Bright Air, Brilliant Fire: On The Matter Of The Mind

BRIGHTAIR, BRILLIANTFIRE

M

GERALD

EDELMAN

BRIGHT AIR) BRILLIANT FIRE -

On the Matter of the Mind

0r j BasicBooks A Division of HarperCollinsPz blisbers

Library of Congresscataloging-in-PublicationData Edelman,Gerald M. Bright air, brilliant fire: on the matter of the mind/ Gerald M. Edelman. P. cm. Includesbibliographicalreferencesand index' ISBN G-465{sZ4s-2 1. Mind and body . z. Neuropsychology. 3. PhilosoPhyof mind. I. Title. 8F161.834 1.992 72E'.2-dc2o Copyright O 1992 bY BasicBooks, A Division of HarperCollinsPublishers PRINTED IN THE UNITED STATES OF AMERICA

DesignedbY EllenLeoine gz 93 94 9s cc/Hc 9 s 7 6 5 4 3 Z r

9t-55454 CIP

To the memory of two intellectualpioneers, CharlesDarutin and Sigmund Freud. In much wisdom, much sadness,

For by earthr,peseeearth,by roaterutaler; by air brightair, and by fire brilliantfire

-Empedocles

And going 0n, rDecome to things like eoil, and beauty,and hope , 'Alhichendis nearerto God; if I may usea religiousmetaphor. Beautyand hope,ar fhefundamentallaws?I think that the right tuny,of course,is to say that what we haoeto lookat is theuthole structuralinterconnection of the thing; and that all the sciences, and notjust thesciences but all theeffortsof intellectualkinds,are an endeaoorto seethe connections of the hierarchies, to connect beautyto history,to connecthistory to man'spsychology,mnn's psychologyto the working of the brain, the brain to the neural impulse,the neuralimpulseto the chemistry,and soforth, up and down, both u)sys,And todayu)ecannot,and it is no usemaking belieoethat u)ecan,drautcarefullya lineall thetrra7from oneend of this thing to the other,because we haoeonly just begunto see that thereis this relatiaehierarchy. And I do not think eitherend is nearerto God, -Richard Feynman

Contents

LIST OF

ILLUSTRATIONS

XI

PREFACE

xiii

ACKNOWLEDGMENTS

xv

PART I PROBLEMS 1 Mind

3

2

9

3

Putting the Mind Back into Nature The Matter of the Mind

T6

PART II ORIGINS 4

33

5

Putting Psychologyon a Biological Basis Morphology and Mind: CompletingDannin's program 6 Topobiology,Lessons from the Embryo 7 The ProblemsReconsidered

42 52

6s

PART III PROPOSALS 8 9

The Sciencesof Recognition Neural Darainism

73 81 lx

CoNTENTS r0 II 12 13 14

Memory and Concepts:Building a Bridge to Consciousness Present The Remembered Consciottst4ess: l^anguageand Higher-order Consciot$ness Attention and the Unconscious Layersand LooPs:A Summary

99 111 r24 1,37 r47

PARTIV HARMONIES 15 16 17 1g 19 20

and lts Claims A Graoeyardof lsms:PhilosophV Memoryand thelndioidualSoul,AgainstSilly Reductionism Emotions Thoughts, Judgments, HigherProducts: self of the Mind: TheReintegrated Diseases ArtifactT to Constructa Conscious Is lt Possible origins of Mind ultimate the on Memory: and symmetry Epilogue Mind without Biology:A CriticalPostscript

SELECTED READINGS CREDITS INDEX

L57 r65 173 178 188 797 209 2'T.T

253 267 273

List of Illustrations

T-T. Ren6Descartes(I59G1650). 7-2. William James(1542-rgrO). 2-r. Galileo Galilei (Is64-I642). 2-2. Diagram of the visual system from Descartes'Traiti de I'Homme. j-T. The exposedsurfaceof the human cerebralcortex. 3-2. Some affangementsof the matter of the mind. 3-3. The developmentof the brain. 3-4. Mupping the eye and its visual fields to the brain. 3-5. The variability of neuralpattems. 4-1. ImmanuelKant (I7 24-1504). 4-2. Gestaltphenomena. 5-r. CharlesDarwin (1S09-7552). 5-2. Population thinking. 5-3. Changesin the frequency of genes may be related to the actual processof natural selection. 5-4. The modem synthesis. 5-5. An excerpt from Darwin's notebooks. s4. The remarkableincreasein cranialcapacityover two million yearsof human evolution. GL The reading of the geneticcode into protein. G2. Protein folding and function. G3. The early developmentof the chick embryo. G4. Cell adhesion. G5. An exampleof aberranttopobiology. 8-1. The immune systemworks as a selectiverecognitionsystem. XI

Ltsr oF IrrusrRATIoNs 8-2. The endlessregressionof homunculi' 9-r. A selectionaltheory of brain function. 9-2. Multiple maps of visual areasof the brain. 9-3. Neuronal grouPs. 9-4. Reentry. 9-s. A global maPPing. as a global 9-6. Darwin III, a recognition automaton that performs mapping. 10-1. Two views of memory. 10-2. Co*ical aPPendages. 11-1. A model of primary consciousness' tz-r. The supralaryngealtract in humans' 12-2. Areas of tn" brain serving speechproduction' 12-3. SemanticbootstraPPing. t2-4. A sciremefor higher-order consciousness' of new fZ-S. The evolution oiconsciourness dependson the evolution morphologY. 13-L. Sigmund Freud (Ls56-1939). ln-t, tivns of biological organization and loops of knowledge' system and diseasesof the mind' ,g-t. Oir"uses of th"l"*out 1.9-1..Jacquesde Vaucanson(170F1782)' z{daptive 1g-r.'A ii.tur" of NOMAD (Neurally Orgmized Multiply Device). 20-L. Some kinds of sYmmetry. 20-2. Some kinds of memory. P-L. Scalesof nature as establishedby physics' P-2. A Turing machine. P-3. Two computers in the real world' P-4. An algorithm for boiling an egg' P-5. Some aspectsof objectivism and functionalism' P-6. Categorization and polymorphous sets' P-7. A typical tree in generative grammar' P-8. An example of processesin a cognitive grammar'

Preface

I have written this book becauseI think its subject is the most important one imaginable.we are at the beginning of the neuroscientiftc."rrolrtion. At its end, we shall know how the mind works, what govems our nature, and how we know the world. Indeed, what is now going on in neurosciencemay be looked at as a prelude to the largesi possible scientific revolution, one with inevitable and important social "orrr"q,r"rr.ur. But this is not a scientificbook, at leastnot in the strict sense.It is a book about scienceand also about my own opinions. To write it I have had to put aside the habits of caution that are n".urr"ry to the working scientist in- order to explain some rather technicalmatters to nonspecialistreaders who occupy themselveselsewhere.so much the worse ftr the habitsparticularly if I can persuade those who are not in the businessto be interested,to lend it their support, and to sharein the excitementof being on the threshold of knowing how we know.

Acknowledgments

Responsibility for the views expressedhere is completely mine, and so is that for any residualfaults this book may possess.Had it not been for the responsesand criticisms of many people, the burden of fault would have been much larger. I expressmy gratifude to all who have been generous, knowing that I cannot list everyone who has been helpful. SusanHassler,Editor of The NeurosciencesInstitute, lent her expertise at all stages and her help was invaluable. Kathryn Crossin, who also contributed to the editing of my trilogy, made a number of important editorial comments on the present book. So did George N. Reeke,lr., my close collaborator, who has made major contributions to the design of recognition automata. Olaf Spoms, my former student and now my colleague,made many imaginative suggestionsand unstintingly lent his artistic talents to the design of many of the figures. His contributions were essential. I am particularly grateful to SusanBorden and Henry G. Walter, Jr.,who independently made felicitous suggestionsabout the organization of the book. W. Einar Gall, the ScientiftcDirector of the Institute, and Institute Fellows Giulio Tononi and ]oseph Gally made important critical remarks. Detlev Ploog did the same during his stay as a Visiting Fellow. This is the only one of my books that my wife, Maxine, read in its incipient stages.Her positive responseshelpedme shapesomeof the views expressedin the final parts. I thank her for them as well as for her encouragementin this and other matters of the mind. Finally, implicit in my title is an acknowledgmentto the Greekdiscovery xv

AcrNowt EDGMENTS a fragment of mind beginning in the sixth century r'c' It comes from philosomaterialist early an written by impedocles, physician, poet, and fit and size of pf,"t .f *i.d. His idea ihat p"r..ition results from the to material entities to particular pores in our bodies is more appropriate (where thought he modem theories of smell than of vision, but his heart his mind was) was in the right place.

PART

PROBLEMS

If we consider that without a mind no questions can be asked,and that there hasnever been a solidly establisheddemonstrationof a mind without a body, the importance of the subjectaddressedhere needsno defense.In this part of the book I want to introduce the reader to some classical thoughts about the mind. I also want to hint at what is attempted later: to describea biological theory of how we come to have minds. To do so I shall go into the organizationof the acfualmatter underlying our mindsneurons, their connections,and their pattems.

CHAPTER

1

Mind

Cogito,ergosum. -Ren6 Descartes

Thedefectof Descartes' Discourseon Metho d liesin his resoluof the realman, the tion to emptyhimse\ of himself, of Descartes, man of fleshand bone,theman who doesnot want to die,in order But the that he might bea merethinlcer-that is,an abstraction, real mqn refurnedand thrust himselfinto his philosophy.. . The truth is sum, ergo cogito-/ afrt, thereforeI think, althoughnot eoerythingthat is thinks,ls not conscious thinking abooeall consciousness of being?ls pure thoughtpossible,utithout of self, without personalityl consciousness -Miguel de Unamuno

on't think of an elephant." Of courseyou did, and so did I. But where is the elephant?In your mind, and certainly not in the room, at least for most people who are reading this book. Nof to think of it, you had to know what it was, to have rememberedit, and even,in somecases,to have entertainedan image of it. Above all, you had to know this language and how to understandthis bit of wordplay. Another piece of wordplay-"What is mindt No matter.What is matter? Never mind"-tells you the conclusionRen6 Descartes(figure 1-I)

PnoBLEMS

1-T FIGURE (15ga-1650), the one of Reni Descartes foundersof modernphilosophyand a great truthcrutician,Thedualismheapousedstillpolains modernthoughtaboutthemitrd' the relationship Cartesiandwlism is litcetyto be-dispelld only whenwe understand and physics' betwemmnsciousncss

came to in his thoughts about the subject. Those thoughts marked the beginning of modem philosophy and they split the mind away from scientificinquiry. To Descartes,the mind was a specialsubstance,one not located in space,not an extended thing the way matter was. This doctrine of dualism has plagued us-if not most of us, then at least many philosophers and some theologians-ver since. What does it mean to have a mind, to be aware, to be conscious?

Mind Everyonehas thought about this at one time or another,but until recently scientistsas scientistshave shied away.Now there is somethingnew on the sc€n€:neuroscience.We have begun to accumulatescientificknowledge about the brain at an explosiverate. It is becomingpossibleto talk in scientificterms about how we see,hear,and feel.The most complicated object in the universeis beginning to yield up its secrets. Why should we think that this will tell us anything about our minds? Becauseof what we have alreadylearned:just as we have recognizedhow matter comesto be in termsof particulararrangementsof things, we should That be able to figure out how minds ariseout of other sucharrangements. is what this book is about: connectingup what we know about our minds to what we are beginning to know about our brains. I will range over a variety of topics: nerves, computers,perception, language,selves.I will try to show how they connectboth to one another and to our being aware.Rather than talk about how we think or reason, I am going to discussthe basisfor thesehigh-level activities.My overall goal is to show that it is scientificallypossibleto understandthe mind. I will try to keep the technicaldetails to a minimum, at the sametime not hesitatingto take on shibbolethsand receivedideaswhen I believe they are in error. Thus, although parts of this book will be concernedwith pointing out positionsI believe are indefensible,I promisethat the main thrust will be positive and constmctive.After all, the subject,like obstetrics, is well nigh indispensableto our being here.It is at the centerof human concern. Let us begin. The word mind prompts thoughtsof abstrusephilosophicaldiscussions. But it also castsa familiar shadow imported from everyday use-"What's on your mind," "Never mind," "Minding the baby." It is not at all clear what is being referredto in theseexpressions.But we can still rely on some commonsensenotions to get started: 1". Things do not haveminds. 2. Normal humanshave minds;someanimalsact as if they do. 3. Beingswith mindscanrefer to other beingsor things;things without mindsdo not refer to beingsor other things. This last property, called intentionality by the German philosopher Franz Brentaho,servesas a good indicator of the existenceof a mental process.It refersto the notion that awarenessis always of something,that it always has an object.I will refer to intentionality often in what follows. But a set of indicatorsis not enough-we want to find out how the mind

Pnonrrvs matter that relatesto matter,particularlyto the specialorganizationof itselfas underliesit. It is not sulprisingthat peoplehavetreatedthe mind from different so a specialthing or rp".i"l form of stuff.After all, it seems "thai its possessormay find it difficult to concludeby ordi*ry -"It", of nonintentional iniorp".tion alonethat it couldarisefrom theinteractions is a process, mind mattei.But asWilliam |ames(ftgure!-2) pointedout, not a stuff.Modem scientiftcstudy indicatesthat extraordinaryProcesses from canarisefrom matter;indeed,matteritselfmaybe regardedasarising reconbeen matterhas In modemscience, of energyexchange. processes as a special reconceived been not has mind ieived in termsof-pro."rr"J,

FIGURET-2 William lames(1842-Iglo), oneof thefoundersof modernphysiologicalpsychologyand an erponentof the philosophyof pragttutism. His thoughtson cottsciousness-thstit is q processnot a substance;tlnt it is personaland reflectsintentionality----shapemuch of our modernoiew of tlw sublect.

Mind form of matter. That mind is a specialkind of processdependingon special arrangementsof matter is the fundamentalposition I will take in this book. If we look at the commonsenselist we startedwith, we seethat biological organisms(specificallyanimals)are the beings that seemto have minds. So it is natural to make the assumptionthat a particular kind of biological organizationgives rise to mental processes.Obviously, then, to pursuethe subject scientiftcallywe must tum to how the brain is organized.It would be a mistake,though, to ignore the rest of the body, becausethere is an intimate relation between animal functions (especiallymovement) and the development of the brain. SinceDarwin, biologists, when facedwith particular kinds of biological organization,have almost automatically consideredhow evolution might have given rise to them. Brain and mind are no exceptions.Therefore we will also want to know something about how the brain structuresunderlying the mind arose in evolutionary history. Above all, what we want to know is how such structureswork. This is where advancesin neurosciencecome to the fore. It is exciting to contemplate the possibility of relating these advancesto the accomplishmentsof psychologists studying behavior and mental processes.The ftndings of neuroscientistsindicate that mental processesarise from the workings of enormouslyintricatebrain systemsat many different levels of organization. How many? Well we don't really know, but I would include molecular levels, cellular levels, organismic levels (the whole creature),and transorganismiclevels (that is, communicationof one sort or another).Eachlevel can be split even further, but for now I will consider only these basic divisions. It is startling to realize how many connectionsproject from any one level to another-from a fear responseinduced by a waming cry to a biochemicalprocessthat affects future behavio4 from a viral infection to a changein brain development that altersmaturation;from a perception of a pattem to the chemistryof changesin a muscle;from any of theseat some critical time of developmentto how a human child developsa self-imagestrong or inadequate,detachedor dependent. To explain thesekinds of changes,I first have to clear up somemisconceptions. These have arisen mainly becauseexperts in various subdisciplines have remainedconfined within their own specialties.But this is not the only reason.Prejudice,the inability to carry out certain experiments, and the traps of language have all made it difficult to tease out the connectionsbetween mental events and events in the nervous system. There is more to studying the problem of mind than these matters of clariftcationindicate.As we will see,methodsof doing scienceon inanimate

Pnorrnvs obiects, while fundamental,are not adequateto doing scienceon animals that have brains and possessintentionality. This is becausescientific observers themselvesare intentional animals,locked into their own experiences of consciousness,who must ensure that their observationscan be communicatedto other observerseffectively, meaningfully, and without prejudice. This meansthey cannot include-indeed, they must deliberately exclude-elements in their own private experienceor awareness.We can say this in a flurry of rhymes and near rhymesrintersubjectivecommunication in sciencemust be objective, not projective. No wonder that magiq vitalism, and animism pervaded prescientificcommunication.The projection of individual wishes,beliefs,and desireswas not only allowed but was a major goal to be achieved in organizing societiesfor defenseagainst natural threats in a sensibleway. None of this means, however, that a scientific study of the mind is impossible.It does mean that such a study will be full of pitfalls, hidden portur"., and receivedideas,many drawn from scienceitself. Even the most intelligent researchersworking on the properties of mind have stumbled. And in studying intentionality, some persist in what can only be considered a p".ody oi ru.."rrful scienceslike physics,sciencesthat are dedicated to the study of objects that lack intentionality. How can we avoid falling into these traps? One way is to take the existing traps apart and ask whether modem neuroscientificresearchcan help uI dismaritle them. Let us tum to these tasks, particularly to the tasi
CHAPTER

2

Puttingthe Mind Backinto Nature

of Galileohasbeenremembered Theway in which thepersecution of themostintimatechange is a tributeto thequietcommencement in outlook,thich the human racehad y;rfL::K::r[d*ritehead

want to go back to the beginningsof modem scienceand considertwo towering figuresof the seventeenthcentury,Galileo Galilei (figure 2-T) and Ren6 Descartes.In Scienceand the Modern World, Alfred North Whitehead observed that in inventing mathematicalphysics, Galileo removed the mind from nature.By this figure of speech,I supposehe meant that Galileo insisted that the observer must be objective, that he must avoid the vexing disputes of Aristotelian philosophers over matters of causation.A scientist should instead make measurementsaccording to a model with no human projection or intention built into it and then search for correlative uniformities or laws that either support or disconfirm his or her claims. This procedurehasworked magnificentlyfor physicsand its companion sciences.Isaac Newton stands as the triumphant figure of its first full flowering. Even today after the Einsteinianrevolution and the emergence of quantum mechanics,the Galileanprocedurehas not been swept aside. Albert Einstein'stheory of relativity showed how the position and the velocity of the observeraltered the measurementof spaceand time, and by taking accelerationinto accountit alteredthe very meaningof the word

PnoBLEMS

2-1 FIGURE thefounderof andastronomer; mathematician, physicist, GalileoGalitei(156a-1642), In 1633hewasforced a truly modernphysicsand,somewouldsay,of modernscience. in hisDialogue riewsespoused by thi RomanCathoticChurchto recanttheheliocentric on the Two ChiefWorld Systems,andspenttherestof his life underhousearrest.

matter. Quantum mechanicsshowed that the operation of measurementin the domain of the very small ineluctably involves the actions of the observer who has to choose,within the uncertainty dictated by Planck's constant,the level of precisionwith which he or she wishesto know either the position or the momentum of a subatomicparticle. This reflectswhat physicists call the Heisenberguncertainty principle. Even with the startling revelationsthat at velocities approachingthat of light or at very small distancesthe observer is embedded in his or her 10

Putting the Mind Back into Nature measurements,the goal of physics remainsGalilean:to describelaws that are invariant. We have no reason to abandon this goal. This is because Einsteinian and Heisenbergianobservers,while embedded in their own measurements,are still psychologically transparent.Their consciousness and motives, despiteoccasionalargumentsabout their importanceto quantum measurementsby philosophersof physics,do not haoeto be taken into accountto practicephysics.The mind remainswell removed from nature. But as Whitehead duly noted, the mind was put back into nature with the rise of physiology and physiological psychology in the latter part of the nineteenthcentury. We have had an embarrassingtime knowing what to do with it ever since.just as there is something specialabout relativity and quantum mechanics,there is something special about the problems raised by these physiological developments. Are observers themselves "things," like the rest of the objectsin their world? How do we accountfor the curious ability of observers(indeed,their compelledneed) to carve up their world into categoriesof things-to refer to things of the world when things themselvescan never so refer?When we ourselvesobserveobservers, this property of intentionality is unavoidable. Keeping in line with physics, should we declarean embargo on all the thought, psychologicaltraits we talk about in everyday life: consciousness, sanitary regimes of behavwe adopt the elaborate beliefs,desires?Should iorism? Should :unorous partners say to each other: "That was good for you; was it good for me too?" The ludicrousnessof this last resort becomes evident when we considerthe denial it entails.Either we deny the existence 'tecome scientists" (for example, our of what we experiencebefore we or we declarethat science(read"physical science")cannot own awareness), deal with such matters. It is here that the secondgreat figure of the scientificrevolution of the seventeenth century, Descartes,comes to the fore. In his search for a method of thought, he was led to declare for "substancedualism." As I mentioned earlier,accordingto this view the world consistedof resertensa (extendedthings) and rescogitans(thinkng things). Galileanmanipulations work on resertensa,the set of extended things. But rescogitans,the set of thinking things, does not exist properly in time and space;lacking location, not being an extended thing, it cannot fall into the purview of an extemal observer. Worse still is the problem of interactionism:the mind and the body must communicate.With an uncharacteristiclack of clarity, Descartes declaredthat the pineal gland (figure 2-2) was the placewhere interactions between res cogitansand res extensaoccurred. Dualism has persistedin various forms to the presentday. For example, while apparentlymonistic,behaviorismis simply dualismreducedby denial 11

PnoBLEMS

FIGURE2_2 y-rop9s9d Diagram of the uisualsystemfrom Descartes'Trait6de l',Homme.Descartes thai the reiinal imagefiom eacheyeroqsproiectedby neroefibersonto the walls of the that binocularuisionwas thenproiected fluid-filtedoentricleiof the brain.It wasassumed 'to (btack proposed this unpairedstructureas the site arrow). Descartes the pineal body with body.We now know that extensa----soul with res interacted at which."r .ogitunt real binocular{rojectionis to the oisualcorterat the backof the brain on both the right and left sides.

of the mind as a scientificobject, and therefore left with one end hanging. Behavioristssolve the dilemmaby examiningbehavior and ignoring intentionality. They do not attempt to put the mind back into nature; they simply deny its validity as a scientificobject. And many nonbehavioristic psychologists,while assertingthat they are materialistsand not substance dualists,are nonethelessproperty dualists.While concedingthat the mind and the brain arosefrom a single substance,they insist that psychological properties must be dealt with exclusively in their own terms,whicl necessarily differ from those used for the physical objects or bodies giving rise to theseproperties.A good exampleof a property dualist is SigmundFreud in his later years. I should point out that even remarkably accomplishedbiologists have been dubious about the enteqpriseof studying the mind. I once discussed theseissuesin a public symposiumwith the distinguishedimmunologist Sir Peter Medawar. He was just the slightest bit dismissive: "Of what use is it?" he asked. I managed to fend him off by pointing out that if we understoodthe brain better we could at leastdisposeof somecrazy notions about how it works. Peter was the enemy of cant and my resPonse quieted him. T2

Putting the Mind Back into Nature I wish I could have said to him what Michael Faraday,I have heard,told the Chancellor of the Exchequer,William Gladstone,after presenting his findings on electricity. When the gentlemaninquired loftily, "Of what use is it?" Faradayreplied,"Sir, somedayyou shall tax it." (On another occasion he said, "Of what use is a newbom baby?") In trying to put the mind back into nature, can we do better than substancedualism or property dualism?Or will we fall into further errors in the attempt?My answerto both questionsis a qualifiedone. We can do better, but we cannot do it by assuming,as have some modem students of cognition, that the structureand biology of the brain are incidental and not centralto the enteqprise.Let us explore this issuefurther, for it has rich implications. In the last few decades,practitioners in the field of cognitive science have made seriousand extensive attempts to transcendthe limitations of behaviorism. Cognitive scienceis an interdisciplinary effort drawing on psychology, computer scienceand artificial intelligence,aspectsof neurobiology and linguistics,and philosophy. Emboldenedby an apparentconvergence of interests,some scientistsin thesefields have chosennot to reject mental functions out of hand as the behaviorists did. Instead, they have relied on the conceptof mental representationsand on a set of assumptions collectively called the functionalist position. From this viewpoint, people behave according to knowledge made up of symbolic mental representations. Cognition consistsof the manipulationof thesesymbols.Psychological phenomenaare describedin terms of functional processes.The efficacy of suchprocessesresidesin the possibility of interpreting items as symbols in an abstractand well-definedway, accordingto a set of unequivocalrules. Suc-ha set of rules constitutes what is known as a syntax. The exerciseof thesesyntacticalrules is a form of computation.(At this point, pleasebear with me on what computation is exactly. For now, let us take it to mean the manipulation of symbols according to a definite procedure.I discussthis in detail in the Postscript.)Computation is assumed to be largely independentof the structure and the mode of development of the nervous system, just as a piece of computer software can run on different machineswith different architecturesand is thus "independent"of them. A related idea is the notion that the brain (or more corectly, the mind) is like a computer and the world is like a pieceof computer tape, and that for the most part the world is so ordered that signalsreceivedcan be "read" in terms of logical thought. Such well-defined functional processes,it is said, constitute semantic representations,by which it is meant that they unequivocally specifuwhat their symbols represent in the world. In its strongest form, this view 13

P n o nr r u s proposes that the substrateof all mental activity is in fact a language of inougttt-" language that has been called "mentalese" (see the Critical Postscript). This point of view---
P u t t i n g t h e M i n d B a c k i n t o N at u r e interestedreadersmight more profitably turn to it after finishing the text of the book. This essayaddresseswhat I believe to be a seriesof category mistakes. The first is the proposalthat the solution to problemsof consciousness will resolution come from the of some dilemmasin physics.The secondis the suggestionthat computation and artificial intelligencewill yield the answers.Third, and most egregious,is the notion that the whole enteqprise can proceed by studying behavior, mental perforrnanceand competence, and languageunder the assumptionsof functionalismwithout first understanding the underlying biology. I will addressthe criticalargumentsin the Postscript.In the chapterthat follows, I review someof the factsand ideasof biology and neuroscience. It is vital to understand the actual matter underlying the mind, and in particularits principlesof organization.Only with suchunderstandingwill it be possibleto dissectthe difficultieswe face when we attempt to study the mind, and to propose some ways out of the predicamentsI have mentioned. The principle I will follow is this: There must be ways to put the mind backinto naturethat are concordantwith how it got therein the first place. Theseways must heedwhat we have learnedfrom the theory of evolution. In the courseof evolution,bodiescameto have minds.But it is not enough to say that the mind is embodied;one must say how. To do that we have to take a look at the brain and the neryous systemand at the structuraland functional problems they present.

15

CHAPTER

3

The Matterof the Mind

Theonly laws of matterarethosewhichour mindsmustfabricate, and the only laws of mind are fabricatedfor it by matter, -James ClerkMaxwell

iven the unique characterof consciousnessand the inability of thought to "see into" its own mechanisms,it is no suqprisethat somephilosophershave proposedthe idea of a thinking substance, or even a kind of panpsychismin which all matter sharesin conThe resultsof modem investigationssuggest,however, that the sciousness. physical matter underlying the mind is not at all special.It is quite ordinary-that is, it is made up of the chemical elements carbo& hydrogen, oxygen, nitrogen, sulphur,and phosphorus,along with a few trace metals. So there is nothing in the brain s essentialcomposition that can give us a clue to the nature of mental properties. what is specialis how it is organized.Those ordinary chemicalelements form parts of extraordinarily intricate molecules,which in tum make up .o^pi"* structuresin the cells of living tissues.In a complex organismlike a human being, the cells come in about 200 different basic types. One of the most specializedand exotic of these is the nerve cell, or neuron. The neuron is unusual in three respects:its varied shape, its electrical and chemicalfunction and its connectivity, that is, how it links up with other neurons in networks. I plan to tell you more about some of these properties,but only just T6

The Matter of the Mind enough to convince you that we are dealing with something unlike anything else in the universe.I shall add to this description as neededand in this way we won't be loaded down with intricaciesright from the start. But before I lay out some descriptive details, it will be useful to give you a feeling for the numbers of neurons in certain brain areas and for the numbers of connectionsthey make with each other. This will be, I think a startling enough beginning. But when I get down to describinga bit of the molphology, you may be even more impressedwith what evolution has accomplishedin selecting for animalswith richly structured brains. Let us begin with the part of the brain called the cerebralcortex (figure 3-l), a structure that is central to what are loosely called the higher brain functions-speech, thought, complex movement pattems, music. If one were to take this comrgated "mantle" that covers the dome and the sides of your brain and spreadit out, it would be the sizeof a large table napkin and about as thick. Counts of the nerve cells making up this structure are not very accurate,but it appearsthat there are about ten billion neurons in the cortex. (There are also other cells called glia that have supporting functions, but I will ignore them.) Eachnerve cell receivesconnectionsfrom other nerve cellsat sitescalled synapses.But here is an astonishing fact-there are about one million billion connectionsin the cortical sheet.If you were to count them, one connection (or synapse) per second, you would finish counting some thirty-two million years after you began.Another way of getting a feeling for the numbersof connectionsin this extraordinary structureis to consider that a large match head's worth of your brain contains about a billion connections.Notice that I only mentioned counting connections.If we considerhow connectionsmight be variously combined,the numberwould be hyperastronomical---onthe order of ten followed by millions of zeros. (There are about ten followed by eighty zeros'worth of positively charged particles in the whole known universe!) So here we have our first clue as to what makesthe brain so specialthat we could reasonablyexpect it to give rise to mental properties.And while the sheer number and density of neuronal networks in the brain are amazing,theseare not the only unique properties of brain tissue.An even more remarkable property is the way in which brain cells are arranged in functioning pattems. When this exquisite arrangementof cells (their microanatomy,or morphology) is taken together with the number of cells in an object the size of your brain, and when one considers the chemical reactions going on inside, one is talking about the most complicated material object in the known universe. I want to say a bit more about some properties of the brain's other T7

PnoBLEMS

S.s :\

=N \\

N

-+_\ -- .*\ \\.. ll---

3_1 FIGURE surfaceof thehumnncerebrslcorter,drawnby thegrmt anatomistAndreas Theerposed anatomy, to bethefounduof,modern considered (1514-156i).He isgenerally vesalir,s art. medical stsndard a new set for and hisDe FabricaHumanii Corpora components.In complex animalssuch as human beings, the brain consists of sheets,or laminae,and of more or lessrounded strucfurescallednuclei. Each of these structureshas evolved to carryrout functions in a complex network of connections,and each consistsof very large numbers of neurons, sometimesmore and sometimesless than in the cortex. The brain is connected to the world outside by means of specializedneurons called sensorytransducersthat makeup the senseorgansand provide input to the brain. The brain's output is by meansof neuronsconnectedto musclesand glands. In additioo parts of the brain (indeed, the major portion of its iissues)receive input only from other parts of the brain, and they give outputs to other parts without intervention from the outside world. The 18

The Matter of the Mind brain might be said to be in touch more with itself than with anything else. How do neurons connect with each other and how are they arranged within nuclei and laminae?As mentioned, the major meansof connection is the synapse,a specializedstructure in which electrical activity passed down the axon of the presynapticneuron (figure 3-2) leads to the release of a chemical (called a neurotransmitter) that in tum induces electrical activity in the postsynaptic neuron. As is suggested in the figure, the strength or efficacy of synapses can be changed-presynaptically by changesin the amount and the delivery of transmitter,and postsynaptically by the alteration of the chemical state of receptors and ion channels,the units on the postsynaptic side that bind transmittersand let ions carrying electrical charge (such as calcium ions) through to the inside of the cell. Neurons come in a variety of shapes,and the shapedeterminesin part how a neuron links up with others to form the neuroanatomy of a given brain area.Neurons can be anatomically arrangedin many ways and are sometimes disposed into maps. Mapping is an important principle in complex brains.Maps relate points on the two-dimensionalreceptor sheets of the body (such as the skin or the retina of the eye) to corresponding points on the sheetsmaking up the brain. Receptor sheets(for example, touch cellson your fingertips and retinal cellsthat respondto light) are able to react to the three-dimensionalworld and provide the brain with spatial signals about pressure or wavelength differences(they react to a fourdimensionalworld if we consider time as well). Furtherrnore,maps of the brain connect with eachother via fibers that are the most numerousof all those in the brain. For example,the corpus callosum,the main fiber bundle connectingparts of your right brain to parts of your left acrossthe midline contains about 200 million fibers. None of this was known in any detail before the nineteenthcenfury. But remarkable surmiseswere made before that time by remarkablemenDenis Diderot, for example.The following is an excerpt from his novel in the form of a play, Le RAoede d'Alembert,in which d'Alembert's mistress, Mademoisellede l'Espinasse,queriesthe physician,Dr. Bordeu,about the causesof d'Alembert's disfurbed dreams. Bonpru: Becauseit is a very different thing to have somethingwrong with the nerve-centrefrom having it just in one of the nerves.The headcan commandthe feet,but not the feet the head.The centrecan commandone of the threads,but not the thread the centre. MepruorsrrLEor r'EsprNAssE: And what is the difference,please?Why don't I think everywhere?It's a question I should have thought of earlier. T9

PnoBLEMS

DENDRITE

Synapses

o o I

o o oo o Vesicles with Neurotransmitter

Receptorfor Neurotransmitter in the Postsynaptic Neuron

AXON

Synapse

FIGURE3_2

'\rons neutons. simplified of themattu of themindaseremplifiedby arrangements Some either a nanron with contact mske that ertensiotrs long are distant'neurons from neari 'on tkons carryelectrical calleddendrites. itsbody(soma)or on its branchingprocesses, the actioityreaches of neuihansmitterwhentheelectrical actiaitythoi ,o*tt therelease neuro' the reryPtor1, ty*pit with anotherneuron.After interactingwith theappropriate A neuronto fire electrically. iansmitter in turn tiggers thi recipient(or postsynaptic) at the in the circle enlarged and this of the bottom at enclrcled is simplifiedsynapse figure is storedin the in whiichneurotransmitter oesicles represent rigit.'fheimall roundsacs (theY'shapedstructures) receptoys neuronptojects fleurofl.Thepostsynaptic piesynaptic bind Thesereceptors into"thi ct4t betwemthi preiynapticandpoitsynapticmembranes. of the respotrses the trigger nnd aesicles presynaptic ihe released to thetrattsmitter frim neuron' postsynaptic

The Matter of the Mind Bononu: Becausethere is only one centreof consciousness. MapruorsulE DEL'EsprNessn: That's very easy to say. Bonpw: It can only be at one place,at the common centre of all the sensations,where memory residesand comparisonsare made.Each individual thread is only capableof registeringa certain number of impressions,that is to say sensationsone after the other, isolatedand not remembered.But the centre is sensitiveto all of them; it is the register,it keepsthem in mind or holds a sustainedimpression,and any animalis bound,from its embryonicstage,to relateitself to this cenhe,attachits whole life to it, exist in it. MaprMotsslrEDEL'EsprNAssE: Supposingmy finger could remember. Bonpru: Then your finger would be capableof thought. Mepruolsnrr or r'EsprN,lssr: Well, what exactly is memory? Bonpnu: The property of the centre,the specificsenseof the centre of the network as sight is the property of the eye, and it is no more surprisingthat memory is not in the eye than that sight is not in the ear. Mepruotsntr ps L'EspINessr: Doctor, you are dodging my questions insteadof answeringthem. Borpru: No, I'm not dodging anything. I'm telling you what I know, and I would be ableto tell you more about it if I knew asmuch about the organizationof the centreof the network asI do about the threads, and if I had found it as easyto observe.But if I am not very strong on specificdetailsI am good on generalmanifestations. Meonuorsnm pn r'EsprNessr: And what might thesebe? Bonoru: Reason,judgement,imagination,madness,imbecility,ferocity, instinct. . . . Bonpru: And then there is force of habit which can get the better of people,suchas the old man who still runs after womery or Voltaire still tuming out tragedies. (Herethe doctor fell into a reverie,and MeoruorsELLE DEr'Esprwessr said:)Doctor, you are dreaming. Bonpnu: Yes I was. Mepsbrorcsnrpr r'EsprNnssr: What about? Bonpru: Voltaire. MeorrraolsilLEos r',EspNessr: What about him? Bonpru: I was thinking of the way great men are made. Even with the discussionso far, we are already in possessionof a number of factsabout the matter of the mind. Nerve cellsare specialized,numerous, and hyperdensein their connections,which themselveshave specialchemical and morphological characteristics.The anatomy resulting from these arrangementsis staggering in its intricary and diversity. But it also has generalorganizing principles:It is madeup of sheetsthat have topographic 27

Pnourrus maps and of rounded nuclei, or "blobs." It sendsmultiple fibers to connect the maps to sensorysheetsand out to the musclesof the body. And maps map to each other. is a.ur,rlt of the stimulation of sensory elements,nerve signalsin the form of electricaldischargesoccur at the membranesof neurons.They are causedby the flow of charged ions. (This means that electrical charges move more slowly in cell membranesthan they do in telephone wires, where the current is carriedby electrons.)Massive numbersof neuronsact in parallel in amazingnumbersof combinations.Their sensitivity to stimulation can be altered by a host of different chemicals,including the neurotransmittersat synapses,other substancescalled neuromodulators,and, of course,by drugs. A piecl of brain tissueis an intricate network that respondsto electrical and chemicalsignals in three-dimensionalspaceand in time' It sendsout dynamic pattems and receivesand respondsto such pattems. These pattems affect eachother and, through other nerve connections,the action of other organs of the body-the heart, kidneys, lungs, muscles,and glands. The brain is a master controller and its rhythmic pattems alter how you breathe, pump blood, digest your food, and move. I discussin a later chapter the principles by which the nervous system (and indeed the whole animal)developsboth its overall and its microscopic shape. But it is useful to anticipate a bit of that discussion here. The anaiomicalarrangementsof the brain and the nervous system are brought about by a seriesof developmentalevents (figure 3-3). In the embryo, cells divide, migrate, die, stick to each other, send out processes,and form synapses(and retract them). This dynamic seriesof events dependsquite sensitively on place (which other cells are around), time (when one event occurs in relation to another), and correlated activity (whether cells fire together or change together chemically over a period of time). P1""" depundenciesin development are quite striking. Nowhere is this seenas clearly as in the formation of mapsduring embryonic development, suchas the map of visual spaceformed by the retinotectalprojection (figure 3-4). In this instance,neurites (fibers)from the ganglion cells of the retina form the optic nerve, the neurites of which then map in a definite fashion to a region that, in an animal like the frog, is called the optic tectum. Stimulation of a particular point on the retina by a point of light leads to the stimulation of ^"uronr in a particular region of the tectum, and the responding cells are arranged in a definite map. Place is critical to the workings of such a maP. The arrangementof this maP is achievedin at least two stepsduring development.The first step, which involves the extensionof neuritesin 22

The Matter of the Mind

Brtil)m+ffi,

FIGURE3_3 of the brain,from theneuralgrooue(top left) to thecerebralcortex(top Thedeoelopment right). Indiuidual neuronsoccur in layers (bottom left) or mo:rein paths (bottom center),finally interactingin synapsesto form an enormouslycomplexneuroanatomy (bottom right). At oneor anothertime in their cnreers all neuronsaregypsies-mouing to theirfinal positions0n othercells,The resultis the most complicatedmaterialobiect in the known unioerse,

overlappingarborsby optic nerve fibers,forms a coarsemap and doesnot require neural activity. The secondstep, in which the map is refined and becomesmuch more precise,requiresneural activity in neighboring ganglion cell fibers that is correlatedwith neural activity in the tectum. Mup formation during developmentin animalslike the goldfish or the frog is dynamic, and the connectionsshift and reassembleas differentialgrowth occursboth in the retina and the tectum. The principlesgoverning these changesare epigenetic-meaning that k"y events occur only if certain previous events have taken place.An important consequenceis that the connectionsamong the cellsare thereforenot preciselyprespecifiedin the genesof the animal. What makesmapsso interestingis that the epigeneticeventsthat create form from place early in embryonic development must to some extent "anticipate"future interactionsof the two-dimensionalsurfacesof sensory 23

Right Eye

Left

Left Eye

temporal

FIGURE3_4 in the Mapping the eyeand its oisualfieldsto the brain. Top: A frog with an electrode oisualpZ* of *s brain katled the tectum).Bottom: Front oiew. !\ rarts.of the retina of the'right-eyemarkedwith letterswereeachilluminatedwith light at the sametime in theftog's tectum.Conespondingneighti.t onittt*iAt measuredelectricalresponses map of respoflsesthat are marked by the same a tectum the left in boring regions form lettei, iote thqt there'isa rotsiion of the map but that neighborsremainthe samein theeyeand in the tectumon theoppositeside.'iemporalregionsof theeyemap to medial regiins of the tectum, inferior regions^op to rostral regiotrs,and n forth'

24

The Matter of the Mind receptor sheets(for example,the retina or skin) with the three-dimensional world in which the animal moves and receivesstimuli. we will see later how the principles of development and evolution account for these phenomena. what I have describedso far may sound like the organization of a vast telephone exclange or perhaps even that of a digital computer. In some ways, the brain does indeed behave like these systems.When we look in detail at the structural feafures and functional properties of the neryous system, however, the analogy breaks down, and we are confronted with a seriesof problems.The implications of theseproblemsamount to a series of interpretive crises for neuroscienceas well as for those sciencesthat depend on its conclusions. The structural crises, which I described in detail in my book Neurnl Darwinism,are those of anatomy and development.Although the brain at one scalelooks like a vast electricalnetworh at its most microscopicscale it is not connected or arranged like any other natural or man-madenetwork. As we have just seen,the network of the brain is createdby cellular movement during development and by the extension and connection of increasingnumbersof neurons.The brain is an exampleof a self-organizing system.An examination of this system during its development and of its most microscopic ramifications after development indicates that precise point-to-point wiring (like that in an electronic device) cannot occur. The variation is too great. Furthermore, although the connectivity of neuronal systems in the centralnervous system (particularly those that are mapped)is more or less similar from individual to individual, it is not identical.Indeed,asfigure 3-5 shows, there is considerablevariation both in the shapesof individual neurons in a classand in their connection pattems. This is not surprising, given the stochastic(or statistically varying) nature of the developmental &iving forces provided by cellular processessuch as cell division, movement, and death; in some regions of the developing nervous system up to 70 percent of the neurons die before the structure of that region is completed! In general,therefore,uniquely specifiedconnectionscannot exist. If one were to number the branches of one neuron and to number in a corresponding manner the neurons it touched, the numbers would not correspond exactly in any two individuals of a species-not even in identical twins or in genetically identical animals. To make matters even more complicated, neurons generally send branchesof their axons out in diverging arbors that auerlapwith those of other neurons, and the same is true of processescalled dendrites on recipient neurons (see figure 3-Z). I gave an example of this when I 25

PnoBLEMS

RAL VARIABILITY STRUCTU

ffi

,t# II

t#

#

,'+ A ,v+ +

$

DYNAMICVARIABILITY

!

Fi

F I N G 3 E R

s

ii

G E R S

2

FACE

FACE

FIGURE3-5

in.fourdifnmt of thenme nerc.e Top left: Patterns Theoariabilityof neuralpattems. Visual-neurons Topcenter: from Goodman). and Corey (tht Pearson Keir work'of Iorurt, and Mncagno (oisualized Flwarlo. by waie, genetically identical but differmt fbot four 'ii, one ot' sides left and ight the on neurons toiitos*d, Nite that ioen corresponding brain rabbit structures neroe Ripeating from I i,naioiiufiari not themme.Top right, -A andoaiability of brain , iiit-nt"a side)areaII diffieni. Bottom: Thedynamism 'r**7o, in brain) location A normalmap(anowindicstes tuuchin anadultowl monkey, the backor of tegions areas)andcorresponding [ii n*ot}t andpalm(lisht numbered "i the cuttins After middle. in the i,n tle.nay shown are t;";; li;;-"i ii; tLia Urif areas) palm, and the of side the part of fingers neie that ientes front lluirless) .the -a.reanangenotonlyof thefront andtheback,butalsoof otherfingets i^i ott"rt in thi map'bordus, his colleagues,) tiip ot nShA.Ehis'is the work of MichaelMerzenichand 26

The Matter of the Mind discussedthe arborization of optic nerve fibers on the tecfum. To put it figuratively, if we "asked" a neuron which input came from which other neuron contributing to the overlapping set of its dendritic connections,it could not "know." The existenceof developmentalprinciplesleading to variancein connections and to overlapping arbors with unidentifiable (and not necessarily repeatable)pattems of synapsescreatesa crisis for those who believe that the nervous system is precise and "hardwired" like a computer. We may ask, "How has this crisis been met, when it has been recognizedat all, by those who believe in the idea of the brain as a computer? First, these explanationsdismissvariations below a certain microscopic level as "noise," a necessaryconsequenceof the developmentaldilemma. Second,they deal with the absenceof uniquely specifiedconnectionsby arguing that higher levels of organization suchas maps either do not need such connectionsor compensatefor their absencein some fashion. And third, they explain the absenceof precisely identified synaptic inputs by assumingthat neurons use a code similar to those used to identify phone credit card or computer users.In neurons,the placeand time codespresumably relate to the frequency,spacing,or type of neuronalelectricalactivity, or to the kinds of chemical transmitters with which they are associated (figure 3-2). Notice, however, that theseexplanationsassurnethat individual neurons carry information,just as someelectronicdevicescarryrinformation. I argue later that this is not a defensibleassumptionand that these explanationsare inadequate.No convincing evidencefor the kinds of codes that humans use in telegraphy, computing, or other forms of human communication has been found in the human nervous system. This brings us to somedeeperriddles for those who would propose that the brain is a kind of computer.Theseriddles constitute a set of functional crisespertaining to physiology and to psychology. The first is this: If one exploresthe microscopicnetwork of synapseswith electrodesto detect the results of electricalfiring, the majority of synapsesare not expressed,that is, they show no detectablefiring activity. They are what have been called "silent synapses."But why are they silent,and how does their silencerelate to the signals,codes,or messagesthat they are supposedto be carrying? A second dilemma concems the functions and interactions of maps of the kind we have already consideredfor the retinotectal system. Despite the conventional wisdom of anatomy books, thesemaps are not ftxed; in some brain areas,there are major fluctuationsin the borders of maps over time. Moreover, maps in each different individual appear to be unique. Most strikingly, the variability of maps in adult animals dependson the availablesignal input (figure 3-5). This might not seemto pose a dilemma 27

P no g rrl"t s on the alteration at first; after all, computerschangetheir "maps" or tables are based on system of software. But thefunctioningmaps of the nervous in the anatomicalmaps-and at this anatomical level, they are changed are maPs neural functioning adult brain only by the death of neurons.If the two gives changing u, u i"r,.,lt of "software" changes,what is the code that or differlni individuals with variant anatomical maps the same output systems altemative are result?One standardexplanationis to say that there hardin the brain that handle changing input, each altemative fixed and wired but switched in or out by changing input. The facts show, howevet that the variance of neural maps is not discrete or two-valued but rather continuous, fine-grained,and extensive.Thus, the number of altematives would have to be very large. Another set of observationsbrings us to psychologicaldilemmasof the most profound kind. They cast doubt on the idea that the complex behav"leaming." ior of animalswith complex brains can be explained solely by Indeed, this crisis highlights the fundamental problem of neuroscience: "obHow can an animal i*tiatv confront a small number of "events" or indefian jects,,and after this exposureadaptively categorizeor recognize nite number of novel obiects(evenin a variety of contexts)asbeing similar or identical to the small set that it first encountered?How can an animal, in the absenceof a teacher,recognize an object at all? How can it then generalizeand "construct a universal" in the absenceof that object or even in it, pr"runce? This kind of generalization occurs without language in animals such as pigeons, as I discusslater on. Explanationr oflhur" challenging problems tend either to rely on the existe-nceof hidden cues,not obvious to the experimenter,or to treat the world of the responding organism as if its "obiects" or "events" camewith labelson them. but in reality, the world, with its "objects," is an unlabeled place;the number of ways in which macroscopicboundariesin an animal's can be partitioned by that animal into objects is very large, "nrrironm"nt if not infinite. Any asiignment of boundariesmadeby an animal is relative, not absolute,and dependson its adaptive or intended needs' What is striking ls that the ability to partition "objects" and their arrangementsdependson the functioning of the maps that we discussed earliei. But how Jo *"pr interact to give definition of objectsand clear-cut action or behavior?In human beings,a considerationof this question leads to a perceiver to what I call the homunculuscrisis:the unitary aPPearance of perceptual processesthat are known to be based on multiple and cornple*'parallel subprocessesand on many maPs.(ln the visual system, there muy be more than thirty interconnectedbrain centers,each with its Who or what organizesa unitary picture?"Computations" or o*n."i.) 28

The Matter of the Mind "algorithms" in the brain-or the homunculus,a little man who has in his headyet anotherhomunculus(seefigure S-2), and so on ad inftnitum?Who is at home?If it is the homunculus,how could he have been constructed during the developmentalwiring of the brain by his cousin,whom we may call the electrician?we have already seenthat if he exists during development, this electricianhas constructed some very odd wiring indeed. Where does this leave us? The short answer is, "with a very great challenge." Unless we wish to pursue brain sciencein a purely empirical fashionwithout concemfor coherentexplanations,we have to confront the crisesI have discussedhere. An obvious altemative is to have a scientific theory that reconcilesthe apparent contradictions and dilemmas and resolves the crises.Clearly any satisfactorydevelopmentaltheory of higher brain function must remove the need for homunculi and electriciansat any level. At the sametime, the theory must account for obiect definition and generalizationmade on a world whose events and "objects" are not prelabeled by any a piori schemeor top-down order. This soundsless like the tasksto which computersare put and more like something utterly unusual and very different from computers. l{hat is specialabout brains that computers,and material particles,and atoms, and rescogitans,and ghosts all lack is evolutionary moqphology. As we have seerythis morphology interactsat many levels,from atoms up to muscles.The intricacy and numerosity of brain connectionsare extraordinary. The maps that "speak"back and forth are massivelyparalleland have statisticalas well as precisefeatures.Furthermore,the matter of the mind interactswith itself at all times. I have not yet mentioned that the dynamic arrangementsof the brain show the system property of memory: previous changesalter successivechangesin speciftedand special ways. Nervous system behavior is to some extent self-generatedin loops; brain activity leads to movement, which leads to further sensationand perception and still further movement. The layers and the loops between them are the most intricate of any object we know, and they are dynamic; they continually change. Indeed, the chemicaland electricaldynamics of the brain resemblethe sound and light pattems and the movement and growth pattems of a jungle more than they do the activities of an electric company. These dynamics result from a specialchemistry.Alterations of that chemistry or destructionof its anatomicalsubstratecan lead to temporary or permanent mental changesfrom elation to unconsciousnessto death. While we recognize that the marvelous matter underlying the mind is like no other, we must beware of a shallow chauvinism.Such a position would assertthat only thosebiochemicalsof which the brain is made could 29

Pnos rnI\ts leadto sucha structure.For evenif that were to someextentthe case,it to creatementalProcesses' of thesesubstances is the dynamicarrangement is It not their actualcomfosition,that is essential. dynamicmorphologyall the way down. But not so far down that we haveto invokevery special physlcaleventsand forcessuchas thosebetweenfundamentalphysical ignorantof brainmorphologyandthe properties p"iti.l"r. Somescientists, of ."*ory, havebeentemptedto explainmentalpropertiesat this level, the quantumlevel (seethe Postscript). chauvinismis out, however,so is the liberalismof tf strict biochemical a brainsoftwarethat actuallydoesnot the computerscientistwho assumes exista priori ^nd then claimsthat it doesn'tmatter what structurethis errors,for thereis no such softwareruns on. He makestwo fundamental thing as softwareinvolved in the operationsof brains,and the evidence overwhelminglyindicatesthat the morphologyof the brainmattersoverwhelmingly. factsin hand,we With ihis backgroundandsomeof the neuroscientific is important to It matters. may now tum to more generalbiological examinethem if we are to avoid the pitfallsin our path toward a better of the matterof the mind. understanding

30

PART II

ORIGINS

One of the temptationsof having a mind is to try using it alone to solve the mystery of its own nature. Philosophershave attempted this since time immemorial.Psychologistsfall back on it, as do we all from time to time. But as a generalmethod to explore the matter of the mind, it just won't do. We havebeen in possessionof an enorrnousinsight into how our minds might work ever sinceDarwin proposedthat minds aroseby evolution. What this meansis that minds havenot alwaysbeenaround;they appeared at somedefinitetime in a seriesof gradedsteps.It alsomeansthat we have to pay attention to animal form, becauseevolution teachesus that the selectionof animalsformed to carry out functionsthat increasetheir fitness is at the very heart of the matter. At "the brain of the matter" is the most complicatedarrangementin the known universe.To understandit will takeus from philosophy to embryolary leap. When we have taken it, we will be ogy, in a curious but necess in a position to return to philosophy via biology in the next two parts.

