Edelman, Gerald - Neural Darwinism

7/28/2019 Edelman, Gerald - Neural Darwinism

1/8

6(101y

Neural DarwinismA GLOBAL BRAIN THE ORY

here is one simple principie that governs howthe brain works: it evolved; that is, it was not designed.As stated, this principie sounds almost simple-minded,doesn't it? But we must not forget that, although evolu-tion is not intelligent, it is enormously powerful. Thepower comes from natural selection acting in complexenvironments over eons of time. A key idea developedby Darwin is embedded in his notion of populationthinking: functioning structures and whole organismsemerge as a result of selection among the diverse variant

individuals in a population, which compete with one an-other for survival. I hold this notion to be central, notonly in considering how the brain has evolved, but alsoin thinking about how it develops and functions.Applying population thinking to understanding how thebrain works leads to a global theory, called neural Dar-winism or the theory of neuronal group selection.What do we mean by the term "global" and whydo we need a global brain theory? An explanation ofconsciousness will necessarily require an understandingof perception, memory, action, and intentioninshort, an overall understanding of how the brain worksthat goes beyond the functioning of one brain regionor another. Given the richness, variety, and range ofconscious experience, it is also important to constructa brain theory that is principled and compatible withevolution and development. By principled, I mean atheory that describes the principies governing the majormechanisms by which the brain deals with informationand novelty. One such theory or model is the idea thatthe brain is like a computer or Turing machine. In con-trast to such an instructive model, which relies on pro-grams and algorithms, models based on populationthinking rely on selection of particular elements or statesfrom a large repertoire of variant elements or states. Ex-planations of consciousness based on one or the otherof these two kinds of models differ greatly. By now, itshould be no mystery that I prefer selectional modelsbased on population thinking.

NEURAL DARWINISM323


2/8

The reason popu lation thinking is important in de-termining how the brain works has to do with the ex-traordinary amount of variation in each individualbrain. This is true at all levels of structure and function.Different individuals have different genetic influences,different epigenetic sequences, different bodily re-sponses, and different histories in varying environments.The result is enormous variation at the levels of neuronalchemistry, network structure, synaptic strengths, tem-poral properties, memories, and motivational patternsgoverned by value systems. In the end, there are obviousdifferences from person to person in the contents andstyles of their streams o f consciousness. The variabilityof individual nervous systems was commented on by thedistinguished neuroscientist Karl Lashley, who admittedthat he had no ready explanation for the existence of somuch variation. Even though there are general patternsexhibited by the brain in the face of this variation, itcannot be dismissed as mere noise. There is too muchof it, and it exists at too many levels of o rganization-molecules, cells, and circuits. It is simply not likely thatevolution, like a com puter programmer dealing withnoise, could have devised multiple error-correctingcodes to assure prese rvation of patterns in the brain bycounteracting this enormous variation.

An alternative way of confronting neural variabilityis to consider it fundamental and to assume that theindividual local differences within each brain make uppopulations of variants. In this case, selection from suc h

a population of variants could lead to patterns even un-der unpredictable circumstances, provided that someconstraint of value or fitness was satisfied. In evolution,fitter individuals survive and have more progeny. In theindividual brain, those synaptic populations that matchvalue systems or rewards are m ore likely to survive orcontribute more to the production of future behavior.This view is in sharp contrast to computer modelsof the brain and mind. According to these models, sig-nals from the environment c arry input information thatis unambiguous, once contam inating noise is averagedaway or otherwise dealt with. These models assume thatthe brain has a set of programs, or so-called effective pro-cedures, which are capable of changing states based onthe information carried by the inputs, yielding function-ally appropriate outputs. Such models are instructive inthe sense that information from the world is assumed toelicit the formation of appropriate responses based onlogical deduction. These models do not deal, however,with the fact that inputs to the brain are not unambigu-ous the world is not like a piece of tape with a fixedsequence of symbols for the brain to read. I have alreadymentioned the challenge to computer models of thebrain posed by the richly variable circuitry of real brains.

There is also a set of functional issues that makecomputer models unlikely. For example, the mappedconnections from the sense of touch in the handthrough the thalamus to the region of somatosensorycortex are variable and plastic, even in adults. The sub-

NEURAL DARWINISM34

NEURAL DARWINISM35


3/8

Figure 4. Il lusory contours in a K anizsa triangle. Mostpeople report the appearance of a distinct triangular shapeand an increase in apparent lum inance within the triangle,but neither of there features exists in the physical image.

received from the two sides of the contour that is per-ceived. Such a contour is called "illusory." The brainconstructs the contour, which, by the way, is not neces-sarily a straight line but can be curved depending on thecontext of the particular figure used.

