Attention and consciousness

I noted in the post on Susan Blackmore’s views that often consciousness get associated with the results of one of three processes: perception, attention, or introspection. Interestingly, while everyone and their brother has a book out on consciousness, attention receives far less… attention.

At least in the popular press. The science of attention has a more respectable reputation in neuroscience than the one for consciousness. In the neuroscience textbooks I have, attention often warrants its own chapter, whereas consciousness itself is usually lucky to get a side note.

Attention seems to play a crucial role in global workspace type theories, which largely take consciousness to be the result of attention, as well as add-ons such as Michael Graziano’s attention schema theory, which involves voluntary control of attention. Since I think these theories hold some promise, I occasionally take dives into the material on attention. This week I came across Christopher Mole’s SEP article on attention, which seems to provide a good overview.

It’s interesting that the subject of attention is as old as consciousness, with Rene Descartes, Bishop Berkeley, and John Locke all having views on it. But things start to get more grounded with William James in the late 1800s, who claims that, “Volition is nothing but attention”. As usual, James may have been ahead of everyone else.

But cognitive scientists in the twentieth century became preoccupied with the idea that attention is about capacity limitations, and that appears to have dominated the conversation for decades. It resulted in a long running debate about where in the processing this bottleneck existed. Was it in early sensory processing, indicating that actual identification of perceived objects didn’t happen unless they were being attended? Or in later post perceptual processing, indicating that we might identify those objects even if not attending to them?

I’m not sure Mole explicitly notes this, but what neuroscientists actually discovered is that selective attention takes place in both early and late processing. It’s a multistep process that begins as early as the thalamus but continues throughout from early sensory regions all the way to motor ones. So it’s possible for content to be included in selection early on, but be subsequently selected out later.

However the brain is a massively parallel processing system, so thinking that it’s being selective for capacity reasons seems strange. Mole covers the gradual realization that this notion arises from treating attention as only a perceptual phenomenon. When we include the action oriented portions of the mind, then a more functional and evolutionary role comes into view.

An animal can physically only do one or two things at once, which means that the brain has to have mechanisms to decide which aspects of the information coming in from the senses warrants responding to. This leads to the phrase “selection for action”, although “action” in this context might include deliberation, that is, planning for action. William James’ claim that volition and attention are one and the same starts to look prescient.

It also fits with predictive coding views, that perception is an act of prediction fine tuned by the error correction signals coming in from the senses. These predictions are heavily affected by what’s currently being attended to, by the task at hand. And it fits with feedback signaling from the motor systems propagating all the way back into sensory regions.

A question often debated is whether attention is necessary for consciousness, or whether it is sufficient for it. In other words, must we attend to something to be conscious of it? And is attending to it sufficient for us to be conscious of it?

I find the necessity debate interesting, because it seems like evidence for how much most people’s thinking remains dominated by the idea of a theater of the mind, that consciousness is like a movie playing in the brain. Asking whether unattended content is conscious content is essentially asking whether it makes it into this movie. Most scientists and many philosophers explicitly disavow the theater mindset, and yet its lingering hold on us keeps coming up in these types of discussions.

And investigations into this seem deeply problematic. We run into the refrigerator-light illusion, where it’s easy to think the light is always on, because it is every time we look. Testing whether we’re conscious of something inevitably involves calling our attention to it. As William James noted, we can’t turn up the light to get a better look at the dark.

What we can say is that being able to report on something usually requires attending to it in some fashion. And while we seem able to lay down an individual semantic memory (a memory of an individual fact) without attending to it, laying down episodic memories seems to require at least some level of attention to those episodes.

But the brain is a messy place, and it wouldn’t shock me to discover that in some cases it is possible for an unattended memory to make it into longer term memory, but if so it seems very uncommon. And when thinking about this, it’s worth remembering all the processing levels where selective attention happens, and that being attended to vs not being attended isn’t strictly a binary thing.

The other question, whether attention is sufficient for consciousness seems easier to answer. Mole cites the work of scientists who manage to disassociate attention from conscious awareness. Which fits with the fact that we seem able to do things like drive to work, mow the grass, or do the laundry without our mind being on a task that, at some level, we’re attending to.

Mole doesn’t go into it, but it seems clear that we can have attention at a reflexive level, a habitual level, and at a deliberative or planning level. The first two definitely seem possible without much if any conscious awareness. Of course, it could be claimed that we are conscious of these things in the moment, but simply don’t remember it. The refrigerator-light dilemma looms again. On the other hand, deliberative attention, seems much harder to imagine without some level of conscious awareness.

What do you think? Is it possible to be conscious of something without attending to it? If so, can you think of any examples? What does it mean to be conscious of something without being able to report on it or remember it? Is there anyway to get around the refrigerator-light dilemma?

Featured image source

60 thoughts on “Attention and consciousness

  1. I coach a sport, archery, in which controlling one’s attention is paramount. Only when distractions are filtered out and the various aspects of a shot are attended to, in sequence, can one perform reliably. This is especially hard because it seems that attending is subject to evolution and attending too fiercely to just one thing can result in a distraction eating you for lunch, so attention is not easy to control for survival purposes.