3I

CHAPTER

PuttingPsycholo gy on a Biologic aI Basis

urasto him n neu)study,and a dark colnerof educaPsychology tion, . . He put psychollgyunderlockand key;heinsistedon maintaining his absolutestandards;ot4aiming at ultimate lJnity. The maniafor handlingall sidesof eoeryquestion,lookinginto etrery windout,and openingeoerydoor, u)as,as Bluebeardiudiciously in pointed out to his wiaes,fatal to their practical usefulness -Henry Adams society,

gnoring the origins of things is always a risky matter. It is even more risky in any effort that purports to explain mental events.But this is exactly what has happenedin much of the history of psychology and the philosophyof mind. I guessthis is so becausethought is a reflexive and a recursiveprocess.It is thereforetempting to think that the nature of thinking can be uncoveredby thinking alone. But if we go back to the earlier chapter on mind, we notice that the biggest differencebetween intentionalobjectsand nonintentionalobjectsis that the former are biological entities.The point is not that all living things are intentional,just that no nonliving things are. As I mentioned in the last chapter, we must accountfor how embodiment occursin each individual. So we must pay attention to biology. But embodimentis not the only reason for doing so. Equally important are the facts of evolution, which suggest that intentionality emergedrather late. What is the basis of the mental, and when did it emergein evolutionary time?The glib answeris 33

OnrcINs that the mental emergedwhen animalsdeveloped nervous systems.That is not quite correct,howevet the mere possessionof nerve cells does not appear sufficient.In this part of the book, I want to look at this question of origins. My goal is to demonstratethat the minimum condition for the mental is a specifickind of morphology. Before I get to that, however, I want to make the more generalcasefor linking psychology to biology. I shall do this in part by consideringhow philosophersproceedingin the absenceof biology have beenmisled.I then want to show how too narrow a view of psychology can also lead us astray.In doing so, I do not wish to claim that the pursuits of philosophy and psychology independentof biology have been worthless. Often they had to be pursued in the absenceof fundamentalbiological data. Even a wrong belief or a wrong theory can lend energy to a science,sustaining it until the appropriate evidence or methodology is available. So this chapter may be looked at as an historical interlpde, shallow and briel but revealing, I hope, of the rich skein of thoughts*that have been brought to the matter of the mind. The practice of ignoring biology when thinking about the mind and about how knowledge is acquired,without making referenceto biology, has a distinguishedhistory. To a great extent, the philosophy of mind has pitched its inquirieswithout conceming itself (exceptanecdotally)with the body or the brain. We have already seenthat the first modem philosophea Descartes,based his form of rationalism on thought itself using his well'method known of doubt," which he outlined in the Discourseon the Method: I thoughtthat I must. . . rejectasif it wereabsolutelyfalseeverythingabout whichI couldsupposetherewasthe leastdoubt,in orderto seeif afterthat thereremainedanythingwhich I believedwhich was entirelyindubitable. So, on the groundsthat our sensessometimesdeceiveus, I wanted to supposethat therewasnot anythingcorresponding to what they makeus imagine.And, becausesomemen makemistakesin reasoning--rvenwith regardto the simplestmattersin geometry-and fall into fallacies, I judged that I wasasmuchsubjectto errorasanyoneelse,andI rejectedasunsound all the reasoningswhich I had hitherto takenfor demonstrations. ...I resolvedto pretendthat everythingwhich had everenteredinto my mind was no more veridicalthan the illusionsof my dreams. Descartes' conclusion that there was a thinking substanceradically sidesteppedbiology, along with the rest of the materially based order. Given his remarkableforays into biology, this is suqprising.One matter Descartesdid not explicitly malyze, howevet was that to be aware and 34

P u t t i n g P s y c h o l o g yo n a B i o l o g i c a l B a s i s able to guide his philosophicalthought, he neededto have language.And for a person to have language,at leastone other person must be involved, even if that person is the memory of someonein one's past, an interiorized interlocutor. This requirementshakesDescartes'notionthat his conclusions dependedon himself alone and not on other people. Moreover, Descartes was not explicit as to when a human being first has accessto a thinking substancein his development.Perhapshe should have ponderedfurther the likelihood of a Frenchbaby concluding, "le pense donc je suis." Philosophical answers to questions of Cartesian rationalism, such as those provided by the British empiricistsIohn Locke George Berkeley,and David Hume, do not fare much better. Locke's notion of the mind as an empty slate, or tabula nsa, was explored in the absenceof knowledge about developmentalor evolutionary events indicating that entire behavioral repertoires can be under genetic control. And Berkeley's monistic idealism--suggesting that inasmuchas all knowledge is gained through the senses,the whole world is a mental matter-falters before the facts of evolution. It would be very strange indeed if we mentally created an environment that then subjectedus (mentally) to natural selection. The most ruthlessand skepticalof the empiricists,Hume, concludedthat no knowledge could be securegiven that it is all based on senseimpressions.Even scientificknowledge appearedto be shakenby his analysisof causeand effect as no more than mental correlationbasedon the repetition of these senseimpressions.But as we will seelater, senseimpressionsare not the issue;the biology of mind involves much, much more. Immanuel Kant (figure 4-1.), whose background in physics and astronomy was greater than in biology, put the matter in larger perspective.He answeredHume by pointing out the existenceof categoriesa priori in the mind, thus assuringtheir coexistencewith sensory experience.But while the existenceof a piori categoriesappearsin better accord with modem evidence on ethologically determined action pattems and on the neurophysiological properties of brain cells, it is not strictly consistent with developmentalsfudies of how babiesgain a senseof space,or even with the physicsof relativity. Ignorant ashe had to be of modem developments in biology and physics,Kant is to be forgiven for not understandingwhat constraints there might be on the a priori. I could give other examples,but these should sufficeto indicate that, in philosophy, a knowledge of psychology based on experiment and an understanding of neurology and evolution are useful to guard against extreme erors. But all this knowledge is a recent acquisition,and one can only admire the courageand persistenceof thesegreat thinkers in keeping important questionsalive. 35

OnIGINS

FIGURE4-7 , His profound of the Enlightenmenf ImmanuelIGnt (1724-1804),thegreatphilosopher rationalistand empiricistpicturesof the mind, The cartoonon thefacing ideasreshaped pagedepictsthe great man preparingmustard,

Psychology itself has not fared very well in the absenceof knowledge of the brain and nervous system. This is not to say that an enorrnous amount of useful and important information has not been accumulated sinceWilliam famesat Harvard in 1878 and Wilhelm Wundt in Leipzig in t879 founded the ftrst laboratoriesof experimentalphysiological psychology. Instead of a unified theory of the mind, however, a seriesof schools subsequentlysprang up, eachwith different views on behavior, consciousness,and on the relative significanceof perceptio& memory, language,and thought. This is no place to review any of theseschoolsat length. But it may be useful to mention some of their main lines of thought to underscorethe need for a biological common denominator.fameshimself was one of the greatest pioneers of modem psychology. ln Pinciples of Psychology,he arguedthat, while paying attention to the brain, psychology could proceed on its own, investigating mental functions by whatever combination of inkospection, experiment,and psychophysicsproved most revealing.Psychophysicswas also advancedby Wilhelm Wundt, Ewald Hering, and the great physicist Hermann von Helmholtz during the sameera in Germany. 36

P u t t i n g P s y c h o l o g yo n a B i o l o g i c a l B a s i s

FIGURE 4-I (continued)

It consistedof carefulmeasurementsof reaction times and of judgments in responseto accuratelymeasuredphysical stimuli. fames's greatest achievementmay have been to point out that consciousnessis a processand not a substancein his characterizationof this elusiveprocessin his essay"Does Consciousness Exist?",a questionhe also pursued inPrinciples.Whitehead has made the claim that, with this inquiry, Jameswas to the twentieth cenfury what Descarteswas to the seventeenth. During James'stime, however, excessiveattemptswere still being made to use introspection to reachconclusionsabout the mind, often with dubious results (as in the case of Edward Titchener, who regarded experimental introspection as the "sole gateway to psychology" and elaboratedgrand theories of sensationand feeling basedon this method). Similarly, studies of human memory (for example,by Hermann Ebbinghaus)usedabstractor nonsensesequencesand syllables,while paying little or no attention to the role of meaning in memory. Ivan Pavlov's early-twentieth-cenfury experiments on conditioned reflexes offered a strong reaction to theseapproaches.Animals receiving an unconditioned stimulus(food) paired repeatedlywith a conditioned stimu37

OnlclNs

lus oell) salivatedwhen presentedlater with the bell alone.Edward Thomdike and Clark Leonard Hull in the United Statesextended and deepened paradigms.Eventually,the extremeposition the study of stimulus-response that the only scientific study possible in psychology was the "-"rguj study-of behavior. As enunciatedby Iohn Watson, behaviorism left consciousness,introspective reports, and the like outside the pale. The fiercest latter-day advocate of this position was B. F. Skinner, who extensively urplor"d- the phenomunonoi operant conditioning. (lnsteadof responding to a classicalconditioned stimulus,an animalis rewardedduring a particular behavior or operant.This behavior is then reinforcedby repeatedreward') Many tefined chains of behavior were analyzedusing behavioral techniques. Clearly, however, at least part of the baby went out with the balhwater. For example,these approachesdid not encomPassthe Gestalt phenomena (figure 4-2) discovered by Max Wertheimer, Wolfgang kohler, and Kurt Koffka. Gestalt pattems were discemed by thinking subjectsin a way that behaviorismwas hard put to explain. Consciousness simply would not go away. And the observationsof Sigmund Freud,who noted the effects of repression on memory and of the unconsciouson consciousbehavior, pointed up the deficienciesof the behaviorist account' The experimentsof Sir FredericBartlett on human memory indicated that more was involved in memory than the rote repetition of meaningless strings of characters,as the previous work of Ebbinghaushad seemed to imply. Biology and human nature were making strong claims that behaviorism had ignored. One important aspectof human nature and behavior that neededto be accountedfor was revealed in the medical clinic. The discovery of brain maps in the nineteenth century by Gustav Fritsch and Julius Hitzig, who notld specificbodily movements in patients after electrically stimulating parts of their brains, and the discovery by Paul Broca that damage to a speciftcpart of the left brain led to motor aphasia(the inability to produce .th"t"nl speech),could not be ignored. In short order, schools of neurophysiology developed,and by the tum of the century scientistswere well tn the way to measuring actual neural activity. Between the two world wars, a seriesof technicalinnovations developed by Sir CharlesShenington made it possible to detect both the individual and collective activity of nerve cells. Thus the picture of psychology was a mixed one: behaviorism,gestalt psychology, psychophysics,and memory studies in normal psychology; ttudi"t oith"-neuroses by Freudiananalysis;clinical studiesof brain lesions and motor and sensory defects;classificationof the psychoseswith their baffling symptoms in medicine;and a growing knowledge both of neuro3E

Putting Ptychology on a Biological Basis

V

3 U-rr I Gestaltphenomena. These r(anizsa,s fisures,::T;'t:rrno work and showhow contert-dependent perception is. (Cooerthe pac-Manfigureson the top qnd uratchthe apparent.:olt:*t.dinpp_nr.),As Kanizsaput it, "seeingand thinkingare clearly distinguishable actioities. with these'pieces' wecanimaginia cube(figure-on thebottom left):.bu!it is aery.dificyltto seeit," Notice, however, lhat thecubiis completed (figure on the bottomnght) behindthethreeopaque stipesand becomes perceptually present.

anatomy and of the electricalbehavior of nerve cellsin physiology, the first brought about by the neuroanatomicalwork of santiago Ram-ony Cajal and the second by the seminalphysiological work of shenington. only occasionallywere seriousefforts made by researcherssuch as Karl Lashley and Donald Hebb to connect these disparateareasin a general way. For the most part, eachwas pursuedindependentryof the others, and research was sometimes accompaniedby truculent denials of the applicability of competing ideas held by practitioners in ,,outside,,fields. I{hat is cwious about these developmentsis their relative separation from the theory of evolution, a theory absolutely essentialto understand39

OntclNs

ing the matter of the mind. Darwin enunciated the theory of nafural selectionin 1859. It was clear to him that evolution had to affectbehavior and vice versa. But only Darwin's contemporariesGeorge Romanesand C. Lloyd Morgan promulgated the idea that a connectionexisted between evolution and behavior. The effects of development on behavior were appreciatedby C. W. Mills and J. M. Baldwin, but their insights,which are part of the foundation for our modem view, did not penetratethe mainstream.Later developmentalstudiesby JeanPiaget of the cognitive behavior of children laid the groundwork for modem studies of cognition in development. Of course there have been elaborations at the other extreme: efforts have been made to explain behavior in terms of social psychology ever sincethose of Herbert Spencer,another of Darwin's contemporaries.These efforts have generally been at a descriptivelevel and have invoked cultural traits and "folk psychology"-the comnonsense evaluation of human behavior. controversies and speculationsabout nature and nurfure, genetics and environment,have abounded.This rich and sprawling set of studies does not lend itself easily to synthesis.Nonetheless,as modem methods of measuringbrain function developed and an increasedunderstandingof brain biochemistry emerged,it becameclear that psychology could not be pursued without being increasinglygrounded in biology. At best it could be provisionally pursued (asit always has been) while awaiting biological interpretation. Once one arrives at this conclusion,however, there is no escapingan even more fundamentalone:The phenomenaof psychology dependon the speciesin which they are seen,and the properties of speciesdepend on natural selection.This view, taken by ethologists suchas Nikolaas Tinbergen and Konrad Lorenz and also by most modem psychologists,inexorably links psychology to biology. That linkage demonstratesthe importance of evolutionary origins in the behavior of species. In consideringour minds, we must also considerboth our kinship with and our differencesfrom other species.As I discussin chapter 16, one difference is that each of us has an individual "soul" based on language. Whatever we find out about the properties of language,however, the sad fact is that neither psychology nor biology will permit the transmigration of souls.The tale is told of a dying man who consoledhis alreadygrieving wife with a promisethat he would retum exactly six weeksafter his demise, at which time she was to visit a medium. Comforted, she waited patiently and went to the medium on the appointed day. A voice from a dark comer ''tlarry," she said, "is that you?" The of the room said, "Hello, darling." 'What do voice said, "Of courseit's me." Somewhatgingerly, she asked, 40

P u t t i n g P s y c h o l o g yo n a B i o l o g i c a l B a s i s you do every day?" The voice replied, "l get up, make love, take a walh make love, eat, make love, nap, make love. The next day, the same old thing." She took this in and said carefully, "But, darling, I didn't know the angels in heaven made love." The voice replied, "l'm not an angel in heaven,I'm a rabbit in Saskatchewan." While the ideas of philosophersand of different psychological schools must be taken into accountin any considerationof the matter of the mind, such ideas have only lately come to grips with the key issuesof biology itself. The messageboils down to this: The fundamental basis for all behavior and for the emergenceof mind is animal and speciesmorphology (anatomy) and how it functions. Natural selection acts on individuals as they competewithin and between species.From studying the paleontological record it follows that what we call mind emerged only at particular times during evolution (and rather late at that). These terse commentscan be used as the basis for a researchProgram to connect psychology with biology-a program to account for embodiment. Given the record of the history of the philosophy of mind and of psychology, the continued avoidance of the biological undelpinnings of such a program is not likely to enhanceour understanding of how the human mind emergedand how it functions.Errors continue to arise when psychology is pursued without strong connections to biology; I discuss some of them in the Postscript. The centerof any connectionbetweenpsychologyand biology rests,of course,with the factsof evolution. It was Darwin who first recognizedthat natural selectionhad to account even for the emergenceof human conLet us turn to some of his insights and their consequences. sciousness.

4I

CHAPTER

5

andMind: Morphology Darwin'sProgram Completing

But then arisesthe doubt' can the mind of mAn, uthich has, as I fullV belieoe,beendeoeloped from a mind as low as that posby the lowestanimal, be trusteduthenit drawssuchgrand sessed -Charles Darwin conclusions?

lfred Wallace,the codiscovererof the theory of natural selection, wrote a seriesof letters to CharlesDarwin expressingwhat he felt was a hereticalview. Wallace denied that natural selectioncould accountfor the evolution of humans,arguing that the capabilities of the human mind could not be explained by natural selection alone. Darwin (figure 5-1) took the opposite position. He saw no reasonwhy natural selectioncould not have given rise to the basicfeaturesunderlying human thought. His books The Descentof Man and The Fspressionof the Emotionsin Man and Animals were dedicatedto this idea. It is important to understandDarwin's ideas on evolution and natural selection.Simply put, they state that evolution occursas a result of competition and environmentalchange,both of which act on variation in populations (figwe 5-2). Variation always exists in living populations, and it results in differencesin fitness.Natural selectionresults in the differential reproduction of those individuals whose variations (read "structural and functional capabilities"-their phenotype)provide them and their progeny with statistical advantages in adapting to environmental change or in competing with individuals of the same or different species.Differential 42

M o r p h o l o g y a n d M i n d : C o m p l e t i n gD a r u t i n ' s P r o g r a m

{ I 'l \':,!

.!

*":"

FIGURE5_I

thefounderof modcmeoolutionnry theory,thetheoreticql Danpin(1809-1s82), Charles was basisof all of biology,Danpin insistedthnt the eaolutionof humansas a species sufiectto thesamekindsof forcesas thoseleadingto theeaolutionof otherspecies. He retiing in his later life, and this dignifiedbut somewhatlugubriouspictureis became indicatioe.

reproduction and heredity enhancethe likelihood that the traits that creasefitnesswill be preserved. What is critically changed in the resulting population is the frequency of the genes that give rise to those traits. (l discussgenes and genetics in more detail in the next chapter.)The fact that evolution has occurred is scoredby the changein gene frequencies.But the meansbywhich it occurs is natural selection on the phenotype (the total structural and functional capabilities)of individuals. The main leoelat which selection occurs is the 43

OnIGINs NewPopulatlonwlth NewVarlance

]aer Populellon

THINKING POPULATION FIGURE 5_2 Natural of organisms. by mutationin a population thinking,Variationoccurs Population of thepopulationthatshow of thosemembers selection fauorsthediferentialreproduction of genesconferring greaterfitnesson theaaerage. Theresultis that therelatioet'requency dark eyes and tor striped population. Notice selection against increases in the fitness bodies.

individual and his behavior. What we need to understand are the rules connecting the ways in which genes are sorted and expressedwith the ways in which genes lead to changesin the phenotype (figure 5-3). This is a formidable tash one that is only partly completed. While Darwin did not grasp the correct genetic mechanisms,he got the principle right. He understood,for example,that phenotypic resemblances between the emotional expressionsof certain animalsand those of human beings were likely. He also understoodthat natural selectionneednot have selectedall emotional expressiondirectly.The sameconsiderationsapply to thought and behavior. His position was that gradual changesin populations could accounteven for the emergenceand descentof human beings. Now there was much in this position that could not be substantiatedin Darwin's time. Many things were unknown to Darwin, including the true nature of genetic inheritance,essentialdata on hominid fossil remains,and a good deal of important information on how animals develop. But his basicapproach,it fums out, was a sound one. It consistedof understanding what we need to know in order to understandthe evolutionary origin of the human mind, what I call Darwin's program. What we need to under44

M o r p h o l o g y a n d M i n d , C o m p l e t i n gD a r u t i n ' s P r o g r a m

Genesin Popufation

-

Fertilized

T}}:ry

T1

T3

Individualsrn Population

OuI Embryo

FIGURE 5-3 of naturalselection. ofgenesmayberelatedto theactualprocess in thefrequency Changes in a populationof thosegmes of eoolutionis a relativeincrease result and indication The occurson individuals,whetheronspermor egg, that haoemhanced fitness,Butselection Sowemrct unilnstandtherules or on theanimalin itsenoironment. or on theembryo, to theefectsofgenes andbehaoior relatingdeoelopment [theaerticallinesfor trattsformaT, represents theconoersion to adult life in deoelopmmt; tions(TrT)1. T, represents the formationof spermand eggs;To represents the T, represents the erutironmcnt; ready new of deoelopment. egg to undergo a rycle fntilized

stand (asidefrom the mechanismsof inheritance)is how the morphology underlying behavior aroseduring evolutionary history, and how behavior itself alters natural selection.I call this Darwin's program, not becauseit representseverything he wished to know, but becausethis was what most concemedhim in his later years. Of course,Darwin did not complete his program. If we accept his position that there is no aspect of human behavior that cannot eventually be accounted for by an evolutionary explanation, then our task is to try to complete it. What would be required to complete this program?First, an analysisof 45

On IclNs the effects of heredity on behaviot and vice versa.Second,an account of how behavior is influencedby natural selection and in tum influencesit. Third, an account of how behavior is constrainedand made possibleby animal form or morphology. And fourth (and most fundamentalof all), an understandingof how animalform arisesand changesduring development. (By form, I mean not only shapr-limbs, symmetry, and so on-but also the microscopic details of tissuesand organs such as the brain that give them their functions.)This last requiremententails an understandingof the relation between evolution and morphogenesisin development. The nature of this relation is the outstanding and central riddle of modem biology-that of morphologic evolution. Here it may be useful to state what is known about the other requirements of Darwin's program. one is fairly complete we know that the basis of heredily rests in the genes and we understanda good deal about how genes are transmitted, modified, and expressed(figure 5-4). It was the coming together of geneticistsand evolutionists in the 1940sthat allowed the genetic discoveriesof Gregor Mendel to be connectedto the theory of evolution in a most fruitful way. This "modem qmthesis" accounted(as Darwin could not) for the origin of genetic variation as mutations in deoxyribonucleicacid (or DNA) as well as for the rearrangementof genetic structuresin a processcalledrecombination.In short, it began successfully to score the results of natural selection in terms of the changesof gene frequenciesin populations. Subsequentdiscoveries about the nature of DNA and the ability to manipulatethis molecule,even to the point of inserting foreign genesinto animals and changing their form or behavior, triumphantly confirmed the position takenin the modem synthesis.Moreover, great progresswas made in extending Darwin's ideasabout how different speciesarosethrough the sexualor geographicisolation of breeding groups of individuals. Following the modem synthesis,a number of scientistsbegan to sfudy behavior in terms of genetics,evolution, and speciesinteractions.This gave rise to the scienceof ethology, the data of which support the notion that some behavior pattems are species-specificand thus subject to genetic influence.The ftndings of ethologists are more subtle than this, however. They indicate that complex behaviors suchas bird song have both genetic and epigeneticcomponents.For example,some aspectsof the motor pattems underlying the song of certain speciessuch as song sParrows are given from birth as part of the phenotype. So are some variations and modifications of vocalization pattems. But to be able to sing the song characteristicof a song sPalrow speciesin a given area,a sparrow needs to hear the songs of conspecifics-mature birds of the same species.In 46

M o r p h o l o g y a n d M i n d t C o m p l e t i n gD a r w i n ' s P r o g r a m

Gregor Mendel

Charles Darwin

GENETICS

EVOLUTION

NEOD ARWINISM r,_4Blrn$trs.4f$ r

0

DNArr+

FIGURE 5-4 Themodernsynthesis. In the 1940sa groupof eoolutionists andgeneticists reconciled GregorMendel's(1822-ISS4) oiginalfindingson heredity with thetheoryof eoolution dacribedby Datwin (1809-1852). Danoin'soriginaltheoryhadan incorrect notionof heredity.Mendel,an Augustinianmonkfrom Austria,kid thefoundationsof modern genetics, but his contibutionwasnot recognized at first and had to berediscouered in 1901.

specieslike song sparrows,birds deafenedfrom birth cannever develop the full-fledged species-characteristic song. Epigenetic events involving interactions with other birds of the speciesare required for that. It is not too difficult to see how pattems of behavior could affect and be affectedby genetic variation and natural selection.Studiesof this kind, which connect the functions of regions of the brain to the rest of the phenotype, go a long way toward filling in parts of Darwin's program. But one must not follow this approachto excess.Developmentsin the field of sociobiology may serve as a waming. Sociobiologistsare concemedwith how behaviorscan be accountedfor by natural selection.Altruism is a case in point. If natural selection occurs to maximize the fitness of indioiduals, it is difficult to seehow the genesof individuals who sacriftcethemselves 47

OntctNs before they breed,or who lose breeding potential in the serviceof others, could be passedon. The genetic analysisof beesundergoing what is called kin selectionindicate that femalesserving a sister queen can,by forgoing their own reproduction, increasethe frequenry of their genesin a population. This finding, which dependson unusual features of the genetics of bees, is an elegant experimental triumph. But attempts to account for human altruism as a direct consequenceof "genesfor altruism" are another matter-and a dubious one at that. Genesdo not act directly, but rather in complex combinations,to alter form. And form alters behavior in subtle ways. More tellingly, subtle changesin form sometimeslead to rather extraordinary changesin behavior. What we want to know is how alterationsin form, either in the whole animal or at microscopiclevels of brain, muscle,or bone affect behavior, and how behavior alters form. This is the part of Darwin's program that remains largely incomplete. One can appreciatehow extraordinary a person Darwin was by poring over his joumals (figure 5-5). In one of them, the M notebook he says: "Origin of man now proved.-Metaphysic must flourish.-He who understands baboon will do more toward metaphysicsthan Locke." Sad in his last several decades,more or less reclusive,he steadfastlycontinued his indefatigable researches.Only now is it becoming clear how much he actually knew. He thought as deeply about behavior as he did about form. He stands as a profound thinker to whom it would not have occurred to attempt to appear "smart." Incidentally,there is no escapefrom the difficulty posedby the idea that genes specify complex behavior directly by attempting to invoke "group selection." This is the notion, for example, that nafural selection acts to favor quick herds of animals rather than by selecting quick individual animalsthat constitute a herd. Darwin raisedand discussedsucha possibility. With few exceptions,however, it appearsthat most natural selection occursnot at the level of genesor groups of individuals,but rather at the level of individuals themselves. These considerationsonly emphasizeagain the part of Darwin's program that needsmost to be completed.This part is concemedwith how animal form, tissue structure, and tissue function could have arisen from ancestors-the problem of morphologic evolution. To seewhy this problem is so important, one need only think about the extraordinary evidence from fossil data on hominids indicating the large increasein hominid cranial capacity and brain size that has occurred over less than a million years of evolution (seeftgure 5-6). How could this have occurred so rapidly? How does it relate to other 48

M o r p h o l o g y a n d M i n d : C o m p l e t i n gD a r w i n ' s p r o g r n m 83.

The possibility of two quite separate trains going on in the mind as in double consciousness may really explain what habit is- In the habitual train of thought one idea. calls up other, & the consciousness of double individual is not awakened.The habitual individual remembers things done in the other habitual state because it will (without direct consciour.r.s?; change its h a b i ts.Aug. l6th. As instance of heredetary mind. I a Darwin & take after my Father in heraldic principle. & Eras a Wedgwood in many respects & some of Aunt Sarahs. cranksr, & so is Catherine in some respects-. good instances.- when educatio My handwriting same as Grandfather.2

84.

Aug. I6th Anger <in worst form is described by Spenser (Faery Queene. CD 25 (Descript of Queen) of Hell Cant IV or V.) as pale & tranbling. & not as flushing & with muscles rigid.-t How is this? dealt with p.2412 Origin of man now proved.- Metaphysic must flourish.- He who understands baboon <will> would do more towards metaphysics than Locke

FIGURE5-5 An ercerptfrom Darwin's notebooks,

hominid traits we can guessabout from the fossil record and from archeological remains?What is the connectionbetweenoverall morphology and behavior and the microscopicmorphology of the brain?How do these evolutionary developmentsconnect with the behavior of hominids in groups and with the developmentof language? Theseare profound and largely unansweredproblemsin paleontology, anthropology, and archeology. They are difficult becausethe record is fragmentary; the soft tissuesare gone, leaving mainly bones, and thus structureand function can only be connectedindirectly. But one thing is clear.Even if we had better evidence,we would still needa theory of how molphology arisesand how it is changedduring evolution. Why is this so? Morphology-the shapeof cells,tissues,organs,and finally the whole animal-is the largest single basisfor behavior.There is much evidenceto support this conclusionin a gross and even a trivial 49

OnIGINs

2000 1800 1600 o o

.= () (! CL (E

1400 1200

o -(g .E 1000 (r o L

800 600 400 4.0

2.0

1.0 0.5 0.1 Millionsof YearsAgo FIGURE5-6 two million yearsof humaneoolution. in cranialcapacityoCIer Theremarkableincrease We are Homo sapiens sapiens;the rest are our presumedancestors,The shaded lioedand its brainsize,TheHomo thetime in which eachspecies quadrilaferalrepresents sapiensrectangleis for Homo sapienssapiensand Neanderthalman, Thedark spindle the rangeand cranialcapacityin modernHomo sapienssapiens.Originally, represents theseskullsluerenot emptylAn adequatebrain theorymust be ableto accountfor such in brain sizeooersucha shortperiodof eoolutionarytime, a largeincrease

sense.To fly you need wings; to think, a brain. But at another level, moqphologyis an extraordinarilysubtlematter.The smallestchangein the position of the insertion of a levator muscle in the jaws of cichlid fish allowed swallowing to occurindependentlyof bait grasping.This provided the basisfor an explosiveincreasein the occupationof different econiches by descendantvariantsof cichlid fish, an explosiveadaptiveradiation that outstripped many competitors.Moqphology matters for us, too: If one at the genelevel,they show 99 percent compareshumansand chimpanzees identity to eachother. But morphologicalchangeleading to the presence of sustainedbipedalism,alteredjaw muscleinsertionsinto the skull,a larger cranium, a supralaryngealspacewith speechorgans, and a part of the cerebralcortex calledthe planum temporaleappearto have been decisive in leading to characteristichuman behavior. There is evidence for a relation (but not a linear one) between the size 50

Morphology and Mind: CompletingDarwin's Program and complexity of the brain and the complexity of behavior.There is much evidence for the coupling of functions of specific parts of the brain to specific skills. And clinical evidence on damaged brains indicates that specific,recognizableloss of functions of the mind occurswhen particular brain regions are damaged.These various findings suggestthat to understand the evolution of mind and behavior, we must first understand the basesof morphologic evolution. Results like these support Darwin's supposition that human mental capacitiesarose by natural selection.There is diffuse evidencefor mentation in the fossil and archeologicalrecords.Evidenceof burial of the dead, for example,may be taken as evidenceof human consciousness and possibly even of self-consciousness. But perhapsa more telling argument can be constructed by looking at how the human brain is structured, how it functions, how its cells arose,what they have in common with those of other species-and what is different and specialabout it. This requiresa theory of the evolution and development of animal and tissueform. It then requiresthat a theory of brain function be constructed based on the first theory, a tall order that must be filled if we are to complete Danrvin's program. what makes the enterprise fundamentally interesting is that it is nol specificto brains.To accomplishit, we need to show how development (embryology) is related to evolution. we need to know how genes affect form through development.We have to ask how this processconshains evolution-how the rules of development, which themselvesevolved, can only be realized in particular ways. This knowledge is necessarybecauseevolution is historical, because only certain combinations of developmental events lead to functioning shapes,and becausethe shapeof an animal's body is as important to the functioning and evolution of its brain as the shapeand funclioning of the brain are to the behavior of that body. In the next chapter, we will look at how embryology and evolution interactto result in brainsand bodies.A word of waming is in order: expect no miraclesof simple explanation.Given what we know about evolution, it is no more likely that a gene can be found for altruism than that a single biochemicalsubstancewill be found to distinguish an ape from a human. The connectionsbetween moqphology and mind that complete Darwin's program will be more indirect and circuitous than that. yet their intricacy will make them all the more intriguing. Not the leastof the intrigue is how a fertilized egg gives rise to a functioning animal,brain and all. This will, I hope, justify a change of pace to present a mini-course in modem molecular biology and development.we will need some of its lessonsin the next part of this book when we face the problem of constructing a brain theory that is consistent with evolution and development. 51

CHAPTER

6

Topobiolo from gy:Lessons theEmbryo

"The Chickenand the Egg,Togetherat Last" -Title of a review of Topobiology, '1.989 New York TimesBookReuiew,lanuary22,

t may seemstrangethat we must concernourselveswith embryology when this book is about the mind. Eggr and sperrnshow no evidence of mind, and neither do very early embryos.But sincewe know that newbom infants do show evidenceof mind, however feebly, it seems reasonableto wonder by what interactionsthe basesfor mental life have been laid down. But why deviate to such issuesas shapeand form? And why concem ourselveswith cells, molecules,and DNA? The straightforward answer is that the rules by which embryos are built govem the way that brains are built. The actualformation of the anatomy of the brain dependson muscles acting on bones,nerves acting on skin in a given order, and so on-that is, it depends on the rest of the phenotype. And as I stated in the last chapter,if we are to understandwhen aspectsof mind arosein the course of evolution, we have to understandthe connectionbetween development and evolution. To examine development, I will of necessity rely on some technical words and details.My suggestionto the readeris to go by the detailsonce, look at the figures, and then retum to the text. Let me dispose of some preliminaries(figure 6-1). The cells of higher organisms(calledeukaryotes) have nuclei that contain DNA, the hereditary material. DNA is made up 52

T o p o b i o l o g y :L e s s o n sf r o m t h e E m b r y o of long strings of four smallermolecules,called nucleotide bases,that are linked in sequences. There are only four types of bases(guanine,cytosine, adenine,and thymine, or G, C, A, and T for short). Thus, the sequenceof a strand of DNA might look like this: . . . GTCGACCTGGCAGGTCAACGGATC . . . It is now known that each strand of DNA containing such a string has a complementary strand coiled with it: . . . CAGCTGGACGTCCAGTTGCCTAG . . . Complementary basespair with each other: Notice that in this "double helix," G from one strand pairs with C from the other, and A with T, whereaswithin each strand the Gs, Ts, and so on are linked by strong chemical bonds, like beads on a shing. By contrast, across strands the Gs pair with Cs and As with Ts by weak cl'remicalforces that allow the two strandsto come apart, for example,as the result of an increasein temperature. The key points are these 1. Using a single strandas a template,a second DNA strand can be built from single basesby special protein enzlnnes, which catalyzeor speedup the chemicallinkage of one base to the next in the strand being formed. The order in the new strand is determinedby the pairing of the right baseto its complementarypartner on the opposite strand. 2. A sequenceof three bases(any combination of G, C, A, and T) on a strand of DNA representsa code word (or codon) telling the cell to incorporate a particular protein building block called an amino acid into a long string of such amino acids, called a polypeptide. This polypeptide chain then folds up to form a protein. If eachcode word is three nucleotides long, sixty-four code words can be constructedfrom four kinds of nucleotides, yet only twenty amino acids occur in proteins. Obviously, then, some code words do not code for amino acids, while others are simply redundant. A piece of DNA of the right length and base sequenceto specify a protein is known as a gene. 3. When a cell divides, it copies the DNA from one of the strands to provide new DNA for its two daughter cells. Normally, each copy will have exactly the same sequenceof code words. If, however, a mistake is made or a DNA strand is cut, say by a cosmicray, replicatingor repair enzymesmay not copy the templatestrand faithfully. This is one way in which mutational changeis incorporatedinto a gene, altering its code. The discovery that DNA was the genetic material was made at The Rockefeller Institute by Oswald Avery and his colleagueg and was reported in an extraordinarily significantpaper in the Journalof Fsperimental Medicinein 1944. Curiously, the paper did not causean immediateexplosion of belief or interest.I used to play music with Stuart Elliot, a microbiologist who worked closely with Avery and also with Fred Griffith in England,another pioneer in this fteld. One day in 1964, Stuart suggested 53

OnrcINs

DNA

Replication

into RNA Transcription Phenylalanine

Translation

Protein

FIGURE6-1 The readingof the geneticcodeinto protein (a mini-coursein molecularbiology). DNA consistsof hno strands held togetherby wmk forcn betweencomplementarynucleotide bases.Guanine (G) pairs with cytosine(C), while adenine(A) pairs with thymine (T). The basa within a singlestrand, linked by much strongerforces,can be readin sequence codewords, eachspecifyinga particular amino aciil (the building as a seriesof three-base block of protein). While DNA itself stays in the nucleusof the cell, its codeis carried in the cell in theform of single-strandedRNA, which is built by specialmzymes elscuthere that hanscibe the DNA sequmce.(The RNA codeusesone diferent base,uracil M, which takesthe placeof thymine, but is otherwisethe sameas that of DNA; sometypical codeutordsare LllJl) : phenyhlanine,CUU : leucine,GGC : glycine,and so 54

T o p o b i o l o g yL: e s s o nfsr o m t h e E m b r y o that I approachPeyton Rous,the editor of the Joumalof Eryeimental Medicine,to proposethat the joumal republishone of Griffitlt'soriginal paperstogetherwith the great paperof Avery and his colleagues. This, Stuartsuggested, wouldappropriatelycommemorate both the twenty-fifth anniversary of Griffith'sdeathandthe twentiethanniversaryof the Avery PaPer. Rous(who in his eightieswon the Nobel plJrzefor his early work on virusesthat causecancer)promisedto "takeup the suggestionwith the boardof the joumal."After hearingnothing for six weeks,I encountered Roussteppingoff a bus,and inquired.He stopped,staredgravelyat me, andsaid,"Ah, yes.I took it up with the boardandthey thoughtit wasan extremelyvulgar suggestion."Surprised, I walkedwith him in silencefor a block,and then I heardhim say quietly,"l neverdid like Avery very much."WhenI askedhim why, he said,'What would you think of a man who got a medalfrom the RoyalSocietyand neverwent to pick it up?" Avery did not get a Nobel Prize.His work was only fully appreciated someyearsafterhis death.Evennow,with our molecularbiologicalunderstandingof DNA, it is not easyto imaginehow momentoushis discovery was.Credit in science, as elsewhere, is not alwaysevenlydistributed. This little pr6cis of molecularbiology allows us to say something alreadyabout the relationbetweenthe genotype(the set of genespossessedby an organism)and its phenotype.The genesconsistof long stringsof codewords,with startandstopsignals(thesearealsopart of the code).Throughthemachineryof thecell,DNA is copiedinto anotherlong string of somewhatdifferentnucleotidescalledRNA. The RNA is then shippedout of the cell'snucleusto be read by a cellulardevice(like a tapehead) that bringsaminoacidscorresponding to the codingsequence togetherto be linkedin theproperorderto makea polypeptideof perhaps severalhundredaminoacidsin length (ftgure6-L). Whenfinished,this polypeptidefoldsup in a complexshapeto form a moreor lesscompactprotein(figure6-2). The orderof the aminoacidsin f*-th) \* KNA codesequmceis then "trarclated" on a specialstructure,the ibosome, whereother enzymeslink eachamino acid in tum ss the codesthat specifuthem are read of the KNA. (The amino acidsare carriedto the siteby another'kini of RNA.) The fglypeptide of linlcedamino acids that resultsfolds into i three-dimmsionalshapethat on its codedseque!1ce. yslt change,or mutation, in the DNA codemay alter 4 the sequence of amino scids,which ffiqy cause,in turn, theprotein'sshapeto chang;e.This cana.lteritsfunction,which is caniedout by an actioesiti on the foldedstructure.Wen a cell ditides, the two DNA stranfu separateand an enzymecoiies eachstrand to gioe identical DNA to eachof the cell's twaodaughter cettsii o prorr* called replicatioT. 55

OnrcrNs

FIGUREF2 Proteinfolding and function-an example,A foldedproteincalledherokinasecan bind the sugarglucosein the cleft that is fhe actioesitefor its enzymefunction,catalyzinga are linkedto glucoseduring metaboderioatir:es chemicalreactionin uthichphosphorous Bottorn: Whenglucoseis nddedthecleft of gluco.se, Iism,Top: Theproteinin theabsence , closes around it, binding it securely

its chain determines the shape of the protein. The shapeconfers phenotypic properties and functions on a protein; for example, certain shapes might allow it to fit together (like blocks) with other proteins to form cell structures, while others might allow it to bind to chemicals and change the speed with which they react. As mentioned before, this is the k"y property of an enzyme. To summarize:

s6

T o p o b i o l o g y , L e s s o n sf r o m t h e E m b r y o 7 . DNA "makes" RNA which "makes" protein (where the quotes mean "specifies"-it is the cell that actually makes the chemicals). 2. The shape of the protein depends on its sequenceof amino acids, which depends in turn on the original sequenceof the code words

in the correspondingDNA. 3. The function of the protein dependson its shape. Sincemuci of an organism'sphenotype dependson the properties of its proteins, these mles might seem to account not only for the shape of proteins but also, by extension,for the shapeof animals. Alas, things are not so simple, for it is not by building up proteinsbut by building up cellsthat an embryo is made.Its shapeand that of its tissues, including the brain, derive from the shapesof collections of cells of a variety of types, each type with different proteins (the differencescome from the fact that different combinationsof genesare expressedin different types of cells). So we must now ask how cells do this. But in doing so, we must not lose sight of a key point-the shapeof the animal ultimately doesdepend on the order of the code words in its DNA. Moreover, changesin shapes over the courseof evolution must have arisenfrom mutations that changed the order of the code words in the DNA of an ancestor.so the questions we want to answer are: 1. How does the one-dimensionalgenetic code specify the shapeof a three-dimensionalanimal (not just a three-dimensionalprotein molecule)? 2. How can we accountfor changesover time in the developmental processesleading to such shapesso that new shapesevolve? To answer these questionsin a provisional way, I wrote a book called Topobiologyexplaining, among other things, how brains could have evolved. Toposmeansplace,and the title refersto the fact that many of the transactionsbetween one cell and another leading to shape are placedependent:They occur only when a cell finds itself surroundedby other cells in a particular place. Let us consider some of the place-dependent events that lead to the formation of an embryo and its organs,pa*icula.ly its brain. An embryo is formed when a sperrn containing DNA from a male fertilizes an egg containing DNA from a female. (By the way, the germ cells-sperm and egg-are enorrnouslyvaried becauseeachmay contain geneswith different mutations.)There are lots of genesand eachhasa long 57

OnIGINs string of code words. The fused sperrnand egg (or zygote, as it is called) now has genes from both parents and it begins to undergo a seriesof divisionsor cleavagesmaking 2, 4, 8, . . . 2n cells. The shapeof the massof daughtercells that emergesis usually that of a ball (althoughin birds it canbe a sheet).Now I haveto stop and consider some of the things cellsdo before I go aheadwith this description: 1.. Cells diaide,passing on the same amount and kind of DNA to their daughter cells.

2. Cells migrate,seParatingfrom their connectionsin sheetscalled epitheliato form a loose,moving collectio. calleda mesenchyme. (fhe sheetsthemselvescan also move by curling up into tubes without releasingthe contactsbetweentheir cells.) 3. Cellsdie in particularlocations. 4. Ce\lsadhereio eachother,aswe havealreadyseen,or they losetheir adhesionand migrateto anotherplace.This migrationo.rurs on the surfacesof otheicells to form layers,or on matrix moleculesreleased by the cells.The cells then readhere,forming new combinations. they expressdifferentcombinationsof the genes 5. Cellsdifferentiate; Th"y .* do this at any time and placebut nuclei. pr"r"r,iin their right cues. Only certain Pl-""::- in the the fnb if they receive right cues.This processof differential the developingembryo have It is what makesliver cells differentiation. is called gun" "*prirrion from brain cells,and different cells skin iiff"r"ni from skin cells,and protein productiory of pattems specific so on. Differentiationmeans on and some tumed are proteins somegenesspecifyingparticular only proteins' many has type are tuired off. U".tt ."[ of a given type' different a someof which are sharedwith cells of Now we may resumeour descriptionof how an embryo is made' Let us considerthe chick embryo (figure 6-t). Continued cell division eventually than leads to a plate of cells called the blastoderm,which contains more the in At this point, cells on either side of the midline IOO,OOO ""llr. middle pori"rio, portion of thetlastoderm detachand migrate through the up end cells these streak. The result is that iortion *tt"d tt u primitive called the feneath the blastoderm, where they adhere to form a layer endoand mesoderm, ectoderm, mesoderm. Three separatelayers-the gastrulation. derm----eventually form through this process, called occurs An amazingevent that combinescell position and cell signaling signals at this stuge.ihis is calledembryonicinduition,and it is the result of in Cells another. in cells layer to a set of f"ssir,g fri* " set of cells in one a placein ih" -""rod"r* send signalsto those in the ectoderm,resulting 58

l

T o p o b i o l o g y : L e s s o n sf r o m t h e E m b r y o

Backbone

FIGURE 6-3 Theearlydeoelop.ment of thechick.embryo layers froma plateof celbthatformsseoeral i1ra process gastrulation. called Cellsmoaethroughtheprimitivestreik(topleftytoiorm layers(top nghl. Thecentralarisof theplatethenfoldsintothenrrroitibr, wkici wiil latergioeriseto thener?Jous systetn.soonafterwari,celkfrom thelowerlayirssegregate (centerleft) andform segmented structures calledsomites(center right). Thenervo1ts systemde.oelops (shiwnin figure J-3). furtherascellsin theneuraltubedeoelopprocesses Theresultis an embryothat beginsto looklikean indioidualanimal(boti3d.

s9

OnrctNs dependent (topobiological) differentiation of a central section of cells to form the neural plate. Cells outside the boundary of the neural plate will form the skin; those of the plate itself will form the neuraltube (figure 6-3) and subsequentlythe nervous system.They do so by rolling up as a sheet into the tube. This not only definesthe axis of the embryo but setsthe head end of the animal. It also sets the position-dependent cues for future induction events. Notice how these primary cellular Processesof division, migration, death, adhesion,and induction vary as a function of time and place.The critical step for consolidating these events is coordinatingthem to present new inductive signalsleading to still further alterationsin a particularplace, one madeas a result of past alterations.As a result of secondaryinductions, for example,nerve cellsin the neuralfube sendout fibers to form networks, eachspecificfor a region of the brain or spinal cord (seefigure 3-3). Other cells will form the eyes,gut, or kidneys.All of this occursin sucha fashion as to yield the shapecharacteristicof the species-in this case,a chick. This matter of shapeis critical. It means that combinationsof genes act to give a heritable shapecharacteristicof that species.It also meansthat the mechanicalevents leading to the rearangement and specializationof cells must be coordinatedwith the sequentialexpressionof the genes.This is the key requirementof topobiology.It explains why genes specifying the shapesof proteins are not enough; individual cells,moving and dying in unpredictableways, are the real driving forces. Making proteins or cell surfacesthat latch on to eachother, eachspecificfor a given cell like a Lego toy, does not account for how genes specify shape.While the cells of an embryo of a speciesresembleeachother on theauerage,the movement and death of aparticularcell at any particularplaceis a statisticalmatter and that cell's acfual position cannot be prespeciftedby the code in a gene. What then does accountfor how shape is achieved in this marvelous sequenceof cellular dances and signals?A clue lies with the molecules called morphoregulatory moleculesthat regulate adhesionand movement (figure 6-4). Theseproteins are specifiedby certainsetsof genesat particular placesin the embryo. Their main function is to causecells to adhereor to link cells in sheets called epithelia. They fall into three families: cell adhesion molecules (CAMs) that link cells together directly, substrate adhesionmolecules(SAMs) that link cells indirectly but provide a matrix or a basis on which they can move, and cell junctional molecules(ClMs) that link cells bound by CAMs into epithelial sheets. The key point is this: The activation of genes for subsets of morphoregulatory moleculesmodifiesthe mechanicsof cells and epithelia.This processis determinedby the chemistry of each cell acting on the intemal 60

T o p o b i o l o g y t L e s s o n sf r o m t h e E m b r y o

CAM

Gell1

Cell2 Border

Tissue 1 (Linkedby one combination of CAMsand SAMs)

Tissue 2 (Linkedby another combination of CAMsand SAMs)

FIGUREG4 Cell adhesion.Top: Celb bind to eachother by meansof specialproteisentin their outer membranescalled cell adhesionmolecules(cAMs). Bottom: other moleculescalled substrateadhesionmolecules(SAMI) form ertracellularmatices on which cellsmooeand rest. The cAMs and sAMs regulatehow celb assembleand disassemble and permit or that underliethe formation of shape,as shown in the forbid mooement.The processes preoiousfigure, are underthe control of specialgenesthat are erpressedin placesat which molecularsignals(morphogens, M, and M) are exchangedbehpemadjoining collections of cells. The topobiologicalprocessis inhicate, but the basic idea is simple: Cells move or stick at a particular place;after a set of genesis turned on or of, thi celtsare either released or keptand thenproducenewsignalsfor new combinations.This resultsin further changesin cAM and sAM erpressionand the alteration of cell and tissuetypes by morphogens.Theseprocesses result in the shapesand the tissuetypes making'up ttre embruo.

61

OnrclNs structures that affect cell shape and movement, and is therefore called mechanochemistry.A specific combination of CAMs and SAMs, for instance,allows some cells to move, controls mechanochemicalevents that fold the sheets formed when cells are linked, and even prevents the movement of certain cells in certain places.Becauseof these altemating permissiveand restraining roles, the expressionof CAMs that are linked to the surfacesof cells,and of SAMs that are depositedby cells to interact with the surfacesof other cells at particular places,can alter the combinations of cells at a given place,leading to different shapes.Particulargenes control the formation of particular CAMs and SAMs, so this role can be inherited for a given species, thus determining its shape-forming mechanism. Even this is not enough,however. Along with CAMs, SAMs, and CfMs expressedat specific places in specific amounts and kinds, other events must also occur. After a place is constructed by inductive signals, new signal combinationsmust tell cells in that place which new CAMs to tum on and which old ones to tum off. Thesesignalsconsist of small molecules or growth factors that interact either directly or indirectly with the appropriate genes to control this expression(figure 6-4). A similar thing must happen to causecells to differentiate or not to differentiatein the right place.Molecularbiologists have identified a special kind of gene known as a homeotic gene. The homeotic gene specifies proteins that bind to portions of other genesand subsequentlyregulatethe production of the proteins specifiedby those genes.In this manner,homeotic genescontrol the differentiation eventsthat makea body region such as a wing, or an eye, or a part of the vertebral column. In a fruit fly, for example,a mutation in a homeotic gene can causea leg to grow where an antennashould be (figure 6-5). Homeotic genesare expressedin gradients acrossthe animal, usually front to back, and in particular regions. To summarize:Cells expressgenes in time and spaceto govem morphoregulatory molecules,which in tum control cell movementsand cell-tocell adhesion.These actions place groups of cells in proximity, allowing them to exchangefurther inductive signals.These alter the expressionof homeotic genes,which then alter the expressionof other genes.The key playersin this topobiological cascadeare the cells,which move, die, divide, releaseinductive signals or morphogens, link to form new sheets,and repeat variants of the process.Genescontrol the whole businessindirectly by goveming which morphoregulatory or homeotic product will be expressed.But the actual microscopicfate of a cell is determined by epigenetic events that depend on developmentalhistories unique to each individual cell in the embryo. 62

T o p o b i o l o g y :L e s s o n sf r o m t h e E m b r yo

FIGURE F5 An erampleof abenanttopobiology, Homeoticgenes,which regulateothergenes,are expressed in particularplacesduringthedeoelopment of theembryo.If a homeotic gene undergoes s mutation,a bodypart suchasa legcanreplace an antennnin thefruit fly Drosophila.Thepictureon theleft showsa normalfly head;theoneon theight shows q mutantflV with legswhne theantennne shouldbe.

The end result is molphology. And becauseeachspecieshas a particular combination of genes,the frequencyof which is establishedon the average by natural selection during evolutioo there is a shapeor tissue structure more or less characteristicof that species. This account provides a provisional answer to the question of how a one-dimensionalgenetic code can specify a three-dimensionalanimal. It also suggestshow evolution leadsto relatively large and rapid changesin morphology. Suppose,for example, that, during evolutio& a mutation affected the timing or binding of a morphoregulatory molecule,delaying its expressionin sufficientamountsuntil cells in the embryonic region had divided more than usual.'A larger structure might result, with a different shape.If animalswith the new size and shapeshowed increasedfitness in a given environment,natural selectionwould lead to differential reproduction of these animals.This would result in an increasein the frequency of the mutant gene in that populatioru and more animalswould be bom with the variant size and shape. Why have I gone into this degree of detaih The reasonis twofold: The nervous systemand the brain are formed by suchprocessesasaredescribed here,and the signaling that occursin the nervous systemis topobiological (seechapter3). The mapsof the nervous systemthat result from nerve cells sending their processesto other regions of cells during development are among the most remarkableof topobiological structures.Their formation 63

OnIcINs has been shown to depend on moqphoregulatorymolecules.Moreover, mapsoften dependon the selectivedeath of the cellsthat competeto make them. They also require specialsignalingprocessesthat locate the branches of active neighboring nerve cells in such a way that they end up as neighbors again in the distant map that is their target. Evidencesuggests that if the chemical or electrical activity of these cells is blocked, their brancheswill not form orderly maps in somedistant place(seefigure 3-4). Imagine now this epigeneticdrama in which sheetsof nerve cells in the developing brain form a neighborhood. Neighbors in that neighborhood exchange signals as they are linked by CAMs and CJMs. They send processesout in a profusefashion sometimesbunchedtogether in bundles called fascicles.When they reach other neighborhoods and sheets they stimulate target cells.Thesein tum releasediffusible substancesor signals which, if the ingrowing processeshave conelated signals,allow them to branch and make attachments.Those that do not either passon or retract. Indeed,if they do not meet their targets,their parentcellsmay die.Finally,as growth and selectionoperate,a mappedneuralstructurewith a function may form. The numberof cellsbeing made,dying, and becomingincorporatedis huge. The entire situation is a dynamic one, depending on signals,genes, proteins,cell movement, division, and death,all interacting at many levels. Notice the main features of this drama. It is topobiological, or placedependent.Eventsoccurring in one placerequirethat previous eventshave occurred at other places. But it is also inherently dynamic, plastic, or variable at the level of its fundamentalunits, the cells.Even in genetically identical twins, the exact samepattem of nerve cells is not found at the because sameplace and time. Yet the collective picture is species-specific that species. characteristic of genes are the ooerallconstraintsacting on the The events I have describedare selectionalones.Certainpattems of cells are selectedfrom a variant mass of cells in a topobiological fashion.This is dramatically the casein the nervous system.Selectionnot only guarantees a common pattem in a speciesbut also results in indioidualdiversity at the level of the finest neuralnetworks. I have alreadymentioned that the diversity or variability of the connectionsat a given place in the nervous system argues against the idea that the brain functions like a computer. Diversity must inevitably result from the dynamicnature of topobiological events. The existenceof diversity at the level of the individual animal is of great importance. Indeed, it is likely to be one of the most important featuresof the morphology that gives rise to mind. But we are not there yetl First, we have to ask how biological systems carryrout recognition events-how, without the transfer of preexisting, specifically coded messages,a biological system nonethelessspecifically distinguishesone thing from another. 64

CHAPTER

7

Reconsidered TheProblems

Man consideringhimselfis the great prodi4Vof nature, For he uthat his body is,eoenlesswhat his spirit is,and cannotconceioe leastof all how body can be unitedutith spirit, That is the peak of his difficulty and yet it is his nery being' -Braisepascar

utting the mind back into naturehasprecipitateda seriesof scientific crises,for the data on the brain, mind, and behavior do not colrespond to the pictures we have been using to explain them. Many people think this is an audaciousconclusion-unwarranted, premature (morefactswill clearthings up!),or even downright unhealthy.I think, on the contrary, that the best time to be working in a scienceis when it is in a crisis state.It is then that one is prompted to think of a new way of looking at the data,or of a new theory, or of a new techniqueto resolve an apparentparadox. One of the most striking crisesof modern science occurred,for example,when it was understoodthat the applicationof the classicallaws of physics to a heated metal block with a cavity (a "black body radiator") led to an impossiblesituation at short wave lengths and high energies;in this so-calledultraviolet catastrophe,energy becomes infinite.The solution compelledby this situationwas given by Max Planck, who suggestedthat energy was not radiatedbV a hot body continuously but rather in packetsor quanta. The crises in considering the matter of the mind are in no way as clear-cut,however. This should come as no surprise,given how subtleand multilayered the businessof brain development,brain action, and mental 65

OnrcrNrs activity is. It begins with moleculesand goes on to genes.It involves vast numbers of cells with electrical activity and chemicaldiversity, an enormously intricate anatomy with blobs and sheetslinked in rich ways, and maps that receive signals from sensory input and send signals to motor output. These structures undergo continuous electrical and chemical change,driving and being driven by animal movement. This movement is itself conditioned by animal shapeand pattem, leading to behavior. Some of this behavior involves communicationwith an animal'smemory, which is in tum affectedby its own products. All of this comes about as a result of evolution-that is, as a result of natural selection operating over hundreds of millions of years. It is no wonder that the crisesof brain scienceand psychology are not as neat or as evident as those of physics. The sheer complexities are much greater than those of physics. Yet as I have shown (see chapter 3), the crises emerge in stark clarity provided that one is willing to reach across the different levels in an effort to relate structure to function. What can we do to reconcilethe interactionsof the various levels and to resolve the crisesof structure and function that they jointly pose?The answer lies in seeing what the critical problems are, avoiding category errors, and constructing a theory. Of course, this theory must be scientific-that is, it must be testableor falsifiableby experimentalmeans.But it need not alwayslead to predictionsat all of its levels,nor need every part of it be immediately or obviously falsifiable.(Had such strict criteria of falsifiability been applied to Darwin's evolutionary theory at its inception, it would have been prematurely abandoned.) The next part of this book will summarize such a theory, already describedat much greater length in my trilogy of books on morphology and mind (seethe SelectedReadingsat the back of this book). I will give the gist of what is containedin that set of volumes. I plan to simplify the task by first describingsome known biological systemsthat have properties analogousto those of the brain. But I want to wam the reader that theseanalogiesareheuristic;they are intendedto help with the comprehension of certain mechanisms,not to be explicit examples of cognitive functions. Before tuming to these analogies,it may be useful to reconsider the problemswe startedwith and to summarizethe argumentthus far. As long as scienceand scientificobserversdealt with physical objects and nafural forces independentof the minds of the observers,a grand set of theories within a group of compatible sciencescould afford to ignore the psychological intricaciesof scientificobservers.While their sensationsand perceptions went into the performanceof their experimentsand into intersubjec66

T h e P r o b l e m sR e c o n s i d e r e d tive exchangeswith their colleagues,these sensationsand perceptions were strictly excludedfrom their theoreticaland formal explanations.Aside from a few difficulties at the boundaries of the very small (in quantum measurement)or of the very fast or large (in relativity theory), the scientific to be from a God's-eyeview. An "objectiobservers'participationappeared vist" picture of nafure developedthat distinguishedthings from eachother by "classicalcategories":categoriesdefinedby singly necessaryand jointly sufficientconditions. Thesewere then mapped onto the physical world in an unambiguousfashionby incorporating experimentaldata into far-reaching physical theories. In many domains,this approachworked enviably well (and still does), but when the mind was put back into nature by nineteenth-centurystudies of physiology and psychology, a series of difficulties began to emerge. One of the first of thesedifficulties was that the observer could no longer neglect mental events and mental experience.He could no longer ignore itself or the fact that consciousexperiencewas intentionalconsciousness always in referenceto an object. The mechanismsof this consciousness were not directly transparent,nor could consciousnessbe studied directly asan extemal object-at best,it could be introspectedor indirectly infened from the behavior of others. One reaction to this state of affairswas to declarethe subjectoff limits and insist that scienceshould concem itself only with behavior that was observablein ways defined by the forms of successfulscientific inquiry concemedwith nonintentional objects.In an attempt to salvagethe "scientific" posture without denying intentionality, and in conhast to this behaviorism, a different position was later taken by cognitive science.The cognitive position was to adopt notions derived from logical and formal analysis,putting an emphasison slmtax. In this view, the mind, like a computer, is organized by rules and operatesby mental representations. Meanings or semanticsare supposedto ariseby mapping theserules onto classicallycategorizableevents and objects.Unlike behaviorism,this view allowed one to look into the mind but then describedit as if it were a formal system. This description floated more or less free of the detailed structure of the brain. The semanticmapping of that description onto the world is objectivist; things and events are unequivocally described as classicalcategories. As I discussin the Postscript,however, proposalsthat the brain and the mind function like digital computersdo not stand up to scrutiny. The idea of mental representationsposited without referenceto brain mechanisms and structuresdoes not fare much better. An examination of how animals and people categorizethe world, and how babiesmentally develop, under67

OnrcrNs cuts the idea that language can be adequately explained by syntactical analysescarriedout in the absenceof an adequateexplanationof meaning. The objectivist view of the world is at best incomplete and at worst downright wrong. The brain is not a computer and the world is not a piece of computer tape. As a young scientist,I believed that physics would explainand exhaust everything, at least someday. I didn't know it, but I was an objectivist. Now, while my regard for physics remains as high, I see that some supplementation is required to get intentionality into the picture. My present view of the structure of things is exemplified by the story of the gentleman who had the paranoid delusion that his girlfriend was seeing another suitor. One hot suruner night he camehome early to their apartment and in a fury of jealousy searchedeverywhere for the hypothetical suitor, but could not find him. Still in his rage,he found himself at the back window of the apartment.He looked out at the fire escapebelow, and saw a man loosening his collar and wiping his brow. Flying into a greater rage, he lifted a very large refrigerator, ftt it through the window, aimed it carefully, and dropped it on the man's head-whereupon he fell dead of exertion. The sceneswitchesto Heaven.Three people are being admitted; Saint Peter tells them that they have fulfilled all the requirementsbut the last, which is to describethe nature of their deaths.The first man said, "Well, I thought there was some hanky-panky going on, so I came home early. I looked all over the place and ftnally found this fellow, and I must have had an adrenaline fit. I lifted a refrigerator I could not ordinarily lift, dropped it on his head,and must have had a heart attack." The secondman said,"l don't know. It was a hot summernight. I steppedout onto the fire escape,loosenedmy collar and wiped my brow, and a refrigerator fell on my head." The third man said, "l don't know. I was just sitting in this refrigerator, minding my own business." The physicsof falling bodies,certainly, and also of intentionality, underlaid by some rather critical morphology in the head,none of which can be disturbed by unverified fantasies or by heavy objects without serious consequences. The notion that we can think about how mental matters occur in the absenceof referenceto the strucfure,function, development,and evolution of the brain is intellectually hazardous.The likelihood of guessinghow the brain works without looking at its structure seemsslim. Certainly, if one agreeswith the ethologists that mental statesare a product of evolution, we must at least sfudy how the brain evolved. Our obligation is to complete Darwin's program. 68

T h e P r o b l e m sR e c o n s i d e r e d When we make even our first halting efforts to do so, we come upon a seriesof intriguing and baffling findings. We see that the development of brains is enormously dynamic and statistical.Developmental analysis suggeststhat the way genesregulate the intricate anatomy of the brain is through epigenetic interactions-particular developmental events must occur before others can occur. Certain adhesionmoleculesregulate collectives of cells and their migration, but do not do so cell by cell in a prescribedor prearrangedpattem. And to some extent, cell migration and cell death are stochastiethey have unpredictable consequencesat the level of individual cells.Thesestatisticalprocessesoblige individual brains, unlike computers,to be individual. The somaticdiversity necessarilygenerated by these meansis so large that it cannot be dismissedas "noise," as one would dismiss the noise in an electronic circuit at normal operating temperafures.(The hiss from your hi-fi amplifier is an example.) Indeed,the circuits of the brain look like no others we have seenbefore. The neurons have treelike arbors that overlap and ramify in myriad ways. Their signaling is not like that in a computer or a telephone exchange;it is more like the vast aggregate of interactive events in a jungle. And yet despite this, brains give rise to maps and circuits that automatically adapt their boundariesto changingsignals.Brainscontain multiple mapsinteracting without any supervisors,yet bring unity and cohesivenessto perceptual scenes.And they let their possessors(pigeons,for example)categorize as similar a large if not endlessset of diverse objects, such as pictures of different fish, after seeing only a few such pictures. If you considertheseextraordinary brain properties in conjunction with the dilemmas createdby the macline or the computer view of the mind, it is fair to say that we have a scientificcrisis.The question then arisesas to how to resolve it. For a possibleway out, let us look to biology itself rather than to physics, mathematics,or computer science.

