Natural Selection of Neurons as Learning


Folksonomies: cognition learning

How Brains Grow Into Bodies

Brain wiring begins with the outgrowth of axons. Once a newborn neuron has migrated, planting its cell body in a permanent position, it sends out a fine axon shoot with an enlarged tip known as a growth cone. At the end of the growth cone are about a dozen long tentacles that shoot out in all directions and act like radar, picking up all manner of navigational signals. They feel out the best-textured surfaces, sniff around for chemical cues, and even use tiny electrical fields to help the axon find its way to appropriate targets. Axons can grow to very great lengths, so long-distance connections, which pose the greatest challenge, tend to get an early start, at a time when the absolute distance between any two parts of the embryo (say, the spinal cord and the toe) is still comparatively short. Axon guidance also makes use of specific chemical attractants, released, much like insect pheromones, by potential target neurons to attract synaptic mates over relatively long distances. Led by their own genetically coded receptor molecules, these axons can\'t help but elongate in the direction of an ever-increasing concentration of the attractant molecule until they reach its source, the target neurons with a matching chemical identity.

Once an axon completes its traverse, whether near or far, it branches out extensively, contacting up to hundreds of target neurons that have released the same potent lure. Contact leads to synapse formation, but these initial connections are promiscuous: both far too numerous and highly unselective. During infancy and early childhood, the cerebral cortex actually overproduces synapses, about twice as many as it will eventually need. The initial wiring scheme is thus quite diffuse, with a lot of overlap that makes information transfer both imprecise and inefficient. It\'s as if all those billions of phones were first connected as party lines; you could dial Grandma at any of thousands of numbers, but it\'s unlikely she\'d be the first to answer.

Why does the brain bother to produce so many excess synapses? Why not save time and energy and simply wire things up precisely from the start? The answers to these questions cut right to the core of the nature/nurture issue.

Up to now, genes have been largely responsible for establishing brain wiring. They prescribe all the early targeting cues—the pheromones that attract one class of axon to a particular class of neuron, the surface receptors that sense these attractants (or in some cases, repellents), as well as the receptors for other chemical, textural, and electrical cues that guide axon growth and synapse formation. But the fact is that there are not nearly enough genes in the entire human genome to accurately specify every one of our quadrillion synapses. There are perhaps 80,000 genes scattered among the miles of DNA in our chromosomes, and even if a generous half of these were allotted to the delicate job of brain wiring (after all, the body does have some other important functions to perform with its genes), we would still be far short of having enough cues to specify an accurate wiring diagram for the entire brain.

This is where \"nurture\" steps in and finishes the job. By overproducing synapses, the brain forces them to compete, and just as in evolution or the free market, competition allows for selection of the \"fittest\" or most useful synapses. In neural development, usefulness is defined in terms of electrical activity. Synapses that are highly active—that receive more electrical impulses and release greater amounts of neurotransmitter—more effectively stimulate their postsynaptic targets. This heightened electrical activity triggers molecular changes that stabilize the synapse, essentially cementing it in place. Less active synapses, by contrast, do not evoke enough electrical activity to stabilize themselves and so eventually regress. (See Figure 2.7.) It\'s \"use it or lose it\" right from the start; like other forms of Darwinian selection, this synaptic pruning is an extremely efficient way of adapting each organism\'s neural circuits to the exact demands imposed by its environment.

Our best evidence for how experience guides synaptic selection comes from studies of visual development... But there\'s another dramatic demonstration—some classic experiments on laboratory rats that were inspired by something Charles Darwin himself described back in 1868.

Ever the careful observer, Darwin rounded up a bunch of rabbits, measured their head and body sizes, and found that those raised in captivity had far smaller brains, relative to body weight, than those that grew up in the wild. Compared to the wild rabbits, Darwin realized, the domestic rabbits \"cannot have exerted their intellect, instincts, senses and voluntary movements, either in escaping from various dangers or in searching for food,\" so that \"their brains will have been feebly exercised, and consequently have suffered in development.\"

A century later, neurobiologists finally started to figure out how a challenging environment stimulates brain growth. Much like Darwin\'s rabbits. laboratory rats that have been reared in an \"enriched\" environment—in a large cage containing several litters and a wide variety of \"toys\" to see. smell, and manipulate—^have larger brains, with a notably thicker cerebral cortex, than those raised in an \"impoverished\" environment—isolated, in a small empty cage, without any social stimulation and a bare minimum of sensory experience. The reason their cerebral cortex is bigger, researchers have found, is that their neurons are larger, with bigger cell bodies, more dendritic branches, more spines, and more synapses than those in the brains of impoverished rats. In other words, the extra sensory and social stimulation actually enhances the connectivity of the enriched rats\' brains, a difference that probably explains why they are also smarter—they learn their way around a baited maze significantly faster—than their impoverished laboratory mates.

