Monday, 2 July 2012

John Campbell: What does Visual Experience Have to do with Visual Science?

    Abstract: The epistemic role of consciousness in sensory experience. Classically, vision science assumed we do not need to appeal to notions relating to sensory awareness to explain how it is that perception generates knowledge of our surroundings.  Sensory experience has often been seen as an epiphenomenon in the generation of knowledge.  This is not the view of ordinary common-sense.  Ordinarily, we take that it is only because we have sensory experience that we can know what the objects and properties around us are.  But where would a role for sensory experience fit in an account of the production of knowledge?  I approach this question by looking at the contrast Huang and Pashler (2007) draw between the roles of visible properties in selecting and in accessing regions or objects in the visual field.  I use this to articulate an account of the way in which an externalist account of perceptual experience relates to a classical account of visual computation.

    Huang and Pashler, 2007, 'A Boolean Map Theory of Visual Attention', Psych. Review 114, 599-631.
    Campbell, John, 2011, 'Visual Attention and the Epistemic Role of Consciousness', in Mole, Smithies and Wu (eds.), Attention:  Philosophical and Psychological Essays (Oxford: OUP), 321-343. [PDF will be provided]

Comments invited


  1. Seeing is not knowing, it's just feeling. What you you see is only probably true about the world. The only thing you know for sure is that that's what it feels like.

    1. As you mentioned in your lecture Sunday, language requires a foundation to allow for any understanding. Likewise, if we can't believe what we perceive, what can we really accomplish?

      You seem to believe that experience in this fundamental sense is all that we can know, and I agree, but only to an extent. If we have nothing to rely upon besides subjectivity, which is unreliable by definition, we cannot prove anything.

      If it is confirmed over and over again that something is true -- take Newton's F=MA -- we decide to call it a law so as to use it to produce new theorems and new laws. This is a game of probabilities and repeated confirmation, and sometimes we're just plain wrong. To dismiss perception as a footnote to phenomenal experience is to defeat the purpose of inquiry and of wonder -- this notion taken to its limit very nearly defeats the purpose of the summer institute itself.

    2. Stevan, the concept of seeing is factive. You can't see that Sally is running if Sally isn't running, you can't see Sally unless Sally is there. If your evidence for the proposition that Sally is running is that you see Sally running, then the probability of the proposition, given your evidence, is 1.

      The talk of 'feeling' in this context needs a lot of explanation. What do you mean? We'd ordinarily contrast seeing with feeling: seeing the kettle involves sight, feeling the kettle involves touch. If you mean 'feeling' in the sense in which one might say, 'Love is more than a feeling', seeing is more than a feeling too. In fact, seeing doesn't seem to involve feeling at all. What you see might make you feel concerned, for example, but the seeing is one thing and the feeling another.

      It's a bit weird to say 'seeing is just feeling'.


      John, it feels like something to see (or hear or smell or taste or touch) something, just as it feels like something to fear or to hate or to love.

      It's just a linguistic quirk that we say we "feel" objects with our hands but we "see" them with our eyes. (English uses "feel" for touch; French uses "feel" for both touch and smell.) ("Sensing" is more polymodal, but also more ambiguous.)

      As to the "factivity" of seeing: I am seeing Sally running if (1) it feels like I am seeing Sally running, (2) Sally is running and (3) I am looking at Sally, running.

      I know that the case is trickier for "I know Sally is running" -- and gets trickIer as it gets Gettier -- but I am not talking here about logical certainty (p = 1), just probability on the evidence; and about seeing and hearing things, not affirming or denying propositions (though it applies there too, because there's something it feels like to believe that P).

      And the fallible probabilistic sense of seeing (or hearing or smelling or tasting -- or believing or wanting or understanding or meaning) is all that the neurobiology (or robotics) of consciousness needs or wants.

    4. Stevan, seeing is one thing and feeling is another. You are running together the feelings that one might have on seeing someone, with the seeing itself.

      You've abandoned the idea that 'seeing is just feeling' and now want to analyze seeing X as looking at X plus having a feeling. The distinction between looking at X and seeing X is subtle, but there are cases, for example I might be looking right at you but not see you, say in that I didn't recognize you. However, merely recognizing someone is one thing, and the feelings that one might have on recognizing them, from rapture to dismay, are another.

      Feelings don't have any place in the analysis of what seeing is. It feels like something to win a gold medal at the Olympics. But that doesn't mean that we have to talk about feelings in explaining what it is to win a gold medal at the Olympics.

    5. WHAT IT FEELS LIKE TO SEE (1 of 2)

      John Campbell: "Stevan, seeing is one thing and feeling is another. You are running together the feelings that one might have on seeing someone, with the seeing itself."

