Image 01

Varela, Thompson and Rosch vs. (?!) Noë and O’Regan

August 23rd, 2011 by Mike Beaton

I’ve finally got round to working through Varela, Thompson and Rosch, The Embodied Mind (VTR) in detail for the first time.

I find it interesting that Ezequiel has said that he finds Noë and O’Regan’s brand of enactivism (N&O) to be interesting, but in need of more theory. Whereas my reaction, on reading VTR, is similar, but opposite!

I find talk of ‘structural coupling’, ‘operational closure’, etc. to be intriguing, but not natural to me, so I have very little idea how to make it mathematical, and useful; whereas talk of ‘perception as action’, in the N&O sense, is very natural to me, and I do have ideas about how to make it mathematical and useful – indeed, making that aspect of enaction mathematical seems to be the main thrust of Kevin O’Regan’s research work at the moment.

I’ve (just!) said this to Ezequiel and he says that part of the problem is that VTR is mainly a manifesto, and that I should look elsewhere “say, in the BBS paper on the enactive theory of colours, on later work on life and mind continuity, the work on long-range neural synchrony in perception and epilepsy, etc” for more theory.

But I think Ezequiel still downplays the significance of N&O’s variety of enactivism. The question here is about how broad ranging these two, self-styled brands of enactivism are. VTR are certainly trying to link (at least) sciences of mind, evolutionary and developmental biology and mindfulness practices. But how broad ranging is N&O’s work? I hope it is OK to quote Ezequiel, in an off the cuff comment, as saying that it is “an account of sensorimotor factors influencing perceptual experience”. For me, it is much more than that. It is an account of the essential nature of perceptual experience, from both the first and third person. (I tried to say something about how one account can be both first- and third-person in my thesis, and in Qualia and Introspection in JCS.)

I do think a lot of what N&O and VTR say is compatible – and the spirit of what they both say perhaps even more so. But there are points where they seem to me to disagree quite fundamentally. One such is hinted at in the Summer School discussion question I settled on (‘what does experience correspond to?’). Others might be disagreements over the import of J.J. Gibson’s work; or over the relation between experience and understanding; or possibly even over whether there is a self – of course not construed as an essential, intrinsic thing, but at least construed as a process.

10 Responses to “Varela, Thompson and Rosch vs. (?!) Noë and O’Regan”

  1. Hi Mike,

    Thanks for these thoughts and comments. I look forward to talking more at the meeting.

    Some historical context might be useful here. Varela and I started writing The Embodied Mind in 1986 and the finished manuscript was delivered in 1990. Kevin and Alva’s BBS article appeared in 2001. So there’s a decade between the two works, a decade that saw huge expansion in embodied cognition approaches. When we started work on TEM, the main discussions were over the relative merits of classical computationalism and connectionism; embodied approaches were just starting to show up. Also, there was no discussion of consciousness; on the contrary, consciousness was considered basically irrelevant to understanding the mind. Our main aim in TEM was to propose a different orientation for the field, one that brought the body and lived experience into the foreground of attention. So, we weren’t concerned with presenting a specific research program for any one phenomenon, as Kevin and Alva did for perception (though, in different ways, each one has also worked on ways to extend sensorimotor theories beyond perception).

    So what, then, are is difference between enaction (in the Varelian sense of the term) and the sensorimotor contingency theory? I’ll restrict myself to Kevin and Alva’s BBS paper, because Kevin and Alva have gone in somewhat different directions since then. The crucial difference (which I spend some time talking about in my 2007 book Mind in Life) is that enaction is grounded on giving an account of cognitive agency in terms of the theory of autonomous systems (see recent papers by Ezequiel, Barandiaran, Rhode, Froese, and others). Without this theoretical grounding, I argue, it’s not possible to give a proper account of sensorimotor know-how, because one winds up treating sensorimotor systems from a heteronomous (external control) perspective. From my perspective, the sensorimotor contingency theory is defective to the extent that it neglects the framework of autonomous systems (again, see Mind in Life for the argument). Kevin disagrees — for example, he’s happy to use a missile guidance system as an example of a system with sensorimotor contingencies, whereas for me such a system, because it’s not autonomous (it doesn’t self-organize its own sensors and effectors from within, so the SM contingencies aren’t autonomously generated and directed) doesn’t count as a system with any genuine or original sensorimotor know-how.