69

PART III

PROPOSALS

We are now in a position to usewhat we know about biology, psychology, and philosophy to postulate a theory of consciousnessthat will be an essentialpart of a theory of how the brain works. Most scientistsconsidersuch efforts premature,if not downright crazy. But the history of sciencesuggeststhat we progressnot by simply collecting facts but by synthesizing ideas and then testing them. It also teaches us that nothing is so effectivein promoting new thoughts and experiments as a theory that one can amend or even knock down. The theory must be a scientific one: Its parts must be testable,and it must help to organizemost, if not all, of the known facts about brains and minds. To accomplishthis for the matter of the mind meanssifting through several layers of organization in the nervous system. It also means rethinking what we mean by "memory," "concepts," "meaning," "consciousas an animal," and "consciousas a human being." At the very least,in attempting to do so, the readerwill leam about some fascinatingbiological phenomenaand findings. At the most, he or shemay glimpse a view of the material basesof mind. The readeris urged to have patience-we are at the frontier, a place where boundaries shift, where although amenitiesmay be lacking, the senseof excitementis heightened.

7T

CHAPTER

B

The Sciences of Recognition

Selectionis betterthan instruction.

-Anonymous

t this point, we are preparedto approachthe matter of the mind from a biological point of view. This is lessa matter of biochemistry, cell biology, or neurophysiology than it is a conceptual mode-a way of looking at mental matters from a vantage point basedon biological concepts. biological It is not commonly understoodthat therearecharacteristically in sciences. required other not even presentor modes of thought that are One of the most fundamentalof these is population thinking, developed largely by Darwin. Population thinking considersvariation not to be an error but, as the great evolutionist Emst Mayr put it, to be real.Individual variancein a population is the sourceof diversity on which natural selection acts to produce different kinds of organisms.This contrastsstarkly which requiresa typology createdfrom the top with Platonicessentialism, down; instead,population thinking statesthat evolution producesclasses of living forms from the bottom up by gradual selectiveprocessesover eons of time (seefigure 5-2).No such idea exists in physics-even the "evolution" of stars does not require such a notion for its explanation. Connectingpopulation thinking to ideasof the mind, I have to go a bit and into those of the evolutionary deeperinto some of its consequences processitself. To do so, I will call on some terms that are probably new 73

Pnoposars to the reader,at least in their specializedusage.Among them are "instruction," "selection," "recognition," and "memory" or 'heritability." Notice that, except for the last, all of thesewords are cornmonly used,but for our purposes,I will use them somewhat differently. Let me begin with the specializednotion of recognition. I will make an abstract statementand then translateit into an evolutionary example.By "recognition," I mean the continual adaptive matching or fitting of elements in one physical domain to novelty occurring in elementsof another, more or lessindependentphysical domain a matching that occurswithout prior instruction. This is quite a mouthful; let me see if I can cut it into digestible morselsby using evolution as an example. In evolution, organisms(elementsof domain 1) are more or lessadapted to events in the environment (elementsof domain 2). This adaptation occurseven when environmentalchangescannotbe predicted(that is, even when the changesrepresentnovelties).The processof adaptationoccursby selectionon those organismalvariants that are on the averagefittest, and what makes them fittest does not require pior explicit information ("instruction") about the nafure of the novelties in the environment. The selectiveenvironmental changesare, in generaf independentof variation in the population of organisms,although selection resulting from such changesmay add to that variation. In sum, there is no explicit information transfer between the environment and organismsthat causesthe population to change and increaseits fitness.Evolution works by selectiorynot by instruction. There is no final cause,no teleology, no pulpose guiding the overall process,the responsesof which occur er postfacto in eachcase. This is an astonishingidea.It remindsme of the lady in the E. M. Forster novel who said,"How do I know what I think until I seewhat I say?"Even more astonishingis the fact that evolution, acting by selectionon populations of individuals over long periods of time, gives rise to selective systemswithin individuals.Suchselectivesystemsacting in one lifetime in one body are called somatic selective systems. Thus, an evolutionary selective system selectsfor a somatic selectivesystem! Now I will describea specificexample, the immune system, which is seenonly in backbonedanimals.Grasping the fundamentalsof immunity will be very useful in understandingselectionin the nervous system.This descriptionwill, I hope,justifu my rather specializedand abstractuse of the word "recognition," for the immune systemis the best understoodexample of a somatic recognition system based on selectionistprinciples. Indeed, before I am through, I want to make the casethat there are sciencesof recognition, sciencesthat study recognition systems.The evidenceis abundant that evolution and adaptive immunity are two such systemsacting 74

T h e S c i e n c e so f R e c o g n i t i o n over different populations and over different time scales.This is by way of leading up to the suggestionthat the brain also acts as a somatic selective system and thus that neurobiology is also a scienceof recognition. But I am getting aheadof myself. I want to describebriefly the immune system, the scienceof which I worked in for some ftfteen years, for it is both intriguing and illuminating. The immune systemis a somaticselective system consisting of molecules,cells,and specializedorgans.As a system, it is capable of telling the difference between self and nonself at the molecular level. For example,it is responsiblefor distinguishing between and responding to the chemicalcharacteristicsof viral and bacterialinvaders (nonsel0,invaders that would otherwise overwhelm the collections of cellular systems in an individual organism (selfl. This responseinvolves molecularrecognition with an exquisite degreeof specificity.An appropriately stimulatedimmune system can tell the differencebetween two large foreign protein molecules composed of thousands of carbon atoms that differ by only a few degreesin the tilt of a single carbon chain. It can tell thesemoleculesapart from all other moleculesand retain the ability to do so once it has initially developed that ability. It has a "memory." If I inject a protein into an individual's body that does not resembleits own proteins, specializedcells called lymphocytes respond by producing moleculescalledantibodies(figure 8-1). Thesemoleculesbind by fitting to specificand characteristicportions of the foreign molecule,or antigen, as it is called.A secondand later encounter allows these antibodies to bind even more effectively to just those antigens.Perhapsmore astonishing is the fact that a specificrecognition event occurs even for new molecules synthesizedby organic chemists,moleculesthat never existedbefore either in the responding speciesor in the history of the earth for that matter. How can an individual's body positively distinguish novel moleculesin such a specificfashion?The theory prevailing before the present one was called the theory of instruction. Its basic assumption was that, in the immune system,a foreign moleculetransferredinformation 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 moleculeswith regions having the shapewith which the impression was originally made. It is obvious why this is an instructive process: Information about a three-dimensionalstructure is posited to be necessary to instruct the immune system how to form an antibody protein whose polypeptide chain folds around that structure (seefigure 6-1, bottom) to give the appropriate complementaryshape. This elegant and simple theory tumed out to be false.The theory that 75

PnoPosALs

ANTIBODYMOLECULE ConstantRegions BindingSite for Antigen

Attachmentto Gell

CLONALSELECTION LymphocyteRepertoire

ClonalCell Division FIGURE8-1 The immune system works as a selectioerecognition system. Your immune system distinguishesforeign molecules(nonself)from tlu moleculesof your bodVbelf) by airtue of their diferent shnpes,It doesn by making proteins calledantibodies.Each immune cell makesan antibody with a diferent oariableregion(top ftgure); eachoariableregion has a binding site with a diferent shape.Wen n foreign moleculeor antigen (bottom hgure, black dots)entersthe body, it is boundby just thoseantibodieson the cellsof the immunesystemthat happento fit parts of its shape(cellsSlZ, 201, I0O, and al. This setof cellsthendioidesand malcesa "clone"-more cellsof the samekind with antibodies of the samekind. The next time the antigenis presented,many morecopiesof thesesame antibodiesare thereto help destroyit. Cellsnumbered542, 201, 100,and 42 are now more prevalent,for erample, and-will recognizethe foreign moleculesmore rapidly the nert time they inhude. Theforeign intruderscouldbemoleculeson a oirus or q bacteium,

T h e S c i e n c e so f R e c o g n i t i o n replaced it is more complicated and even belies common sense,but it appearsto explain a wide variety of facts;indeed, few, if any, present-day immunologists would dispute its essential correctness.This theory is known as the theory of clonal selectionand was first proposed by the late Sir Frank MacFarlaneBumet. Bumet maintainedthat, prior to a confrontation with any foreign molecule, an individual's body has the ability to make a huge repertoire of antibody molecules,each with a different shapeat its binding site. When a foreign molecule (say on a virus or bacterium) is introduced into the body, it encountersa populationof cells,each with a differmt antibody on its surface.It binds to those cells in the repertoirehaving antibodieswhose combining sites happen to be more or less complementaryto it. When a portion of an antigen binds to an antibody with a sufficiently close fit, it stimulatesthe cell (called a lymphoryte) bearing that antibody to divide repeatedly (figure 8-1.).This results in many more "progeny" cells having antibodies of the sameshapeand binding specificity. A group of daughter cells is called a clone (the asexualprogeny of a single cell) and the whole processis one of differential reproduction by clonal selection.In other words, as a result of the selection of cells with appropriately specificantibodiesfrom a large repertoireof diversecells,the kinds of antibodies speciffcfor a foreign shape are increasedin number becausethe selective binding event er post facto caused those cells to multiply. The composition of the lymphocyte population hasbeenchanged by selection. An analysisof the complete strucfureof an antibody was carried out in my laboratory severaldecadesago. It showed that the polypeptide chains of an antibody (figure 8-1, top) consist of constant regions (similar or identicalfrom moleculeto molecule)and variableregions (differentfor each kind of moleculeand comprising the binding site.)We now know that this diversity is generatedsomatically(that is, within an individual's lifetime) in the lymphocytes of eachindividual's body. The processinvolves a kind of jumbling within eachlymphoryte of the genetic code specifuingthe variable antibody regions that may someday happen to bind an antigen. I hope I have said enough to show you that the immune system

Thisis a selectiae systembecaose oastnumbers of diferentantibody-binding shapes are present(eachoneon a diferentcell)beforetheantigmsenter.Theseantigercselectonly a fat of theshapes, andantibodyproductionis oastlyamplifielby clonaldioisionof the cells(2,4,8,16.. . .) to enormous numbers,Thus,thepopulationis changed asa resultof erperimce. 77

Pnopos^ers correspondsto my definition of a recognition system. It exists in one physical domain (an individual's body) and responds to novelty arising independently in another domain (a foreign molecule among the millions upon millions of possible chemically different molecules)by a specific binding event and an adaptive cellularresponse.It does this without requiring that information about the shape that needs to be recognized be transferredto the recognizing systemat thetime whm it makestherecogniz.er moleculesor nntibodies.Instead, the recognizing system frsf generatesa diverse population of antibody molecules and then selectser post facto those that fit or match. It does this continually and, for the most part, adaptively. The immune selectivesystemhas someintriguing properties.First, there is more than one way to recognize successfullyany particular shape. Second,no two individuals do it exactly the same way; that is, no two individuals have identical antibodies. Third, the system has a kind of cellular memory. After the presentation of an antigen to a set of lymphocytes that bind it, some will divide only a few times, while the rest go on irreversibly to produce antibody specific for that antigen and die. Becausesomeof the cellshave divided but not dl the way to the antibodymaking end, they constitute a larger group of cells in the total population of cells than were originally present. This larger group can respond at a later time in an acceleratedfashion to the sameantigen. As I mentioned before,the systemthereforeexhibits a form of memory at the cellularlevel. Here is a molecularrecognition system that is noncognitive and highly specific,the explanation of which is a marvelous example of population thinking-the essenceof Darwinism. Like evolutiory it has a generator of diversity (the 'jumbley'' of DNA in each lymphocyte), a means of perpetuating changesby a kind of heredity (clonal division), and a meansof differentially amplifying selectionevents (differentialclonal reproduction). Unlike evolutioru it occurs in cells over short periodsof time and does not produce many levels of form-just different antibody molecules.It is a somatic selectivesystem. Notice that in evolution itself, diversity is generatedwithin a population of animals by mutations in DNA. These are transmitted hereditarily through germ cells (sperm and egg). Selectionthen occurs on individuals continually over eoolutionarytime to give rise to different species,depending on many variables in the environment. Both systems,evolution and immunity, deal with novelty by similar selective principles but by very different mechanisms.It is conceptuallyimportant to distinguisha selective principle from the mechanismsused to expressit in any particularphysical system. 78

T h e S c i e n c e so f R e c o g n i t i o n What these two sciencesof recognition, evolution and immunology, have in common is not found in nonbiologicalsystemssuchas "evolving" stars.Suchphysical systemscan be explainedin terms of energy transfer, dynamics,causes,and even "information transfer."But they do not exhibit repertoiresof variantsready for interactionby selectionto give a population responseaccording to a hereditary principle. The application of a selectiveprinciplein a recognition system, by the way, doesnot necessarily mean that genesmust be involved-it simply meansthat any stateresulting after selectionis highly correlatedin structurewith the one that gave rise to it and that the correlationcontinuesto be propagated.Nor is it the casethat selectioncannot itself introduce variation. But a constancy or "memory" of selectiveeventsis necessary.If changesoccurredso fast that what was selectedcould not emergein the population or was destroyed, a recognitionsystemwould not survive.Physicsproperdoesnot dealwith recognition systems,which are by their nature biological and historical systems.But all the laws of physics neverthelessapply to recognition systems. Leo Szilard,a greatphysicistwhose experimentswith EnricoFermiled to controllednuclearfission,was fascinatedwith both immunology and the brain. He used to visit my laboratory often to see what was new with antibodies.Usually, he would start by saying,"All right, what's the problem?I have fifteen minutes."Once Szilardattendeda meeting at which an unfortunate researcherproposed a theory of memory. Thought, he explained,causednew proteins to be madeby our brains.After a time, these being new, they would stimulate antibodylike proteins that represented memories.Leo rose and said, with a merry and mercilesssmile, "Maybe that's how your brain works." Do brainsconstituteselectiverecognitionsystems?Will describingthe fundamentaloperationsof brainsin suchtermsbe revealingand useful?As you have undoubtedly surmised,I think it will not only be useful,but it will alsoremovemuch of the paradoxand senseof crisisthat one confronts when reviewing the data on brain structureand function. Indeed,I believe that neurobiology is a scienceof recognition.But even though after our antibody work my colleaguesand I were excitedto discoverthat neuralcell adhesionmoleculesor "brain glue" arethe evolutionary precursorsof the whole immune system,I am hardly suggestingthe kind of proposalmade by Szilard'sunhappycorrespondent.The resemblance betweenthe immune and neryous systemsis only in principle,not in detail. I have defendedthe notion that brainsare selectiverecognitionsystems becausethinking about brain function in selectionaltermsrelievesus of the hbrror of the homunculus(figure 8-2). Becausediversity existsbeforehand 79

PnoPosALs

FIGUREF2

Themdlessregression of homunculi.Theideaof inshuctionor informationprocessing or something, to readit. Buta similarentity is thm requiredto read requiressomeone, the resultingmessagea and so on, endlessly. in a selectivesystem,and becausespeciftcityarisesas a result of selection er poslfacto,we are no longer facedwith an endlessregressof information processorsin the head.To defendthis statementadequately,I will describe a theory of brain function that follows selectionalprinciples.The cJrallenge is to show how evolution and development give rise to a somatic selectional system in the brain. Having accomplishedthat, I will attempt to show that the selectionalmechanismsproposedcan accountfor psychological functions-perceptiorl memory, even consciousness. Let us tum to the various parts of this task.

CHAPTER

9

NeuraIDarwinism

If a term has to be usedfor the utholeset of ideasI utould suggest Neural Edelmanism, -Francis H.C.Crick

f we considerrecognition to be a kind of adaptivematching,then it is obvious why it appliesto evolution and immunity. In both instances, population thinking provides a meansfor understanding.What is the justification for applying population thinking to the workings of the brain, for neural Darwinism? This is not the place to go into all the intricaciesof sucha position,but I believeit will clarify much of what I have to say in the rest of this book if I give someof the reasonsfor considering brain sciencea scienceof recognition. The ftrst reason is almost too obvious: Brain scienceand the study of behavior are concemed with the adaptive matching of animals to their environments.In consideringbrain scienceas a scienceof recognition I am implying that recognition is not an instructive process.No direct information transfer occurs,just as none occurs in evolutionary or immune processes.Instead,recognition is selective. Some justifications for this position may be found in my previous criticisms (chapter 3) of various category mistakesin thinking about the brairy an extended argument supporting these criticismsmay be found in the Postscript.I have already argued that the world is not a piece of tape and that the brain is not a computer. If we take such a position, we have 81

Pnoposers to show how a behaving animal neverthelessadaptively matchesits responses to unforeseennovelty occurring in such a world. There is an additional set of reasonsfor assumingthat recognition cannot be instructive. We have already seenthat the individuality and strucfural diversity of brains even within one speciesis confounding to models that consider the brain to be a computer.Evidencefrom developmentalsfudiessuggests that the extraordinary anatomical diversity at the ftnest ramifications of neural networks is an unavoidableconsequenceof the embryological process.That degreeof individual diversity cannot be tolerated in a computer system following instructions. But it is just what is needed in a selective system. A potent additional reason for adopting a selective rather than an instructive viewpoint has to do with the homunculus.You will remember that the homunculusis the little man that one must postulate "at the top of the mind," acting as an interpreter of signals and symbols in any instructive theory of mind. If information from the world is processedby rules in a computerlike brain, his existenceseemsto be obliged. But then another homunculusis required in ftls head and so on, in an infinite regress (seefigure E-2). Selectionalsystems,in whidr matching occursexpostfacto on an already eristing diverse repertoire, need no special creations, no homunculi, and no such regress. If we assumethat brain functions are built according to a selectional process,we must be able to reconcilethe structural and functional variability of the brain with the need to explain how it carriesout categorization. To do so, we need a theory with a number of essentialcharacteristics.It must be in accordwith the facts of evolution and development;accountfor the adaptive nature of responsesto novelty; show how the brain's functions are scaledto those of the body as the body changeswith growth and experience;accountfor the existenceand functions of maps in the brainwhy they fluctuate,how multiple maps lead to integrated responses,and how they lead to generalizationsof perceptual responses,even in the absenceof language.Eventually,sucha theory would also need to account for the emergenceof language itself. And finally, such a theory must accountfor how the various forms of perceptualand conceptualcategorization, of memory, and of consciousnessarose during evolution. To be scientific,the theory must be based on the assumption that all cognition and all consciousexperiencerest solely on processesand orderings occurring in the physical world. The theory must therefore take care to explain how psychological processesare related to physiological ones. The theory I have proposed to account for these matters is known as the theory of neuronal group selection (TNGS). Its basic tenets are de82

Neural Darutinism scribed in this chapter and those of its features that provide a bridge between psychology and physiology are stressed.This will enable us to come to grips with the daunting problem of consciousness, and it is one of the main reasonsfor explaining the theory in any detail. In the course of doing so, I deal with perceptualcategorization,concepts,memory, and leaming. I want to wam the reader that I have to explain a seriesof complex processesthat must be graspedin order to understandbrain function. The main ideasto grasp are calledneuronalgroup selection,reentry, and global mapping. I provide examplesfor each.If they are mastered,they will serve you well, becausewe will use them over and over again in later chapters. In the senseof requiring all these processes,the TNGS is a complex theory; its basic tenets,however, are only three in number. No additional tenets are required to explain even so remarkablea property as consciousness.What is requiredto explain sucha property, however, is the evolution of new kinds of morphology in both the body and the brain. So I will take up some featuresof these moqphologiesas we go along. The three tenets of the TNGS (ftgure FI) are concemedwith how the anatomy of the brain is first set up during development,how pattems of responsesare then selectedfrom this anatomy during experience,and how reentry, a processof signaling between the resulting maps of the brain, gives rise to behaviorally important functions. According to the first tenet, developmentalselection the dynamic primary processesof development discussedin chapters.l and 6 lead to the formation of the neuroanatomy characteristicof a given species.This anatomy obligatorily possessesenorrnousvariation at its finest levels and ramifications. This is becauseof the dynamic regulation of CAMs and SAMs, the stochasticfluctuation of cell movement, cell processextension, and cell death during development,and the activity-dependentmatching of connectionsthat is superimposedon neuralbranches(or neurites)asthey explore a developing brain region. This entire processis a selectionalone, involving populations of neuronsengagedin topobiological competition. A population of variant groups of neuronsin a given brain region, comprising neural networks arising by processesof somatic selection,is known as a primary repertoire.The genetic code does not provide a specific wiring diagram for this repertoire. Rather, it imposes a set of constraintson the selectionalprocess.Even with such constraints,genetically identical individuals are unlikely to have identical wiring, for selectionis epigenetic. The secondtenet of the TNGS provides anothermechanismof selection that, in general,does not involve an alteration of the anatomicalpattem. It assumesthat, during behavior, synaptic connectionsin the anatomy are 83

PnoPosALs

Developmental Selection (Yielding Primary Repertoire)

1

LTY

d

\

oo

/

Cell Division Cell Death Process Extension and Elimination CAM Action

Time 1

Experimental Selection

Changesin Strengthof

t a a

0 t.

(Yielding Secondary Repertoire)

Population of Synapses Time 1

Time 2

Reentrant Mapping

Stimuli to Map2

to Mapl

to Map2 Time 2

FIGURE FI groupselection, it A selectional theoryof brainfunction.Calledthetheoryof neuronal asa resultof themolecTop: Developmental selection.Thisoccurs hasthreetenets. growthfactorsignaling, andselectiae cell ular effects of CAM and SAM regulation, that makeup networksin eachindiuidual,nehporks deathto yieldaaied anatomical or the primary repertoire.Center:Experientialselection.fulectivestrengthening of populations of synapses as a resultof behaoiorleadsto theformationof wealcening of aariouscircuits,a secondaryrepertoireof neuronalgroups.The consequences by dashedpaths. synapticstrengthmingare indicatedby bold paths;of wealcening, snd Bottom:Reentry.Thelinkingof mapsoccursin timethroughparallelselection groups,whichindependently and disjunctiuely the correlation of the maps'neuronal Dotsat the providesa basisfor perceptual categorizntion, receioeinputs.Thisprocess indinteparallelandmoreor lesssimultaneous connections endsof theqctioereciprocal may wish to refreshtheir in reentrantpaths.(Readers of the synapses strengthening (or weakening) cnn by referringto figure 3-2.) Strengthening knotaledge of synapses connections. occurin bothintrinsicand extrinsicreentrant

84

Neural Darwinism selectively strengthenedor weakenedby specific biochemical processes. This mechanism,which underliesmemory and a number of other functions, effectively "carvesout" a variety of functioning circuits(with strengthened synapses)from the anatomicalnetwork by selection.Such a set of variant functional circuits is called a secondaryrepertoire, To some extent, the mechanismsleading to the formation of primary and secondary repertoires are intermixed. This is so becauseat certain times and places the formation of the primary repertoire depends on changing synaptic strengths, as in the activity-dependent matching of connections(for an example, see figure 3-4). Even in a developed brain, "sprouting" can occur, in which new neural processesform additional synapses.And in some cases,such as the development of bird song and frog metamorphosis,the formation of new parts of the nervous system involving simultaneousprimary and secondaryrepertoireformation occurs during behavior in the world. The third tenet of the TNGS is concemed with how the selectional events described in the first two tenets act to connect psychology to physiology. It suggestshow brain maps interact by a processcalled reentry. This is perhapsthe most important of all the proposalsof the theory, for it underlies how the brain areasthat emerge in evolution coordinate with each other to yield new functions. To carryrout such functions, primary and secondary repertoires must form maps.Thesemaps are connectedby massivelyparallel and reciprocal connections.The visual systemof the monkey, for example,hasover thirty different maps, each with a certain degree of functional segregation (for orientation, color, movement, and so forth), and linked to the others by parallel and reciprocalconnections(figure 9-2). Reentrantsignaling occurs along theseconnections.This meansthat, as groups of neuronsare selected in a map, other groups in reentrantly connectedbut different maps may also be selectedat the same time. Correlation and coordination of such selectioneventsare achievedby reentrantsignaling and by the shengthening of interconnectionsbetween the maps within a segment of time. A fundamentalpremise of the TNGS is that the selectivecoordination of the complex pattems of interconnection between neuronal groups by reentry is the basis of behavior. Indeed,reentry (combinedwith memory, which I discusslater) is the main basisfor a bridge betweenphysiology and psychology. I have not yet mentioned what the unit of selectionis for the brain. In evolutiorl the main unit of selection is the individual animal (the phenotype) and in immunity, it is the individual lymphocyte. According to the TNGS, however, the unit of selectionis not the individual nerve cell, but 85

PnoPosALs

FIGURE 9-2 Multiple mapsof oisualareasof the brainarereentrantlyconnected to eachother(see doubleanowslinkingoisualmapsV1-V5, thetemporalareas,and theparietalarms). Eachmapserces in a functionallysegregated manner-for color,motion,oientation,and soforth. No "supmtinr map" eriststhst summaizes"information"on thue properties. But as a resultof reentryfthe doubleanows),the mapsact coherently to respondto combinations of properties. Eoenthe regionknown as the hteral geniculatenucleus (LGN),an infracorticalregionthat receionsignalsfrom the opticneroeof the eye,is reentrantlyconnecteil to thepimary uisualmap,VL

is rather a closely connected collection of cells called a neuronal group (figure *-3). The reasonfor this has to do with limitations on the properties of neurons, with developmental constraints, and with neuroanatomical requirementson reentrantcircuits.Individual neuronsare either excitatory for other neuronsor inhibitory for other neuronsbut not both. In contrast, groups, which consist of mixtures of both kinds of neurons,can be both. During the formation of the primary repertoire,neighboring neurons tend to connect more extensively to each other to form circuits containing varying proportions of eachkind of newon. This lends a richly cooperative property to the activity of neurons in groups, activity that one would expect to be different in different areasand maps becauseof differencesin their primary repertoires. There is an even more compelling reason to suppose that neuronal groups are units of selection.When mapsare connectedby reentrantftbers, 86

Neural Darwinism the individual fibersgenerally extend their arbors over many locally linked neurons(figure F3). When secondaryrepertoiresare formed, the strengthening of synapseswithin thesearborsmay then selectgroups of neighboring neurons,changing borders over smaller dimensionsthan those of the arbors.We may summarizeby saying that, in general,no individual neuron is selectedin isolation; no individual neuron in a map reentersto only one other neuron in another map; and no individual neuron has the properties alone that it shows in a group. Theseconstraintsarisefrom the density of the interconnectionsamong neurons,and they make it highly unlikely that a single neuron could function as the unit of selection. With the three tenets of the theory in hand, we are now ready to see how the ability to carryrout categorization is embodied in the nervous system.I shall use the exampleof perceptualcategorization-the selective discrimination of an object or event from other objects or events for adaptive purposes.Rememberthat this occursnot by classicalcategorization, but rather by disjunctive sampling of properties. (For a detailed discussionof the psychologicalbasesof categorization,seethe section on categoriesin the Postscript.) To explain how categorizationmay ocor, we can use the workings of what I have called a "classificationcouple" in the brain. This is a minimal unit consisting of two functionally different maps made up of neuronal groups and connected by reentry (figure F4). Each map independmtly receivessignalsfrom other brain maps or from the world (in this example, the signals come from the world). Within a certain time period, reentrant signaling strongly connectscertainactive combinationsof neuronalgroups in one map to different combinationsin the other map. This occursthrough the strengthening and the weakening of synapseswithin groups in each map and also at their connectionswith reentrant fibers. In this way, the functions and activities in one map are connectedand correlatedwith those in another map. This occurs even though each map is receiving independent signalsfrom the world, One set of inputs could be, for example,from vision and the other from touch. If the maps in question are topographically connected,they correlate happenings at one spatial location in the world without a higher-order supervisor.(By "topographic," I refer to the situation in which a sensory receptor sheet receiving signals from the world connects to its recipient map in sucha way that neighboring locations in the sensorysheetare also neighboring locationsin the recipientmap.)This connectivity is not limited to a pair of mapsor to any one moment of time: The interactionsof multiple maps can be coordinated in the samefashion. This is a very important property. Neuronal group selectionoccurring 67

PnoPosALs

Group5

Group1

Group4

Group2

V

I

\,/

I

I

EXTRINSICINPUTS

o ct) (r

< ( - - r i r r \ r r r \ t

=

I

I

to'l't'"'lt"ti"'l'

r ,z_. /'..

l / l l l i l l < * --i l r r r * --^*--/,," --4--.t*f*l *-lf--,'l <' .-^r*^-{t rEL.EcrRoDE i ^ l _Firirig l (spikest I i I lN CELL

Cell

Y

EXTRINSICINPUTS

Time of a second) 1r7u

'F"'fi Tr:H,?i"'#l'ft"8

FIGURE9-3 and Neuronalgroups. Top: How neuronsare connectedin groups (intinsic connections) how groups are connectedto eqch other (erhinsic connectiota).Each group shows a differmt aspectof connectittity.Groups 1 and 5 show that eachcell contactscells in its outngroup and in othergroups.Group 2 shoucsthe denseintrinsic connectioityof groups. Group 3 shows that eachgroup also receivesinputs from a set of ooerlappingertrinsic inputs that can beselectioelystimulated.(ln gmeral, such inputs ertendooerdistancesof many cell diameters.)Group 4 showsthat eachcell thereforereceioesinputsfrom celb in its own group, from cellsin othergroups,and from ertrinsic sources.Groupsdiffer in size 88

Neural Darwinism within mapsleadsto the production of new kinds of signals,which can then be reenteredinto earlier maps along with signalsfrom the outside world. This property of reentry allows for what I have called recursivesynthesis: Not only are events correlated topographically across different maps without a supervisor,but new selectionalproperties emerge through successive and recursive reentry across maps in time. This property has been simulated in a computer model, the RCI (reentrant cortical integration) model of the cerebral cortex, which I describedin detail in The Remembered Presml. This model successfullycorrelates the activities of many different maps by reentry and produces coordinated responsesto complex visual figures. How canreentry accountfor perceptualcategorization,the function that the TNGS takes to be fundamentalin any attempt to relate physiology to psychology?The brief answeris: By coupling the outputs of multiple maps that are reentrantly connectedto the sensorimotorbehavior of the animal. This is achievedthrough a higher-order structure called a global mapping. A global mapping is a dynamic structurecontaining multiple reentrantlocal maps (both motor and sensory)that are able to interact with nonmapped parts of the brain (figure F5). (Thesenonmappedparts of the brain include parts of specialized structures known as the hippocampus, the basal ganglia, and the cerebellum,the functions of which will be discussedlater.) I want to stress here that a global mapping allows selectional events occurring in its local maps (the kind illustrated in hgure 9-4) to be connected to the animal's motor behavior, to new sensory samplingsof the world, and to further successivereentry events. Such a global mapping ensuresthe creation of a dynamic loop that continually matchesan animal's gesturesand posfure to the independent sampling of severalkinds of sensorysignals.Selectionof neuronal groups within the local mapsof a global mapping then resultsin particularcategorical responses.Categorizationdoes not occur according to a computerlike program in a sensory area which then executes a program to give a

(r.onSrrg and in actualconnectioity, from perhaps50 to 10,000neurons) which is determined by the localneuroanatomy of theareasin whichtheyarefound. Bottom: Eoidence of groups,An electrode in a oisuslneuronrecordsitselechical for theeristence (spike!as well as theresponses response (fieldpotentials)in thegroup. of its neighbors when a oisualstimulusof theright type(a lit barmooingupandto thengfu)is presmt, t-heresponses of the newonand its neighbors all oscilliteat thesamefriqumcv fforfu hertz,or-forty, cyclesper second). when the stimulusis remooed, the spikesi"i f"k potentialsno longerconelate. 89

PnoPosALs

CorrelatedFeatures

Features

REENTRY

FIGURE9-4

inputs(I andz). Eachmap independent Reentry.Two mapsof neuronalgroupsreceiae to local that is, map 1 responds isfunctionallysegregated; features(forerample,oisually (for erample,an angles)that arediferentfrom thoseto whichmap2 responds detected by neroefibersthat carry The two mnpsare connected object'soverallmooement). anddense andserveto "map them.These signalsbetween reentrant fibersarenumerous groups indicatedby thecircles themaps"to eachother.lf within sometimeperiodthe groups the in map2, these to the indicated by squares reentrantly connected in tnap1 are ,Asa resultof reentrantsignaling,and W meansof may bestrengthened. connections with patternsof responses patternsof responses in map1 areassociated synapticchange, to prewnt couple."Because of synapticchange,responses in map 2 in a "classification inputsare alsolinkedto preaiouspattemsof responses. particular motor output. Instead, sensorimotor activity over the whole mapping selectsneuronalgroups that give the appropriateoutput or behavior, resulting in categorization.Decisionsin such systemsare basedon the statisticsof signal correlations.Notice the contrast with computers;these changesoccur within a selectionalsystem rather than depending on the carriage of coded messagesin a processof instruction. But what is "appropriate" with respect to behavior, and how does perceptualcategorizationmanifestitselft The TNGS proposesthat categorization always occursin referenceto intemal criteria of value and that this referencedefinesits appropriateness.Suchvalue criteria do not determine specific categorizations but they constrain the domains in which they occur.According to the theory, the basesfor value systemsin the animals of a given speciesare already set by evolutionary selection. They are exhibited in those regions of the brain concemed with regulating bodily functions:heartbeat,breathing, sexualresponses,feeding responses,endocrine functions, autonomic responses.Categorization manifests itself in 90

Neural Darutinism

Motor Gortex

MultipleReentrant SecondaryMaps

Modified Output

Primary LocalMap ---=l

Sensorv Sheets

I

t--.;--^^.-^

Muscles andJoints i

Muscles

MOVEMENT

SENSORY

SAMPLINGs"rnpf,lt3'#onl'o?Lr"n* FIGUREF5 A global mapping. This structureis madeup of multiple maps(of the kind shown in the preoiousfigure).The mapsare also mnnectedto brain rcgionssuchas the hippocampus and cerebellum.Noticethat signalsfrom the outsideworld enter into this napping, and signals of outputleadto movement.Thk in turn altershow sercory* that multiplesources with time arepickedup. A global mappingis thusa dynamicstructure,onethat changes and behaoior.Its reentrantlocal maps,which conelatet'eaturesand moaement,make perceptualcategorizationpossible.Perturbationsat diferent leoelscausea global mapping to rearrange,to collapse,or to be replacedby anotherglobal mapping.

behavior that appropriately fulfills the evolutionarily selectedrequirements of such life-supporting physiological systems. A specific example of categorizationconstrained by value may help connecttheseideas.My colleaguesand I have simulatedcomplex automata based on the TNGS in supercomputersto demonstrate that perceptual categorizationcan be carriedout on value in a global mapping (figure 9-6). In automata such as Darwin III, value is seen to operate for the visual system,for example,in circuits that favor light falling on the central part of the eye. (Value : "light is better than no light"; light and stimulation 9I

PnoPosALs

Environment

System Categorization

@Eye BeforeTraining

tr .9 864

to *

AfterTraining

o o c o CL o o E CD

r o+ ro ro

t *

il

*o

tt ol E tr t* a-

4

(g

tr loll

/ *1rl**

llrrrl/il* olrcl

0/8

0' X- Position

*H * t ** tr*I ^t ililil

Numberof Trials

8t8

FIGURE9_6 Darwin III, a recognitionautomatonthat performsas a global mapping,This automaton lt hasa singlemooableeqe,a four-jointedarm utith was simulatedin a supercomputer, (joint sense)signaledby neuronsin its joints as touchat the last joint, and kinesthesia eachresplnsiblefor they moue,Its nerooussystemis organizedinto seoeralsubsystems, of its behaoior(top). It containsmany mapsof thekindsshoutnin figures differentaspects g-4 and %5, What is programmed in the simulationis the "euolutionary"phenotype, includingneuroanatomy.But the behaoiorof the simulation is not programmed(see 92

Neural Darwinism at the center of vision are favored over light and stimulation at the periphery.) In Darwin III, the action of these value circuits enhancesthe probability that synapsesactive when suc} circuits are engaged will be strengthenedin preferenceto competing synapses.The net result is that with selectionand experiencethe eye of the automaton trackssignalsfrom lit objects. This defines one form of "appropriate" behavior as acquiredbehavior that is consistentwith evolutionarily set values.But percepfualcategorization occurs only when, after dis;unctive sampling of signals in seoeral modalities (vision, touch, joint sense),Darwin III activates an output through its reentrant maps. This occurs,for example,when as a result of explorations with its "hand-arm" and "eye" it "decides" that something is an object, that the object is striped, and that the object is bumpy. Given that Darwin III has a higher-order value system for output on such a categoricaldecision,it then activatesa neural circuit that flails its arm. This output reflectsthe categorizationthat results from multiple synaptic selectional events that have occurred as a result of experiencein all its reentrantly connected maps. These selectionalevents occur during ongoing behavior and are nof specifiedby value. As shown in ftgure F6, Darwin III "decides" by its flailing responseamong a large set of objects, distinguishing those that are striped and bumpy from those that are striped or bumpy but not both. Like an animal,it does this in an individually variant manner,not by predefinedclassicalcriteria. That is, it categorizesonly on the basisof experience,not on the basisof prior programming. Darwin III is a model of a global mapping that caniesout categoizntionon oaluc in a fashionthat might be calledembodied.lngeneral,a global mapping such as Darwin III is the smallestorganizationcapableof categorization.Of course, the brain of a real animal has the capacity to assemblemany more mappings of this kind. Incidently, the movements of the "am" of Darwin III are also selected chapter19).After erperience with randomlymooingobjects that it "sees,"its eyewill out to "touch"an object,and with each follow any object.Similarly,its arm reaches selection of mooernents it increases its success in achieaing this touching(lower left). In theerprimmts shownat thelowerleft, the tip of thearm alwaysstartsin a standard location(thepointof originof thetraces).lts motiontowarda targetarea(thesquarebox) is plotted.Noticethat beforetraining,thearm movesin manydirections. After training (bottomset of traces),its mooements inooloingselection aretargeted, Darwin Ill was confronted with fifty-fioedifermt objects andwasgiom eighttials in whichto categoize eachofiect. Theresults(lower right), plottedas a positioeflailing response oersusthe numberof successful tiab, showthat Daruin III dioidedthiscollectionof obiectsinto two classes. 93

Pnoposers rather than instructed. For example, by defining value as a factor that increasesthe likelihood of the strengthening of synaptic connections in those movement repertoires that are active when the automaton's arm moves to the center of the visual field, one obtains a system in which originally undirected movements eventually yield to those that successfully target an object (figure F6). According to the TNGS, the driving forces of animal behavior are thus evolutionarily selectedvalue pattems that help the brain and the body maintain the conditions necessaryto continue life. Thesesystemsare called homeostats.It is the coupling of motion and sensorysampling resulting in behavior that changesthe levels of homeostats.Aside from those occabehavior pattems that have been selectedfor directly sionalspecies-specific by evolution, however, most categorization leading to behavior that changes homeostatic levels occurs by a somatic selection of neuronal groups in each animal. Categorization is not the sameas value, but rather occurs on value.lt is an epigenetic developmental event, and no amount of value-basedcircuitry leads to its occurrencewithout experiential selection of neuronalgroups. But it is also true that without prior value, somatic selectionalsystemswill not convergeinto deftnite behaviors.This hasbeen shown, for example,by cutting off the value circuits in Darwin III: Not surprisingly, the convergent behavior shown in the figures does not occur if this is done. This hasbeena strenuoustour. I hope neverthelessthat I have conveyed something of the flavor and self-consistencyof the TNGS. Now I want to say somethingabout the evidencethat hasbeen gatheredin support of the theory sinceit was first proposed.I shall not be exhaustive,but becausethe theory has been occasionallyattacked or misunderstood,it may be useful to clarify someissuesand mention someexperimentalfindings that corroborate it. The two conceptsof the theory to have come under the most intense attackarethoseof neuronalgroupsandof selectionitself.HoraceBarlow and, separately,FrancisCricl have attackedthe notion of the existenceof groups. Barlow'scriticismis basedon the claimthat neuronalgroup formation would require a Malthusian population dynamic. Thomas Malthus was an inspiration for Darwin, who recountsthat when he read Malthus's suggestionthat populations grow geometrically while food suppliesgrow only arithmetically, he saw how competitive struggle could lead to natural selection.But Barlow'sthinking is not asclearasthat of his distinguishedforebear.Neither natural selection in evolution nor neuronal group selection necessarily requiresa growing population. What is required is difrerentialreproduction (for evolution) and differentialamplification(for neuronalgroup selectionby 94

Neural Darwinism changesin synapticstrength).Barlow compoundshis error by assumingthe wrong mechanismof synaptic change and concluding on this basis that, during selection,neuronal groups would have to incolporate increasing numbersof cellsor elselosetheir selectivity.Explicit modelshaveshown that neither of theseconsequences results,provided one choosesrulesof synaptic changethat more closely resemblethose found in real experiments. Crick's claim is that neuronal groups have little evidence to support them. He also assertsthat neuronal group selection is not necessaryto support ideas of global mapping. Finally, he claims that he has not found it possibleto makea worthwhile comparisonbetween the theory of natural selection and what happensin the developing brain. Contrary to these claims, experimental findings have emerged since the TNGS was first proposed that directly demonstrate the existenceof neuronal groups and the functions of reentry. One of the main ftndings concems the behavior of neurons in the "orientation columns" of the primary visual cortex (see ftgures 9-2 and 9-3). When moving bars of light are presented to an animal's visual system, particular neurons in these columns are known to fire in responseto bars that are oriented in a particular direction. Different neurons have different "orientation tuning" or speciftcresponsesto bars moving at different angles.Wolf Singer and his colleagues,and ReinhardEckhom and his colleagues,have shown that the probability of firing of a single cell in this part of the visual cortex is very closely correlated with the firing of neighboring cells in the same column. The activity of these neighboring cells was measured by recording local fteld potentials, which representthe summed activity of many cells in a small area.These experimentsshowed that the presentation of an appropriately oriented bar causeda group of cells within the responding orientation column to fire together in time, with a predominant oscillatory component at forty hertz (forty cycles per second; see figure F3). ltVhen the stimulus was removed, the coherent oscillatory response of the group of cells disappeared.fust as strikingly, when a stimulus bar was presented and field potentials were recorded in two separatevisual maps, Vl and V2 (see hgure 9-2), the groups of neurons in the two maps showed a mutual phase-coherentoscillation at about forty hertz. That is, despite the distance between them, groups of neurons in separatemaps oscillated at the same frequency and phase when the bar of light was presented.The distant neurons in these groups are known to be linked by reentrant fibers. Thus, the phasic reentrant signaling proposed in the original TNGS appears to be confirmed. My colleaguesand I modeled these findings in a computer. We found that cutting even one leg of the reentrant pathway between the two areas

9s

Pnoposars served to make the oscillation of simulated neuronal grouPs in the two areasgo out of phase and become incoherent. These findings corroborate: (1) the notion of a group---cooperatively interactive neurons more or less tightly coupled by synaptic connections, firing together and responding as units to selectionby particular stimuli; and (z) the concept of the correlation by reentry of selective events occurring in different maps.Both ideashave been shown to apply experimentally to secondaryrepertoireslike those of the visual system. We have also successfullymodeled the plastic changesthat occur in the map boundariesof another secondaryrepertoire,one located in the part of the cortex subserving touch. These changeswere discovered experimentally by Michael Merzenich and his colleagues(seechapter 3, particularly the bottom part of figure 3-5). The notion of neuronalgroups undergoing competition for interaction with neighboring neuronsby strengtheningtheir synaptic connections has nicely explained this plasticity. Changing the pattems of light touch or cutting the nervesthat mediatefinger touch leads to rapid changesof map boundariesin the somatosensorycerebralcortex subserving these functions. The findings are entirely compatible with a Darwinian notion of selection among groups competing within a map. It seemsthat the criticismsleveled at the notion of neuronal groups do not stand up. Criticisms of the theory at the level of primary repertoire formation have also revealeddeep misunderstanding.Dale Purves'sinterpretation that the theory is "regressive" becauseit suggestsselection of nerve cells solely by elimination during development is simply a misrepresentation.The description of primary repertoire formation in Neural Darwinism explicitly states that eliminative selection is insufficient. While eliminative selectionundoubtedly occursduring the formation of the nervous system, it is only one of many selectionalmechanisms.Others of equal or greater importanceare the formation of new anatomicalpaths by the expression of new adhesion molecules and the formation of signal loops by activity-dependent synapseformation. Crick s position (seehis quote at the beginning of this chapter)that the theory should not be calledneuralDarwinism but rather neuralEdelmanism becauseit bears no relation to Darwin's work is simultaneouslyderisive and flattering. But it is also misplaced.As pointed out by Richard Michod and also by myself, the theory has deftnite parallelsto Darwinian notions. In other words, it employs population thinking quite stringently. Corresponding to the idea of fitness, for example, a neuronal group has a likelihood of responseto an input and this likelihood hasto do with variant The connectivity of variant groups directly affects structuralcharacteristics. this likelihood. Furthermore,there is a relationship between the idea of 96

Neural Darwinism heritability and neuronal group selection.In a selectivesystem,there must be some correlation higher than background noise between parent and offspring entities. In evolution this is assuredby inheritance,and in the TNGS by synaptic change. Neuronal grouPs that respond initially to a stimulushave,on the average,a higher likelihood of respondingto a similar stimulus when it is subsequentlypresented,but that likelihood is modulated by value systems.In evolution, differencesamong various organisms' adaptations to the environment lead to differencesamong reproductive processes,which lead in tum to changesin the frequenciesof genesin the population. In neuronalgroup selection,differencesin connectivity, synaptic structure, and the morphology of neurons in the primary repertoire, after confrontation with different correlated pattems of signals from the environment, lead to differencesin the probabilities of their resPonsesas groups. This reflects cJrangesin the pattems of their synaptic strengths. There is differential reproduction in one case,differential amplification in the other. Crick's misprision is probably basedon his faulty notion that the elements in a repertoire cannot vary as a result of selection and that therefore they must be absolutely fixed entities. This is true neither for natural selection nor for neuronal grouP selection. It is crucial to recognize that while the pinciples of the sciencesof recognition (evolution, immunity, and brain science)are shared,their mechanismsmust obviously be different. What is stunning about these sciences during evolution produced two completely different is that naturalselection somaticselectionsystemscapableof recognition. If we acceptthese ideas,a small loop consisting of the events of neuronal grouP selection leads to diverse phenofypic behaviors in different individuals of a species.These diverse behaviors provide the basis for ongoing natural selection in the grand loop of evolution. The two selectivesystems,somaticand evolutionary, interact. Everything in scientificinquiry should be exposed to remorselesscriticism.What is curiousabout the criticismsof the TNGS is the level at which they have been aimed.One would have expectedthat most criticism would have been aimed at the attempt to bridge psychology and physiologythat is, at the proposed mechanismsof perceptual categorization and memory. These mechanisms,along with the proposalsfor concept formation discussedlater, are at the true heart of the theory, and they remain to be tested. As matters stand now, however, neither the experimentalfindings on which the TNGS is basednor the actual proposals of the theory itself have been displaced.Indeed, severalpredictions of the theory have alreadybeen confirmed.It would be enormously valuableif either the facts I have presentedwere shown by scientiststo be false or if an altemative 97

Pnoros.trs theory basedon themwereforthcoming.We couldthen look forwardto more constructivecriticismsand to developmentsthat might further sharpenour vision of brainfunction. So far this hasnot beenthe caseand accordinglyI now tum to those partsof the theory concemedwith bridgingthe gap betweenphysiology and psychology.It is acrossthat bridge that a biological accountof consciousness must pass.

98

1O

CHAPTER

Building MemoryandConcepts: a Bridg.to Consciousness

'Concept' is a nagueconcept,

-Ludwig

'Nhat is an ideaT It is an imagethat paints itself in my brain,

Wittgenstein

-Voltaire

his is a good point at which to take stock and to look ahead.We have been trying to seehow mental functions are embodied,how psychology maps onto physiology. We have argued that natural selectionhas given rise to somatic selectionalsystems-the immune systemand the brain.The major basisof brain function is morphology. The appropriate neuroanatomydevelops accordingto topobiologicalprinciples.Indeed,the brain is a topobiological consistingas it doesof mapsand mappingsystems systempar excellence, in which place is critical for performance. Two apparentlyunrelatedobservationshave compelledus to take a new look at how the brain might function as a recognition system.The first is the enorrnousdiversity and individuality of brain structure.The secondis that the world, although constrainedbV the laws of physics,is an unlabeled place.To considerhow a brain so constitutedmight categorizesuch a world, the TNGS was formulated. Its tenets-developmental selectioil,

P n o p os t r s experiential selection,and reentry-are consideredto be the fundamental grounds for developing psychological functions. This does not mean, however, that new morphological anangements are not necessaryfor emergmtpsychologicalfunctions.It just meansthat the TNGS assumesthat no additional major principleshave to be added to assurethe evolution of new functions. I will argue here that somatic selection acting in global mappings, with new kinds of mappings added to old ones during evolution, is a powerful means of acquiring new functions such as specialized memories and conceptualabilities. Before tuming to a considerationof memory and conceptsthemselves, it may be useful to look at how we expect 'higher brain functions" to be related.Vfhat psychologicalfunctions should this selectionistview explain? And how can they account for consciousnessand intentionality? The fundamentaltriad of higher brain functions is composedof perceptual categorization,memory, and leaming. (While thesefunctions are often heated separatelyfor the convenienceof discussion,it should be kept in mind that, in fact, they are inseparable aspects of a common mental performance.)We have already seenhow classificationcouplesand global mapping can carry out perceptualcategorization.Perceptualcategorization is generally necessaryfor memory, which is, after all, about previous categorizations.The functioning of both may be tested by analyzing behavior.We will seein this chapterthat any kind of memory, while based on changesin synaptic strength, is a dynamic systemproperty, one whose characteristicsdependon the actualneural structuresin which it occurs.To servethe adaptive needsof an animal facedwith the unforeseenjuxtapositions of events affecting survival, however, leaming that affectsbehavior is also necessary.Thus, the three fundamental functions
M e m o r y a n d C o n c e p t s :B u i l d i n g a B r i d g e t o C o n s c i o u s n e s s Leaming in any speciesresults from the operation of neural linkages between global mappingsand the value centersmentioned above.It serves to connectcategorizationto behaviorshaving adaptive value under conditions of expectancy.Physiologicalsystems,like somecontrol devices,have set points (think of a thermostat).What is meant by expectancyis simply the condition under which the set points of the physiological structures making up portions of the hedonic system are not yet satisfied.Leaming is achievedwhen behavior leads to synaptic changesin global mappings that satisfy the set points. We can now seewhy the operation of memory, as it relatesperceptual categorizationto leaming, strongly underlinesthe adaptive value of neuronal group selection.Increasingthe size of the primary repertoires or the reentrant connectivity between repertoires, or enhancing the means of synaptic change by adding new c}emical mechanismsduring evolution, increasesthe number of categoricalresponsesthat may enhanceleaming. Leaming is adaptive, and by this reasoning, having more nutnerous or more diversified neuronal groups would also be adaptive. Nevertheless, whatever the degree of leaming, behavior is constrainedby ethological factors, among the most important of which are the value systems and homeostatic requirementsselectedfor during the evolution of a species. Memory is at the centerof all theseevents,and I shall spenda good part of the rest of this chapter analyzing its wo*ings and requirements.As important as the basic triad of perception,memory, and leaming is, however, their functioning together cannot generate the kinds of capabilities that connectperceptualcategorizationstogether to yield generalrelational properties. These properties emerge from the acquisition of conceptual capabilities-the ability to categorizein terms of general or abstractrelations. So I shall also have to discussthe subjectof concepts.My goal is to show how, with the physiological bases for our central triad and for conceptual capabilities in place, we can account for the emergence of consciousnesswithout invoking any new principlesbeyond those already contained in the TNGS.

MEMORY Let us begin with memory. One difficulty in dealing with memory is that so many different kinds have been described,and so many of them are so closelyrelatedto linguisticcapabilitythat it becomesdifficult to teaseout 101

Pnoposlrs the fundamentals.As a result, a physiological basisfor memory-synaptic change-is often mistakenly equatedwith memory itself in an attempt to simplify matters. To clarify the issue,let us agree that, whatever form it takes,memory is the ability to repeat a performance.The kind of performancedependson the structure of the system in which the memory is manifest,for memory is a systemproperty. As such,memory in the nervous systemis a dynamic property of populations of neuronal groups. In computers,memory depends on the specificationand storage of bits of coded information. According to the TNGS, this is not the case in the nervous system. For example,while the behavior of Darwin III obviously shows signs of a form of memory after its encounterswith various objects (seethe last chapter, especiallyfigure 9-6), its memory cannot be describedin terms of a static configuration of bits. The memory exhibited by a global mapping such as Darwin III is not a store of fixed or coded attributes that can be called up and assembledin a replicative fashion, as is the casein a computer. The TNGS proposes instead that memory is the specificenhancement of a previously establishedability to categorize(figure 1O-1).This kind of memory emerges as a population property from continual dynamic drangesin the synapticpopulationswithin global mappings--changesthat allow a categorizationto occur in the fust place.Alterations in the synaptic strengths of groups in a global mapping provide the biochemicalbasis of memory. In such a system, recall is not stereotypic. Under the influence of continually changingcontexts,it changes,as the structureand dynamicsof the neural populations involved in the original categoization also change. Recall involves the activation of some, but not necessarilyall, of the previously facilitated portions of global mappings. It can result in a categorization responsesimilar to a previous one, but at different times the elementscontributing to that responseare different,and in generalthey are likely to have been altered by ongoing behavior. Since perceptual categoriesare not immutable and are altered by the ongoing behavior of the animal, memory, in this view, results from a processof continual recategorization. By its nafure, memory is procedural and involves continual motor activity and repeatedrehearsalin different contexts.Becauseof the new associationsarising in thesecontexts,because of changing inputs and stimuli, and becausedifferent combinations of neuronal groups can give rise to a similar output, a given categorical responsein memory may be achievedin several ways. Unlike computerbased memory, brain-basedmemory is inexact, but it is also capableof great degreesof generalization. 702

M e m o r 7 a n d C o n c e p t s :B u i l d i n g a B r i d g t t o C o n s c i o u s n e s s Stored Item

Ad d rero Label

OLDVIEW

0 0 0 1| 0 1 0 0 1 1 1 0 1 0 1 0 0 1 0 1

Repllcatlve

0 0 1 0 | 0 0 1 0 0 1 1 1 0 1 0 11 0 1 i

0 0 1 1 11 0 1 1 0 1 1 1 0 0 0 0 1 1 1 0 l l 0 1 0 0 10 1 0 1 0 0 1 1 1 0 1 1 1 0 1 0 l \ ,r

rt-I

rl

/l@-'o'oruo'ffioil I OUTPUT

INPUT

0 1 0 0 | 0 1 0 1 00 1 1 1 0 1 1 1 0 1 0

NEWVIEW

OUTPUT

I

stiltLAR OUTPUT

Dynamlc R e e n t r e n tl l a p r

R e e n t r a n tl l a p r

=ffi-

FIGURE 1T1 Two oiewsof memory,Top: An erampleof memoryas thestorageof preciselycoded information(replicatioe.memory). I call it replicatioebecause recallmusi reproiucethe samecodedpatternwithoutenor and thusreplicateit faithfully as in a computer.one is sn enor.Bottom:An erampieo/ dynamicmimoryin a 4ol7g i" a bit anywhere global mappingof the kind illustratedby Darwin lll (figure9-6) after it caniesout categorization onoalue.Many similarlycategorized objects cangioethesameoutput,and mistakes canbemade,Thismemoryis a property of the eniire system,attlnugh its i2 changein syniptii strength,as indicatedby changes fundammtalmechanism ln the Iinesbetwemtheneuronalgroups(small ciircles) iniide themaps.

