This is an unedited preprint of a review article that was publised in 1999 in the South African Journal of Philosophy, volume 18, number 2, pages 258-274. This version does not have the footnotes that appeared in the published version. Were I to write such a review article now I suspect that I would be far more spirited in my criticisms - DS.

Cilliers, P. (1998) Complexity and postmodernism: Understanding complex systems, London: Routledge.
A Review Article

David Spurrett
Philosophy Programme, University of Natal, Durban, 4041, South Africa.


Again, regarding the brain, they will not be able to imagine anything more plausible than that it is composed of many tiny fibres variously interlaced..."

Descartes, Treatise on Man (1985, vol. 1: 107).

(1) Paul Cilliers’ Complexity and postmodernism (1998) develops a philosophical framework for understanding complex systems, and argues that, properly deployed, this understanding shows the way out of a number of old philosophical disputes and confusions about meaning, representation and reductionism by showing that certain traditional philosophical expectations relating to those problems are misguided. The traditional philosophical expectations in question, which are connected to some programmes in the philosophy of science and in scientific research, are paradigmatically modern conceptions of the nature of language, rationality and thought. These include the notion that meaningful signs are tokens of some sort which "have" a meaning of the form of a positive content, the related view that thought is essentially representational, and the idea that rationality is essentially algorithmic or computational.

Parallel to this line of enquiry, and partly independently, Cilliers argues that at least some "postmodern" philosophy (especially Derrida and Lyotard, although Baudrillard also features) shows itself to be especially sensitive to complexity, and to have already moved away from the problematic notions of meaning, representation and rationality, and towards conceptions which are much closer to those suggested by the study of complex systems. The overall thesis of Complexity and postmodernism is thus that there are possibilities for mutually enlightening conversations between postmodern philosophers and researchers into complexity, an opportunity which Cilliers both defends and exercises.

Cilliers explains that complex systems have certain important characteristics, including being constituted by a large number of elements interacting richly, locally and non-linearly, containing feedback loops and being "far from equilibrium" (CP: 3-4). For research purposes a popular type of complex system is a "neural network" where a collection of nodes are linked up in various ways, and the network then processes "synapses" according to a distribution of weights for the connections between the neurons and transfer functions for each individual neuron. There are various types of neural network, and Cilliers discusses some of their differences and similarities (CP: 26f). A short discussion of one type will illustrate what is at stake.

A "back propagation" network is one designed in such a way that, given an input at one end, having some intermediate layer(s) of neurons, and then having the correct output "forced" at the other end, will internally adjust the weights and transfer functions between input and output so as to make the forced output a normal response for the network.

By means of presenting a variety of inputs and forcing the appropriate outputs, such networks can be "trained" to perform a very wide variety of tasks. One of the examples Cilliers discusses (CP: 31f) is the network developed by Rumelhart and McClelland which was trained to generate the past tense forms of English verbs when given the present tense as input. As long as the networks are sufficiently rich (i.e. have enough neurons and connections) then, as with the Rumelhart and McClelland example, they can be trained to be both very accurate, and also flexible, which is to say that they perform well after training when presented with "new" examples not in the original training data set. A crucial point is that when the distribution of weights and transfer functions is analysed after training what is found is that there is nothing in the network which counts as a representation of that kind of thing which is processed, and no rigorous way to describe its processing terms of rule-following. Cilliers notes that Rumelhart and McClelland concluded that the notion of grammatical rules seemed to have only descriptive application (CP: 32), and found no reason to think that rules were essential to either learning or competence. (These findings in models of complex systems are mirrored in research on actual biological mechanisms, such as Thelen and Smith’s work (1993) on how such competencies as walking require no "essence of locomotion" (1993: 17) to be encoded in the living organism.)

Cilliers points out that there are striking and instructive similarities between Saussure’s model of language (especially as developed by Derrida) and some of the conclusions of connectionist research (CP: 38f). As is well known, rather than focusing on a notion of language where meaning was regarded as the positive content of the meaningful sign, Saussure emphasised the role played by differences and relationships between signs, and saw meaning as distributed through a system rather than located in the individual tokens. Although Saussure himself tended to regard developed systems of signs as relatively stable and unchanging, Cilliers explains how Derrida has contested this view (CP: 42f), and argued for the need to recognise an unavoidable slippage or non-identity in any deployment of a sign in virtue of both the essential iterability of the signs themselves, and also the philosophical indefensibility of any notion that signs or collections of signs could be rigorously compared with intentions or meanings so as to fix their content, or to contexts to as to fix their proper interpretation (Derrida 1977).

Cilliers observes that the connectionist and post-structuralist programmes have arrived at a position where meaning or content is both distributed through a system, and also the consequence of networks of difference and relationship, and where the notion of representations as being the main currency of thought has been powerfully contested. (Cilliers is careful not the draw the radical and probably premature conclusion that representation has been thrown out as such (CP: 58-88), since the more important philosophical task here is challenging the traditional status of representation, rather than trying to get rid of it altogether.) The poverty of the rule-based approach to rationality and intelligence are elegantly demonstrated through a discussion of Searle’s Chinese Room thought-experiment in a chapter called "John Searle befuddles" (CP: 48-57). This occasion also serves as an opportunity to show how the traditional speech-act approach to language fares in the face of Derrida’s criticism of the notion that speech acts can compared with intentions and situated in specifiable contexts so as to fix meaning and interpretation.

