure and hierarchy, has found many applications fields ranging from the study of animal micro societies (ethology) to the study of human organisations on a social or cognitive level. Using examples from our analysis of human work activities, we will show how the concept of complexity can improve the methods of modelling and the design complex socio-technical systems. This paper concludes by trying to find an intermediate position between the analytical and complexity approaches which would allow us to understand real situations in better way.