Monday 2 July 2012

David Freedman: Brain Mechanisms of Visual Categorization and Decision-Making


      Abstract: We have a remarkable ability to recognize the behavioral significance, or category membership of a wide range of visual stimuli. While much is known about how simple visual features (such as color, orientation and direction of motion) are processed in early stages of the visual system, much less is known about how the brain learns and recognizes categorical information that gives meaning to incoming stimuli. This talk will review a series of neurophysiological and behavioral experiments aimed at understanding the neuronal representations underlying visual categorization. We have found that the activity of individual neurons in both the posterior parietal and lateral prefrontal cortices can reflect the learned category membership of visual stimuli, and that these two areas play distinct roles in category-based decision making.

    Freedman D.J. and Assad J.A. Experience-Dependent Representation of Visual Categories in Parietal Cortex. Nature, 443: 85-88, 2006.http://www.cns.upf.edu/jclub/freedman_assad2006.pdf
    Swaminathan S.K. and Freedman D.J. Preferential encoding of visual categories in parietal cortex compared to prefrontal cortex. Nature Neuroscience, 15: 315-320, 2012.http://monkeylogic.uchicago.edu/Swaminathan_Freedman_nature_neuroscience_2012_with_SuppInfo.pdf
    Freedman D.J. and Assad J.A. A Proposed Common Neural Mechanisms for Categorization and Perceptual Decisions. Nature Neuroscience, 14:143-146, 2011.http://library.ucls.uchicago.edu/FirstDay201112/learning/categorization%20and%20perceptual%20decisions.pdf
    Freedman D.J., Riesenhuber M., Poggio T., and Miller E.K. A Comparison of Primate Prefrontal and Inferior Temporal Cortices During Visual Categorization. Journal of Neuroscience, 23: 5235-5246, 2003. http://www.neuro.cjb.net/content/23/12/5235.full

Comments invited

10 comments:

  1. Freedman: L.I.P. neurons categorizing as well as the subject does.
    Haggard: before, during or after the subject feels like doing it...

    ReplyDelete
  2. Doctor Freedman has very precisely shown quite interesting activation profiles for PFC and LIP neurons and I am quite convinced by his presentation. I would be interested to know whether PFC and LIP neurons are active differently while the organism is learning new categories. Indeed, while the category selective profiles of the neurons in these areas are interesting in and of themselves, I am intrigued by the dynamics of encoding new categories, as opposed to the functioning of already-learned ones. I imagine, however, that empirical evidence for the mechanism through which such a complicated function is realized would be immensely difficult to observe and measure—any idea how one would go about it?

    ReplyDelete
    Replies
    1. This is a very good question, and one that we are very interested in pursuing in our ongoing and future work. We are currently working on studies in my laboratory in which we are recording from parietal and prefrontal neurons during the category learning process itself.

      One recent study from Earl Miller's laboratory at MIT examined prefrontal and striatum during a different category learning task. Here's the reference:

      Antzoulatos,E.G. and Miller, E.K. (2011) Differences between neural activity in prefrontal cortex and striatum during learning of novel, abstract categories. Neuron. 71(2): 243-249.

      Delete
  3. This is very cool research, but it was not made clear (at least not to me) why this topic is relevant to a summer school on consciousness. I hate to be ignorant; does anyone else have some insight into how it's connected?

    ReplyDelete
    Replies
    1. Freedman extensively discussed categorization, specifically in the context of neurones in the PFC, ITC, and LIP. LIP neurones were much more specialized for a binary response, as were PFC neurones to a lesser extent (ITC neurones deal with a prototypical category that they fit the stimulus to). As interesting as this is, it's true that the research has little to do with consciousness. Perhaps, as he postulated later on, the fact that judgement of ambiguous decisions changed the firing rate in LIP neurones specifically are integrally involved in the spacial attention and memory trace functions. Our concept of consciousness (at least mine) is predicated upon the ability to recognize time outside the present (memory trace) and selection of information (attention).

      Delete
    2. I guess the research described has more to do with decision-making and visual-categorization than with conscious experience per se, but one can suppose that the monkeys are reporting what they are consciously perceiving. And assuming this, the inferred perception correlates better with neurons' responses in LIP rather than PFC. (We've seen farther stretches from consciousness research than this in the last few days I think...)

      Delete
    3. Thank you for your comments. Our primary motivation for doing these experiments is to examine whether and how the brain learns and represents "abstract" information about the behavioral significance or meaning of sensory stimuli. For our purposes, abstract representations are those that are not directly linked with sensory input or motor output. And in our studies, we know that these abstract representations were learned over the course of weeks or months of behavioral training. The categorical representations that we have described are interesting to us since they provide a signal about the learned significance of sensory stimuli. And, the animals need to base their decisions upon these learned category rules.

      It has been interesting for me to hear everyone's point of view regarding the relationship between our categorization studies and consciousness. From my point of view, the link seems to be that the neuronal category signals that we have observed are likely very closely related with the animals' categorical perception and decisions in the context of the behavioral task that they perform in our experiments. Since these tasks require the monkeys to classify stimuli on each trial according to a learned rule (and sometimes for ambiguous stimuli), this seems to touch at least somewhat on the issue of conscious perception. As always, I'd be interested in any other thoughts from the group.

      Delete
  4. Unrelated to consciousness, the talk made me ask whether or not categories learned as an adult organism are represented by the same networks as those learned during development.

    ReplyDelete
  5. This is an interesting question. Unfortunately, we don't yet know the answer to this, since the experiments have not been done. However, an interesting related question is whether newly learned and highly familiar categories are represented in the same manner.
    This is something that we are beginning to examine in my lab, and we hope to have something to say about this in the near future.

    ReplyDelete
  6. (I thought I had posted this before, but it did not work apparently. So my post is a little redundant considering the last two comments.) Fascinating work, but I wonder whether the experiment involving macaques who had to distinguish motion direction with a single category boundary is relevant to understanding conceptual categorization (as it occurs when a subject distinguishes between a cat and a dog, to use one of the example that was given at the beginning of the talk). I see many differences between the kind of categorical learning involved in this experiment and the kind of conceptual learning that occurs in our everyday lives: (1) the monkeys learned to distinguish the categories over a very long period of time and a very large number of trials, whereas humans can often learn new concepts with a few trials, and sometimes in a few minutes or seconds; (2) rewards played a great role in the macaques' learning process (if they identified a motion direction properly, they had a reward), whereas humans generally don't receive a rewards when they learn new concepts. So it still seems possible that conceptual learning might activate very different parts of the brain than those that were activated during this experiment and that the monkeys acquired a sort of purely automatic discrimination ability.

    ReplyDelete