Many other functional responses of the perceivinganimal could be described to illustrate why an a prioriprogram is not a likely explanation for physiological orpsychological properties. I shall mention only two more.The first is the remarkable tendency of brains to seekout closure and avoid gaps. In daily life, for example,you do not see the blind spot in your visual field occa-sioned by the presence of the optic nerve near the centerof your retina. Even more striking phenomena come

NEURAL DARWINISM37

regions in the somatosensory cortex mapping the fingersdynamically shift all their boundaries as a result of exces-sive use of even one fingera shift in the context of use.Similar phenomena reflecting such context dependenceand dynamic circuit variation are seen for other senses.Furthermore, in sensory systems such as that for vision,there are multiple cortical regions that are each func-tionally segregated, for example, for color, movement,orientation, and so on. These functionally specializedareas can exceed thirty in number and are distributedafl over the brain. Yet there is no superordinate area orexecutive program binding the color, edge, form, andmovement of an object into a coherent percept. Thisbinding is not explicable by invoking a visual computerprogram operating according to the principies of artifi-cial intelligence. A coherent percept in fact neverthelessemerges in various contexts, and explaining how this oc-curs constitutes the so-called binding problem. A globalbrain theory must provide a cogent solution to thisproblem by proposing an appropriate mechanism. Itwill soon become clear that such a solution is central toour understanding of consciousness.

To emphasize the dependence of perception oncontext, we may call upon the huge phenomenologyof illusions, visual and otherwise. One example is theKanizsa pattern, which consists of the angular portionsof a triangle, disconnected, but appears to show an over-lying triangle with sharp boundaries (Figure 4). Yetthere is no true energy difference in the light that is

NEURAL DARWINISM36


4/8

from the field of neuropsychology, which, among otherthings, studies responses to strokes. This field is repletewith examples of closure phenomena that can even bedelusional. A most exotic example is anosognosia, a syn-drome in which a paralyzed patient does not recognizethe existence of paralysis even if it involves his or herentire left side. In such cases, we see extraordinary adap-tation and integration by the damaged brain as it re-sponds to the loss of cortical areas.

In addition to construction and closure, and possi-bly in connection with them, the brain's capacity to gen-eralize is astonishing. A case in point is the ability ofpigeons, when appropriately rewarded, to look at nu-merous photographs of various fish species in differentscales and contexts and learn to positively recognize thesimilarity in the photographs. Pigeons trained at thistask can recognize that these diverse pictures have some-thing in common more than 80 percent of the time. Itis highly unlikely that this behavior is the result of afixed template or a set of predetermined algorithms inthe brains of pigeons. Nor can it be explained by naturalselection for the positive recognition of fish. Pigeonsneither evolve with fish nor live with them, and theydon't eat them either.

I could cite many more examples ranging from thedevelopmental anatomy of the brain to the individualvariation of brain scans in humans carrying out similartasks. But the conclusion is clear: the brains of higher-level animals autonomously construct patterned re-

NEURAL DARWINISM38

sponses to environments that are full of novelty. Theydo not do this the way a computer doesusing formalrules governed by explicit, unambiguous instructions orinput signals. Once more, with feeling: the brain is nota computer, and the world is not a piece of tape.

If the brain is in fact not a computer and the worldis not a piece of tape, how can the brain operate so as toyield adaptive and patterned responses? As I have alreadysuggested, the answer lies in a selectionist theory that Ihave called the theory of neuronal group selection, orTNGS (Figure 5). This theory has three tenets: (1) De-velopmental selectionduring the early establishmentof neuroanatomy, epigenetic variations in the patternsof connections among growing neurons create reper-toires in each brain area consisting of millions of variantcircuits or neuronal groups. The variations arise at thelevel of synapses as a result of the fact that neurons thatfire together wire together during the embryonic andfetal stages of development. (2) Experiential selection-overlapping this first phase of selection and alter the ma-jor neuroanatomy is built, large variations in synapticstrengths, positive and negative, result from variationsin environmental input during behavior. These synapticmodifications are subject to the constraints of value sys-tems described in the previous chapter. (3) Reentry-during development, large numbers of reciprocal con-nections are established both locally and over longdistances. This provides a basis for signaling betweenmapped areas across such reciprocal fibers. Reentry is