    There are other differences too — some of which you mention in your post — and that we’ll probably have occasion to discuss during the week. But from a nitty gritty theory and modeling perspective, I’d say it’s the issue about autonomy that’s the crucial difference.

    • Mike Beaton Mike Beaton says:

      Dear Evan,

      Thank you very much for the comments, and for the historical perspective.

      One quick side point: I personally have never found the issue about whether or not consciousness is explicitly in play too concerning, as long as any works being compared are discussing ‘mind’. But that is because I like to argue that (qualia) consciousness is the same thing as experience (in some more general, agent-world interaction sense), and that both are the same thing as ‘cognition’ (when considered as an activity of the agent, NOT as brain information processing), or at least ‘sense making’.** (I am sure Alva, at least, would be a fan of equating experience with cognition/sense making/understanding – though perhaps not of both these with qualia consciousness.)

      On the grounding issue – this helps considerably in seeing what the key issue is that you (and perhaps Ezequiel?) find to be fundamentally missing from Noë and O’Regan’s brand of enactivism.

      But would I be right to construe the theoretical work concerning ‘operational closure’, autopoiesis, and so on, as the search for a grounding for action in non-action terms? Or is that over-simple? Or plain wrong?!

      Thank you for the response, and for the opportunity to ask some of these questions!


      ** That said, there were are couple of points in TEM where I thought I read you as addressing the issue of (at least roughly) qualia consciousness, and suggesting that colour conscious experience corresponds to particular, high-level emergents in the internal activities of an operationally closed system. I wonder if I was right to take it this way?

      • Hi Mike,

        Lots to talk about here.

        I approach the whole issue of consciousness very much from a phenomenological (in the big P Husserl/Merleau-Ponty sense) perspective, which is to say, the important issue is to understand the kinds of intentionalities (mental directedness or modes of sense-making) proper to modes of experience such as perception, memory, empathy, thought, and so on. It’s precisely this conception of consciousness that was missing from old-school cog sci but that now has emerged much more into prominence in recent years.

        I see Kevin and Alva as mainly dealing with perception in this context, but not so much other modes of experience (or sense-making if you like).

        On the matter of qualia, ‘qualia’ is a term I tend to avoid because it’s often theoretically loaded in ways I find problematic (e.g., as qualitative properties intrinsic to internal states decoupled from action or sense-making, etc.). But setting that aside, I’d say I’m very much in sympathy with Kevin and Alva’s approach to understanding qualitative aspects of perception in terms of dynamic sensorimotor activity (here I’m also thinking of Alva’s paper on neural plasticity with Susan Hurley).

        One point where I wouldn’t agree with Kevin and Alva (I don’t think Alva holds this anymore) is that all perception is attentive perception and all consciousness is transitive (object-directed) consciousness. In Mind in Life, I follow phenomenologists in arguing that transitive consciousness of an object also necessarily involves intransitive self-awareness (of the subject as agent and experiencer).

        I’m happy to equate experience and sense-making, although there are some difficult questions here: At one end, does bacterial sense-making count as experiential? At another end, are there forms of objectless awareness or “pure consciousness” (as spoken of in meditative traditions) where sense-making is no longer in play? Tough questions that seem open to me.

        On grounding, the concepts of autopoiesis, autonomy, closure, etc., can be seen as general system-theoretic ways of thinking about requirements for grounded cognition, but I wouldn’t say that they aim to ground action in non-action terms. I’d say they’re attempts to understand the organizational conditions for something to be a sense-making system, but they have no concrete explanatory purchase apart from action contexts.