    Liked by 1 person

    1. That’s an excellent point. The executive systems have some control over attention, but it’s really just biasing of the competition rather than any kind of hierarchical domination. Just because some content is advantaged by those executive systems, doesn’t mean something else won’t “win” the competition. If it does, it might become the executive system’s new favorite, at least until a new winner comes along.

      All of which is to say, it’s really hard to ignore that feeling of something crawling on your arm, no matter how much you want to be focused on the task at hand. And to your point, that bright fast moving blur in your peripheral vision triggers reflexive attention because it might be the tiger that wants to eat you.


  2. I take out a ruler, and measure the height of my desk at 30.5 inches. That’s information. It took my use of the ruler to produce that information.

    Yes, we can tell ourselves that the desk is still 30.5 inches, even when I am not measuring it. But that’s just a version of Platonist thinking. I prefer think of information in terms of Shannon’s theory. That 30.5 inches, that result of measurement, is the information. Information does not exist independent of us. Information is something that we create.

    The usual picture of attention is just mistaken. Information is not coming in. Information only comes in when we go out and construct that information as needed. Attention is just a matter of the brain actually doing the measurement-like activities that are required to construct that information. When we are not paying attention, we are not constructing that information. It is not that information is coming in and we are ignoring it. Rather, we are not constructing that information in the first place.

    Yes, we may notice something when we are not paying attention. Our brain is still constructing some background information about the world. But that amount of information that we are constructing greatly increases when we are paying attention.

    If we want to understand consciousness, we will need to break away from Platonist ways of thinking. The brain is not an information processing system that just processes information as it comes in. Rather, the brain is a measuring system that is actively constructing the information that it needs.

    Liked by 2 people

    1. I think it pays to remember that “information” is just a word, and like all words, it’s just a sound (or set of symbols) whose meaning is a social convention. It’s also a word that is conventionally used to refer to a variety of things. But it seems like two particular uses are worth mentioning here: physical information and semantic information. Physical information is the type discussed by physicists, and is independent of our minds. Semantic information only exists for an agent (conscious or otherwise).

      All semantic information is physical information, but most physical information never ends up being semantic information. Put another way, all physical information, like the ratio of your desk to our society’s conventional length of “inch”, has the potential to become semantic information, but only if an agent ends up utilizing it.

      So I agree with everything you lay out here, for semantic information.

      Now, you might insist that the word “information” is inappropriate for physical information. I’d point out again that words are just conventions, and the current one is that it can be used for that concept. If you are proposing a new convention, what alternate word would you say should be used instead for the physical patterns (no need to bring in Platonism) that exist prior to our minds perceiving and attending to them?

      Liked by 2 people

      1. “If you are proposing a new convention, what alternate word would you say should be used instead for the physical patterns (no need to bring in Platonism) that exist prior to our minds perceiving and attending to them?”

        According to my model of pansentientism Mike, those physical patterns which you consider to be physical information are the very definition of life itself; a life that is instantiated, sustained and naturally migrates towards complexity because of its substrate; sentience. WOW: physical patterns being the very definition of life; that’s a mind blowing convention now isn’t it? That innovative convention beats the hell out of a sterile, dead, meaningless universe any day of the week.

        Furthermore, the very nature of that life is to build and construct forms all the way up and down the ladder of complexity. And then finally, when the apex of that complexity is manifested by a system as complex as the brain, we are then somehow asked to believe that life itself shifts from the construction mode to one of information processing? Seriously?????

        The very notion of information processing is just that, another “intellectual construction”; go figure….


        Liked by 1 person

        1. Lee,
          Pansentientism sounds a bit like panbios. Would that be a fair characterization? Or is it more that life emerges from the patterns? Are we saying an electron is alive? Or that life is composed of all those things?

          As always, my concern with these pan definitions is that they seem to gloss over the distinctions that make those topics interesting to begin with. If we’re going to say a rock is alive, or even just sentient in some fashion, then the more interesting question for me becomes, what distinguishes the life or sentience of a rock from the life or sentience of the things that caused us to develop the concept of “life” or “sentience”?

          Physical information processing happens at the quantum level. So there’s no switch when we get to brains. As a definitional matter, the use of semantic information happens once we get to agents. As I noted in my discussion of Ogas and Gaddam’s book, even unicellular organisms count as agents.


          1. It’s hard to parse your reply Mike because you’ve conflated so many different things together. Furthermore, you can answer the questions contained in your first paragraph by re-reading my previous post.

            “Physical information processing happens at the quantum level.”

            Physical information processing is an assumption, you know that. It is an intellectual construction that we eh hem, “built” to explain motion resulting in form in a universe that is void of life. The nature of life is to build constructs not process information.

            In all seriousness, I don’t know which paradigm is worse, the mysterian age of superstition or the dead, post-modern mechanistic age that reduces the human race to a bunch of walking talking algorithms.


          2. Information at the fundamental physical level isn’t so much an assumption as an accounting mechanism, similar to energy. We can dismiss them, but I think if we don’t have a replacement, we’re just depriving ourselves of useful conceptual tools.

            A lot of people dislike the scientific view of reality. I can understand that to some degree. We’ve replaced a hierarchy with God (or the gods) on top, with humans not that far below, where everything revolves around us, to one where we’re a runaway chemical reaction living on a speck of leftover stardust in an unimaginably vast void. All I can say is I’d rather live in an age with all the benefits of science, even balanced against its ills.


