Thursday, 5 July 2012

Wolf Singer: Consciousness: Unity in Time Rather Than Space?

      Abstract: The search for neuronal correlates of consciousness (NCC) often relies on comparisons between neuronal activation patterns associated with conscious and non-conscious processing, respectively, of physically identical stimuli. This strategy is known as the subtraction method and thought to isolate neuronal processes specific for conscious experience. However, this approach does not allow one to clearly separate the NCC proper from processes that just permit access to consciousness such as fluctuations in excitability at early stages or from processes that follow conscious experience such as storage of perceived items in working memory and response preparation. This problem can be reduced but not eliminated by considering the precise temporal sequence of events, using methods that capture brain activity with high temporal resolution such as time frequency analysis and event related potentials extracted from EEG or MEG signals.
    Applying these methods we find as an early NCC a brief burst of oscillatory activity in the beta/gamma frequency range that occurs about 180 ms after stimulus presentation and is synchronized across a widely distributed network of cortical areas. This suggests as NCC not the activation of a particular, higher order cortical area but a dynamic state that is characterized by the coherent activation of a widely distributed network. This agrees with Baars and Dehaene's hypothesis of a work space and also with Sherrington's view that the unity of conscious experience does not require convergence in space (anatomical convergence) but results from coherence in time (temporal convergence, phase coherence). Indications for a special role of precisely synchronized oscillatory responses in the high frequency range have been obtained previously in animal experiments, using the paradigm of binocular rivalry.

      Fries, P., Roelfsema, P.R., Engel, A.K., Koenig, P., and Singer, W. (1997) Synchronization of oscillatory responses in visual cortex correlates with perception in interocular rivalry. Proceedings of the National Academy of Sciences of America. 94(23), 12699-12704
      Singer, W. (1998) Consciousness and the structure of neuronal representations. Phil. Trans. R. Soc. Lond. B 353: 1829-1840
      Engel, A.K., P. Fries, P.Koenig, M. Brecht, and W. Singer (1999) Temporal binding, binocular rivalry, and consciousness. Consciousness and Cognition 8: 128-151 
      Engel, A.K., P. Fries, P. Koenig, M. Brecht, and W. Singer (1999) Concluding Commentary. Does time help to understand consciousness? Consciousness and Cognition 8: 260-268 
      Melloni, L., C. Molina, M. Pena, D. Torres, W. Singer, and E. Rodriguez (2007) Synchronization of neural activity across cortical areas correlates with conscious perception. The Journal of Neuroscience 27(11): 2858-2865
      Melloni, L. and W. Singer (2010) Distinct characteristics of conscious experience are met by large-scale neuronal synchronization. In: E. Perry, D. Collerton, F. LeBeau and H. Ashton (Eds.). New Horizons in the Neuroscience of Consciousness. Advances in Consciousness Research 79. John Benjamins, B.V., Amsterdam 2010, 17-28
      Aru, J., T. Bachmann, W. Singer and L. Melloni (2012) Distilling the neural correlates of consciousness. Neuroscience and Biobehavioral Reviews 36(2): 737-746

Comments invited


  1. Prof. Singer proposes that saliency of a stimulus can be transmitted either by discharge rate (i.e. neural firing) or by gamma synchrony between neurones. If this synchrony truly is an alternate mechanism (as his data indicate), how does it appear at the appropriate times, only? It seems to me that preexisting circuits or perhaps neuromodulators could mediate this effect.

    Also, when Singer mentioned that Schizophrenics have a decreased observed effect in gamma synchrony, does this indicate a lesser form (or intensity) of consciousness?

    1. Regarding the schizophrenia patients: I don't know the answer to your question, but (if I remember correctly) Singer commented that they often misattribute stimuli, are confused and not aware of reality. I think this could be interpreted as a lower intensity of consciousness (of course depending on how we define consciousness).

  2. Personal comment: I really liked when W. Singer mentioned that studying neurons on an individual basis (through patch-clamp techniques) wouldn't have revealed to us that "macro"-neuronal gamma phenomenon depend on the firing of myriads of neurons. Very interesting!

    1. I dont think that these things are exlcusive. I think we know from the hippocampus recordings that individual neurons can spike by "riding" on a background oscillation. Also, i think this could be the way to understand the role of the brain stem structures: they provide that upregulation of the system, so different neurons in different areas can spike.

