What are prefrontal pyramidal neurons doing?

[Thinkor]

I've been reading up on neurology lately and I'm mystified by pyramidal neurons.
The pyramidal neurons in the prefrontal cortex, for example, have maybe 20,000 inputs through their dendritic trees and 1 output through the axon.
The output is an action potential that always has the same waveform, up to the usual
imperfections.
Strength of a signal is indicated not by a change in the waveform but by emission of more waveforms (essentially all the same) per unit of time.
So, these neurons are getting a tremendous amount of information and then reducing it to (1) one waveform (always the same) and (2) an integer frequency, which is probably not more than 100hz.

I cannot imagine what such a device is good for, but these pyramidal neurons dominate the part of the brain usually associated with consciousness.
Where are the outputs used and what are they used for?

I'm also mystified why no one else comments on this. The wiki article on pyramidal neurons
says this regarding cognition (consciousness):

"Pyramidal neurons in the prefrontal cortex are implicated in cognitive ability. In mammals, the complexity of pyramidal cells increases from posterior to anterior brain regions. The degree of complexity of pyramidal neurons is likely linked to the cognitive capabilities of different anthropoid species. Because the prefrontal cortex receives inputs from areas of the brain that are involved in processing all the sensory modalities, pyramidal cells within the prefrontal cortex appear to process different types of inputs. Pyramidal cells may play a critical role in complex object recognition within the visual processing areas of the cortex."

OK, but how do pyramidal neurons actually do anything useful?
Are we just dumb about the brain?

 

[DiracPool]

So, these neurons are getting a tremendous amount of information and then reducing it to (1) one waveform (always the same) and (2) an integer frequency, which is probably not more than 100hz.

Most cortical neurons do not fire at above 3-4 hz. It's the cooperative action of hundreds of millions of neurons all firing together that form complicated chaotic attractors over the cortical regions that give rise to cognition and consciousness. Every neuron in the cortex plays a role in this action, not just the pyramidal cells, so when you read things in Wiki like "Pyramidal neurons in the prefrontal cortex are implicated in cognitive ability," I wouldn't read that much into it. Pyramidal neurons are ubiquitous about the neocortex so of course they're going to be important in cognition, but that doesn't tell you much.

 

[Pythagorean]

Prefrontal pyramidal neurons are doing what every other neuron is doing: processing information. You can actually do quite a bit, in terms of information processing, with the setup you describe (thousands of inputs being decoded to one output). Put many such objects together, and you can process a lot of information in a lot of different ways depending on connectivity and the functional properties of the neurons themselves.

This is all helpful for understanding the physiological aspect of neural function and information processing, but we still have no idea how consciousness arises in the first place, so we cant' really comment on that.

 

[Thinkor]

Most cortical neurons do not fire at above 3-4 hz. It's the cooperative action of hundreds of millions of neurons all firing together that form complicated chaotic attractors over the cortical regions that give rise to cognition and consciousness. Every neuron in the cortex plays a role in this action, not just the pyramidal cells, so when you read things in Wiki like "Pyramidal neurons in the prefrontal cortex are implicated in cognitive ability," I wouldn't read that much into it. Pyramidal neurons are ubiquitous about the neocortex so of course they're going to be important in cognition, but that doesn't tell you much.

That sounds interesting. Do you have a reference that explains "chaotic attractors" and their connection to cognition?
I've read some books and papers by prominent experts (e.g., Baars, Dehaene, Dennett, Crick & Koch) without coming across that concept.

 

[DiracPool]

Do you have a reference that explains "chaotic attractors" and their connection to cognition?
I've read some books and papers by prominent experts (e.g., Baars, Dehaene, Dennett, Crick & Koch) without coming across that concept.

Here's a great book that just came out this year:

https://www.amazon.com/dp/3319244043/?tag=pfamazon01-20

If you want a more introductory classic text you can try these:

https://www.amazon.com/dp/0262611317/?tag=pfamazon01-20
https://www.amazon.com/dp/0231120087/?tag=pfamazon01-20

 

[Thinkor]

I will check them out. Thank you very much.