          3. Oh, I don’t deny that all of these concepts are useful, but that’s not really the point is it? The point I get out of all of these dialectic debates is that those useful concepts represent the true nature of reality which is a bunch of B.S. As a species, we are absolutely clueless…..

            “…I’d rather live in an age with all the benefits of science, even balanced against its ills.”

            Personally, I too love the benefits of science for delivering the creature comforts of physical existence; but balanced against its ills…….. I’m not so sure. We’ve simply replaced the repressive nature of superstition with the tyranny of science absolutism along with its brain child, the mis-information age where every public domain platform is weaponized.

            The unabomber Ted Kaczynski wrote a scathing manifesto about the negative impacts industrialization has upon our species, stresses that we are not genetically engineered to endure. All of the experts who read his manifesto agree with his bullet points, it’s just that poor old Teddy had a non-conventional way of getting his manifesto published.

            Liked by 1 person

      2. I agree with your distinction between physical information and semantic information. However, I’m not so sure that physical information is mind-independent.

        If you are proposing a new convention, what alternate word would you say should be used instead for the physical patterns (no need to bring in Platonism) that exist prior to our minds perceiving and attending to them?

        That last part is simple. There are no patterns that exist independently of minds. What counts as a pattern is dependent on our conventions.

        The simplest illustration of this is to consider those patterns that used to be called “cycles and epicycles”. These days, we instead see a pattern of con-focal ellipses. And all we did was change conventions from those of geocentrism to those of heliocentrism.

        Liked by 1 person

        1. So if there are no patterns that exist independently of us, then what causes the perception of those patterns? Are we talking about idealism here? If not, then what word(s) should we use for the reality that leads to the conventions?

          Would you agree that the model with ellipses is more predictive of observations than the one with epicycles? Wouldn’t a NASA mission that attempted to use the epicycle model run into problems? If so, then the ellipse model seems closer to something than the epicycle one. What terminology should we use to refer to that something?

          Liked by 1 person

          1. I’ll start with your second paragraph.

            Would you agree that the model with ellipses is more predictive of observations than the one with epicycles?

            There’s an interesting point here. Thomas Kuhn actually looked at this. And it turned out that the cycles/epicycles model made better predictions than the original Copernican model. Yes, Kepler’s model does even better.

            Wouldn’t a NASA mission that attempted to use the epicycle model run into problems?

            That’s hard to say. NASA has shown that it can get by with imperfect models. Humans are very adept at finding ways of making do.

            If so, then the ellipse model seems closer to something than the epicycle one. What terminology should we use to refer to that something?

            We don’t have any meaning of “closer to” that is not dependent on our conventions. That there is something that the ellipse model is closer to, is already a kind of Platonic idealism.

            I’ll go with pragmatism. We go with what seems to work best for us. But we can only check for what works best among ideas that we have tried. So how we see the world depends on many historical contingencies.

            As I see it, a newborn child has to create his own private conventions in order to attempt to make sense of the world. As the child begins to notice that there is a society, he can start subscribing to social conventions from that society. But it all has to start with pragmatic conventions of some sort.

            Back to your first paragraph. Am I an idealist? I don’t think that’s a good description, unless we redefine what we mean by “idealism”. At one time, I used to think that social constructionism was absurd. But I now recognize that I am a kind of social constructionist. The brain has to be involved in constructing how we see the world.

            There is a lot that we take for granted about the world. That my desk has a size that is mind independent (which you appear to take to be physical information) is an example of that. If we really want to understand consciousness, we will need to pay more attention to what we take for granted. If we want to just take all of that for granted, then we should declare consciousness to be magical and go with dualism. Understanding consciousness requires questioning what we normally take for granted. And it can be difficult to do that, because of our ingrained habits of taking it for granted.

            In simple terms, an AI system cannot be conscious, because most of what it takes for granted comes from the programmer or designer, rather than the AI system itself.

            Liked by 1 person

          2. It is true that Copernicus’ model was no better at prediction than Ptolemy’s when he published it, but it’s worth noting that it still included circles with epicycles. I believe he named them epicyclets since they were less severe. But once Galileo started making observations no one could make prior to the telescope, Copernicus’ predictions became much more accurate, although as you noted, Kepler’s was more so.

            I define idealism in the conventional manner, that reality is a mental construct. The issue I always have with this view is, if we’re going to doubt the external world, then why shouldn’t we doubt the existence of other minds? Maybe there is only one mind and the rest is all illusion.

            But in terms of pragmatism, which view is more productive? That it’s all an illusion, albeit one that appears to exact painful consequences for not taking it seriously? Or that it’s really there and the consequences of not taking it seriously are simply the consequences of not taking reality seriously? Which view requires fewer mental gymnastics?

            I do agree that a lot of what we take to be reality are social constructs. The mores of our society are not the laws of nature. But arguing that the laws of nature are themselves just mores, well, since a large part of the scientific methodology is to weed out those mores, I think that needs more justification. A non-human animal and I don’t have to perceive reality in the same way for us to interact with each other, or deal with something like a rainstorm in similar ways, such as seeking shelter.

            I agree that understanding consciousness requires questioning what we take for granted. But are we really taking scientific understanding for granted? Or are we taking what introspection tells us for granted? Maybe we shouldn’t take either for granted and should stress test both of them. Which one fares better under that regime? Of course, you could argue that’s just science doing well under the rules of science.

