III

If so far the most coherent attempts to apply basic concepts of a theory of nonlinear non-equilibrium systems to social phenomena had their specific point of reference in the notion of autopoiesis, a self-organizing mechanism rather than a process leading from order to chaos, we may now consider in more general terms the theoretical promise that chaos theory (both in the wider and the narrower sense) holds for sociology. Walter Bühl (1990) argues that the sociological relevance of natural science theories of nonlinear dynamics rests in particular in their application to the analysis of discontinuous social change - of fluctuations, catastrophes, and chaos, as the subtitle of his book reads. According to Bühl, such change processes have been neglected in the recently dominating sociological theories, which have rather been concerned with problems of social order, social differentiation, and patterned structural change. As we have seen in section 1 of this paper, this diagnosis is correct if we think of explicit post - World War II macro theories. It is not quite correct with respect to the broad mass of sociological work; nor does it apply to such classical theorists as Karl Marx. We do, however, lack an integrated, modern sociological theory of discontinous social processes that might provide a general framework for the wide variety of specific cases that have been analysed by sociologists - from riots and mass violence to revolutions and other forms of structural disruption. The question is to what extent natural science models of nonlinear dynamics in general and chaos theory in particular would be helpful in the development of such a sociological theory. Doubtless natural science theories and mathematical models of nonlinear dynamics may help us to get a better understanding, or at least a better formal grasp, of certain kinds of social discontinuities. However, keeping in mind what has been said earlier about fruitful and abortive forms of theory transfer, what is needed in any case is the development of a genuinely sociological theory, instead of a mere reformulation of existing insights in the fashionable terminology of "chaos."

Theories and models of nonlinear dynamics hold still another, even more general theoretical promise for sociology, over and beyond their ability to direct sociological theory building to the mechanisms underlying various forms of discontinuous processes. This promise derives from a feature that all of the natural science theories and mathematical models that come under the heading of nonlinear dynamics have in common: they all deal with the generation of macro events or macro patterns from micro processes. By virtue of their dealing with the micro - macro link in systems behaviour, these models also contribute to a better understanding of the hierarchy problem. Hierarchy in this connection does not refer to vertical control relationships, but to a succession of different levels where the phenomena at each higher level have properties that cannot be derived by summation from the properties of the lower level phenomena. To the extent that the natural science models can be applied to social reality, they might help to clarify the micro - macro link, i.e. what is probably the most crucial general problem in present social theorizing. This special theoretical relevance may have been most evident in the case of the self-organization paradigm, but it holds in principle for all models of nonlinear dynamics that generate macro events from the uncoordinated (though interdependent) micro behaviour of the system elements, whether these macro events are phase transitions to chaos or to new stationary states far from equilibrium.

That a theory of self-organization should be able to contribute to a specification of the famous micro - macro link is explicitly recognized by its natural science proponents. The point they all emphasize is that self-organization produces qualities at the macro level of the systems considered that cannot be derived from, or explained by, reference to the measurable properties of the elements. With reference to ferromagnetism and the laser, Haken for instance states: "Thus the order on the microscopic level is a cause of a new feature of the material on the macroscopic level" (Haken 1978: 3). In the case of iron, it is the quality of magnetism, in the case of the laser, a light beam of particular intensity that is generated. Similarly, in the case of the Bénard instability, the macro quality is a pattern of fluid motion that obviously cannot be derived directly from the qualities of the H₂O molecules. Since, in Haken's words, "to describe collective behavior we need entirely new concepts compared to the microscopic description" (Haken 1978: 13), he introduces the term "order parameters" to refer to such macro-features. An order parameter represents a macroscopic property emergent from interactions at the microscopic level.⁷ Of course the properties of the elements are important because they imply specific capacities for influencing each other and being influenced, but it is their interaction that produces the new structure. It is evident that these formulations can also be applied to the relationship between micro behaviour (individual behaviour) and macro properties of social systems. The promise to gain a better understanding of emergent effects in social systems resulting from the behaviour and interactions of the system members is probably the main reason for the attractiveness of the self-organization paradigm to social scientists, and it plays a focal role also in the transfer of mathematical models of nonlinear processes.

Though micro - macro relations have been a perennial theoretical problem in sociology, this problem seems to have special salience in present-day sociological discourse. This special salience has to do with the renaissance of action (or agency) theory in today's macro sociology. As long as structural change at the systems level is explained by reference to functional imperatives, changing values, or technical development, macro events are explained by macro variables, and there is no need to go back to the micro level of individual action. Only when attempts are made to explain macro events by micro action, to paraphrase the title of a famous book by Thomas Schelling (1978), are we confronted by the challenge of linking micro and macro levels.

About 40 years ago, Talcott Parsons (1951) made the attempt to link the micro level of social action and the macro level of societal change in one theory, but his conception of a hierarchy of system levels and of the interpenetration between social and personal systems is better suited to explain the macro determination of micro events than the generation of emergent macro phenomena from micro interactions. More relevant is the concept of unintended consequences, which have often been identified as the proper object of sociological enquiry. A generation ago, Hayek formulated this view as follows: "If social phenomena showed no order except in so far as they were consciously designed, there would be... only problems of psychology. It is only in so far as some sort of order arises as a result of individual action, but without being designed by any individual, that a problem is raised which demands theoretical exploration" (Hayek 1955: 39).

Where sociologists are interested to explain emergent macro phenomena by reference to the actions and interactions of individuals, they adopt a perspective that has a long tradition in economics. Not surprisingly, therefore, the most explicit contributions to the formulation, if not to the theoretical solution of the micro - macro problem in sociology, have not been made by systems theorists, but by authors who do not only use an actor's perspective, but are at the same time methodological individualists using a rational choice approach (Lindenberg 1977; Boudon 1977). The adoption of such a perspective is also motivated by a particular extra-scientific cognitive interest - they query whether man by his own actions will unintentionally bring about the annihilation of society.