NEURAL DARWINISM39


5/8

Cell divisionCell deathProcess

extension

Developmentalselection(Yieldingprimary

repertoire)

Experientialselection(Yielding

secondaryrepertoire)

Changes instrength ofpopulationso synapses

Reentrantmapping

InputnputtooMap 1ap 2 49/InputnputtooMap 1ap 2Figure 5. The three main tenets of the theory of neu ronalgroup selection, or neural Darwinism: (1) Developmentalselection leads to a highly diverse set of circuits, one of

which is shown. (2) Experiential selection leads to changes inthe connection strengths of synapses, favoring some pathways(thickened black Enes) and weakening others (dashed unes).

(3) Reentrant mapping, in which brain maps are coordinatedin space and time through ongoing reentrant signaling across

reciprocal connections. The black dots in the maps on theright indicate strengthened synapses. As a result of (1) and

(2), a myriad of circuits and functioning pathways is createdconstituting a repertoire for selectional events. The furtherand ongoing even ts of reentry in (3) must be thought of as

dynamic and recursive, mapping the maps over time.

the ongoing recursive interchange of parallel signalsamong brain areas, which serves to coordinate the activi-ties of different brain areas in space and time. Unlikefeedback, reentry is not a sequential transmission of anerror signal in a simple loop. Instead, it simultaneouslyinvolves many parallel reciprocal paths and has no pre-scribed error function attached to it.

The consequence of this dynamic process is thewidespread synchronization of the activity of widely dis-tributed neuronal groups. It binds their functionally seg-regated activities into circuits capable of coherent out-put. In the absence of logic (the organizing principie ofcomputers as instructive systems), reentry is the centralorganizing principie that governs the spatiotemporal co-ordination among multiple selectional networks of thebrain. This solves the binding problem that I mentionedearlier. Through reentry, for example, the color, orienta-tion, and movement of a visual object can be integrated.No superordinate map is necessary to coordinate andbind the activities of the various individual maps thatare functionally segregated for each of these attributes.Instead, they coordinate by communicating directlywith each other, through reentry.

The three tenets of the TNGS together form a se-lectional system. Prominent examples of selectional sys-tems include evolution, the immune system, and com-plex nervous systems. All follow a set of three guidingprincipies. The first principle assumes a means for gen-erating diversity in a population of elements, whether

NEURAL DARWINISM41


6/8

of individuals or of cells. The second is a means allowingextensive encounters between individuals in a variantpopulation or repertoire and the system that is to berecognized, whether it is an ecological environment, aforeign molecule, or a set of sensory signals. The thirdprincipie is some means to differentially amplify thenumber, survival, or influence of those elements in thediverse repertoire that happen to meet selective criteria.In evolution, these are criteria of fitness allowing thedifferential survival and breeding of certain individu-alsthe process of natural selection itself. In immunity,amplification occurs through the enhanced division ofjust those clones of immune cells having antibodies ontheir surface that bind particular foreign molecules orantigens weli enough to exceed a certain critica' energyof binding. In neural systems, amplification consists ofenhancing the strengths of those synapses and circuitsof neuronal groups that meet the criteria set by valuesystems. It is the neuronal groups made up of excitatoryand inhibitory neurons in particular anatomical patternsrather than individual neurons that are selected.

Notice that while these three different selectionalsystems obey similar principies, they use different mecha-nisms to achieve successful matching to various unfore-seen inputs. Evolution is, of course, special and over-arching because it is also responsible for actuallyselecting the different mechanisms used by the immuneand nervous systems. It tends to favor those individualsthat successfully utilize such mechanisms to improve

their fitness and allow more of their progeny to sur-vive.

Since the proposal of the TNGS in 1978, a grow-ing body of evidence has supported the notion that neu-ronal groups connected by reentrant interactions are theselectional units in higher-level brains. This evidence ispresented in a number of books and papers and willnot be reviewed here. Instead, I will consider certainconsequences of the theory that are particularly impor-tant for understanding the mechanisms underlying con-sciousness.