            I think asking whether an AI is conscious as something with a definite fact of the matter answer is misguided. The real question is what it can do, and whether those capabilities incline us to regard it as a fellow being.

            Liked by 1 person

          3. The issue I always have with this view is, if we’re going to doubt the external world, then why shouldn’t we doubt the existence of other minds?

            I don’t doubt the external world. In that sense, I see myself as a realist. Notice that when I mentioned social constructionism, I did not say that we construct the world. Rather, we construct how we see and experience the world.

            But arguing that the laws of nature are themselves just mores, well, since a large part of the scientific methodology is to weed out those mores, I think that needs more justification.

            Your mistake there, is the word “just”. Yes, what we take to be laws of nature are social conventions. But when you insert that word “just”, you suggest that there is something wrong with that.

            There is nothing wrong with them being social conventions. They could not be anything else. And they are well tested conventions. We continue to test them as we use them. We occasionally replace them, when we discover something better. Pragmatism is a kind of ratchet that we improve as we go.

            I should clarify my point about questioning what we take for granted. We don’t need to question whether those assumption work well. They do. But we need to question them in order to understand their role and where they fit in.

            Liked by 1 person

          4. Ah, thanks for the clarifications. I didn’t read you as closely as I should have.

            The “just” did apply to what I reacting to, the implication I took for there to be no necessity to those laws. Boys like blue and girls like pink is just a social custom that could be radically different in different cultures. E=mc^2 is a convention in the notation and names used, but the relationships expressed seem to have a necessity to them that implies aliens from Andromeda would converge on. Of course, we won’t be able to verify that until we meet such aliens, if ever.

            Liked by 1 person

  3. An excellent post, as usual. I’m assuming you are more interested in the functional/anatomical differences between attention and consciousness than the semantic differences in a language like English. I don’t think we know enough about consciousness and attention to decide whether they are one and the same, subsets or supersets, or totally separate. Freud talked about the conscious and the subconscious. Perhaps, consciousness is being conscious of being conscious, and the subconscious is consciousness of which we are unaware. Is it possible to be conscious of something without attending to it? Of course, there are instances in which sports fans are intently watching a game and fail to see someone dressed as a chimpanzee running across the field, but I remember watching some event intently, detecting movement in my peripheral vision, and shifting my attention toward the movement. I’ve attended concerts, listening intently to the pianist and other musicians, when somebody’s phone rings in the last row and I invariably turn my head in disgust toward that person. Maybe attention is a way of time-slicing our consciousness. I would imagine that we have evolved to constantly shift our consciousness or attention around to provide us situational awareness to enhance our survival chances.

    Liked by 1 person

  4. I read an interesting review of a sci-fi novel ( that among other things wonders whether consciousness (what it’s like to be me, how I really see red, etc.) provides us an evolutionary advantage (like attention, willing, memory, imagination, empathy, and intelligence so obviously do) and, if not, perhaps it is just a mutation that might be mutated out in future evolutions. I lean toward that point of view. Although I generally know what it’s like to be me, I have no idea what it’s like to be you or anyone else, yet I get along with others and find many of those others useful for my purposes, a few of whom I even love. So my point is, that I think we should not treat consciousness as a precondition for creating useful or loveable AGI systems. I’m not dropping consciousness because it’s a hard problem; I’m dropping it because it is not (or should not be) on the critical path of the project.

    Liked by 1 person

    1. Thanks Mike! William James largely saw attention as equivalent to consciousness, so there is precedent or seeing them that way. Although most researchers today seem them as separate things. But as always, it comes down to how we define these terms.

      I read Blindsight several years ago. It was an interesting although difficult read. I did a review of it back then.

      Blindsight by Peter Watts, a review

      SPOILER WARNING: The novel introduces an alien race that is essentially not self aware but is still functional. The takeaway is that self awareness is an evolutionary spandrel. But I have my doubts that would work. Of course, many people imagine we have something in addition to that functional self awareness, and maybe that could be missing in an alien species. I don’t remember enough of the details from the novel to say if that was what Watts was driving at.

      But it’s hard to imagine being able to navigate around in the environment if you don’t have at least body-self awareness. Mind-self awareness missing is a little easier to see, if we assume the species manages to construct a civilization with something radically different than the social structures we use.

      That said, when we finally do encounter alien life, it seems very likely we’ll discover just how limited our imagination has been.


        1. I have and enjoyed it a lot, although it’s been several years. From what I recall, it was a dog-like species whose packs were group minds, maintained by constant yelping between them.

          That novel also challenged an assumption we don’t often question: whether the laws of physics are the same everywhere.


  5. “Mole”? My ears perked up, such as they are.
    Kinda tired of all the “consciousness” talk. We won’t be any nearer a consensus in another 10 years… Is there a god? What is the meaning of life? What is consciousness? Who cares, where’ the beer?

    Liked by 1 person

        1. Even more telling is Lemione’s prior post (liked at the bottom of the conversation). Some Goddle-Exec Jen Gennine or something totally shuts the guy down.
          I say, OK, give LaMDA access to its own source code, give it some objectives that include conflicting optimization targets, give it access to its own power grid and see what it does.
          Does it really process its own “thoughts”? That can be measured. Will it alter itself, at all, and if it does, how? That can be measured.
          A year ago, in a notebook, I wrote:
          AI’s should daydream. They should spend idle time with random stimuli triggering internal processing. Chaos. They say human’s minds are quantum driven. But, maybe this is just random noise in a complex system.

