Posts tagged ‘neurons’

An interesting but somewhat problematic paper from the Blue Brain project claims that the application of topology to neural models has provided a missing link between structure and function. That’s exciting because that kind of missing link is just what we need to enable us to understand how the brain works.  The claim about the link is right there in the title, but unfortunately so far as I can see the paper itself really only attempts something more modest. It only seems to offer  a new exploration of some ground where future research might conceivably put one end of the missing link. There also seem to me to be some problems in the way we’re expected to interpret some of the findings  reported.

That may sound pretty negative. I should perhaps declare in advance that I know little neurology and less topology, so my opinion is not worth much. I also have form as a Blue Brain sceptic, so you can argue that I bring some stored-up negativity to anything associated with it. I’ve argued in the past that the basic goal of the project, of simulating a complete human brain is misconceived and wildly over-ambitious; not just a waste of money but possibly also a distraction which might suck resources and talent away from more promising avenues.

One of the best replies to that kind of scepticism is to say, well look; even if we don’t deliver the full brain simulation, the attempt will energise and focus our research in a way which will yield new and improved understanding. We’ll get a lot of good research out of it even if the goal turns out to be unattainable. The current paper, which demonstrates new mathematical techniques, might well be a good example of that kind of incidental pay off. There’s a nice explanation of the paper here, with links to some other coverage, though I think the original text is pretty well written and accessible.

As I understand it, topological approaches to neurology in the past have typically considered neural  networks as static objects. The new approach taken here adds the notion of directionality, as though each connection were a one-way street. This is more realistic for neurons. We can have groups of neurons where all are connected to all, but only one neuron provides a way into the group and one provides a way out; these are directed simplices. These simplices can be connected to others at their edges where, say, two of the member neurons are also members of a neighbouring simplex. Where there are a series of connected simplices, they may surround a void where nothing is going on. These cavities provide a higher level of structure, but I confess I’m not altogether clear as to why they are interesting. Holes, of course, are dear to the heart of any topologist, but in terms of function I’m not really clear about their relevance.

Anyway, there’s a lot in the paper but two things seem especially noteworthy. First, the researchers observed many more simplices, of much higher dimensionality, than could be expected from a random structure (they tested several such random structures put together according to different principles). ‘Dimensionality’ here just refers to how many neurons are involved; a simplex of higher dimensionality contains more neurons. Second they observed a characteristic pattern when the network was presented with a ‘stimulus’; simplices of gradually higher and higher dimensionality would appear and then finally disperse. This is not, I take it, a matter of the neurons actually wiring up new connections on the fly, it’s simply about which are involved actively by connections that are actually firing.

That’s interesting, but all of this so far was discovered in the Blue Brain simulated neurons, more or less those same tiny crumbs of computationally simulated rat cortex that were announced a couple of years ago. It is, of course, not safe to assume that real brain behaves in the same way; if we rely entirely on the simulation we could easily be chasing our tails. We would build the simulation to match our assumptions about the brain and then use the behaviour of the simulation to validate the same assumptions. In fact the researchers very properly tried to perform similar experiments with real rat cortex. This requires recording activity in a number of adjacent neurons, which is fantastically difficult to pull off, but to their credit they had some success; in fact the paper claims they confirmed the findings from the simulation. The problem is that while the simulated cortex was showing simplices of six or seven dimensions (even higher numbers are quoted in some of the media reports, up to eleven), the real rat cortex only managed three, with one case of four. Some of the discussion around this talks as though a result of three is partial confirmation of a result of six, but of course it isn’t. Putting it brutally, the team’s own results in real cortex contradicted what they had found in the simulation. Now, there could well be good reasons for that; notably they could only work with a tiny amount of real cortex. If you’re working with a dozen neurons at a time, there’s obviously quite a low ceiling on the complexity you can expect. But the results you got are the results you got, and I don’t see that there’s a good basis here for claiming that the finding of high-order simplices is supported in real brains. In fact what we have if anything is prima facie evidence that there’s something not quite right about the simulation. The researchers actually took a further step here by producing a simulation of the actual real neurons that they tested and then re-running the tests. Curiously, the simulated versions in these cases produced fewer simplices than the real neurons. The paper interprets this as supportive of its conclusions; if the real cortex was more productive of simplices, it argues, then we might expect big slabs of real brain to have even more simplices of even higher dimensionality than the remarkable results we got with the main simulation. I don’t think that kind of extrapolation is admissible; what you really got was another result showing that your simulations do not behave like the real thing. In fact, if a simulation of only twelve neurons behaves differently from the real thing in significant respects, that surely must indicate that the simulation isn’t reproducing the real thing very well?