It is Cilliers’ contention that his analysis suggests the possibility of mutually beneficial dialogue between the two (connectionist and post-structuralist) programmes, and his own book is an example of what can be achieved when the two fields are carefully and significantly related. Cilliers’ book is a good one, and I see no advantage in constructing a series of criticisms narrowly directed at his argument. In what follows, therefore, I begin by suggesting a few areas of detailed disagreement which I think are worth discussing (part II), and then move on to address a number of wider questions in areas raised by the book (part III).


(2) My first specific point of disagreement concerns chaos theory. I am not likely to be the only reader who is surprised, on reading the preface, to find Cilliers stating his relative disinterest in chaos theory. Noting that the field was, at least for a while, surrounded by "hype" he contends that the possible contribution of chaos theory to study of complexity would be "extremely limited". The justification for this view is that the sensitivity to initial conditions characteristic of chaotic systems is "not such an important issue" for complex systems, with respect to which the property of being robust, i.e. the "capability to perform in the same way under different conditions" is crucial (CP: ix).

While I fully support Cilliers’ refusal to be side-tracked by the popular and misleading metaphor of the "butterfly effect," though, I am nonetheless perplexed by his wider position. Before explaining why it must be noted that Cilliers is careful to point out that he does not want to sound "too dismissive", and formulates his position as the claim that "chaos theory, and especially the notion of deterministic chaos and universality, does not really help us to understand the dynamics of complex systems." This apparently follows because chaotic behaviour "results from the non-linear interaction of a relatively small number of equations" whereas in complex systems "there are always a huge number of interacting components" (CP: ix).

I’m just not convinced that the distinction between the two fields is as clear or straightforward as these remarks suggest, nor that Cilliers’ own presentation observes such a strict separation. For a start, at various points in his argument Cilliers makes use of concepts and vocabulary which are central to chaos theory, most strikingly in his chapter 6 on "Self-organisation in complex systems" (CP: 89-111). But this should not be surprising if we bear in mind that the study of complex systems and the study of deterministic systems which exhibit stochastic behaviour are connected in various ways. Both are parts of the contemporary process of rethinking classical dynamics and recognising new kinds of dynamical systems and new relationships between determinism and predictability (e.g. Lorenz 1963). Both rely massively on computer models for research purposes. The phenomenon of robustness is central to both fields as well, since research in chaos theory is at least as concerned with stability and structure, for example the long-term coherence of the Red Spot on Jupiter (Marcus 1988), as with instability and disintegration, and especially with the transitions between these regimes. And there are some research problems which seem to belong equally to both areas, such as the study of cellular automata which can be seen as dynamical experiments, fractal generators, exercises in self-organisation or formally equivalent to neural networks, and in some cases all of these (Pines 1988, Stein 1989). Indeed there is at least one sense in which fractal geometry and non-linear dynamics are inseparable in principle, since the phase portraits of non-linear dynamical systems (such as the function now known as the "Lorenz attractor") are themselves fractal objects.

So, contrary to Cilliers’ position, I would suggest that complexity and chaos are not readily separated from one another, and that it is more likely that not that the two fields of chaos theory and the study of complex systems would be mutually illuminating.

(3) Related to the question of chaos is that of the distinction between "complex" and "complicated" drawn early in the book. Cilliers point out, quite correctly it seems, that not all things which are complicated, in the sense of having a large number of parts, say, are in virtue of that also complex (CP: 3f). In this context, of course, "complex" is a term of art specifically referring to systems which have certain specified characteristics, some of them noted above. Since a Jumbo jet, for example, does not have those characteristics it follows that a Jumbo jet may well be complicated but can not be complex.

What concerns me here is not that Jumbo jets fail to be complex "by definition". There is nothing wrong with setting up a contentful set of parameters for calling something "complex" and then using that to point out differences between things which are complex and others which might seem similar but are relevantly different. In any event Cilliers makes clear that whether a given system counts as complex or not depends partly on how we mark off the limits of the part of the world we call the "system" and also on the level of detail at which we consider the area so marked out (CP: 3).

This, though, is precisely what does concern me. Consider the antiquated laptop on which I am typing this article. It seems to be a clear enough example of a merely complicated system which is not also complex. It has a large number of parts performing different functions, but its components interact linearly. The "feedback" loops which occur in it are not of the right kind for complexity since they are neither significantly enhancing nor inhibiting. Partly in consequence of this, even though it has "storage" capacities, it does not have what deserves to be called a history. And it is completely and routinely rule-based to the extent that it could be modelled on a Universal Turing Machine.

For all that, though, it is capable of running software simulating chaotic weather systems or neural networks. In the first case we would have a deterministic simulation of a deterministic system that had unusual dynamical properties. This can show why in real systems with the relevant properties prediction is impossible. But the "chaotic" simulation itself is supposed to be deterministic. In the second case, though, that of the neural network, the result seems somewhat more perplexing, because we end up with a complex system which is in some sense "inside" a merely complicated one.

Its possible that I’m missing something here, and that this isn’t a problem at all. The reason I think that there is a problem is that it seems to me as though there is a real question about whether a given system, even if we allow, as we must, that how the boundary and level of detail for the "system" are to be specified by us, is simple, complicated or complex. But this notion appears to be undercut by the apparent implication that at least some systems which are considered complex for some purposes are at other levels of description or explanation merely complicated. I return to this question in the discussion of post-modern philosophy of science in section (7) below.

(4) Before moving on to some of the wider issues raised and questions provoked by Cillier’s book I want to make a few remarks in what may seem to be a pedantic or perhaps just disastrously unfashionable spirit. They concern some of the references to matters in the history of philosophy which occur from time to time in the course of Complexity and Postmodernism.