        On colour, here my view is a relational one: I think being coloured is a property of the environment taken in relation to the perceiver, so I wouldn’t identify colours with emergent internal states. My book Colour Vision (1995) goes into that in much more detail than either TEM or the 1992 BBS article on colour vision with Varela.


  2. I’m happy to see the discussion starting! Thanks Mike and Evan.

    I don’t have much to add, but since Mike mentions our conversation I thought I’d say a few words.

    I largely concur with Evan on the differences between sensorimotor and enactive (life&mind) approaches. The issue (Mike’s initial question) should revolve about what counts as a theoretically-loaded account of X (perception, action, mind, consciousness, autonomy, etc.) In my opinion, that would have to be an account that doesn’t leave yawning gaps. Now, it may be incomplete, hazy, but if it is workable and you can “see ahead” towards feasible explanatory possibilities, that will be the account I’d prefer. Even just attempting to address the issues, giving them their due relevance, is preferable than not doing it, even if your attempt doesn’t quite succeed.

    I think that the sensorimotor account should be most fruitfully understood as focusing on certain aspects of transitive experience and their dependence on an active body in the world. There are several interesting proposals in this account and it could be worked out into new empirical tests and models and so on. Hence, I agree that it is not theoretically-poor at all. But, the enactive, life&mind approach (which could easily incorporate many aspects of the sensorimotor account; so it is not right to see the options – after perhaps enough delimitation – as strictly opposed) aims at providing a radical answer to questions of autonomy, agency and self-affection (and I would add my recent interest in the concepts of activity and passivity of the body). These answers are definitely not hand-wavy, even if work is still ongoing; they are based on systemic concepts and, as you know, some people (at least I) think that the result is a radically non-functionalist approach to the mind.

    The sensorimotor account, like cognitivist accounts before it, can be criticised for not having an concept of autonomy, agency, or actually (being a bit bold here :), not even of action or perception! The enactive approach attempts to provide answers to these fundamental questions. I think that there would be a lot of agreement here. I’ve made this point to O’Regan once and he sort of agreed that questions such as how is experience integrated, what makes it meaningful to you, etc. are not resolved in the sensorimotor approach and something like a concept of a “self” would be necessary. I read this as: bring on the idea of autonomy and sense-making.

    Now, I would say that for these reasons the enactive approach is also a theoretically-rich option, one actually with a more ambitious scope.

    (On the point made by Evan that in the SM approach one could well consider a guided missile as enacting sensorimotor contingencies, I suggest an excellent essay by Hans Jonas that can shed light on this question: Cybernetics and Purpose: A critique – it appears in The Phenomenon of Life, but also published in a journal; I can try to make it available here if people are interested).

    (On the other point raised by Evan about transitive and non-transitive experience; I’ve been working on linking the concept of self-affection in phenomenology and the concept of autonomy in enaction – this is ongoing work, but actually quite exciting).

  3. Mike Beaton Mike Beaton says:

    OK, I’ll try to do a short post this time, and to stay focussed!

    I certainly agree, Ezequiel, with your comments about what makes a good research programme. I think the disagreement is about how well the two approaches fare, according to criteria which we share.

    I think it’s worth emphasizing that we are discussing here, amongst friends, the merits and demerits of two related approaches, both of which we agree are much better than the standard alternatives.

    Now, I know that many people (a lot of them ex-Sussex) have done a lot of work to flesh out autonomy in a theoretical way. And it is without doubt important work, come what may. But I have serious (and theoretically based) doubts about whether this research programme is “workable” in one important sense (which we would both agree matters): I am unsure whether this approach *can* be completed, in the way required to augment N&O’s enactive perception with what the VTR approach says is missing.