The researchers also looked at the celebrated roundworm C. Elegans, the only organism whose neural map (or connectome) is known in full, and apparently found evidence of high-order simplices – though I think it can only have been a potential for such simplices, since they don’t seem to have performed real or simulated experiments, merely analysing the connectome.

Putting all that aside, and supposing we accept the paper’s own interpretations, the next natural question is: so what? It’s interesting that neurons group and fire in this way, but what does that tell us about how the brain actually functions? There’s a suggestion that the pattern of moving up to higher order simplices represents processing of a sensory input, but in what way? In functional terms, we’d like the processing of a stimulus to lead on to action, or perhaps to the recording of a memory trace, but here we just seem to see some neurons get excited and then stop being excited. Looking at it in simple terms, simplices seem really bad candidates for any functional role, because in the end all they do is deliver the same output signal as a single neural connection would do. Couldn’t you look at the whole thing with a sceptical eye and say that all the researchers have found is that a persistent signal through a large group of neurons gradually finds an increasing number of parallel paths?

At the end of the paper we get some speculation that addresses this functional question directly. The suggestion is that active high-dimensional simplices might be representing features of the stimulus, while the grouping around cavities binds together different features to represent the whole thing. It is, if sketchy, a tenable speculation, but quite how this would amount to representation remains unclear. There are probably other interesting ways you might try to build mental functions on the basis of escalating simplices, and there could be more to come in that direction. For now though, it may give us interesting techniques, but I don’t think the paper really delivers on its promise of a link with function.

Can we, one day, understand how the neurology of the brain leads to conscious minds, or will that remain impossible?

Round here we mostly discuss the mind from a top-down, philosophical perspective; but there is another way, which is to begin by understanding the nuts and bolts and then gradually working up to more complex processes. This Scientific American piece gives a quick view of how research at the neuronal level is coming along (quite well, but with vastly more to do).

Is this ever going to tell us about consciousness, though? A point often quoted by pessimists is that we have had the complete ‘wiring diagram’ of the roundworm Caenorhabditis elegans for years (Caenorhabditis has only just over 300 neurons and they have all been mapped) but still cannot properly explain how it works. Apparently researchers have largely given up on this puzzle for now. Perhaps Caenorhabditis is just too simple; its nervous system might be quirky or use elegant but opaque tricks that make it particularly difficult to fathom. Instead researchers are using fruit fly larvae and other creatures with nervous systems that are simple enough to deal with, but large enough to suggest that they probably work in a generic way, one that is broadly standard for all nervous systems up to and including the human. With modern research techniques this kind of approach is yielding some actual progress.

How optimistic can we be, though? We can never understand the brain by knowing the simultaneous states of all its neurons, so the hope of eventual understanding rests on the neurology of the brain being legible at some level. We hope there will turn out to be functions that get repeated, that firm building blocks of some intelligible structure; that we will be able to deduce rules or a kind if grammar which will let us see how things work on a slightly higher level of description.

This kind of structure is built into machines and programs; they are designed to be legible by human beings and lend themselves to reverse engineering. But the brain was not designed and is under no obligation to construct itself according to regular plans and principles. Our hope that it won’t turn out to be a permanently incomprehensible tangle rests on several possibilities.

First, it might just turn out to be like that. The computer metaphor encourages us to think that the brain must encode its information in regular ways (though the lack of anything strongly analogous to software is arguably a fly in the ointment). Perhaps we’ll just get lucky. When the structure of DNA was discovered, it really seemed as if we’d had a stroke of luck of this kind. What amounted to a long string of four repeated characters, ones that given certain conditions could be read as coding for many different proteins; it looked like we had a really clear legible system of very general significance. It still does to a degree, but my impression is that the glad confident morning is over, and now the more we learn about genetics the more complex and messy it gets. But even if we take it that genetics is a perfect example of legibility, there’s no particular reason to think that the connectome will be as tractable as the genome.

The second reason to be cheerful is that legibility might flow naturally from function. That is, after all, pretty much what happens with organs other than the brain. The heart is not mysterious, because it has a clear function and its structure is very legible in engineering terms in the light of that function. The brain is a good deal more complex than that, but on the other hand we already know of neurons and groups of neurons that do intelligibly carry out functions in our sensory or muscular systems.