P no r o s , c r s The properties of association,inexactness,and generalizationall derive from the fact that perceptualcategorization,which is one of the initial bases of memory, is probabilistic in nature. It is no sulprise that different individuals have such different memories and that they use them in such different fashions. Writing about theseproperties,I am reminded of the contrast between different gifted individuals in their approachto memory and performance. A story told about Fritz Kreisler,the great violinist, and SergeiRachmaninoff, the great pianist, provides a casein point. In 1930, they met in Berlin to record the Grieg C Minor Sonatatogether. Rachmaninoffwas a meticulous worker and wanted to practiceright away. Kreisler,who didn't practice much, was not so assiduousand went out on the town. The next moming, at Kreisley'sinsistenceand with Rachmaninoffs reluctance the recording took place; it went well. (lt is still available,I believe, and is stunning.) Nonetheless,Rachmaninoffwas not pleased. Somewhat later that year, the story goes, the two played together in New York and the program included the samesonata.Somewherein the course of a movement, Kreisler had a memory lapse. Being Kreisler, he simply made up some cadences,probably in the hope of picking up the thread later. When that didn't happenafter a minute or so, he leanedover, still playing, and asked,"Sergei,where are we?" Rachmaninofflooked up from the keyboard and said, "Camegie Hall." If one considersmemory to be a form of recategorization,it is obvious that one can only understandits workings by consideringthe entire system in which it operates.(Refer back to Darwin III in the last chapter for an example.) One of the dynamic characteristicsof the system of global mappings in the brain is the ability to order successivechanges.Memory would be uselessif it could not in some way take accountof the temporal successionof events-of sensoryevents as well as pattems of movement. To seehow all this works would immerseus in a seaof technicaldetails. But it is valuableto understanda bit about the meansby which the cerebral cortex and its appendagesdeal with the time and spacerequirementsof memory. Rememberthat the cortex is an interconnectedsixJayered sheet of about ten billion neurons with about a million billion connections. Besidesbeing arrangedin functionally segregatedmapsthat are reentrantly connectedand that subserveall the different sensorymodalitiesand motor responses,the cortex is connected to three structures I have called the organsof succession,obviously becausethey have to do with ordering the output of the brain. Each of these structures-the cerebellum,the hippocampus,and the basal ganglia-is concemed with a different aspectand scaleof ordering IO4

M e m o r y a n d C o n c e p t s :B u i l d i n g a B r i d g e t o C o n s c i o u s n e s s (figure 1o-2). The cerebellum is a remarkable structure surrounding the upper brain stem. It consists of a distinctive set of neural circuits with a rather stereotyped structure, and receivestwo main kinds of inputs from the cerebralcortex and spinalcord. A variety of sfudiessuggeststhat while the cerebellumis not absolutely requiredfor the initiation of movement, it plays a very specificrole in the timing and smoothing of successionsof movements. Together with the cerebral cortex, it provides the basis for producing and categorizing smooth gestures.In the absenceof portions of the cerebellum,otherwise smooth pattems of movement becomejerky and discoordinated. But what has this to do with memory? Rememberthat categorization dependson smooth gesturesand postures as much as it does on sensory sheets.The cerebellumand motor cortex together undergo the synaptic changesyielding the smooth movementsthat underlieboth categorization and recategorization. The longer-range execution of a sequenceof motor events, called a motor program, dependson another set of cortical appendages,the basal

.-- BasalGanglia

Hippocampus.

Gerebellum

CORTICAL APPENDAGES FIGURE1O-2 Cortical appendages-the organsof succession. The brain containsstructuressuchas the cerebellum,the basal ganglia, and the hippocampusthat are concernedwith timing, successionin mooement,and the estqblishmentof memory. They are closelyconnected with the cerebralcorter as it carries out categoization and conelation of the kind performedby global mappings(seefigure 9-5). The diagram is simply to help the reader Iocatethese"organsof succession"in a cartoonof the brain. 105

Pnopos.lrs ganglia.Theseare a large and complex set of strucfureslocated deep at the center of the brain. They connect to the cerebral cortex in a series of parallel circuits involved in eye and body movements, as well as to the frontal portions of the cortex, the function of whicjr is related to behavioral planning and emotions. It appearsthat the basal ganglia are involved in planning for movement and thus in choicesof the types and successions of motor output. They not only help regulate movement in a motor program by coupling sensory and motor responsesbut also help direct what is to be done according to a motor plan. Notice that, unlike the cerebellum,which smooths and coordinatesmore immediategestures,this appendageworks over longer time scalesand helps correlate whole sequencesof gesturesin a plan. The central role of the basalganglia in motion and motor plans is seen in disease.In Parkinson'sdisease,for example,destruction of a particular set of basalganglia (substantianigra) neuronsthat produce the neurotransmitter dopamineleadsto difficulty in initiating motions, to tremors,and to alteredgait. The basalganglia are also intimately connectedto the hedonic centers of the brairu and as I discusslatea they very likely play a role in attention. A third important cortical appendage the hippocampus,also has intimate connectionswith the hedonic centersfound in the midbrain and the hypothalamus.Its main characteristicis its important role in relating shortterm memory to the establishmentof long-term memory. It sits at the inner edge of the skirt of the temporal cortex, a sausage-shaped structure with a double-nested,C-shapedcrosssection (its appearanceprompted its Latin name, which means "sea horse"). what is exhaordinary about the hippocampusis that it receivesinputs from practically all regions of the cerebralcortex, through a smallerregion known as the entorhinal cortex (figure i.o-z).These inputs run through the hippocampusin a sequenceof three successivesynapses.Having passed through thesestructwes,the signalsloop back to the entorhinal cortex and are relayed back by reentrant fibers to the cortical areas that originally connected with it. Cells inside the hippocampal loop all receive simultaneous connectionsindirectly from the midbrain and hedonic areas,areas subserving value. What is all this circuitry for? It appearsthat the hippocampusis necessary for the laying down of long-term memories. The famous patient H. M., whose hippocampus was removed becauseof potentially lethal epilepsy, had a clear-cutsyndrome after surgery. He could recall all longterm memoriesup to a time somewhatbefore the removal of his hippocampus. But he was unable to rememberevents that had occurredjust a short 706

M e m o r y a n d C o n c e p t s :B u i l d i n g a B r i d g e t o C o n s c i o u s n e s s time before he was asked about them. If intemrpted in the recital of his address,for example,he lost the capacityto recallit. He was fully conscious and could even leam some motor seguences.He was, however, perrnanently crippled and could not store any new ongoing long-term memories becauseof his hippocampaldefect. Studies with animals have confirmed the role of the hippocampus in transforming responsesto short-term tasks for storage into long-term memory. It appearsthat the role of this cortical appendageis to help order events that have been immediately categorized by the cortex and then ensurethat thesecategorizedevents effect further synaptic changesin the cortex to enable long-term memory. What all this means in terms of the TNGS is that while classification couplesand global mappings undergo neuronal group selectionby synaptic change,this in itself is not enough to assurethe relationship between short-term and long-term memory. Unlessorgans of successionsuchas the hippocampusintervene and order the results,it appearsthat severememory defectswill ensue. I have had to burden this discussionwith some of the anatomicaland physiological details to emphasizethe interactive role of different brain structuresand brain dynamicsin carrying out psychologicalfunctions.Had the strucfuresand neuroanatomy of the cerebellum,or something like it, not evolved, smoothly coordinated and rapid motion would be compromised.Without the basalganglia and their specificanatomy, animalswould not be able to orchestrate whole symphonies of movements in a plan. Without the functions provided by the hippocampus, whole suites of categorization in a time range between the immediate and those forever stored could not be linked. And without that linkage, no long-terrn memory could be coherent. In theseexamples,we seethat evolutionarily developedbrain morphology and circuitry, modulated by biochemistry at the synapses,can yield new functions and new kinds of memory. It is clear from studies of invertebrates and "lower" vertebrates that the nervous system acts to regulate bodily functions and behavior. Particulartypes of memory based on synaptic change certainly occur in such animals.But in tenestrial life (and for the precursors of hominids, in arboreal life), great and novel environmentalchallengesoccurred.The further evolution of the cortex for carrying out perceptualcategorizationsand of the organs of successionfor ordering thesecategorizationsallowed for muci richer setsof psychological functions with which to deal with complex environments.Thesedevelopments altered the meaning of what it is to have a memory. Notice, however, that no new principles beyond selectionand reentry are necesr07

PnoPosALs sary to gain new memory functions.What is neededare new structuresor moqphologies-new orderings of connectionsin the brain such as those seenin the cortical appendages.

CONCEPTS The sameprinciple appliesto categorizationitself. Indeed,the embodiment of broad categorical capabilitiesrequires another evolutionary development in addition to memory as recategorization.I have called this the ability to have concepts,and the view I have taken, like the one I have taken of memory, departsfrom the conventional picture. The word "concept" is generallyusedin connectionwith language,and is usedin contexts in which one may talk of truth or falsehood.I have used the word concept, however, to refer to a capability that appearsin evolution prior to the acquisition of linguistic primitives. What is this capability? An animal capableof having conceptsidentifiesa thing or an action and on the basis of that identiftcation controls its behavior in a more or less generalway. This recognition must be relational:It must be able to connect one perceptualcategorization to another, apparently unrelated one, even in the absenceof the stimuli that triggered those categorizations.The relations that are captured must allow responsesto general properties"object," "up-down," "inside," and so on. Unlike elementsof speech,howevet concepts are not conventional or arbitrary, do not require linkage to a speechcommunity to develop,and do not dependon sequentialpresentation. Conceptual capabilities develop in evolution well before speech. Although they depend on perception and memory, they areconstructedby the brain from elementsthat contribute to both of these functions. It is difficult to know which animals beside humans have conceptual abilities. Certainly the evidenceon chimpanzeesis persuasive.These animals generalizeand classify relations-whether of things or of actions. Decisionsabout the conceptualcapabilitiesof other animalsare harder to make,howevet becauseunlike the casewith the chimpanzee,our corrmunication with other animalsis severelyrestricted.The best we may be able to do is to comparethe strucfuresand functions of their brain regions with those of humans and make guessesto guide further study. How did conceptualabilities arise?The TNGS proposesthat the evolutionary development of specializedbrain areasis required before conceptual abilities emerge.The argument supporting this proposal is based on 108

M e m o r y a n d C o n c e p t s :B u i l d i n g a B r i d g e t o C o n s c i o u s n e s s the notion that a simple increasein the number of reentrant maps capable of perceptualcategorizationis insufficientto accountfor concepts.Conceptual categorizationsare enorrnouslyheterogeneousand general.Concepts involve mixtures of relationsconcemingthe real world, memories,and past behavior. Unlike the brain areasmediating perceptions,those mediating conceptsmust be able to operate without immediate input. What brain operationsgive rise to theseproperties?The TNGS suggests that in forming concepts,the brain constructsmapsof its oaz activities,not just of extemal stimuli, as in perception.According to the theory, the brain areasresponsiblefor concept formation contain structuresthat categorize, discriminate,and recombinethe various brain activities occurring in different kindsof global mappings.Such structuresin the brain, insteadof categorizing outside inputs from sensory modalities, categorize parts of past global mappings according to modality, the presenceor absenceof movement, and the presenceor absenceof relationships between perceptual categorizations(figure 1O-Z). Structuresable to perform these activities are likely to be found in the frontal, temporal, and parietal cortices of the brain. They must represent a mapping of fupesof maps. Indeed, they must be able to activate or reconstruct portions of past activities of global mappings of different types-for example, those involving different sensory modalities. They must also be able to recombineor comparethem. This meansthat special reentrant connectionsfrom thesehigher-order cortical areasto other cortical areasand to the hippocampusand basalganglia must exist to carry out concepts. Brain areasgiving rise to conceptsmust be able not only to stimulate parts of past global mappings but also to do so independently of cunent sensory input. They must also be able to distinguish classesof global mappings (for instance,those correspondingto objects from those corresponding to movements).They must then be able to connect reactivated portions of global mappings and mediate the long-term storage of such activities. This is necessarybecauseconcept formation requires memory. The frontal cortex is a prime exampleof a conceptualcenterin the brain. Not enough is known about how its maps are organizedto be surewhether concept formation in this cortical area requires topographicmapping, as perceptual categorization does. It seemslikely, however, that maps that map the types of activity occurring in other cortical maps would be required; in higher-order maps, topography may not be so important. Given its connectionsto the basalganglia and the limbic system,including the hippocampus,the frontal cortex also establishesrelations subserving the categorization of values and sensory experiencesthemselves.In this

ro9

Pnoposars way, conceptualmemoriesdre affectedby values-an important characteristic in enhancing survival. With this notion of concepts,in which the brain categorizesits own activities (particularlyits perceptualcategorizations),it becomespossibleto seehow generalizedcategoriesand imagesmight be embodied.It is also possible to see how events may be categorizedas "past" without necessitating their being played out in presentbrain activities, as they must be for short-terrn memory and for the hippocampal successionleading to long-term memory. Furthermore,one can seehow conceptareas,by recursively restimulating portions of global mappings containing previous synaptic changes,give rise to combinations of relationshipsand categories. There is no need for any inherent logical order, classicalcategorization,or prior explicit programming.Yet the meansof conceptformation described here could quite naturally be responsible for establishing the complex categoriesthat I take up in the Postscript.Finally, becauseconcept formation is basedon the central triad of perceptualcategorization,memory, and leaming, it is, by its very nafure, intentional. This discussionof memory as recategorization,and of conceptsas the products of the brain categorizing its own activities,provides the bridging elementsrequired for reachingour goal: a biological accountof consciousness.Building on the tenets of the TNG$ the fundamentaltriad of percepfual categorization,memory, and leaming were linked to the emergenceof conceptualcapabilities.Notice, however, that no new theoretical assumptions were made,only assumptionsabout evolutionary changesin cortical morphology that alter the pattems of reentrant connectivity. It will tum out that an additional alteration of reentrant connectivity also provides a key to understandinghow we came to be aware.

110

C HAPTER

11

Consciousness: TheRemembered Present

Somethingdefinitehappenswhento a certainbrain-statea certain 'sciousness' corresponds, -William James

ost people,askedwhat it is about the mind that is truly distinctive and strange,would probably hark back to Descartes'lonely music We are now at that point in of the self and say,"Consciousness." our excursionwhen we may profitably ask whether we can do better than posfulate a thinking substancethat is beyond the reach of a scienceof extended things. What is daunting about consciousnessis that it does not seemto be a matter of behavior. It just is--winking on with the light, multiple and simultaneousin its modes and objects,ineluctably ours. It is a processand one that is hard to score.We know what it is for ourselvesbut can only judge its existencein others by inductive inference.As Iames put it, it is something the meaning of which "we know as long as no one asksus to define it." Indeed,it is initially best defined by consideringsome of its properties (of cowse the temptation is to indulge in a circulardefinition, madein terms of "awareness").Considerwhat I call its "famesian"properties(after)ames, who discussedthem): It is personal(possessedby individuals or selves);it is changing,yet continuous;it dealswith objectsindependentof itself; and it is selectivein time, that is, it does not exhaustall aspectsof the objects with which it deals. 1.1"I

Pnoposars Consciousnessshows intentionality; it is of or about things or events. It is also to some extent bound up with volition. Some psychologists suggestthat consciousness is markedby the presenceof mental imagesand by their use to regulate behavior. But it is rof a simple copy of experience (a "mirror of reality"), nor is it necessaryfor a good deal of behavior.Some kinds of leaming, conceptualprocesses,and even some forms of inference proceed without it. I have madea distinction, whicjr I believe is a fundamentalone, between primary consciousness and higher-orderconsciousness. Primary consciousnessis the state of being mentally awareof things in the world-of having mental images in the present. But it is not accompaniedby any senseof a personwith a past and future. It is what one may presumeto be possessed by some nonlinguistic and nonsemanticanimals(which ones they may be, I discusslater on). In contrast, higher-order consciousnessinvolves the recognition by a thinking subject of his or her own acts or affections.It embodiesa model of the personal,and of the past and the future as well as the present.It exhibits direct awareness-the noninferentialor immediate awarenessof mental episodeswithout the involvement of senseorgans or receptors.It is what we as humans have in addition to primary consciousness.We are consciousof being conscious. There are other resonancesin the term "consciousness"-these are revealed,for example,in the criteria used by cliniciansto assesswhether a traumatizedpatient is "conscious"or not---{riteria concemedwith alertness,orientation, self-awareness, and motivational control. Physicianstalk "clouded," of consciousnessas being in which state perceptualacuity and memory capacity are diminished.In extreme casesof disease,the lamesian properties, the "flights and perchings of consciousness,"become random, automatized,or show perseveration,with no evidenceof the existenceof inhospection or any attention to novelty. And in the last extremenothing, nothing to report. There is no end of hypotheses about consciousness,particularly by philosophers.But most of these are not what we might call principled scientiftctheories,basedon observablesand related to the functions of the brain and body. Severaltheories of consciousnessbased on functionalism and on the machinemodel of the mind (seethe Postscript)have recently been proposed. These generally come in two flavors: one in which consciousnessis assumedto be efficacious,and another in which it is considered an epiphenomenon.In the first, consciousness is likened to the executive in a computer systemsprogram, and in the second,to a fascinatingbut more or less uselessby-product of computation. In none of these notions, however, is there a direct appeal to biology or to the nature of embodiment. Such an appeal would obviously be TT2

C o n s c i o u s n e s sT: h e R e m e m b e r e dP r e s e n t essentialto any theory of consciousnessthat is based on evolution. A theory of this kind must propose explicit neuralmodels that explain how consciousnessarises. It must of necessity explain how consciousness emergesduring evolution and development.It must connectconsciousness to other mental matters suchas conceptformation, memory, and language. And it must describestringent tests for the models it proposesin terms of neurobiological facts. These tests should be undertakerypreferably with real experiments,or at least with what are called gedankenerperimentsthought experiments. In the latter, any properties postulated must be completely consistent with presently known scientific observationsfrom whatever field of inquiry and, above all, with those from brain science. Given the presentstate of affairs,this is a tall order becauseanalysesof consciousnessin biology are a bit like analyses of early cosmological events:Right from the beginning, certain manipulationsand observations are just not possible. Under these circumstances,one must be careful to spell out the assumptionsunderlying any proposed theory. I will spell out three that are part of the underpinnings of my theory of consciousness. Two of theseassumptionsare straightforward,but the third is a bit tricky. I call them the physics assumption,the evolutionary assumption,and the qualia assumption(the tricky one).I have to make theseassumptionsclear beforehandto avoid certain pitfalls, for example,into the Cartesianposition, into panpsychism,or into the cognitivist--objectivist quagmire that I discussin the Postscript. The physics assumptionis that the laws of physicsare not violated, that spirits and ghosts are out; I assumethat the description of the world by modem physics is an adequatebut not completely sufficient basis for a Modem quantum field theory provides a descriptheory of consciousness. tion of a set of formal properties of matter and energy at all scales(see figure P-1). It does not, however, include a theory of intentionality or a theory of namesfor macroscopicobjects, nor does it need them. What I mean by physics being just adequateis that I allow no spooks-no quantum gravity, no action at a distance,no superphysics(seethe Postscript)to enter into this theory of consciousness. The evolutionary assurnptionis also reasonably straightforward. It is that consciousnessarose as a phenotypic property at some point in the evolution of species.Before then it did not exist. This assumptionimplies that the acquisition of consciousnesseither conferredevolutionary fitness directly on the individuals having it, or provided a basisfor other traits that enhancedfitness.The evolutionary assumptionimplies that consciousness is effcacious-that it is not an epiphenomenon("merely the rednessof the melting metal," when pouring is what counts). Now, however, with the third assumption,we come to more subtle 1'1"3

P no r o s r r s issues.They are methodological ones,forced on us by the peculiarway in which consciousnessis made manifest. To explain the difficulty, I must make a detour here to discussphenomenal or felt properties, otherwise known as qualia. Qualia constitute the collection of personal or subjective experiences, feelings,and sensationsthat accompanyawareness.They are phenomenal states-'/how things seem to us" as human beings. For example, the "redness" of a red object is a quale. Qualia are discriminableparts of a mental scenethat nonethelesshas an overall unity. They may range in intensity and clarity from "raw feels" to highly refined discriminanda. These sensationsmay be very precise when they accompanypercepfual experiences;in the absenceof perception,they may be more or lessdiffuse but nonethelessdiscemible as "visual," "auditory," and so on. In general, in the normal waking state,qualiaare accompaniedby a senseof spatiotemporal continuity. Often, the phenomenalsceneis accompaniedby feelings or emotions, however faint. Yet the actual sequence of qualia is highly individual, resting on a seriesof occasionsin one's own personalhistory or immediate experience. Given the fact that qualia are experienceddirectly only by single individuals, our methodological difficulty becomesobvious. We cannotconslruct a phenomenalpsychologythat cnn beslnred in the sameway as a physics can be shared.What is directly experiencedas qualia by one individual cannot be fully sharedby another individual as an observer.An individual can report his or her experienceto an observer,but that report must always be partial, imprecise,and relative to his or her own personalcontext. Not only are qualia fleeting, but interventions designed to probe them may change them in unforeseenways. Furtherrnore,many consciousand nonconsciousprocessessimultaneouslyaffect eachperson'ssubjectiveexperience. Individuals may have their own private theories of the totality of their indioidual conscious experiences, but these can never be scientific theories.This is becauseother observersdo not have adequateexperimental controls available to them. The paradox is a poignant one: To do physics,I employ my conscious life, perceptions,and qualia. But in my intersubjective communication, I leave them out of my descriptiorLassuredthat fellow observerswith their own individual consciouslives can carq/ out the prescribedmanipulations and achievecomparableexperimentalresults.When for somereasonqualia do affect intelpretations, the experimental design is modified to exclude such effects;the mind is removed from nature. But in investigating consciousness,we cannot ignore qualia. The dilemma is that phenomenal experienceis a first-person matter, and this T'T.4

C o n s c i o u s n e s sT: h e R e m e m b e r e dP r e s e n t seems,at first glance,to prevent the formulation of a completely objective or causalaccount. Is this a completely hopelesssituation? I think not. But what altematives are open to us if we want to pursue One altemative that definitely does a scientificanalysisof consciousness? not seemfeasibleis to ignore completely the reality of qualia,formulating aloneto convey to a a theory of consciousnessthat aimsby its descriptions hypothetical "qualia-free"observerwhat it is to feel warmth, seegreen,and so on. In other words, this is an attempt to propose a theory based on a But no scientifictheory of whatkind of God's-eye view of consciousness. ever kind can be presentedwithout already assumingthat observershave sensationas well as perception. To assumeotherwise is to indulge the errors of theoriesthat attempt syntacticalformulationsmappedonto objectivist interpretations-theories that ignore embodiment as a source of meaning (seethe Postscript).There is no qualia-freescientific observer. If we exclude such an avenue,what other recourseis there?I believe there is one, based on the fact that human beings are in a privileged position. While we may not be the only consciousanimals,we are, with the possibleexception of the chimpanzee,the only self-consciousanimals. We are the only animalscapableof language,able to model the world free of the present, able to report on, study, and correlate our phenomenal states with the findings of physics and biology. This suggestsan approach to the problem of qualia. As a basis for a it is sensibleto assumethat, just as in ourselves, theory of consciousness, human beings,whether they are considered conscious qualia exist in other as scientific observers or as subjects.(lt does not matter whether these qualia are exactly the samein all observers,only that they exist.) We can then take human beings to be the best canonicalreferent for the study of consciousness.This is justified by the fact that human subjective reports (including those about qualia),actions, and brain structuresand function can all be conelated.After building a theory basedon the assumptionthat qualia exist in human beings, we can then look anew at some of the properties of qualia basedon these correlations.It is our ability to report and correlate while individually experiencing qualia that opens up the possibility of a scientific investigation of consciousness. distinguishesbetweenhigher-orderconsciousness This qualiaassumption and primary consciousness.Higher-order consciousnessis based on the occurence of direct awarenessin a human being who has languageand a reportable subjective life. Primary consciousnessmay be composed of phenomenalexperiencessuch as mental images,but it is bound to a time around the measurablepresent,lacksconceptsof self, past, and future, and lies beyond direct descriptive individual report from its own standpoint. 115

Pnopo src,rs Accordingly, beings with primary consciousnessalone cannot construct theories of consciousness--rven wrong ones! A researchprogram built on the assumptionsI have discussedobviously has a number of inherent difficulties. We must ftrst build a model for primary consciousness, build on that a model for higher-order consciousness,and then proceed to check the connectionsof each of these models with human phenomenalexperience.To be consistentwith the evolutionary assumption,this procedure must explain how primary consciousness evolved, and then explain how it was followed by higher-orderconsciousness. The order of the experimental enterprise (which, according to the qualia assumption, must be based on correlations obtained mainly on human subjects)must thereforebe exactly opposite that of the theoretical one, which must begin with the evolutionary precursorsto humans. I hope it is now clear why a biological theory based on our three assumptionscannot take a God's-eye view. To be scientists,we cannot expect any theory of consciousnessto render obvious to a hypothetical qualia-freeanimalwhat qualiaareby any linguistic description.To maintain intersubiectivecommunicationand carry out scientificcorrelation,which is a human activity, we mtst assumequalia. Qualia cannot be derived as experiencesfrom any theory. This does not mean,however, that different qualiacannot be theoretically discriminatedin terms of modality, intensity, continuity, or their temporal and spatialproperties.Nor does it mean that, after making the qualia assumption,we cannot consider the actual mechanismsby which qualia arise.Our cosmologicalcomparisonis not so far afield;we may askmodem physicsto explain certainaspectsof cosmology beginning at the earliestmoment, consistentwith the understandinggiven to us by modem physical theory. But we cannot ask a theory of physics to give a satisfactoryanswer to Gottfried Leibniz'squestion of why there is something rather than nothing. As it will tum out after we consider models for primary and higherorder consciousness,qualia may be usefully viewed as forms of higherorder categorization, as relations reportable to the self and then somewhat less satisfactorily reportable to others with similar mental equipment.Such a terse statementhardly satisftes.But instead of expanding on it now, I will describea model of primary consciousness, basedon our three assumptions,that appears to be consistent with the facts of brain structure and function. The elementsof this model include several systemsalready discussed,ones that give rise to value, to perceptualand conceptual categorizatioruand to memory. The dynamics of the model depend on a special kind of reentrant circuit. This is why I have explained these matters at length in previous chapters.(l will keep qualia to 11,6

C o n s c i o u s n e s sT,h e R e m e m b e r e d Present one side for now, but I will retum to them later when considering higher-orderconsciousness.)

PRIMARY CONSCIOUSNESS The model I have proposedhas a number of parts. (Would you believe a model of consciousness that had only one part?)Before describingtheir interactions,I want to say a few things about eachpart that might make an explanation of their interactionsclearer.There are,grossly speaking, two kinds of nervous system organization that are important to underevolved. Thesesystemsare very different in standinghow consciousness their organization,even though they areboth madeup of neurons.The first is the brain stem, together with the limbic (hedonic)system,the system concemedwith appetite,sexualand consummatorybehavior,and evolved defensivebehaviorpatterns.It is a value system;it is extensivelyconnected to many different body organs,the endocrinesystem,and the autonomic nervous system. Together, these systemsregulate heart and respiratory rate, sweating,digestive functions,and the like, as well as bodily cycles relatedto sleepand sex.It will comeasno suqpriseto learn that the circuits in this limbic-brain stem system are often arrangedin loops, that they respondrelatively slowly (in periodsranging from secondsto months),and that they do not consistof detailedmaps.They have been selectedduring evolution to match the body, not to match large numbersof unanticipated signalsfrom the outside world. Thesesystemsevolved early to take care of bodily functions;they are systemsof the interior. The secondmajor nervous system organization is quite different. It is calledthe thalamocorticalsystem.(The thalamus,a centralbrain stmcture, consistsof many nuclei that connectsensoryand other brain signalsto the cortex.)The thalamocorticalsystemconsistsof the thalamusand the cortex acting together, a system that evolved to receive signals from sensory receptorsheetsand to give signalsto voluntary muscles.It is very fast in its responses(taking from millisecondsto seconds),although its synaptic connectionsundergo some changesthat last a lifetime. As we have seen, its main structure,the cerebralcortex, is arrangedin a set of maps,which receiveinputs from the outside world via the thalamus.LJnlikethe limbicbrain stemsystem,it doesnot containloops so much as highly connected, layered local stntctures with massively reentrant connections.In many placesthese are topographically arranged(seefigure 9*2).The cerebral 717

P n op o s l r s cortex is a structureadapted to receive a denseand rapid seriesof signals from the world through many sensory modalities simultaneously-sight, touch, taste, smell, hearing, joint sense(feeling the position of your extremities). It evolved later than the limbic-brain stem system to permit increasinglysophisticatedmotor behavior and the categorizationof world events. To handle time as well as space,the cortical appendages-the cerebellum,basal gangli4 and hippocampus (see figure 1O-2F-evolved along with the cortex to deal with successionboth in actual motion and in memory. The two systems,limbie-brain stem and thalamocortical,were linked during evolution. The later-evolving cortical system served leaming behavior that was adaptive to increasinglycomplex environments.Because this behavior was dearly selectedto serve the physiological needs and valuesmediatedby the earlier limbic-brain stem system,the two systems had to be connectedin such a way that their activities could be matched. Indeed,suchmatching is a critical part of leaming. If the cortex is concemed with the categorizationof the world and the limbic-brain stem system is concemed with value (or with setting its adjustments to evolutionarily selectedphysiological pattems), then leaming may be seen as the means by which categorization occurs on a background of value to result in adaptive changesin behavior that satisfy value. Leaming certainly occursin animalsthat show no evidenceof conscious behavior. But in some animal specieswith cortical systems,the categorizations of separatecausallyunconnectedparts of the world can be correlated and bound into a scene.By a sceneI mean a spatiotemporally ordered set of categorizationsof familiar and nonfamiliar events, somewith and some without necessnry physicalor car.salconnections to othersin thesamescene.The advantageprovided by the ability to construct a sceneis that events that may have had significanceto an animal's past leaming can be related to new events,however causallyunconnectedthose events are in the outside world. Even more importantly, this relationshipcan be establishedin terms of the demands of the value systems of the individual animal. By these means,the salienceof an event is determinednot only by its position and energy in the physical world but also by the relative value it has been accordedin the past history of the individual animal as a result of leaming. It is the evolutionary development of the ability to createa scenethat led to the emergenceof primary consciousness. Obviously, for that emergence to have survived, it must have resulted in increasedfitness. But before considering how, let's consider the model itself. The appearanceof primary consciousness,according to the model, dependson the evolution of three functions. Two of these evolutionary 118

C o n s c i o u s n e s sT: h e R e m e m b e r e dP r e s e n t developmentsare necessarybut not sufficient for consciousness. The first is the developmentof the cortical systemin sucha way that when conceptual functions appearedthey could be linked strongly to the limbic system, extending already existing capacitiesto carry out leaming. The secondis the development of a new kind of memory based on this linkage. Unlike the systemof perceptualcategorization,this conceptualmemory systemis able to categoriztresporlses in the diferent bruin systemsthat carry out perceptual categorization and it does this according to the demandsof limbicbrain stem value systems.This "value-category" memory allows concepfual responses to occur in terms of the mutual interactions of the thalamocorticaland limbic-brain stem systems. A third and critical evolutionary development provides a sufficient means for the appearanceof primary consciousness.This is a special reentrant circuit that emerged during evolution as a new component of neuroanatomy. This circuit allows for continual reentrant signaling between the value-categorymemory and the ongoing global mappings that are concemedwith perceptualcategorizationin real time. An animal without these new reentrant connectionscan carry out perceptual categorizations in various sensory modalities and can even develop a conceptual value-categorymemory. Such an animal cannot, however, link perceptual events into an ongoing scene.With the appearanceof the new reentrant categorization of concurrentperceptiotts circuits in eachmodality, a conceptual can occur beforetheseperceptualsignalscontribute lastingly to that memory. This interaction between a special kind of memory and perceptual Given the appropriate categorizationgives rise to primary consciousness. 'tootstrapping process" takes place in reentrant circuits in the brain, this all sensorymodalities in parallel and simultaneously,thus allowing for the constructionof a complex scene.The coherenceof this sceneis coordinated by the conceptualvalue-categorymemory even if the individual perceptual categorizationevents that contribute to it are causally independent. My useof the word "scene"is meantto convey the ideathat responsesto roughly contemporaneousevents in the world are connectedby a set of reentrant processes.As human beings possessinghigher-order consciousness,we experienceprimary consciousness asa "picture" ora "mentalimage" of ongoing categorized events. But as we shall see when we examine higher-order consciousness, there is no actualimage or sketchin the brain. "image" The is a conelatlozr between different kinds of categorizations. To summarize:The brain carries out a process of conceptual "selfcategorization."Self-categoriesare built by matching past perceptualcategories with signals from value systems,a processcarried out by cortical systemscapableof conceptualfunctions. This value-categorysystem then 179

Pnoposrq,rs interacts via reentrant connectionswith brain areascarrying out ongoing perceptual categorizationsof world events and signals. Perceptual(phenomenal) experiencearisesfrom the conelation by a conceptualmemory of a set of ongoing perceptualcategorizations.Primary consciousnessis a kind of "rememberedpresent." These notions are illustrated in figure LL-1. While the diagram hardly conveys the complexity of the neural circuits involved, it does highlight several points. The first concems what we may call self and nonself components.(By self in this context I mean a unique biological individual, not a socially constructed"human" self.)The self, or intemal systems,arise from interactionsbetween the limbic and the cortical systems.This differentiates them from outside-world systemsthat are strictly cortical. The second point concems the formation of value-category memory.

SELF

NONSELF

Internal Homeostatic Systems

World Signals lncluding Proprioception

CurrentRegistration of InternalStates and Values

PRIMARY CONSCIOUSNESS ReentrantLoopConnecting Value-CategoryMemory to CurrentPerceptual Categorization

(lmpliespreviousExperienceand NeuronalGrouDSelection)

FIGURE11-1 A modelof primary consciousness. Pastsignals relatedto oalue (set by intemal control systems)and.categorized signalsfrom the outsideworld are conelatedand leadto memory in conceptualareas.This memory,which is capableof conceptualcategoizntion,is linkid by re,entrantpathsto cunent perceptualcategorizationof world signa| (hrovv lines).This resultsin primary consciousness. when it occursthroughmany modalities(sight,touch, and so forth), primary consciousness is of a "scene"madeup of objech and eients, some of which are not causallyconnected. An sninul with pimary consciousness can nonetheless connect these objects and eomts through memory aia its preoious aalue-laden erpeience. 120

C o n s c i o u s n e s sT: h e R e m e m b e r e dP r e s e n t This conceptualmemory dependson constantinteractionbetween self and world systems.The third point concems the occunencein real time and in parallelof perceptualcategorizationsfor eachsensorymodality via the cortical system, including the organs of succession.The final and critical point heraldsthe appearanceof primary consciousness: A correlative scene results from the function of reentrant connectionsbetween those cortical systemsmediating conceptualvalue-categorymemory and those thalamocortical systemsmediating ongoing perceptual categorizationsacross all the senses. Notice that primary consciousnessas I have characterizedit has the necessaryfamesianproperties:It is individual ("sell' systemscontribute to it), it is continuous and yet changing (as both world and intemal signals evolve), and it is intentional (referring necessarilyto intemally given or outside-world signalsderived altemately from things and events).If figure 11-1 were to be reiteratedin a seriesof time steps,it would serveto stress theseIamesianpropertiesof primary consciousness and the kind of percepfual bootstrapping that primary consciousness represents.famesianproperties stressthe flow of consciousness, its "before" and "after." In the consciousprocess,current value-freeperceptual categorizationinteracts with value-dominatedmemory. This occurs beforeperceptualevents contribute further to the alteration of that memory. When such events do contribute to the alteration of that memory, they are, in general, no longer in the speciousor rememberedpresent, that is, they are no longer in primary consciousness. What is the evolutionary value of such a system?Obviously, primary consciousnessmust be efficacious if this biological account is correct. Consciousnessis not merely an epiphenomenon. According to the TNGS, primary consciousnesshelps to abstract and organize complex changes in an environment involving multiple parallel signals. Even though some of these signals may have no direct causal connection to each other in the outside world, they may be significant indicatorsto the animal of danger or reward. This is becauseprimary consciousnessconnects their features in terms of the saliency determined by the animal's past history and its values. Primary consciousnessprovides a means of relating an individual's presentinput to its actsand past rewards.By presentinga correlativescene, it provides an adaptive way of directing attention during the sequencing of complex leaming tasks.It also provides an efficient meansof correcting effors. Theseperforrnancesmight conceivably be carried out without the construction of a scene.But it seemslikely that an animal with primary consciousnesswould have the ability to generalize its leaming abilities 72r

PnoPosALs acrossmany more cuesmore quickly than an animalwithout it. Consciousness,I repeat,is efficaciousand likely to enhanceevolutionary fitness. Primary consciousnessis required for the evolution of higher-order But it is limited to a small memorial interval around a time consciousness. chunkI call the present.It lacksan explicit notionor a conceptof a personal self,and it doesnot afford the ability to model the past or the future aspart seesthe room of a correlatedscene.An animalwith primary consciousness the way a beam of light illuminatesit. Only that which is in the beam is explicitly in the rememberedpresent;all else is darkness.This does not cannot have long-terrn mean that an animal with primary consciousness memory or act on it. Obviously, it can,but it cannot,in general,be aware of that memory or plan an extendedfuture for itself basedon that memory. I have Where are the actualbrain loci mediatingprimary consciousness? written elsewhereabout the possibility that certaincircuitsin the thalamus, between the cortex and the thalamus,and connectingone cortical region to anothermay be the sitesof the k"y reentrantcircuits.I will not overload this discussionwith the actual neuroanatomy(see ftgure 11-1 for the namesof the areasinvolved). Nevertheless,it may be useful to mention here that, as revealedby cognitive testing,certainbrain lesionslead to the recognitionof a signalwithin a given selectiveloss of the explicit conscious perceptualdomain that is nonethelessimplicitly recognized,as shown by psychologicaltests of the affectedperson. A good exampleis provided by stroke patients who have prosopagnosia-the inability to recognizefacesas such. Although they have no awarenessof faces,some of these patients will, while denying that they recognizetheir spouse'sface,perform on testsin sucha way as to indicate strong discriminatory knowledge of that face.Another example is blind sight. Individualswith lesionsin their primary visual cortex report blindness-no awarenessof vision-but can locate objects in space when tested.These matters will be discussedfurther in chapter 18. I mention them hereto point out that they may be explainedby assumingdisruptions (within the appropriateperceptualdomains)of the reentrantloops that I (figure 11-1). Let have postulatedas important for primary consciousness until later. for consciousness us defer the discussionof tests a few Beforeturning to the developmentof higher-orderconsciousness, words about some sticky mattersare in order. The first is: Which animals I really cannot answerthis exceptby relating have primary consciousness? it to the human referent that we agreed on. Going backward from the humanreferent,we may be reasonablysure (for reasonsthat will be made clearlater) that chimpanzeeshave it. In all likelihood, most mammalsand some birds may have it, although we can only test for its presencein122

C o n s c i o u s n e s sT:h e R e m e m b e r e d Present directly. Unfortunately,such tests are only neuroanatomicalor behavioral (not by sign communicationor report).If the brain systemsrequiredby the Presentmodel representthe only evolutionary path to primary consciousness,we can be fairly sure that animalswithout a cortex or its equivalent lack it. An amusingspeculationis that cold-bloodedanimalswith primitive corticeswould face severerestrictionson primary consciousness because their value systems and value-categorymemory lack a stable enough biochemicalmilieu in which to makeappropriatelinkagesto a systemthat could sustainsuch consciousness. So snakesare in (dubiously,depending on the temperature),but lobsters are out. If further study bears out this surmise,consciousness is about 300 million years old.

123

CHAPTER

12

LanguageandHigher-Order Consciousness

Human consciousness is a perpetualpursuit of a languageand a style,To assumeconsciousness is at onceto assumeform. Eoenat leoelsbelout the zone of definition and clarity, measuresand relationshipsexist,The chief characteristic of the mind is to be constantlydescribingitself, -Henri Focillon

he last two chaptershave involved a strenuousmarch through variegatedand difficult terrain.But if you will bearwith me through the next march, I believe you will be able to look back and see things more clearly-to make things "click." This is not quite possible at this juncture-to " see" clearly how primary consciousness works requiresseeinghow higher-orderconsciousness emergesand differs from it. It is curiousthat w€, as humanbeings with higher-orderconsciousness, alone. Creatures cannot "see the world" with our primary consciousness have no images, while possessingmental with primary consciousness, capacity to view those imagesfrom the vantage point of a socially constructed self. Yet one who has such a self as a result of higher-order needsit to link one mental image to the next in order to consciousness Higher-orderconsciousappreciatethe workings of primary consciousness! nesscannot be abandonedwithout losing the descriptivepower it makes possible.(l often wonder whether this abandonmentis what somemystics seek.) 124

Language and Higher-Order Consciousness

What we can usefully do before taking up the origins of higher-order consciousness is to seewhat "function" there is in our proposedmodels for various kinds of categorization.Perceptualcategorizatioo for example,is nonconsciousand can be carried out by classificationcouples,or even by automata. It treats signalsfrom the outside world-that is, signals from sensory sheetsand organs. By contrast, concepfualcategorization works from within the brain, requiresperceptualcategorizationand memory, and treatstheactivitiesof portionsof globalmappingsas its substrate.Connecting the two kinds of categorizationwith an additional reentrant path for eacl sensory modality (that is, in addition to the path that allows conceptual leaming to take place)gives rise in primary consciousnessto a correlated scene,or "image." This image can be regeneratedin part by memory in animalswith primary consciousness, but it cannot be regeneratedin referenceto asymbolicmemory. By this I mean a memory for symbols and their associatedmeanings.And so an animal with primary consciousnessalone is strongly tied to the successionof events in real time. How can the tyranny of this remembered present be broken? The impreciseanswer is: By the evolution of new forms of symbolic memory and new systems serving social communication and transmission.In its most developedform, this meansthe evolutionary acquisitionof the capability for language.Inasmuchas human beings are the only specieswith language,it also meansthat higher-orderconsciousness hasflowered in our species.But there are strong indications that we can see at least some of its origins in chimpanzees.Both speciescan think not just have concepts, and chimpanzeesalso appear to have some elements of a self-concept. Certainly, the basis for recognizing a subject-predicate relationship in humansrequiresan emerging consciousness of the distinction between the self (in the socialsenseof "selfhood") and other entities classiftedasnonself. Chimpanzeeshave behaviorsindicating that they make the distinction, but they lack true language and so I claim that what I call higher-order consciousnesscannot flourish in them as it does in us. Higher-order consciousness obviously requiresthe continued operation of the structuresserving primary consciousness. In addition, it involves the ability to construct a socially basedselfhood,to model the world in terms of the past and the future, and to be directly aware. Without a symbolic memory, these abilities cannot develop. To tracehow theseabilitiesmay have developedthrough the evolutionary emergenceof a symbolic memory, it will be necessaryto considerhow speechevolved and how it is acquired.Therefore,I will first considerhow the emergenceof true language required the evolution of the vocal tract and the brain centersfor speechproduction and comprehension.I then will 725

PnoPosALs confront an issuecentralto this €SS?!:whether conceptsare formed prior to speech.In doing so, I will concludethat a model of self-nonselfinteraction probably had to emergeprior to true speech.

THEORY AN EPIGENETIC SPEECH: The considerations presented so far suggest that a model-for speech Furthermore,the development acquisitionrequiresprimary consciousness. without the prior improbable is highly and grammar syntax of a rich evolution of a neuralmeansfor concepts.If this tums out to be true, it will be obvious why computers are unable to deal with semantic situations. Their embodiment is wrong; it does not lead to consciousness. I propose that before language evolved, the brain already had the necessarybases for meanings in its capacities to produce and act on concepts.The evolution in primates of rich conceptualmemories,and in hominids of phonological capabilities and special brain regions for the productiorl ordering, and memory of speechsounds,then opened up the possibility of the emergenceof higher-order consciousness.(Although I will not discussthe details of grammaticalsystemshere,a pertinent discussion of some aspectsof grammar may be found in the Postscript.) Speechis special and unique to Homo sapims.Can we account for its evolutionary emergencewithout creating a gulf between linguistic theory and biology? Yes,provided that we accountfor speechin epigeneticaswell as genetic terms. This means abandoning any notion of a genetically programmedlanguage-acquisitiondevice.It does not mean,however, that specializedheritable structures were not necessaryfor speech to arise. Indeed, the evidence for the existenceof specializedheritable structures related to speech is not hard to find. After the assumption of bipedal posture by hominids, changesoccurredin the basicranialstructureof their skulls (figure l2-7). This provided a morphological basisfor the evolution of a uniquely human piece of anatomy, the supralaryngealtract or sPace. This tract becomesmature in human infants when the larynx descends.(To avoid choking, a structure called the epiglottis must close when humans eat. Indeed, unlike other animals,we cannot phonate and swallow at the sametime without potential disaster.)As part of this evolutionary development, the vocal folds emerged and the tongue, palate, and teeth were selectedto allow fuller control of air flow over the vocal cords, which in fum allowed the production of coarticulatedsounds,the phonemes. 126

Language and HiSher-Order Consciousness

N

E 5 ... OralCavity

o c o a cr o tt

... Eplglotils ... Pharynx "'

Larynx

c 6

E o lr 4681012 Length

of Back

Cavity

(l U)

@mm

FIGURE 12-1 The supralaryngeal tract in humans,one of the main anstomicalbasesfor speech productionand theeoolutionof language. Thiscomplexstructurefunctionsbecause the larynx(ooicebor)hasdescmded sothat erhaledair effectioely causes theoibrationof the oocalcords,whicharein tum alteredin tercionand apposition by exquisite muscular (left).Modulationbyotherelements-the changes tongue, teeth,lips, andsoforth-leads to a setof coarticulated sounds(nght top). Thereis a "isk of chokingto deathif the epiglottisdoesn'tclosetheairutaywhenswallowingoccurs(right bottom). These changes presuryose that changes haoealreadyoccurred in thebaseof theskullduing eoolution.

At the sametime or shortly after in evolution, specialcerebralcortical regions emerged on the left side that are now known as Broca's and Wemicke's areas (ftgure 1,2-2). These cortical regions linked acoustic, motor, and conceptual areas of the brain by reentrant connections. Through theseconnections,Broca'sand Wemicke's areasservedto coordinate the production and categorizationof speech.Most importantly, they provided a system for the development of a new kind of memory capable of recategorizing phonemes (the basic units of speech)as well as their order. We may reasonablyassumethat phonology arosein a speechconrmunity that used primitive sentences(perhapsresemblingpresent-daypidgin languages)as major units of exchange.In such an early community, utterancescorrelatednouns with objectsand led to the beginnings of semantics (figure t2-3). Verbs followed. Note that the preexisting capacity for conceptsprovided a necessarybasisfor thesesemanticdevelopments.In early humans,the presyntacticalorganization of gesturesmay have allowed a simple ordering of nouns and verbs. Further development of Broca'sand Wemicke's areas allowed the more sophisticatedsensorimotor ordering that is the basis of true syntax. According to the theory of speech acquisition that I favor, syntax 727

PnoPosALs \

l

Superior SpeechCortex

(Supplementary Motor)

Anterior SpeechCortex (Broca)

SPEECHAREAS IDEATIONAL Evidence Stimulation

FIGURET2-2 Areas of the brain sentingspeechproduction(top). If thesebrain regionsare damaged, aphasiaoccursin a oariety of forms, Picturedis the brain of oneof Paul Broca'spatients who had a lesionin what is calledBroca'sarea (bottom). Its ourner,ushenalioe, had motor aphasia.

728

Language and HiSher-Order Consciousness

a a a I a a

v a

a a a t a a l r a a a a a a a a a a a a a a a a a a a a I a a

Objectr (Cat) Objectz (Repeat) a I

i

P

r

I Multiple Objbctr'r{ Occasions, Uhrases

FIGURE 12_3 crudelyshowinghow affect,reward,and learning A scheme Semanticbootstrapping, proaidesthe leadto speechacquisition.Phonology underconditionsof categorizntion aremadewith to semantics. As reentrant connections objects categorized meansto connect are occurs.,* a lexiconis built and sentences conceptcentus,semanticbootshapping leadsto syntar. thecategoizationof their anangements experienced,

emergedepigeneticallyin a definite order (figure 1.2-3).First, phonological capabilities were linked by leaming with concepts and gestures,which allowed for the developmentof semantics.This developmentpermitted the accumulationof a lexicon: words and phraseswith meaning.Syntax then emergedby connectingpreexisting concepfualleaming to lexical leaming. A similar idea has been proposed by Steven Pinker and others within the framework of a grammardevelopedby foan Bresnan,which shecallslexical functional grammar.They call this processsemanticbootstrapping. In the extended TNGS, I provide explicit evolutionary, anatomical,and physiological argumentsto support the notion that an infant already has concep\29

P n o po s l r s tual categoriesthat are not defined or originated by semanticmeans or criteria.Thesecategoriesare requiredif semanticbootstrapping is to occur, and they are also required to support the related proposals of Ronald Langacker,George Lakoff, and others of what have been called "cognitive grammars" (seethe Postscript). Thus, to build syntax or the basesfor grammar, the brain must have reentrant structuresthat allow semanticsto emergefirst (pnor to syntax) by relating phonological symbols to concepts. Becauseof the special memory provided by Broca's and Wemicke's areas, the phonological, semantic,and syntacticallevels can interact directly and also indirectly via reentrant circuits that are formed between these speechareasand those brain areasthat subservevalue-categorymemory. When a sufficientlylarge lexicon is collecte{ the conceptualareasof the brain categonze lhe order of speechelements,an order that is then stabilized in memory as syntax. In other words, the brain recursively relates semantic to phonological sequencesand then generatessyntactic correspondences, not from preexisting rules, but by treating rules deoelopingin meffiory as objects for conceptualmanipulation.Memory, comprehension,and speechproduction interact in a great variety of ways by reentry. This permits the production of higher-order structures(suchas sentencesin a grammar)and obviously helps with the elaboration of lower-order sequences(suchas phrases).Of course, once achieved, the sequencingbecomes automatic, as do many other motor acts. Chimpanzees,unlike humans, have no brain bases for the complex sequencing of articulated sounds. They appear to have concepts and thought and are even capableof a simple "semantics,"but inasmuchas they lack an elaboratedsyntax, they have no true language or speechper se. It is obvious why the acquisition of true speechleads to an enorrnous increasein conceptualpower. The addition of a specialsymbolic memory connectedto preexisting conceptualcentersresults in the ability to elaborate, refine,connect,create,and remembergreat numbersof new concepts. It is not the case that the language centers "contain" concepts or that concepts"arise" from speech.Meaning arisesfrom the interaction of valuecategory memory with the combinedactivity of conceptual areas and speechareas.And although vocal speechwas probably necessaryfor the evolutionary selectionof the necessarymorphological changesin the brain after their emergenceany gestural system in a speechcommunity (suchas sign language)could be employed if necessary.Moreover, like many brain systemssubjectto epigeneticdevelopment,the system underlying speech acquisitiondiffers in children and in adults;it is subjectto a developmental critical period. In all likelihood, this time period is related to extensive 130

L a n g u a g ea n d H i s h e r- O r d e r C o n s c i o u s n e s s after synaptic and neuronal group selectionoccurring up to adolescence, in a different and less extensively occur much changes time such which fashion. This theory of speechis a nativist theory insofaras it requiresthe prior evolution of specialbrain structures.But it invokes no new principles beyond thoseof the TNGS. It is not a computationaltheory, nor one that insistson a languageacquisitiondevicecontaininginnategeneticallyspecified rules for a universal grammar. Syntax is built epigeneticallyunder genetic constraints,just as human faces(which are about as universal as grammar)are similarly built by different developmentalconstraints.The principlesof topobiology (seechapter6) apply to both cases. This proposalis compatiblewith the capacityto constructand interpret a potentially infinite number of sentencesfrom a finite number of words. This is so becausethe generalizingand categorizing power of a concePtual system interacting reentrantly and recursivelywith specializedlanguage areasis well-nigh unlimited.Inasmuchas syntax is constructedfrom semantics (underconstraints),local grammaticalrelationscanbe constructedeven from sentencefragmentsfree of the strict order of sentences.A grammar so built is necessarilymapped onto the continual activities of a very definite set of brain structures,among which the most important may be Indeed,if this theory is correct, thosegiving rise to primary consciousness. languageis impossiblewithout primary consciousness.

CONSCIOUSNESS HIGHER-ORDER With this theory of speechin hand, we may return to our main subject: How does one become "consciousof being higher-orderconsciousness. conscious?"In order to acquirethis capacity,systemsof memory must be related to a conceptualrepresentationof a true self (or social self) acting on an environmentand vice versa.A conceptualmodel of selfhoodmust be built, aswell asa model of the past.A numberof stepsof developmental learning that alter the individual's relation to the immediatepresent are ary for this to take place. necess Brain repertoiresare requiredthat are able to delay responses.(Repertoires of this type are known to be present in the frontal cortex.) These repertoiresmust be able to categorize the processesof primary consciousnessitself. This is achievedlargely through symbolic means, by comparison and reward during socialtransmissionand learning.During the acquisir37

P no p o s a r s tion of semantics,that reward arisesby relating speechsymbols to the gratification of affective needs by conspecificsin parental, grooming, or sexual interactions. The figure (figure 72-4) showing the relation of speechareasto conceptual areas,which allows for the development of a concept of self and of higher-order consciousness, must be supplementedwith one showing social relations (see figure 1,2-3). Long-term storage of symbolic relations, acquiredthrough interactionswith other individuals of the samespecies,is critical to the self-concept.This acquisitionis accompaniedby the categorization of sentencesrelated to self and nonself and their connection to events in primary consciousness. The correspondingelaboration achieved by the leaming of elements in phonemic and symbolic memories also allows more effective categorizations through verbs of various actsin relation to the self and others. The interaction between this specializedset of memoriesand conceptual value-categorymemory allows for a modeling of the world. And given the

SELF

NONSELF

Internal Homeostatic Systems

WorldSignals IncludingProprioception I

*

Y

CurrentRegistration of InternalStates and Values

Conceptual Categorization

CurrentPerceptual Categorization Semantic Bootstrap

PRIMARY CONSCIOUSNESS ReentrantLoop Connecting Value-CategoryMemory to CurrentPerceptual Categorization

HIGHER-ORDER CONSCIOUSNESS

(|mpr ies t':r8Hfi €:,'i:i:ffi i" FIGURE1.2-4 A scheme (Thereadermay relatethisto thescheme shown for higher-orderconsciousness. in figure 11-1 for primary consciousnessj Theacquisitionof a new kind of memoryaia (figure12-3) leadsto a conceptual semnnticbootstrapping erplosion,As a result,concepts of theself, thepast,and thefuture canbeconnected "Consciousto primary consciousness, nessof consciousness" becomes possible, 132

L a n g u a g ea n d H i g h e r - O r d e r C o n s c i o u s n e s s emergenceof the ability to distinguish such conceptual-symbolicmodels from ongoing percepfualexperience,a concept of the past can be developed. This frees the individual from the bondage of an immediate time frame or ongoing events occurring in real time. The rememberedpresent is placed within a framework of past and fufure. While the embodiment of meaning and referencecan be related to real objects and events by the reentrant connectionsbetween value-category memory and perception(primary consciousness), simultaneousinteractions can also occur between a symbolic memory and the same conceptual centers.An inner life, based on the emergenceof language in a speech community, becomespossible. This is tied to perceptual and conceptual structures,but it is highly individual (indeed,it is personal)and it is also strongly tied to affect and reward. It is higher-orderconsciousness, capable of modeling the past, present, future, a self, and a world. One of the astonishing featuresof higher-order consciousnessis how rapidly it appeared.Paleontologicalstudieshave shown that these developments occurred over very short periods of evolutionary time. The topobiological principles underlying brain development and the mechanismsof the TNGS can account for this rapid emergence,for they allow for enormous changesin brain size over the relatively short evolutionary periods in which Homo sapiensemerged.According to topobiology, morphological changesof relatively large extent occur through changesin the timing of the action of morphoregulatory genesasa result of relatively few mutations (seechapter 6). And the premisesof the TNGS allow for the rapid incorporation of new and enlargedprimary repertoiresinto existing brain structures. A synoptic picture of how consciousnessis related to evolutionary moqphology is diagrammed in figure 12-5. While this hardly gives the detailsand even lacksa time frame,it does suggesthow two successioe setsof (perceptual bootstrapping eoents and semantic),eachinvolving the evolution of new morphology (memory circuits and new forms of reentry) could give rise first to primary consciousness and then to higher-orderconsciousness. This evolutionary panorama provokes additional questions about the adaptive advantagesof consciousness. Primary consciousness provides the ability to determine by intemal criteria the salienceof pattems among multiple parallel signalsarising in complex environments.That salienceis largely but not completely determinedby the previous history and leaming of the individual animal. Higher-order consciousnessadds socially constructedselfhood to this picture of biological individuality. The freeing of parts of consciousthought from the constraintsof an immediatepresent and the increasedrichnessof social communicationallow for the anticipa733

PnoPosALs

NATURAL SELECTION (See chapter 9)

DEVELOPMENT (Seechapter10)

NEURONAL GROUP SELECTION

Form and Tissue PatternLeadingto Behavior (Changes in Genesinvolvedin Morphoregulation and Differentiation)

Morphoregulatory and \ Historegulatory Genesin I Interacting CellCollectives \ \Subjectticnrrl Cyclesand fSAM Networks("Topobiology") | Leadingto SomaticVariationJ

Primary Repertoire of VariantNeuronal Groupsin Brain

Brain Stem, Hypothalamus, AutonomicSystems

(Seechapter13)

ReentrantMapping (Perceptual Categorization )

PRIMARY CONSCIOUSNESS (See chapter 15)

Frontal,Temporal, ParietalCofiices (Conceptual Categorization) Semantic Bootstrap

HIGHER.ORDER CONSCIOUSNESS (Presentchapter)

FIGURE12-5 The eoolution of cottsciousness dependson the eoolution of new morphology.Here, an and eoolutionarysequence of eoentsis shownin which thepinciples of natural selection deoelopmmtleadto neural recognitionsystemsand result in consciouserpeience.No new pinciples besidesthoseof the theory of neuronalgroup selectionare required.But newly eooloedanatomical structuresselectedfor function are required. Theseinclude those shown in thefirst two figuresof this chapter,The principlesunderlyingthefunction of eacharea nre erplained in the chaptersindicated. Notice that a "perceptualbootstrap" producespir*ry and a "semanticbootstrap" produceshigher-orderconcottsciousness Both bootstrapsrely on the eaolutionof approgiate reentrantpathways in the sciowness. brain.

