Changing Maps: Empirical Legal Autopoiesis John Paterson and Gunther Teubner

these theories stimulated empirical research. But while such projects have flourished in areas as diverse as economy, psychotherapy and flamenco, in ‘law and society’ there has so far been a paucity of empirical research on self-organisation. This anomaly can perhaps be traced back to certain peculiarities of legal sociology as a field of knowledge. It seems that the long-lasting and deep hiatus between theory construction and empirical research is actually deepened by the emergence of theories of self-organisation and autopoiesis. This is our first thesis. Secondly, if we look more closely at concrete, detailed, historical research carried out in the name of autopoiesis, we can discern clear discontinuities with ‘normal’ practices of empirical research. Autopoiesis calls for a redefinition of empirical work and requires different empirical tools—tools that are capable of analysing the transformational dynamics of recursive meaning processes. As a consequence, everything changes: research questions, the phenomena to be identified, the concepts to be made operational and the analytical instruments. And there are even stronger anomalies in the socio-legal relationship between the empirical and the theoretical. The constructivist orientation of legal autopoiesis, we submit in our third thesis, works against the fantasies of omnipotence inherent in the process of empirical falsification. Legal autopoiesis is not anti-empirical, but it does suggest a role for empirical research that is different from straightforward Popperian theory-killing. It suggests, instead, a quasi-therapeutic relationship between the speculators and the data collectors. But who, then, is the therapist and who is the patient?


Why is there a structural hiatus between theory and empirical research in law and society? In the classics of legal sociology, Marx’s historical methods, Durkheim’s choses sociales and the ideal-typical method in Weber’s interpretive sociology were guarantees of the unity of empirical research and grand theories of law. But the introduction of modern empirical methods signalled problems for theory.

‘The dissolution in data and their recombination with the help of newly developed methods of data analysis destroyed the high level of theorising which had been built up in the classics without being able to substitute it adequately.’4

Today, the field is still suffering from this deep hiatus, which renders theory rather empty and empiricism rather blind. Or to put it more mildly, empirical research in law and society has developed a highly sophisticated methodology which is, however, based on poor and rather ad hoc theorising, while theorising about law and society has become more and more philosophical and speculative relying, however, on poor and rather ad hoc empirical support.

And today the hiatus is deepening. Empirical legal sociologists are giving in to the temptation of trying economic models and theories for their data with the predictable result that they are losing their sociological identity. Meanwhile, legal theorists are tempted to follow the famous ‘linguistic turn’ in sociology and thus to question the validity of systematic data collection and patient data analysis.

Usually it is the micro-macro problem that is held responsible for the empirico-theoretical gap. Empirical methods are good at gathering individual data at the micro level of legal action and aggregate data at the macro level of socio-legal relations. But they fail when it comes to analysing law’s ‘organised complexity’ which good theory regards as central to understanding law as a social phenomenon. Without denying the importance of the micro-macro difference, we prefer to identify another famous petite différence as responsible for the great hiatus: the difference between law as operation and law as observation,5 which has sharply divided socio-legal theoreticians and empiricists. Empirical analysis has opted for first-order observation of the law. It takes legal action as simple operations, as spatiotemporal events, which can be correlated in empirical models with other social events. This drives empirical analysis of law in two directions: towards models of logical and mathematical formalisation on the one hand and towards attempts at causal explanation and prediction on the other.6 In contrast, ambitious sociological theories of law are usually second-order observations. They see legal action itself as observation, as a trinity of utterance, information and understanding, as the recursive transformation of differences, as constructing a special space of meaning and an autonomous world of knowledge. This drives socio-legal theories deeper and deeper into the hermeneutic tradition, which allows for sophisticated analyses of the ‘operation called Verstehen’, but which ridicules attempts at formalisation, causal explanation and prediction. And attempts to combine both traditions are sucked into the black hole bounded by formalisation, causal explanation and hermeneutics.

