JOURNAL CLUB: Crossing the threshold to consciousness
Alfredo Pereira Jr
Thursday, 28 February 2008 11:48 UTC
Introduction
The study of visual consciousness in cognitive neuroscience is based on establishing correlations between three parameters: properties of visual stimuli, brain activity and the conscious state of the subject (communicated by means of verbal or non-verbal behavioral reports, also called “psychophysics”). Progress in this line of research is often defined by how well the researchers control the timing of stimulus presentation, how accurately brain activity is measured and whether the behavioral reports are precise.
The paper by Del Cul, Baillet and Dehaene (2007; free PDF here ) introduces a new methodology to sharpen the control of visual stimuli and examine the respective effects on brain function. After a previous study on subliminar (unconscious) perception elicited by brief stimulation (ref. 21 in the paper), Stanislas Dehaene and colleagues measured brain events and behavioral reports elicited by the presentation of supraliminar stimuli combined with backward masking. The main goal was to find the sequence of brain events necessary for the formation of a reportable conscious visual state.
The authors attempt to address a controversy in their field concerning which brain areas are essential for human visual consciousness. Researchers in one camp claim that the activation of visual areas in occipital cortex is sufficient to generate visual conscious states. A second group suggests that the activation of other cortical associative areas (especially frontal visual areas) is also necessary. The timing of brain events supporting conscious processes is crucial to distinguish between the diverging views, since processing in occipital areas alone would take around 100-200 ms to occur, while the formation of a large-scale integrative process involving more brain areas (frequently related to the Global Workspace Theory, proposed by Bernard Baars in 1989) would take longer.
Methods and Results
The target stimulus consisted of a single digit (a number) projected for 16 ms (a brief, but supraliminar stimulus). The mask was a group of numbers projected soon after, for 250 ms, at the same visual location. The authors varied the time interval (called SOA, “target-mask stimulus onset asynchrony”) between stimulus onset and the presentation of the mask, from 16 to 100 ms. Shorter intervals were predicted to cause backward masking of the stimulus (basically, relegating the stimulus below the threshold of consciousness), by means of a perturbation of the sequence of brain events necessary to generate the corresponding conscious state. Longer intervals would not perturb the brain processing of the stimulus, allowing the propagation of excitation to higher cortical areas.
High-density event-related potential (ERP) recordings were collected to determine whether a change (or a “transition”) in brain activity occured between the lower and higher SOA values. The temporal location of this transition was defined behaviorally, using two measures: a forced-choice comparison of the presented digit with another one (checking for both sub- and supraliminar perception), and a scale of visibility, which roughly assesses the conscious access the subject had to the stimulus.
This methodology produced several behavioral and physiological (ERP) results. Behaviorally, a “significant nonlinearity” was found for the SOA interval from 33 to 66 ms. Below 16 ms, the performance was at chance level, suggesting that the presented digit was completely masked by the second. In the 16-33 ms interval, performance in the forced-choice task was above chance, while in the 33-66 ms interval both the forced-choice and the (conscious) visibility ratings increased non-linearly. Above 66 ms there was not a significant change in visibility and the subjects consistently had conscious access to the presented stimulus. The authors concluded that “a major transition in processing occurs around SOA = 50 ms”.
The authors then looked for ERP components temporally correlated with the transition in visibility elicited by the 50 ms SOA. Using statistical testing across subjects, they found activity within a fronto-parieto-temporal network (called the P3 component in the paper) to be strongly correlated with the non-linear increase in visibility. The change in P3 that correlates with the transition occurs about 270 to 300 ms after target onset. This finding is consistent with the hypothesis that conscious access to a visual stimulus involves the sequential activation of several cortical areas. Thus, we now have some fairly strong evidence suggesting that the conscious process of visual stimulus sensation involves more than the simple activation of occipital visual areas. So, in order to cross the threshold to consciousness, we need network activity that is likely greater than the sum of its parts.
Questions:
1) Why is activation of frontal visual areas necessary for (conscious) visibility?
2) What are the physiological mechanisms supporting the Global Workspace?
3) Does conscious processing in other perceptual modalities also display similar timing?
-
Replies
-
Another interesting study that uses the same concepts of stimulus presentation and backward masking to determine conscious presentation.
Neurons in the medial temporal lobe seemed to react in an all-or-none fashion to conscious reaction to a stimulus. I wasn’t able to get the full version of the paper to see how they did the masking, but even short presentations of a stimulus (33 msec) could elicit a response in single neurons.
Alfredo, could you check the methods and see how this paper relates to the one above? Thanks!
-
Dear Noah:
The second study (by Quian Quiroga et al.) has important differences in methodology, but the results seem to be compatible with the Del Cul. et al. study.
One of the differences is that Quiroga made single and multi-unit recordings (measurement of axonal activity by means of invasive microelectrodes implanted in epilepsy patients), while Del Cul used ERP.
