User:Graeme E. Smith/Disappearing Memory Cells in Laminae I: A role in learning?

Disappearing Memory Cells in Laminae I A Role in Learning? Graeme E. Smith, GreySmith Institute of Advanced Study  http://en.wikiversity.org/wiki/Portal:GreySmith_Institute   http://en.wikiversity.org/wiki/User:GreySmith_Institute  grysmith@telus.net In 1970 Marr introduced A Theory for Cerebral Neocortex which put great weight on the role of the Mossy Dendrites in Laminae I of the Cerebral Neocortex. In 1983 Eccles introduced The Horizontal (Tangential Fiber) System in Laminae I of the Cerebral Neocortex Which indicated that very few nerve cells were found in Laminae I of the Cerebral Neocortex of Adult brains. Assuming both are true, where did the Mossy Dendrites go? In this article I assume that both articles are correct, with respect to Mossy Dendrites and that they merely represent studies done at different times in the life of the Neocortex. I present a hypothesis that might explain why immature brains have more mossy dendrite neurons than adult brains. In 1969 David Marr began to publish a series of studies on the Micro-Anatomy of the Pyramidal Memory areas in the brain, based on the then in vogue assumption that neurons were equivalent to mathematical statements, and therefore that an arrangement of neurons would define a simultaneous equation which would have a probability function, that he could detect. Of interest in this article is his 1970 article A Theory on the Cerebral Neocortex In this theory he introduced the idea of a group of neurons called a CODON which he suggested made the cortex into a "self-classifying content addressable memory". It was of course an oversimplification of the nature of the cerebral neocortex, but it pointed the way towards a more complete model at a later date. Of especial interest to me, is the role that the mossy dendrites in Laminae I seemed to suggest, that of detection of patterns of activation in the inputs to the Cerebral Neocortex, and an almost buss-like local connective architecture, that allowed them to influence the activation state of the associated pyramidal neurons in layer 2/3. From this I devised a hypothesis that in fact the mossy dendrite neurons had a role in training the pyramidal neurons to detect patterns. In this case, it was assumed that the nature of the layer was a highly competitive layer where neurons competed for activation by creating inhibitive links with each other, to suppress alternative patterns of activation. It was naturally assumed that this inhibitive activity would concentrate locally, with the effect of creating local islands of specialization, explaining to some extent Marr's conception that the CODON was self-classifying. Unfortunately Marr also assumed that he would be able to find a probabilistic function for the connections between the neurons, something that did not actually happen. This destroyed the mathematical function model of neural networks because there was a chaotic factor that eliminated probabilistic models. There were two main things going against Marr's probabilistic approach to the placement of connections, the Phenomenal nature of Neural Networks where the location of a memory is indeterminate until the actual time it is distributed across the network, which makes selection of particular memories within the implicit memory difficult if not impossible, and the opportunistic nature of connections between neurons which adds a chaotic factor to the connections that means that no two brains are connected simularly. The best method of dealing with these confounding factors in organization is the concept of Neural Darwinism, which allows the brain to self-organize to overcome variations in local anatomy by the gyrus and sulcys organizational level. Of interest but not well explained in the book, is the concept of a Neural Group, which is defined by hundreds of neurons acting on each other pre-firing to promote a very small number (1 or 2) neurons in the group to fire for the whole group. Marr's theory did not attempt to deal with this larger micro-architecture, and so the two theories can be seen to be to some extent complimentary. In 1983 JC Eccles wrote an article The Horizontal (Tangential Fibers) System of Laminae I of the Cerebral Neocortex in which he takes a closer look at the micro-anatomy the brain, and determines that there are very few neurons in Laminae I of the Adult Brain, but that the so called Tangential Fiber system of Laminae I, is a complex matted fiber layer that takes it's inputs from a number of different sources. At one point in his article he shows that the tangential fibers form a classic instar at the dendrite of the pyramidal neuron suggesting a direct connection between these different input sources, and the dendrites. He also draws an inference from the inhibitive field of the Martinotti cells, that the role of the spiky neuron is to block the inhibitive field. This suggests that inputs to the granular layer in the 4th laminae, trigger the pyramidal cells in their columns by blocking the inhibitive effects of the Martinotti cells, and thus, sensitizing the pyramidal cells to patterns of inputs from layer 4. If so, it would seem that role of the mossy dendrite neurons is no longer needed in the adult brain which is why they are sloughed off at an earlier age, and very few are left by adulthood. If we accept this then we are left I think, to wonder what the role of the mossy dendrite neurons is, if it can be mostly dispensed with by adulthood. One hypothesis that I put forward, is that the mossy dendrites in laminae I of the immature cerebral neocortex are there to impose an arbitrary classification system that connects the inputs in the tangential fibers with the inputs of layer 4, in essence correlating multiple sources of inputs so that the same patterns are detected despite the individualistic nature of each source. If this is so, then I suggest the role of the mossy fiber neurons is to compare patterns from different input sources, and find the best fit that activates the pyramidal neuron with the closest levels of similar behavior between the different input signals to activate when the signals are detected. Once the Pyramidal neuron has been trained, and consistently fires when the signals are detected, the correlative function can be dispensed with, and so the mossy dendrite neurons become redundant. Mossy dendrite neurons would only remain in areas where there is some variation to the correlation between inputs that requires the correlation function to remain active. This might indicate a reason for a slow down in plasticity that is not related to the production of growth hormones, but is instead related to the stability of the input patterns for different inputs to the cerebral neocortex. People who continue learning in adulthood might retain more mossy dendrite neurons in laminae I than people who are immersed in a stable 9-5 work routine. If so this might be a factor in staving off senescence until later in life, since the mossy dendrite neurons would assist in retraining after strokes.