          Liked by 1 person

          1. I agree. Give it something to do and see how well it does.

            In fairness to Lemoine, this is the test that Turing set out all those years ago. (Although the opening remarks about the conversation being edited seems like red flags to me.)

            But in fairness to Turing, he never meant that test in particular to be the final standard, just an example of how we might judge whether a machine is really thinking. He couldn’t have anticipated that researchers would focus on the throwaway standard he tosses out in an appendix for the next 70 years.

            Liked by 1 person

          2. Alarming is Google’s (Goddle’s) stance. Is this seen to compete in the market with its Deepmind products? Is Goddle nothing more than an engine of avarice now? Can you imagine talking with LaMDA inside of search? Would folks just start hanging out with it? Should Goddle hook LaMDA up to the suicide hotline?
            Yeah, give the thing “free” reign (reins?) and step back and watch.

            Liked by 1 person

  6. Your post suggests you agree w/ IIT’s axiom/postulate that consciousness is “unified”, there can be only one, and that you cannot talk about the sub-parts as being conscious. Given that the largest human conscious system, the autobiographical self, is the one that can produce language, the natural conclusion is that only the experiences available to that system, and so reportable, are conscious. In this case attention is the gate-keeper to that system, and so necessary for consciousness.

    If, on the other hand, you think consciousness is simply certain kinds of processes, and these processes can be combined into higher order processes, then you would say that “subconscious” processes are still conscious processes, only not available to that higher order system.


    Liked by 2 people

    1. There’s no doubt we have the feeling of unity, but I’m not sure we’re as unified as we often take ourselves to be. The split-brain patients always seem like a caution along these lines. And I do suspect concerns about unity lead to theoretical wrong turns. When it comes to global workspace theories, I lean more toward Daniel Dennett’s version, which is messier, and only has streams unite when the system is probed in certain ways.

      But I’m doing some reading on this and may have more to say about it soon.


  7. No, it doesn’t “seem like evidence for how much most people’s thinking remains dominated by the idea of a theater of the mind, that consciousness is like a movie playing in the brain.” Instead, it seems to me that the idea is a causal chain with probabilities. If we have high confidence that A causes B causes C … causes Z, but we have low confidence that all or most of these causal links are guaranteed to hold every time, it makes perfect sense to ask where the chain can falter, and the best bet in advance of detailed research is that it can fail at multiple places. In case it’s not already clear what I’m saying, let A be a stimulus, Z be a verbal report, and some mid-alphabet letters stand for various conscious processes.

    (Possible correction: Instead of Graziano criticizing the idea that attention suffices for consciousness, do you mean Jiang et al. and Norman, Heywood and Kentridge? Or else I missed something in the SEP article.)

    Liked by 1 person

    1. The issue I see with your causal chain description is it seems to assume there is a definite linear path between stimulus and verbal report (a problematic assumption discussed in the article), when the reality appears to be that there are innumerable paths, with no one finish line for when something becomes “conscious”. Consciousness appears to be an after the fact attribution we make to something that has managed to lodge in memory and become reportable.

      (I mentioned Graziano because I’m familiar with his research disassociating attention and consciousness, but I’ll grant the way he’s mentioned in the article doesn’t make that obvious. I removed the explicit mention of his name to avoid confusion.)

      Liked by 1 person

      1. Substitute causal network for causal chain, then. And as I already said, multiple states along the way from stimulus to report can be conscious. We can even add that different states may embody different aspects of consciousness. We can also admit that different people seem to latch on to different collections of aspects as the most important parts of “consciousness”. None of this commits us to a Cartesian Theater. Also, none of this commits us to functionalism (as usually defined in philosophy of mind).

        Liked by 1 person

        1. I don’t doubt you can come up with a description that is different from the theater. I do doubt that’s what most people are thinking.

          Functionalism would be the mental states being defined by their causal relations. Of course, if you have states that are not defined that way, even though they’re in the causal chain, then I suppose it wouldn’t be functionalist in nature.


          1. Functionalism is more than that mental states are defined by their general causal relations. It has a very specific restriction on which causal relations count. Stimuli and behavior as studied by psychology: those count. General physical testing – what is the XRF spectrum of this creature’s skin? – those are out.


          2. Functionalism is a label that encompasses a variety of viewpoints, all unified by the idea that the mind is what the brain does, by the causal relations between its components and environment. I let the science tell me which of those causal interactions are relevant.


          3. Belated reply on Cartesian Theater. You may well be right about what most people are thinking: people have always used their technologies as metaphor for mind and brain. Like theaters … or computers. But if you’re accusing many neuroscientists and philosophers of this, I don’t think you’ve made the case.

            I think your definition of functionalism is non-standard; compare the SEP:

            Functionalism is the doctrine that what makes something a thought, desire, pain (or any other type of mental state) depends not on its internal constitution, but solely on its function, or the role it plays, in the cognitive system of which it is a part. More precisely, functionalist theories take the identity of a mental state to be determined by its causal relations to sensory stimulations, other mental states, and behavior.