Thus, for example, Heraclitus is held up for approval in virtue of the fact that unlike Thales and Anaximenes, who can be seen as "privileging a specific element," he emphasises "strife" (CP: 107) and hence shows himself to have been closer to systemic thinking. The surviving Heraclitus fragments, and references to his thought by other ancient authors (Cilliers’ bibliography includes no entries on pre-Socratic philosophy) indicate that he did indeed give a central place to strife, and that he did not ground phenomena in a single element the way Thales and others seem to have done. Having said that, he was also of the view that the element of Fire was the "archetypical form of matter" (Kirk and Raven 1971: 200), and more importantly that all strife was of the same kind (involving binary oppositions). Finally recall that his pronouncements on these and other matters were supposedly licensed by his direct contact, or as we now say metaphysical ‘presence’, with the Logos (op cit: 187). Even at the level of illustrative metaphor, what we know of the historical Heraclitus puts him a long way from being a likely ally of post-modern agonistics (e.g. Lyotard 1979: 23-4).

Kepler, in contrast, apparently said "to measure is to know" (CP: 35). Again there is no reference to a primary text or source. Nevertheless Kepler’s view is related to his position early in the period during which science (allegedly) allowed consideration of "relationships" only in the qualitative sense, and did not permit them to play a significant role in its quantitative or mathematical work. This claim, though, is just impossible to square with the facts. Kepler, after all, said that "Geometrical reasons are co-eternal with God" (Kepler 1952: 836) and made this statement in the course of an attempt to argue, inter alia, for the greater divinity of straight lines compared to curves, the perfection of regular polyhedra, and the ways in which a sphere was only properly understood when regarded as an image of the Holy Trinity. In Kepler’s neo-Pythagorean case it is simply not possible to clearly separate the mathematical from the relational, or the qualitative from the quantitative, except at the expense of inaccuracy.

At a different scale entirely are some of the assertions made about philosophy as a whole. There are a number of these dotted through Complexity and postmodernism, but one example suffices for illustration. The notion of a "unifying metanarrative" is, we are told, a "dream central to the History of Western metaphysics" (CP: 114). In some philosophies it is admittedly quite simple and diverting sport to find and tag metanarratives. Marx, Hegel and Comte, say, are easy starters, with most philosophy tied to the Christian narrative of creation, fall, incarnation, sacrifice and second coming thrown in for free. But does it really make any sense beyond name-calling to attribute a metanarrative to Descartes, Montaigne, Ockham or Hume (even with his narratives concerning the natural state, and his genealogy of the ethical). What about Aristotle with his never-ending cosmos? Or, for that matter, Parmenides? And does it really help us to understand the historical particularity of what is being defended (complexity, post-modernism) by projecting a broadly nineteenth century notion of historical coherence and linear development (although there are surely exceptions even there) onto the entire history of philosophy?

Perhaps it does in some cases, since the suggested motivation for resisting post-modernity is "nostalgia" (e.g. CP: 114, 130) for Western metaphysics. If Western metaphysics has merits beyond its capacity to inspire nostalgia they are not mentioned. Now it must certainly be the case that nostalgia could play a role sometimes. But it seems somehow uncharitable and dismissive to grant one’s opponents’ position so little philosophical substance. This gesture has the same unfortunate effect as the common ploy by many contemporary analytic philosophers of saying "post-modernist" when they mean "the most idiotic kind of relativist". And the effect in both cases is to reduce the likelihood of open and useful exchange before the other side has even had a chance to speak.

Such, admittedly passing, remarks are neither in the spirit of the rest of Cilliers’ argument, nor his concluding commitment to "openness, provisionality and adventure" (CP: 142). Nor, for that matter, are they called upon to carry a significant argumentative load. Given both of these considerations, though, it seems that they could have been omitted without loss.


As noted above Cillier’s book covers a lot of ground and raises many questions. There are a number of key areas where I either do not share his optimism, or think that there are serious and interesting questions and concerns not directly covered in the argument of Complexity and postmodernism. None of what follows is direct criticism of Complexity and postmodernism. Rather all of it is suggested by that work and offered in the spirit of an ongoing philosophical conversation. In section (5) below I discuss the question of evolutionary explanations of ethics, and some related questions concerning post-modern ethics. In (6) the problem of poorly executed references to natural science by post-modern writers other than Cilliers is raised, which leads in (7) to a discussion of the relationship between post-modern thought and the philosophy of science.

(5) Cilliers is bold enough to suggest that his study of complex systems could be usefully related to problems in ethics. He makes some brief remarks on this point at the end of chapter 6, referring to the work of Axelrod (1984), and suggesting that he can see the a way to answer a "traditional criticism of evolution" (CP: 111) which maintains that the evolutionary accounts are ones which give no significant place to values, in virtue of their all too mechanistic notion of human existence. Cilliers suggests that such a line of criticism is "only possible on the basis of some metaphysical commitments" and that it is "closely related to the position that there is no such thing as postmodern ethics."

I am not convinced that the objection to what is sometimes called "evolutionary ethics" and the claim that there can be no post-modern ethics are really that closely related. One can surely maintain that metanarratives have collapsed, post-modernity dawned, and valence gone poly quite independently of holding any particular views about evolution. Indeed it is prima facie plausible to think of evolution as being one of the collapsed metanarratives. Consider the ways in which medical and social factors have so undermined the purchase of natural selection in developed countries. There it is increasingly the case that life expectancy and prospects of reproductive success have more to do with wealth and less to do with favourable adaptation. Not only that, genetic engineering promises more tricks and fixes by the hour, even further reducing the direct pressure for genetic adaptation by natural selection.