If this is an adequate sketch of the intellectual map, how does self-reference and autopoiesis change the somewhat desperate outlook for law and society? At first sight it looks like Columbus’s egg, as François Ewald has called it.7 It nourishes hope for a recombination of both the empirical-analytical and the normative-hermeneutic traditions. It seems to promise a bridge between law as operation and law as observation since it compels us to combine first––and second––order analysis. Since law is defined as a closed system of self-reproductive observing operations, legal action is seen as being at the same time both operation and observation. This requires the normative tradition to leave Popper’s World III and to search for ‘law in action’ as its social base, and it requires the empirical tradition to include in its observations the complex chains of normative observations of the ‘law in the books’.

But a closer look reveals that autopoiesis offers no easy synthesis. It burdens the three traditions—the hermeneutic, the formal and the causal orientation—with an almost unbearable task. How to cope with self-reference? Hermeneutics, with its long tradition of dealing with self-referential relations, reflexivity, paradoxes and hermeneutic circles, is obviously in the best position. This explains the rapid development of autopoiesis in hermeneutically oriented theories of law. In a view of law as a concatenation of communicative events based on a code which deparadoxifies a basic self-referential relation autopoiesis has strong (s)elective affinities with discourse analysis as developed by the maître-penseurs of poststructuralism: Foucault, Lyotard and Derrida.

The tradition of formalisation in legal theory has much greater difficulties with autopoiesis. The reason is that the paradoxes of self-reference pose an enormous challenge for a formal calculus. It is true that Hofstadter’s famous book on the enigmas of reflexivity and self-reference has had a certain impact on legal theory.8 However, sophisticated attempts to come to terms with self-reference, such as Spencer-Brown’s Laws of Form,9 the development of a multi-value logic by Günther10 or ‘A Calculus for Self-Reference’ by Francisco Varela11 have up to now only found one resonance in legal sociology which is Niklas Luhmann’s discussion of the legal paradox and the binary coding of law.12

However, the situation for causal explanation and prediction, the precious hope of orderly empirical work in law and society that would transform it into a real science, is disastrous. For causal analysis, self-reference is an explosive. The blast comes from a theory of recursive systems and from a concept of non-trivial machines; and the blast is so strong because these explosive concepts were developed not just from the hermeneutic softies of the Geisteswissenschaften but also from the hard-liners of the exact sciences. According to the sociologists Krohn and Küppers, who deal with problems of the legal regulation of social fields, the results look like this:

[i]n non-linear systems with a recursive dynamics … there are only few cases in which prediction of the system’s development is possible, even if their mechanism is known, the systems are deterministic and disturbances do not occur … Due to recursion, even very small deviations in the initial conditions are reinforced in such a way that similar starting constellations lead after a very short time period to totally opposite system developments … . In the case of a non-linear and recursive system dynamics … no prediction of the system’s development is possible.13

And if law as a social system is correctly defined as one of these ‘non-trivial machines’ (that is, as one of the deterministic systems whose input-output relationship is not invariant, but is determined in a self-referring way by the machines’ previous output), then, in the words of von Förster, ‘for all practical reasons they are unpredictable: an output once observed for a given input will most likely not be the same for the same input given later’.14 The only hope for causal explanation and prediction is a trivialisation of law and society, their social construction as trivial machines—something that happily coincides with the triviality of certain results of attitude and impact research, results that everyone familiar with the fields already knew in advance.

So what does this mean for the chances of empirical research in the autopoietic framework? Well, they look excellent for all kinds of historical analysis, for genealogical and archaeological digging in historical texts, and for qualitative research techniques, case studies of formal organisation, ethnomethodological types of socio-legal interaction, discourse analysis, for ‘critical empiricism’. And indeed these are the research techniques that are mainly used in the empirical projects. For static correlations (of ‘the more x, the more y’ variety), however, the chances look rather bleak.


But before we get carried away, is it not the case that autopoiesis is simply incompatible with the dominant working orientation of orderly empirical research, where the task of theory is causal explanation and prediction of empirical facts, and the task of empirical research is the reality test of hypotheses derived from theoretical constructs? Indeed, it is incompatible. Viewed from the constructivist position of autopoiesis, every element of this statement about the empirico-theoretical relationship is flawed.

