Sunday, 1 July 2012

Michael Graziano: Consciousness and the Attention Schema

    Abstract: One possible explanation of awareness is that it is a construct of the social perceptual machinery. Humans have specialized neuronal machinery that allows us to be socially intelligent. The primary role for this machinery is to construct models of other people--minds thereby gaining some ability to predict the behavior of other individuals. In the present hypothesis, specific cortical machinery, notably in the superior temporal sulcus and the temporo-parietal junction, is used to build the construct of awareness and attribute it to other people. The same cortical machinery, in this hypothesis, is also used to attribute awareness to oneself. Damage to this cortical machinery can lead to disruptions in consciousness such as hemispatial neglect. In this theory, the value of the construct of awareness, and the value of attributing it to a person, is to gain a useful predictive model of that person--attentional processing. Attention is a style of information processing in the brain in which neuronal signals compete. One interrelated set of signals rises in strength at the expense of others, and thereby dominates the control of behavior. Awareness, in the present hypothesis, is a construct, a useful schema, that models the dynamics and essential properties of attention. To be aware of X is to construct a model of one--attentional focus on X. A brain concludes it is aware of X, and assigns a high degree of certainty to that conclusion, and reports that conclusion, because of an informational model that depicts awareness of X.

    Graziano MSA and Kastner S (2011) Human consciousness and its relationship to social neuroscience: A novel hypothesis. Cognitive Neuroscience, 2: 98-113.

Comments invited


  1. Inman Harvey's horse meets Mike Graziano"s squirrel!

    But when I have a headache, what I have (Cartesian) certainty about is what *feels like* a headache (right now) -- no certainty about having a head: it might be referred back pain, or phantom head pain, or I may not have a head, nor a body; there may be no external world -- just Cartesian certainty about what it "feels like* right now. And feeling is not a squirrel, nor a description, nor magic. It's feeling. All mammals -- and probably all vertebrates and invertebrates too -- have it. And if they can doubt at all, if they are doubting that they are feeling when they are feeling doubt, they are contradicting themselves (hence not making sense).

  2. "Awareness, in the present hypothesis, is a construct, a useful schema, that models the dynamics and essential properties of attention."

    What does the model afford?

    Attention has a clear function: it selects important features of our environment. Couldn't we access and benefit from "the dynamics and essential properties of attention" simply by having attention, without this extra step of awareness (feeling)?

    It's as if Graziano is saying something like, awareness is what attention feels like to us. In other words, feeling is what attention feels like to us. This is like what Damasio says about homeostasis: consciousness is what homeostasis feels like to us. Maybe they wouldn't agree with these statements; but if they do they are begging the (Hard) question.

  3. Good points. One cannot simply pick an obviously adaptive function -- doing -- and baptize it "feeling," saying *that's* what feeling really is. I can say the same thing about moving, or breathing: That's what feeling (consciousness) really is. Why? How? (And, by the same token, why and how is selective attention -- indisputably useful -- *felt* selective attention, rather than just selective attention [i.e., doing]?)


    What *is* the squirrel argument (Graziano)? Try to repeat it to show what it assumes and what it implies. I feel there is a squirrel in my head. There is no squirrel in my head. Maybe I have no head. But if I feel it, I feel. That feeling is not a squirrel, it is a feeling. How does the non-existence of the squirrel imply the non-existence of the feeling? And what has this to do with description, or magic? (You have been taken in by a bit of entertainment, like Mike's ventriloquist dummy, Kevin, and like Inman's horse. Shake off the joke and root out the sense, if any.)

    1. While I find Doctor Graziano’s explanation of consciousness quite appealing—the idea that it could be an “attention schema”—I do find that it leaves the hard problem completely untouched. I am intrigued by the phrase: “Awareness is representation; attention is what is represented.” However, I do not see what it is about being a description or representation of a particular object that would make the latter conscious. Why would such a representation be felt?
      I agree with Doctor Harnad. Indeed, what is it about representation that would yield something like “feeling”? I understand that the squirrel argument would have it that the mysterious properties we attribute to consciousness are elaborate illusions—but this simply avoids the question entirely. From a Cartesian point of view, I cannot be in error about my feeling what I feel. How would one account for the feeling of a representation itself—and supposing that feeling was indeed an elaborate illusion, what accounts for the illusion?

