Crisis Thresholds: network demarcation and the Kuhnian turning point

Crisis Thresholds: network demarcation and the Kuhnian turning point

In at least some of his work (e.g. 1962), Kuhn refers to a ‘crisis’ within the scientific community, which occurs when the build-up of anomalies becomes so substantial that most of the community begin to search for an alternative to the dominant paradigm. The crisis precipitates a period of revolutionary science.

Some commentators view this as the introduction of an ineliminable subjective and sociological dimension to scientific theory-change. The point at which crisis occurs is determined by sociological and subjective psychological factors within the community and the scientists who make up the community. An account of demarcation by Reisch (1998), however, suggests an alternative view. We can turn Reisch’s ‘Network Demarcation’ account on the crisis threshold. Thinking about the scientific network more broadly allows us to see why a crisis threshold is epistemically warranted. It also accounts for a changing threshold over time, depending on the state of the scientific network, and offers a testable account of scientific theory-change which yields empirical predictions about the rate of acceptance of a change amongst different scientists and groups.

Reisch views science as a network of theories, methods, processes and phenomena. As proponents of pluralistic views like Mitchell (2002) and Brigandt (2010) have argued, to solve complex scientific problems, we must draw upon many theories and fields. Duhem (1906) pointed this out obliquely; Duhem’s thesis is that, in order to derive an empirical prediction from a theory, we need both background theoretic assumptions and ways of specifying the experimental conditions. Pushing on this point, we’ll find that the background theories and methods for controlling experimental conditions come from a broad range of theories and fields. For instance, consider the complexity of cooling a system down to milikelvin in order to perform an experiment. Regardless of what that experiment is, or what field it is in, we will need some serious physical and chemical theory and techniques to achieve super-low-temperature cooling. We’ll certainly draw upon quantum theory. The point: to do science today, we need a network of interconnected theories, methods and techniques. The techniques of a field may be substantiated by the theories of another, and so on.

From the perspective of theory-change, this makes a key point salient: if you change part of the network, the rest of the network could be disrupted to some extent. If I have to change a step along the way to my derivation (in Duhem’s terms), then maybe my derivation will no longer work. We see this all the time—Newtonian physics forced us to adopt a new theory in cosmology, and postulate the existence of Neptune. Astronomical data forced us to question Newtonian physics when another postulated planet, Vulcan, could not be found. Reisch uses the notion of the scientific network, and the effect of changing a part upon the rest of the web, to try to solve the demarcation problem. Reisch argues that anything which cannot be suitably integrated into the network without disrupting too much of existing science (or rather, without disrupting it negatively—that is, without replacing the old functionality and adding more besides, akin to Lakatos’ (1980) notion of progressiveness) cannot be scientific. The network is protected from major disruption by demarcation. We cannot, for instance, accept Creationism into the scientific network, because it would sever links between fields like biology and paleontology, geology and biogeography. The same applies to astrology and homeopathy—the former disrupts a social scientific network of intentionality and personality, as well as a physical network of exertion of forces, while the latter butchers the physical scientific network, chemistry and medicine with the claim that the potency of a solution increases as the solution is diluted, even past Avagadro’s number.

Reisch’s account does not solve the problem of demarcation. Reisch relies on a central assumption—the unity of science. This assumption is controversial. There is no obvious justification for the assumption that there is one single united scientific network. Science need not be a single consistent network. There may be many different networks which are joint and severally scientific. Of course, it’s possible to draw links between different fields and maybe include all scientific fields in that spider’s web. But this will not really connect all of science as a single coherent network. For different problems, in different fields, we will draw upon different sub-networks. We may also see competing networks. There could be two or more networks of theories and techniques attempting to explain the same (or overlapping) phenomena. This can often be seen in fundamental physics, for instance. These competing networks, importantly, may incorporate many shared components. Rival theories of fundamental physics could draw upon the same theories in computer science for testing models or of chemistry in setting conditions for experiments. They may employ the same techniques. But they may have different theories occupying the same node in the network. Crucially, each may exclude the other from any network which contains it. As such, they are not just two parts of a larger network, but two rival networks.

If we admit the existence of rival networks and multiple scientific networks—i.e. we adopt scientific network pluralism—then Reisch’s network demarcation strategy will fail. Nothing in his account prevents a pseudoscientific discipline establishing its own rival network, incorporating some scientific theories and techniques, and calling this a scientific network. Reisch does not provide demarcation criteria to distinguish scientific from non-scientific or pseudo-scientific network. Network demarcation uses networks to perform demarcation, but is not a demarcation of networks. With this in mind, we can see the application: ‘Christian science’ has established its own purportedly scientific network, which includes techniques and theories from mainstream science, but also endorses intelligent design theory, divine intervention, sometimes views from alternative medicine, and even Noachian geology. We can debate whether this allegiance of religious theories constitutes a consistent network. But if it does, we’d need more apparatus than Reisch provides to argue that this network is not a scientific one. The same goes for a network of scientific alternative medicine, or a new age network. While I, like Reisch, am convinced that these are not scientific networks, we will need more work to argue the claim, particularly the provision of a demarcation criterion for the scientificity of networks.