Conversely, it seems as though the view that post-modernism is not congenial to serious treatment of ethics could arise in a person quite confident of the validity of evolution, but either concerned that the imperative to respect local differences of discourse undermines any attempt at non-parochial moral evaluation, or distressed that the notion that the "dropping out of the real" similarly undercuts principled resistance to crimes or atrocities which must be admitted as real for the criticism to be significant. As an example of the first possibility note Rorty’s apparently cheerful admission that he has nothing that he considers morally compelling to say to someone who keeps slaves (1995: 140). And with respect to the second, recall Baudrillard’s contention that the Gulf War did not happen, discussed and criticised at length in Norris (1992).

The objection to the view that opposition to evolutionary ethics is much the same thing as the contention that there can be no post-modern ethics is more than a quibble over details. It is always important precisely to identify the target of an argument or criticism. In this case the lumping together of two separable views does not help the argument, and the additional contention that both are "only possible on the basis of some metaphysical commitments" (CP: 111, emphasis mine) stands in need of justification. It seems clear that even from a broadly naturalist perspective which eschews any attempt at moral metaphysics there are reasons to be cautious about the promises made on behalf of "evolutionary ethics".

Let us consider, then, the suggestion that evolutionary models can show the way to a respectably non-metaphysical account of at least some ethical phenomena. Cilliers’ view is that "values" can be "understood as emergent properties of the social system" (CP: 111) and he gives two illustrative examples.

The first notes that some research into the question of "selfishness" and "altruism" (my words and scare quotes) has shown that "purely selfish behaviour by members of a system is detrimental not only to the system, but also to the particular individuals." So we see, supposedly, that altruistic behaviour can be accounted for as a precondition for the "survival and flourishing of the system" (CP: 111).

The second example draws a lesson from attempts at centralised control and notes that distributed and decentralised control are preferable from the point of view of systemic efficiency and flexibility. Hence we can say that "autocratic management or a fascist government" are not "‘bad’ per se" but instead that "they will ultimately lead to the degeneration of the system in question" (CP: 111).

So, if this is correct, evolutionary analyses of various behavioural strategies show, in the words of Groucho Marx, that "time wounds all heels". Not only that, it seems reasonable to think that this fact may be of some use in attempting to show those behaving in ways we might normally call "selfish" or "dictatorial" that they are not acting rationally, or even wisely from the narrow point of view of their own optimal self interest.

We need to be extremely careful about what we accept here, and about where we might want to draw a line. It is indeed a fact that in many cases behaviour that we would perhaps (at least in our more soft-brained moments of nostalgia for moral metaphysics) call "altruistic", "just" or plain vanilla "good" are also the kinds of behaviour which can be seen to have pragmatic utility from the perspective of the agent, or of the group of which the agent so behaving is a part, and hence, more or less indirectly, to the agent itself. The evolutionary version of this argument is strengthened by noting that many of these forms of "good" behaviour, such as those involving loyalty to blood relatives, are more likely to result in the success of the genes coding for such behavioural dispositions than in success for any particular creature having such genes (e.g. Dawkins 1989: 202-33). I do not propose to deny or contest any of this.

I do want to maintain, though, that it would be a significant mistake to think that these results gave us more than they do. To relinquish the baggage of moral talk referring to "right and wrong", "justice" and so on, in favour of appeals to what "works", or has been shown to pragmatically optimal by its historical success, has consequences which are, at least potentially, surprising.

So we can agree that cuckoos, for example, are not "bad". They do, it must be conceded, reproduce by means of a process which is parasitic on the endeavours of breeding birds from other species which are duped by the cuckoo egg in their nests, a process which further results in disbenefits for themselves and their own young. But we tend, I think correctly, to regard someone who attributes "selfishness" either to individual cuckoos or the entire species as making a kind of mistake, perhaps even an anthropomorphism. From the evolutionary point of view the particular strategy characteristic of cuckoos is no more value laden than the fact that sycamore trees produce propeller-like seeds with the effect of those seeds landing further away from the parent tree than might otherwise be the case. So far so good.

To return to the cuckoo, though, it seems clear that at least one specific result from a particular evolutionary example suggests that extremely "selfish" behaviour can be highly successful in the event that it is carried off with due attention to details. (If evolutionary biology is even broadly correct then cuckoos as we know them would not even exist if such behaviour was not successful.) But if we are trying to explain "ethics" (or "the evolved mechanism formerly known as ethics") by reference to systemic success, then it seems that we would have to say that there is nothing "bad" about, for example, certain transnational corporations manufacturing their products in countries where cheap underage labour is readily available. Considered as systems these corporations are highly adaptive and successful. Furthermore, as long as the products meet the standards expected by the more wealthy parts of the world where they are actually sold and spectacularly advertised, they will remain successful.

The point here is simply that the attempt to recast the moral in terms of the pragmatic or the effective is permanently in danger of foundering on a mis-match between what is pragmatic or effective and what we might normally call good. At this point, of course, there might be some who would be willing to bite the bullet. If you are extremely impressed with the credentials of evolutionary explanations, and suspicious of what does not fit into such schemes then categorical moral talk might be just as well lost as geocentrism or palmistry. Those who would rather not follow that line might feel that the most an evolutionary and engineering approach based on efficiency can show is that from the perspective of selfish attempts to maximise utility, it is rational to co-operate or be "altruistic" in cases where one could be held accountable for not co-operating or suffer some other consequences for one’s selfishness. But, the objection goes, why should we think that this is the same thing as saying what is at issue in referring at something as "good", "right" or "just"?