    2. This talk certainly resonates with many of my own perspectives. Though I think that Dr. Graziano has falsely described the hard problem, I certainly share his belief that we should expend our efforts explaining how the brain produces consciousness before we are fully able to understand the why we experience consciousness, or its adaptive functions.

      To Dr. Harnad:
      I believe Dr. Graziano intended the squirrel argument as an illustrative analogy. The non-existence of a squirrel in the head does not imply the non-existence of the feeling, and Dr. Graziano has not claimed such a thing.

      Because only one patient feels as though there is a squirrel in his head, we approach this phenomenon in a very different way than we approach any other feeling. If one were to explain what it was in the brain that causes this man to feel as though there was a squirrel in his head, we would understand how the man came to feel as though a squirrel were in his head. However, suddenly when we replace the word squirrel with consciousness, many no longer believe that explaining all of its neurobiological mechanisms is sufficient to explain the presence of consciousness itself. This is an untenable position.

    3. The squirrel is analogous to consciousness, not the feeling that there is a squirrel.

      However, unlike the existence of a squirrel and a feeling that there is a squirrel there, consciousness and the feeling of consciousness are one and the very same thing. So the analogy doesn't work.

      Dr. Graziano is asserting that consciousness doesn't exist which we know with incorrigible certainty is false.


    Notice all the weasel-words one falls into if one does not call a spade a spade (i.e., call feeling a feeling)? What I "know" for sure is that I feel something (anything) when I am feeling it. I know it because I am feeling it. I can be wrong about what that feeling implies about the world (e.g., there is something wrong with my tooth, I have a tooth, there is a squirrel in my head, I have a head, there is a world, etc. etc.). But I cannot be wrong that it feels what it feels like (while it is being felt). [Descartes].

    So (many of) you have indeed been taken in by the entertainment. But it is a useful exercise -- if you can now reason your way clearly through the whole thing.

    Ditto for the ventriloquist dummy, "Kevin": What does it actually show, apart from the fact that a few cues are enough to make us think (and feel!) there's another thinking/feeling/talking person there? [Other-Minds Problem]. Did that teach you anything new about anything?

    And what about the suggestion that feelings are really "descriptions"? Descriptions are symbol-strings: Have we not already gone through the limitations of computation as the physical substrate of cognition (Searle and the symbol grounding problem)? Mike Graziano is simply declaring that he is a computationalist here.

    And magic? What's magic here? Mark Mitton is the magician. But he is showing you ways that the mind can play tricks on you, making you feel the wrong thing about the world -- not that you might be wrong that you are feeling what you are feeling when you are feeling it.

  5. A couple points on Graziano's *extremely* interesting talk (they're long points, so I'll split them up into two posts):

    First, I suspect that the squirrel argument is somewhat misleading as to Graziano's actual position on consciousness. Graziano seems to suggest by the argument, and by other comments he made, that he thinks consciousness *doesn't exist.* But he also seems to suggest, confusingly, that consciousness just is an 'attitude schema,' which is something like an attribution of attention (either to oneself or others). So it would appear Graziano *does* think consciousness exists; it' just that it just is an attribution of attention.

    Perhaps what Graziano is thinking is that the concept of consciousness is incompatible with its *being* merely an attribution of attention, so his position is not really that consciousness is an attribution of attention but that *the phenomena we commonly (and mistakenly) describe as instances of consciousness* are in fact not instances of consciousness at all, but rather instances of attributing attention to oneself.