  3. Waaaay back in the 1960's, I attended a colloquium by a recent PhD who was looking for a computer science posting at the Univ. of Texas. (Alas, his name fell out of my head long ago). He had explored, via computer models, the computing power of neurons. He built a few simple models with a few neurons in each to discover how they might work together. His conclusions have stuck with me ever since - neurons are very poor at transistor-like logic: and gates, OR gates, etc. Neurons are amazingly good at anything to do with rhythms (which, of course, involves synchrony).

    So it came as no surprise to me when Dr Singer suggested that consciousness may actually be not a "noun" (i.e. a specific set of neurons) but rather, a "verb" (a specific kind of behaviour of a [very] large number of neurons).

  4. Dear Wolf,

    I am sorry to miss a chance to say hello to you in Montreal. I hope we can find another venue, some time soon.

    Spatial and temporal coding are not mutually exclusive, of course. Spatiotopical sensory, motor, and cortico-thalamic arrays seem to generate oscillations. You and your students have contributed very greatly to those insights, and you deserve a great deal of credit for making the world of neuroscience sit up and take notice. The idea that it takes 180 ms to build up the required recurrence seems perfectly plausible.

    Finally, a question: Antti Revonsuo and his coworkers seem to find a Visual Awareness Negativity around the P3b part of the ERP. Gaillard, Dehaene, et al, also seem to find something a little later than your 180 ms. Both of them use "contrastive analysis" experimental designs, comparing brain activity between closely matched conscious and non-conscious stimuli.

    I would distinguish that from the notorious "subtraction method," which I thought came from early PET and fMRI studies. I was taught in graduate school that subtraction between sample means (for example) was not a sensible method, since the variation within the two samples was not accounted for. That was the whole point about partitioning the variance in ANOVA and similar methods using the General Linear Model. In any case, I think the kind of sophisticated analytical tools used by Gaillard et al to pull causal relationships out of time series of brain events look a lot cleaner.

    By "contrastive analysis" I simple meant (and still mean) comparing two conditions as best we know how, methodologically.

    I also did not mean to rule out comparisons between "preconscious" and "post-conscious" moments in the flow of brain events. Nor did I mean to ignore simultaneous comparisons between cortical (e.g.) events believed to underlie conscious contents, and cerebellar (e.g.) events that are believed NOT to underlie conscious contents.

    We should just surround the problem from all sides, and let the evidence speak!

    One interesting possibility raised by the work of Dehaene, Gaillard, Revonsuo (and his coworkers) and others is that the "novelty P3b" involves a preconscious build-up of precisely the kind of "...dynamic state that is characterized by the coherent activation of a widely distributed network" that you suggest. Then at some point perhaps one finds a phase-change (Freeman), or a moment of ignition (Dehaene, et al), or a coherent activation of a widely distributed network.

    I know that a number of important people have made proposals along these lines, but we never quite believe it until we work out all the steps ourselves. I think I'm finally getting close to that point, and it's great fun, also to be doing this in a widely distributed international network of scientists thinking in similar directions.

    Antti's VAN igniting from a P3b gathering of preconscious recurrence is a little bit off your number of 180 ms. I have no idea which is right, or whether they are aspects of the same phenomenon.

    I did not expect a couple of decades ago to ever have this discussion, or to be getting within a few hundred ms of a plausible answer.

    What fun!

    Best wishes,


  5. At the end of his talk, Singer mentioned consciousness without content, such as in people meditating. This is not something that has come up so far in the summer school. We can add it to the list of phenomenon that a theory of consciousness should explain.

    1. This is indeed very interesting! Like you said it's a new element! I'm also very interested in the "degree of consciousness". I know that Mr. Harnad argue that you are either conscious or you are not (sorry for repeating that again!), but I wonder about evidence like the meditation you mentionned, or hypnosis that is supposed to put you in an hyper-conscious state (if I remember correctly), or even when you are sleeping and you hear something in your dream that wakes you up, you were sleeping so you weren't really conscious, but still you got conscious about this stimuli. So many ways to explore!!!