      John, I'm not talking about the feelings you have when you see someone, I'm talking about what it feels like to see someone, or see anything (or hear someone/something, etc.). Sensory experience in every sense-modality feels like something. If it is unfelt, it is not seeing but mere optical transduction, detection, processing, as in a robot.

John Campbell: "You've abandoned the idea that 'seeing is just feeling' and now want to analyze seeing X as looking at X plus having a feeling."

      No, I really mean what it feels like to see X. ("Looking at X" is rather equivocal. What you mean is "directing your line of gaze at X," which you could do, presumably, when your eyes were closed, or even perhaps while fast asleep.)

      If you are seeing X, then you are feeling something. And what you are feeling is what it feels like to see X.

      The only sense modality in which I don't have to do this tortured circumlocution is when you are touching X. There too you are feeling X, where what is meant is what it feels like to touch X.

      Feeling X via touch is the same sort of thing as seeing X via sight. And both are different from the sense of feeling when I feel tired -- not to be confused with feeling *that I am tired*: both being feelings, but not quite the same feeling.

      It also feel like something to be looking at X (awake, eyes open, and paying attention) when you don't recognize that it's X. Hence you're seeing then too, and it feels like something to be doing that seeing. It just doesn't feel like seeing X.

      And if, in a few seconds, I recognize or remember that it's X, then it feels different from what it felt like when I was looking at X, and seeing someone or something, but it didn't feel like the someone or something I was looking at was X.

      (part 2 follows)

    6. WHAT IT FEELS LIKE TO SEE (2 of 2)

      John Campbell: "The distinction between looking at X and seeing X is subtle, but there are cases, for example I might be looking right at you but not see you, say in that I didn't recognize you."

      As I said, your line of gaze may be aimed at me, but if you are fast asleep or your eyes are closed, you are not seeing at all. And if your eyes are wide open but you are listening so intently to something that you are not seeing me (or maybe not seeing anyone, or anything) then you are not seeing. But if you are seeing anything at all, you are feeling, because it feels like something to see. It's just that what you are seeing may not look to you like me (and hence it feels like something, but not like seeing me).

      John Campbell: "However, merely recognizing someone is one thing, and the feelings that one might have on recognizing them, from rapture to dismay, are another."
      Feeling rapture or dismay are like the example of feeling tired that I mentioned above. They can accompany my seeing you, and they are feelings too, but they are not the same feeling -- or rather, although I don't really want to play phenomenological games, dissecting feelings into parts -- they are not the same "part" of the total feeling I am feeling when I am looking straight at you: I may simultaneously be feeling tired, stressed, under pressure from a tight neck-tie, smelling the dinner that's almost ready, seeing that the sun is setting -- and also seeing something that either does or does not feel like I am seeing you, depending on whether I am or am not recognizing you. (Needless to say, it's also irrelevant to lapse into pondering whether perception is phenomenologically serial or parallel…)

      So just as it feels different to see you when I'm feeling tired, compared to seeing you when I'm not tired, it feels different to see you when I recognize you compared to seeing someone whom I don't recognize to be you, but it's you.

      In other words, it also feels like something to recognize you -- another add-on to what it feels like to be looking at you.
      John Campbell: "Feelings don't have any place in the analysis of what seeing is. It feels like something to win a gold medal at the Olympics. But that doesn't mean that we have to talk about feelings in explaining what it is to win a gold medal at the Olympics."

      Not in explaining what it is to win a gold medal in the Olympics. That's just doing (Turing-robot style).

      But feelings certainly have a place in explaining how and why it feels like something to win a gold medal in the Olympics.

    7. Stevan, you began by saying 'seeing is just feeling'. Then it seemed that you wanted to analyze seeing as feeling + something else. Now you seem to have abandoned the attempt to analyze seeing. But you still want to hold on to the idea that feelings are somehow constitutive of seeing. I can't find any evidence for that at all.

      Seeing is a relation between the perceiver and the scene observed. It's a subjective, experiential relation. That doesn't mean that 'feelings' are involved. Certainly introspection provides no support for the idea that feelings are involved. When I introspect my experience of seeing, what I encounter is not some 'feeling', but the ordinary concrete objects, the tables and chairs, and their properties.

      You make much use of the phrase 'what it feels like to see'. The fact that there is such a thing as 'what it feels like to X' doesn't show that there are any feelings that are constitutive of X-ing. The most it can show is that there are feelings that accompany X-ing.

      I'm concerned with the experiential relation of seeing. 'Feelings' are no doubt very important. But they're a different topic.


      Well we have a clear, crisp disagreement, John:

      I think seeing is what it feels like to see, hearing is what it feels like to hear, smelling is what it feels like to smell. (I have thought and said that all along; I don't thing I changed a thing!)