134

L a n g u a g ea n d H i g h e r - O r d e r C o n s c i o u s n e s s tion of fufure statesand for planned behavior. With that ability come the abilities to model the world, to make explicit comparisonsand to weigh outcomes;through suchcomparisonscomesthe possibility of reorganizing plans. Obviously, these capabilitieshave adaptive value. The history of humanity sincethe evolution of hunter-gatherersspeaksto both the adaptive and the maladaptive properties of the only specieswith fully developed higher-order consciousness. Some anthropologists have proposed (somewhat fancifully) that our brains enlarged so rapidly because,after a certain point, higher-order consciousnessconferred on us the ability to deceive ourselvesin sudr a way as to allow us to deceive others more "sincerely" to our own advantage. In a socially bound animal, this deceptivenessmight, according to theseauthors,have selectiveadvantages.An old tale tells of Boris and Ivan at the train station. Ivan says,"Boris, where are you going/" Boris replies, "To Minsk." Ivan then says, "Boris, I know you. If you were going to Minsk you would tell me you were going to Pinsk. Now, I happen to know you are going to Minsk. So, why are you lying to me?" Given this picture of a human animal simultaneouslyand interactively capableof higher-order and primary consciousness, we may retum to the vexing question of qualia.Recallthat our theoreticalanalysisof consciousnesswas founded on three assumptions:the physicsassumption,the evolutionary assumption,and the qualia assumption.Having already assumed that human beings have qualia,why retum to the issue?We know that a God's-eye view-in which the theory would, through the communication of its strucfure, allow an imaginary qualia-freeanimal know what qualia ar*-is not feasible.We have said enough about the mechanismsof consciousnessto indicate that only through direct possessionby an individual of the appropriate molphology and experiencedo qualia arise.Nevertheless, our elaboratedpicture provides certain refinements. First of all, it is clear how diferent qualia are discriminated-through differencesin neural structure and behavior in different sensorypathways. This has been known for a long time, since the doctrine of specificnerve energiesadvancedby ]ohannes Miiller. We can add that an animal with higher-order consciousnessis likely to call a given phenomenalstate by a different name than another one evoked by a different neural pathway (if not "green" ot "warm," then at least "ween" and "grarm," but in any case, in a consistent fashion). If animals having only primary consciousnessalso have qualia, they cannot report them explicitly either to a human observeror to themselves, for they lack conceptualselves.Like flashlights illuminating a room, their qualia,if they occur,exist only for the duration of the rememberedpresent 735

Pnoposlrs of the scene.We can only adducetheir possiblepresenceby observing the behavioral responsesof these animals. But with us,it is different.Qualia, individual to eachof us, arerecategorizations by higher-order consciousnessof valueladen perceptualrelations in eachsensorymodality or their concepfualcombinationswith eachother' We report them crudely to others; they are more directly reportable to ourselves.This set of relationshipsis usually but not always connectedto value.Freedom from time allowsthe locationin time of phenomenalstatesby a suffering or joyous self. And the presence of appropriate language improves discrimination enorrnously; skill in wine tasting, for example, may be consideredthe result of a passionbasedon qualia that are increasingly refined by language. With this view of higher-order consciousness,it is possible to see roughly what lies beneath the self that connectsphonology to semantics in a naming sentence.Once a self is developedthrough socialand linguistic a world is developed that interactionson a baseof primary consciousness, requiresnaming and intending. This world reflects inner events that are recalled,and imagined events,as well as outside events that are percepfually experienced.Tragedy becomespossible-the loss of the self by death or mental disorder, the remembranceof unassuageablepain. By the same token, a high drama of creation and endlessimagination emerges. Ironically, the self is the last thing to be understood by its possessor, even after the possessionof a theory of consciousness. Given the way in which higher-orderconsciousness arisesand naming occurs,that should be no surprise, except to each of us as a possessor.Embodiment imposes ineluctablelimits. The wish to go beyond theselimits createscontradictioo fantasy, and a mystique that makes the study of the mind especially challenging,for after a certain point, in its individual creationsat least,the mind lies beyond scientific reach. Scientific study recognizes this limit without indulging in mystical exercisesor illusions.The reasonfor the limit are is straightforward:The forms of embodimentthat lead to consciousness unique in eachindividual, unique to his or her body and individual history.

136

CHAPTER

13

Attentionand the Unconscious

Illusionscommendthemselues theysaoeus pain and to us because allow us to enjoy pleasureinstead,We must thereforeacceptrt without complaint uthen they sometimescollide with a bit of reality againstwhich they are dashedto pieces, -Sigmund Freud

reignsbut doesn'tgoaern, Consciousness

_paul Val6ry

here are two objectionsthat may be raisedto what I have said so far. The first is that I have not really explained"what it is like" to be conscious.The secondis that there appearsto be much I have failed to explain, for example, the fact that a good deal of our behavior is unconsciously driven. Sigmund Freud spent much of his life trying to understandthis, particularly trying to understandthe repression of experiencesthat were threatening,painful, or unpleasant.I will consider the first objection briefly and the second at greater length. Given the central importance of consciousnessto knowing that we exist-Descartes'claim-it is no suqprisethat much would be expectedof any account that presumesto explain it. Of the present one it might be 'You may think you have explained how memory, percepfualcatesaid, gorization, reentry, and so on, work to give the properties of consciousness, but you have not explainedhow Ifeel being conscious,or why Ifeel myself to be conscious.Consciousnessis shange, mysterious, the ultimate mysr37

P n o po s a r s tery." To reply I have to point out the limits of any claim to scientific explanationand then show what is specialabout any explanationproposed for consciousness. Scienceis concemedwith the formal correlationsof properties,and with the development of theoretical constructs that most parsimoniously and usefully describeall known aspectsof that correlation,without exception. It must couch its descriptionsin terms that can be exchangedand understood between any two human observers.Any description that does not assumea conscious,understandinghuman observer as its target, one who can object to flaws in logic, repeat experiments,and construct new ones, is not a scientificdescription.An example of a nonscientificdescription is a personalaccountof my particular sensations,apparentrecollections,and emotions during a drug-induced trance. At best, someonemay correlate the reports of twenty subjects (including me) in such a trance and find regularities.But he or she will not be able to correlate reliably my actud, feelings,particularhistory, and mode of forgetting in any detail or with any general certainty. So sciencefails for individual histories even though it may succeedin disceming what is common among twenty chronicles. There is something peculiarabout consciousnessas a subjectof science, for consciousnessitself is the individual, personalprocesseach of us must possessin working order to proceed with any scientificexplanation.Even though I may be unaware of what I have forgotten or repressed,or ot unconsciousfactors that drive my behavior, I feel as if the process of consciousnessis all of a piece, at least in my healthy state. And so it is nafural that I demand an explanation of my own consciousnessin terms satisfactoryto myself. But I must realizethat it is not a scientificact to do so, nor would I expect it to be. After all, no one says to a physicist, "You have explainedenergy and matter in terms of symmetry relations,and you have even approachedthe beginnings of the universein your theories.But you have not reallyexplainedwhy thereis somethingrather than nothing." To attempt such an explanation would be fruitless; under these circumstances,no sciencebasedon experiment could recommenditself as better than any other. A scientific explanation cannot be given. Well then, why are we tempted to demand a scientific explanation of how it feelspersonally to be conscious?It is the certainty of consciousness to ourselves and its relation to the idea of self that makes us want to demand more of a psychologist than of a physicist or a cosmologist. But the demand is not a scientificallyreasonableone. A reply to the questionwould have exactly the sameform as one given by a physicist would have. "l have offered you a theory in terms of known structwes and relationships,one based on experimentalfacts. The theory 138

Attenlion and the Unconscious says that, if you perform an operation on structuressaid to be important those propertieswill be predictably altered for propertiesof consciousness, or may even disappear.lf, for example,you cut a reentrantloop connecting a part of the brain essentialfor carrying out facerecognition (and just that part), a consciousperson will experienceprosopagnosia.That person will of the fact that he or she can (asshown by becomeand remain unconscior,s examination of implicit memory) still recognize faces that an observer knows the person has seenbefore, and he or she will quite sincerelydeny recognizing them. So it is with other tests; a theory of consciousnesscan have operationalcomponents.But it must, if it is a good theory, also unify all kinds of pertinent facts and deepenour understanding.(For example,it should explain how some processescan be unconsciousand yet motivate behavior, which is the task of this chapter.) Why then do we insist on a God's-eye view, even in the face of these mystique-a desirefor univerexplanations?Why is there a consciousness sal explanation,for conservationof consciousnessas an individual experience, time without end? A reasonableanswer seems to be that each consciousnessdependson its unique history and embodiment.And given that a human consciousself is constructed,somewhat paradoxically, by social interactions,yet has been selectedfor during evolution to realize mainly the aims and satisfactionsof eachbiological individual, it is perhaps no surprisethat as individuals we want an explanationthat sciencecannot give. It is also perhapsno suqprisethat we desireimmortality. But there is no more mystery to our inability as scientiststo give an explanation of an than there is to our inability to explain why there individual consciousness is something rather than nothing. There is a mystery perhaps,but it is not a scientificone. If one stays solely with one's own mind, the mystery rests in imagining how that particular mind arises with regard to its own personalhistory. We are "locked in." There is one real but remote possibility of dealing scientificallywith the "locked-in" property of consciousnessin an individual, the source of this "mystery." If an artifact could be built that had structuresand experiences allowing it both to becomeconsciousand to have language,one could test for the presenceand absenceof qualia.If it reported a certainfeeling,would it be reasonable(and ethical) to rebuild it without the structuressurmised to be essentialto that feeling? After such a procedure would the same artifact then feel "strange" and "different," given that its self would have been built from the interaction of unconsciousprocessesand conscious "social" interactions?We must wait and see,but as fantasticand as improbable as sucha proposalsounds,it is at leasttheoreticallypossible.The same cannot be said of experimenting with the creation of the universe. 739

PnoPosALs Before turning to the unconsciousprocessesthat give rise to and alter it may be useful to consider what is difficult to imagine consciousness, about the mind and what is easy.As we have seen,one overriding difficulty hauntsany attempt to explainthe mind: It is that the mind arisesas a result of physical interactionsacrossan enorrnouslylarge number of different levels of organization,ranging from the molecularto the social.Furthermore, these interactions are often idiosyncratic or irreversible,and the structuralfeaturescentralto their workings includeparallel,one-many, or many-many mappings.Our brains (and particularly philosophers'brains) are not very good at visualizingsuchcomplex orderings.But the situation may not be hopeless;as I discusslater,the advent of increasinglypowerful computersmay help us build heuristicsthat can let us seehow things go together. Until this field is more fully developed,we may ask' What is easy to imagineabout the mind?I believemost people would agreeto the following list: The workings of brain circuits grossomodo in terms of their inputs and outputs. The example is classicalneurophysiology. 2. The interaction between an animal's pattern of behavior and the physical world of stimuli. The example is descriptive psychology. 3 . Certain acts of social transmission. The example is the study of imprinting in ethology and the accepted propositions of folk psychology-what people seem to believe, desire, or intend. L

What is hard to understand? 1 . The net result of the simultaneous action in parallel of complex neural populations. An example is the difficulty of predicting the outcome of the activity of a large number of neuronal groups. 2. Memory as a dynamic process and system property, one that is not equivalent to the sum of the synaptic changes that underlie it. An example is the overall responseof an automaton like Darwin III after training. 3 . More complex psychological phenomena such as consciousness. Numerous examples have been given in the last two chapters. 4 . The idea of a socially constructed self resulting from the interactions of both unconsciousand consciousprocesses.The example is discussedin the rest of this chapter.

Undoubtedly, one may be able to think of alternativelists. To understandpsychologicalprocesses(particularlythosein the secondlist) in terms 1.40

Attention and the Unconscious of a brain theory, however,one must not only have a good theory but also of be able to use synthetic computermodels to checkthe self-consistency its mechanismsand analyzemultilevel interactions.Unfortunately,mathematicsalonewill not sufficientlyaid our verbal constructionsas it doesour physics.At present,computer simulationsseemmore promising. In consideringthis chapter,I suggestthat the readermake every effort to understandhow classificationcouplestogether with reentry yield perceptualcategorizationand how a dynamic memory functionsas a system property (seechapters9 and 1O).With thesetwo processesunderstood,I believe the reader can move step by step through the models I have and "see" how they describedfor primary and higher-orderconsciousness might yield Jamesianproperties,changesaliencein a scene,and allow plans to be formulated.Perhapsthe best way to try this is by examining the figuresin the appropriatechapters.The understandinggainedwill make it much easierto see how attention and the unconsciousoperate.

ATTENTION hasnot explicitly dealt with attentioil, which My accountof consciousness "the taking possessionby the mind, in clear and vivid form, Jamescalled of one out of what seemseveralsimultaneouslypossibleobjectsor trains but its relationship of thought." Attention is not the sameasconsciousness, to consciousnessposes some of the most difficult problems for theory. , for Attention must,for example,be discriminatedfrom overall wakefulness directional a it lends alertness; vigilance or of matter it is not simply a to the componentto behavior,and it modulatesan animal'sresponsiveness It environment.Indeed,attention revealsthe "fragility" of consciousfl€ss: sulrounding attenuates and or obliterates focusesour mind on its objects "irrelevancies."It doesnot seempossibleto pay specificattention to more than a few objectsor lines of thought; attention is highly selective,apparently obligatorily so. Many theoriesof selectiveattention arebasedon the notion of "filtering out" input signals,either early or late. But there is a variety of evidence suggestingthat such filtering does not occur. I have favored the notion, posited by others, that brain mechanismsof attention were originally derived from evolutionary pressureon an animalto selectone out of a set of appropriateactions.An animal that is hun gry or being threatenedhas r47

P n o p os l r s to selectan object or an action from many possibleones.It is obvious that the ability to choose quickly one action pattem to be carried out to the exclusion of others confers considerableselective advantage.Possessing suchan ability makesit possibleto achievea goal that would otherwisebe interfered with by the attempt to undertake two incompatible actions simultaneously.Survival may depend critically on this ability. 'motor" This theory of the origin of attention does not imply that perceptual components are not important. Indeed, it is obvious that the mechanismsof attention are multiple, ranging from perceptualcompetition to volitional choice. But if the end result is the formulation of a sequence of actions or motor plans,whether executedor not, then, accordingto the TNGS, global mappingsand the basalganglia are likely to be involved (see chapters 9 and fO, particularly figure 1.0-2). In an animal with primary consciousness, a balancemust be struck between responding to intemally determined salienceand extemally produced novelty. With higher-order consciousnessthe situation becomesmuch richer. Volitional statesrelated to the selection of plans, values,and temporal projections can all change the relative contribution of different parts of a global mapping. In both cases,large portions of the nervous systemare likely to be involved when a global mapping is modified to alter attention. According to this view, we would expect attention to be altered by changesin severallevels of a global mapping: by unconsciousas well as by consciousactivity. How, specifically,could such a system work? Any model proposed to explain attention must accountfor its selectivity;for the fact that, after an animal leams a skill, it becomesautomatic;for the ability to intemrpt automatic acts by attention to novelty; and for the ability to direct attention specificallyby consciousmeans. We all know that consciousattention plays a large role in the leaming of complex skills.But in many cases,successfulleaming allows us to carry out skilled actions without attention. Performancethen remains unconsciousuntil either novelty or threat makesadditional demands.Remember that, in presentingthe TNGS, I suggestedthat the basalganglia were major organs of succession,acting with the cortex to choosemotor plans.Motor plans, which may be consciouslyformed in humans,are executedvia the motor cortex as it sendssignals to the spinal cord. But the output of the cortex is also routed to the basal ganglia. These structureshave only an indirect connection back to the cortex, but it is a very significantone. The output from the basalganglia is inhibitory, and thereforeit can alsoinhibit inhibition.In other words, it candisinhibit target areasin the cortex. This either excitesthem or preparesthem for excitatory input, a stateimportant for attention. 142

Attention and the Unconscious In accord with a given plan, the basal ganglia selectively disinhibit thalamic nuclei projecting to the cortex. This leads to anticipatory and selective arousal of cortical areas corresponding to the motor program. These cortical areasthen become more sensitive to those sensory inputs that are consistentwith the performanceof the task via a global mapping. Such a mechanismcan explain focused attention. What about automatic activity intemrpted by novelty? If the task is not completed within a certain time, or if a novel event is detected and categorized,"alarr:." signalsmay passdown to the midbrain value systems that connectback to the cortex and the basalganglia. These systemsmay then send back signalsto intemrpt the motor plan in the cortex and block the execution of a motor program. As long as an automatized action is accomplishedwithout a hitch these midbrain nuclei are not engaged. Otherwise, as in the caseof a shouted "watch out!" during a conversation while driving, they will causea shift in attention to occur. alter attention and alter priorities in the But how carr conscior,rsness it mappings? In the caseof primary consciousness, global construction of can alter attention and priorities by a change in the salienceamong the parallel reentrant loops connectedto the basalganglia in a processsimilar verbal to the one outlined above.In the caseof higher-orderconsciousness, schemasin conceptualareascan,through the activities of the frontal cortex and limbic system, dominate the apportionment of disinhibition by the basal ganglia, whicl have strong connectionsto such regions. The fragility of attention is a particularly interestingissue.How is it that consciousattention is so narrow-usually able only to focus on one or at most two targets at the sametime?An answer is suggestedby the motor theory, which looks at attention as arising from evolutionary needs.Motor plans and programs are more or less exclusive (that is, they will not accommodate contradictory actions that are simultaneous).Moreover, given the large amount of nervous tissueinvolved in eachglobal mapping, it seemsunlikely that one could sustainmore than a few complexmappings at the same time without their interfering with each other. Such a view of attention still concedesa major overriding significance to nonconsciousmechanismsand to the orienting behaviors mediated by global mappings in responseto emergencies.Yet becausehaving intentional consciousstatesdependson values,categories,and memoriesas well as plans,this selectionalview of attention allows us the ability to entertain consciouslyan "intention to attend" to what is planned or envisioned.But this capacity is always subject to competition from unconsciousand nonconscious elements (the latter being those that can neoerbecome conscious).We are all aware of parapraxes-slips of the tongue-and of 743

PnoPosALs actions committed "not as planned." These suggest the intervention of unconsciousprocesses.

THEUNCONSCIOUS Freud(figure 13-1) was the singlemost important figure in pointing up the role of unconsciousprocessesin our behaviorand feelings.In his "Project

FIGURE13-1 SigmundFreud(1856-1939),founderof psychoanalysis and erplorerof the mechanisms of repressionin memory, 144

Attention and the Unconscious for a ScientificPsychologyi'he tried to write an explicit neuralaccountof the relation betweenconsciousand unconsciousprocessesand behavior, but soonabandonedthe attempt.His laterformulationswerepsychological explanationsof behavior that emphasizedintentionality but were at the sametime ruthlesslydeterministic. The postulation of an unconsciousis a central binding principle of Freud's psychologicaltheories.Since his time, ample evidencehas accumulatedfrom the study of neurosis,hypnotism,and parapraxesto show that his basicthesesabout the action of the unconsciouswere essentially correct.As he used it, the term unconsciousreferredto elementsthat can be easilytransformedinto consciousstates-"the preconscious"-a5well as those that can be transformed only with great difficulty or not at all-"the unconsciousproper." Freud suggestedthat threatening events could be repressedin memory so that they were not ordinarily available for consciousrecall. We must not forget that theseare psychological,not structural,terms. My late friend, the molecularbiologist JacquesMonod, used to argue vehemently with me about Freud, insisting that he was unscientificand quite possiblyu charlatan.I took the sidethat, while perhapsnot a scientist in our sense,Freudwas a great intellectualpioneer,particularlyin his views on the unconsciousand its role in behavior.Monod, of stern Huguenot stock,replied,"l am entirely awareof my motives and entirely responsible for my actions. They are all conscious."In exasperationI once said, "Jacques, let'sput it this way.EverythingFreudsaidappliesto me and none of it to you." He replied,"Exactly,my dear fellow." Freud'snotion of repressionis consistentwith the modelsof consciousnesspresentedhere.The extendedTNGS strongly implicatesvalue-dependent systemsin memory formation. Self-nonselfdiscrimination(seefigure 11-1) requiresthe participationof memory systemsthat are forever inacRepression,the selective inability to recall, cessibleto consciousness. would be subjectto recategorizationsthat are strongly value-laden.And given the socially constructednature of higher-orderconsciousness, it would be evolutionarily advantageousto have mechanismsto repress thoserecategorizationsthat threatenthe efficacyof self-concepts. Circuitry that interactswith value systemsexistsin the hippocampusand the basal ganglia.In a linguisticanimal,symbolsmatter,and the evolution of a way of reducing accessto states consideredthreateningto the self-concept would have selectivevalue.This providesa great clue to the propertiesof emotions,a subjectto be touched on in a later chapter. My generalconclusion,important for all theoriesof mind, is that given the existenceof acts driven by the unconscious,conclusionsreachedby 745

P n o p os e r s consciousintrospection may be subject to grave error. In other words, Cartesianinconigibility is incompatiblewith the facts.Descartes,an adult genius with mastery over language,did not take several things into account. The first is the developmentallydeterminednature of higher-order (Recallthat Frenchbabies,even gifted ones,are unlikely to consciousness. 'Je assert, pense,donc je suis.")The secondis that his linguistically based consciousnessis nof self-sufficientand beyond doubt. Given that it is linguistic, it is always in dialogue with some "other," even if that interlocutor is not present. The third is that unconsciousmechanismsblock and intervene with what we consider to be transparentand obvious lines of thought. By opening up the questionsthat prompted his method of doubt and by fearlesslyexposing his thoughts on the nature of the mind, Descartesbecamea great pioneer of modem philosophical and psychological investigation. Since the announcement of his method, however, accumulatedknowledge hasforced us into a much humbler posfure about the certainty of what we know. This is probably a good point at which to state again how little it is we actually know. Given the difficulty of imagining the mind and the complex layerings of its wo*ings and processes,we should not be surprisedat this. But we have many resourcesopen to us now that Descartesdid not have. They should, in the end, allow us to appreciatehow fruitful his "wong" theory has been in stimulating our attempts to understandthe mind.

1,46

CHAPTER

I1

LayersandLoops:A Summary

It seemsto me that the humanracestandson the brink of o major breakthrough,We haoeadoancedto the point wherewe can put our hand on the hem of the curtain that separatesus from an that we understanding of the natureof our minds,ls it conceioable will withdrar,Dour hand and turn back throughdiscouragement and lack of oision? -PercyWilliamsBridgman

t is high time for anotherview of the mental,for a neuroscientificmodel of the mind. What makesthe one proposedhere new is that it is based remorselesslyon physicsand biology. It is also basedon the ideasof evolutionary morphology and selection,and it rejectsthe notion that a syntacticaldescriptionof mental operationsand representations(seethe Postscript)sufficesto explain the mind. Others have held similarpositions but have not united them in a single evolutionarily basedtheory, one that connectsembryology, morphology, physiology, and psychology. Only such a physically based theory of mind is open to disconfirmationby scientificmeans. The road connectingthesedisciplinesis a bumpy one and,asthe readerhas see& traveling the route is occasionallystrenuous.This is because,in the constructionof the mind, so many levels of organizationarerequiredand so many interactive loops have to be made to link what at first appearto be disparatelayers of description.Given that the mind is a result of evolution and not of logical planning, I would not expect a different outcome. This r47

PnoPosALs profusionof levels,and not someesotericnew principle,whether of physics or of theosophy,is why it is difficultto think aboutthe mind.The braingiving rise to the mind is a prototypical complexsystem,one more akin in its style of constructionto a jungle than to a computer.This analogyfailsat onepoint: While plants in junglesare selectedfor during evolution, the jungle itself is of selection,naturalselection not. But the brainis subjectedto two processes and somaticselection. The result is a subtle and multilayered affair,full of loops and layers. From genesto proteins,from cellsto orderly development,from electrical activity to neurotransmitterrelease,from sensory sheetsto maps, from shapeto function and behavior,from socialcommunicationbackto any and all of these levels, we are confronted with a system of somatic selection that is continually subjectedto natural selection(figure I4-I). Is it any wonder that philosophers,thinki.g about the problem of the mind without this knowledge, were tempted to postulate entities, that physicistshave been tempted to postulateexotic new material fields,and that those in hope of immortality continue to postulate etemal spirits? It may come as a disappointmentto such thinkers that the answersto many of the fundamentalproblems of mind will come from analyzingthe complexity of its organization, which is governed by novel ordering principles.But consideredagain,how rich, how full of surprise,how much

LAYERS Genetic Code

p

Particles+ Fields Atoms Molecules

Tissues Organs Neuronal H Organisms Group r-+ >> Sefeition7Humans

Actions s- Societies l/

Physics-€cosgology

re

Natura.f7""u" selection

Gultureq)

LOOPSOf KNOWLEDGE

Languages Mathematics Art Religions Science

(

]Chemistry

(

Geology

\Biorosy

(Evolution)

,l {""'T'Tce\ -

Psychology

Anthropology, | Psychiatry,etc. |

Linguistics
\

/ ,/

FIGURET4-1 Layersof biologicalorganizntionand loopsof knowledge,Seefigure P-l for the scales affectingthe layersand loops. -1,48

L a y e r s a n d L o o p s :A S u m m a r y more of a piece with the grand and compatible theories of evolution and physics this altemative is! The untangling of the complexity has barely begun and the present effort at synthesiswill undoubtedly seempaltry when it is over. But even at its early stages,the whole businessof the matter of the mind requires a global view if we are to get anywhere.A simple analogicalcomparison will not do, nor will rationalism,nor will physics alone. A theory of the matter itself is needed.A major aim of the one describedhere is to provoke the construction by others of altemative theories within the same constraints-not philosophical hypotheses,not highJevel formulations, but biological theories that challengeeither the facts I have presentedor the ways in which I have interpreted them. In the meantime, I hope the reader will welcome a surnmary of the layers and the loops from my point of view. By now I assumethat the strange vocabulary is familiar enough to be used here to sum up my position. I will proceed in reverse,from the theory adopted to my criticisms of the altematives.Pleasekeep in mind the generalizationthat as a selective system, the brain (especiallythe cerebralcortex) is a correlator. It conelates temporal inputs during its own development, and it correlates the properties of signals and scenesin its adult functioning to give rise to consciousness. My ftrst premise is that consciousnessappearedas a result of natural selection.The mind dependson consciousnessfor its existenceand functioning. A related notion is that consciousnessis efficacious,enhancing fitness in certain environments.Consciousnessarisesfrom a specialset of relationshipsbetween perception,concept formation, and memory. These psychologicalfunctions dependon categorizationmechanismsin the brain. In additioo memory is influencedby evolutionarily establishedvalue systems and by homeostatic control systemscharacteristicof each species. Primary consciousnessis achievedby the reentry of a value-category memory to current ongoing percepfualcategorizationsthat are carried out simultaneouslyin many modalities.It links parallelstimuli in time and space (including those not necessarilycausallyconnected)into a correlatedscene. In an individual animal,the feafuresof that sceneachievesaliencefrom that animal'spast values and leaming history. Primary consciousnessis limited to the rememberedpresent. It is necessaryfor the emergenceof higherorder consciousness,and it continues to operate in animals capable of higher-order consciousness. Higher-order consciousnessarises with the evolutionary onset of semantic capabilities, and it flowers with the accessionof language and symbolic reference.Linguistic capabilitiesrequirea new kind of memory for r49

Pnoposlrs the production and audition of the coarticulatedsounds that were made possibleby the evolution of a supralaryngealspace(seefigure 12-1). The speechareasmediating categorizationand memory for language interact with already evolved concepfualareasof the brain. Their proper function in a speechcommunity connectsphonology to semantics,using interactions with the conceptualareasof the brain to guide leaming. This gives rise to a syntax when these sameconceptualcenterscategorizethe ordering events occurring during speechacts.As a syntax begins to be built and a sufficiently large lexicon is leame4 the conceptualcenters of the brain treat the symbols and their referencesand the imagery they evoke as an "independent"world to be further categorized.A concepfualexplosionand ontological revolution-a world, not just an environment-are made possible by the interaction between conceptualand language centers. By these means,concepts of self and of a past and a future emerge. Higher-order consciousnessdependson building a self through affective intersubjectiveexchanges.Theseinteractions-with parental figures,with grooming conspecifics,and with sexualpartners-are of the samekind as those guiding semiotic exchangeand language building. Affectively colored exchanges through symbols initiate semantic bootstrapping. The result is a model of a world rather than of an econiche,along with models of the past, present,and future. At the same time that higher-order consciousnessfreesus from the tyranny of the rememberedpresent,however, primary consciousnesscoexists and interacts with the mechanismsof higher-order consciousness.Indeed, primary consciousnessprovides a strong driving force for higher-order processes.We live on severallevels at once. The Jamesianproperties of these consciousprocesses(seechapter 1.1) dependon the function of the cerebralcortex and its appendages.The latter constitute the organs of succession-the cerebellum for smooth movement, the hippocampusfor laying down long-term memory, and the basal ganglia for choosing motor pattems and attentional plans.Their functioning depends on the motion and action of the organism exploring its environment. The resulting properties of subjectivity, intentionality, continuity, and change occur together in an apparent unity. These properties can be explainedby the extendedTNGS, requiring no assumptionsbeyond those of developmentalselection,experimentalselection,and reentry. New functions, including consciousness,are made possible by new evolutionary morphology connectedin new ways to existing brain strucfures. "Objective" scienceand languageboth depend on the metastability or constancy of objects in the physical world. The consciousnesstheory 150

L o y e r s a n d L o o p s :A S u m m a r y assumesthat physicsand evolution, supplementedby the assumptionsof the TNGS, aresufficientto constructa scienceof mind. No scientifictheory of a single actual mind is possible,however, any more than a scientific accountof all historical events in the world is possible. To remain scientific,the extended TNGS must assumethat both the human subjectand the human scientificobserverwho studiesthat subject experiencequalia.This assumptionis necessaryto assurethat meaningful intersubjectivescientificexchangeoccurs.According to the theory, qualia are categorizationsby higher-order consciousnessof the "scenes" and "memories"provided by primary consciousness. They involve recategorical relationshipsthat are ultimately governed by how evolutionarily selected values interact with memory. Creaturesthat have primary consciousnessalone can neither report qualianor reflect on them. If they experiencethem (and we can only infer that they do), they experiencethem solely in the rememberedpresent.On the basisof moqphologicalcomparisons,we may conceiveof three levels of sensorypropertiesin the evolution of animalswith n€urons: 7 . Responsesto stimuli with aversive and consummatory responses directly governed by selection for evolutionary values. An example is the lobster, capableof learning and long-terrn memory but not of primary consciousness. 2. Stimuli eliciting responsesin animals having primary consciousness. Mental life consisting of Jamesianscenescorrelating value and perceptual categorization, but no socially constructed self. Anatomical bases for qualia and their discrimination according to different modalities. No categorization of qualia over time by r subject,but long-term memory (nonconscious as such) based on qualia in the remembered present. Example: dogs. 3 . Stimuli with aversive and appetitive significance transformed by animals having higher-order consciousnessinto a world, not just an econiche. Full-blown qualia capable of being refined, remembered, altered, and reported, as in wine tasting. Examplerhumans. Extreme example: sainthood, with denial of all biological imperatives, including unusual responsesto painful qualia, on the basis of deeply held belief.

For now, we can only speculateon suchmatters.But we do know that higher-order consciousnessleads to the construction of an imaginative domain, one of feeli.g, emotion, thought, fantasy, self, and will. It constructsartificialobjectsthat are mental.In culture,theseactslead to studies of stablerelationsamong things (science), of stablerelationsamong stable 151

PnoPosALs mental ob;ects(mathematics), and of stable relationsbetween sentences that are applicableto things and to mental objects (logic).One possible reasonfor the incompletenessof suchdomains,as shown for mathematics by Kurt Godel, is that pattern formation in the mind always requiresthe higher-order bootstraps that are necess ary for consciousness. Thinking occursin termsof synthesized patterns,not logic, and for this reason,it may always exceedin its reachsyntactical,or mechanical,relationships. The analogy between the mind and a computer fails for many reasons. The brain is constructedby principlesthat assurediversity and degeneracy. Unlike a computer,it has no replicativememory. It is historicaland value driven. It forms categoriesby intemal criteriaand by constraintsacting at many scales,not by meansof a syntacticallyconstructedprogram. The world with which the brain interacts is not unequivocally made up of classicalcategories.(lt is true,however,that some"naturalobjects"appear to follow thesecategoriesbecauseof the interactivefeaturesof our phenotype and the physicalpropertiesof theseobjects.) The world, therefore,is not like a pieceof computertape.Physics,which studiessuch a world, describesits formal correlativepropertiesbut does not contain a theory of unique categoriesfor the partitioning of macroscopicobiects.As I point out in the Postscript,objectivismfails. Categorizationmechanismswork through global mappingsthat necessarily involve our bodiesand our personalhistory. Perceptionis therefore not necessarily veridical (see,for example,the Kanizsatriangle in figure 4-2).ln our behaviorwe are driven by u recategoricalmemory under the influence of dynamic changes of value. Beliefs and concepts are individuated only by referenceto an open-endedenvironment,the description of which cannotbe specifiedin advance.Our modesof categorization and the use of metaphorin our thinking (mappingone thing to anotherin a different domain) reflect these observations. I argueat length in the Postscriptthat the cognitive scienceview of the is ill-founded. mind basedon computationalor algorithmicrepresentations Mental representationsthat are supposedlysyntacticallyorganized (in a "languageof thought") and then mappedonto a vaguely specifiedsemantic model or onto an overly constrainedobjectivistone are incompatiblewith the factsof evolution.The propertiesproposedby thesecognitivemodels are incompatiblewith the propertiesof brains,bodies,and the world. The extendedTNGS puqportsto explainhow embodimentof mind takesplace and thus connectscognition to biology. It providesa consistentbasisfor explaininghow meaningarisesfrom embodimentas a result of referential interactions.A rich field of study concernedwith exactly how our concepts map onto our bodies is presentlyin its earlieststages. 1,52

L o ye r s a n d L o o p s : A S u m m n r y Why have I rejected as a basis for mind the apparent eleganceof axiomaticand syntacticsystems? Axiomatic systemsoften seemto provide the right clue as to how the mind works, especiallywhen taken together with physics. But they are social constructionsthat are the resultsof thought,not the basisof thought. Their roots lie in the mathematicallogic of the nineteenthcentury.They flowered with David Hilbert, were modulated and circumscribedby Godel,and are often conceivedof in a typological or essentialist fashion.They arenot a good model for the mind, for the mind must preexisfto createand drive them. Consciousness is essentialfor their formulation and also for the Platonismthat they sometimesinspire, but the factsshow that consciousness aroseby evolutionary,not typological,means.Darwin was right: Morphology led to mind, and on this issue Wallace,who felt that naturalselectioncould not explain the humanmind, was wrong. Plato is not eT)en wrong; he is simply out of the question. It may be useful here to mention the obvious-that the evolution of dependedon certain temperatures.The stability of any consciousness physicalobject that scienceand the consciousscientistdescribeis the same stability that "glues together" a naming event. There could be no consciousnessat 1O6oC.It emergedat a certain time, place,and at a much lower temperature,one that allowed chemistry to occur. To say so is to reject panpsychismas a theory of mind. (l discussthis matter further in chapter 20, which considersthe ultimate origins of mind.) is central to human behavior, society, language,and Consciousness science.Imaginethe oppositeand you haveto postulatea prescribedworld tape, a "brain-computer,"and a very boring "world programmer." The TNGS, with its complexity of layersand loops,appearsto be more in line with the facts of biology and seemspreferablebecauseit fits much more of our own experience. With this statementof an obvious personalpreference,I turn to the final part of this book. I have entitled it Harmoniesto underscorethe fruitful interactionsthat a scienceof mind must have with philosophy,medicine, and physics.Thesefields are all different,but truly interestingharmonies arebasedon the consonanceof differententities,not on identity or unison. We all hope for a resolution of conflicting visions, for clarification of thought, and for harmoniesbetweenideas.I am no exceptionand I have no intention of missingthe chanceto philosophize,first, about philosophy itself;second,about the ideaof the self,its thoughts and its disorders;third, about the possibility of making consciousartifacts;and last, about the grand themesof a future sciencethat will reveal more clearly the connection betweenphysicsand psychology. In Modes of Thought,Whitehead pointed out that philosophy is the 153

PnoPosALs attempt to make manifest the fundamentalevidenceas to the nature of things. In the same work, he remarkedthat scientificreasoningis completely dominatedby the presuppositionthat mental functioningsare not properly part of nature.He deploredthis and hoped that a proper connection betweenthe mental and physicalcould be forged within scienceitself. That was in 1933.Now, in the !990s, a glimpseinto how that could come about may be possible without closing the door on philosophy, which above all is an attitude of mind.

1,54

PART IV

HARMONIES

This final set of chapters asks about the implications of our new brain theory for human (and some inhuman) concems.It pleads for an openmindednessabout the mind. It suggeststhat our knowledge is not incorrigible, that we are deeply embeddedin the matter of the world as well as in the matter of the mind, that we are eachof us unique as individuals (and importantly so), that our thinking in a culture is a critical matter for our being human and for our grasping of meaning, and that, even in disease, our minds are marvelously adaptive. It also suggeststhat the time is not hopelesslyremote when we may be able to build artifactsthat sharesome of our own psychological properties. Above all, it suggeststhat constructing an adequatetheory of the brain promises to offer basesfor new harmonies,including those according to which we may placeourselvesin the universe.In the ftnal chapter,I attempt to answer the question:If one were to name two grand scientificideas or concepts that together capture how we may ground ourselvesand help determinewhere we are in the order of things, what would those ideasbe?

155

CHAPTER

15

A Graveyard of Isms: andlts Claims Philosophy

can tell eachother all they know in tuto Any tuto philosophers hours. hedoesn'tseea beard I don'tseewhy a manshoulddespairbecause on his Cosmos.If he belieoesthat he is insideof it, not it inside and ideals purpose,significance, him, he knotpsthat consciousness, are among its possibilities. , , and the businessof philosophyis luant to do. to shor,nthat we are not foolsfor doing uthat T,pe -Oliver WendellHolmes,Jr.

concernwith the mind and its workings has perrneatedphilosophy from its beginnings.In his review of the conceptsof philosophy, Arthur Danto definesalmost all philosophicalpositionsin terms of what he callsa basiccognitive episode.This notion harksback to Descartesand is expressedas a relationshipbetween three components:a subject, a representation,and the world. The relationship between the world and the subject is that of causality.The relationshipbetween the world and representationis that of truth, and the relationshipbetween the subjectand representationis that which the subjecthaswith him- or herself. Danto calls humansrepresentingbeings or representationalbeings,and I believe he falls into the trap we have already warned against when he advancesthe view that the body is "sententiallystructured,"meaningthat is "to show how nervous tissuerepresents." the task of the neurosciences If one doesnot hold him too closelyto this view, which skirtsdangerously 157

HlnuoNrrs close to ideas of coding and instruction, his triad is a useful one, for it can help us seehow, as individuals, are(not our brains) "represent" the world. To a scientist, philosophy can be a disconcerting business.Scienceis supposedto provide a description of the laws of that world and of how they may be applied.Philosophy by contrast,has no proper subjectmatter of its own. Instead,it scrutinizesother areasof knowledge for clarity and consistency.Furthermore,unlike science,it may be calledimmodest.There is no partial philosophy; it is complete with eachphilosopher.Like a child exploding into a grasp of language, the philosopher must not simply describean environment but construct a whole world. Each time a philosophicalconstruction is attempted, there is a world view behind it, and a personalone at that. The entitling of such world views as various "isms" (table 15-1) leads to an interesting collection of ways in which Danto's triad hasbeendissectedwith respectto the importanceaccordedits intemal relations. I will not discussthem all here; the shallow account given in chapter4 should give you a feeling for a few of them. Readersmight want to amusethemselveswith a dictionary or an encyclopediaof philosophy to checktheir ramifications.The trouble is that each "ism" is likely to spell the rejection of the last, as each philosopher constructsa unique point of view. Philosophy is a graveyard of "isms." Why bother with it then?Becausephilosophy attemptsto apply thought to all aspectsof our individual and collective existence;becauseits history

TABLE I5-7 SomePhilosophical"Isms"* Empiricism Rationalism Phenomenalism Reductionism Objectivism Operationalism Instrumentalism Logical positivism Foundationalism Pragmatism Evolutionism Selectionism

Monism Dualism Pluralism Epiphenomenalism Materialism Panpsychism Determinism Compatibilism Incompatibilism Occasionalism

Realism Idealism Foundationalism Essentialism Behaviorism (Philosophicalbehaviorism) Representationalism Functionalism Interactionism Internalism Externalism Existentialism

.This list might go on beyondall decentboundszt)ereone to includemoral, aesthetic,clinical, religious,and political ideologies. Isms mostdefinitelyruledout by scientificstudy includegeocentrism, uitalism,and mechanism.Of course,not all doctrinesare isms(andpossiblyuiceuersa)but they could be madeto be, and that is the danger.

158

A G r n o e y a r d o f I s m s : P h i l o s o p h ya n d l t s C l a i m s is closely intertwined with that of psychology; and becausea new scientific view of the mind basedon biology may help give philosophy a new lease on life. An anonymousuniversity presidentis quoted by lohn Barrow and Frank Principle: Tipler in their book TheAnthropic Cosmological Why is it that you physicistsalwaysrequireso muchexpensiveequipment? Now, the Departmentof Mathematicsrequiresnothing but money for paper,pencils,andwastepaperbaskets,andthe Departmentof Philosophy is better still. It doesn'tevenaskfor wastepaperbaskets. Contrast this with Einsteio who remarkedthat the theoreticalphysicist's most important tool is the wastebasket! Philosophy in this century has been claracterizedby a retreat from the grand synthetic goals of its past, goals touched on in chapter 4. Since Wittgenstein, a good part of philosophy has been concemedwith tidying up logic and language. Since Edmund Husserl, another part has been concemedwith a deliberatelynonscientificset of reflectionson consciousnessand existence,or phenomenology,as it is called.It is certainly worth askingwhether a biologically basedtheory of mind would invigorate these areasof thought and perhaps even give philosophy a new tum. Let us join the game and see how many isms must fall if we take a scientific position on the mind. Of course, this means saying something about the limits of scienceand of knowledge itself. We should first state the assumptionsof a scientificview: There is a real world-one describedby the laws of physics, which apply everywhere. (This is the physics assumPtion.) 2 . We are embedded in that world, follow its laws, and have evolved from an ancient origin. The mind arose on the basis of new evolutionary morphology. (This is the evolutionary assumPtion.) 3 . It is possible to put the mind back into nature. A science of mind based on biology is feasible.The way to avoid vicious circles and dead ends is to construct a brain theory based on selectionistprinciples. (This is the central argument of this book.) L

If we accepttheseassumptionsand the previousargumentsof this book, we can immediately add to the graveyard. Dualism, panpsychism,epiempiricism,and essentialism idealism,representationalism, phenomenalism, are all incompatiblewith both the assumptionslisted above and the evios well as biology itself. I shall dencefrom psychology and neuroscience, 159

HenraoNrss not belabor the argumentsbut simply state,in the samesequenceas shown in this list, the following presumably lethal sentences:There is no res clgitans;particlesare not conscious;consciousnessis evolutionarily efficacious; the world exists and persistsindependent of mind and preexisted before its appearance;the brain is a selective system and not a Turing machine;sensedata are not the basis of the mind; the "world" does not consistof classicalcategories;typology is destroyedby biology. And over the last 300 years,sciencehas already destroyed the more parochialideas of geocentrism,vitalism, and simple mechanism. So much for destruction, at least for the moment. What can we say constructively about science and about the possibility of a theory of knowledge based on biology-a biologically based epistemology? To construct a reasonableaccount, we must recognize that modem particle physics and field theory have eliminated the notion of the world as a deterministic or clockwork mechanism.This does not mean that mechanismscannot be describedor be useful (as they are to both macroscopic physics and biology). It meanssimply that the universe cannot be sensiblyconsideredat all scalesin suchterms (seeftgure P-1). We must also recognizethat a death blow was dealt to essentialismand Platonismby the Darwinian theory of evolution. Finally, we must take into accountthe fact that systemsof natural selectionin evolution (which are historical systems) gave rise to somatic selective systems capable of dealing with novelty within an individual's lifetime. This last point is not so securelybased-the idea that there are sciences of recognition, of which neuroscienceis a central one has not yet been generally acceptedfor studiesof the brain. But if we assumethat the main premisesof neuralDarwinism are correct (and evidencein support of them is mounting), severalinteresting conclusionsmay be drawn. First of all, we need not reachbeyond biology itself to mount any exotic explanationsof the mind. Rememberthat the assumptionsof the TNGS-developmental selection and variance, synaptic selection, and differential amplification within reentrant systems-are all of the principles proposed by that theory. No new principles need be adducedto account for consciousnessonly new evolutionary moqphologies.Second,thesenotions, if conect, rule out a gmeral description of the workings of the brain as a Turing machine or computer.And third, while substancedualism(the Cartesianvariety) and property dualism (the notion that psychology can be satisfactorily describedonly in its own terms)are ruled out, we must admit to a distinction between selectiveand nonselectivematerialsystems.This distinction identifies living and mental systemsas selective.In other words, there is a real distinction between biology (or psychology) and physics.While admitting r60

A G r a o e y a r d o f I s m s : P h i l o s o p h ya n d l t s C l a i m s that the laws of physics apply to both intentional and nonintentional systems,this position at the sametime deniesthat fancy physics----*uchas quantum gravity or other specializedconceptsof fundamentalphysicsare required to explain mind. Where, after this, do we stand with the isms? By taking the position of biologically based epistemology, we are in some sense-realistsand also sophisticatedmaterialists.Given the facts of developmentand evolution, we deny teleology (the doctrine of final causes or ultimate goals).But at the sametime, we admit that evolution canselect animalsin such a way that they have generalgoals, purposes,and values, so that they embody what have been called teleonomic systems.As Mayr put it, teleonomy is a prediction of the past. The past experienceof natural selectionadjuststhe set points of value systems(for example,those related to hunger, thirst, sexual responses)that are adaptive for survival. In our case,the brain of a conscioushuman being, serving as a somatic selective system,usesvalue constraintsto project the fufure in terms of categories and goals. And by assumingthat the brain is a somatic selectivesystem, we rule out the idea of the little man or homunculusin the head.He is no more necessaryto the sciencesof somaticrecognition than specialcreation or the argument from design is to evolution. If he is ras cogitans,he is exorcised. Mind, which arose from material systemsand yet can serve goals and purposes,is neverthelessa product of historical processesand of valuebasedconstraintsrelated to evolution. What bounds does this place on our knowledge and our freedom? arosein the material order does not restrainintellecThat consciousness tual trade; philosophy itself is witness to this conclusion.But it does limit us, despiteour capacityto extend our sensesand our Powersof calculation through physical devices.Given how meaning is defined in this book, we must accepta position of qualifiedrealism.Our description of the world is qualifiedby the way in which our conceptsarise.And although there may be infinite freedom within a grammar, our language and our ideas of meaninggo far beyond the rules of grammar.I have alreadydescribedhow I think meaningarisesfrom embodimentthrough neuronalgrouP selection and reentry. Despite the remarkableextensionsof meaningby our calculations and our experiments,we must admit that we may well be limited in our thought by the way in which we are constituted as products of evolutionary morphology. We can add three more important elementsto this picfure of qualifted realism and biologically based epistemology. These are (1) an extraordinary density of real-world events exists;even given our ability to catego76r

HeRMoNrEs nze a large number of them, we can hardly exhausttheir description;(2) many events are irreversible;and (3) in each individual, sensationand perceptionfollow unique, irreversible,and idiosyncraticcourses. According to biologically based epistemology and qualified realism, knowledge mustremainfragmentaryand corrigible.There is no Cartesian certainty.But, you may object,what of mathematicalcertainty,of analytic relations and tautologies?This is no place for an extendeddiscussionof thesematters.It is useful,however, to make it clearthat such systemsare artificialones,createdby the mind through socialinteractionsand individual manipulationsof symbols.The most basicof thesesysteffis,arithmetic, hasbeen shown by Godel to be incomplete.I would characterizethe study of mathematics,?s Philip Davis and Reuben Hersh aptly put it in The MathematicalErperience, as the study of stableor invariant mental objects. Although our subtlematerialismhardly assertsthat theserelationsare not meaningful,it deniesthem a separatePlatonic existence. We have already describedsome other limits. For example,given the limitations on our knowledge and our "locked-in" state,the exact idiosyncratic and irreversiblepath of an individual's qualia is not accessibleto another individual. (lndeed,it can becomeinaccessible even to its owner). Moreover, we must accept that death means the irrevocableloss of an individual and of that individual'sbeing.Death is not an experiment:There is nothing to report. Minds do not exist disembodied.Theseare obvious limits to the knowledge of personsand to science.What we are trying to do, nonetheless,is placeourselvesclearly within the world view given by science. In addition to qualifying our realism,w€ must consider questionsof history and culture and onesrelatedto value and purpose.This may seem strangein a discussionof science,which is supposedto be value-free.But the sciencetouted as value-freeis that basedon the Galileanposition, a physical sciencethat quite deliberatelyand justifiably removed the mind from nature.A biologically basedepistemologyhas no such luxury. It is worth dwelling on this matter a bit, for it revealsmuch about the placeof sciencein our lives. Conscioushumanexperiencehas given rise to culture, and culture to history. History is not simply a chronicle but an interpretation, encompassingsuspectedcausesand values. Sciencehas emerged within history, and it attempts to describe,with considerably more certainty,the boundariesof the world-its constraintsand its physical laws. But these laws cannot replacehistory or the actual coursesof individual lives. A set of laws is not a substitutefor experienceand it is certainlynot equivalentto a set of events.Laws do not and cannotexhaust experienceor replacehistory or the eventsthat occur in the actualcourses 1.62

A G r a o e y a r d o f l s m s : P h i l o s o p h ya n d l t s C l a i m s of individual lives. Events are denser than any possible scientific description. They are also microscopicallyindeterminate,and, given our theory, they are even to some extent macroscopicallyso. It may seem that I am attempting to limit a pioi the capabilitiesof scientificdescription.Nothing of the sort. I am simply pointing out that, even if sciencesucceedsin putting the mind back into nature, it will not, according to the description we have given, be able to describeindividual or historical experience adequately. But it does provide a satisfactory (indee4 the best) description of the constraintson experience. What are some of those constraintswhen viewed through a selectionist theory of the mind? One comesout of the extendedTNGS: No selectionally basedsystemworks value-free.Valuesare necessaryconstraintson the adaptiveworkings of a species.In our species,the commonalitiesof physiological function, hunger,and sex imply a set of mutually sharedproperties' The brain is structured so as to play a key role in regulating the evolutionarily derived value systems that underlie these properties. Undoubtedly, thesevalue systemsalso underlie the higher-orderconstructionsthat make up individual aims and purposes.We categorizeon value. arose,values at the biological level Once higher-orderconsciousness could be modified, though only to a degree. Of course, ?s mentioned earlier,w€ must admit the possibilityof an almosttotal denialof biological values on the part of those organismswe call martyrs and saints.Only can so transcendthe creaturesendowed with higher-orderconsciousness dictatesof biology. If one agreesto omit saintsfrom consideration,the insertion of aims and purposes and ethical values into social systems, however far they are from basicbiological value systems,almost certainly resultsfrom the original need for value in guiding the selectionalsystems of the brain. In any culture,decisionsinvolvi.g socialvalue must come before those that elevate the interests of science,however important scientificknowledge may be. Sciencehas tumed out to be eminentlypractical,as it must be, given its serviceto the verifiabletruth. Modem society and its economicsdepend increasinglyon scientifictechnology,and scientificbeliefsarebeing assimilated by increasingnumbersof people.In addition to curiosity, however, greed now often drives the searchfor knowledge.The good to be derived from that search,however motivated, is that sciencemay usefully transform our material conditions of being (provided we remain clear about values).But a constant tension remainsin balancingprivate and public good, as the history of this industrialand atomic century has dramatically taught us. Power is not insight, and the shift from a sciencebasedsolely on physicsto one basedon physicsand biology is likely to lend us deep 1,63

HenuoNrrs insight and to help change how important social decisions are made. A biologically based epistemology will have valuable things to say about such decisionsas we discover more about our brains. Knowing our place in the world on the basis of a biologically based theory of the mind will also reveal our limits and restrainow philosophical ambition. But in certain directions the limits are hardly very constraining. The imagination of conscioushumans in culture is potentially limitless. Constrainedas we are (eachof us locked into our own consciousexperience), mortal as we may be, and qualified as our realism is, the future remains open; it is not predetermined.We do not have the security of foundationalism or a ftrst philosophy, nor the ability to know with certainty all that we can appreciateor place into a pattem. Most of us cannot deny our evolutionarily selectedbiological values, nor should we, given that they provide a common ground for our moral decisions. But the history of scientificdiscovery and the achievementsof human imagination promise constant surprise and, with the rise of brain science,provide an increasingly solid basis for attempts to place ourselves within our own world description. In the end, therefore,we must concludethat we have not been able to kill all the isms.We have suggesteda favored set: qualifiedrealism,sophisticated materialism,selectionism,and Darwinism. Indeed,consideringtheir significanceand relating them to what physics and biology together have to offer should enrich philosophy and ensure its harmony with science. After all, thinking is not the sameas a theory of mind, and there is much thinking to be done about selective systems. By its very nature, the position on biologically basedepistemology that I have taken here implies that science-freephenomenology and grammatical exercises,whatever their value, placetoo narrow a set of limits on the philosophicalenterprise. Philosophy needs a new tum. I believe that neurosciencewill play a centralrole in sucha development. But it will not be a development in which a simple reductionism to quantum fields,to strangeparticlesor the like has any sway. It will instead be one in which the task is to see how selectionalsystems of the brain grounded on value give rise to meaning and selfhood and how the self construesthe boundariesof the world. This task feeds back onto physics and forward onto socialviews of the worth of the individual. Let us explore a few of its implications.