    2. Good point. And to follow up on Marjorie's post: Shimon Edelman, who was borrowing from dynamic systems theory (and who was thus also fond of the idea that consciousness has to be understood as the result of the dynamics of an entire system over time), claimed explicitly that there are degrees of consciousness. So I am wondering: is the view that there are degrees of consciousness a direct corollary of a dynamics approach to consciousness?

    3. I so much agree! I am also struck by how little is said at this conference simply concerning a description of how consciousness is actually experienced, i.e. what is the phenomenon that we seek to describe. A while back, looking back at how my own consciousness had evolved over my life time, it struck me that it had actually grown over the years to encompass a wider and wider realm. And I thought that perhaps that could be the reason for adults' anxiety, because our consciousness has grown to a level to encompass problems that we can't actually deal with, and whether that could be the reason that children are so happy, because they can still deal the realm of what they are conscious about, and whether that is not the reason that meditation (focusing on a much limited realm) does not appease us? Another thing to consider about the conscious experience, it seems to me, is its power to feedback upon specific mental processes, as in when we think of something and it makes us cry or laugh or decide to think of something else...

    4. I’m with you on this point, and you make some excellent observations – especially coming from a non-cognitive science or philosophy perspective (although, yes, I have done my own investigations onto the definition of conscious experience), it is difficult to integrate all the possibilities falling under the realm of conscious experience. Even creating definitions, and even when focusing on humans alone, becomes as much a logistical nightmare as it is a semantic nightmare.

      As well as serving to answer the ‘nature vs nurture’ question which has arisen periodically during this conference, examining the conscious experience of children serves to uncover incredible insights on how a fully actualized adult forms conscious experience. It not only becomes wider, it becomes richer and more influential as new semantic and episodic memories are incorporated. As our past experiences grow in magnitude across a lifetime, they seem to exert disproportionately more control over things as elemental as feelings.

      I am kind of just rambling, but I suppose a point I could make is that as this body of conscious experience (and repertoire OF conscious experience) grows over one’s lifetime, all elements seem to be more intertwined. I can no longer feel elemental joy at a cute newborn due to the looming threat of global catastrophe caused by overpopulation (I don’t actually think this, just making an extreme point ☺).
      It becomes increasingly difficult to point to a neural correlate or cause of consciousness as a single defined loci, especially a single visual neuron making a binary choice. Coming back to Wolf’s ideas, I think it makes more sense for consciousness to exist as a global state of small world networks, interconnected across time and space. Each mini network might represent specific features of the external, internal or episodic memory world- but alone insufficient to qualify as an NCC.

    5. I also liked his thoughts/questions at the end of the talk about consciousness without content, as Martha mentioned. I was also very interested in his discussion of schizophrenic patients. Could they be considered to have different levels of consciousness? What about patients with personality disorders? I am wondering if it would be a good idea to look into different disorder populations in studying consciousness. Where would these fit in terms of evolutions of humans? So many models/talks have used animals and/or robots to study consciousness, it would seem interesting to also compare different types/levels of consciousness within humans.

      Izabo Deschênes

  6. The zero phase-lag in oscillations measured at distant cortical points... was explained by a common pacemaker, nonlinear interactions and/or something about having three points in space... how about good-old volume conduction? Also, I wondered what Dr. Singer thought about long-range inhibitory connections such as those identified between hippocampus and septum, do they also play a role in inter-region synchronization?

    Do schizophrenics with positive symptoms (hallucinations) have too much integration of info, perhaps displaying increased gamma power and increased synchrony compared to normals?

    I think it's important to establish how fMRI, EEG and MEG signals correlate with each other. I was happy to hear Dr. Singer say that increased gamma power in MEG correlated well with fMRI signal. Generally, this will be difficult to do, possibly because it may require applying two techniques in the same study if not simultaneously. But I think it would be important if we are to have a unified idea about what is happening in the brain when given a certain cognitive task, since we see more and more that experimental conditions make a huge difference to brain activity.

  7. This comment has been removed by the author.

  8. Very interesting idea! I am not sure I understand well, but could this be related to the fact that after all consciousness is about putting different kinds of information together? Hm... Further, synchrony in time, could it be a general feature of biological systems, and explain why biological systems are so robust to perturbation (of specific points in space)?