There are big problems when it comes to the higher cognitive functions though. First, we don’t already understand consciousness the way we already understand pumps and levers. When it comes to the behaviour of fruit fly larvae, even, we can relate inputs and outputs to neural activity in a sensible way. For conscious thought it may be difficult to tell which neurons are doing it without already knowing what it is they’re doing. It helps a lot that people can tell us about conscious experience, though when it comes to subjective, qualities experience we have to remember that Zombie Twin tells us about his experiences too, though he doesn’t have any. (Then again, since he’s the perfect counterpart of a non-zombie, how much does it matter?)

Second, conscious processing is clearly non-generic in a way that nothing else in our bodies appears to be. Muscle fibres contract, and one does it much like another. Our lungs oxygenate our blood, and there’s no important difference between bronchi. Even our gut behaves pretty generically; it copes magnificently with a bizarre variety of inputs, but it reduces them all to the same array of nutrients and waste.

The conscious mind is not like that. It does not secrete litres of undifferentiated thought, producing much the same stuff every day and whatever we feed it with. On the contrary, its products are minutely specific – and that is the whole point. The chances of our being able to identify a standard thought module, the way we can identify standard functions elsewhere, are correspondingly slight as a result.

Still, one last reason to be cheerful; one thing the human brain is exceptionally good at is intuiting patterns from observations; far better than it has any right to be. It’s not for nothing that ‘seeing’ is literally the verb fir vision and metaphorically the verb for understanding. So exhibiting patterns of neural activity might just be the way to trigger that unexpected insight that opens the problem out…

Picture: Astrocyte. Alfredo Perreira Jnr has kindly let me see an ambitious paper he and Leonardo Ferreira Almada have produced: Conceptual Spaces and Consciousness: Integrating Cognitive and Affective Processes (forthcoming in the International Journal of Machine Consciousness).

The unifying theme of the paper is indeed the integration of emotional and neutral cognitive processes, but it falls into two distinct parts.

The first, drawing on the work of Peter Gärdenfors, sets out the heady vision of a universal state space of consciousness.  Such a state space, as I understand it, would be an imaginary space constructed from a large number of dimensions each corresponding to one of the continuously variable aspects of consciousness. In principle this would provide a model of all possible states of consciousness, such that anyone’s life experience would form a path through some area of the space.

The challenges involved in actually populating such a theoretical construct with real data are naturally daunting. Perreira and Almada suggest that it could be approached on the basis of reported states of consciousness. An immediate problem is that qualia, the essence of subjective experience, are widely considered to be unreportable: Perreira and Almada meet this head on by adopting the heterophenomenology of Daniel Dennett: this approach (which I think implies scepticism about ineffable qualia) is based on studying phenomenal experience indirectly, through what subjects say about their own: the third-person perspective. Perreira and Almada note that Dennett adopted this stance mainly as a means of refuting first-person approaches, but I’m sure he would (or will) be delighted to hear of its being adopted as the explicit basis of serious research.  It’s implicit in this approach that we’re dealing with states that are capable of ‘inter-subjective validation’, that is, that they’re states which are accessible to all conscious entities.  This rules out objections on the grounds that, say, Andy having experience X is not the same as Bill having experience X, though in so doing it may appreciably impoverish the scope of the exercise. It could be that the set of experiences common to all conscious beings is actually a significantly restricted sub-set of the whole realm of conscious experience.  For that matter, can we afford to ignore the unconscious or the subconscious?  At times the borderline between conscious states and their near relations may be blurry.

I think two other worries are worth a mention. The state space model suggests that all trajectories are equally valid, but it seems unlikely that this is the case here. Consciousness is a stream, both emotionally and cognitively: certain kinds of state naturally follow other kinds of state. In fact, it doesn’t seem too much to claim that some states refer to previous states: we can’t repent our anger intelligibly without having first been angry. The business of reference, moreover, is a problem in itself.  We’ve already excluded the possibility of Andy’s anger being different from Bill’s, but we can also be in the same state of anger about different things, which seems a material difference. Me being angry about my tax return is not really the same state of consciousness as me being angry about receiving a parking ticket, though in principle the anger itself could be identical. Because, thanks to the miracle of intentionality, we can be angry about anything, including imaginary and logically absurd entities, this is a large problem. Either we exclude these intentional factors – and put up with a further substantial impoverishment of our state space – or the size of our state space balloons out infinitely in all directions.