      I agree completely that feelings are only a correlate. But what is left, if you remove that correlate, is just doings (not seeing, hearing, or smelling).

      And that is the (hard) problem.

    9. Excellent. The question is this. Should we characterize the subjective dimension of seeing as (a) an experiential relation that you stand in to the scene observed, or (b) a feeling?

      We agree, I think, that (b) leads to the hard problem of consciousness, and in fact the conclusion that experience has no evident causal role.

      The point about (a) is that it acknowledges a causal role for consciousness, in giving us knowledge of our qualitative surroundings, and frees us up from the idea that we have to explain the qualitative aspects of experience in neural terms. That was the point of my talk.

      The idea that 'seeing is just feeling' does not seem to me to be an insight, but a source of insoluble problems. Why stick to it?
      For anyone who wants a quick introduction to the alternative, but doesn't want to listen to the whole lecture, there is a brief introduction here:


      Seems simpler to just say it hurts...

  2. Campbell's focus on color makes me think: even within a single sensory modality, there is still integration of different features before our conscious experience of an object. During yesterday's discussion, an audience member asked about robots integrating sensory information from different modalities. To me, this question was getting at the idea that one function of consciousness may be to integrate complex sensory stimuli. Now I consider that complexity does not require different sensory modalities, but can come from different features of the same modality.

    1. Martha, I strongly agree. It's a very natural idea that consciousness has something to do with the integration of the senses. But it would surely be pressing the idea too hard to say that you couldn't have consciousness unless you had more than one sensory modality.

      There is, of course, a lot more to say about this. You might argue that the reason why consciousness evolved was because it can function in integrating information from different sensory modalities. That would be consistent with your point that an individual creature with only one sensory modality could have enough complexity to sustain consciousness.

  3. I found the epistemic role of conscious visual experience to be a fascinating theme for discussion. Indeed, while phenomenological methods have increasingly been operationalized in cognitive science (either implicitly in qualitative research methods or explictely, in approaches like neuro- and front-loaded- phenomenology), very little has been in the opposite direction, namely: the “phenomenologization” of cognitive science.

    1. Maxwell, thanks for that. I think this area is wide open for further discussion. Traditional epistemology, and common-sense itself, take it absolutely for granted that perceptual experience is the foundation of knowledge. In cognitive science, though, we find not only that (a) there isn't any working role given to experience, but (b) the concept of knowledge itself seems to drop out. All cognitive science needs is the notion of a probability being assigned to a proposition by the system (if that).
      It seems to me most unlikely that experience and knowledge should be epiphenomena or some kind of illusion. But that means the challenge is to explain what experience and knowledge have to do with the phenomena described in cognitive science. That's the question that is wide open.

  4. As I understand him, one of Campbell's main motivation for defending an externalist theory of phenomenal experience comes from his dissatisfaction with the state of research on the nature of consciousness. (Remember what he says at the beginning of the talk: the 17th-century project of 'analyzing the qualitative world in terms of its relation to consciousness' has been a massive failure, whereas the project of explaining physical phenomena in terms of mathematical laws has been an amazing success.) But it seems to me that it is too early in scientific research to take this massive failure as a reason to adopt an externalist theory (which, in my mind, doesn't really explain anything about consciousness, but only presupposes it exists). Newton didn't have the theoretical tools to explain the photoelectric effect in his time, but it didn't mean he should have simply ignored it had he been aware of its existence. (That being said, Campbell has other arguments to defend his externalist theory. So this is certainly not meant as a knock-down argument or anything of the kind.)

    1. Alexandre, there are two questions here. One is whether the project of analyzing the qualitative world in terms of its relation to consciousness has been a massive failure. I'm not sure whether you're willing to grant that it has. My point about that was that after all this time we don't have a coherent model of consciousness that lets us see how consciousness could 'project' the qualitative world onto an underlying physical reality.
      Talk about 'qualia', 'sensations' and so on seems incoherent, because on the one hand qualia are theoretical characteristics or objects, postulated to explain something about experience, they're not something we know about on the basis of experience because experience only even seems to give us knowledge of characteristics of the world around us, but we're never given any explanation of what qualia are; and on the other hand, they're supposed to be the most immediately known and understood aspects of the world, and hence not to need any explanation.
      Talk about 'representation of (e.g.) color' as the basis for the projection is currently the most popular option, but again it makes no sense. The only theory of representation that anyone has been able to think of is a causal theory. But with no colors 'out there' to cause the representations, and no colors 'in here' to cause them, we have no way of explaining what it means to talk about representations of color at all.
      We need something better than these tired models. The second question is whether the externalist approach I recommend is the only alternative. I think it would be great to explore further possibilities. What are they?