One important consequence is that the brain is soversatile in its responses because those responses are de-generate. Degeneracy is the ability of structurally differ-ent elements of a system to perform the same functionor yield the same output. A clear-cut example is seen inthe genetic code. The code is made up of triplets ofnucleotide bases, of which there are four kinds: G, C,A, and T. Each triplet, or codon, specifies one of thetwenty different amino acids that make up a protein.Since there are sixty-four different possible codons-actually sixty-one, if we leave out three stop codons-which makes a total of more than one per amino acid,the code words are degenerate. For example, the thirdposition of many triplet codons can contain any one ofthe four letters or bases without changing their codingspecificity. If it takes a sequence of three hundred co-dons to specify a sequence of one hundred amino acidsin a protein, then a large number of different base se-

NEURAL DARWINISM42 NEURAL DARWINISM43


7/8

map to those of others to form a functioning circuit.Simulations show that the neurons that yield such cir-cuits fire more or less in phase with each other, or syn-chronously. But in the next time period, different neu-rons and neuronal groups may form a structurallydifferent circuit, which nevertheless has the same out-put. And again, in the succeeding time period, a newcircuit is formed using some of the same neurons, aswell as completely new ones in different groups. Thesedifferent circuits are degeneratethey are different instructure but they yield similar outputs to solve thebinding problem (Figure 6).

Within each particular circuit, the different neu-ronal groups fire synchronously. The different circuitsyielding the same output are not, however, synchronousor in phase with each other, nor do they have to be.As a result of reentry, the properties of synchrony andcoherency allow more than one structure to give a simi-lar output. As long as such degenerate operations occurin succession to link distributed populations of neuronalgroups, there is no need for an executive or superordi-nate program as there would be in a computer.

The formulation of a global brain theory like theTNGS, while essential to understanding how the brainworks, does not solve all of the detailed mechanisticproblems related to the local operations of networks inthe various nuclei and regions of the brain. But it doesremove the paradoxes that arise if one assumes that thebrain functions like a computer. One such paradox

NEURAL DARWINISM45

quences in messages (approximately 3 1 0 0) can specify thesame amino-acid sequence. Despite their different struc-tures at the level of nucleotides, these degenerate mes-sages yield the same protein.

Degeneracy is a ubiquitous biological property. Itrequires a certain degree of complexity, not only at thegenetic level as I have illustrated aboye, but also at cellu-lar, organismal, and population levels. Indeed, degener-acy is necessary for natural selection to operate and it isa central feature of immune responses. Even identicaltwins who have similar immune responses to a foreignagent, for example, do not generally use identical combi-nations of antibodies to react to that agent. This is be-cause there are many structurally different antibodieswith similar specificities that can be selected in the im-mune response to a given foreign molecule.

Degeneracy is particularly important in helping tosolve major problems in complex nervous systems. Ihave already mentioned the binding problem. How canit be that, despite the absence of a computer program,executive function, or superordinate map, up to thirty-three functionally segregated and widely distributed vi-sual maps in the brain can nevertheless yield perceptionthat coherently binds edges, orientations, colors, andmovement into one perceptual image? How do differentmaps for color, orientation, object movement, and soon correlate or coordinate their responses? As I suggestedaboye, the answer lies in mutual reentrant interactionsthat, for a time, link various neuronal groups in each

NEURAL DARWINISM44


8/8

OUTPUT munculus, a little man who lives in the brain, tointerpret the meaning of a percept. Just as Darwin's the-ory of natural selection disposed of the argument fromdesign, the TNGS disposes of the need for either a fixedinstructional plan or a homunculus in the head.

These issues are directly relevant to my next task,which is to show how the principies and mechanismsof the TNGS can be used to understand the origin ofconsciousness.

Outputs: At time t, CAt time t+1, AAt time t+2, B

Figure 6. I llustration of the degene racy of reentrant circuitsin the brain. Even though the three overlapping circuits inA, B, and C are different, as shown by the shading, theycan yield a similar output over some pe riod of time.

would have us imagine a cell with a designated categori-cal function that dominates the function of all subord-nate neurons connected to itfor example, a cell thatfires when you think of a particular person, a so-calledgrandmother cell. Such a cell is not necessary in thistheory. Different cells can carry out the same functionand the same cell can, at two different times, carry outdifferent functions in different neuronal groups. More-over, given the selectional nature of higher-order inter-actions in the brain, one does not have to invoke a ho-

NEURAL DARWINISMNEURAL DARWINISM46 47

Documents

Edelman, Gerald - Neural Darwinism