The practical problems are not necessarily fatal, of course: it’s not as if Perreira and Almada were actually proposing a fully-documented description of the universal state space. What they do suggest is that if we assume another state space (wow!) corresponding to all the possible biophysical states of the brain, we can then hypothesise a mapping of points in one space to points in the other, which would give us the prize of a reduction of conscious experience to physical brain function.  Now I think a biophysical state space of the brain faces formidable difficulties of its own: for one thing we really don’t know exactly which biophysical features of the brain are functionally relevant; for another different brains are not wired the same way – and of course the sheer complexity of the thing is mind-boggling. The biophysical state space of a single neuron is a non-trivial proposition.

However, at a purely theoretical level, this is a nice rigorous statement of what the much-sought Neural Correlates of Consciousness might actually be. If we merely claim that there is a mapping between the two state spaces, we have a sort of rock-bottom version of NCCs, a possible statement of the minimum claim.  We would expect there to be some more general correspondences and matches between regions and trajectories in the two spaces – though I think it would be optimistic to expect these to be simple (and constructing the two state spaces and then observing the regularities would be a remarkably long way round to discovering correspondences between brain and mind activity). Still, the fact that these pesky NCCs turn out to be more abstract and problematic than we might have hoped is in itself a conclusion worthy of note.

All these heroic speculations are in any case just the hors d’oeuvres for Perreira and Almada: the state space of consciousness would have to represent emotional affect as well as rational cognition: how would that work?  They proceed to review a series of proposals for integration emotion and cognition. Damasio’s Somatic Marker Hypothesis, which has emotional affect deriving from states of the body is favourably considered, though criticised for elements of circularity. The alternative view that emotions come first avoids such problems but is ciriticised for not squaring with empirical evidence. Perreira and Almada suggest a better third alternative might be based on mapping the complex inter-relations of cognition and effect, and give a friendly mention to the oft-quoted Global Workspace theory of Bernard Baars. Now we can begin to see where the discussion is going, but first the paper brings in a new element.

This is a discussion of what actually makes mental states conscious – embodiment, higher order states, or what? Perreira and Almada look at a number of proposals, including Arnold Trehub’s retinoid system and Tononi’s concept of Phi, a measure of integrated information. Briefly, they conclude that something beyond all these approaches is needed, and something which puts the integration of affect and cognition at the heart of the system.

Now we come to the second major part of the paper, where Perreira and Almada introduce their own proposal: step forward the astrocytes.

Astrocytes are the most common form of glial cell, which are ‘the other brain cells’. Neurons have generally had all the glory in the past; historically it was assumed that the role of glia was essentially as packing for the neurons – in fact ‘glia’ is the Greek word for ‘glue’.  In recent years, however, it has become gradually clearer that glia, and astrocytes in particular, are more important than that. They form a second network of their own, across which ‘calcium waves’ are propagated.  I think it would be true to say that the standard view now sees astrocytes as important in supporting and modulating neural function, while still reserving the main functional signficance for all those showy synaptic fireworks that neurons engage in. Perreira and Almada want to give the neural and glial networks something like parity of esteem. The proposal, in essence, is that plain cognition is done by the neurons, while feelings are carried by large astrocytic calcium waves: only when the two come together does consciousness arise. Consciousness is the “astroglial integration of information contents carried by neurons”.

What about that?  It’s a bold and novel hypothesis (something we certainly need); it’s at least superficially plausible and has a definite intuitive appeal.  But there are objections.  First, there seem to be other established candidates for the role of feeling-provider.  We know that certain parts of the brain are required for certain kinds of affect – the role of the amygdala in producing ‘fear and loathing’ (or perhaps we should say ‘reasonable distrust and aversion’) has been much discussed. Certain emotions are almost proverbially (which of course is not to say accurately) associated with hormones and the action of certain glands. This needs to be addressed, but I don’t think Perreira and Almada would have too much difficulty in setting out a picture which accomodated these other systems.

More difficult I think are two more fundamental questions. Why would astrocytic calcium waves cause, or amount to, feelings?  And why would those feelings, when associated with cognitive information, constitute consciousness? Damasio’s and other theories can offer a clearer answer on the first point because it’s at least plausible that emotional states can be reduced to the pounding heart, the watering eyes, the churning stomach: calcium waves rippling across the brain somehow don’t seem as obviously relevant. And then, is it really the case that all conscious states have emotional affect? Perreira and Almada suggest that if neurons alone are involved (or astrocytes alone) all you get is proto-consciousness: but intuitively there doesn’t seem anything difficult about completely dispassionate but fully conscious thought.

One strength of the theory is that it seems likely to be more open to direct scientific testing than most theories of consciousness: a few solid experiments would probably relegate the kind of objection I’ve mentioned to secondary status. So perhaps we’ll see…