Now it might be suggested at this point that my line of objection presupposes just that kind of hoary distinction between the "real" thing and the ubiquitous "simulation" which identifies me as no more than a nostalgic modernist unable to cope with the excitement and unpredictability of the post-modern condition. This seems like a very poor answer, though. I am expressly not making a pitch for some kind of metaphysical essentialism about "the good" here, any more than Lyotard is when distinguishing a litigation from a wrong (1988: xi). Even so, it seems as though what needs to be explained, if it is explaining the ethical, or the possibility of the ethical, that we want to do, is that there is something characteristically categorical about ethical statements, and that this fact does not change no matter how deeply we might be committed to naturalism.

Some recognition of this very point seems to be at work in Cilliers’ own proposal, later on in Complexity and Postmodernism, that post-modern ethics could involve moral rules deployed in what Derrida and Cornell call a "quasi-transcendental" (CP: 139) fashion, which does not involve calculation, and where the rules are re-legitimated in their actual use. It is worth pointing out in this regard that the attack on ethics based on "calculation" seems to apply more clearly to consequentialist systems such as classical utilitarianism, and by no means applies to all traditional ethical thought. Kant’s ethics, for example, was not especially concerned with calculation, and the application of the categorical imperative was explicitly tied to "the same time" as the action being chosen (Paton 1948). Kant’s thought is tremendously important in Lyotard’s own thinking on ethics as well (see especially Lyotard 1988), and not in ways which fit at all well with the notion of the "ethical" as translatable into the vocabulary of systemic efficiency. Indeed Lyotard states in more than one place in The Postmodern Condition (1979: 8, 34-5) that the drive for efficiency can have totalitarian consequences. Furthermore his opposition to this consequence is motivated by more than the concern that there are better ways of being efficient than totalitarian ones. (It should be clear, by the way, Lyotard’s work is but one of many examples showing the idiocy of the view that post-modern positions have no place for ethics, on which point Cilliers is entirely correct.)

But if we take the richer side to Cilliers’ argument seriously, which it clearly deserves, then it seems as though there is an interesting and provocative gap in that argument. Something is needed to connect, on the one hand, the undeniable value of evolutionary studies in illuminating plausible pragmatic sources for some kinds of behaviour, and, on the other, the more rich ("quasi-Kantian" perhaps?) notion of the ethical which he discusses at other points. Something is also needed to resolve the tension between the hostility to ethics based on calculation just noted, and the earlier endorsement of the evolutionary studies which derive most of their force from being paradigmatic examples of algorithmic strategies intended to maximise some quantifiable form of utility. I see no a priori reason why such bridges could not be constructed.

(6) A striking feature of Cilliers’ book, noted above, is that the argument shows a considerable command of both post-modernism and the scientific fields under discussion. This happy state of affairs is all too unusual. Indeed it is equally striking that so few of those who are apparently serious about their post-modernism are especially careful about their science. The Sokal hoax has popularised an especially nasty manner of reading utterances on logic, mathematics and natural science by made post-modern theorists. Even if we reject the tone and spirit of that particular intervention, though, it is important to acknowledge the fact that the heroes of post-modern thinking have indeed been prone to making wildly uninformed and unjustifiable statements about the natural and formal sciences.

Cilliers himself is cautious about the practical prospects for fruitful dialogue between post-modernists and scientists despite the fact that his own book amply demonstrates the existence of exciting opportunities. He notes, for example, that "[p]ost-structuralism (deconstruction) is often presented in anti-scientific terminology that stresses the proliferation of meaning, the breaking down of existing hierarchies, the shortcomings of logic, and the failures of analytical approaches," and admits that one "cannot blame scientists for being sceptical" about a body of thought which tends to present itself in this way (CP: 22). Although correct, I think that these remarks are far too diplomatic, and make the problem seem smaller than it is. It is not merely that the "post-modern style" is alien and perhaps even anathemic to the "scientific style". More than that, a great deal of what is said by post-modern authors on scientific topics is opportunistic, dangerously speculative or plain wrong.

Sokal and Bricmont (1998) offer an often superficial and uncharitable analysis of aspects of this tendency. Unfortunately their grasp of post-modernism, and philosophy more widely, is at least as poor as the understanding of science which they criticise (see especially their dogmatic and unsubtle chapter on the philosophy of science, 1998: 49-95). Their book is saved from being completely useless by the extensive catalogue of pseudo-scientific pronouncements by people who should have known better which they have assembled. And for one collecting such prizes there is truly an embarrassment of riches to be found. Two examples will have to suffice for now. Thus Irigaray:

Is E=Mc2 a sexed equation? Perhaps it is. Let us make the hypothesis that it is insofar as it privileges the speed of light over other speeds that are vitally necessary to us. (Irigaray 1987: 110)

And Virilio:

This drifting of figures and geometric figuring, this irruption of dimensions and transcendental mathematics, leads us to the promised surrealist peaks of scientific theory, peaks that culminate in Gödel’s theorem: the existential proof, a method that mathematically proves the existence of an object without producing that object. (Virilio 1991: 66)

In a discussion of relativity it would not usually be necessary to point out that the speed of light is not "privileged" over any other speed, or to explain that the speed of light in a vacuum was not chosen to appear in the equation after the speed of sound in air, say, or the top speed of a Model T Ford had been rejected for not being masculine enough. Similarly it would normally not be necessary to draw attention to the fact that Gödel’s proof is certainly not existential (in the technical sense applying to proofs) and hardly surrealist. Furthermore Gödel (see Nagel and Newman 1958) does the opposite of what Virilio suggests: he "produces" an object which entails a self-referential paradox, but which defeats the particular ideal of proof which presupposes that the true statements of a formal system are all either axioms or members of a consistent set of provable theorems. Unfortunately such mathematical and scientific corrections and objections are required all too often in the face of "post-modern" writing on all manner of topics.