    1. Synchrony brings cells' activity in distant networks to fluctuate together, so it's thought to be involved in binding of information processing. Consciousness is also a unified percept across modalities. But I am not sure about equating consciousness with synchrony, just because they appear to do similar things.

      I don't understand what you mean by systems being robust to perturbation?

  9. Some principle questions concerning synchrony:
    What makes a neuron a candidate of being a member of a synchronized ensemble of neurons? Are there certain processes which enable that?
    In other words: What is the criteria which parts/neurons are bound together via synchrony? Even if it's a self organizing process then the affected neurons must have a certain "property" or be in a certain state (I guess).
    If synchrony emerges from coupling oscillators then establishing synchrony takes some time.
    It seems that synchrony is the resulting (dynamic) state. Hence it can be used to "label" many different areas from different regions as belonging / being bound together.
    But how do the synchronizing neurons "know" that they have to participate in the synchrony process? Which states/properties of neurons exclude them from synchrony ?

    1. addon:
      Could one even suppose that conscious perception does not just correlate with the synchronization of neural activity across cortical areas but that conscious perception is used to indicate which parts/areas/neurons participate in synchronization?
      Unfortunately I've got no idea how this hypothesis could be tested.

    2. Hi Peter,

      As for being a member of a neuronal ensemble, a cell signals similar information as the rest of the group (as far as place cells in the hippocampus are concerned, and that's the only system I know!). For example, place cells in the same 'assembly' signal similar spatial locations. (But I must admit that the neuronal ensemble/cell assembly idea is man-made and very much up to debate about the exact definition.)

      As for a cell being part of synchronized rhythmic activity, that's actually a different question because cell assemblies and rhythmic activities are somewhat dissociable but related. For example, you can get rid of the theta rhythm but you may still have cell assemblies. But, the information contained by these cell assemblies is degraded because they cannot fire together as effectively without the rhythm.

      Many cells are synchronized by the same rhythm because they are driven by the same pacemaker (other mechanisms of synchrony exist of course). The pacemaker input can come from a local synaptic network or from a subcortical structure that provides a rhythmic input. I guess you may be left out of the rhythm if you don't receive such an input, thus the membership/exclusion is dictated by synaptic circuitry. (A bit like if you don't belong to the same facebook group, you won't know where everyone's going out for drinks after the conference! :) And we have certain key people - the pacemakers - that disseminate this knowledge.)

      I think your idea bout conscious perception and what constitutes the contents of it by looking at which regions are synchronizing together is feasible. I'm probably going to do my paper on this subject!

  10. A quite general question poped-up in my head when Wolf Singer presented briefly the two hypotheses commonly found in neurosciences about the neurobiological structures underlying consciousness, mainly that consciousness might be either produced and controlled (1) in some specific parts of the brain; or (2) in the whole brain. However, I was wondering whether it might not be a mistake to see consciousness as a "thing" with neural correlates (such as the visual pathway, the auditory pathway, underlying structures of memory, etc...) that is produced and controlled by some specific structures (as suggested by the first hypothesis) rather than seeing consciousness as a general neural state that is the consequence of the combination of all our cognitive abilities interacting together (those cognitive abilities having themselves their own underlying structures), which would lead to say that consciousness doesn not really have a proper neural circuitry?

    Dr. Singer sustains that the neural correlate of the global workspace of consciousness may lie in the oscillatory synchrony between neurons across the brain. If so, then maybe we could design a test for the presence of consciousness in animals based on the presence of such global oscillatory synchrony in the brain. For example, if no such global synchrony phenomena occur in aplysia, then we could conclude that the aplysia nervous system does not generate conscious experience. Of course, this test would need to pass many tests before considering its validity.

  12. Xavier Dery ‏@XavierDery

    Signer: correlation between "synchronized" oscillating neurons is at best 0.3, and he just called 0.25 a "strong" correlation... #TuringC

    9:27 AM - 5 Jul 12 via Twitter for Android

  13. In Dr. Singer's talk he highlighted the dynamic state that is characterized by coherent activation of different networks. He mentioned schizophrenic patients as having a reduced gamma synchrony as compared to control patients. I have read that people with autism also have a reduction in high frequency oscillations, which I assume is also associated with a reduction in oscillatory synchrony. I am curious if in this population this would be attributed to a a confusion or misinterpretation of stimuli similar to that in schizophrenic patients.