To put the counter-position bluntly:

  1. Empirical research is by no means closer to the reality of the outside world than theory. Even from empirical experience we know that often the opposite is true. The hard facts about the external world that empirical research pretends to produce are in reality highly artificial constructs, excessively selective abstractions, mere internal artefacts of the scientific discourse that are both as real and as fictional as are theoretical constructs.
  2. The real role of empirical research does not lie in dull falsification. It is in the ‘surprise value’ of its self-produced data. Empirical world constructions in law and society do not need to be destructive of theories. Rather, they could play a maeutic role in the birth of theories in the spirit of empiricism.
  3. Causal explanation and prediction are grossly overestimated in law and society. They are only special cases of theoretical work, which are indeed very rare, and they by no means exhaust the potential of theoretical explanation.
  4. For autopoiesis, theoretical explanation of empirical results means that the theory reformulates these artefacts of perception in new contexts in order to analyse the transformational dynamics of recursive meaning processes.

Let us take a concrete example of the social effects of legislation in order to discuss this counter-position. Occupational health and safety in Britain’s offshore oil industry constitutes a well-defined area that has seen considerable regulatory development over its 40–year history. Traditional empirical research on the effectiveness of law and implementation research suggest the construction of a network of dependent and independent variables among which we can identify correlations and find out their causal connections.15 The usual causality chain—as Renate Mayntz, for example, tells us—works like this: political goal definition → legislative act → legal norm → motivation of implementation staff → motivation of actors in the field → deviation/sanction/incentive → social behaviour → social effects.16 In our concrete example of offshore health and safety regulation, it is possible to trace this sort of causality chain as follows.

During the early days of the offshore industry in the mid–1960s there was no detailed regulation of occupational health and safety, simply an instruction from the government that those involved should follow an industry code of practice. When a serious accident occurred in 1965, the inadequacies of this approach became evident and an Inquiry chaired by a lawyer recommended that ‘a statutory code with credible sanctions’ be implemented to provide for the safety of workers in the industry.17 This recommendation was accepted by the government of the day, which introduced a Bill to Parliament that eventually became law as the Mineral Workings (Offshore Installations) Act 1971. This provided a framework for the development over a period of years of detailed regulations by the regulators (mainly the Petroleum Engineering Division of the Department of Energy) covering every aspect of the industry from the design and construction of offshore installations to the content of first aid kits. These regulations were then implemented and the oil companies they were aimed at complied with or deviated from them ultimately producing an effect on the level of safety that existed in the industry.

We might summarise this on the basis of the foregoing causal chain as: political goal definition by the Ministry of Power Inquiry → introduction of a Bill to Parliament by the government → passing of the Mineral Workings (Offshore Installations) Act 1971 → development and implementation of detailed regulations by the regulators → compliance/deviation by the industry → effects on safety. In accordance with this understanding of the regulatory chain, when questions are asked about continuing safety problems in the industry, we find concern about delay in getting detailed regulations into place and about the toughness of the regulators’ enforcement.18 In other words, control of safety will be achieved when detailed regulations are in place telling the industry what it must do and when these are being enforced by the regulators.

Autopoiesis, however, forces us to break up this causal chain of events and to replace it by a multitude of autonomous but interfering fields of action in each of which, in an acausal and simultaneous manner, recursive processes of transformation of differences take place. To put it more simply, a single horizontal chain of causal relations is replaced by a multitude of vertical chains of recursions. We can indicate this shift graphically in Figure 1.