    That is because I do indeed think that there is an answer to what autonomy is. And I do agree that the sensorimotor approach needs such an answer to be complete. But I think that the answer lies in an interpretationist view of ‘rationality’/sense making/practical understanding. Thus, for instance, an amoeba has mind (/consciousness/agent-level cognition/autonomy) to the extent that we (can reasonably) view it that way.

    A standard aspect of such interpretationist views (which I believe is correct) is the claim that the levels of interpretation are self-standing: they cannot be grounded in anything simpler; in this case, either you already know what mind is, or you don’t. (Of course, we do!) You can certainly flesh out the terms; you can examine how they relate to each other; but you can’t ground them in anything else. Therefore, I strongly suspect, an attempt to ground embodied cognition in something simpler – autonomy – is a chimera. Autonomy (in the relevant, loaded, sense) is just another word *within* the charmed circle: you know what it is if you know what understanding is, and you don’t if you don’t. Crucially, the nature of autonomy can’t be grounded in *operational* concepts like autopoiesis. Which is to say, any theory of what autopoiesis (or similar) is – which is meant to ground cognition – will either miss the point, or will sneak in mind-level concepts in order not to miss the point.

    If I wanted to be controversial, and to attempt to prompt someone else to chip in… then I would say that the person I believe I know of, from within the VTR ‘side’ of things, who is closest to agreeing with me on this, is Xabier…!

    PS Please understand, as I know such things are rightly very important to you, Ezequiel, that the above is *not* just a claim that “autopoiesis is wrong”; it is a claim that another, different, positive research programme – that of elucidating the (self-standing) nature of practical, embodied rationality (but *not* grounding it in something else) – is what is needed, either instead, or as well. (If you want to know how I view the incomplete, hazy, but workable next steps in this research programme, ask me about Edwin Jaynes…)

    • I don’t think we disagree about “what needs to be done”, but I’ve got the suspicion that we may be reading or writing the same sentences and making different sense out of them.

      I will (really!) be short and someone else might chip in. But it all seems to hinge on what is meant by terms like “grounding” or “self-standing”. I interpret “grounding” not as a reduction that explains, or explains away, all the phenomena at a certain given level (say the rationality/normativity of actions). But precisely as a way of explaining how such self-standing levels get off the ground and what are the requirements that enable them, and how it happens that the grounding conditions are not fully determining of the new levels. I think the notion of autonomy is precisely so powerful because it allows us to query whether autonomous organisations exist at different levels, and if we find them, we can often speak of them as self-standing in some sense, which is not meant as “independent” of its context, processes and conditions.

      I wonder if you’re not interpreting “self-standing” as “ungroundable” and if you do, then can we still have a naturalistic approach to the mind?

      What do Xabier and others say? :)

    • Miriam Kyselo Miriam Kyselo says:


      Thanks for your comments – it’s great that the discussion has set off and that it is actually so close to my own work. I generally agree with Ezequiel’s and Evan’s stance on the issue in that I do consider the notion of autonomy to be a promising concept for grounding cognition and that there are ways to integrate the SM approach with Enactivism (in line with VRT and the “Sussex-school”). But I also think that adding the concept of autonomy and including consciousness proper is not enough.
      I would like to add two thoughts. The first is concerned with Mike’s general doubt about autonomy as a useful concept to ground cognition. The second is dealing with his initial question whether and to which extent VTR and Noe and O’Regan are compatible.