1,64

CHAPTER

16

Memoryandthe IndividualSoul: AgainstSillyReductionism

Sciencecannotsohtethe ultimate mysteryof Nature, And it is because in the last analysisr,Deourseloes are part of the mystery we are trying to soloe, -Max Planck

If I had to lioe ooer again, I'd lirseooer a delicatessen. -Woody Allen

rom the last quarterof the seventeenthcentury to the last decadeof the eighteenth,an explosion of creativity called the Enlightenment changed the history of ideas. Its reigning views were many, but above all it was dedicated to reason, to science,and to human freedom and individuality. Its underlying sciencewas physics, the system of Newtory and its philosophy of society was, in large measure,that of Locke. Yet the Enlightenment ideas of causality and determinism, along with its mechanisticview of science,undermined hopes for a theory of human action basedon freedom. If we are determinedby natural forcesby mechani which a free individual makesmoral choices.Moreover, while the ideasof the Enlightenmentpaid much attention to the role of reasonand culture in such choices,there was no general notion of how deeply the minds of all humans(including those of "reasonable"human beings-that is, the "cultured") were influencedby unconsciousforces and by emotion. 165

HenMoNrns Whaiever forms it took at various times and places, the oveniding Enlightenment view was a secular one that forged many of the ideas underlying modem democracy.But despite its valuable heritage, the Enlightenment is over. The ftrst great blow to its ideas came with Hume's damaging attacks on both rationalism and the notion of human progress as linked to nafural science.Its major fault was its inability to create an adequate scientific description of a human individual to accompany its description of a machinelikeuniverse.Its social failure was its inability to go beyond the concept of a sociely composed of self-seeking,commercially successfulindividuals with a shallow view of "humanism."Certainly, Enlightenmentthinkers attempted to provide us with a larger,more inspiring view of ourselves.But its sciencewas a mechanisticphysics and it had no body of data or ideas with which to link the world, the mind, and society in the style of scientificreason to which it aspired.Whatever the Enlightenment'sfailures and inconsistencies,however, it left us with high hopes for the place of the individual in society. Can we expect to do better with a sound scientific view of mind? In this chapter I hope to show that the kind of reductionism that doomed the thinkers of the Enlightenment is confuted by evidence that has emerged both from modem neuroscienceand from modem physics. I have argued that a person is not explainablein molecular,field theoretical, or physiological terms alone. To reduce a theory of an individual's behavior to a theory of molecular interactions is simply silly, a point made clear when one considers how many different levels of physical, biological, and social interactions must be put into place before higherorder consciousnessemerges.The brain is made up of 1011cells with at least 1015 connections. Each cell has a fantastically intricate regulatory biochemistry constrained by particular sets of genes. These cells come together during morphogenesisand exchange signals in a place-dependent fashion to make a body and a brain with enormous numbers of control loops, all obeying the homeostatic mechanismsthat govem survival. Selection on neuronal repertoires leads to changesin myriad synapsesas cells die or differentiate.An animal's survival and motion in the world allow perceptual and conceptual categorization to occur continually in global mappings. Memory dynamically interacts with perceptual categorizationby reentry. Leaming involving the connection of categorization to value (in its most subtle form within a speechcommunity) links symbolic and semantic abilities to conceptual centers that already provide embodied structuresfor the building of meaning. A calculation of the significant molecular combinations of such a sequenceof events,even in identical twins, is almost impossible,and in any L66

Memory and the Indioidual Soul case,useless.The mappingsare many-many, and the processesare individual and irreversible.I wonder what Enlightenmenthumanistswould have made of all this. Diderot, who as we saw in chapter3 speculatedabout the nervous system of his friend in Le RAaede d'Alembert,might have been pleased.Diderot's view of human consciousness opened up the possibility that to be human was to go beyond mere physics. I have taken the position that there can be no complete science,and certainly no scienceof human beings, until consciousnessis explained in biological terms. Given our view of higher-order consciousness, this also meansan accountthat explainsthe basesof how we attain personhood or selfhood.By selfhood I mean not just the individuality that emergesfrom geneticsor immunology, but the personalindividuality that emergesfrom developmentaland social interactions. Selfhood is of critical philosophical importance.Some of the problems related to it may be sharpenedby the selectionistview I have taken on the matter of mind. Pleaseremember,however, that no scientifictheory of an individual self can be given (our qualiaassumption).Nonetheless,I believe that we can progresstoward a more completenotion of the free individual, a notion that is essential to any philosophical theory concemed with human values. The issues I want to deal with are concemed with the relationship between consciousnessand time, with the individual and the historical aspectsof memory, and with whether our view of the thinking conscious subject alters our notion of causality. I also want to discussbriefly the connection between emotions and our ideas of embodied meaning.All of theseissuesultimately bear upon the matter of free will and thereforeupon morality under mortal conditions. According to the extended TNGS, memory is the key element in consciousness, which is bound up with continuity and different time scales. There is a definite temporal element in perceptual categorization,and a more extended one in setting up a conceptually based memory. The physical movements of an animal drive its perceptualcategorization,and the creation of its long-terrn memory dependson temporal transactionsin its hippocampus.As we have seen,the Jamesianproperties of consciousness may be derived from the workings of such elements.But in human beings,primary consciousness and higher-orderconsciousness coexist,and they each have different relations to time. The sense of time past in higher-order consciousness is a conceptual matter, having to do with previous orderings of categoriesin relation to an immediatepresent driven by primary consciousness.Higher-order consciousnessis based not on ongoing experience,as is primary consciousness, but on the abilily to model 167

HeRMoNrEs the past and the future. At whatever scale,the senseof time is first and foremost a consciousevent. and "experienced"time are thereforeclosely The ideasof consciousness intertwined.It is reveali^gto comparethe definition of William James,who is somethingthe meaningof which "we know as statedthat consciousness long as no one asksus to define it," with the reflectionsof St. Augustine, "What then is time?If no one asksIn€,I know who wrote in his Confessions, what it is. If I wish to explain to him who asksrn€,I do not know." The notion of continuity in personal,historical,and institutionaltime was a central one in Augustine's thought. An intriguing suggestionabout the connecTime involves succession. tion betweentime and the idea of numbershas comefrom L. E. J. Brouwer, He suggeststhat all mathemata proponentof intuitionismin mathematics. (and of ical elements particularlythe sequence naturalnumbers)comefrom what he calls "two-icity." Two-icity is the contrast between ongoing as a large element)and consciousexperience(with primary consciousness (requiring higher-orderconsciousthe direct awarenessof past experience ness).What is intriguing about this is that it suggeststhat one's concept of a number may arise not simply from perceiving sets of things in the outside world. It may also come from inside-from the intuition of twonessor two-icity plus continuity.By recursion,one may cometo the notion of natural numbers. Whatever the origins of such abstractions,the personal senseof the sacred,the senseof mystery, and the senseof ordering and continuity all have connectionsto temporal continuity as we experienceit. We experience it as individuals,each in a somewhatdifferent way. Indeed,the flux of categorization,whether in primary or higher-order consciousness, is an individual and irreversibleone. It is a history. Memory grows in one directioru with verbal means,the senseof duration is yet another form of categorization.This view of time is distinguishablefrom the relativisticnotion of clock time used by physicists,which is, in the microscopicsense,reversible.Aside from the variation and irreversibility of macroscopic physicalevents recognizedbV physicists,? deep reasonfor the irreversibility of individually experiencedtime lies in the nature of selectivesystems.In suchsystems,the emergenceof pattern is expostfacto. Given the diversity of the repertoiresof the brain, it is extremelyunlikely that any two selectiveevents,even apparentlyidenticalones,would have Eachindividual is not only subject,like all material identicalconsequences. systems,to the secondlaw of thermodynamics,but also to a multilayered set of irreversibleselectionalevents in his or her perceptionand memory. Indeed,selectivesystemsare by their nature irreversible. '1,69

Memory and the lndioidual Soul This "double exposure"of a person-to real-world alterationsaffecting nonintentional objects as well as to individual historical alterationsin his The or her memory asan intentional subject-has important consequences. flux of categorizationsin a selectivesystem leading to memory and consciousnessalters the ordinary relations of causationas describedby physicists. A person, like a thing, exists on a world line in four-dimensional spacetime.But becauseindividual human beings have intentionality, memthey can samplepattems at one point on that line ory, and consciousness, u.i or, the basisof their personalhistories subject them to plans at other points on that world line. They can then enact these plans, altering the cu,rrulrelations of objects in a definite way according to the strucfuresof their memories.It is as if one piece of spacetimecould slip and map onto another piece.The difference,of course,is that the entire transactiondoes not invJve any unusualpiece of physics,but simply the ability to categorize, memorize, andform plans according to a conceptualmodel' Such an historical alteration of causalchainscould not occur in so rich a way in any combination of inanimatenonintentional objects,for they lad< the aPPropriate kind of memory. This is an important point in discriminatingbiology hom physics, an issue I discussfurther in chapter 20' In certain memorial systems,unique historical events at one scalehave causalsignificanceat a very different scale.If the sequenceof an ancient ancestors genetic code was altered as a result of that ancestor'shavels through u ,*u-p (driven, say, by climatic fluctuations),the altered order of nucleotides, i] it contributed to fitness, could influence present-day selectionalevents and animal function. Yet the physicallaws goveming the (the actualchemicalinteraction of the genetic elementsmaking up the code level chemical the at nucleotides)are deterministic.No deterministiclaws could alone,however, explain the sustainedcode changethat was initiated and then stabilized over long periods as a result of complex selectional events on whole animals in unique environments. Memorial events in brainsundergoing selectionaleventsare of the same ilk. Becausethe environment being categorizedis full of novelty, because selection is er postfacto, and becauseselection occurs on richly varied historical repertoires in which different structurescan produce the same result, many degreesof freedom exist. We may safely concludethat, in a multilevel conscioussystem, there are even greater degrees of freedom. These observationsargue that, for systemsthat categorizein the manner that brains do, there is macroscopicindeterminacy.Moreover, given our previous argumentsabout the effectsof memory on causality,consciousnesspermits "time slippage" with planning, and this changeshow events come into being. r69

HrC,nI\,IOhftrS

Even given the successof reductionismin physics,chemistry,and molecular biology, it nonethelessbecomessilly reductionism when it is applied exclusivelyto the matter of the mind. The workings of the mind go bevond Newtonian causation.The workings of higher-order memoriesgo beyond the description of temporal successionin physics. Finally, indiJidual selfhood in society is to some extent an historical accident. Theseconclusionsbear on the classicalriddle of free will and the notion of "soft determinism," or compatibilism,as it was canedby Iames Mill. If what I have said is correct, a human being has a degree oi fruu will. That freedom is not radical,however, and it is curtailed by a number of intemal and extemal events and constraints.This view does not deny the influence of the unconscious on behavior, nor does it underestimatehow small biochemical changes or early events can critically shape an individual,s development.But it does claim that the strong psychologicaldeterminism proposed by Freud does not hold. At the very Last, our freedom is in our grammar. Thesereflections,and the relationshipof our model of consciousness to evolved values bear also on our notion of meaning.Meaning takes shape in terms of concepts that depend on categorizationsbar"J o. value. It grows with the history of rememberedbody sensationsand mental images. The mixture of events is individual and, in large measure,unpredictaLle. when, in society, linguistic and semanticcapabilitiesarise and sentences involving metaphor are linked to thought, lhe capability to create new models of the world grows at an explosive rate. But one must remember that, becauseof its linkage to value and to the concept of self, this system of meaning is almost never free of afrect;it is chargej with emotionr. thi, is not the place to discussemotions, the most complex of mental objects, nor can I dedicatemuch spaceto thinking itself. I considerthem in the next chapter. But it is useful to mention them here in connection with our discussionof free will and meaning. As philosophers and psychologists have often remarked, the range of human freedtm is restricied uy tt e inability of an individual to separatethe consequencesof thoughl and emotion. Human individuals, created through a most improbable sequenceof events and severelyconstrainedby their history and morphology, can still indulge in extraordinary imaginative freedom. They obrrio.,rly of a "r" different order than nonintentional objects. They are able to refer to the world in a variety of ways. They may imagineplans,propose hopesfor the future, and causallyaffect world eventsby choice.Thly are linked in many ways, accidentaland otherwise,to their parents,their society,and the pasi. They possess"selfhood," shored up by emotions and higher-order con170

Memory nnd the lndioidual Soul sciousness.And they are tragic, insofar as they can imagine their own extinction. Often it is saidthat modem humanshave sufferedineversible lossesfrom severalepisodesof decentration,beginning with the destruction of earlier cosmologiesplacing human beings at the center of the universe.The first episode,according to Freud, however, took place when geocentrismwas displacedby heliocentrism.The secondwas when Darwin pointed out the descentof humanbeings.And the third occurredwhen the unconsciouswas shown to have powerful effectson behavior.Well before Darwin and Freud, however, the vision of a Newtonian universeled to a severefatalism,a view crippling to the societalhopesof Enlightenmentthought. Yet we cannow see arecorrect,this fatalistic that if new ideasof brain function and consciousness justified. is not pregnant with a fixed The present view is not necessarily programmed future, and the program is not in our heads.The theories of modem physicsand the ftndingsof neurosciencerule out not only a machine model of the world but also sucha model of the brain. We may well hope that if sufficiently general ideas synthesizing the discoveriesthat emergefrom neuroscienceare put forth, they may contribute to a secondEnlightenment.If such a secondcoming occurs,its major scientiftcunderpinning will be neuroscience,not physics. The problem then will be not the existenceof souls,for it is clear that eachindividual person is like no other and is not a machine.The problem will be to acceptthat individual minds are mortal. Given the secularviews of our time, inherited from the first Enlightenment,how can we maintain morality under mortal conditions?Under present machine models of the mind this is a problem of major proportions, for under such models it is easy to reject a human being or to exploit a Person as simply another machine. Mechanism now lives next to fanaticism: Societiesare in the hands either of the commercially powerful but spiritually empty or, to a lesserextent, in the handsof fanaticalzealotsunder the sway of unscientific myths and emotion. Perhapswhen we understandand accept a scientiftc view of how our mind emergesin the world, a richer view of our nature and more lenient myths will serve us. How would humankind be affectedby beliefs in a brain-basedview of how we perceive and are made aware? What would be the result of acceptingthe ideasthat eachindividual's "spirit" is truly embodied;that it it is mortal and unpredictablein its creativity; that we is precious because must take a skepticalview of how much we can know; that understanding the psychic development of the young is crucial; that imagination and toleranceare linked; that we are at least all brothers and sistersat the level of evolutionary values;that while moral problems are universal,individual 777

Hrq,nuONtrS instancesare necessarilysolved, if at all, only by taking local history into account?Can a persuasivemorality be establishedunder mortal conditions? This is one of the largest cJrallengesof our time. What will remain unclearuntil neurosciencegrows more mature is how any of these issuescan be linked to our history as individuals in a stillevolving species.In any case,silly reductionismand simple mechanismare out. A theory of action basedon the notion of human freedom-just what was missing in the days of the Enlightenment-appears to be receiving more and more support from the scientificfacts.We may now examinethe connection of these facts to thought itself.

CHAPTER

17

HigherProducts: Thoughts, Ernotions Judgments,

Thereis in us somethingwiser than our head, -Arthur Schopenhauer

ow can a book on the matter of the mind pay so little attention to thinking, willing, and judging, or to feeling, emotion, and dreaming?Partly, this has to do with my original intentions,which were and meaning in ary basesfor consciousness to describethe necess and more further faith that in the this have attempted I fashion. scientific a sufficientpsychologicalexplorationscan be launchedonce this description is substantiated.To pursueany one of thesehigher products of the mind's working would requirea separatebook. Nevertheless,I want to comment hereon how our thesesabout the mind may be connectedto psychological activities. is consideredby someto be the sameas thinki.g. I think Consciousness this is too crude an identification, for thought has additional acquired components:a complex of images,intentions,guesses,and logical reasoning. It is a mixture of severallevels of mental activity. At its highest and most abstractreaches,it is a skill, one that dependson symbolic abilities. With the exceptionof the spatialabilitiesexhibited in artistic thinki.,g and the tonal and rhythmic activities of musicalthinking, higher thought depends strongly on both languageand logic, on an inner dialoguebetween the thinker and another"interlocutor" of whose existencethe thinker may not be aware.This is the "two in one" to which HannahArendt refersin 173

HenuoNrrs her book The Life of the Mind. She points out the distinction in German between Vernunft,pure thought or reason, md Verstand,understanding with a direct connection to the cognitive processesof perception,feeling, and the like. I am not surethis distinction is usefulin scientificterms but it does serve to emphasizehow far thought can go. The thinker in the mode of pure thought is so immersedin a specificattentive state related to the project of thought that he or she is truly "abstracted"-unaware of time, space, self, and perceptualexperience.One may say that in the pursuit of these levels of meaningand abstraction,"thought is nowhere." But this is simply a metaphor to expressthe individual's degree of removal from awareness of other parallel activities of the mind. Whatever the skill employed in thought-that of logic, mathematics, language,spatialor musicalsymbols-we must not forget that it is driven by the Jamesianprocesses,undergoesflights and perchings,is susceptible to great variations in attentioru and in general,is fueled by metaphorical and metonymic processes.It is only when the results of many parallel, fluctuating, temporal processesof perception,concept formation memory, and attentional states are "stored" in a symbolic object-a sequenceof logical propositions, a book, a work of art, a musicalwork-that we have the impressionthat thought is pure. Becausethoughts are driven by other thoughts, by images,and by an imaginedgoal, we have the impressionthat there is a domain of Vernunft-a place where the thinker (in an absorbed attentional state)is nowhere and in no definabletime. The path from this impressionto Platonismand essentialism,both biologically untenable,is a short one. Thought cannot be pursuedexcept againsta consciousbackdrop.But a biological theory of consciousness provides only a necessarycondition for thinking, not a sufficientone. Thinking is a skill woven from experienceof the world, from the parallel levels and channelsof perceptualand conceptual life. In the end, it is a skill that is ultimately constrainedby social and culfural values.The acquisitionof this skill requiresmore than experience with things; it requires social, affective, and linguistic interactions. Thoughts, concepts, and beliefs are only individuated by reference to events in the outside world, and by referenceto social interactions with others, particularly those involving linguistic experience. What this means is that no amount of neuroscientificdata alone can explain thinking. There is nothing mysterious or mystical about this statement. A neuroscientiftcexplanation is necessarybut is not sufficientas an ultimate explanation. This is comparable to the statement that, while a completeembryologicalaccountis necessaryto explain how I look and act T74

H i S h e r P r o d u c t s : T h o u g h t s ,J u d s m e n t s ,E m o fi o n s like a man, it will never explain why I am a man. Only an additional evolutionary account involving historical events and natural selectioncan provide a sufficient exPlanation. At a certainpracticaipoint, therefore,attempts to reducepsychology to neurosciencemust fail. ii r"n that the pursuit of thought as a skill depends as on on social and cultural interaction, convention, and logic' as well insuffimetaphor, purely biological methods as they presently exist are and recursive is thought at its highest levels .iuni. fn puit, thi, is bec-ause symbolic.Becausewe are eachidiosyncraticsourcesof semanticinterpretathought tion, and becauseintersubjectivecommunication is essentialfor sh'rdy and (*t with a real or imaginary interlocutor), we must use "tL", however, ih.r" ."pu"ities in their own right. This necessity does not, properly be cannot contradict our position that cognitive psychology of consciousunderstood without a sound,biologically basedexplanation embodied' is meaning ness and of the Processesby which the central ln Acts of Meaning, |erome Bruner makes a strong casefor He emphasizes role of the construction of meaning in human psychology' under the culture in a interactions how the self arisesfrom interpe.sor,al interpretive influence of narratives. He urges that we employ rigorous our own scientific methods in social psychologyJn this instance,we are present work The devices. instruments,not to U" t"ptfud by measuring underlymeaning to provide a biological basisfor the constructionof "irn, how consciousness' irrg such efforts' With this foundation, we can see leads to the language' balsedon evolved value systems and driven by extension and modification of those systemsin a culture' and also what If we wish to investigate what dioes thinking, however, individual's that accompaniesit in an individual, we must still examine level of state and explore that individual's memory at the fi.f"SlA By its nature' this thought as it prompts tr motivates other thought' also to studies of uppr]".n i, li*it"d one. we must therefore look " as mathematics, iit'rinri. meaning, of invariant mental constructions such in logic, and found as substitution of invariance with respect to lexical derived' We g"""r"if' to a set of -les that are socially and experientially their leave room for philosophers without ceding to them i.'"y "rr"n of thought as miiaken but time-honored irivilege of applying methods the solemeansof understandinghow the mind arises' eachother but are not fully Justas different sciencesare compatiblewith for the reducible to one another-one being necessarybut not sufficient for the basis provides_a next-so a description of the matteiof the mind one analysisof relationaland symbolic matters.In making that description, nature of cannot help being struck Ly the multiple, parallel, and shifting 175

HanMoNrss consciousstates.Cognitive science'stask is to find out how to interpret statesconcemedwith symbolic modes of reasoningand statesof ludging or willing in which the subjectis more directly aware of his or her relation to time. Even if an analysisof such matters were to succeed,however, it would not give an adequateaccount of the potentially limitless use of recursivemodes of reasoning-induction, analogy, and formal logic. And, in any event, such an analysiswould not serve to exhaustexplanation in historical matters. Biological regularitiesunderlie all theseactivities.Theseregularitiescan and should be studied.But until, at somedistant time, we have conshucted consciousartifactscapableof speech,biological methods are too dumsy to be used to make neural correlationswith the meaning of the thoughts of a "pure thinker" during a processof reasoning.We can,however, study the fundamentalneural processesthat underlie these acts, and we can do so without becoming property dualists.But practically speaking,it would be foolish to use only biological methods in the name of scientific purity. Much more could be said, but it would not be illuminating for our concems,which are to consider the biological basesof mind. It may be usefuL however, to comment on a few related issues,particularly those concemedwith feelings and emotions. Feelingsare a part of the conscious state and are the processesthat we associatewith the notions of qualia as they relate to the self.They are not emotions,however, for emotions have strong cognitive components that mix feelings with willing and with judgments in an extraordinarily complicatedway. Emotionsmay be considered the most complex of mental statesor processesinsofar as they mix with all other processes(usually in a very specificway, depending on the emotion). They are not made simpler by the fact that they also have historical and social bases. What is perhapsmost extraordinary about conscioushuman beings is their art-their ability to convey feelings and emotions symbolically and formally in extemal objects such as poems,paintings, or symphonies.The summaries of conscious states constrained by history, culture, specific training, and skill that are realizedin works of art are not susceptibleto the methods of scientiftcanalysis.Again there is no mystification in this denial, for understanding and responding to these objects requires referenceto oursehtes in a social and symbolic mode. No extemal, objective analysis, even if possible, supplants the individual responsesand intersubjective exchangethat takesplace within a given tradition and culture. A beautiful analysis of these psychological processeshas been given by Suzanne Langer in her chef d'oeuvre, Mind: An Bsay on Human Feeling. In addition to Vemunft and Verstand,another set of Germanwords used 176

H i g h e r P r o d u c t s : T h o u g h t s , J u d g m e n t s ,E m o t i o n s to characterizehuman knowledge makes a distinction first clearly set out refers to knowledge concerned by Wilhelm Dilthey. Naturutissmschaften with what usedto be calledthe natural sciences-physics, biology, and the refers to fields of knowledge concemed with the like. Geistenoissenschaften social sciences,with culture, with abstract reasoning,and with studies of historical events basedon symbols and feeling. In making this distinction we must not succumbto the idealismof Georg Hegel, who was among the most distinguishedproponents of Geist.Nor must we think that psychology falls outside evolutionary biology, that there is a separateGeist and a separateNatur, for this leads to endlessunnecessarycomplications. ]ames,a reflective investigator of the subjectsof this chapter,had this to say about some philosophical positions on such matters: is,it seemsto me, speculation Thewholelessonof Kantianandpost-Kantian the lessonof simplicity.With Kant, complicationboth of thought and of by the musty academicism statementwasan inbom infirmity,enhanced With Hegelit wasa ragingfever.Terribly,therehis Konigsbergexistence. fore, do the sourgrapeswhich thesefathersof philosophyhaveeatenset our teeth on edge. Given what I have said here, I expect philosophical psychology to continue to go its own way, with this qualiftcation:Despite the methodoand Naturwissenschaften, logical differences between Geisteswissenschaften psychology can no longer declareits autonomy from biology, and it must always yield to biology's findings. I used to wonder why there were so many subjects in a university catalog. Why is knowledge so heterogeneous?The view presentedhere offers a possible reason.Given the parallel, constructive brain processes that underlie consciousness,given the recursive symbolic properties of language, and finally, given the irreversible historical bases fo1 specific symbolic and artistic realizations in society and culture, there c{n be no fully reducible description of human knowledge. But different spheresof knowledge and different subject domains can be compatible with each other, and their basesin biological and cultural evolution can be understood. Human beings, at least in their pursuit of these different domains, seem to be doing just about what they should be doing. A more poignant situation ariseswhen human beings are afflicted with neural disorders.As I hope to show in the next chapter, these afflictions also reveal the enormous range of responsesand the layered complexity of which the nervous system is capable.

177

CHAPTER

IB

Diseases of the Mind: TheReinte gratedSelf

A satisfactorygeneralcomprehension of neuropsychotic disturbancesis impossibleif one cannot make connectionsto clear assumptions about normal mentalprocesses, -Sigmund Freudto Wilhelm Fliess

ental diseasehas always seemedmysterious.It affectsthe individual "soul," and as an aberrationfrom a person'spreviously witnessedhistory and behavior,it seemsstrangeto those who know the Personif not to him- or herself.It is often difficult to trace its causes,and while one may be convinced that it is the result of alterations in brain function, it does not have the symptomatological "directness"of many of the neurological disorders that also result from altered brain function. U/hat is the difference, and where is the dividing line? The differenceis a subtle one, but it almost always has to do with changesin intentionality, consciousness,value, or symbolic function. A theory of mind such as ours makes it clear that all mental diseasesare based on physical changes. I have no intention of dealing at any length with this fascinating and difficult subject.But to neglect all of its many facetsis to lose an opportunity provided by nature to checksome of our models of the mind. Accordingly, I will deal here with some time-honored medical subjectsfrom the viewpoint taken in this book. There is no better way to reveal the multilevel controls affecting the brain and mind. First I want to discusswhat 778

D i s e a s e so f t h e M i n d : T h e R e i n t e g r a t e dS e l f mental diseasesare diseasesof-whether they are physical or not. Then I want to consider why they seem different from neurological diseases. Following that, I want to take up some diseasesof consciousness-both neurological and psychiatric-because they shed light on our models. Finally, I want to look at mental diseasesas adaptations,as reintegrations of the self under crippling conditions. I discussedsome of this material in TheRemembered Present but not in quite the sameway. The readeris invited to compare. Freud was deeply concemedwith a problem that is central to the issues with which we have been concemed.While investigating the neuroses,he strode into the thickets of intentionahty. Freudthought that neuroseswere those alterations of behavior or emotion that did not involve losing the ability to test reality but that did impair function or satisfaction.It did not seemto him that thesedisordersresultedfrom brain disorders.Rather,they appearedto be functional disorders,stemming from psychological factors, from symbolism both consciousand unconscious.They were, in his view, He constructedhis psychological disorders of psychological deoelopment. theory basedon the results of a therapeuticmethod, psychoanalysis.This method involves symbolic interactions between a patient and a trained analyst who has a definite theory of how the human personality is formed and how a person'sego is developed.The patient is encouragedto explore, with the analyst, mechanismsof defense and repression by using the techniquesof free association,dream analysis,and the like. Although Freud initially adopted a severeeliminative materialism(just the sort of reductionismI deplored in the last chapter),he later resorted to a kind of property dualism. While remaining a severe determinist and materialist,he held that the neuroseswere neverthelessto be considered only in psychological terms. Freud's position on psychosis, in which a patient's reality testing is truly impaired and for which organic causescan be found, was more equivocal.Certain psychoseswere and are considered to be "functionaL" schizophrenia,someforms of manic-depressivepsychosis, and paranoia, for example, have causal or etiological histories quite different from those of organic psychosesor the degenerativebrain diseasesthat can also end in psychosis.Theseobservationspose difficultiesfor any brain theory. According to the TNGS, thesedifficultiesarisebecauseof the intricacies faced in sorting out the levels at which brain function is controlled. A further intricacy stemsfrom the population nature of synaptic responsesfrom their diversity and individuality. The real problem, however, is not intricacy but the misassignmentof levels of causation. All psychiatric disorders,even those traceableto difficulties in individual and social com179

HenvoNrss munication, have physical causes.Given the multilevel reentrant systems of the brain that control consciousand unconsciousstates,it is no surprise that different causesof diseaseyield overlapping or similar derangementof responsepattems. Sooner or later, all aberrationsare reflectedat synapses. But at the sametime, complex signals and environmental interactions are linked to memory and behavior in pattems acrossall the levels discussed. It is often more useful to considermental diseasesas disordersof categorization,memory, reentry, and integration rather than asdisordersof "reality testing." One way to visualize this is to adapt the diagrams for consciousness given in chapters 1.1and 12 to our present purposes(figure 18-1.).If we consider,for example,that factors destroying or affecting neurons in parts of the basal ganglia lead to Parkinson's disease,or that other factors affecting the motor cortex yield a motor paralysis,we have no difficulty diagnosingthe affliction asneurological-that is, as "not mental." But if the diseaseinvolves interactionsamong the highly parallelreentrantcircuits of the brain, or alters the connections between value systems and those driving behavior (seefigure I2-4), we are likely to diagnosethe disorder as an alteration of mental function. In both cases,physical causesare sufficient to account for the disturbances. The problem of mental diseasemay be usefully looked at in terms of alterationsin reentrant pathways and in categorization.Derangementand diseasesof consciousnessrepresent rearrangementsand adaptations to alterationsin reentrantmaps,homeostaticregions,and the cortical appendages responsiblefor perceptualawareness,symbolic conceptualfunctioning, and emotional responses. In any event, the individual history of a person with such afflictions virtually assuresthat no two patientswill be alike.Higher-order consciousnessinvolves concepfual,semantic,and socialintegrations, all of which are involved in the construction of a socialself.Many of theseintegrationsare mediated or modulated by particular synaptic populations. It is therefore not sulprising that drugs that changethe functions of synapseshave been found to be enormouslyusefulin the treatmentof mental disorders.But the individuality of the consciouspatient results from an enormously complicated pattem of synaptic efficaciesunique to him or her. The task of communicatingwith the patient by verbal and emotional meanswill therefore not be abrogated by the use of drugs alone. A combination of drugs and psychotherapy is still likely to be required in most cases. A brain theory that views categorization,memory, and concept formation in theseterms can even be useful in purely psychotherapeuticformulations. An appraisalof the TNGS by a psychiatrist formulating a theory of 180

Diseasesof the Mind,

The Reintegrated Self

\

--.....""Cortgx MotorCortex Stroke, (Paralysis)

"' BasalGanglia

(Parkinson's Disease)

""''|Cgrgbgllum (Ataxia,Discoordi nation)

Hippocampus

(MemoryFailure, Amnesia)

Brain Stem

NEUROLOGICAL DISEASE

GeneralizedFailure of Coordination of ReentrantLoops (e.g., Schizophrenia)

PSYCHIATRIC DISEASE FIGURE18-1 'Neurological Dbeasesof the neroow systemand diseases of the mind. disease"rdus to dkruptiotrs of sight, mooement,and soforth, and is the resultof alterationsin the regions of the hain inoolved in thne functions. "Psychiatric disease"refers to alteratiorc in categorization,mental activity, qualia, and n forth, in which responses are symbolically deoiantor in which "reality testing" is compromised.Thesediseases resultfrom functional alterationsat many leoels,from synapsesthrough reentrant loops.Both categoies are physical in oigin, and they ooerlap.Psychiatic diseasesafect categoization, memory, and symbolic processes more ertensioelyoia reentrant loops. 181

HlnuoNrrs psychoanalytic treatmentwaspublishedseveralyearsago.ln OtherTimw, Amold Modell employsthe ideaof memoryasrecategoriOtherRealities, zationto reevaluate betweenthe patientand thenatureof the transactions therapist.He revivesthe termNach*iiglichlceif, whichFreudappliedto the experience. ideathat a memoryis rehanscribed as a resultof subsequent Modell pointsout that the egois a structureengagedin theprocessing and reorganizingof time,whichhe linksto the ideaof memoryasrecategorization (seechapters10 and 16). In his critiqueof conceptsrelatedto the transference relationin the psychoanalytic setting,Modell proposesthat a selectionistview of brain functionoffersan altemativeinterpretationof the repetitioncompulsion describedby Freud.Modell proposesthat the recreationof a categorical memoryis a fundamental biologicalprinciple,one that undercertaincircumstances supersedes Freud'spleasureprinciple.He suggeststhat "the compulsionto repeatrepresents a compulsionto seeka perceptual identity betweenpresentand past objects."In doing so, the patient is awareto varyingdegreesof the relationof the selfto time,a subjecttouchedon in cJrapter16. Modell also points out that, in the psychoanalyticprocess, thinkingin metaphoris the currencyof the mind.Finally,he suggeststhat the view of memoryproposedin the TNGS,which replacesFreud'sview of fixed memorytraces,opensup a new way of looking at the treatment setting.In this view, the treatmentsetting is designed"to accentuate multiplelevelsof reality,whichin tum enlargesthe potentialfor both old perceptionsand retranscriptions of new perceptions." Another psychiatrist,EdwardHundert,has also usedthe TNGS as a toucistonefor relatingpsychiatry,philosophy,and neuroscience. He has proposedthat all threedisciplinesmustbe relatedthroughsucha theory to assurethat individualsareviewedin a sufficientlybroadway. Interested readersareencouraged to comparehis remarkswith the viewsI presented in the previouschapterand throughoutthis book. AlthoughI amhardlyqualifiedto speakwith authorityaboutmattersof psychoanalytic theory,it seemsto me to be importantthat attemptsare being madeto relatepsychoanalytictheory to a physicallybasedbrain theory concemedwith problemsof categorization. Attempts to relate pharmacological effectsto braintheoriesaresomewhatmoreprevalent.We needboth kindsof effortif we areto havea psychiatrythat is solidlybased in biology. With thesegeneralcomments, I will tum to a specificsetof neurological diseases andonepsychiatricdisease. I do so because both kindsof disease shedlight on the proposalsI havemadeaboutthe importanceof reentrant circuitsto consciousness. TheneurologicaldisordersI will discussarea set 782

D i s e a s e so f t h e M i n d : T h e R e i n t e g r a t e dS e l f of syndromes that involve what has been called implicit memory. The psychiatricdiseaseis schizophrenia,the most florid, varied, and mystifying of psychoses. First, the neuropsychologicalsyndromes (here the reader will have to bear with some clinical jargon; I will translate as I go along). Daniel Schacterand his colleagueshave studied a set of dissociationsbetween explicit, consciousaccessto knowledge and the implicit ability to perform a task. This kind of dissociation-near-normal implicit knowledge with severelyimpaired explicit knowledge-has been observedin patients with a variety of diseases.Before listing them and describing some cases,it might be best to give a concreteexample:blindsight. Patientswith blindsight distinguish visual stimuli in spaceeven though they are blind in the part of the visual field in which thesestimuli are presented.Thesepatients perform tasksof perceptualdiscriminationeven though they are perceptually unaware or unconsciousof being able to do so. Such dissociationshave been observed in patients with amnesia,dyslexia (the inability to read certain texts), aphasia(the inability to expressor produce intelligible speech),prosopagnosia (the inability to recognize faces),hemineglect(the inability to attend to the egocentric side of space opposite to a damaged cerebralhemisphere),and anosognosia(the unawarenessof or denial of gross neurological defectseven when presented with direct evidence for their existence). This is an extraordinary list. Amnesiacs show leaming responsesto priming cues during performance that indicate the presenceof specific knowledge even though the patients deny awarenessof such knowledge. Prosopagnosicsshow implicit knowledge of faces that one knows they have seenbefore the onset of their disease,but of which they now have no consciousawareness.Under certain circumstances,dyslexics read texts without the awarenessthat they can do so. And in patients with hemineglect, information presented on the neglected side affects the patient's performanceof a task without his or her consciousknowledge. The lesions that underlie these disorders are all different. There is no global change of consciousnessin any of these cases.Indeed,the patients behave quite normally outside the domains exhibiting the defect; in other words, in each casethe defect is domain specific.There does not seemto be any evidencethat languageimpairmentsare responsiblefor the patients' dissociativeresponses.Above all, there is no evidencethat patients with these syndromes suffer from neurosis or psychosis. Thesediseasesare diseasesof consciousness. They can be explained by assumingthat what has been affectedis the specialreentrant loop connecting a value-categorymemory to classificationcouplescarrying out percep183

HenMoNrrs tual categorization (figure 18-1; see also figure I2-4). Notice that this interference will not, in generaf affect other pathways connecting with value-categorymemory and mediating the performanceof a specifictask. The result is a domain-specificdeletion from the conscious"scene"but not from the repertoire of the individuals' capabilitiesto perform the task. If thesereentrantpathways were clipped one by one acrossall the modalities, one wonders whether any primary awarenesswould remain. This is unlikely to occur except as a result of massiveinjury, and on ethical grounds the experiment could hardly be performed even if some less traumatic means of bringing it about were available. I have describedtheseneurologicaldisturbancesto show that they, like mental diseasesproper, can also be mental-that is, they can affect intentionality. To illustrate thesedisturbanceseven more dramatically,I end this chapterby describinga caseof dissociationthat is downright mysterious, that of anosognosia.But first, let us fum to the most flagrant,polymorphic and mysterious of psychoses:schizophrenia.Schizophreniais characterized by mixfures of bizarre symptoms. These include third-person auditory hallucinations,delusionsof control by alien forces,ideasof influence,ideas of reference,thought echo,thought broadcast,thought bloch and thought withdrawal. In the acutestagesof this syndrome,the symptoms are accompanied by a dreamlikestate and a slight clouding of consciousnessas well as by pelplexity. Patientsexperiencea barrage of signalsthat make sense only in fragments and across islands of awareness.Some patients show blunting of affect. Some are easily distracted,misjudge perceptualsignals, have visual illusions, or are not able to discriminategestalt figures.Others show poor judgment, are slow to respond,or perseverate.In extremeforms of the disease,patients are sometimes catatonic; they are unresponsive while maintaining bizane posturesfor long periods of time. On recovery from acute episodes,some patients rememberwhat was said or done by others during the catatonicstate,while others have a poor memory of the acute experience. Schizophreniais a diverse and protean diseaseof consciousness, affecting perception, thought, and qualia. It is a moving and sorrowful experience to seea patient, with his or her unique constellationof symptoms, in the grip of this affliction. It has not been easy to accountfor the diversity of symptoms, the individual characteristics,and the bizarre featuresof this disease. I have suggested the possibility that schizophreniais a generalized diseaseof reentry (see figure 18-1, bottom). I have hypothesized that a disorder in the production of or response to several neurotransmitters could cause a general disabling of communications between reentrant maps.If there is a failure in appropriatemapping or an asynchronybetween T84

D i s e a s e so f t h e M i n d : T h e R e i n t e g r a t e dS e l f maps,imaging may predominateover perceptualinput, or different modalities may no longer be coordinated.This could lead to hallucinationsand failure to coordinate real-world signals. The same disorder in another patient may lead to disturbancesin the reentrant engagement between different conceptual systemsand the organs of succession.In yet others, there may be disordersin reentrantlinkagesbetween areasinvolved in the lexicon, conceptualcenters,and those that mediate imagery. Historical and individual factors involving repertoire variation explain the different responsesfound in different schizophrenic patients. Each patient has a unique pattem and respondsdifferently to sites of reentrant disorder. I am not saying that all neurons are normal in schizophrenia(l doubt that this is so) but only that the main psychological defect is the result of a defect in reentrant mapping. This may be causedby *y of the factors that alter individual maps or their connections,including neuronal disorderor loss. It is not difficult to see how a patient who still has higher-order consciousnesswould, if afflicted with this disorder, try to adapt as a self to what he or she perceives.This behavior would obviously appearabnormal by normative and physiological criteria, but it is likely that the patient's overall responseis still an attempt at adaptation,at reintegration. That it is not the best adaptation or that it may be destructive to the self and others is not in question.But the mind of a schizophrenic,viewed from the extendedTNGS, does make sense,particularly if one knows the history of the individual patient. Alas, the "sense"is made in terms of the predictions of the theory, not in the socialor affectiveterms that are speciftcto a given society. We underestimatethe faculties of psychotics as readily as we misread the apparenteccentricitiesof normal people. I do not know where I heard or read of the man in Parisduring the Nazi occupation,who, knowing that the Gestapowas closing in on him, decidedthat the last place they would searchfor him would be in a hospital for the insane.So he cultivated the art of behaving crazily,and after a 'hallucinatory" episodein the street he was committed. Life went on without terror for quite a while, and from time to time he displayed deliberately bizane behavior in front of the doctors and his fellow patients. One day two sinister men in long black leather coats appearedat the door of his room, accompaniedby the chief superintendent.Certain that the Gestapo had arrived, he leaped up, assumed abizane posture,rolled his eyes,and began to emit strangeyelping noises.Whereupon the man in the next bed, who spent most of his time in a trance-likestate,openedhis eyes and said to him firmly, "Taisez-vous, simulateur" ("Shut up, faker"). As for normal "simulation" (but not real faking), I recall my dear friend, 185

Hanlraorutrs the late Lars onsager, one of the most extraordinary physical chemistsof our time, indeedthe most phenomenallygifted scientistI have ever known. within ftve minutes of taking his seat at some arcane lecture, his head would loll to one side in a posfure of sleep.But if the lecturer erred in an equation, Lars was likely to get up, stroll to the board, erase the error, correct it, smile,go back to his chair,and fall asleepagain.I once askedhim, "Lars, when people ask you deep questions,why do you grin and giggle and nod your head and say incomprehensiblethings?" He became 9rave, almost stem, and said, "l'm lazy." "Lazy!" I exclaimed,"l don't get it." To which he responded,"l want to answer zy questions,not their questions." I have deliberatelydiscussedtwo extremeexamplesfrom the "neurological camp" and the "psycJriatriccamp"-implicit-explicit dissociationand We must schizophrenia.Both are examplesof diseasesof consciousness. disease. either type of with patients in analyzing of Cartesianism beware To drive that point home, I will describe a case reported by the Italian neurologist EduardoBisiach.This casepresentsas significanta challengeto brain theory as the one that facedFreudin the early days of psychoanalysis when he confronted his patient Anna O. Bisiach'spatient had anosognosia;that is, he denied the existenceof a neurological defect even when presentedwith direct evidencefor its existence.In this case,the syndrome accompaniedleft hemiplegia(paralysison the left side) and left hemianopia(inability to seethe left visual field). The causewas a suddenvascularepisodeaffecting his right temporal, parietal, and occipital brain areas.(Roughly speaking,these cortical areasmediate visual tasks as well as some motor capabilities.) The patient was intelligent, responsive,and in no obvious way emotionally upset. He showed no evidence of a speech disorder. But he was anosognosic to his deficienciesboth of sight and of movement. I{hen questioned about left-sided tasks that he did not actually perform, the patient claimedto have performedthem. (His left-sidedparalysisobviously made their performanceimpossible.)The patient's paralyzedleft hand was then placed in the hands of his examiner and positioned in the patient's right visual field so that he could seeit. He was then askedwhose hand it was. He claimedit was the examiner's.When askedabout the three-handed 'A hand is an discrepancy,the patient's responserevealedflawlesslogic: extremity of an arm. Sinceyou have three arms, it follows that you have th'reehands." If we disqualifu suggestionsthat this patient was neurotic,psychotic, or had a language disorder, we are forced to admit to an extraordinary basedon languagecan be alteredby conclusion:This is that consciousness the removal of brain sourcesof nonverbal signalsin the individual. What is striking is that this patient reintegratedhis entire semanticinteqpretation r86

D i s e a s e so f t h e M i n d : T h e R e i n t e g r a t e dS e l f of reality without emotional disfurbance.He underwent a radicalconceptual rearrangementand reintegration.One would have thought that if his body image and categoricalcapability were "stored" in memory, the contrast of that stored memory with his present perceptual and motor state would have led either to a consistent "realistic" report of his actual plight or to great conflict and emotional disturbance.But if our theory is correct, the patient has no such ftxed memory. He cannot attend to parts of personal space,and he has undergone a concepfualand semanticreintegration that not only reflects this incapacity but in some sense builds an adaptive picture around it. This caseclearly indicateshow changesin intentionality can accompany neurologicaldisease.It alsoservesasa deepchallengeto our notions of mind. How is the self reintegrated?In this connection,I should also mention the remarkableresponsesof a split-brainedindividual, the left brain of whom appears to belong to a more or less normal person with higher-order consciousness. In somecases,the right brain respondsto visual words with left-handed responsesspelled out in Scrabbleletters. Sometimesthe responsesareconcordantwith thoseof the left brain,sometimesnot. One must at least consider the possibility that the right brain is mediating primary consciousness. Unfortunately, a completeanalysisis blurred by suchfactors as past leaming and the previous connectivity of the hemispheres.In any case,thepersozl reporting with higher-orderconsciousness is seatedin the left hemisphereand interprets all eventsadaptively. These observationspose problems closely related to those studied by Freud.What govems the dispositionof consciousandunconsciousresponses so asto allow integration into personhood?What is the minimum apparatus required for the appearanceoi high"r-order consciousness? What govems the courseof reintegration when the individual is affectedby disease? The study of mental diseaseat all levels is obviously as important to an understandingof how the brain works asit is to an understandingof what it meansto be an individual in a society.Given its culturalsignificanceto us,the study of mental diseasehas obvious practical significance.However, the complexitiesof thesestudiesareenornous, and it is unlikely that the normal and abnormalworkings of the brain can be unraveledby psychiatry alone. Many disciplinesareneeded.At the furthestreachfrom psychiatry,attempts have been made to synthesizeobjectsand artifactsthat have psychological functionsand manifestintentionality. If theseefforts are successful, they will play an important part in helping us understandour placein nature,both in health and in disease.So I tum now from the study of intentional humansto the possibility of creatingintentional things.This is an exciting prospectand, whatever its limitations, one of great practicaland theoreticalimportance. r87

CHAPTER

19

to Constructa Is It Possible Artifact? Conscious

in this world, and that It is clearthat thereis but onesubstance man is its ultimate erpression,Comparedto monkeysand the cleoerestof animals he is just as Huygen'splanet clock is to a watch of King Julien,lf more tpheelsand springsare neededto shoutthe motionof theplanetsthan are requiredfor showingand repeatingthe hours;and if Vaucansonneededmore artistry in producinga flautist than a duck, his art rnould haoebeenet)en 'talker', in and sucha machine,especially harderput to producea must no longer be the handsof this neu) kind of Prometheus, thoughtof as impossible, -lulien Offray de la Mettrie

fter the discussionof human issuesin the last severalchapters,the title of this one may seemincongrllous.My purposeis to consider whether a knowledge of brain function will allow us to construct intentionalobjects.I alsowish to raisethe possibilitythat the only way we may be able to integrate our knowledge of the brain effectively, given all its levels,is by synthesizingartifacts.To do this, we needthe most advancedkind of computers.The construction of "conscious"artifactshas a meagerbut definite history, one to whidr the epigraph and figure IFI bear testimony. Indeed,is it not contradictory to suggesta need for computersin a book maintaining that the brain is rof a computer?To get at the answersto this question, I have to say a bit about computers.Then I want to consider 188

I s I t P o s s i b l et o C o n s t r u c t a C o n s c i o u sA r t i f a c t ?

FIGURElFI Jacquu de Vaucatron (1709-1782), a famousconshuctorof artifacts imitnting belnoior, shown with his duck. The constructionquacked,waddled,and had "intestinalfunction."

whether it is possible to construct severalkinds of artifacts: a perception machine,an artifact with primary consciousness, and one with higher-order consciousness.If the answer is yes to any of these questions, there are moral issues to be considered,as is the case in the application of any scientiftcfinding. Computers are logic engines;in principle, they can carryrout any effective procedure that is speciftedunequivocally by a set of instructions and producesa unique result for a given problem. I have said that the brain is not a computer and that the world is not so unequivocally speciftedthat it could act as a set of instructions.Yet computerscan be used to simulate parts of brains and even to help build perception machines based on selection rather than instruction. I can resolve the apparentcontradiction by pointing out what a simulation does. In a simulation,a program is written that specifiesthe required structuralproperties and operating principlesof the entity to be simulated. The program is so constructedthat when it is run, parts of the entity that is simulated in its entirety will carryr out their proper functions. lf, for example,I want to simulateaT47 airplaneflying into furbulent conditions, I have to put into the program the design featuresof the airplaneas well as the principles that allow it to fly-the properties of the airfoils, the power characteristicsof lift for a certain weight, and so on. If the program is well designed,I can "fly" the aircraft under both smooth and furbulent conditions.My goal might be to seewhether the plane losescontrol under certain conditions, or whether a wing vibrates too much and shearsoff. If 189

HaRMoNIEs successful,the whole exerciseis cheaperand more informative than mnning a real-world model of the plane in a wind tunnel. Can a selectionalsystem be simulated?The answer must be split into two parts.If I take a particularanimalthat is the result of evolutionary and developmental selectioh, so that I already know its structure and the principles goveming its selectiveprocesses,I can simulate the animal's structurein a computer.But a systemundergoing selectionhas two parts: the animal or organ, and the environment or world (seechapter 8). No instructions come from events of the world to the system on which selectionoccurs.Moreover, eventsoccurringin an environmentor a world are unpredictable.How then do we simulateevents and their effectson selection?One way is as follows: 1 . Simulate the organ or the animal as described above, making provision for the fact that, as a selective system, it contains a generator of diversity-mutations, alterations in neural wiring, or synaptic changes that are unpredictable. 2 . Independently simulate a world or environment constrained by known physical principles, but allow for the occurrenceof unpredictable events. 3 . Let the simulated organ or animal interact with the simulated world or the real world without prior information transfer,so that selection can take place. 4. See what happens.

Now we have a situation in which unpredictablevariation is occurring in each of two separatesystemsor domains.Moreover, if we allow these domains to interact in the simulation, there is no way of predicting the outcome beforehand (other than by performing the same simulation in another computer). Therefore, no effective procedure simpler than the simulation itself can predict the outcome. Put another way, the program used for each domain doesspecifuconstraintsand structures,but it cannot specify the exact way in which thesewill act under conditions of variation and selection. Variational conditions are placed in the simulation by a techniquecalleda pseudo-randomnumber generator.This is a formula that producesnumbers in a way that simulatesrandomness!It doesn't succeed absolutely;if we wanted to capturerandomnessabsolutely,we could hook up a radioactive sourceemitting alpha particles,for example,to a counter that would thm be hooked up to the computer. But it appearslikely that if we choosetwo separaterandom number generators,one for variation in the animal and one for variation in the environment, and allow selection 190

l s I t P o s s i b l et o C o n s t r u c t a C o n s c i o u sA r t i f a c t ? to occur, we are likely to avoid introducing a predictable bias into the system. After all, the two systems have no way of "knowing" what variation will match what, and we can always keep changing eachrandom number generator. Under thesecircumstances, we cannot specify an effectiveprocedure/or the consequences of selectiozthat is independent of our choice of pseudorandom number generators.Insofar as this is so, it is not meaningful to describethe systemand its future resultsasa wholeasa computer (or Turing machine).Any particularpastresult of selectiononcen)elenowit canof course be so specified.By theproceduredescribedabove we have addedfeaturesto our simulatingcomputerthat convert its performanceinto somethingthat is not strictly that of a computer. This is the result of choosing methods for generatingrandom numbersthat do not couple in predictableways to the sequencesof events in the systemsbeing simulated,so that the theoretical probability of a future event lies in the sequenceof pseudo-randomnumber generatorsand not in the simulation itself. Now we are in a position to tackle our ftrst question: Is it possible to construct a perception machine?Yes, although the ones that have been constructedso far are primitive. I have already presentedsomedata on the performanceof one the first of thesein chapter 9. This artifact, Darwin III, has a four-jointed arm with touch receptorson the part of its arm distal to the last joint, kinesthetic neurons in its joints, and a movable eye. It contains simulatedneurons in numerousrepertoiresthat show diversity in both local connectionsand synaptic strengths.Although it sits still, it can move its eye and arm in any pattem possiblewithin the bounds imposed by its mechanicalarrangement.Objects in a world of randomly chosen shapesmove at random past its field of vision and occasionallywithin reach of its arm and touch. The synaptic strengths of its neurons are initially set by a random number generator.After encounteringobjects (and responding to them), it displays behavior that appearsvery much like perceptual categorization(seefigure F6). This is the caseas long as its neural circuits have beenconstructedto respondto value (for example,light is better than darkness,or touch is better than no touch). Thus, it carriesout categorization on value. A word about this is in order-the basisfor value is programmed into the design of the machine.But this value is not the sameas any category, which is not so programmed. The programming of value is allowed becausevalue is consideredto have resulted from the eoolutionaryselection of preferencesof a particular type becausethey conferred selectiveadvantageson the individuals of a species.If we were to simulatea cat, we might put in value systemsor circuit constraintsthat made movements leading T9T

HenuoNIrs to fur licking (asdetectedby simple parts of the cat's own nervous system) more likely to be rewarded. The effects of such constraints on behavior would ultimately have to be observed,not programmed.In our automat4 we have nof programmed in the kinds of categorization that result from actual somatic selection,becausethese changesare epigenetic. There is nothing, by the way, especiallymysterious about value. For instance,to give Darwin III valuespositive for lighted objectsin its central vision, specializedneuronsare constructedwith inputs more denselywired in the central part of its "retina" and less densely in the periphery. The image of a lighted object falling on the central retina produces strong responsesin these neurons. These responsesare then transmitted to the vicinity of the synapsesthat connect visual neurons with the motor neurons that move the eye.Activity in thesesystemsleavesa "chemicaltrace." While these traces last and following any movement that brings the stimulus toward the central region of the eye, a value signal leads probablistically to the strengtheningof the synapsesinvolved. This increasesthe likelihood that similar motions will take place the next time a stimulus appearsin a similar position. The behavior of Darwin III is quite limited. It does not categorizeacross a broad set of stimulus characteristics,and it does not show true leaming, although experimentsproviding it with a "taste" system have suggested that it modifies its selectionpattems for objects after the alteration of its values.In any case,to test suchbehaviors,one wants to have a much richer environment, one composedof real-world objects. A perception machine with its nervous system simulatedin a supercomputerand with its "real" eyes and motor apparafusin a "creafure" "living" in a different room is currently under construction (figure t9-2). This artifact, NOMAD (Neurally Organized Multiply Adaptive Device), is in touch by television and radio with its brain, a brain more complex than that of Darwin III but still designed as a selectionalsystem.The tasks set for NOMAD couple categorization to leaming-picking out variously shapedobjects it considers to have value, for example.Unlike what happensin ordinary robots, these choicesare not programmed;they are the result of training. I have called the study of such devicesnoelicsfrom the Greek noein,to perceive. Unlike cybemetic devices that are able to adapt within fixed design constraints,and unlike robotic devicesthat usecybemetic principles under programmed control, noetic devices act on their environment by selectionalmeansand by categorizationon value. This field is in its early stages,but already it promisesto teach us much about how to study the layers and loops of neurally organized systems.In time, noetics will also have considerablepractical importance. 192

I s I t P o s s i b l et o C o n s t r u c t a C o n s c i o u sA r t i f a c t ?

Two-WayTelevision and RadioLink Antenna

ta/

I

Transceiver

RF Baseband Unit Computer

Contact Sensor

TV Cameras (#1 and #2)

Roomwith Objects and Stimuli

Snout

NOMADin its HOMEENVIRONMENT FIGURE79-2

NOMAD (NeurallyOrganizedMultiply AdaptioeDeoice)is a real-worldartifact constructed on pinciplessimilar to thoseof DaraninIIL AlthoughNOMAD's brain is simulatedon a powerfulsupercomputer, this "brain" doesnot act like a computer. NOMAD "lipes"at the Neurosciences Instituteand is thefirst nonlioingthing capable o! "lgarning"in thebiologicalsense of theword.With its "srrot4t" it picles up mngnetic blocksof diferentshapes and cohrsthat yield "oalue"(electicstimulation)on contact, WhileNOMAD loolcslikea robot,it doesnotoperate likea robotunderthe strictcontrol of a program.It operates like a noeticdeoice,onethat is neurallyorganized and worles according to selectionist pnncrples. Theneuralimpulsestrsnsmittedto N)MAD from its simulatedbrain (impulses thst in an animalwouldactioatemuscles) are translated insteadinto signalsfor NOMAD's wheelsby an on-boardmmputn,

Can we extend these notions to the construction of a primarily consciousartifact?The answer is not straightforward. But one might hazard a guessand say, in all likelihood, yes. In principle there is no reasonwhy one could not by selective principles simulate a brain that has primary consciousness,provided that the stimulation has the appropriate parts. But 793

HenuoNlrs there is much to be done before a consciousartifact can be successfully designed.For example,no one hasyet been able to simulatea brain system capable of concepts and thus of the reconstruction of portions of global mappings. This in itself is a very challenging task. Add that one needs multiple sensorymodalities,sophisticatedmotor appendages,and a lot of simulated neurons, and it is not at all clear whether presently available supercomputersand their memories are up to the task. If the proposed model for primary consciousnessis correct, such a simulation,possiblein principle, can be used to test the self-consistenryof these ideas. How? By enabling us to see whether an artifact capableof correlating a sceneby reentry between value-categorymemory and perceptual categorizationsbehavesin such a way as to choosecombinations of causallyunconnectedoutside eventsfor its own adaptioeneeds-according to its own assignmentof salienceand its own history. The test of efhcacyinvolves leaving the artifact's circuits intact and then cutting key reentrant loops, one at a time, to seewhat, if any, deteriorativeeffectssuch a disruption has on the artifact'sadaptive behavior. (Sucha procedureis a bit like testing the implicit-explicit dissociations discussedin the last chapter.)This illustratesone of the main valuesof simulations,one already proven in the case of perception machines like Darwin III. Given the complexity of neural pattems and behavior and their many levels of interaction, only a fast computer with a huge memory storage could hold all the pattems necessaryat eachlevel for scientiftcreview. Computersare not appropriatemodels of brains,but they are the most powerful heuristic tools we have with which to try to understandthe matter of the mind. Given what I have just said,you have probably guessedthe answer to the last of our questionsconceming artifactswith higher-order consciousness.It may be possibleto construct suchartifactssomeday,but right now it is so unlikely as to be unworthy of too much reflection. Not only will an artifact with primary consciousness alreadyhave to have beenmade,but we will also have to have understood how at leasttwo such systemscould intend something "to be what it is for each other under the auspicesof a symbol," as the writer Walker Perry put it (see the Postscript).In other words, artifacts with higher-order consciousnesswould have to have language and the equivalentof behavior in a speechcommunity. A great deal still remainsto be understood about the organization of linguistic memories, and a quick solution to this problem does not seemlikely. For now, we can relax in the knowledge that, so far, we remain the only known systemswith linguistically basedhigher-order consciousness, and competing artifacts remain a long way off. In principle, however, there is no reasonto believe that we will not be T94

I s l t P o s s i b l et o C o n s t r u c t a C o n s c i o u sA r t i f a c t T able to construct such artifacts someday. Whether we should or not is another matter. The moral issuesare fraught with difficult choices and We have enough to concemourselveswith in unpredictableconsequences. justify suspensionof judgment and thought on the human environment to the matter of consciousartifacts for a bit. There are more urgent tasks at hand. In thinking about these matters, we must remember how young a truly integrated scienceof the mind is. Of course,observationalpsychology is one of the oldest of "sciences."But psydrologically sophisticated neurobiology is in its infancy. So we may have to wait a while, indeed perhaps a long while, for the kinds of developmentsI have discussedin this chapter. My personal belief is that the construction of conscious artifacts will take place, under enlightened circumstancesand with due concem for human welfare. But it will take a long time. The pop artist Andy Warhol once approachedme at a party and told me that he collected scientificjoumals, but couldn't understandthem. He drifted away, then came back and said, "Do you mind if I ask you a question?" "Of 'lVhy does sciencetake so long?" I said, coursenot," I replied. He asked, "Mr. Warhol, when you do a picture of Marilyn Monroe, does it have to be exactly like her, as close to being her as you can make it?" He said, "Oh no. And anyhow, I have this place called the Factory where my 'Well, in scienceit has to be exact, as exact as you helpers do it." I said, can make it." He looked at me with limp sympathy and said, "lsn't that terrible?" One of the issuesI have left out of the discussionso far is what some philosophershave called "chauvinism" versus "liberalism." Mrct artifacts of the kind I have describedbe made of organic molecules?For perception machines,the answer is already in hand: no. But the close imitation of uniquely biological structureswillbe required.If our position on the mind is correct,however, liberalismwill never be the order of the day. Even with a completeknowledge of brain structures,the bet is that we will not be able to run successfullyon any sufficiently to design software for consciousness powerful computer in the manner demanded by functionalism (see the Postscript).The constraintsof morphology and selectionismrun counter to thesehopes. So the answer to the question posed in the title of this chapter is: in principle, yes, but the practical problems involved in "making" higherorder consciousnessare so far out of reach that we needn't concem ourselveswith them right now. As for the idea of a primarily consciousartifact, a somewhat stronger yes but with the caveat that much remains to be leamed about how a neural system mediatesconceptsin a body. As for 195

HlnuoNrus artifactswithout consciousness and capableof categorizing,prototypical perception"machines"alreadyexist. We havecomea long way with computersin lessthan fifly yearsby imitating just onebrainfunction:logic.Thereis no reasonthat we should fail in the attemptto imitateotherbrainfunctionswithin the next decade or so.Given the promiseof researchon syntheticneuralmodelingof the nervoussystem(of the kind usedin Darwin III), we may soonbe ableto considerwhatkind of performance wouldemergeif we hookedtenperception (or fl macJrines capableof categorizingnovelty to a Turing (or T) machinecapableof logic. The combination,a PT machine,may behave with respectto the recognitionof novelty roughly like a hunterand his dogs,providedthat the P machinesarewell trainedand the T machineis well programmedby a human operator.The resultsfrom computers hookedto NOMADs or noeticdeviceswill, if successful, haveenormous practicalandsocialimplications. I do not knowhow closeto realizationthis kindof thing is,but I do know,asis usualin science, that we arein for some surprises.