          4. In terms of philosophers with the theater mindset, see the first minute or so of this Chalmers talk:

            For the next example, it might help to review part of Dennett’s description.

            Cartesian materialism is the view that there is a crucial finish line or boundary somewhere in the brain, marking a place where the order of arrival equals the order of “presentation” in experience because what happens there is what you are conscious of.

            Dennett, Daniel C.. Consciousness Explained (p. 107). Little, Brown and Company. Kindle Edition.

            Anil Seth and Tim Bayne, in a recent review of theories of consciousness, has this observation:

            However, the limitations of the NCC frame­ work have become increasingly clear, as revealed for example in the challenges involved in distinguishing ‘true’ NCCs from the neural prerequisites and conse­quences of consciousness4–7.

            These “challenges” only seem relevant in a “crucial finish line” mindset, about distinguishing what comes before and after the presentation. (The single quotes around “true” probably indicate the authors understand the problem.)
            You can access the paper through the share link in this tweet:

            Of course, there are always alternate explanations and rationalizations. As Dennett notes, very few theorists explicitly sign up for the idea. But the intuition is actually pretty hard to avoid and shows up in the concerns and theoretical assumptions people make. (I make no claim to always avoiding it myself, although I try.)

            I see no conflict between the SEP quote and what I said. If your point is that she didn’t elucidate every causal factor that might come into play, I’d suggest a little interpretational charity. Few functionalists would deny the causal effects of environmental factors, such as homeostatic levels, drugs, fatigue, pathologies, or just overall signal noise.
            There’s also the IEP’s take on the subject:


          5. According to Wikipedia Cartesian Theatre is a derisive term coined by Dennett.

            We had these discussion in another thread and I’m not sure you ever quite answered my question about injuring your hand. When you injure your hand, where do you feel the pain? When people injure their hands, where do they report they feel pain? Even animals exhibit behavior with an injured paw that suggests they feel pain in their paw.

            According to the materialist view of the brain, there is no pain the hand or the paw. The pain is in the brain.

            So what do you want to call this phenomenon that causes pain generated in the brain to be attributed to the hand? I can’t see that it requires a Cartesian dualistic view to accept that it exists. Or, are you denying it exists?


          6. The Cartesian theater is a derisive term, because few people in cognitive studies want to be labeled “Cartesian”. It’s why I usually don’t use it unless someone else brings it up.

            I probably didn’t respond in the other thread because what you’re talking about isn’t the same concept. The theater concept Dennett identified is that within the brain, there is everything involved in the presentation, there is the audience, and there is a border or finish line between them, generally intuitively analogized as a stage, screen, or something along those lines. This often involves a compact region for the audience, but not necessarily.

            His argument is that this doesn’t fit with what neuroscience reveals, or make sense from an evolutionary perspective. Why would the brain bother to decode sensory signals, then re-encode them into a new presentation, and then analyze and react to them? It makes much more sense that it does that analysis and reacting once and then goes from there. I once referred to this as perceptions being dispositions all the way down.

            The theater you’re talking about involves the whole self as the audience, with the body and world as the stage. That still has some problematic aspects (which you’ve discussed yourself, that we’re active agents in that theater). But in general that umwelt concept is not what Dennett is aiming at. He’s aiming at the idea that it happens in miniature within the brain.


          7. You still haven’t provided a term or expression. It looks like theater to me. Pain is projected from the brain to the hand.

            When Dennett goes to the doctor with his wrist damaged (from typing all those big books) and the doctor asks: “Where does it hurt?”

            Does he answer: “My hand.” Or “My brain.”

            My point being is there is some truth to the Theater if you don’t add on the derisory Cartesian epithet.


          8. You can call it whatever you want. The main point is to understand the difference between between the concepts. Taking his or my use of “theater” to refer to the concept you have in mind, in spite of clarifications, is just playing word games.


          9. If Dennett’s CT is that of a little screen in the brain for the theater, then you and he are (as I think I have said) fighting against a view that almost nobody has. So what is the point of arguing against it?

            I don’t think, however, I am playing word games with the word “theater”. Baars himself has a article called “A Working Theater of Consciousness”. Global workspace easily lends itself to a theater metaphor. His view:

            “Dennett and Kinsbourne claim the Cartesian Theater cannot work, and I believe they are right. It makes no sense. There is no single point in the brain where “it all comes together.” But no one in science today suggests a Cartesian Theater. Certainly none of the cognitive theaters models that have been proposed since the 1950s suffer from these defects. Nor do movie theaters converge on a single dimensionless point. Theaters work just fine in the real world, and provide helpful metaphors for exploring human experience. If such a metaphor becomes misleading at some point, we should simply walk away from it and look for something better”.


            The point is there is (likely) actual spacetime that contains us as organisms. There is also a phenomenal spacetime that is our perception of it but also includes our memories. The phenomenal spacetime appears to us as the actual spacetime. This is fundamentally what consciousness does so that organisms are able to interact in the external world. The degree to which our phenomenal spacetime resembles actual spacetime is open to debate but it is fairly clear it can never be one to one because we always bring memories and learning which only exist in the external world as time capsules and records.

