Specializations > Psycholinguistics
To what extent do the neurons in the brain follow a 'design'?
(1/2) > >>
Is the distribution of neurons random? Do they develop connections at random (based on input)?
It's hard to imagine how the actual locations and specific connections of neurons could be encoded in our DNA. It probably varies a LOT by person and also we adapt so easily to change (such as when some neurons die from damage-- the vast majority of the time we don't even notice-- drinking alcohol kills a few neurons each time, but it isn't, in moderation, problematic).
We have about 86 billion neurons, each connected to about 1000 others. If any information about specific location or connections were encoded in our DNA, the information encoded in the DNA would be vast! I came across an article the other day that suggested the number of 'switches' in a single human brain outnumbers the entire number of equivalent 'switches' in all computers on the planet.
So... I conclude, just by simple and superficial logic, that the details of the arrangement of neurons is not genetic. There is no "design". Clearly there are certain distributional aspects to us genetically, such as where large clusters of neurons go (in the head, not the feet), but I doubt it gets into any sort of detailed neuron-by-neuron level at all.
It is possible that the pattern is somehow determined by growth: an initial connection is pre-established genetically and from that the rest automatically follow based on the programming attached to each individual neuron. Or maybe they're just randomly distributed to general areas and then random connections occur until something fits right.
Either way, if what I'm suggesting is remotely accurate, then the implications for modeling the human brain are profound: all that is required is (roughly) the number of neurons in each broad region and the behavior of neurons. The rest follows directly from those two.
Computer models of neural networks are often crudely designed: arrange neurons randomly (or just in arbitrary rows/layers) and then allow them to connect to others at random and until something fits, plus some input to the system or training. This seems superficial and like a silly programming experiment, but in fact it actually might be exactly what the brain does! Wouldn't that be crazy?
In short, the information held in the brain is probably around (86 billion / 2) * 1000 = 43 trillion direct, binary connections [each neuron has 1000 half-connections]. And as for the number of potential connections throughout the brain, I think it would be something like (43trillion^[43trillion-1]) [assuming no repetition (cross through the same connection twice), which might mean infinity]. Some ridiculously huge number. And that may be absolutely all there is to intelligence. One neuron alone is basically a 0 or a 1. And perhaps that's all our minds are: some huge number of ones and zeros.
[Edit: just fixed some math in there. It's probably not very accurate, but I'd hope vaguely representative.]
If that's all true, then what does it mean for Linguistics? How can "UG" possibly exist, and if so how can it be biological? (Perhaps UG is actually a mathematical requirement on communication systems or symbolic representational systems, if anything at all.) And, even though this questions some of that... how does it all work out anyway? Are we just "smart enough" to do language? We just have enough neurons?
Something about the structure of the brain must be genetic, since it develops in the womb in the first place. The brain isn't made out of one homogenous type of material either, and it is genetics that create these different areas of the brain and map for the proteins and hormones that send messages and allow our body to 'know' how to grow. Most of our thinking doesn't come from learning input, either - neurons are firing that allow us to breathe, and our organs to function. This doesn't all come from learning or the development of a completely blank slate brain. So if all that can organize itself from genetics - I imagine a whole lot else can too.
But this is all from a product of long slow evolution. I agree that 'universal grammar' is unlikely to have evolved or even be selected for - but I don't know if the immense amount of neurons supports this.
Well, there must be some design to the whole system, right? I mean we all have specific regions that host different clusters of cells in the brain that display similar functions across the species. The Brodmann Classification System would just make no sense if there wasn't an underlying structure that tied clusters of neurons that carry out a specific function to a specific area. Having said that, we all know that this isn't fixed and damage to one locus of the brain can result in even the other hemisphere taking on the load and rewiring of those specific functions, all demonstrations of Hebbian theory ("Neurons that fire together, wire together.")
There would have to be a basis, but on top of that basis, a loose system of inter-connectivity that had the freedom to join and be moulded based on specific electrical input. The architecture must be there and if your question is about what happens after that architecture is in place, then I understand your question a bit more. When you look at genetic motor disorders, there is a lot of convincing evidence that this caused by the distribution of neurons in an atypical fashion (genetically inherited) which is likely to have been inherited in the genetic code and therefore mapped via some building block (plan) that would imply a pretty specific plan of neuronal distributions. Having said that, there is clearly some crazy acrobatic qualities to the system and the interplay of those systems is not something I believe is well understood in neuroscience in general.
I'd like to share a word that I may have coined in a discussion with a close friend of mine — it probably has been used before — "neuroplasticity".
Both physically and mentally our minds are quite "plastic", or mouldable, responding to both internal and external "stimuli" of various kinds. So although we most certainly do have some kind of inherent "design" to our brains, like any other part of our body, it wouldn't surprise me if the brain was able to "reshape" itself to environmental pressure.
By this I mean people who have suffered brain damage, particularly in the language specific regions, yet who seem to have regained their abilities as the brain "moulds" itself, with undamaged regions taking on the functions of the damaged regions.
So although there is an inherent "design" and a seemingly basic foundation for our higher-level functions, it might not necessarily be the only arrangement that works. The result of evolution, but not set in stone.
Or at least when it comes to the neocortex? The deeper you go into the brain, the closer to autonomous functions, this probably doesn't hold anymore. This might even be the evolutionary advantage of our large neocortex.
Or? Neurology isn't exactly my specialty, lol
Neurobiology is not my speciality either, but I understand that recently neuroscientists have found that neurons tend to be organized in "motifs" (AKA "neural legos"), which are small groups of neurons (10 or fewer neurons) with the same patterns of connections between them. I blogged about this a few months ago, talking about Markram and Perin (2011) and Perin, Berger, and Markram (2011). The general idea is that Hebbian learning relies on having varying connection strengths between neurons, but neurons in these motifs appear to have the strongest connections that are biologically possible (the connections are saturated). Thus, meaningful learning occurs only between the motifs: it cannot happen within a motif. This means that a motif is effectively a deterministic transformation of its inputs, and different motifs may perform different transformations.
I agree that it is unlikely for individual connections to be genetically coded. However, I think it is possible for the relative prevalence of various motifs in different regions to be genetically coded; some regulation sequence makes a particular motif (= deterministic transformation of inputs) more or less likely at some point in development. I'm not sure how this would fit with the particular view of UG that has been developed in the Generativist literature. However, if we are willing to consider a broader view of UG, then UG may just be a particular probability distribution over these motifs, or over structures of these motifs.
All of this is highly speculative, of course. As far as I know, these motifs have not been directly implicated in language behavior, and may be too low-level to have any discernible reflex in language.
 Message Index
[#] Next pageGo to full version