196

CHAPTER

20

Symmetry andMemory: On the UltimateOriginsof Mind

thing about the uni:rerseis that it is The most incomprehensible comprehensible, -Albert Einstein

'L'Homme pense;doncje suis,'dit l' Unioers,

-Paul Val6ry

osmology has formed part of the myth and the scienceof many civilizations.In it the mind has always played a central role, whether interior, exterior, or ulterior. It is natural for creatureslike ourselvesto wonder how everything came about, how we ourselvesgot here,and how we could come to be awareof the world in which we find ourselves. The religious cosmologies of the past have been replaced in some culturesby a scientificcosmology, one with remarkableties to the farthest reachesof theoretical physics. But as grand and mysterious and beautiful as this scientificcosmology is, it has no inherent principle that would lead us to ourselves:observerswho are conscious,who formulate physics and relate it to cosmology, and who have the urge to place ourselveswithin the scientific world view we have constructed.Even a "theory of everything," as somephysicistscall it, would be incomplete if it did not provide us with such a principle. 197

H^rnuoNrrs In this book I have maintained that mind has arisen in a very definite way thrroughthe wo*ings of evolutionary morphology. I have attempted to show that consciousnesshas arise& at least in this little speck of the cosmos,at a particular historical time. That it emergesfrom definite material arrangementsin the brain does not mean that it is identical to them, for, as we have seen,consciousness dependson relationswith the environment and, in its highest order, on symbols and language in a society. Higher-order consciousnessleads to a rich cognitive, affective, and imaginative domain-feelings (qualia),thought, emotions, self-awareness, will, and imagination. It can construct artificial mental objects such as fantasies.In culture,it leadsto studiesof the stablerelations among events (science)and among mental objects (mathematics),as well as to sfudiesof the relations among sentencesthat refer to events and mental objects (logic). The way I have proposed that the mind arose in nature may seem strange.This is partly becauseit does not seemto have come about in the same way that our most cherished constructions and inventions have arisen-through orderly relations suchas those guiding the logic, arithmetic, and physics that have led us to build computersand other informationbased devices. This does not mean,however, that a deep principle in nature underlying the evolution of consciousness cannot be found. In this final chapter,I want to speculateon what that principle might be and connectit to a more firmly grounded principle that most physicists would agree is among the most fundamentalin physics and cosmology. Then I want to ask how the two principles together might govem future scientificthinking and our view of how we fit into the cosmos. Physics and biology will "correspond" with each other in an intimate way, certainly in the next century and possibly even sooner than that. At the very least,they will have an exchangeabout how the human observer influencesphysical measurementsand how the observers'perceptionsrelate to their physical descriptions. This is a key problem in quantum mechanics(seethe Postscript). Thesepredictive remarksmay sound vague and utopian, but readerswill have to make their own judgments after they think about what has been and will be said here. As is the case for every scientific specialist,my knowledge and experiencehave sometimesbeen less than satisfactoryin trying to judge the relations between various fields of knowledge. Indeed, comparisonsof specializedexperienceamong scientistssometimesleadsto impasse.I was once askedby the man who taught me quantum mechanics and statistical mechanics,George Uhlenbeck, a very great physicist, to 198

Symmetry and Memory: On the Ultimate Origins of Mind introduce him to an equally great biochemist,Fritz Lipmann, who tutored me in graduate school. I arranged a dinner for the three of us. Fritz was eating his soup with gusto, when George, a man of seriousdignity, said, "l have concludedfrom a calculationof Gibbs potentials in various phases that life happenedonly once in the entire history of the universe." Fritz kept pushing spoonfulsof soup into his mouth, but between two of them, he looked up and said emphatically in his charming accent,"Happens all ze time." George remonstratedwith physical argumentsof great sophistication. Fritz listened,kept eating, and ftnally emptied his bowl. He put his 'Ven ve ground up ze pigeon breastmuscle,zey said spoon down and said: ve vouldn't get ze oxidative phosphorylation. Ven ve got ze microsome, zey said ve couldn't get ze protein synthesis. Ve did. Don't vor4r, it happensall ze time." They smiled at eachother but I am surethat eachleft the table frustrated with the intellectual limits of the other's specialty. This past century has seenperhaps the greatest revolution in scientiftc thought of all time. The revolution is not just in the adoption of unusual conceptsthat are necessaryto understandthe universeand its fundamental particles.It is in our notion of the scientificobserver and in the extension of the generality of scientificthinking. Beginning with Plancks discovery of the quantum,and continuing through the early formulations of quantum theory by Bohr and Heisenbergand of the theory of relativity by Einstein, observerswere no longer seenas fully detachedfrom their measurements. In quantum measurements,the way in which observerschooseto arrange their apparatusdetermines the outcome. In relativity theory, obseryers' measurementsof time and length depend on their relative velocity and acceleration.Thus, the observers'conscious choices in the one case and their physical location in the other must explicitly be taken into account. The outcome of efforts to rationalize these findings is well knownr Quanfum mechanicsand the general theory of relativity are the two grandest theoretical constructionsin science.They range in their descriptionsfrom the smallestand most shortlived fundamentalparticlesto the edge of the measurableuniverse. What is perhaps not appreciatedoutside the community of physics is that underlying both of thesedescriptionsis a key mathematicalprinciple: symmetry. This is not the place to go into the mathematics,but I shall try to say enough here to give you a glimpse of it. Symmetry is a stunning exampleof how a rationally derived mathematicalargumentcanbe applied to descriptionsof nature and lead to insights of the greatest generality. I want to discusssymmetry a bit becauseI plan to contrast it with another principle I believe underlies the mind, and indeed all of biology, the principle of memory. Later I will argue that an understandingof thesetwo 799

Hlnr'roNlrs principles,interacting in a tenseharmony, will allow us to seemore clearly the place of our minds in nature. We are all familiar with symmetry from daily experience.As creatures, we are roughly bilaterally symmetrical.We know that our mirror images have certain properties and that our right and left hands are mirror images of each other (figure 2fI). No operation in the real world will convert a right hand into a left without destruction. But we can convert a righthanded glove into a left-handedone if we fum it inside out. This suggests that certain operationsare necessaryto reveal certain types of symmetry. Symmetry principles and the rules goveming these operations are embedded in the mathematicaltheory of groups, which plays an essentialrole in the construction of advancedphysical theories.This mathematicaltheory was formulated in the early nineteenth century by a young French genius, Evariste Galois, whose life was cut short at age twenty and a half in a duel over a woman. Galois's ideas about groups revealedthe general insolubility of quintic equations (polynomial equations of the fifth degree; most nonspecialists will have stopped at quadratics,degree two). But it tums out that his ideas also have the most generalapplicability. The group of mirror reflectionsis concemedwith discontinuouscJrange,aswe have seen(figure ZO-l).Other groups deal with continuous symmetries-for example, translations in space.(This theory was advancedspectacularlyin the second half of the nineteenth century by the Norwegian mathematician Sophus Lie.) The highest symmetries are generally possessedby relatively feafurelessobjects such as a circle in two-dimensional space and a sphere in threedimensionalspace.In general (but not always), the addition of featuresto such objects results in a lower symmetry. Here we come to one of the formal constraintsthat bring deep insight to the laws of physics.This is the connectionbetween the ideasof symmetry and the so-calledconservationlaws of physics.The study of physicshas revealedthat a number of fundamentalquantitiesare conservedin mechanics and in both electricaland particle fields.Mass--energy,momentum, and spin are eachgovemed by conservationlaws requiring that eachis neither creatednor destroyed within the whole context of a physical description. Electric charge follows a conseryation law: The number of positively charged particles in the universe is equal to the number of negatively chargedparticles.This law has analogiesamong fundamentalparticles.The number that counts protons and other particles is conserved, as is the number that counts elechons and related particles. The consequencesof applying the principle of symmetry are truly beautiful, for the different laws set limits on the wavs in which these 200

Sy*metry

and Memory:

On the Ultimat e Origins of Mind

re o '\)/

\

-€

\

p_

/()t \-/

-"

a

-€

o

C

\

Kd

/\ nr\

\

n\l

v/

\ , t f !

"/\

\

\/ffi Y W0, \/

o

RotationalSymmetry

"Broken"RotationalSymmetry

(Sixturnsbringone back to the originalplacesetting)

(Theswitchof a knifeand a fork breaksthe symmetry)

Axis

'tr'.n - Axis

w -l-; I

i

+(-+--Axes '.J.

z r'

J,''i .r' MirrorSymmetry FIGUREzFT Somekinds of symmetry,

20r

,,"

i

t r..

HeRMoNrEs

particleswill interact with eachother. In other words, the rules describing the interactiozsbetween particles are constrainedby conservation principles. As a result, in some casesparticlescan only be createdor deshoyed in pairs,while in others particlescan be createdor destroyedwithout these constraints. Thus, we arrive at one of the grand themes of physics: Thereis a deep connectionbetween consercationlaws andsymmetry.Emptyspaceand time are symmetric; that is, they appear the same under many kinds of change. Space is the same regardless of translations, rotations, and changes in direction. When reversed,time is the samein either direction (l mean the time of the physicists, not your own personal senseof time). Indeed, in quantum mechanicsand relativity theory, the laws of motion themselves are invariant under symmetry operations such as rotation and translation. These invariances assure that the outcomes of physical events do not depend on the systemsof coordinatesused to measurethem. Without the application of a force, a body or a particle will not alter its velocity and direction of motion (its momentum) or its energy. The German mathematicianEmmy Noether first showed that the conservation of these quantities can be formally identified with symmetry principles. For example,the conservationof momentum correspondsto the symmetry of space under hanslation. The conservation of angular momentum corresponds to the symmetry of space under rotation. The conservation of energy correspondsto the symmetry of time under reversal of direction. (Time reversalcannot acfually be canied out, but the physical laws can be checkedfor their invarianceunder such operations.) It was Einsteinwho ftrst understood the significanceof the invarianceof the laws of physicsand, thus, of their symmetry. Indeed,his generaltheory of relativity may be considereda meansby which to searchfor conditions of absoluteinvariance! More recently, a seriesof discoverieshas made it possible to envision unifying all particle interactionsunder a single theory, a grand uniftcation theory (or GUT). This has not yet been accomplishedfor all four forces of nature-the strong, weak" electromagnetic,and gravitational forces. But partial theories have been posited that are stunning, theories not even envisionable twenty years ago. If there is a major language of these theories, it is the language of symmetry. The hope is that eventually the whole of nature (read"physics") will be describedby a symmetry that leads to all fields and forces in a unique manner. It would take us too far afield to discussall thesematters.I will, however, mention two notions that arefundamentalto physicists'efforts to construct a unified field theorv. These notions-also essential to modem cos202

Symmetry and Memory: On the Ultimate Origins of Mind mology-are the ideasof local gauge symmetry and spontaneoussymmetry breaking.Local symmetry can be contrastedwith global symmetry. To leave a global symmetry invariant in a given domain, any transformation that takesplacemust occtJreaerywhere. Local symmetry, by contrast,allows different transformationsto occur in different parts of spaceand time. A theory about local symmetry developed by C. N. Yang and F. E. Mills played a key role in the successof later unification attempts.For example, if one considers a fteld and if one wants to achieve invariance under a changeof local symmetry, things must be arrangedso that anotherfield will act to compensateexactly for any local changesintroduced by the first operation. To understandthe gist of symmetry breaking,considerthe bottom half of an empty wine bottle, a symmetrical upward dome with a gutter for sediment.If a ball is poised exactly at the peak of the dome, the sifuation is symmetrical.But the ball may spontaneouslybreak this symmetry and roll down the dome to some point in the gutter, a point of the lowest energy. The overall symmetry has spontaneously broken, although the bottle and the ball retain their individual symmetries.Applied to any given physical theory, this idea implies that a particularsolutionto the equations of the theory can be less symmetrical than the theory itself. Suchnotions underlie the electroweaktheory and the theory of strong interactions,two recent theoretical triumphs of modem physics. Another discovery of our century is that the laws of physics are stunningly general.Thus, notions of symmetry can be applied even to theories about how the universe came to be the way it is today. If one examines modem cosmologicaltheories (for example,those of inflation and the hot big bang) one sees that, as a function of decreasing temperature and increasingtime, the universe evolved to give rise to fundamentalparticles by symmetry-breakingevents.At sometime far into this process(and long after the particlesand fields we know were formed),galaxies,stars,and our solar system with its planets appeared.On Earth, by a process still unknown in detail, life originated and evolution occurred,leading eventually to the emergenceof mind. How Fritz Lipmann would have loved to have known what that processwas! What may we offer as a new principle underlying the evolutionary development of mind and intentionality in this set of events?I submit that the new principle is one of memory, one that takes many forms but has general characteristicsthat are found in all its variations. I am using the word "memory" here in a more inclusive fashion than usual.Memory is a processthat emergedonly when life and evolution occurredand gave rise to the systemsdescribedby the sciencesof recognition. As I am using the 203

HenuoNrrs term memory, it describesaspectsof heredity, immune responses,reflex leaming, true leaming following perceptualcategorization,and the various forms of consciousness(figure 2O-2). In theseinstances,structuresevolved that permit significantcorrelations between current ongoing dynamic pattems and those imposed by past pattems. These structuresall differ, and memory takes on its properties as a function of the system in which it appears.What all memory systems have in cornmon is evolution and selection.Memory is an essentialproperty of biologically adaptive systems. This extension of the term may seemhopelesslybroad. But let us see what all thesephenomenahave in common, for it is actually quite specific. What they have in common is relatioestability of structureunderselectioe mappingeomts.To make myself clear, I shall say something here about structure, stability, and mapping. The physical law concemedwith structure and stability is the secondlaw of thermodynamics.This law statesthat entropy, a measureof the disorder of a system, must spontaneouslyincreaseor remain the same but never decreasein a closed system. (By a closed system I mean one in which energy and matter neither enter nor leave.)The most orderly possible system is that of a perfect crystal (one with absolutely identical spacingof its atoms in a symmetricallattice) at a temperatureof absolute zero. Sincethe earliestevolution of the universeits entropy hasbeen increasing. But in parts of the universe that are open systems (ourselves,for locally as a result of the transferof example),entropy can actually decrease give rise to stable strucinteractions matter and energy. Various chemical tures, including those of moleculesin living forms. The stability of structures and their energetictransactionsare govemed by the laws of thermodynamics, including the second law. It is now clear that stable chemical structurescan exist in the absenceof life or living forms. Indee4 even in outer space,evidencehas been found of organic moleculesthat are similar to those in our bodies-molecules formed by the collision of nitrogen, oxygen, and carbon, for example.The conditions of their formation and dissolution, of their stability, are determined by energy and entropy. However stablethesemoleculesmay be, they lack a hereditary principle. They do not show any ability to replicatethemselves-to makemolecules that might be called their progeny by using their own structure as a template. I want to be clear about how I am using the word stability in connection with memory. After all, periodic crystals exist in nonliving domains (for example, in rocks) that add atoms to their structures to becomelarger, following the samerules of symmetry. Suchcrystalsdo not replicate;they grow. What is the difference? 204

S ym m e t r y a n d M e m o r y: O n t h e U l t i m a t e O r i g i n s o f M i n d

TYPESOF MEMORV HEREDITARY (covALENT)

Genetic Code

GTC GAC CTG GCA

Replication

+

GTC GAC CTG GCA

IMMUNE

1 2 0 3 1 5 1 1 8 2 2 . . . . . .M.

@

eo0@o@

n 9sr nn

LYMPHOCYTE

@

x e @8RX' o'0 by

e

// Cells 1F 1F Antibody-Producing

REFLEX (NEURAL) SpinalCord

Synaptic Change-d'r/

;'{/ SensorySheet

Muscle

RECATEGORICAL NeuronalGroup Selectionin ReentrantBrainMaps

COMPLEX BRAINS

FIGURE2F2 Somekinds of memory,

Han N,rowrss In a replicating system, there is an apeiodic structure that undergoesa kind of mapping; think of the sequencesof DNA in chapter 6. A chemical reaction faithfully copies the aperiodic strucfure, resulting in daughter structures.But this fidelity is not absolute;mutant strucfuresarise that are also copied. The result is a variant population. Finally, the stableaperiodic strucfure mapsthrough additional chemistry to make other kinds of structures that contain it, so that favorable variants have a selectiveadvantage when further copies are made. This abstracteddescription correspondsto that of a living system: a self-replicatingsystem undergoing natural selection (see chapter 6). The aperiodic structureis DNA or RNA" and the container to which they map is made of various protein products. But notice that the main process, which is lacking in nonliving forms, is the hereditary principle. Notice also that this hereditary principle, which allows an increasein the population of favored variants over time, dependson the stability of chemicalbonds. In the caseof DNA, theseare the covalent bonds linking different nucleotide basestogether to make a genetic code of linked triplet codons, each one correspondingto one of twenty amino acids that make up protein chains. The energy and entropy conditionsin the temperaturerangeunder which life flourishesassurethat a hereditary processtakesplace.But it is historical selection events that result in the actualsequencesfound in the population. The appearanceof this hereditary processis a new kind of event-a form of memory. Aside from variations introduced into the sequencethat have proven favorable, it is the ability to retain much of the order or mapping of the parent aperiodic structure that enablesthese systems to continue. They have stability of structureunder selectivemapping events. But notice that this "memory" is not perfect (asit must be, by contrast,in computer messages).Indeed, to some degree, it musl contain errors (changesin entropy) or mutants for the system to be a selectiveone-to be one that is able to respond adaptively to unforeseenenvironmental events becausethere is population variance. As thesestructuresevolved and cellularpopulationsformed into animals with many linked cells and with nervous systems,a new kind of memory appeared.This occuned as a result of synaptic changesin the nervous systemsof these animals.Becauseof neuronal group selection,behaviors that proved adaptive were stabilizedby selectionwithin a single animal's lifetime. Memory basedon synaptic changeis essentialfor suchbehaviors. In vertebrates,the requirement that their immune systems make the distinction between self and nonself resultedin the selectionof individuals who had a variant of the gene for the neural cell adhesion molecule, N-CAM. By introducing somatic variation into what were to become 206

Symmetry and Memory: On the LII|imate Origins of Mind immunoglobulin (antibody) moleculesand by combining that processwith the faithful replication of cells selectedby foreign molecules,a new recognition systemappeared(seefigure S-1). This systemhad immune memory: The selection of lymphocytes by antigens led to changesthat were retained for the entire lifetime of the individual. Yet again,the evolutionary elaborationof sensoryreceptorsand motor sheetsin animalswith increasinglysophisticatedbrain maps made memories based on perceptual categorizationpossible.With the appearanceof conceptual capabilities and even more sophisticated mapping, synaptic change in responseto novelty occurring within populations of neuronal groups led to additional kinds of memory. Each memory reflects a system property within a somatic selection system.And eachproperty servesa different function basedon the evolution of the appropriateneuroanatomicalstructure.Thesehigher-order systems are selectiveand are basedon the responsesto environmentalnovelty of populations of neuronal groups arrangedin maps.They are recognition systems. At some transcendentmoment in evolution, a variant with a reentrant circuit linking value-category memory to classiftcationcouples emerged. At that moment, memory becamethe substrateand servant of consciousness. With the emergenceof language in the speciesHomo sapiens,the iteration of this same principle in specializedlinguistic memories made higher-order consciousness possible.And within culfure,higher-order consciousnesseventually gave rise to a scientificdescriptionof nature,one that allows us to study the origins of our own existencein the universe. This description of the developmentof memory is so different from the previous one describingthe developmentof the cosmosfollowing symmetry principles as to seemincommensuratewith it. The biological story is a local saga so far told only on Earth: It is historical, it occurs in a very narrow temperature range, it is extraordinarily complex and specific to particular structures, it takes unexpected and different forms, and it is dizzying to considerin detail. But the sagabegins in a world govemed by symmetry. Only with symmetry breaking only with the formation of chemistry, only with the appearanceof large, stable molecules,only with the appearanceof irreversible selectionevents, only with the evolution of meansdescribedby the sciencesof recognition, could memory lead to the appearanceof mind. symmetry principlesgovem the possibility that memory can arise, but only after symmetry breaking occurred, leading to chemistry and to living and evolving organisms,could memory develop. Memory underliesmeaning.With the transformationof meaning made possible by the embodiment of concepts as describedby the TNGS, it 207

HlnuoNrrs becamepossible within human cultural history to develop true information-processingsystems.The historical development of scienceby social transmissionin human culture has made it possiblefor us to loop back into the truth thrroughchains of knowledge (see ftgure 1.4-1). But unlike the development of memory, this explosive transmissionis no longer Darwinian. It follows Lamarckian rules because of the character of informational systemsand the nature of meaning itself. The contents of informational systems are transferred by use;no genetic hereditary principle is needed.The transferis to somaticsystems,eachone unique,and the results have been stunning-the transformationof the environment by the human mind in ways that are both valuableand honendous. They are resultsthat should inspire caution as well as deserving pride. I have tried in this book to develop a view of the mind based on scientific evidence. Given the state of our knowledge, this view must it is still remain speculative.Although it has philosophical consequences, our ignorance basicallya scientificview, subjectto disconfirmation.Despite

of many of the detailedworkingsof the brain,I believeit is importantto theformulationof suchviewsnow. One of themwill helppoint encourage us in the right directionfor a while.This is the most one canexpectof a it providesus.A theoryexists scientifictheory,beyondthe understanding theory. so that we may build a better Lateasit is in thisjoumey,it will not hurt to stressagainthatwhatI have one.It remains beenconcemedwith hereis a theory-and not anaccepted doing so in my previous ways of to betestedrigorously,andI haveproposed trilogy on morphologyand mind. Like all the theoristsI have known,I Theunit of selection believethatmy theoryiscorrectuntilprovenotherwise. Onehundredof usgo theorycreationisusuallya deadscientist. in successful to our gravescertainthatwe areright,but only onetumsout to beso.Rarer Butwe eachmustactas suchacquiescence. still is a living scientistaccorded if theoriesareasimportantasany otherscientificpursuit,risky asthey are' Hope and belief are as importantin scienceas they are elsewhere;the they mustyield to experiment. is that in science difference The theoryof the mind I haveput forth heredisclaimsthe possibilityof knowledgebeyonddoubt.Thisshouldnot disappointus,giventhehistory of sciencein the last threecenturies.If the futurecourseof of the success at allby its presentreach,we mayexpecta remarkable is determined science in thenextcentury.Buta "theoryof everything"will certainlyhave synthesis to includeboth a theory of the mind and a fuller theory of the observer. of the will unitein amorecompletecomprehension Physicsandneuroscience relationbetweentheprinciplesof symmetryandmemory.Theywill existin not only the a tenseharmony,one that will makeit possibleto understand andtheirplacein it. world but alsohumanobservers 208

I

EPILOGUE

I started this book by telling you I thought its sub;ect was the most important one imaginable.This statement is obviously true in the sense that without a mind there is neither a subject (you or I) nor any subject matter. But I hope our trip thrroughthe layers and loops-from molecules to mind and back again even to fundamentalparticles-has persuadedyou of another, less obvious aspect of the importance of neurobiology: that without an understandingof how the mind is basedin matter, we will be left with a vast chasm between scientific knowledge and knowledge of ourselves. This chasm is not unbridgeable.But biology and psychology teach us that the bridge is made of many parts.The solution to the problem of how we know, feel, and are aware is not contained in a philosophicalsentence, however profound. It must emergefrom an understandingof how biological systemsand relationshipsevolved in the physical world. When that evolution resultedin language,the imaginableworld became infinite. There is great beauty and much hope in the realization of this open-endednessof imagination. But we must continually retum from that world to the world of matter if we are to seehow as consciousobservers we are acfually placed within our own descriptions.Analyzing that placement will be one of the major goals of the scienceof the future, What form this sciencewill take, it would be foolish to predict. It is sfficient and consoling to know that, whatever form it takes,the conscious life it describeswill always remain richer than its description.

209

MI N D w I THCOUT B I OL o GY A C RI T I C A L P S TS C RIP T

No one likes to spendmuch time being critical when there is creativework to do. But in order to explain why the kind of biological theory put forth in this book is needed,I have to do a bit of bashing-to criticize several received ideas and establishedpoints of view. As I stated in the body of this book a number of prevailing views about consciousness and the mind are simply untenable,however well establishedthey may be. Why bother with them at all? There are two reasons.First, they are dangerously seductive;sooner or later even the uninitiated reader will run into one or another version of them. And second,a critical analysisof these notions helps to define further the nature of our tash which is to show how the mind is embodied. There is a third reason: wrong as they may be, these views-that shange physics may hold the key, that the brain is a computer, that we have a kind of built-in language machine in our head-are interesting, whatever their deficiencies.But to convey that interest involves presenting some tedious detail and some rather abstractargumentsthat would have intemrpted my descriptionsof the biology of the brain. Therefore I have decided to save my critique of these views for this Postscript. My goal is to dispel the notion that the mind can be understood in the absenceof biology. What I am presentinghere are not afterthoughts;they are extensionsof points made in the body of the booh intended for the experts but also for the curious who may want to know more. Readersshould not be suqprisedthat the discussionencompasseslarge numbers of disciplines and jumps from one to the next. The hardest to graspare perhapscognitive scienceand linguistics,both abstractmultidisci2'1,7

MrND WrrHour

BloLoGy: A

CnrrrcAL PosrscRrpr

plinary areas.But once the obstaclesare cleared,they are also fascinating and enolrnously challenging.Before taking them up, let us turn again to physics.

PHYSICS: THESURROGATE SPOOK A spook is a specteror ghost, a disembodiedspirit that haunts or scares you. It must seem strange that I am calling that most rationally based of sciences,physics, a sunogate spook. But this is what it becomeswhen it is applied directly to the mind. Let me explain what I mean. One way out of the dilemma imposed by the embodiment of mind and the apparentmysteriesof consciousness is to makemind and consciousness direct properties of matter. In its most extreme form, this becomes the philosophical doctrine called panpsychism. Panpsychismproposes that all matter, even the finest particles, is a bit conscious,or even that the whole universe is conscious.After all, the reasoninggoes, we want to be able to say that mind and matter are connected. If we get a sufficient number of very slightly conscious particles together in the right way, the end result is a conscioushumanbeing. This view doesnot say how one would determine that a particle is conscious,much less a human being. Sucha position "scientizes"another view originally basedon the philosophy of idealism.In it the world is perceived only through the mind and thus perhaps,as Bishop Berkeleyproposed,there is no matter, only mind. On hearing this, Dr. Johnsonkicked a stone and stated, "l refute it thus." A better refutation comesfrom the theory of evolution: If natural selection gives rise to sentient animals, it is difficult to see how the selecting environment and the brain can both be mental events in a single sentient animal that also has progeny undergoing natural selection.The mind reels trying to comprehend how such a complication would ever come to pass. The theory of natural selection did just as much damage to Plato's idealist notion of essentialism-that there is a world of perfect essencesof which the exemplars in the actual world are merely flawed examples. Speciesare not essencesor types; they are the result of selection from variation. Somevery intelligent people have been attractedto panpsychism,idealism, and essentialism.One was the Irish poet William Butler Yeats, who 272

Mind Without Biology: A Critical Postscript wrote the mystical tract "A Vision" and someextraordinary poems reflecting his thoughts on occult matters. Brains and intellecfual gifts are no guarantee against attraction to the spooky and mystical. Under some circumstancesit is consoling to have suchbeliefs,particularly if one clings to notions of immortality. But as my mother said as she lay dying, "l'm in no hurryr." When asked why, she smiled and said, "Becauseno one has come dancing back to tell me what a good time they've been having." Most good physicistsare hardly committed to notions of panpsychism or disembodiedspirits. But some very good physicists have nevertheless reachedbeyond the biological facts and have supposedthat the answers for example,will reside in a new theory of to the riddle of consciousness, physics,such as a theory of quanfum gravity. To explain why they might be tempted to do so, and why I think they are simply providing us with a surrogate spook I have to say a few more words about the differences between physics and biology. Physicsis the mother of all the sciences:the earliest,the most general in its reach,the most fundamental.It differs from biology in its generality: it applies equally well to all intentional objects (including human beings) and nonintentional objects. In contrast,biology as we know it is specific. It concemshappeningstaking placewithin a very narow range of temperature (or energy), pressure,and chemistry. Even more specificis the fact that biology is historical. Evolution is based on a particular historical sequenceof natural selectionfrom populations of variant organisms.Nothing of the sort has to be consideredin formulating the general laws of physics. This century has seen an astonishing intellectual revolution based on Plancks finding that energy is radiated from matter in finite, discrete packets,or quanta, and on Einstein'stheory of relativity, which replaced spaceand time with the notion of spacetimeand advancedthe notion of gravity and matter as representing the curvature of a four-dimensional spacetimemanifold. The elaboration of these revolutionary ideas led to changesin our ideasof measurement(figure P-1) and radically challenged ordinary notions about the simultaneity of events and about causation. These ordinary notions were replacedby strange, or at least unfamiliar, ones. The elaboration of Planck's and Einstein's work also led to some extraordinary problems that remain unsolved to this day. Their "strangeness"has tempted some scientiststo tuck the problem of consciousnessin with them. The ideasbehind thesebasic physical laws can indeed be strange (read "unfamiliar," in the senseof "not commonsensical").Unlike the ideas of biology, they are oery generalandare often best expressedin mathematical 213

MrND WrrHour Subatomic Particles

BroLoGy: A CnrrrcAL PosrscRrpr

Large Molecules

Universe (Farthest Known Reach)

Atoms

-io,1o'ffi

en'emisilt

Hn!qceq!r'1qq-

-)rr{

n'"Eliaayrlfo-t

Scalesof NatureandApplicable Theories Scales ofnature asestablished r, ,rrii1i}r'rltlo-r,

cm,ttwtheory ofretatirity and

quantumtheorydo not hold.At the leaelof molecules and below,qwntum thiory is essmtial(andof course it applies at all scales abooe10-33 t^).at iny highvelocities andaccelerations onemustapplyrehtiaifu theory,Butat thelevelof mauoiopic objects (indudtnshumansandtheirbrains),one deicrip,mayapyoximatenicelywith classical tions;theconoergence of quantumand classical theoiesat largi scales is lenownas ihe c3rrespoydmce pi"t pP.Notethat thesizerangeof brainsandthetemperature rangeof liuing ,thingsare both quite nanow. The siale is in powersof tm-that is, it is logaithmic.

theoriesthat have great power and beauty. one exampleis that of symmetry, which I discussedin chapter 20. sugh notions of physics,with their generality and predictive power, are beguiling. As powerful as they are, however, p.ofouttd problems arise in understanding their application. An example comes from the theory of quantum measurement,which must be taken into account when one attempts to measurethe position or momentum of a fundamentalparticle.In facing the paradoxesthat arise from these attempts, distinguishedmathematicianssuch as fohn von Neumann and equally distinguishedphysicists such as Eugenewigner were tempted to propose that consciousness itself causally intervenes in the processof quantum measurement. There are many issuesrelated to these proposals,and to discussthem all here would take us too far afield. But let me sketch out one aspectof the quanfum measurementproblem to show why these scientists were tempted to bring consciousnessinto physics.As I do so, however, please keep in mind that physicsis concemedwith formal correlationsof the most general properties of things in spacetime.Theories of physics are not 214

Mind Without Biology: A Critical Postscript concemedwith the sensesproper, with categorizingnameablemacroscopic objects,or with intentionality. If one delves into quanfum mechanics,it is easy to forget these restrictions, becausethe decisions of the observer appearto affect the measurementshe or shemakes.To understandthis, we have to consider a few salient featuresof quantum theory. Quantum theory is the most generally applicable of all theories. In dealing with enormous energiesand very small particles,this theory has revealedbehavior that confoundsordinary expectations.For example,one particle cannotbe identiftedas distinguishablefrom another.Particlesshow duality of behavior: Under one set of circumstancesthey arebest described as waves, in others as particles.Indeed, as Max Bom first suggested,the fundamental wave function r[ in the Schrodinger wave equation, when taken as an absolutevalue and squared,is a measureof the probability of finding a particle in a given position of space-anywhere! If one attempts to determine that position in an experimental setting, however, one loses forever the possibility of determining the momentum to the sameprecision.This so-calledHeisenberguncertainty is fundamentd; there is a conjugate relation between the position and the momenfum (the mass times the velocity) of a particle, and this relation sets the precision of the product of thesevariablesto a value no lessthan Planck's constant.This is not iust becauseto measurea particle'sposition precisely one must use particles or waves of much smallerwavelength and thus of higher energy, inevitably "kicking up" the particle's momentum. It is a fundamentalproperty of the theory. In consideringthis relationship oPera' tionally, one begins to get a feeling for the strange flavor of quantum theory. If one (the physicist observer)choosesto measurethe position of a particle to a certain precision,the act of setting up and carrying out the measurementprecludesforever and irreversibly the measurementof the momentum to a similar precision.According to the theory, however, no bias exists before the measurement:The wave function rf is a linear combination of functions describingall possibleoutcomesof the measurement, and when a measurementis made the wave function "collapses"or "projects onto" one of the possible outcomes. As von Neumann pointed out, the macroscopicmeasuringinstrument is also describedby a quantum mechanicalwave function (practically speaking, we do not need quantum theory to describesuch objects physically). He then formally showed that one cannot draw a line from the wave function of the particle all the way up to the act of the observerto establish the value of rf at any scale. The "collapse of the wave function" is determined just when the apparatusand the particle interact to give a definite measurement.This collapsewas attributed by Wigner to be the 215

MlNp Wrrnour

Blorocy: A Cnrrrcer Posrscnlpr

result of the intervention of the observer's consciousness.After all, the observer decides to set up the apparatus,decides whether he or she is interestedin position or momentum, and actually makesthe measurement! To determine the state of this apparatusin von Neumann's view, one apparatus needs another, and that one needs another one, and so on, regressingin an infinite fashion. In Wigner's schemea phenomenononly becomesactual (that is, the regressis ended)when the observer becomes consciousof it. In all faimessit should be said that other distinguishedphysicists have interpreted the quantum measurementproblem without calling the consciousnessof the observer into play. Niels Bohr, the father of quantum theory, declaredthat there is no ultimate or deep reality; one simply applies the principle of complementarity (of which Heisenberg'sprinciple is perhaps the most elegant expression)and then obtains the result dictated by the entire situation of measurement,particle, apparatus, and observer. Bohr's "Copenhageninteqpretation"is the position taken by most physicists who use the theory. It gives a formula describingwhat one observes with an apparatus,one that is ultimately made up of the same kind of quantum particlesone is measuring.Other physicistshave even proposed that there is no "collapse" of the wave function. Insteadthey conceivethat there are "many worlds," in each one of which the function takes on a possiblevalue altemative to the one in this world with this observerwhom we seehere and now. Still othershave proposeda "quantumpotential" that might even involve faster-thanJight signaling, something that contradicts Einstein'stheory of relativity! I once discussedthis problem at lunch with my friend Isidor Rabi,a great physicist, just five months or so before his death. He looked at me with a puckish smile and said, "Quantum mechanicsis just an algorithm. Use it. It works, don't worryr." I nagged him, saying "Rab, don't tell me you are getting like Einstein,dubious of the whole thing." He replied with a laugh, "Listen, if I'm having trouble with Go4 why shouldn't I have trouble with quantum mechanics?" This brings us to the issue at hand: with such strangeness,why not get a little stranger and propose that additional, as-yet-undiscovered physical ftelds or dimensions might reveal the true nafure of consciousness?This is a subtle but also more off-putting way of proposing physics as a surogate spook. A good example is the position taken by the mathematician and cosmologist Roger Penrosein his wide-ranging book The Emperor'sNew Mind, which takes as its theme the nature of consciousness.The book abounds in clear examplesof paradoxesin physics and in descriptions of the axiomatic limitations of mathematics. On intuitive 216

Mind Without Biology: A Critical Postscript grounds alone, based on his personal experienceas a mathematicianand in appreciationof theseaxiomatic limits, Penroserejects the notion of the brain as a computer. He points out the limits of quanfum mechanicsand relativity in domains where the dimensions are so small (below the so-33 cm) that such theories cannot apply. And called Planck length of 10 he calls for a theory of quantum gravity that would extend these theories. Then, by a remarkableleap, he proposes that the mystery of consciousnesswill be resolved when a theory of quantum gravity is satisfactorily constructed. I suppose the reasoning is as follows: The decisions of the observer and the operator are intimately involved in quantum mechanicaland relativistic measurements.The observe/s mind constructsand appliesmathematical theories, the statementsof which transcendthe ability of formal axiomaticsto prove or disprove them, yet he or she can cirecktheir truth and meaning as a computer cannot. Like everything else, the observey's brain is ultimately made up of particles obeying quantum laws, particularly at its synapses,where most of the action is. Physical laws as currently formulated do not account for consciousness.Neither can they explain quantum gravity. Perhapsan explanation of quantum gravity will provide the clue to consciousness,which seemsto hover around all of our theories! Ttuly, this is physics as the surrogate spook-more reasone4 perhaps, than many a spook in religious tracts or in occult accounts,but no more rewarding in the end. Indeed, while Penrose'sbook contains many fine descriptionsof physics, it bears little on the problem of consciousnessas intentionality, for it ignores both the psychological and the biological knowledge essentialto understandingthe problem. Penrose'saccount is a bit like that of a schoolboy who, not knowing the formula of sulfuric acid askedfor on an exarn,gives instead a beautiful account of his dog Spot. \{hat is missing from his and other accountsis a sober scientificanalysis of the proximate structures and functions related to awareness:an account of real psychology, of real brains, and of their underlying biology. While physics obviously provides the necessarybasesfor biology, it doesn't concem itself with biological structuresand processesand principles. These are quite special and much more demonstrably connected with the mind than are the general ideas of symmetry and quantum measurement,important as these are to understandingthe existenceof all things. Indeed, it is a much more sensiblething to construct and test a theory of mind based on biological processesthan to posfulate exotic physics as an explanation.There is plenty of direct evidence,after all for anatomy affecting consciousness. 277

MrNp Wlrnour Blorocyr A Cnlrrcer posrscnrrr until we reacha biologicalimpasse,therefore,we would do wen to rejectas a categoryerror the notion that exoticphysicsitself will give a descriptionof theobserver's consciousness. we mustnot confusethebases of the workingsof our mindswith our minds'fine intellechralconstmctions, suchas theoriesof physics.(An irreverentdescriptionof a horse showmay focusour attentionon the natureof the categoryerror:a horse showis a bunchof horsesshowingtheir assesto a bunchof horses'asses who are showingtheir horses.) We mustnonetheless begratefulto Penrose, a greatscientist, at thevery leastfor drawing renewedattentionto an even more commonlymade categoryerror:that whichassumes thebrainis like a computer.Letus tum to it, for its consideration bringsus much closerto the fundamental issue than doesany furtherconsideration of physicsitself.

DIGITATCOMPUTERS: THEFAISE ANALOGUE If physics won't do as a surrogatespook what about an unusualphysical object or construct-the digital computer?After all, this most remarkable of all inventions of the twentieth century seemsto carryrout a remarkable number of functions that at ftrst glance appear to be mindlike. Extraordinarily silly things have been proposed about the capacity of machinesto think. For the most part, the sillinessarisesfrom the analogy between thinking and logic. The indisputablefact is that computerscarr5r out logical operations.The rub is that logic alonecarriedout on a computer no more constitutes thinking than the physical events of adding up numbers on an abacusresemblewhat goes on in the brain during the performance or creation of arithmetic by a mathematician. To seewhy this confusionhas ariserl I must explore a bit of the theory behind the digital computer. That theory owes itself largely to the work of the late Alan Turing, a British mathematicianwho committed suicideby biting a poisoned apple. As a discoveredhomosexual,he had been given a forced choice by the British courts either to go to jail or to take the feminizing hormone estrogen.He chosethe latter, with feminizing effects on his body, and who knows what effect on his brain. That brain gave rise to a powerful set of mathematicalideas,one of which is known as a Turing machine. 218

M i n d W i t h o u t B i o l o g y' A C r i t i c a l P o s t s c r i p t Turing defined an abstract class of automata and showed that any memberof that classcan computeany of a large classof functions.(All but computersare Turing machines.)A Turing machine a few special-p.r{pose (figure P-2) is a finite-statemachinewith an infinite tape;in a given square on the tape it can write either a 0 or a7, and it can shift the tape one square (containing one such digit; to the left or to the right. It has instructions

"Head"

Step 1 Rgad

rrrrrrrrrrrrr

Step2 Move left one "box"

0- 1 -0- 1

Read Erase W ri t e Move Halt

0-1 -1 -Qrrrrrrrrrrr,

n

-!r-- =

Illrlllllrlll

0 - 1 - 0 t 1 { 0 - 1 - 10-- 1

ltrrttrllrllr

0-1-0

Infinite

Tape

-1 -0rrrrrrrlrrlr

Step3 Erase1 Write0 I I

0-1'1'

I

: etc.

StateTable

Program lf a 0, move left one box

S=\

lf a 1, erase

S=/

Write a zero

S=

Continue

etc.....

I I I I

A TURINGMACHINE FIGUREP-2 A Turing machine.This abstractionhasbeenshown to representthefunctional operations of practically all computers.Turing's analysisholdsfor real-world computerseoenthough a Turing machine(unlikea real-world computer)hasto go through many morestepsthan is conoenientto carry out a simple information procnsing Tocedureor algorithm. The idea is a triumph of clear reasoning. 2-1,9

MrNp Wrrnour

B r o r o c y : A C n r r r c , c , rp- o s r s c n r p r

containing conditions and actions,and it carriesout an action if a particular condition is satisfied.The condition is determined by the symbol on its tape under the tapeheadand by the stateof the machine,and a given action is any one of the four describedabove,after which it shifts to the next state specifiedby the program. A "universal Turing machine" can simulate any particular Turing machine.(ParticularTuring machinescan have different mechanismsand parts, as long as they obey Turing's description.) Now comesthe temptation to commit a category error. A persuasiveset of argumentsstatesthat if I can describean effective mathematicalprocedure (technicallycalled an algorithm; seeftgure p-4), then that proiedure can be carriedout by a Turing machine.More generally,we know that any algorithm or effectiveproceduremay be executedby *y universalTuring maclrine.The existenceof universalmachinesimplies that the mechanism of operation of any one of them is unimportant. This can be shown to be true in the real world by running a given program on two digital computersof radically different constructionor hardwaredesign and successfullyobtaining identical results (seefigure P-3). on the basis of these properties, the workings of the brain have been considered to be the result of a "functional,, process, one held to be describablein a fashion similar to that used for algorithms. This point of view is calledfunctionalism(and in one of its more trenchantforms, Turing machine functionalism). Functionalism assumesthat psychology can be adequately described in terms of the "functional organization of the brain"-much in the way that software determines the performance of computer hardware. Functionalismis concemed not only with functions performed by various systemsbut also with the relations between their components,particularly as they causeother relations to take place.Functionalist theoriesare indifferent to the mechanicalinstantiation bf a system, and thus they deal in abstract terms with such relations. In the functionalistview, what is ultimately important for understanding psychology are the algorithms, not the hardware on which they are executed. According to functionalism,what the brain does may be adequately describedby algorithms.Furthermore,the tissueorganization and composition of the brain shouldn't concem us as long as the algorithm "runs" or comesto a successfulhalt (figure P-a). (This "liberal" position affirming the absenceof any need for particular kinds of brain tissue suffusesmuch of present-daycognitive psychology.) If we accept this position, an analysis from formal logic known as church's thesis suggeststhat if any consistentterminating computational method exists to solve a given problem, then a method exists that can run on a Turing machineand give exactly the sameresults.For problems that 220

M i n d W i t h o u t B i o l o g y' A C r i t i c a l P o s t s c r i p t

FIGUREP-3

22r

MrND WrrHour

BloLoGy: A CnrrrcAL PosrscRrpr

. AN ALGO ALGORITHM FORBOILINGAN EGG -T O

Put water in pot

@ Turn on heat O

ff water is not boiting go to qtep O otherwisecontinuet" ilt6-'='

@ Put egg in water

O

Set timer to 3 minutes

@ ff timer has not gone off goJg step @ otherwisecontinueto step Q) @ rrtn off heatand cool @ Finished... retrieve,peel,and eat egg

FIGUREP-4 An algorithmfor boilingan egg,Onefor addingtwo numbersrpouldhaoeequallyexplicit instructions,

can be solved consistentlyin a specifiedfinite amount of time, a Turing machine is as powerful as any other entity for solving the problem, includingthebrain,According to this analysis,either the brain is a computet or the computeris an adequatemodel or analoguefor the interestingthings that the brain does. This kind of analysisunderlieswhat has becomeknown as the physical symbol systemhypothesis,which provides the basisfor most researchin artificial intelligence.This hypothesis holds that cognitive functions are carriedout by the manipulationof symbolsaccordingto rules.In physical 222

Mind Without Biology: A Critical Postscript symbol systems,symbols are instantiatedin a program as statesof physical objects.Strings of symbols are usedto representsensoryinputs, categorieg behaviors,memories,logical propositions, and indeed all the information that the system dealswith. The operationsneededto transform shings of input symbols into strings of output symbols are computations, and according to the physical symbol system hypothesis,they may therefore be carried out by any suitably programmed Turing machine.As I have said, these operations are purely formal in nafure; that is, they may be carried out without referenceto the meaningsof the symbolsinvolved. (As we saw in chapters 2 and 1.2,a set of these rules is known as a syntax.) The particular design of the computing device responding to these syntactical rules is of concem only in that it must meet certain requirementsof speed and memory capacity in order to be able to complete its work in a reasonableamount of time. Why won't this position do? The reasonsare many, but before I take them up, remember the connection of physical symbol systemswith the argument for functionalism (which has many variants, all of which sharethe formal causalposition). If any of the forms of functionalism is a correct theory of the mind, then the brain is truly analogous to a Turing macirine. And in that case,the relevant level of description for both is the level of symbolic representationsand of algorithms, not of biology. Not all forms of functionalist theory impose such a large degree of identification of processesin the mind with processesin Turing machines. The strongestpositioo originally formulatedby Hilary Putnamand known as "Turing machine functionalism," posfulates that the two are entirely equivalent. This view is no longer widely accepted;indeed, it has been rejected by Putnam himself. Weaker forms of functionalism do not require a strict equivalencebetween brain statesand Turing mac-hinestates.However,all forms of functionalismhold that two systemshaving isomorphous functional statesmust be in identical cognitive states,irrespectiveof any differencesin their physical makeup.This conclusion is a close cousin to some of Turing's results on universal computation. These results amount to the assertionthat two computershaving identical abstractstate transition tables and identical symbols on their tapes (seefigure P-2 for definitions and examples)are carryringout the samecomputation, regardlessof the physical form taken by the processorand the tape. And now the coup de grace(actuallymultiple coups)!An analysisof the evolution, development, and structure of brains makes it highly unlikely that they are Turing machines.As we saw in chapter 3, brains possess enorrnous individual structural variation at a variety of organizational levels.An examinationof the meansby which brainsdevelop indicatesthat 223

MrNo Wrrnour

Brorocy: A Cnlrrcer Posrscnrrr

each brain is highly variable. Indeed, a simple calculation shows that the genome of a human being (the entire collection of an individual's genes) is insufficientto specify explicitly the synaptic structureof the developing brain. Moreover, each organism'sbehavior is biologically individual and enormously diverse, whether or not that organism registers or reports subjective experiencesas human beings can. More damaging is the fact that an analysisof ecological and environmental variation and of the categorization procedures of animals and humans (which I discussin the next section) makes it unlikely that the world (physicaland social)could function as a tape for a Turing machine. Arguing along similar lines, Putnam has repudiatedhis original and other dependentmodels of functionalism.His central point is that psychological states including propositional attitudes ("believing that p," "desiring that p," and so on) cannot be describedby the computationalmodel. We cannot individuate conceptsand beliefswithout referenceto the environment.The brain and the nervous systemcannot be consideredin isolation from states of the world and social interactions.But such states,both environmental and social, are indeterminate and open-ended.They cannot be simply identified by any software description. Functionalism,construed in this context as the idea that propositional attitudes are equivalent to computational states of the brain is not tenable. Another philosopher, Iohn Searle,has also been a strong critic of the functionalist position. His opposition is based on the idea that no purely computational specificationprovides sufficient conditions for thought or for intentional states.His argument (which applies to higher-order consciousness,the kind we have as humans)is that computer programs are defined strictly by their formal syntactical structure,that syntax is insfficient for semantics,and that in contrast,human minds arecharacteriz.edby having semantic contents. Semantic contents involve meanings, and a syntax doesnot in itself deal with meanings.The rejection of functionalism in this position is obvious. Moreover, Searlemaintains that, inasmuchas consciousnessis identified in humanswith a type of intentionality that is inexorably accompaniedby subjective experience,by definition no organism canhave intentional statesif it lackssubjectiveexperience.Computers lack such experience.Certain functionalists(possibly the majority) restrict their claims to statementsthat precludesubjectiveor phenomenalproperties. Given his arguments,Searlewould reject their claims(rightly, I think) as having no bearing on the origins of consciousnessor thinking. What is at stakehere is the notion of meaning.Meaning, asPutnamputs "is it, interactional.The environment itself plays a role in determining what a speaker'swords, or a community's words, refer to." Becausesuch an 224

Mind Without Biology: A Critical Postscript environment is open-ended,it admits of no a pioi inclusivedescription in terms of effective procedures.Moreover, we have seen in this book that the acfual body of the speakerplays an equally great role in determining meaning.Arguments conceming semanticsand meaning are important for any theory of consciousness(and thinking) that takes as its canonical referenceour own phenomenalexperienceas humans and our ability to report that experienceby language. Notice the differencewhen we tum to computers.For ordinary computers,we have Iittle difficulty acceptingthe functionalistposition becausethe only meaning of the symbols on the tape and the statesin the processor is themeaningassigned to themby a humanprogrqmmer.There is no ambiguity in the interpretation of physical statesas symbols becausethe symbols are representeddigitally according to rules in a syntax. The system is designedto jump quickly between defined states and to avoid transition regions between them; electrically, each component always goes to a "zero" or a "one." The small deviations in physical parametersthat do occur (noiselevels,for example)are ignored by agreementand design.One purpose of all theseconventions is to assurethat any differencesbetween two systems that occur because of their different ways of physically representingsymbols should indeed have no meaning.Different hardware is not an issueas long as the hardware performs.Remember,though, that this portability of functionalist systems across different hardware implementations is bought at the price of requiring primitive functional processesto operate on symbolic representationsof information. Now we begin to seewhy digital computersare a falseanalogueto the brain. The facile analogy with digital computers breaksdown for several reasons.The tape readby a Turing machineis markedunambiguouslywith symbols chosenfrom a finite seb in contrast,the sensorysignalsavailable to nervous systemsare truly analoguein nature and therefore are neither unambiguousnor finite in number. Turing machineshave by definition a finite number of intemal states,while there are no apparentlimits on the number of statesthe human nervous system can assume(for example,by analog modulation of large numbers of synaptic strengths in neuronal connections).The transitions of Turing machinesbetween states are entirely deterministic, while those of humans give ample appearanceof indeterminacy.Human experienceis not basedon so simple an abstraction as a Turing machine;to get our "meanings" we have to grow and communicate in a society. The abstract beauty of Turing machines is beguiling. But one must watch out for excessiveabstractioneven in science,where it usually lends great power to much of our thinking. Abstraction in somecontexts is quite 225

MlNp WrrHour

Brorocv: A Cnlrrcer Posrscrlpr

foolish. The story is told of a race track that was failing financially. The track managementconsultedthree experts:an accountant,an engineer,and a physicist. The accountant recommended restructuring of the balance sheets;the engineersuggestedthat the slow track could be fixed by slight banking and better drainage.When his tum came,the physicist went up to the blackboard,drew a circle, and said, "Let us replace the horse with a sphere." In contrast to computers, the pattems of nervous system response dependon the individual history of eadr system,becauseit is only through intnactionswith the world that appropriate responsepattems are selected. This variation becauseof differencesin experienceoccursbetweendifferent nervous systemsand within a single system acrosstime. The existenceof extensiveindividual variation in cognitive systems(seechapter3) negates the fundamentalpostulateof functionalismthat representationshave meaning independentof their physical instantiation.Thus, it would appearthat the independenceof physical instantiation that is such a prized feafure of functionalist systemsmust be abandonedif a nontrivial level of cognitive performanceis to be achieved.(This doesnot mean that in abandoningthe liberal position of functionalism, we must adopt the extreme chauvinist position: that carbon chemistry, wet tissues, and so on, are absolutely necessaryfor cognition to occur. Were that so, the artifacts discussedin chapter 19 could not be constructed.) Whatever type of intemal representationsa functionalist system may employ, a procedureis neededfor establishingthe meaningsof the individual units (symbolsor their generalizations)and of combinationsof units in those representations.It is not easy to see how, in the absenceof a programmer,a mechanismcould be constructedthat would assignmeaning to syntacticrepresentationsand still preservethe arbitrary quality of those representations,a quality that is an essential part of the functionalist position. But that is our poignant position: we have no progr.lmmet no homunculus in the head. I could not close here without mentioning that, in recent years, a large amount of work has been done on "connectionist" or "neural network" models of perceptual or cognitive processes.These are formal models in which the connectionsbetweennetwork elementsare modified in a fashion loosely analogous to synapses.I suppose this justifies the metaphoric "neural," but in other respectsthe metaphor is strained,as I point out in what follows. Theseconstructionshave been usefulin a number of applications.Many of the models begin with assumptionsabout the nature of intelligent systemssimilar to those made by workers in artificial intelligence.Unlike 226

Mind Without Biology:A Critical Postscript classicalwork in artificial intelligence,however, these models use distributed processesin networks, and changesin connectionsoccur in part without strict programming.Nonetheless,connectionistsystemsneed a programmeror operatorto specifytheir inputs and their outputs,and they use algorithms to achievesuch specification.While the systemsallow for alterationsas a result of "experience:'themechanismof this "learning" is instructional,not selectional.Unlike selectionalsystemscarrying out cate(not the values)of connectionistsystems gorizationson value,the responsr"s are specifiedin advance and are imposed on the system by a human operatorunder appropriateconditionsand with appropriateerror feedback to establishthe training. The architecturesof neural networks are removed from biological reality, and the networks "function" in a manner quite unlike the nervous system."Neural nets" use symmetricaland densematrixlike connections. In general,they do not at all resemblethe neuronal stmctures and the anatomy of which I have written in this book. If neural networks were adopted as the standardmodel of brain structureand function, we would have to say that they support the view of the brain as a Turing machine. Whatever their interest and usefulness,neural networks are not adequate models or analoguesof brain structure.(For readersinterestedin pursuing theseissues,I have placeda referenceto two collectionsof paperson this topic in the SelectedReadingsat the back of the book.) Whether digital computersor connectionistmodels are used as a base, we are left with the same embarrassment. In consideringthe brain as a Turing machine,we must confront the unsettling observationsthat, for a brain, the proposedtable of statesand state transitions(seefigure P-3) is unknown, the symbols on the input tape are ambiguous and have no preassignedmeanings,and the transitionrules,whatever they may be, are not consistentlyapplied.Moreover,inputsand outputsarenot specifiedby a teacheror a programmerin real-worldanimals.It would appearthat little or nothing of value can be gained from the application of this failed analogy between the computer and the brain. But the field is not abandonedso easily.There remainsa large body of work in cognitive psychology based on similar confusions concerni.g what can be assumedabout how the brain works without bothering to study how it is physicallyput together.Let us tum now to some of the difficultiescreatedin cognitive psychology by one of its centralnotions: the idea of mental representations.