But while Reisch’s network demarcation fails to solve the big-name problem, it does suggest an approach to theory-change. Network alterations can jeopardize the links between theories and fields. As such, before we alter the network we are rationally justified in requiring assurance that the resulting changes that ripple through the web will be for the better. We cannot just jump on a bandwagon of a new theory without considering what it does to the network (or networks) in which it must be embedded in order to solve complex scientific problems. We want reassurance that we will still be able to solve the problems our unmodified network allowed us to solve. Moreover, we want reassurance, before we go about resolving old problems and doing the hard work of re-forging inter-theoretic links and adapting techniques, that the network will be better for it. There is no point in exerting a great deal of scientific labour updating our network to fit a new or revised theory at an important node if the resultant network can’t do anything more than the old network could.

Hence, there are two tests which much be met before a theory-change could go ahead: (1) the re-networking test, and (2) the network superiority test. The re-networking test requires that proponents of a new theory or programme demonstrate the ability to rebuild severed links within the new network, suitably equipped to allow previous work to be adapted to continue or to identify any work which is rendered problematic by the change and needs revising. Those proponents must provide tools to do this work. The network superiority test needs to show, at least in some domains, that the new network is capable of some feats which the old was not. Meeting these tests takes time, energy, and considerable interdisciplinary coordination, all of which sets a high bar for theory change (albeit a rationally defensible high bar).

Part of the sociological and subjective bent to Kuhn’s notion of a crisis within the scientific community comes from the individualised and localised way in which crisis arises. For sure, crisis occurs when a large proportion of the scientists in a field are uneasy about core theories. But these scientists reach that point locally—they are influenced by one another, but the factors which convince each individual or each local group can vary, as can the weight they attach to each factor. If this is purely or mainly subjective or sociological variation, then crises seem to be purely or mainly social and subjective.

But considering networks changes this perspective. Suppose we accept that scientists should rationally wait until they are convinced that a change meets these tests sufficiently to justify the exertion that re-forging inter-theoretic links and re-solving old problems brings. Now, suppose that different scientists and different groups within the same field work on different problems, using different methods and different equipment. Then, they will draw upon different sub-networks to do so. Different links within the broader scientific network will be more important for them than for others in the field. Network positioning creates a different set of priorities for different scientists and groups, and so sets in motion different rates of satisfaction with the change. We would predict, then, different groupings of scientists who would share common characteristics in terms of their willingness to accept a change. This is predictable on the basis of their network positioning, in terms of the network links which are critical to their work. The groups whose acceptance of a change is easiest to achieve are those for whom links are rebuild most easily, naturally and parsimoniously, and particularly those for whom early network superiority results can be demonstrated. Those who lag will tend to be those for whom rebuilding is slow, threatens the parsimony of their network, or presents persistent obstacles to rebuilding, particularly where local network superiority cannot be established, or the local network change results in inferior performance.  

The inter-theoretic links which are important to particular scientists’ work vary between individuals and groups. These links may be easier or more difficult to reconcile with a particular theoretic development at a node within a scientific network. Re-forging a link for which the supplanted theory is a central node may be very time-consuming. It may not be obvious that the link can be reformed until others working in other parts of the field have made suitable adjustments to other parts of the network.

Ultimately, then, network positioning in Reisch’s framework offers a way to think about Kuhnian transformations which gives first a defence of the high bar for crisis thresholds and second which predicts purely on the basis of their perspective on the network, a reasonable and scientifically-motivated staggering of the acceptance of the change between groups and scientists. Moreover, this approach would offer empirical predictions which could be tested to evaluate the accuracy of the account. Evaluating scientific networks during periods of crisis, this account predicts a strong correlation between the speed of acceptance of the new theory, and: the level at which the change severs links which are crucial to a group, the speed and complexity of rebuilding links, and the location within the new network of superior and inferior performance.


  • Brigandt, I. 2010. “Beyond Reduction and Pluralism: Towards an Epistemology of Explanatory Integration in Biology.” Erkenntnis no. 73:295-311.
  • Duhem, P. 1906. The Aim and Structure of Physical Theory. Princeton: Princeton UP.
  • Kuhn, T. 1962. The Structure of Scientific Revolutions (1st ed.). Chicago, IL: The University of Chicago Press
  • Lakatos, I. 1980. “Falsification and the Methodology of Scientific Research Programmes.” In The Methodology of Scientific Research Programmes: Volume I: Philosophical Papers, edited by Worrall, J. Cambridge: CUP.
  • Mitchell, S. 2002. “Integrative Pluralism.” Biology and Philosophy no. 17:55-70.
  • Reisch, G.R. 1998. “Pluralism, Logical Empiricism and the Problem of Pseudoscience.” Philosophy of Science no. 65 (2):333-48.