Moreover, our framework gives us a new understanding of social regulation through law. Understanding these vertical chains of recursions as operationally closed means that each constructs information internally: there are no input-output relationships between, say, the regulators and the industry. As a consequence, attempts by the regulators to steer the industry by means of prescriptive regulations backed by sanctions and incentives must be understood in a fundamentally different way. Such attempts can only ever be a multitude of self-steering processes. More specifically, this self-steering must be understood as the minimisation of a difference, an attempt to reduce the difference between the current situation and the desired one. This definition is consistent with all forms of steering but in the context of a recursively closed system of communicative operations the difference is itself internally constructed. Thus, offshore safety regulators construct the current situation according to their own code and similarly construct a desired situation and apply their own programme of difference minimisation in an attempt to arrive at it. Given that the industry constructs reality according to its own code and steers according to its own difference minimising programme, the limits of regulatory ambition become clear. In other words, regulation is possible only as self-regulation within each of these recursive processes. Regulation over the boundaries of action fields is impossible. Chains of causality need to be replaced by simultaneous events of structural coupling. This is not to say, of course, that regulatory attempts produce no effects, only that those effects cannot properly be regarded as steering in the sense implied by traditional theories. Instead, these effects arise from the construction of differences by the regulators and their attempts to minimise them but depend on the internal construction of differences by the industry and its attempts to minimise them.19


Figure 1: The Shift from a Horizontal Chain of Causal Relations to Vertical Chains of Recursions.

This is a suggestive idea, but can it be made empirical? The task for empirical research in these circumstances would become one of inquiring into several chains of difference minimisation and into their interferences. We would have to retell in detail several divergent stories of self-regulation in the political arena, in the legislative chambers and courtrooms, in the offices of the regulatory agencies and in the managerial suites of corporate actors, and on the drill-floors of offshore installations. The question would be one of how, in each of these stories, the events common to them are idiosyncratically reconstructed and processed in the meaning context of their specific difference minimisation programmes. To be clear, such a division of the regulatory chain into divergent stories does not imply that autopoiesis is bound to discover regulatory failure. Autopoiesis is not in some sense the opposite of regulatory success, as Nahamowitz seems to believe.20 Instead, understanding steering as self-steering means that the theory accounts for regulatory failure and success in ways different from theories where linear causality is assumed. So, if we find that our different stories of recursive operations travel together for a time in a common direction instead of diverging then we can readily speak of regulatory success.

The crucial question, then, is how to disentangle the connections of these multiple cascades of concatenated differences. To repeat, we do not mean causal influences, but the acausal synchronisation of ongoing parallel processes. And our theory tells us that there is not one magic formula of structural coupling; rather there are several types of synchronisation. In order to find out how the different recursive processes are interrelated we need first of all to find out how they are closed to each other. L’ouvert s’appuye sur le fermé21—this is not a matter of theoretical definition but a matter of empirical variation. Autopoiesis theory suggests a variety of closure mechanisms in the relations between meaning systems to which correspond a variety of ways in which they are open to each other: from ad hoc contacts to systematic linkages and long-term co-evolution. Success or failure of regulation depends— this is our guiding hypothesis—on the specific qualities of interwovenness of several recursive meaning processes, which in turn depend on the qualities of their mutual closure.

This compels us to ask a twofold question when it comes to detailed empirical research:

  1. How can we identify concretely the multitude of elementary acts— meaning operations—that constitute the autopoietic closure of the various processes involved?
  2. How can we identify the different types of mutual recontextualisation that are responsible for a meeting of these closed discourses?

Applying the first question to our example: are the legislative process and the implementation field autopoietic systems? Although we have so far spoken as if they are for the sake of the argument, this is not in fact a question we can answer theoretically but only by empirical observation. Autopoiesis theory does not impose a set of pre-existing systems but rather compels us to observe the concrete interactions in legislative chambers, lobby halls and the technological processes in our implementation field in order to discover the systemicity of our research object. Strangely enough, this reliance on empirical knowledge runs counter to the opinion of empirically-minded researchers who tend to treat this as an ‘analytical’ question, namely the identification of a ‘system’ as the somewhat arbitrary conceptual selection of the field of inquiry according to the concrete research interests. In contrast, the system concept of autopoiesis is much closer to empirical reality than the abstract models of empirical research.