    I myself doubt that the concept of consciousness is so inflexible that it precludes consciousness' being no more than an attribution of attention, but this may be merely a semantic sort of issue, so I'm not much interested in pursuing it. My point is just that it's not necessarily an appropriate response to Graziano's point to insist that there obviously *is* 'something it's like.' Insofar as I understand him, Graziano's point is just that that this 'something is is like' just *is* an attribution of attention, not that there is *nothing* it is like.

    If Graziano's position is, as I am suggesting, not that there is *nothing* it is like, but rather that that what it is like is just an attribution of attention (which may or may not be compatible with that 'what it's like' being consciousness), then it's not as clear that Graziano's position really frees us, at least as much as he thinks it does, of having to account for some of the more 'magical' components of consciousness. Rather, it may simply move the explanatory work to the questions of *what* attribution and attention are (this latter point was made by someone in the Q&A, whose name I unfortunately did not catch, apologies to that person!)

    So that's the first point: Graziano's position may not be as radically eliminativist about 'what it's like' as he seems to suggest. It may rather be a kind of revisionism and this sort of position, while perhaps more attractive to some than eliminativism, does not obviously decrease our explanatory load, contrary to what Graziano suggests.

    1. When I feel a headache, is it an attribution of attention? (What on earth for?) I think it's just feeling a headache...

    2. Yes, I am myself sympathetic to the idea that a headache isn't just (or even in part) an attribution of attention.

      My point in this post was just that, contrary to what might be suggested by Graziano's 'squirrel' analogy/argument, Graziano doesn't actually deny *that* you have a headache, he just thinks it's nothing more than an attribution of attention. And this view may not be *so* radical, since attributions (or schemas or representations, as he also calls them) and attention aren't themselves very nailed down sorts of things. Before we know more about the sorts of schemas involved and maybe even about attention, it would be hard to pass judgment in the view.

      I suggest in the following post one reason why Graziano should probably *not* say that attributions are just judgments, but there are many things other than judgment that the relevant sort of attribution/schema/representation might turn out to be.

  6. My second point is about the predictions Graziano's theory would make about blindsight patients, in particular about blindsight patients who have learned that by considerng their own guessing behavior, they can (with at least some reliability) attribute to themselves attention to some object. For example, a blindsight patient might learn to correctly attribute to herself attention to some object that is tilted at such-and-such an angle. If attribution of attention to some object is sufficient for being conscious of that object, is the trained blindsight patient then conscious of an object tilted at such-and-such an angle?

    Also, and in a more critical vein, if the trained blindsight patient is conscious of (say) an object tilted at such-and-such angle, what is the difference between the trained blindsight patient, who apparently cannot *see* the relevant object, and the normal subject, viewing the same object? Is it merely a difference of degree, one that derives from the fact that the trained blindsight patient's information about the object is quite impoverished, whereas the normal subject's information about the object is comparatively rich? I'm a bit skeptical myself about whether this sort of difference could entirely account for the difference and am interested in what Graziano might say about such a case.

  7. Dr. Graziano's theory on attributing awareness to others brought to my mind autism. Is autism not described as a deficit of relating to others? If the TPJ area is such an important part it would be interesting to look at functional and/or anatomical integrity of this area in autistic people. Given the theories relating autism with dysfunctions in fundamental synaptic mechanisms and presumably these deficits have widespread consequences on many brain areas, it is interesting that a specific area should be given a function which appears intruitively somewhat complex, higher-order and multimodal?

  8. The only trouble is that this post-modern view contradicts all the neurobiological evidence.

    The brain substrate of conscious perception, cognition, volition, etc., is the thalamo-cortical system. The evidence for that is available in many places, but permit me to recommend Baars & Gage, Fundamentals of Cognitive Neuroscience: A Beginner's Guide. (Elsevier/Academic Press, 2011). Chapter 8 has all the latest information about consciousness and attention in textbook form.

    The above is true in mammals, and very plausibly in birds, which have analogous brain structures. Much evidence suggests that the same point applies to at least some cephalopods --- which have mammal-size brains, many of them.

    It follows that consciousness (or awareness) cannot be a social construct, either in ourselves, or in interpreting others. In the case of mammals and birds, the biological evidence is rock solid.