It must be said at once that at least two of the great "post-modern" thinkers emerge relatively well from all this. On the one hand there is Derrida, who says very little about technical aspects of mathematics and science, which are almost never of direct reference to the arguments he is developing. And on the other is Lyotard, who seems in comparison relatively well informed. Even Derrida, though, makes occasional unhelpful gestures at Gödel (e.g. Derrida 1978: 251-77) and allowed himself to make some especially impenetrable and misguided remarks about Einstein in the discussion following Structure, Sign and Play (Derrida 1970: 265-7). And, although better informed than many post-modern philosophers, Lyotard’s discussion of "postmodern science as the search for instabilities" in his The Postmodern Condition stands up rather poorly in the face of detailed criticism, a point which I have discussed at length elsewhere (Spurrett, forthcoming ).

It could be asked at this point just what kind of "problem" this is. After all political philosophers, sociologists and dromologists are not required to be scientific specialists. Nor, in most cases, are philosophers. The reason that there is a problem, though, is that these references to the natural and mathematical sciences by post-modern thinkers are usually supposed to carry some argumentative weight. Perhaps some aspect of a "hard" science is supposed to illustrate or be an example of an important point, which may be what Virilio is getting at in his remark about Gödel. More significantly a critical point might be in the offing, as is at least supposedly the case with the statement by Irigaray about relativity quoted above. And perhaps most seriously of all the post-modern argument being developed might actually rely, at least in part, on scientific claims, as I contend is the case with Lyotard’s analysis of contemporary science, which he says shows that today science is legitimated "solely on paralogy" (1979: 97). All of these kinds of argument are important, but if they are to work then it cannot be seen as an optional extra to get things right.

To the extent that the problem remains widespread we stand to lose out on at least some of the exciting opportunity represented by the kind of work Cilliers has shown is possible. It may be an important "post-modern" lesson that playfulness and seriousness are entirely compatible. But to be playful with science is not an alternative to being accurate.

There are those who might respond that such criticisms presuppose a distinction between "narrative" and "scientific" utterances which is no longer appropriate. That, though, would be indefensible. Without being guilty of any obvious slide into metaphysical madness or the viewing of philosophy as riding shotgun for natural science, it is possible to urge an important difference between scientific and non-scientific claims. Science is indeed political in nature, and is an appropriate object for social scientific study. But it is not "just" political, and that is why it is so clearly idiotic to treat it as though it is, a way of thinking which leads to excesses such as Feyerabend’s protest that modern societies are Copernican by means of scientific coercion since the question whether the Earth orbits the Sun was never put to a vote or referendum (1978: 74), or Irigaray’s reading of the "sexed" nature of the mass-energy equation quoted above. Now, though, we are no longer talking about the pitfalls of inaccurate scientific statements by post-modern writers, but the more interesting question of how all of this relates to the philosophy of science.

(7) The first thing which needs to be noted here is that there is not really any such thing as the "post-modern philosophy of science". There are at least two reasons for this. On the one hand much of the traditional content of philosophy of science (with its debates over rationality, progress, verification, realism and the like) does not represent the kind of problem which a post-modernist would be interested in solving, as much as it represents a model of philosophical activity rooted in modernity, and which is largely no longer recognised as legitimate. On the other hand those remarks by post-modern thinkers which relate to philosophy of science, whether explicitly or implicitly, are considerably varied and do not add up to a single body of thinking. This is not the place to attempt an account of those different tendencies and streams, so for the purposes of brevity I will concentrate on the implicit philosophy of science to be found in Lyotard’s The Postmodern Condition, since that work plays such a significant role in Cilliers’ argument.

What I am calling Lyotard’s "philosophy of science" (he does not use the term himself) is his "method of language games" (1979: 20). This is a strategy for engaging with science based on paying close attention to the discursive practices of those called "scientists" and especially, at least for the purposes of the argument of The Postmodern Condition, to those practices which relate to questions of legitimation. ("Legitimation" in this context refers roughly to the processes by means of which statements are accepted into or excluded from some or other discourse, on the basis of who makes them or how as much as for reasons relating to what is said.)

This approach yields some interesting returns, most notably in my view the analysis of the pragmatics of scientific teaching and research (1979: 43-48) and their role in the ongoing re-creation of the collection of recognised experts which is needed for the institutions by which science is authenticated (peer review, for example). For all that, though, this "method" has serious limitations as well. The most striking example of this is the analysis of "postmodern science as the search for instabilities" in section 13 of The Postmodern Condition. Lyotard’s survey of the contemporary scientific scene can be criticised by reference to questions of the accuracy of his specific scientific citations and the tenability of his interpretations of the theories and results he refers to. Rather than focus on that here, though, I want to ask whether his conclusion, noted above, that post-modern science is "legitimated solely on paralogy" could possibly be defended on the basis of an analysis of scientific "language games".