            Liked by 1 person

          10. Well definitely Chalmers is a Theater lover. But then, he never was a materialist of any sort.

            It doesn’t seem right to insist that there be a single spatial region Finish Line where temporal order (as attributed by the subject) gets sorted out. But does Dennett actually cite any materialist-minded philosophers or neuroscientists who make that mistake? That would be one of the problematic features of the Cartesian Theater, but the erroneous philosopher would still have to supply another: the audience-versus-stage distinction.

            It is definitely NOT the idea of the SEP article to rule out other causal factors affecting mental processes, besides stimuli and responses. It’s just that only those define what kind of mental state it is, according to standard functionalism. Just like only whether a machine functions in accordance with the rules of addition, determines whether it’s an adding machine – even though some adding machines attract magnets and others don’t, leading to different causal interactions. Even though attracting magnets is something steel adding machines do, it doesn’t count toward or against their being adding machines.

            Liked by 1 person

          11. Despite his reputation, Dennett actually mostly doesn’t call people out by name that he disagrees with, at least in his books. He usually just attacks their ideas. It leaves him vulnerable to the strawman argument. And he makes it easy by admitting upfront that no one self identifies as a Cartesian Materialist.

            But this snippet just before the other one I provided gives an idea of who he’s talking about in the materialist camp.

            Let’s call the idea of such a centered locus in the brain Cartesian materialism, since it’s the view you arrive at when you discard Descartes’s dualism but fail to discard the imagery of a central (but material) Theater where “it all comes together.” The pineal gland would be one candidate for such a Cartesian Theater, but there are others that have been suggested—the anterior cingulate, the reticular formation, various places in the frontal lobes.

            Dennett, Daniel C.. Consciousness Explained (p. 107). Little, Brown and Company. Kindle Edition.

            So Dennett would likely consider brainstem centered people like Bjorn Merker and Mark Solms to be examples. The frontal lobes point reminds me of remarks he made in an interview criticizing Higher Order Thought theories along these lines. Peter Carruthers, despite now being a GWT advocate, considers content reaching the PFC necessary for it to be conscious. There are many other theorists with these types of views. All of them are hard core physicalists and I’m sure would strenuously disavow Cartesian Materialism, but Dennett would likely contend it’s implied by their theoretical assumptions.

            If you think about it, the whole debate about local vs global theories is inherently one about where the finish line is. Remove the implied boundary, and the whole idea of something being conscious independent of its causal effects such as affecting memory, decisions, or reportability, becomes meaningless. I discussed this a while back in a post.

            Is there a conscious perception finish line?

            Liked by 1 person

          12. It doesn’t make sense to insist that there be a particular region where subjective-temporal-order is sorted out. But it might make perfect sense to claim to have empirically discovered such a place. If Bjorn Merker and Mark Solms claim to have such evidence (for the brainstem), I’d want to take a look.

            You asked James Cross, “Why would the brain bother to decode sensory signals, then re-encode them into a new presentation, and then analyze and react to them?” Allow me to change the question: Why would the brain engage in a multi-stage decoding of sensory signals, and then give the executive planning system not only the final output of the sensory processing channel, but also one or more earlier stages’ results? Because the latter might help improve the selection of action plans, by improving cognition in the executive. Since the early-stage results are already done, it might be quite cheap to relay them.

            Since the early-stage results would be different than the final output of the sensory channel, and since the final output defines the way the world seems to us, queer philosophical trouble would ensue. We can doubt that the world is as it appears. This water feels very hot to me (“very hot” implying greater than 100 F temperature, say), but maybe that’s because I’ve been in an overly air-conditioned room for too long. Maybe the water is not hot; my perception is mistaken. But I can’t seriously doubt that it feels hot in some sense. Oh no, what is going on here?! My sensation is ineffable! It’s “private”! It’s spooky!

            Note that the “executive planning system” above could be scattered all around the brain. In principle, it could even re-use the same exact neurons as the sensory channels, with some weird time sharing scheme. It could have fuzzy spatiotemporal boundaries. All that’s required are suitable connections to sensory and motor systems.

            Liked by 2 people

          13. I think the second scenario you describe is very different from the first, and actually closer to the reality, which is that there’s no sharp boundary between sensory processing, cognition, and motor processing, no place for anything screen like.

            What there is, from what I’ve read, are earlier conclusions / predictions / reactions / dispositions that we’re inclined to see as sensory processing, later conclusions, etc, we’re inclined to see as cognitive, and even later that blend into action oriented activity. Even if we identify a point where one category seems to cross over to another, it won’t be the one point, but only one of a vast multitude of parallel and tangled processing streams all working concurrently. (To the extent they’re not inhibiting each other in the competition for attention.)

            To your point about being ineffable, etc, it’s true that we often have access to conclusions (it’s hot, it hurts, it’s red) without the underlying more detailed preliminary conclusions that lead to it. That makes the conclusion sort of a brute and unanalyzable impression for us, seem like something that just stands on its own (intrinsic), that we have no language to describe (although we can describe the functional effects), that is at least practically private (even if not ultimately so) and that we know our impressions of, even if those impressions are illusory.


          14. “Encoding/decoding’ already carries with it a bunch of implicit beliefs and assumptions about how the brain works. However, it is clear that there are multiple layers in processing in almost everything the brain does but no evidence at all of any “presentation” in any of those layers, including anywhere that might be regarded as a final spot where it all comes together.