The second difference – related to the first – is that Del Cul measured neural activity (including both dendritic and axonal signals) in the whole cortex, while Quiroga was restricted to axonal firing in the medial temporal lobe.
The third difference is that Del Cul presented simple stimuli for a minimal supraliminar perceptual time duration (one digit number for 16 ms) while Quiroga presented complex stimuli for longer and varied times (pictures of faces, WTC, etc. during 33 to 264 ms).
The fourth (and central) difference is that the stimulus and mask duration varied in the Quiroga experiment (the total time of stimulus and mask presentation was always 500 ms), while Del Cul varied the time interval between them. This methodological divergence makes a huge difference in the interpretation of results.
My first impression is that the Del Cul’s methodology assures better correlation of measured brain activity with visual consciousness.
The results are compatible, because the medial temporal response measured by Quiroga (at approximately 300 ms after stimulus onset) occurs simultaneously with or soon after the frontal response measured by Del Cul. (at 270-300 ms).
There are, of course, several possible interpretations of this compatibility:
a) both responses (frontal and medial temporal) occur at the same time and both support visual consciousness equally;
b) one of the responses occur first and supports visual consciousness; the other serves to another function (e.g., the Del Cul response may be related to short-term memory and the Quiroga response may be related to triggering the formation of long-term memory).
In the case of interpretation “b” I am more inclined to trust the Del Cul result, because the frontal response was abolished by shorter time intervals between stimulus and mask (in this case, there was only subliminar perception). The Quiroga methodology does not allow the usage of the masking procedure to check the necessity of medial temporal responses for visual consciousness (they assume that the masking procedure only affected occipital visual areas, allowing them to rule out a crucial role of these areas for visual consciousness).Best
Alfredo
-
Another comment about the Quiroga paper can be found here
-
Thanks Alfredo, it did seem to be compatible, but since I couldn’t see the methods, I wasn’t sure.
It is interesting that the PNAS paper suggests a sparse representation of these neural correlates in humans because that is a theory put forth by researchers examining sensory processing in rodents. This is particularly studied in auditory cortex (see the work of Tony Zador).
So, sparse representation may be involved in several different coding processes in the brains of mammals.
-
Dear Noah:
The hypothesis of sparse coding supporting consciousness may be an artifact of the single cell technique. People (including scientists) see mostly what they are looking for. The proponents of this hypothesis seem to ignore other kind of data indicating widespread activation of Local Field Potentials (post-synaptic activity, corresponding to the hemodynamic response measured by BOLD fMRI) and large-scale gamma oscillatory synchrony (measured with EEG).
I understand conscious processing as beginning with the formation of LFPs at several brain locations, involving the sustaining of these fields until they get synchronized, and something more (that we still do not know) following the synchronized phase.
In this view, axonal activity is just the neuron´s way to broadcast local field activity to other neurons. How could consciousness be based on the spiking activity of a small group of cells? Possibly sparse coding is central to the process of recognition, but it does not seem to fit an explanatory role for consciousness.
There are some complex, technical issues involved in this problem that I will not mention in this message, but I will discuss if someone is interested. Basically, the issue is that LFPs includes subthreshold and inhibitory activity, besides the suprathreshold activity that generates axonal firing. Subthreshold and inhibitory activity may contribute even more than axonal activity for consciousness. Gamma oscillatory synchrony is based on subthreshold oscillations. Of course, the onset of synchrony increases spiking rates in the neuronal population, and this is the reason why in some studies oscillatory synchrony has been related with synchronous axonal firing.Best
Alfredo
-
A PubMed search on “Zador sparse” retrieved two interesting abstracts (below).
In the first (2008) study they used glass electrodes to record (single unit) firing rates, and found that “the population response is sparse, with stimuli typically eliciting high
firing rates (>20 spikes/second) in less than 5% of neurons at any instant”.
In the second study (2006) the authors recorded subthreshold potentials. The title of the paper is a bit enigmatic (do they mean that the synchronous “bumps” are sparse?). The suggestion they make at the end does not seem to follow from the collected data.
Alfredo1) PLoS Biol. 2008 Jan;6(1):e16.
Sparse representation of sounds in the unanesthetized auditory cortex.
Hromádka T, Deweese MR, Zador AM.
How do neuronal populations in the auditory cortex represent acoustic stimuli?
Although sound-evoked neural responses in the anesthetized auditory cortex are
mainly transient, recent experiments in the unanesthetized preparation have
emphasized subpopulations with other response properties. To quantify the
relative contributions of these different subpopulations in the awake
preparation, we have estimated the representation of sounds across the neuronal
population using a representative ensemble of stimuli. We used cell-attached
recording with a glass electrode, a method for which single-unit isolation does
not depend on neuronal activity, to quantify the fraction of neurons engaged by
acoustic stimuli (tones, frequency modulated sweeps, white-noise bursts, and
natural stimuli) in the primary auditory cortex of awake head-fixed rats. We find
that the population response is sparse, with stimuli typically eliciting high
firing rates (>20 spikes/second) in less than 5% of neurons at any instant. Some
neurons had very low spontaneous firing rates (below 0.01 spikes/second). At the other
extreme, some neurons had driven rates in excess of 50 spikes/second.