      I believe that the discussion between Ezequiel and Mike implies two very different readings of the idea “cognition should be grounded in autonomy”.
      According to the first, we are considering cognition at an already complex level, say the “mind-level” or the level of “practical, embodied rationality” (as Mike has it) and then we ask whether these sophisticated cognitive processes are grounded in autonomy. We are looking at the mind from a developmental perspective and argue that at some stage autonomy is at play. Crucially, mind and autonomy appear to be two different things. Such a perspective could be based on the implicit assumption of a cognitive gap, as it has been recently discussed by Hanne and Tom. While Enactivism might establish a good alternative to classical representational views on cognition for low-level processes in “concrete and situated practices” (Di Paolo et al., 2011), it is difficult to see how it may so for cases of higher-level human cognition. There seems to be a “cognitive gap”, between action and perception, which are explained by means of sensori-motor engagements, and the level of more complex cognitive phenomena in human beings (De Jaegher and Froese, 2009) on the other. It seems to me that Mike’s worries are based on something like this. For him mental phenomena just are those very sophisticated, rational processes, which arise in the individual itself; and in human beings.
      The second reading of “cognition should be grounded in autonomy” has already accepted the continuity of mind and life. According to this, autonomy and mind are not seen as two different things that only need to be correctly related. They are identical. Mind just is autonomy. Wherever there is a “mind-level” there must already be an “autonomy-level”. This is a much more radical use of the concept and I believe that this is what Ezequiel has in mind. Autonomy can ground cognition in a naturalistic and non-reductive way because we find the property of autonomy at different levels of complexity (in life forms).
      The task to make the notion of autopoiesis (as the primary and biological form of autonomy) workable would not require to “sneak in mind-level concepts”, but to consider a) autonomy at different levels of complexity and b) the interrelations between autonomy. This does not rule out that there are different kinds of autonomy or aspects of more complex autonomous systems that we cannot find at the lower levels. While the boundaries of an autonomous system such as the cell for instance arise as a result of metabolism and can be physiologically determined, the boundaries of autonomy in human beings are encompassing psychological and social processes, which cannot be materially located. There is a crucial difference between the two: one conceives of identity-construction as metabolic and organism-bound, the other includes non-organismic aspects as well. Yet, this does not mean that both kinds of systems are not usefully related to one another as being operationally closed, teleological and engaged in sense-making activities.
      On adopting the latter reading one can still have an interpretationist stance. In this more radical form of Enactivism the level of interpretation is autonomy, not the mind. This is self-standing and not grounded in anything further – it’s autonomy all the way down (and up).

      I agree that it is possible to bring together some variant of a sensory-motor skill approach with Enactivism. But I do not think that this is possible by “only” adding to Noe and O’Regan’s approach the concept of autonomy. An integration requires elaborations on both sides and this may have to include an elaboration on the concept of sensory-motor skills and what it means to possess them (which is something Ezequiel and I are working on). In Noe and O’Regan the notion of sensory-motor skills is quite restrictive. According to them sensory-motor skills are based on bodily movement. But why should sensory-motor skills require movement? Do they need to biological, i.e. can there be sensory-motor interactions that are artificially realized? Noe and O’Regan are also restrictive with regards to the temporal dimension of sensory-motor skills and thus their general relevance for cognition. For them to master a sensory-motor skill does not necessarily require a momentarily enactment, but only a practical knowing how. I perceive based on the knowledge of what would happen where I to move in such and such a way. This concept of practical knowledge has recently been criticized by Hutto, who worried it renders the sensory-motor approach representationalist. But there are other worries, too. If sensory-motor interactions are only momentarily (as in cases of TVSS) or developmentally relevant to cognition, then there is the danger of trivializing the body’s role for cognition. If sensory-motor skills were to play a role in an enactive (and general) approach to cognition, then one would have to account for these limitations.

    • Hi Mike,

      I am not really sure how much we agree but we definetly do on some aspects. First of all is the critique to the narrow conception of operational closure and structural coupling that TEM provides. Then I think we agree that much more than plain operational closure of the nervous system is needed for cognition. But I am not sure I will push myself into interpretationism. No! If anything Autonomy should stands on its own, as much as you and me, at least ;-) I am not sure how far can we operationalize the notion of “genuine” autonomy but I think that there still a lot to be genuinely learned on its modelling. In this sense the last 20 years have witnessed an true progress from Bittorio as a model of autonomy to much more detailed and richer models in robotics and systems biology.