It is no great stretch to see the implication of these experiments for human development: A young child\'s environment directly and permanently influences the structure and eventual function of his or her brain. Everything a child sees, touches, hears, feels, tastes, thinks, and so on translates into electrical activity in just a subset of his or her synapses, tipping the balance for long-term survival in their favor. On the other hand, synapses that are rarely activated—whether because of languages never heard, music never made, sports never played, mountains never seen, love never felt—will wither and die. Lacking adequate electrical activity, they lose the race, and the circuits they were trying to establish—for flawless Russian, perfect pitch. an exquisite backhand, a deep reverence for nature, healthy self-esteem— never come to be.

The magnitude of this synaptic sorting is enormous. Children lose on the order of 20 billion synapses per day between early childhood and adolescence. While this may sound harsh, it is generally a very good thing. The elimination of stray synapses and the strengthening of survivors is what makes our mental processes more streamlined and coherent as we mature; the party lines sort themselves out into clear, private, efficient channels for information transfer. On the other hand, it may also explain why our mental processes become less flexible and creative as we mature. Although the brain continues to exhibit certain more subtle forms of plasticity in adulthood (which is, after all, the way we learn or remember anything at all), it is never as malleable as in childhood.

Notes:

Best description yet of the synaptic \"pruning\" human brains go through as the brain wires up to the body and best reason yet for why children should have rich, mentally-nourishing environments in which to grow so that their synapses don\'t get unnecessarily pruned, resulting in smaller brains.

Folksonomies: nature parenting genetics development dna nature vs nurture nurture

Cause and Effect

Instruction vs. Selection

The main difference between an instructional system and a selectional system is that the instructional system uses information from the environment to change the properties of the object in question, but a selectional system has a large and varied population of objects, and the ones that are most fit for the environment are differentially reproduced. Hopefully an example Edelman uses from immunology will help clear this up.

The theory prevailing before the present one was called the theory of instruction. Its basic assumption was that, in the immune system, a foreign molecule transferred information about its shape and structure to the combining site of the antibody molecule. It then removed itself (the way a cookie cutter would be removed from dough) leaving a crevice of complementary shape that could then bind to all foreign molecules with regions having the shape with which the impression was originally made. It is obvious why this is an instructive process: Information about the three-dimensional structure is posited to be necessary to instruct the immune system how to form an antibody protein whose polypeptide chain folds around that structure to give the complementary shape. 
(Edelman, Bright Air, Brilliant Fire)2.7.1

However, it was discovered that this theory was wrong. The immune system was in fact a somatic selection system. Most laypeople do not realize that evolution is really just a search algorithm, and the most efficient one ever found in the proper problem domains. It is a specific example of a selection system.

In evolution, organisms are more or less adapted to events in the environment. This adaptation occurs even when environmental changes cannot be predicted (that is even when the changes represent novelties). The process of adaptation occurs by selection on those organismal variants that are on the average fittest, and what makes them fittest does not require prior explicit information ("instruction") about the nature of the novelties in the environment. The selective environmental changes are, in general, independent of variation in the population of organisms, although selection resulting from such changes may add to that variation. In sum, there is no explicit information transferred between the environment and organisms that cause the population to change and increase its fitness. Evolution works by selection, not by instruction. There is no final cause, no teleology, no purpose guiding the overall process, the responses of which occur ex post fact in each case. 
(Edelman, Bright Air, Brilliant Fire)2.7.2

The algorithm can be run on many different kinds of things and not just biological organisms in evolution. In the immune system it is run on a population of different antibody molecules. Those that encounter foreign molecules are reproduced, and those that do not are not magnified in this way. The algorithm works in the immune system on a much shorter time scale than it does in evolution. The thing that selectional systems give us in this debate that instructional systems do not, is that selectional systems do not have to have any prior knowledge of the environment, and require no explicit information transfer from the world. Whereas with the instructional system you are left with the question of who or what decides what is important for the system to learn!

Notes:

Folksonomies: natural selection selection cognitive development instruction