According to Lyotard, under conditions of modernity science normally legitimates itself by performance but at other times, when it "does not restrict itself to stating useful regularities and seeks the truth" it is legitimated by a "discourse called philosophy" which generates metanarratives containing science (1979: 7). If this is all that is at stake in the legitimation of science then it seems perfectly reasonable to imagine that we can decide questions of the status of science by reference to what amounts to the sociology of science. But "stating useful regularities" is surely too positivistic a notion of science, and in any event has far more to do with a perspective in the traditional philosophy of science than with most of science itself. Furthermore, the suggestion that philosophy legitimates science (if that is what it does at all) by means of metanarratives is too vague and general to be of much use at all.

What is missing is any notion of what it is about science in virtue of which it does more than simply state regularities, and which is distinctly different from being situated in a grand unifying metanarrative. Without attempting to deny the truth that science is political, we need to ask what it is that makes it science, rather than any other social activity with a political aspect. Such a distinction is missing in Baudrillard’s assertions concerning the dropping out of the real, or, in a phrase which Cilliers quotes, (CP: 84) the "liquidation of all referentials", and equally in Feyerabend’s complaint about Copernicanism noted above. There are all manner of things we could decide to do about what people say and do concerning the motion of the Earth around the Sun. We might decide to burn their books, lock them up, threaten them with torture and so on. But we surely cannot decide by means of any political process whatsoever whether or not the Earth actually is the unmoving centre of the universe.This is a distinction simply not captured by the emphasis on what is said and done, when and by whom.

Cilliers certainly does not endorse the kind of view suggested by Feyerabend’s remarks on Copernicanism, and rejects the specific suggestion that in post-modern science "anything goes" (CP: 115). He does, though, find some use in Feyerabend’s notion of "anarchy" by means of an analogy with Lyotard’s paralogy (CP: 118). Lyotard does not explicitly say that "anything goes" either, although The Postmodern Condition includes some favourable references to Feyerabend, and the crucial appeal to paralogy as the strategy for legitimating science in a post-modern world. If there is an important distinction between legitimate paralogy and a dispensation where "anything goes" it is not clear what it might be. Paralogy is a process modelled on both the avant garde in art and, in turn, on the Kantian notion of the sublime, where some system or other is confronted with that which cannot be represented or interpreted with its categorical and expressive resources (See Kant 1987: 103f, Lyotard 1988: 161-71). On the other hand living in the spirit of "anything goes" means, for Feyerabend at least, "to step outside the conceptual circle and either to invent a new conceptual system, for example a new theory, that clashes with the most carefully established observational results and confounds the most plausible theoretical principles, or to import such a system from outside science, from religion, from mythology, from the ideas of incompetents, or the ramblings of madmen" (1993: 52-3). Cilliers’ own account of the project of scientific "anarchy" attributes a constructive intention, involving finding "meaningful relationships among the different discourses" (CP: 118) which is at least difficult to square with the clearly competitive and confrontational mood suggested by Feyerabend, who at times suggested that his approach might better be called "Dadaism" than "anarchism" (Feyerabend 1993: Chapter 1).

Here too, in order to distinguish the merely capricious from the novel and relevant (which may, of course, only be possible some time after the fact) we need some distinction between the content of science and its social aspects. Such a distinction need not be an instance of the kind of metaphysical thinking from the bad old days. Nor need it take the form of a fully worked out theory of the distinction which has to be developed a priori. All that is important is that it be acknowledged as a regulative principle with some significant force.

Furthermore, to the extent that post-modern thinking holds out the prospect of significant and novel criticism of scientific practice on philosophical, social and political grounds, it needs to keep this distinction far more than it needs to subvert it. What makes it possible to criticise some science for socially motivated reasons is precisely the fact that its conclusions outreach its findings in ways which we can make sense of in terms of ideology, prejudice or some other interest. As Dupré puts it "The only way to put the genuine insights of the sociology of knowledge programme, not to mention a good deal of orthodox history of science, into a proper perspective is to see the kinds of forces they describe as interacting with a real and sometimes recalcitrant world" (1995: 12).

Indeed, such kinds of distinction are surely presupposed, at least implicitly, by Cilliers when he argues for the limitations and inferiority of reductionistic and analytic approaches to complexity, of representational approaches to thought, of algorithmic approaches to language and metaphysical essentialism concerning ethics. Similarly when he urges the relative superiority of approaches to all of these problems which are sensitive to complexity, whether arising from connectionist research or post-structuralist/post-modern thought. But then, to return to a point I raised in section (3) above, it seems as though to support these claims we need something stronger than we allow ourselves if we are too modest, and limit the force of our claims to their being contingent features of our ways of describing. Even if I am wrong about the specific point regarding the distinction between complicated and complex, I would still urge the need for a more substantial post-modern philosophy of science.

If there is to be a significant post-modern philosophy of science, one which can make a defence of this kind of claim without falling into positivism or naive realism on the one hand, or undercutting itself by excessive and self-defeating relativism on the other, then a suggestion of its outline may better be found in Derrida, rather than Feyerabend, Lyotard or Baudrillard. For in White Mythology he notes the need to "account for the specific divisions that epistemology cannot overlook, the divisions between what it calls metaphoric effects and scientific effects" (Derrida 1982: 263, emphasis added).


(8) As noted several times now Complexity and postmodernism is a wide-ranging and provocative book. For all the criticisms and reservations I have raised here, that fact really is the one with which to conclude. There are reasons to be cautious about some points in Cilliers’ argument, and to be sceptical about the possibility of further significant work combining detailed scientific content with post-modern theoretical approaches. Nonetheless, I hope it is clear that this kind of poly-disciplinary work has a great deal to recommend it. Not only that, Complexity and postmodernism is not entirely alone. Related research in fields such as evolutionary robotics, even though growing out of initially analytic and reductionistic approaches, has found that many of the metaphysical assumptions of modernity simply fail to fit the facts of what is involved in navigating through the world. Those sympathetic to post-modernism, or to "Continental" philosophy more widely will not be surprised to find that at least some of such researchers have found congenial ways of thingking in Heidegger’s phenomenology, as Being There, the title of Clark’s (1997) book shows.