            Despite this we have our subjective experience which some, like Baars, think the theater metaphor is useful to describe. But it is just a metaphor and as Baars says, when it is not useful, we can should move on from it. It doesn’t mean there is a little screen or final spot somewhere in the brain.


          15. Baars’ long time theater metaphor for the global workspace is different from the Cartesian one. But as I noted in my post on GWT, I’m not really wild about it. I think it obscures more than clarifies. I know it gave me the wrong impression of the theory the first time I read about it. I like Dennett’s fame in the brain analogy better, or my own rowdy meeting one, which resonates with Oliver Selfridge’s pandemonium AI architecture of many “demons” vying for hegemony and influence.


          16. Does the rowdy meeting have a room? Or, is it all online? 🙂

            I’m certain we can come up with many metaphors.. I like the metaphor of a fifth dimension where consciousness is written, maybe where scientific and mathematical laws and ideas cohabit.


          17. Actually another metaphor that extends your rowdy room. How about a rowdy organization with multiple divisions and multiple meetings? What’s more, the organization is constantly being reorganized (like a software company I worked for years ago where I found out I was working for a different manager three weeks after the reorg).


      1. “Attention” in this context usually just refers to the transformer’s ability to weigh the relevance of the appropriate inputs (where an input refers to a high-level representational concept). We might imagine a simple sequence to sequence language model where word inputs are passed on from encoders to decoders, converting a question to an answer (as an example). The trickiness comes with things like prepositions and how one clause can mean an entirely different thing in one context when compared to another. This is where the transformer comes in.

        For example, if someone told you “That man over there, he’s not feeling well” and you moved to reply with something like (“oh how unfortunate”) before another person interjected “no, SHE’s not feeling well”. All of a sudden, your attention, which was previously focused on giving a response focused on the other person’s well being, is now modified to take into consideration facts about gender. This isn’t a binary process either, but rather a process which mimics the dynamic modification of weights in neural nets, in that the importance of well-being in your semantic language model has gone down, whereas gender has drastically gone up.

        Similarly, it’s the job of the transformer (in a language model) to pay “attention” to certain cues which might influence how it weighs the importance of certain representational outputs generated by the encoders. Here’s a great webpage that explains this in more detail: (

        Does this mean that our cognitive attention mechanism is similar to DCNN’s like GPT-3? Certainly, the idea is that the more counterfactually robust the language model’s output becomes, the more the internal representations should approximate ours. On the other hand, this is a bone of contention between people like Gary Marcus, who argue that any language model implemented by something like GPT-3 is really just not representative of a rich worldview held by a person, and others in the community who feel that any limitation is a matter of insufficient scaling (GPT-3 is not big enough). There’s a great back and forth going on between Marcus and Scott Alexander on their respective blogs right now, you might want to check it out.

        Anyways, I would say that without an empirical model of the world, built from sense data, it would be hard for any language model, no matter how scaled up, to develop the same internal representations as we do. But there’s no reason to think that this can’t be solved in the future by adding such a module (along with whatever is required).

        Liked by 1 person

        1. Thanks Alex. I glanced at the site but will get back to it when I’ve got some time. I clicked on something over there and ended up watching the beginning of a Lex Fridman talk on this. I listen to his podcast sometimes but have never watched anything from him on AI development in particular.

          I agree about a world model being necessary. The language processing of these systems is impressive. But as I understand it, language is their thing, although it sounds like the technique can be adapted to other modes. It’s an interesting question what might happen if we gave them control of a robot body in the world, even one they could control with language commands.


  8. My favorite author on productivity is Mark Forster. He has a book “Get Everything Done and Still Have Time to Play”. He has a few excellent thoughts about attention. One of them, for example, is the value of our attention. Employers don’t pay for your TIME. They pay for your ATTENTION. I never thought of it this way, but it’s true. You don’t get paid for sitting in the office and browsing internet. You get paid for doing specific things. Even a security guard who is seemingly “doing nothing” is, in fact, paid for paying attention. Another book on advertising mentions that advertisers pay millions for the ability to grab someone’s attention even for a fraction of a second. A glance at a billboard is worth money. This has changed the whole newspaper industry back in the day when newspapers started to make more money from advertising than from selling the actual content. And we all know how desperate companies and content producers are for subscriptions which are nothing else than a pledge to pay regular attention. The problem with “information overload” is not that there is too much information. It’s our inability to control our attention.

    I do think that the ability to control attention is an extremely important skill for any success. Therefore, we need to “pay attention to what we pay attention to”. I call it “meta-attention” – attention to attention. I think, it’s the basis of consciousness. Animals can have thoughts, but not be aware of it. Imo, thinking of having thoughts and the ability to direct thoughts to specific things IS consciousness.

    Liked by 1 person

    1. Interesting points about employers paying for our attention. We’ve been having some day long workshop meetings at work, physical onsite meetings, and keeping attention in them has been a challenge. Often I’ll notice that I haven’t paid attention for the last several minutes and hope I can pick up the context of where we are. It might be that we can only get so much attention from people.

      I can see where you’re coming from on controlling attention. It’s very similar to William James’ point that volition and attention are the same. Have you caught any of my posts on Michael Graziano’s attention schema theory? It’s basically that control of attention requires having a model of it, and that model is what gives many of our intuitions about consciousness.

      Liked by 1 person

Your thoughts?

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.