 

[atyy]

http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1000803
Bistable, irregular firing and population oscillations in a modular attractor memory network.
Lundqvist M, Compte A, Lansner A.

Just a note: don't get too into "attractor" networks. They can be good models, but fundamentally there are no attractors in biological systems if one includes enough detail in the model, because all biological systems have finite lifetimes.

 

[Pythagorean]

http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1000803
Bistable, irregular firing and population oscillations in a modular attractor memory network.
Lundqvist M, Compte A, Lansner A.

Just a note: don't get too into "attractor" networks. They can be good models, but fundamentally there are no attractors in biological systems if one includes enough detail in the model, because all biological systems have finite lifetimes.

I don't like dirac's assertion, in general, " It's the cooperative action of hundreds of millions of neurons all firing together that form complicated chaotic attractors over the cortical regions that give rise to cognition and consciousness". It is stated with too much confidence, and what I don't like about it is that attractors are abstract mathematical objects, they don't give rise to anything so much as they are a tool to predict. Further, we have no idea what gives rise to consciousness, only correlates of consciousness.

The next step for modelling, given what you said, is to turn to transient chaos, systems in which the manifold of interest has many stable (attractive) dimensions but at least one repulsive one. Thus, trajectories get caught in a chaotic saddle. In the case of spatiotemporal transient chaos (where we consider spatial extensions to more neurons) the time the trajectory spends caught in the saddle before collapsing to some attractor in the network grows with the number of neurons. There is also the idea of chaotic itinerancy, as proposed by Tsuda, in which there are several such "attractor ruins" linked together and trajectories can jump from one to the other.

To some extent, your interpretation comes down to model inclusiveness, too. We *could* model our system as being a series of true chaotic attractors, where we externally control a parameter that causes them to jump and then somehow say that this parameter is determined by another controlling network in the brain that may or may not, itself, be made up of neurons - some other candidates are astrocytic networks and expression/molecular processes that lead to changes in synaptic strength between neurons and intrinsic firing properties within individual neurons.)

 

[DiracPool]

I don't like dirac's assertion, in general... It is stated with too much confidence, and what I don't like about it is that attractors are abstract mathematical objects, they don't give rise to anything so much as they are a tool to predict.

Well, you could make the exact same argument about the quantum wave function. What alternative to using an abstract mathematical object to model cortical function (a physical process) do you have in mind?

The next step for modelling, given what you said, is to turn to transient chaos, systems in which the manifold of interest has many stable (attractive) dimensions but at least one repulsive one. Thus, trajectories get caught in a chaotic saddle. In the case of spatiotemporal transient chaos (where we consider spatial extensions to more neurons) the time the trajectory spends caught in the saddle before collapsing to some attractor in the network grows with the number of neurons.

You may be confusing the "neurodynamic model" (aka Freeman's "action-perception model) that the books in my post #5 refer to with more commercial "attractor network" or Hopfield network models:

https://en.wikipedia.org/wiki/Attractor_network

I'm not sure what model you are talking about, but in the neurodynamic model trajectories don't converge to point attractors, which amounts to brain death in real brains. Instead, it models a basal state of brain activity as existing in a high-dimensional chaotic state. This state is referred to as a "background attractor" whose activity is not random, but that shows the distinctive sign of chaos when you phase plot the local field potentials of a locally circumscribed cytoarchitectonically brain region, such as V1, S1, V2, etc. This instability is largely an interareal property whereby bidirectional signals between two different cortical regions disrupt each others activity through phase differences brought about via the cable delays of convergent and divergent fiber tracts between the two regions. For instance, typically, the background attractor of the olfactory network vanishes if you sever the connections from the olfactory bulb to the olfactory cortex.

There is also the idea of chaotic itinerancy, as proposed by Tsuda, in which there are several such "attractor ruins" linked together and trajectories can jump from one to the other.