  9. POSTED on FB during the talk

    If " we attend of what we do" and accordingly to the upper model on the first slide, which show that we are conscious of what we attend to, what we are suppose to conclude about the syndrome de la main étrangère?

    Not familiar with that syndrome, can you expand? Along the same line, I can easily imagine a situation where someone acts on something that she isn't aware of, despite that she's still in a conscious state. Clark says this doesn't happen but he hasn't given us evidence to support this claim.


    Its like youre not conscious of mouvement of a part of your body, you dont have any feeling about it, you cant control those movements. So what does it tell about intention/attention and feeling?

    Actually, I'm surprised and a bit desapointed that the summer school doesn't have any talks on consciousness-related "alterations" as found in pathological cases such as patients with "main étrangère" syndrome or some crazy ones such as synesthesia, somatoparaphrenia, apotemnophilia, anosognosia, out-of body experiences and others (see the book of VS Ramachadran for more details

    1 juillet, 14:46

  10. I just watched this talk again, I love it. There is no squirrel in our head, haha!


    To all who have commented on my talk: thank you for the time and thoughtfulness!  I have been wandering the jungles of a hiking trail, out of contact. Here I give a very brief general response to the kinds of comments that seem to come up frequently and that are also represented in this thread.

    Although we “know” we have an inner experience, I just don’t see how an information processing machine can arrive at that conclusion, and configure that information for speech, unless the basis of it IS information. The question to me is: how, where, and why (in the sense of utility) does the brain compute that information? The attention schema theory answers those questions.  The construct “I am aware of X” serves as a useful way to model the physical act, “this brain is focusing attention on X.”

    I would not equate consciousness with social cognition. Social cognition covers a lot of ground, much of which is not directly relevant to the question. Autistic people can be impaired on many aspects of social cognition and yet presumably be fine on other skills. I would suggest that all brains use the data-sorting method of attention. Some brains construct an internal model of attention, in order to better predict and monitor their own output. This internal model of attention is, according to the theory, what we call awareness. Even fewer brains in the animal kingdom, perhaps only the most socially sophisticated, use this internal model of attention not only to model the self, but also to model others. Thus awareness is handy for social cognition. But I would not equate them.

    I do not think there is a hard problem. I think it is a huge mistake, a hole into which the field of study had gotten its leg stuck. It is like asking how many angels dance on the head of a pin. There are no angels. Brains compute that they have an inner feeling. The computation has a utility. If there were a utility to computing that we have an inner nuclear Gandalf with six fingers and no conscious experience attached to the item, then we would all compute that, and all "know" that we have it.


    Michael has expressed a solid credo about information processing and speech: He thinks they would not be possible unless the brain was "aware": But what does Michael mean by aware? If it just means being in possession and command of information (data), then that's just doing (and even today's toy robots already have it). If he means that it feels like something to be in possession and command of that information, then he still owes an explanation of how and why that should be: Doing (i.e. Turing) sounds like it ought to be enough...

    Michael writes "Brains compute that they have [an inner] feeling. The computation has a utility."

    Sound promising: But how do brains "compute" feeling?

    (And Is there any other kind of feeling than "inner" feeling"?)

    And why do brains compute feeling? What is that "utility"?

    (The utility of being able to "compute" everything we are able to do is indisputable -- but that's all just doing, not feeling: Why felt doing?)

    I couldn't quite follow the Gandalf bit...

  13. Originally posted on facebook

    A fundamental question that all scientists working on consciousness need to agree on an answer to is: do all animals have consciousness, or do only humans have consciousness and other animals only evolutionary precursors? The multitude of opinions on this matter confuses the hell out of me.