Finally, the areas where I have urged caution are all areas where further exciting philosophical work is possible, and where the contours of such work are suggested rather than excluded by what Cilliers does say, and by the arguments he advances in his timely and original book.


Axelrod, R. (1984) The Evolution of Cooperation, New York: Basic Books.

Cilliers, P. (1998) Complexity and Postmodernism: Understanding complex systems, London: Routledge.

Clark, A. (1997) Being There: Putting Brain, Body and World Together Again, Cambridge, Mass. and London: MIT press and Bradford Books.

Dawkins, R. (1989) The Selfish Gene (2nd edition), Oxford: Oxford University Press.

Dennett, D. (1995) Darwin’s Dangerous Idea: Evolution and the Meanings of Life, London: Penguin.

Derrida, J. (1970) "Structure, Sign and Play in the Discourse of the Human Sciences", In Macksey and Donato (1970).

Derrida, J. (1977) "Signature Event Context", Glyph, vol. 1 (Baltimore: Johns Hopkins University Press), pp 172-97.

Derrida, J. (1978) Writing and Difference, (translated by Bass, A.), London: Routledge and Kegan Paul.

Derrida, J. (1982) Margins of Philosophy, (translated by Bass, A.), Chicago: University of Chicago Press.

Descartes, R. (1985) The Philosophical Writings of Descartes, 2 vols., (translated and edited by Cottingham, J. G., Stoothoff, R., and Murdoch, D.), Cambridge: Cambridge University Press.

Dupré, J. (1995) The Disorder of Things: Metaphysical Foundations of the Disunity of Science, Cambridge, Mass: Harvard University Press.

Feyerabend, P. (1978) Science in a Free Society, London: Verso

Feyerabend, P. (1993) Against Method (3rd edition), London: Verso.

Fuller, S. (1988) Social Epistemology, Bloomington: Indiana University Press.

Goodman, R. B. editor (1995) Pragmatism: A Contemporary Reader, New York: Routledge.

Irigaray, L. (1987) "Is the subject of science sexed?" (translated by Bové, C. M.), Hypatia 2(3), 65-87.

Kant, I. (1987) Critique of Judgement, (translated by Pluhar, W.), Indianapolis: Hackett.

Kepler, J. (1952) Epitome of Copernican Astronomy, Books IV & V, and The Harmonies of the World, (translated by Wallis, C. G. in Great Books of the Western World, edited by Hutchins, R. M., vol. 16: Ptolemy, Copernicus, Kepler), New York: Encyclopaedia Britannica.

Kirk, G. S. and Raven, J. E. (1971) The Presocratic Philosophers: A Critical History with a Selection of Texts, Cambridge: Cambridge University Press.

Lorenz, E. N. (1963) "Deterministic Nonperiodic Flow." Journal of the Atmospheric Sciences (20) 1963, 130-141.

Lyotard, J-F. L. (1979) La Condition Postmoderne, Paris: Les Éditions de Minuit.

Lyotard, J-F. L. (1984) The Postmodern Condition: A Report on Knowledge, (translated by Bennington, G and Massumi, B.), Manchester: Manchester University Press.

Lyotard, J-F. L. (1988) The Differend: Phrases in Dispute, (translated by Georges Van Den Abbeele), Minneapolis: University of Minnesota Press.

Macksey, R. and Donato, E. editors (1970) The Languages of Criticism and the Sciences of Man: The Structuralist Controversy, Baltimore: Johns Hopkins University Press.

Marcus, P. (1985) "Numerical Simulation of Jupiter’s Great Red Spot", Nature, 331, pp 693-696.

Nagel, E. and Newman, J. R. (1958) Gödel’s Proof, New York: New York University Press.

Norris, C. (1992) Uncritical Theory: Postmodernism, Intellectuals and the Gulf War, London: Lawrence and Wishart.

Norris, C. (1997) Against Relativism: Philosophy of Science, Deconstruction and Critical Theory, Oxford: Blackwell.

Paton, H. J. (1948) The Moral Law, (A translation of Kant’s Groundwork to the Metaphysic of Morals), London: Hutchinson.

Pines, D. editor (1988) Emerging Synthesis in Science: Proceedings of the Founding Workshops of the Santa Fe Institute, Santa Fe, New Mexico, New York: Addison-Wesley.

Rorty, R. (1995) "Feminism and Pragmatism" in Goodman (1995: 125-148).

Sokal, A. and Bricmont, J. (1998) Intellectual Impostures, London: Profile Books.

Spurrett, D. "Lyotard and the Postmodern Misunderstanding of Physics", Theoria, (forthcoming).

Stein, D. L. editor (1989) Lectures in the Sciences of Complexity: The Proceedings of the 1988 Summer School held June-July 1988 in Santa Fe, New Mexico, New York: Addison-Wesley.

Thelen, E. and Smith, L. B. (1993) A Dynamic Systems Approach to the Development of Cognition and Action, Cambridge, Mass.: MIT Press.

Virilio, P. (1991) The Lost Dimension, (translated by Daniel Moshenberg), New York: Semiotext(e).

Wheeler, M. and Clark, A. "Genic Representation: Reconciling Content and Causal Complexity." British Journal for the Philosophy of Science (forthcoming).