Tsuda's model I'm more familiar with, and does fall under the neurodynamic model framework. Getting back to the background attractor, it's utility is actually to prevent the entrainment of any given local cortical region's activity to converging to a point attractor or the stable orbit of a limit cycle. This keeps the brain in a "ready state" to respond instantly to sensory input, which would be difficult or impossible if the collective behavior of hypercolumns in the pool existed at rest in some entrained state. What sensory input achieves is the effect of coordinating a critical subset of Hebbian assemblies in primary sensory cortex which, in turn, place that cortex into a basin of attraction whereby the cooperation between the columns begin to converge onto limit cycle behavior...but not quite, it doesn't quite get there. It converges to a "near limit cycle" state which allows that cortex to achieve a temporary "burst" state which sends the consensus of what the sensory cortex has interpreted that the sensory stimulus was, say the image of a flower. This consensus forms the identity of the percept and is what is sent out to other cortical regions. In the visual system, a burst in V1 can send this consensus statement to many extrastriate regions simultaneously. So, when it's comes to cognition, this consensus statement serves as a fundamental "frame" of information. In Tsuda's model, these frames are the chaotic states that form the "ruins" you are talking about. In Freeman's model, these frames are modeled as frames on a film reel. A currently active near limit cycle attractor is modeled as the frame currently in the "shutter" that is being viewed on the screen, or in the case of the brain, the "mind's eye," whereas the frames that have just passed through the shutter are reflective of an itinerant path or trajectory of "attractor ruins," ruined as in they've already seen their day (until perhaps next time when a memory of the percept is re-ignited.)

Collectively, the version of the action-perception cycle I described above is referred to as the "cinematographic model":

http://sulcus.berkeley.edu/wjf/FG Freeman Phase 'shutter'2007.pdf
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2288954/pdf/11571_2006_Article_9001.pdf
http://www.ncbi.nlm.nih.gov/pubmed/16513196

To some extent, your interpretation comes down to model inclusiveness, too. We *could* model our system as being a series of true chaotic attractors, where we externally control a parameter that causes them to jump

In the neurodynamic model, this "external control parameter" would be sensory stimuli coming into a primary visual cortice from the sense receptor relays.

and then somehow say that this parameter is determined by another controlling network in the brain that may or may not, itself, be made up of neurons - some other candidates are astrocytic networks and expression/molecular processes that lead to changes in synaptic strength between neurons and intrinsic firing properties within individual neurons.)

We've discussed the participation of astrocytes as relevant drivers of cortical activity in earlier threads. I'm still not convinced that whatever contribution they may be making is anything significant.

It is stated with too much confidence

You may be right here. The neurodynamic model is one among many and maybe I should have been more explicit with that statement above. But the OP seemed to already know this as he was familiar with competing models. However, I've studied many different brain models and the neurodynamic model has far more explanatory power than any other model I've come across, so I've focused mainly on that one in my own studies. I do have an open mind, though, please link me to other models you feel show promise and I will look at them.

 

[DiracPool]

I don't like dirac's assertion, in general, " It's the cooperative action of hundreds of millions of neurons all firing together that form complicated chaotic attractors over the cortical regions that give rise to cognition and consciousness". It is stated with too much confidence, and what I don't like about it is that attractors are abstract mathematical objects, they don't give rise to anything so much as they are a tool to predict. Further, we have no idea what gives rise to consciousness, only correlates of consciousness.

Ok, I re-read this passage and think I understand now what you're getting at as far as the attractors "giving rise" to consciousness and cognition. Perhaps I could have worded that differently but I'm just assuming that nobody is really going to think that a human-created abstract mathematical object would actually give rise causally to a physical or biological process. I think that just goes without saying. It's the same thing for the "neural correlates of consciousness" thing. I always hated that phrase. I thought and still do that it's ontologically weak. To me, using the phrase "neural correlates of consciousness" implies some sort of dualist or least agnostic stance. It's saying that we don't know that what is going on in the biophysics of neural networks is actually equated with consciousness, all we can say is that it is "correlated" with consciousness. Again, I think that is a weak stance and I take the position they are one in the same, the mind is what the brain does, as far as we will ever be able to know.