Interestingly, the overall population response was well described by a lognormal
distribution, rather than the exponential distribution that is often reported.
Our results represent, to our knowledge, the first quantitative evidence for
sparse representations of sounds in the unanesthetized auditory cortex. Our
results are compatible with a model in which most neurons are silent much of the
time, and in which representations are composed of small dynamic subsets of
highly active neurons;2) J Neurosci. 2006 Nov 22;26(47):12206-18.
Non-Gaussian membrane potential dynamics imply sparse, synchronous activity in
auditory cortex.
DeWeese MR, Zador AM.
Many models of cortical dynamics have focused on the high-firing regime, in which
neurons are driven near their maximal rate. Here we consider the responses of
neurons in auditory cortex under typical low-firing rate conditions, when stimuli
have not been optimized to drive neurons maximally. We used whole-cell
patch-clamp recording in vivo to measure subthreshold membrane potential
fluctuations in rat primary auditory cortex in both the anesthetized and awake
preparations. By analyzing the subthreshold membrane potential dynamics on single
trials, we made inferences about the underlying population activity. We found
that, during both spontaneous and evoked responses, membrane potential was highly
non-Gaussian, with dynamics consisting of occasional large excursions (sometimes
tens of millivolts), much larger than the small fluctuations predicted by most
random walk models that predict a Gaussian distribution of membrane potential.
Thus, presynaptic inputs under these conditions are organized into quiescent
periods punctuated by brief highly synchronous volleys, or “bumps.” These bumps
were typically so brief that they could not be well characterized as “up states”
or “down states.” We estimate that hundreds, perhaps thousands, of presynaptic
neurons participate in the largest volleys. These dynamics suggest a
computational scheme in which spike timing is controlled by concerted firing
among input neurons rather than by small fluctuations in a sea of background
activity. -
Another nice paper from this group was published in Nature in with Mike Weir as the first author. This study examined sparse coding in an anesthetized prep using whole-cell patch clamp. A technical tour-de-force, but obviously cannot inform on consciousness.
-
Dear Noah:
I am inclined to accept that sparse coding in higher associative areas has a central role in conscious processing, when combined with cortical oscillatory synchrony.
A chronology of events that include all relevant and compatible findings would be the following:
a) Under 80 ms after stimulus onset, stimulus-evoked receptor field responses occur in primary sensory areas, without conscious perception of the stimulus;
b) Around 100-120 ms after stimulus onset pre-conscious priming occurs. This phase is necessary to trigger the conscious process but does not generate the full conscious visual state yet. The priming is done by a transient stimulus-evoked gamma synchronization, encompassing primary sensory, higher sensory and associative cortical areas, as described by Melloni et al. (2007);
c) Between 130-270 ms, a stimulus-evoked, feed-forward gamma synchronous firing from sensory to associative areas occurs (see single unit recordings, in visual areas of the anesthetized cat, by Samonds and Bonds, 2005);
d) Around 270-300 ms, sparse responses to the visual stimulus occur in higher associative areas (as registered by Quiroga et al., 2007), which are related to the P3 component recorded by Del Cul et al. (2007) and the P300 component referred by Melloni et al. (2007);
e) Beyond 300 ms, reentrant signaling from higher associative back to sensory areas reach previously potentiated neuronal assemblies. From this moment on, visual consciousness of the stimulus occurs. Gamma oscillations, sustained with the participation of astrocytes in glutamatergic tripartite synapses (Pereira and Furlan, 2007), becomes phase-locked with alpha and theta (Palva and Palva, 2007), generating brain-wide coherence of post-synaptic potentials.
Best,
Alfredo
References:
Quiroga RQ, Mukamel R, Isham EA, Malach R, Fried I. (2008) Human single-neuron responses at the threshold of conscious
recognition. Proc Natl Acad Sci U S A 105(9):3599-604.Melloni L, Molina C, Pena M, Torres D, Singer W, Rodriguez E. (2007) Synchronization of neural activity across cortical areas correlates
with conscious perception. J Neurosci. 27(11):2858-65.Palva S, Palva JM. (2007) New vistas for alpha-frequency band oscillations. Trends Neurosci. 30(4): 150-8.
Pereira Jr A, Furlan FA (2007) Meta-Potentiation: Neuro-Astroglial Interactions Supporting Perceptual Consciousness. Available from Nature Precedings
Samonds, JM and Bonds, AB (2005) Gamma oscillation maintains stimulus structure-dependent synchronization in cat visual cortex. J. Neurophysiol. 93 (1): 223-36.
-