      Regarding Consciousness and phenomenological approaches, I don’t think we are close enough to talk about such things, not yet. Human brain, body and behaviour afford for many levels of autonomy and if we could only get a grasp on the most simple behavioural one (cf Piaget’s sensorimotor stage) I would die happy as a naturalist-philosopher ;-)

  4. Marieke Rohde says:

    Hey Mike,

    not sure this is the “killer contribution” that you were waiting for but here it is anyways.

    The main tangible difference between VTR and N&O is that the latter are not very concerned about metabolism or other value-generating mechanisms. On a practical level, N&O implement the idea that sensation and motion (and experience) are inseparably intertwined but metabolism/value generation could happen in a different system or not at all. For VTR, the mechanisms of self-production also have to be taken into consideration, as in the famous example of the sugar-seeking bacterium, whose sensation and motion is entangled with its metabolism. In the Barandiaran (&Di Paolo & Rohde, 2010) variant of this, we suggest that processes /like/ metabolism could be at work also on other levels of self-organization, a suggestion that is not yet backed with concrete examples or models. To suggest this is not to explain mind /as/ amoeba, it is to make a case that, like in an amoeba, sensorimotor behaviour and self-production of sorts can be linked and you shouldn’t just assume they’re separate because it’s convenient.

    This is rephrasing some of what Evan and Ezequiel just said, I just try to make it clearer because you seemed initially to interpret “we need autonomy” as something that concerns the subjective level, not so much as something that concerns mechanisms or models.

    For the large common grounds – i agree. It is not of crucial significance to dwell on how N&O enactivism differs from VTR enactivism at this point. For any concrete scientific problem, we can probably be glad if either of the approaches offers a way of appraoching it, and if they do, it is likely to be the same in most cases. Such a problem-oriented way of thinking will likely save us a lot of idle arguing – but that’s probably the wrong thing to say to a philosopher.

  5. Hi Everyone,

    Very exciting to see the discussion starting … I might reply to specific comments inside the main threads latter but I wanted to reply to Mike’s original post first, since I already had my own thoughts about Autonomous Enactivism (AE) vs. Sensorimotor Enactivisms (SME) before this discussion started.

    The first thing to note is that not only was TEM written almost 10 years (and very exciting and productive years!) before the N&O’s BBS paper. It happens that the foundations for the notions of operational closure and structural coupling where developed in 1969 by Maturana and latter by Maturana and Varela in 1973! So the temporal gap on some crucial aspects is really big. But no matter how old a theory or approach is it should stand on its own conceptual architecture and I really believe that the notions of “operational closure of the nervous system” and the notion of “structural coupling” are seriously in need of revision if the Autonomous Enactivism is to stand still. I believe TEM did a good job on its time but we really need a new textbook and despite the valuable update and development of “Mind in Life” we still need something more. But I fundamentally agree with Evan on the essential requirement of autonomy for the SME approach.

    IN SHORT: Here is, IMHO, what we really need, and I think it merges both Enactive approaches: we need a theory of sensorimotor autonomy, which is not a theory of the operational closure of the nervous system, but a theory of the autonomous organization of brain-body-environment dynamics (something like Piaget’s sensorimotor stage). Little progress will be made if we don’t develop this level of autonomy in more detail. My last 5 years of research have been focused on that and I think we are close to achieve it. Only a theory of Sensorimotor Autonomous Enactivism can ground a theory of behavioural-self, sense-making, development, etc. In this sense I really thing the AE and SME are complementary and require each other if we are to make some progress. There is no knowledge or skill without a subject and no subject without a sensorimotor domain from which to emerge.

    PD: I have argued elsewhere why I believe a bacterium has no mind because it does not have a properly behavioural level of autonomy, it has metabolic autonomy and it is a genuine instance of biological agency but not of cognition. The specific manner in which autonomy is constituted on the level of behaviour is of crucial importance to

Leave a Reply

You must be logged in to post a comment.