    I'm still confused by that too. At first I thought of feelings or consciousness as in a way "aknowledging" something that is happening like "I feel pain", which is like being in the moment and feeling what is happening (your attention is focused toward what you are feeling?). I guess animals can be in the moment too and feel. But then came the octopus and all the talks about vision and it confused me. For the octopus Mr. Edelman said that it was a primary consciousness but is being aware of your environment and reacting to it is consciousness (even a precursor) ? Isn't it just an innate reactions for an octopus to get the crab? Today, Mr Harnad wrote somewhere that it's not just detecting/reacting and I agree with that, but some talks seems to point toward that. So this summer school is a big learning path and I guess it will be like that until the end :D

  14. Originally posted on facebook

    Carey Huh:
    Dr. Graziano's theory on attributing awareness to others brought to my mind autism. Is autism not described as a deficit of relating to others? If the TPJ area is such an important part it would be interesting to look at functional and/or anatomical integrity of this area in autistic people. Given the theories relating autism with dysfunctions in fundamental synaptic mechanisms and presumably these deficits have widespread consequences on many brain areas, it is interesting that a specific area should be given a function which appears intruitively somewhat complex, higher-order and multimodal?

    Marjorie Morin:
    Indeed autists are impaired in theory of mind (on tests like false beliefs and Happe's strange stories) and TPJ is an important region for this ability. I would think that autists do perceive the social clues (so their attention must be focused on them) but are unable to make sense of them/integrate them together and they won't necessary see their role in your feelings (like if they make you cry they won't understand that it's their "fault").

  15. The paradox (or pseudo-paradox) of the internal simulation hypothesis: in order to know what you know, I need to know what you know.

    In other words, if want to know what you know/believe, I need to process your inputs in my own "cortical machinery". But for that I need to know what those inputs are, and I don't have direct access to your inputs. So I need to use my ToM to know what to input in my ToM.

  16. How are we to make sense of Dr. Graziano's claims that awareness (phenomenal consciousness) is a depiction of attention (or an attention schema, as he also puts it) in light of Dr. Tallon-Baudry's claims that attention and consciousness are two distinct things and that they have different neural correlates. Her work seems to provide a pretty strong, emprically-motivated objection to the kind of view that Dr. Graziano defends.

  17. I think it would be helpful to clarify the meaning of "awareness" as Dr. Graziano uses it. He appears to equate "awareness" with the sense of an agent, of an "I" that is aware, and to be saying, in effect, that there is no agent/squirrel/ghost in the machine/apart from the neuronal structures and processes (I don't like the word "machinery" in this context) that give rise to the sense of an "I".

    He therefore identifies this sense of an "I" with the product of processes that 1) model what has been attended to and 2) "reports" that model in the sense of saying/feeling that "I am aware." Without this report, there would be no sense of the source of awareness, and thus no sense of an "I".

    While I am not qualified to evaluate this argument, I would say that it seems attractive and sensible to me.


      You're nearby in Princeton Junction: Ask Mike!

      What he had to say seemed to me to apply equally to felt and unfelt doings, hence it did not seem to have anything to do with consciousness at all...

    2. I believe that he addresses the "feeling" issue as follows in his 2011 discussion paper, "Human consciousness and its relationship to social neuroscience," coauthored with Sabine Kastner:

      In summary, we speculate that even though conscious experience is information computed by expert systems, much like the information in a calculator or a computer, it nonetheless can “feel” like something in at least the following specific sense: A decision process, accessing the self-model, arrives at the conclusion and triggers the report that awareness shares associations with somesthesis. This account, of course, does not explain conscious experience in its entirety; but it helps to make a central point of the present perspective. ‘Experienceness’ itself may be a complex weaving together of information, and it is possible to tease apart at least some of that information (such as the similarity to somesthesis) and understand how it might be computed."


      This not only "does not explain conscious experience in its entirety": It does not explain consciousness at all.

      "awareness shares associations with somesthesis": Is that like sharing a sandwich? What on earth does it mean? Who/what is sharing what with whom/what?

      "‘Experienceness’ [!] itself may be a complex weaving together of information". Really? Could you step me through it please, slowly, for, say, a headache?

      It's simple: Explain why and how (some) information-processing (or any state at all) is felt, rather than just done.

      Anything else is just word-play and question-begging.