227

MrND WrrHour

BloLoGy: A CnrrrcAL PosTscRrpr

SOMEVICIOUSCIRCLES IN THE IANDSCAPE COGNITIVE The blend of psychology, computer science,linguistics, and philosophy known as cognitive sciencehas grown enormously.As with all vigorous efforts, ill-founded or not, much has emerged that is of great interest to scientistsand nonscientistsalike. Not the least of the positive results has been the routing of simplemindedbehaviorism.But at the sametime, an extraordinary misconceptionof the nature of thought, reasoning,meaning, and of their relationship to perception has developed that threatens to undermine the whole enterprise. To trace the nature of this misconceptiontakes some doing, for it has complex historical, intellecfual,and practicalroots. I must wam the reader that this field delves into some complicatedmatters,and I cannot simplify their description altogether. Before I tackle the details,let me give you a short characterizationof the misconception.It stemsfrom the notion that objects in the world come in fixed categories,that things have essential descriptions,that conceptsand languagerely on rules that acquiremeaning by formal assignmentto fixed world categories,and that the mind operates through what are called mental representations.Theserepresentationsare supposed by some to be expressedthrough a language of thought, or "mentalese,"asthe philosopherferry Fodor callsit. Meaning consistsof the assignment of symbols in such a language to correspond eractly with entities or categoriesin the world defined by singly necessaryand lointly sufficientcriteria (classicalcategories).Thus, a specificationof the rules by which representationsare manipulated(constituting a syntax),if complete, can be carried out by a computational device. The brain in this view is a kind of computer.(Note the similarity of some of theseassertionsto those in the last section.) The acceptanceof this view or versions of it is widespreadin psychology, linguistics,computer science,and artificial intelligence.It is one of the most remarkablemisunderstandingsin the history of science.Indeed,not only is it not in accord with the known facts of human biology and brain science,but it constitutes a major category error as well. We have fooled ourselvesin part as a result of our successin removing the mind from nature in the "hard" sciences.The error hasbeen to attribute the characteristicsof humanmental constructions(suchaslogic and mathematics)to human reasoningand to the macroscopicworld in which we live. Whenever I think of this carving of vicious circlesof rational design onto 228

M i n d W i t h o u t B i o l o g y' A C r i t i c a l P o s t s c r i p t the surfaceof the cognitive landscdpe, I cannot help but think of the conversationbetween two mice in a psychology laboratory. After a successfulrun in the maze,one mouse says to the other, "You know, I think I have my psychologisttrained.Every time I run the mazesuccessfully, he gives me a piece of cheese." To show you why the ideasof "mentalese i' of rulesand representations, and of computation will not work, I must take up some of the stricter assumptionsof the functionalismunderlying cognitive psychology. Then I must considera view of the world (and particularly the scientificworld) called objectivism. Finally, I must look at a central issue:the evidence concerninghow we actually categorize the world, both perceptuallyand conceptually.That done,we will be able to seethe errorsin reasoningthat threaten to underminethe cognitive enterprise.The argumentsand data consideredhere will not be exhaustive;the readeris encouragedto consult the appropriateworks in the SelectedReadingsfor more information.I will attempt to sketchthe issuesin a minimal but incisivemanner.They are at the heart of any effort to understandthe matter of the mind. It appearsthat the majority of those working in cognitive psychology hold to the views I attackhere.But there is a minority who hold contrary views, in many ways similar to mine. These thinkers come from many fields' cognitive psychology, linguistics, philosophy, and neuroscience. They include Iohn Searle,Hilary Putnam, Ruth Garret Millikan, George Lakoff,Ronald Langacker,Alan Gauld,Benny Shanon,Claesvon Hofsten, JeromeBruner, and no doubt others as well. I like to think of them as belonging to a RealistsClub, a dispersedgroup whose thoughts largely convergeand whose hope it is that somedaythe more vocal practitioners of cognitive psychology and the frequently smug empiricistsof neurosciencewill understandthat they have unknowingly subjectedthemselves to an intellectualswindle. The views of this minority will be reflectedin what I have to sa1l,but obviously they vary from person to person.The readeris urged to consult these scholars'works directly for a closer look at the diversity of their thoughts and inteqpretations.

FunctionalistViewsand the Semantic Representation of Meaning The centralidea underlying much of modem cognitive psychology is that of mental representations. Theserepresentationsare abstractand symbolic 229

MrNp Wlruour

Brorocv: A Cnrrrc.nr Posrscnrpr

(that is, they stand for some thing or some relation), are formed in a well-defined manner, and follow rules that constitute a syntax. They are supposedto be semanticallyrelated to the world by fixed and determinate relationships and by the semantic assignmentof symbols to objects in classicalcategories.They are essentialto the formation of "inner models of the world." The idea of inner modelsfollows the early suggestionsof K. I. W. Craik and in this view, intemal representationsparallel extemal structuresin the world. The representationsare propositional,involving conceptsand their relationships,or they are mental images.The origin of images is perception, which in this view is a form of computation,accordingto the seminal and influential ideas of the late David Man. The computationsthat occur on mental structuresare govemed by a system of rules (or a syntax) and by the representationsthemselves.The whole system of representation forms a lingua mmtis or mentalese,a language of thought. How does this point of view deal with the problem of intentionality? Presumablyby declaring that meaning arisesfrom the mapping of rulegooernedslmtacticalstructures onto definedandfircd world objects or relations. Such a semanticsis exhaustiveand determinateand, together with its underlying syntax, it provides a framework for modeling the mind. How did this functionalist, computational view of the mind, which is highly formal and disembodied,arise?How could anyone accept so abstract a notion of human knowledge, reason,and mental activity? Before I criticize this view of the mind, let us considerthe correspondingview of the world that forms one of its foundations.

Objectivism The term "objectivism" has been used to characteri ze a view of the world that appearsat first sight to be both scientificallyand commonsensically unexceptionable.(One analysis I will follow is that of Lakoff, see the SelectedReadings.)Objectivismgoes beyond the hypothesisof scientific realism,which itself assumes:(1) a real world (includi.g humansbut not dependingon them); (2) a linkage between conceptsand that world; and (3) a stable knowledge that is gained through that link. Objectivism assumes,in addition to scientificrealism,that the world has a definite structure made of entities,properties,and their interrelationships(figure P-5). Theseare capableof definition accordingto classicalcriteriaof categoriza230

Mind Without Biology,A Critical Postscript

OBJECTIVISM

Obiects

DescriptiveTerms

(and Evcntr)

(Wordr and Sclcntlflc Thcorlcr)

Syntax

Rulesbuiltin ( takeverband place afternounin phrase) "The man ls slttlng In a chalr" TakemassM andmultiply by acceleration to get force

t b Y

fI II

Semantlcs Slngly nessssaryand folntly sufflclent condltlons for classlcal categorlzatlon

I -

l.E'..' Il=*P-l I I

I

tt It

t t

I

ut |

I

1. Aboutthe sizeof a human

I

tl |

2. An objectwith a platform andlegs

l-l

3. Allowsa person(see 1) condition to sit etc...

Hardware

Software MACHINE FUNCTIONALISM

FH,@ /",,R'ryl 1.,..::::" 1 qFry l=;__4+ )@)

I-

Output

q a

"lnformation"

Brain

Q+ \

FIGUREP-5 Someaspectsof objectiaismand funcfionalism, 23-1,

Output

MrNp Wrrnour

Brorocv: A Cnlrrcer Posrscnlrr

tion that are singly necessaryand;ointly sufficientto defineeachcategory. The world is arrangedin sucha fashion that it can be completely modeled by what mathematiciansand logicians would call set-theoreticalmodels. These kinds of models, which are seen in mathematicallogic, consist of symbolic entities appearingsingly or in sets,together with their relationships.Symbolsin thesemodelsare mademeaningful(or are given semantic significance)in a unique fashion by assuming that they correspond to entities and categoriesin the world. Some of the categoricalproperties of things in the world are considered to be essential;others are seen as accidental. Becauseof the singular and well-defined correspondencebetween settheoretical symbols and things as defined by classicalcategorization,one can,in this view, assumethat logical relationsbetween things in the world exist objectioely.Thus, this system of symbols is supposed to represent reality, and mental representationsmust either be true or false insofar as they mirror reality correctly or incorrectly. According to objectivism, this correspondenceto things in the world gives meaning to linguistic expressions;meaning is basedon this "correct" or "incorrect" definition of truth and thought itself is a manipulation of symbols. This view cancertainly be held outside of science.Indeed,the objectivist position seemsin accord with much that is commonsensical.But when it is held inside science,it comesclose to the Galileanposition we discussed in chapter 2. In that sense,human concepts,assertions,and languagesare valid only if limited to physics, chemistry, and parts of biology. We will seethat, however sensibleit seemsat first, this view is woefully incoherent and not in accord with the facts. How then did it arise?Well, one can go a long way with this view in the hard sciences.Removal of the mind from nature is a sensibleprecaution for much of classicalchemistry and physics. And many of the major developments in physics have depended strongly on the use of the rigorous formal reasoning at the core of mathematicsand logic. In the late nineteenthand early twentieth centuries,deep investigations into mathematicallogic by Gottlob Frege, Giuseppe Peano, and Alfred North Whitehead and BertrandRussell,followed by the work of Stephen Kleene, Emil Post, Alonzo Church, Alan Turing, and Kurt Godel, were triumphs of human analysis of the "mechanics" of reasoning by logic. When I was in college, I was enrapfuredby the eleganceof all of this. I spent long evenings with those great dark blue compendia of logical hieroglyphics, the PrincipiaMathematicaby Whitehead and Russell.Their very dryness convinced me that I was on the inside track. It was too bad that I had no one to tell me about the human side of these authors at the time. I have sinceheard that during the writing of thesetomes the usually 232

Mind Without Biology:A Critical Postscript mild-manneredWhitehead once said to his feistier colleaguer"Bertie, the world is divided into the simplemindedand the muddleheaded,and I shall leaveit to you to decidewhich one you are." The mathematicianG. C. Rota hasrecently mounted a scathingattackon the excessiverelianceon formalism and axiomaticsof somephilosopherswho ape the clarity of mathematics by adopting a symbolic mode of discussion(seethe SelectedReadings, '1.4, chapter for a referenceto his work). The subsequentdevelopment of the computer, which was partly based on theseinvestigations,reinforcedthe ideasof efficiencyand rigor and the deductive flavor that had already characterizedmuch of physical science. The "neat" deductive formal background of computers, the link with mathematicalphysics,and the successof the hard scienceslooked endlessly extensible.There was a natural tendency to stop a philosophical analysis of scientificexploration at the surfaceof the human body (the skin and its receptors).Behavior could be analyzed,but not phenomenalexperience.In this way, sciencecould remain "extensional," as W. V. euine put it, and one could declarewith him that "to be is to be a value of a variable." The computational or representationalistview is a God's-eye view of nafure. It is imposing and it appearsto permit a lovely-rooking map between the mind and nature. such a map is only lovely, however, as long as one looks away from the issue of how the mind acfually reveals itse* in human beings with bodies.when applied to the mind rn srfu, this view becomesuntenable.

The difficultiesof the computational or mentalrepresentational view of mindaremanifoldandcanbeusefullygroupedinto eighttypes(tablep-1). Thegroupingis not just a matterof convenience but alsopiovidesa battle plan for an assaultagainstthis view of mind. I recommendthat the interestedreaderconsultthe selectedReadingsto obtaina deeperunderstandingof the works of the authorsmentionedin table p-i. on the assumptionthat the readerwill do so,I treat the issueslistedin the table only brieflyhere.My goal is to sketchthe majorcriticalarguments against functionalismand objectivism,not to exhaustthem.

categories:A crisis for Functionalist Viewsof Cognition One of the largestchallengesto the functionalistview of mental representations comes from philosophical and psychological work on how we categonzethings.Most of this work is concernedwith conceptualcategori233

MrND WrrHour

BroLoGy: A

CnITIcAL PosrscRlpr

TABLEP-1 SomeProblemsutith the Idea of Mental Representation* 1. Perceptionand reason are not govemed by classicalcategories.Biology (particularlyDarwin's work) shows essentialismto be false (Rosch,Wittgenstein, Lakoff, Mayr). Similarity is not the sameas categorization. Z. Thought is not transcendentbut dependson the body and brain.It is embodied, Meaning arisesfrom relationsto bodily needsand functions.The mind does not mirror nature (Putnam,Millikan, Langacker,Lakoff,]ohnson,Searle,Edelman). 3. Memory cannot be describedby intemal codes or syntactic systems.Moreto account for its full over, one needsa self and higher-orderconsciousness linguistic manifestation(Searle,Shanon,Gauld, Edelman). 4. Languageis acquiredby interacting with others in learning events, which initiate thu formation of connectionsbetween semanticsand phonology. It dependson having conceptualsystemsand values already in place (Pinker, Johnson,Edelman). 5 . Minds createtheir own versionsof reality by socialand linguistic interactions, and reality, like biology itself, dependson historical events (Searle,Putnam). 6. Computation is not only disembodied;it cannot by itself provide a meaningful relation between symbols and world entities (Searle). 7. Cognition gets its content from the identification of proper functions in a ,yJ"rn thaidepends on evolutionary history. Eachpart of a ProPerfunction has a "normal" explanation,which tells how that systemhasmanagedhistorically to perform that function. "Meaning rationalism," the assignmentof meaning from above, is untenable(Millikan). its 8. The structure,function, and diversity of the nervous system, ?s well as view functionalist the evolution and development, are incompatible with (Edelman).

Rndings *Thenames in thelist.ofSelectd worlaappear ofauthwswhose arethose in parmtheses arguments, e*ended the readings these consult Please book. the t'or iitita at thi endof

zation in humans,although some of it also concemsperceptualcategorization in both humans and animals. The single most striking conclusion arising from a variety of analyses and studies is that people do not categorizethings or eventsin terms of classicalcategories.Classicalcategories are those in which membershipis defined in terms of singly necessary and jointly sufficient conditions (figure P-5). Wittgenstein had some of the earliestcritical thoughts on this subject. He refle-ctedon family resemblance,noting that category memberscan be related to each other even if some members do not have any of the properties that classically define the category in common (figure P-6' ighO. (lmagine that there are r propertiesdistributed among the members 234

M i n d w i t h o u t B i o l o g y' A C r i t i c a l p o s t s c r i p t

oo%

8

?'

A'A, ,A.

Y %oo A.A, A. A' A

AA

oo

AA.A

o ,,CHAIR"

N

categorization andpolymo,rp^* *r!]f![!rioi" are.notnecessariry characterized by singlynecessary andjointly sffiient uiteia (classicat *tti*iiil,liwt,'A pay*oiphousrulefor setmembership,wly gtayica|categoizntionlottixffi, t"ti^ii* or the,.e\eyfka,ft markedY for "yn") hoJ,ory twoof thepripritiu ,orndnur, .(gro.up (groupmarked yt!,lltlr: * bl?tu!wymetry, Nonmembers N for ,,n0,,) haoeonlyone ol theseyoperties.Thefgure is from theetperiments of lan Dennisand his ,owoikrrs. of the set and that m properties suffice to allow membership,where n is larger than m.lf m of the n properties permits membership,i*o membe* need not have any of the sameproperties in cornmon.This partly deftnes what is called a polymolphous set.) wittgenstein also consideredseveral other intriguing ideas-that certain categoriesmay have degrees of membershipbut no clearboundariesand that olhers may have -"*b"r, that are more central or prototypical than others. since wittgenstein's time, psychologistshave done a number of sfudies establishingevidenceto support his ideas.Notable among them are Brent Berlin and Paul Kay, who showed that human color categorieshave degrees of membership and centrality; Roger Brown, who showed that children first name things at a level that is neither the most generalnor the most specific;and EleanorRosch and her coworkers, *hori studies were perhapsthemost generaland who developedthe analysisof categorization as a wide-ranging researchtool. 235

MrND WlrHour

BroLoGy: A CnrrrcAL PosrscRlpr

centrality,and Rosch'swork showsthe existenceof family resemblance, prototypicality. Categorieslike "red" have fuzzy boundariesbut nonethelesscontain centralmemberswhose degreeof membershipon a scalefrom zeroto one would be scoredas one. Theseare gradedcategories.Categories like "bird" have sharp boundaries,but within theseboundariessome birds are judged to be better examples than others-to be more "prototypical." Knowledgeof categorymembersis often organizedaround a basic level-a level which, in subjectstested by Rosch, reveals itself in terms of the easeof imagining and rememberingmembership,actions, and use."Horse" would be a basic-levelcategory and "quadruped"would not be. it is not suqprisingthat there is often If we acceptfamily resemblance, no definitehierarchicalrelationshipbetweensuperordinateand subordinate categories.Consistentwith this is the fact that categoriesare heterogeneous in origin: the actual properties humansuse to determinecategory membershipare interactional and they depend on different biological, cultural, and environmentalvariables. This empiricalwork was done on human subjects.Although aspectsof it have beenchallengedfrom time to time, it hasbeengenerallyconfirmed. More recently, Lance Rips showed that neither similarity nor typicality of categorymembershipand that the reasoning fully accountfor the degree involved in placing membershipis often nondeductive.LawrenceBarsalou additionally showedthat particularcategoriesare not even representedby invariant concepts.There is great variability in the conceptsrepresenting a category; different individuals do not representa category in the same way, and the sameindividual changeshis or her views of category membership in different contexts. Consistent with these ideas, the seminal studiesof Daniel Kahnemanand Amos Tversky showed that human decision making and humancategoryjudgmentsoften violate probability rules suchas the conjunctionrule, which statesthat a conjunctionis never more probable than either of its constituents.In certain contexts some people actually do not doubt that the conjunction is more probable. I have been concemedhere with conceptualcategories;perceptualcategories were discussedin the body of this book. But what we have at hand is enough to state that if this work is correct,the objectivist model of the mind-world relationshipis in hot water. For example,if categoriesthat have centrality and prototypicality, such as those for color, exist in addiWorse tion to classicalcategories,then the objectivistview is inadequate. than that, the objectivist model cannot deal with the fact that certain in theworld.Psychologicalwork indicates,for symbolsdo not matchcategories views of the mind cannot deal with categoriesof example,that computer the mind and of language(seeany poem) that fail to reflect categoriesin 236

Mind Without Biology: A Critical postscript the world. Individuals understandevents and categoriesin more than one way and sometimesthe ways are inconsistent.As Mark Johnson points out, metaphor and metonymy are major modes of thought. Metaphor is the referralof the propertiesof one thing to those of another in a different domain. Metonymy allows a part or an aspectof a thing to stand for the whole thing. Both are incompatible with the objectivist view. All of this spellstrouble for mental representation.In order to function, mentaleserequires an accurateunambiguous link to the extemal world. often meaning cannot be so establishedand such a link cannot exist. objects in the world arenot labeledwith dimensionsor codes,and the way they are partitioned differs from person to person and from time to time. Indeed,the fixed semanticsof mental representationcannot accountfor the occurrenceof novelty in the world and, as will be apparentin my discussion of language,well-defined codes cannot exhaust the meaning of linguistic expressions.Meaning simply refusesto be bound by a fixed set of terms in a specificcoding system.while representationsmust remain fixed, behavior changesin new contexts (unaccountably,in the objectivist view). If this holds, the mind is not a mirror of nature. Thought is not the manipulation of abstract symbols whose semanticsare justified by unambiguous referenceto things in the world. Classicalcategoriesdo not serye in most casesof conceptualcategorization and they do not satisfactorily accountfor the actualassignmentof categoriesby human beings.There is no unambiguousmapping between the world and our categorizationof it. Objectivism fails.

MemoryandLanguage Another source of embarrassmentfor the computational or functionalist view of mind has to do with memory and its connection to the self and to language.I considersomespecialaspectsof languagein the next section, but for now note that the words of a naturallanguageare not like the terms of a computer language.I pointed out in the last section that all computation is syntacticalin nature and thus, unlike the use of words in a speech community, it cannot have meaning without a programmer. Moreover, functionalists often speak of propositional attitudes-beliefs, desire+ wishes. But as Putnam has pointed out, beliefs and desires cannot be individuated without referenceto an open-endedenvironment,one that is not characterizedbeforehand. There is an additional problem: Human memory is not at all like computer memory. As we have already noted, intemal codes and syntactic 237

MrND WrrHour BroLoGy: A CnrrrcAL PosrscRrpr systemscannot adequatelydescribehuman memory. Memory has been variously and sometimesconfusinglydescribedasepisodic(relatingto past events in a life), semantic(relating to language),procedural (relating to motor acts),declarative(referringto statements),and so on. Memory is a systemproperty:lt differs dependingon the structureof the systemin which it is expressed.In biological systems,memory must not be confusedwith the mechanismsthat are necessaryfor its establishment,such as synaptic change.Above all, biological memory is not a replica or a trace that is coded to representits object. In whatever form, human memory involves an apparentlyopen-ended set of connectionsbetween subjectsand a rich texture of previous knowledge that cannotbe adequatelyrepresentedby the impoverishedlanguage of computer science-"stora ge," "retrieval," "input," "output." To have memory, one must be able to repeat a performance,to assert,to relate mattersand categoriesto one'sown position in time and space.To do this, one must have a self, and a consciousself at that. Otherwise, one must postulate a little man to carry out retrieval (in computers,it is w€, the programmers,who are the little men). How is the proposed functionalist model of an algorithmicmind to be accessed without an infinite regressof homunculi,one inside the other? With the homunculus,we come to one of the great problemsin considering the matter of the mind: the problem of accountingfor intentionality itself. We have already shown that formal semanticscannot refer unambiguously to real statesof affairs.But many of the causalaspectsof our mental statesdependon semanticcontents.As Searlehas stressed,semantic contents are meaninglesswithout intentionality or the ability to refer to other statesor objects.To carry out referral,a formal representation must becomean intentionalone.In humanbeings,this requiresa consciousnessand a self-a biologically basedpersonalawareness, a first person.No theory of mind worth its salt can evade this issue,which is not only a matter of languagebut also a great biological problem. Let us pursueour quarry relentlessly,tuming finally to some biological matters that cannot be reconciledwith the functionalistpicture of the mind.

The Lessonsfrom Biology A great revolution in thought camefrom Darwin's efforts to understand the origin of species.In his theory of natural selection,he gave the world 238

Mind Without Biology: A Critical Postscript its first exampleof population thinking. Population thinking, as Emst Mayr puts it, considersvariance to be real, that is, not an error (see hgure S-2). Natural selection acts on variation acrossindividuals in a population. As Mayr has shown, speciesoften ariseas a result of the occurrenceof sexual and geographicbarriersto the propagation of variantsor even by accident. The speciesconceptarisingfrom this part of population thinking is central to all ideasof categorization.Speciesarenot "natural kinds"; their definition is relative, they are not homogeneous, they have no prior necessary condition for their establishment,and they have no clearboundaries. Thus, population thinking dealsa death blow to typological thinking or essentialism,the notion that "essences"of speciesexist before particular organismsor exemplarsdo. Essentialism,most clearly formulated by plato and reflectedin most idealisticphilosophiesever since,has a deep kinship to the notion of classicalcategories.But biology shows us that, even though taxonomy is possible for living creatures,essentialismis false. Given my previous remarks,it is also likely to be false in thinking about the mind. Searle,Lakoff, Johnson,and others including myself have pointed out that thought is not transcendentbut dependscritically on the body and the brain. This position is exactly opposite to that of functionalism, which assumesthat the realizationof software is independenf of hardware.According to those who reject functionalism,the mind is embodied.lt is necessarily the casethat certaindictatesof the body must be followed by the mind. Gestalt perception is such a dictate; the categoriesof a gestalt (see,for example,hgure 4-2) are not validated by a unique pattem in theworld, and yet they are often inconigible. Gestalts,mental images,bodily movements, and the organization of knowledge must all to some degree be the result of evolutionary and developmentalconstraints. The syntax and semanticsof natural languagesare not just specialcases of formal syntax and semantics,the models of which have strucfurebut no meaning.In the biological view, symbols do not get assignedmeaningsby formal means;insteadit is assumedthat symbolic structuresare meaningful to begin with. This is so becausecategories are determined by bodily structure and by adaptive use as a result of evolution and behavior. The symbols of cognition must match the conceptual apparatuscontained in real brains. The basesfor truth and knowledge come from this apparafus and have their earliestfoundationsin evolutionarily derived value systems. According to the purveyors of this view, including Lakoff, Johnson,Modell, and myself, when symbols fail to match the world directly, human beings use metaphor and metonymy to make connections,in addition to imagery and the perception of body schemes.Minds create aspectsof 239

MrNp Wtrsour

BtotocY: A CnIrlc.lr Posrscnlpr

reality through cultural and linguistic interaction. Like biology itself, this interaction dependson historical events.I deal directly with thesematters in the next section,when I discusslanguage and its acquisition. Besidesembodiment,there is one more key issue-that of function. A deep insight into function was henchantly expressedby Millikan in connection with minds, languages,and "other biological objects," as she calls them. Biological objects under evolution have functional properties that differ from, say, those of molecules.One does not speakof the "abnormal" function of a molecule as a chemicalobject. But a biological obiect has a proper function that dependson its evolutionary history. A heart has a proper function to pump blood. There is alsowhat Millikan callsa'hormal" explanation for the production of such an item in a species,and this accounts for the resemblanceof this organ to "normal" hearts in that species.Hearts work well or not; badly functioning ones are abnormal.In contrast,organic chemicalsdo whatever they do, and whateoerthey do is part of their "working." During evolution, functions that account for the proliferation of survivors are proper functions and associatedwith them is a "notmal" explanation accounting for how they have hisloically managedto perform that function. The funny thing is that states and activities can have ProPer functions without performing them, and they can even have proper functions without contributing to further proper functions in accord with a "normal" explanation. This is becausehistorical phenomenain selective systemscan lead either to failure or to unexpectedsuccess. Millikan regardspsychology as a branchof biology, I think properly so. she claims that cognition gets its content from the identification of proper functions.This is an important claim. Eachset of functions has a "normal" explanationthat relateshow the systemmanagesto perform that function. Millikan's view of cognition allows it to be placed in the context of physiology (for example,that of the value systemsI discussedin this book) and still provide grounds for a theory of beliefs and desires.unlike the propositional attitudes of the functionalist,such a theory of intentionality does not differ dramatically from the usesand referencesof ordinary folk psychology (the way we usually characterizemental function in everyday iif"). tn Millikan's view, the brain is thought of asa symbol manipulatorand 'hormally" as a semanticengine. This is becausebeliefs and desiresare manipulated in terms of significant (that is, bodily significant) differences betwlen them and also in terms of differencesin their proper functions.The appraisal of meaning and truth comes from this path, according to her that are anilysis, not from the assignmentof semanticsby correspondences antithetical with those calls made by the "meaning rationalists," as she views. 240

Mind Without Biology: A Critical Postscript The upshot of the argumentsI have describedis that the factsof biology force us to conclude that the mind is not transcendental.There is no God's-eye view of the world. Essentialismis not tenable and neither are functionalism, objectivism, or the form of "computational realism" that considersthe mind to be a machine.Moreover, there is another fundamental sourceof difficulty that I describedin the early chaptersof this book and elsewhere The variation in the structures and functions of the neryous system and the way in which the brain develops its anatomicalconnectivity by depending on correlations with events in the world are both incompatible with the functionalist view. The vicious circlescarved in the cognitive landscaPeare broken by the evidenceunderlying the foregoing analysis.But it is not enough to say that the mind is embodied to account for meaning and memory. The question is, How? And how, after explaining how, does this explanationaccountfor This was the task I the development of the self and of consciousness? it, I had to consider To accomplish the book. the body of undertook in In addition, however, language in relation to higher-order consciousness. there are sometechnicalmatters specificallyconceminglanguagethat must be placed in the context of the argumentspursued in this Postscript.Let us fum to them.

IANGUAGE:WHYTHEFORMAL FAIIS APPROACH First, I want to considerhow formal views of languageare inconsistent with what I have already said about categories.Then I want to touch on some proposalsfor cognitive models and for grammarsthat are more in accordwith what is known about categorization.My purposeis to contrast these two views, the formal and the cognitive, in an attempt to give the readera glimpse of how different their premisesare. The study of languageis vastly challengingand the field of linguistics proper is intricate in the extreme. I do not attempt to penetrate these studieshere, for their full extent is beyond my expertise.Fortunately for our purposes,we need to know only a few guiding facts.I will describe them briefly and move on to my main point: that formal approachesto grammar fall under the same ax as felled the obiectivist and the strict functionalistapproachesto psychology. To know a language is to be able to produce sounds or gestures conveying meaning and to understandthem as they are produced by 241

MIND WlrHour

BroLoGy: A CnrrrcAL PosrscRlpr

others. In general,the relation between form and meaning in a language is arbitrary.A striking featureof languageis its creativity,a personcompetent in a languagecan make and understandcompletely new phrasesand sentences.It is also striking that, most of the time, a person can tell the differencebetween proper and improper grammaticallocutions. Linguists attempt to constmct theoriesof a speaker'sgrammar.In this sense,being grammaticalmeansto conform to descriptiverules derived from use. In the broadestsense,however, grammarincludesthe study of the laws governing phonology (the sound system),moqphology (in this context, word formation),and semantics(the systemof meaning).All the laws togetherare saidto makeup a "universalgrammar:'a phrasethat has been adopted from Noam Chomsky.According to his seminalproposals, all languageshave a set of grammaticalpropertiesin common that constitute this universalgrammar. Chomsky has also suggestedthat inasmuchas languageis unique to humans,and children'sactuallinguisticperformanceis underdeterminedby their testablecompetence,there must be a "languageacquisitiondevice" that is innate to humans.It is important to note here that speakingis an acquiredskill that developsfrom belonging to a speechcommunity. A lot of categorizingmust be done in order to speak.One must develop conceptsor intentions,formulateexpressionsaccordingto grammarand phonology, and articulate,comprehend,and monitor one'sspeechproductions in exchangeswith others. To do this as an interlocutor requiresa cooperativeprinciple,as describedby H. P. Grice. One must be informative at just the level required and flo rnor€; one must be brief, orderly, and unambiguous.One must be receptiveto cuesfor turn taking.In addition,one must definethe "here and now" or the "there and then." This so-calleddeixis locatesboth interlocutors and objects in space.One's communicativeintentionmust also be expressedin appropriatespeechacts.Speakingin generalis a tactful aswell as a tacticalact. Language acquisitionin a speechcommunity differs for children and adults.Moreover, to acquirea languageand to use it are not necessarily the samething. The study of linguistic knowledge, or psycholinguistics, and the study of the biological and neuralbasesof language,or neurolinguistics,must both be brought into play here. We confronted the differencesbetween languageacquisitionand use in chapter12. Recallingthem here will keep us from confusingacquisitionwith rehearsedpractice. All of this is by way of introducing the problem of how thought and language are connected.A clear picture must be drawn of the relation between concept systemsand language.Does the mastery of language 242

Mind Without Biology: A Critical Postscript depend on the existenceof a ricl'r and embodied concept system?Or is language mastery more or less autonomous, developing by means of a language acquisition device? One of the most pervasive and influential approachesto these critical questionswas pioneeredby Chomsky. In his formal systemsapproach,the principal assumptionis that the rules of syntax are independentof semantics. Language,in this view, is independentof the rest of cognition. I must take issuewith this notion. The set of rules formulated under the idea that a grammar is a formal system are essentially algorithmic. In such a system, no use is made of meaning. Chomsky's so-calledgenerative grarnmar (figure P-7) assumes that syntax is independent of semanticsand that the language faculty is independentof extemal cognitive capabilities.This definition of grammar is impervious to any attempt to disconftrm it by refening to facts about cognition in general.A languagedefinedasa set of strings of uninterpreted symbols generated by production rules is like a computer language.To give the symbols semanticmeaning, they must be mapped onto the real world or onto a language of thought or mentalese. The previous discussionhaspreparedus for the conclusionthat underlying this view is the objectivist position: categoriesare classicaland semantics are generatedby unambiguousassignmentto entities in the world. This amounts to a definition of languageand grammar.Under this definitioo languagefalls afoul of all the difficultiesfacedby the objectivist view. The problem is not just that this view does not agree with the empirical facts conceming categorization. It also ignores the fact that language serves to convey the thoughts and feelings of individuals who already think independently of language. The languageacquisitiondevice was proposed by Chomsky to answer the question of how a child who does not appear to understand many simple things can master the complexities of language.But a number of observationsseem inconsistentwith the Chomskian view. They concem thought and languageacquisitionin children as described,for example,in Margaret Donaldson'sbook Childrm's Minds. Donaldson points out that Chomsky directed the attention of his field toward studiesof how a child acquiresa knowledge of grammar. Consequently,linguists collected and interpreted data on what a child said in terms of a set of rules by which the child's utterancescould have been generated.But in this enterprise,a good deal was ignored-often including what the child actually meant and what he or she understood. As recounted by Donaldson, Iohn Macnamarahas proposed that chilthey ftrst make senseof situations dren are able to leam languagebecause 243

MrND WrrHour

BroLoGY:A CnITIcAL PosrscRIPr

GRAMMAR GENERATIVE PHRASEMARKERS: Symbolizethe analysisof a sentence,"the girl was nice"

s.

I l

NSiE ENS

NP /\

DET

N

I

l

P,AST

girl

the

V

/\

be

AP I A

nice

SymbolsUsedin PhraseMarkers S NP DET N TENSE VP V AP A

= = = = = = = = =

Sentence Noun Phrase Determiner Noun Tense Marker Verb Phrase Verb Adiectival Phrase Adiective

FIGUREP_7 A typical tree in generatioegrummar, uthich is used to deoelopand analyzesyntar, in Accordingto Clwmsky, the rulesof a uniaersalgrammar are assuredby the presence humansof an inborn languageacquisitiondeoicethat operateson sucha syntar or on one of its moderueremplars,The relation to semanticsis assuredby the obiectirtist assumption (seefigure P-5), This grammatical analysis has been supersededby Chomsky'smore recentgoaernmentbinding theory, but the underlyingassumptions remainunchanged.

244

Mind Without Biology:A Critical Postscript involving human interactions. Children make sense of things first and, above all, they make sense of what people do. Donaldson's summary makesit clearthat children can seethings from another'spoint of view, not just their own. They reasondeductively and carry out inferenceat age four or so, much more skillfully than had been previously supposed.It also seemsthat a child first makes senseof situations and of human intentions and then of what is said. This meansthat language is not independent of the rest of cognition. Therefore we need to account for languageacquisition not only developmentally but also evolutionarily. I discussedthis '1.2, where I consideredhow both conceptual problem at length in chapter are embodied. linguistic systems and Before I tum to altemative ways of looking at language, I want to mention a prescientaccountby the novelist Walker Percy,whose collected essays on language were published in a book called The Messagein the Bottle. My impression is that the attempt to understand language and meaningwas at the center of his life and of his work. Percy was aware that generativeor transformationalgrammar did not explain languageand that it was merely a formal description of competence:No relationship is necessarybetween this collection of algorithms and what goes on in a person'shead.He also understood that individual awarenessis symbolicas well as intentional. Higher-order consciousness,as I have called it (see '12), is a "knowing with" (con-sciousness).Percy faulted both chapter behavioristic and semiotic approachesto languagethat do not pay attencharacterof any linguistic act. He also faulted the tion to the intersubjectiae philosophy of phenomenologyfor "leaving out the other guy." He insisted that all symbolic exchangesinvolving meaningshow a tetradic relationship between symbol, obiect, and at least two humans.In a denseand resonant sentence,Percy put it thus: "The act of consciousnessis the intending of the object asbeing what it is for both of us under the auspicesof a symbol." He describesHelen Keller's rapture when she leamed that water was "water" and her urgent desire to know then what other things "were." Language,as Percy put il createsa world, not just an environment. That world is loaded with intentionality, with projections,with feelings, with prejudice,and with affection.The story is told of two Jewishtourists who visited Israelfor the first time. After an exhaustingbut enjoyableday in Tel Aviv, they decidedto go to a nightclub. A comedianwas on stage telling oneJinersin Hebrew. After a few of thesejokes, one of the tourists fell off his chair,laughing uncontrollably. His companionlooked down and 'lVhat asked, are you laughing at? You don't even understandHebrew." The man on the floor clutched his sides and said, "l trust these people." Formal semanticscannot accountfor such richness.Well then. what can 245

MrND WrrHour

BroLoGy: A CnrrrcAL PosrscRlpr

we do?One approachis to constructwhat is calleda "cognitive granunar"l an approachthat beginswith the factsof cognition rather than with formal analysis.An early pioneerwas Ronald Langacker,whose book Foundations Grammarmay be consultedfor a history of this work and the of Cognitirse principlesguiding it. As in all nascentsubjects,terminologiesdiffer. Rather which is in a sensethe "coining" terminology,l will, than use Langacker's, for the sake of convenience,discussand follow the proposalsof Lakoff, which are closelyrelatedto Langacker'sand which provide examplescloser to my own work in brain theory. Let us consider as an example this linguist'sattempt to provide a model of cognition adequateto the available facts of categorizationand to build a semanticsbased on the idea that meaningis embodied.

CognitiveModelsandCognitiveSemantics: Returningto Biology Lakoff has approachedthe subjectof grammar and semanticsin a way that appearsto be more in accord with the biological and psychological facts than are generativegrammars.He startsfrom actualdata on categorization and proposesthat meaningresultsfrom the intrinsic workings of the body and the brain. He suggests that individual humans construct cognitive models that reflect conceptsconcemedwith the interactionsbetween the body-brain and the environment. It is this conceptual embodiment, he claims, that leads to the formulation of basiclevel categoriesof the kind describedby Rosch. Cognitive models areueatedby humanbeings,and in this sensethey are idealized-that is, they are abstractions.But they dependon the formation of images as a result of sensory experience and they also depend on kinestheticexperienc-the relation of the body to space.Lakoff suggests that the exerciseof thesefunctions leadsto various image and kinesthetic schemas.Schemashave properties that are reflected later in the use of metaphor and metonyms.Recallthat a metaphor is the referralor mapping of one thing to another in a different domain, while a metonym is the use of some part or aspect of a thing to stand for the thing itself. Lakoffs example of a metaphor is "Anger is a dangerousanimal"' His example of a metonym is "The ham sandwich left without paying'" The important thing to grasp is that idealizedcognitive models involve conceph,ralembodiment and that conceptual embodiment occurs through 246

Mind Without Biology: A Critical Postscript bodily activities pior to language.Conceptualembodiment is used in categorization and it allows for the heterogeneity and complexity of real human categorization.Categoriesof mind correspondaccordingly to elements in cognitive models. Some of these models have different degrees of membership.Others include classicalcategoriesand are formed according to singly necessaryand jointly sufficientconditions (note that there is no contradiction here,provided that not all the models are classical!).Some models are metonymic. But the most complex cognitive models correspond to what Lakoff callsradial categories.Theseconsist of many models linked around a center.Although noncentral models (and categories)cannot be predicted by a knowledge of the central category, they do have a relationship to the centeq they are said by Lakoff to be "motivated" by it. Suchproperties permit degreesof membership,degreesof relationship to the central model, family resemblances,nonhierarchicalrelationships with basic categoriesdominating, and prototype effects.Prototype effects are not fundamental but arise from many sources-/'scalar," "classical," "metonymic," and "radial." With this background,Lakoff attempts to mount a structure for cognitive semantics(figure P-S). Notice ftrst that meaning is already based in embodiment by meansof image schemas,kinestheticschemas,metonyms, and the categoricalrelationsthat underliemetaphor.But this is not enough: Languageis supposedto be characterizedby symbolicmodels. These are models that pair linguistic information with the cognitive models that themselvesmake up a preeristingconceptualsystem.Inasmuchas preexisting conceptualmodels are already embodied through their link to bodily and social experience,this link is not an arbitrary one. In contrast, the attribution of such a linkage to generative grammar in terms of mental representationsis arbitrary; it is made from on high by the grammarian. In this view of cognitive semantics,linguistic categories naturally show shong structural resemblancesto their underlying cognitive models. Language makes use of general cognitive mechanismsto construct propositional models, image schematicmodels, metaphoric models, and metonymic models. As we have said, metaphoric models involve a mapping from a strucfure in one domain to a corresponding strucfure in another domain. This mapping involves either propositional or image schemas.Metonymic models use these schemasand metaphor to map a function from one element of a model to another (for example, a partwhole relationship). In Women,Fire, and Dangnous Things: What CategoriesReoealAbout the Mind, Lakoft usesthe work of his colleague Johnson (The BodVin the Mind: The Bodily Basisof Meaning, lmagination, and Reason)to construct a series 247

MrND WrrHour

BroLoGy: A

CnITIcAL PosrscRIPT

GRAMMAR COGNITIVE

Conceptswith MetaphoricalUnderstanding

Schemasbased on Embodiment

Categories Hierarchies Relations

Semantics Propositions Scenarios

ldealized CognitiveModel (lcM)

STAGE2 (Build Syntaxfrom Semanticsand the Categorization of Sentenceand PhraseRelations)

rt

; I

F

SyntacticICM I Syntactic usingsameschemas: Constructed HierarchicalSyntacticStructure GrammaticalRelations SyntacticCategories

Part-Whole Schema

Link Schemas ContainerSchema

FIGUREP-8 grammar according to Lakofr. In contrast to a cognitiae in An exampleof Tocesses and experience (see areacquiredthroughlinguisti.c rules P-Z), grammar generatioe figure -meanings irise becauseconceptsare embodied.This kind of gtammar has not .yet been attemptssuchas theleical slnwn'iutty to haoetheanalyticpowerof mainlinegeneratioe gioe a scheme But it does for the relationof meaning functionalgrammar of Bresnan. -(through structure.The "stages"are not embodiment)to categoizntionand sentence nutoirily sequentialin time i"a tn y overhp, oboiously they haoe-the leastooerlap duing eirly kng*gt acquisition.ICM : idealiz*d cognitioemodel.

248

Mind Without Biology' A Critical Postscript of schemas,basedon embodiedconcepts,that provide a basisfor linguistic meaning.Theseinclude containerschemas(defining a boundary, or "in and out"), a part--uholeschema,a link schema(one thing connectedto another, sdterna(asin body centerversusarms and asby a string), a center-periphery legs),and a sourcrpath-goal sdrema (starting point, directional path, midpoint) including uVdown andfront-baclcschemas.He then goes on to show that metaphorsare motivated by the structuring of experienceresulting in schemas. The sourcrpath-goal schema, for example, emerges from our bodily functioning, pervadesour experience,is well structured,and is well understood. The source domain as well as the target domain of any metaphor based on it will be conelated experientially in terms of this schema.The prior basicJevel and image schematicconcepts are directly meaningful and provide the basis for the schema.They also provide starting points for the rules of semantic composition that form more complex conceptsfrom simpler ones. Theseideasare encapsulatedin Lakoffs "spatializationof form hypothesis." According to this hypothesis, categoriesare understood in terms of containerschemas,hierarchical structure is understood in terms of partwhole anduy-down schemas,relational structure is understood in terms of lirekschemas,the radial structure of categoriesis understood in terms of schemas,foreground-background structure is understood center-periphery in terms of front-back schemas,and linear quantity scalesare understood in terms of urdown and linear order schemas.All involve a metaphorical mapping from physical (or spatial) structuresto conceptualstructures. But where, specifically,does language come in? With ifualizedcognititte models,some of which are structuresconsisting of symbols' Thesemodels are of five types: image schematic,metaphoric,metonymic, propositional, and symbolic. Of these, the ones that lead to linguistic function are the propositional and symbolic idealized cognitive models. A propositional idealizedcognitive model doesnot usemetaphor,metonymy, or mental imagery. Instead,it uses basiclevel concepts---entities, actions,states,and properties.Simple propositions follow the part-whole schema:the proposition is the whole, of which the predicateis one part and the arguments(agent,patient, experiencer,instrument,locatioo and so on) are the other. Semanticrelations are built from link schemas,and complex propositions are then formed from simple propositions by modification, quantification,conjunction, negation, and so on. Moreover, scenarioscan be built of an initial state,a sequenceof events,and a final state structured by a sourcrpatlr2oal schema. idealizedcogniWhen linguistic elementsare associatedwith conceptunl They can models. symbolic idealized cognitive models, these become tive 249

MrNp Wrrnour

Blorocv: A Cnrrlclr- Posrscnlpr

then be characterizedin terms of the moqphemesand words of particular languages.A noun, for example,is a radial category (centralcategoriesare people, places, things; noncentral categories are abstract nouns like "strength"). A oerb is also a radial category (central categoriesare basiclevel physical actions like run, hit, give). Remaining members of these categoriesare motivated by relations to these central members.The relation to semanticsis obvious. And what of syntax itself?Lakoff claimsthat the principleshe discusses allow us to provide a semanticbasisfor syntactic categories.According to his theory, hierarchicalsyntactic structure (seefigure P-7 for an example) is itself characterizedby part-whole schemas,head and modifier structures are characterizedby center-peipheryschemas,grammatical relations are clraracterizedbylink schemas,and syntacticcategoriesare characterizedby containerschemas.Notice what is happeninghere:Grsmmaticalconstructions are thmrseloesidealizedcognitioemodels.Thus, semantic pairing with syntax is a direct pairing of an idealized cognitive model for syntax with a pior idealized cognitive model for semanticsor meaning. Regularitiesin the strucfure of grammar and in a lexicon can be describedin terms of radial categories,and words with multiple meanings can be explained in these terms. In this view, and in Langackey's,languageis basedon cognition-that is, on cognitive models that can be understoodin terms of bodily functioning. This cognitive baseis constrainedby the nature of physical reality and also dependson imagination and socialinteractions.Meaning derivesfrom embodiment and functio& understandingariseswhen conceptsare meaningful in this sense,and truth is consideredto arisewhen the understanding of a statementfits one's understandingof a situation closely enough for one's own puposes. (Notice the pragmatism!)Thus, there is no absolute truth or God's-eye view. Our view of what exists (metaphysics)is not independentof how we know it (epistemology).As Lakoff puts it, "Truth is a bootstrapping operation, grounded in direct links to preconceptually and distinctly structuredexperienceand the conceptsthat accordwith that experience."This fits with my proposals related to qualified realism in chapter 15. Knowledge, like truth, is a radial concept.It dependson our understanding, on basiclevel concepts,and also on socially acceptedunderstanding. It is secureto the extent that human understandingcan be secure,but it is always subject to revision. Objectivity is not absolute but dependson looking at a situation from as many points of view as possible and by distinguishing basicJevel concepts and image schematic concepts from concepts that are only indirectly meaningful. 250

Mind Without Biology: A Critical Postscript Obviously the exampleprovided by Lakoffs cognitive grammar (figure P-8) is radically different from the more widely acceptedgenerativegrammars (seefigure P-7). It differs in philosophy, style, and methodology. It is in closer accord with the biological basesof brain and bodily function and with the psychological data on categorization.It avoids the category mistakesof the "language of thought" proposal and the objectivist error inherent in generative grammar. It is an imaginative and important proposal. But in proposing embodiment as the origin of meaning,it does not show fioo this might come to pass. Nor does it show how symbolic idealizedcognitive models of languageariseas a result of the mechanisms of perceptualand conceptualcategorization.For these tasks,one needs a generalbiological theory of brain function and a theory of consciousness, both based on the facts of evolution and development. That is what I attempted to construct in my trilogy and to review in this volume. It may be useful to comment on the relation between Lakoffs cognitive grammar and the theory of speechacquisition described in chapter 12. Cognitive grarnmaris basedon the notion of embodiment,but it does not specify how such embodiment takesplace.Instead it searchesfor signs of radial categories, metaphor, and metonymy as guiding structures for speech.And, similarly, it usescategorizationto accountfor the emergence of syntacticalrelationships.In all these respects,it is compatible with the epigenetic theory presentedin chapter L2. This theory clarifiesthe issues related to evolution and to the acquisitionof speechin a way that Lakoffs theory, lacking a description of mechanismsof embodiment, cannot. Indeed,the epigenetictheory provides additional grounds for taking aspects of an extensivestructuralgenerative theory of grammar such as Bresnan's (which stresseslexicon) and linking them to a categoically basedtheory such as Lakoffs. Langackey'streatment,while rejecting the generativeaspectsof Bresnan'sapproach,resemblesit in stressingthe importance of lexicon. A complete understandingof grammaticalformulations requires an analysis of the brain mechanismsfor concept formation, value-categoryformation, connection to the phenotype, and connection to the mechanismsof consciousness.It also requiresexploration of the granunarsof particular languages in the terms describedby Langacker,Bresnan,and others. A rich field of study might be opened up by exploring how a theory of embodiment such as the one I have presented might bridge and relate these important but different approachesto linguistic theory. Lakoffs Women,Fire,and DangerousThingscameout at about the same time as my book NeuralDarwinism,which attempted to provide a basisfor a global brain theory. I know that I was unawareof his book and surmise that he was unaware of mine. The central problem confronted by Neural 251

MIND WlrHour

BroLoGy: A CnrrrcAL PosrscRlpr

Darainism was perceptualcategorization.In a subsequentwork, The ReI extended the brain Present: A BiologicalTheoryof Consciousness, membered theory to perceptualexperience,conceptformation,and language.In retrospect, it appearsthat these two books nicely complement Langacker's, Lakoffs, and Johnson'swork, providi.g an essentialbiological underpinning for many of their proposalsconcerningthe importanceof embodiment to grammarand cognition.But neithertheir work nor mine deniesthe significanceof the efforts of other linguists to understandsyntacticstructure. The importanceof their efforts and of the efforts of cognitive psychologistsis very great. But without biology, they remaininsufficientand even, at times, in error. This is what I have attempted to show in this Postscript. For those who have read both the text and the Postscript,I hope the challengehas been made sufficientlyclear.We must incoqporatebiology into our theoriesof knowledge and language.To accomplishthis we must developwhat I have calleda biologically basedepistemology-an account of how we know and how we are aware in light of the facts of evolution and developmentalbiology. A fuller realizationof this goal will expandour scientifichorizons.And through its connectionsto what makesus uniquely human,a biologically basedepistemologywill enrich our lives.

252

Readings Selected

I have prt together the following referencesfor those readerswho want to obtain additional background or follow up an idea. Where I have mentionedsomerecentoriginal work in the text, I have attemptedto cite a paper and hold at least to the minimal requirementsof scholarship. For extensivebibliographies,the readermay consultthosein my trilogy NeuralDarainism, and TheRememon morphology and mind: Topobiology, beredPresent(all publishedby BasicBooksand listed below). The comments following the given referencesmay be useful to some readers.They reflect my personalopinions but are not at all complete.

CTIAPTERI '1,890. Reprint. New York Dover, 1950. W. The Principlesof Psychology, JnrraEs, A monumental work by one of the founders of experimental psychology. Contains penetrating descriptions and analyses,as well as strong personal opinions. Jnvns,1 y'."Does ConsciousnessExist?" In The Writings of William James,ed. J. I. McDermott. Chicago: University of Chicago Press,L977, 169-83. is a process,not a thing A seminalwork that makesthe casethat consciousness or substance. FreNncaN,O. ]., In. TheScience of theMind, Znd ed. Cambridge,Mass.:MIT Press, T99T. A nice survey of the thoughts and work of modern psychologists, with a balancedassessmentof their present standing. 253

SnrEcrED RneDTNGS BnrxrnNo, F. Psychologyfrom an Empirical Standpoint,ed. O. Kraus and L. L. McAlister, trans. A. C. Rancurelloet al. Highlands,N.1.,Humanities,1973. The k.y work of the psychologist,philosopher,and ex-priestwho emphasized the importance of intentionality. A professorin Vienna, he influencedFreud, who attendedhis lectures. GnEcoRy,R. L. Mind in Science, Cambridge:Cambridge University Press,T987. An historical account of the uneasy relationship between scientific methodology and the matter of the mind. Discursive, inconclusive,but rich in suggestions. GnrrnN, D. R. Animal Thinking, Cambridge, Mass.: Harvard University Press, 1984. A spiritedcasefor animalawarenessmadeby a noted ethologist.In other works he has gone so far as to suggestthat beesare conscious.I believe the casehas scarcelybeen made (and indeed is highly unlikely),but Griffin's way-out position is usefully provocative. PnEuncK,D., and A. J. PnEnancr.TheMind of an Apt, New York Norton, 1983. This is a splendid and clear accountof the mental capabilitiesof chimpanzees. Current, expert, and stimulating. McCuttocn, W. S. Embodiments of Mind, Cambridge,Mass.:MIT Press,7989. Essaysby r brilliant forerunnerof modem neurobiology, reprinted.Useful and imaginative. Identity, BmrEuoRE,C., and S. GnnnNFrELD, eds.Mindwa:res:Thoughtson Intelligence, New York Blackwell, 1987. and Consciousness, A collection of paperson these subjectsby neuroscientistsand philosophers. Displays the major issues,the confusions,and the different positionsof various practitioners. Gnsconv, R. L, ed. The Orford Companionto theMind, Oxford' Oxford University Press,1987. A small encyclopediawith articles on many issuesby a variety of experts. Spotty, but useful and lots of fun for the intellectually curious. Great for browsing.

CIIAPTER2 Wnnnrtep, A. N. Sciencennd the Modern World. New York Macmillan, 7925. The classicaccount by a logician, historian of science,and metaphysician.Puts the relation between the scientific observer and the subjectively reflective individual in a rich historical perspective. Gelnu, G. "The Assayer," 1623, trans. S. Drake. ln Discooeiesand Opinionsof Galileo.New York: Doubleday, 1.957.Dnert, S. Galileo.Oxford: Oxford University Press,L980. In both books, the "founde/ is on display and his thoughts summarizedby one of his most devoted historians. Those who feel that scientiststoday are too 254

SelectedReadings concernedwith priority should read Galileo's complaintsin the beginning of "The Assayer."Incidentally,Galileo understoodthe nature of secondaryqualities (color, warmth, and so forth) almost a century before Locke. R. Meditations and Passionsof the Soul, In The PhilosophicalWorks of DEscenrES, vols. 7 and 2, ed. E. Haldane and G. Ross.Cambridge:Cambridg. Descartes, University Press,797E. If Galileo is the founder of modem science,Descartesis the founder of modern philosophy. His thoughts are proof that genius,even genius leading to wrong conclusions,can be of continuing major significance.We still wrestle with the questionsthat Descartesposed.

CHAPTER3 New York: Oxford University Press,7983. G. Neurobiology. SHspHsRp, A standardelementaryaccount of modem findings. Many others exist, but this has most of the basics,with some discussionsat the level of principles. New York: Lurun, A. R. The Working Brain: An Intraductionto Neuropsychology, BasicBooks, 7973. The late, great clinician and neurologist from the Soviet Union gives a clear accountof what happenswhen parts of the brain are disturbed.In following the courseof his descriptions,a newcomerreceivesa larger picture of the functions of the brain proper than can be culled from a book like Shepherd's. DpEnor, D. Le RAoede d'Alembert,7769. Reprint. New York: Penguin, 1966. The little classicquoted in the text. The original is in St. Petersburg;Diderot was an adviserto Catherinethe Great.My understandingis that his collaborator on the Encyclopedia,d'Alembert,was none too pleasedto have his relationship with Mlle de l'Espinasseso openly on display. CHaNcEUx, I.-P. NeuronalMan: TheBiologyof Minl. New York Oxford University Press,7986. A popular accountby a neurobiologistwho believes,as I do, that the brain is a selectionalsystem. Contains quick surveys, historical matters, and a short account of connectionsbetween cognitive psychology and neurophysiology.

CHAPTER 4 For a history of philosophyat the elementarylevel,seeB. Russnrr,A Historyof WestemPhilosophy. New York Simonand Schuster,rSaS.This is dear,prejuastringentlydone,see dice4 and stimulating.For moremodemdevelopments in the TwentiethCentury.EastHanover,N.J.,Vintage A. ]. Avrn, Philosophy 1,9s4.A good (if rathertechnical)setof historicalaccountsof modempsychol-

2s5

S E T E c T E DR u a D r N G s Psychologv, ogy may be found in E. HnaRsr,ed. The FirstCenturyof Experimental Hillsdale,N.l.: LawrenceErlbaumAssociates,1979. KeNzs.e'G. Organizntionin Vision:ksays on GestaltPerception, New York Praeger, rg7g. Wonderfully revealing accountof the world of so-calledvisual illusions.Masterful and a minor classic,with marvelousillustrations.

CHAPTER5 or the Preseroation DnnwtN, C. On the Origin of Species by Meansof Natural Selection of FaoouredRacesin the Strugglefor Life. London: Murray, 1359. The masteqpiecethat provided the foundation for modern biology. BrRRErr,P. H., P. J.GnurRuy, S. HrnnERr,D. KonN, and S. SuttH, eds. Charles Metaphysical Daranin'sNotebooks,183F1844: Geology,Transmutationof Species, Enquiries,Ithaca:Cornell University Press,1957. Views into the mind of a great scientistand thinker. Rovnuns, G. J. Mental Eaolutionin Animals,New York Appleton, IE84. Mental Eoolutionin Man. New York: Appleton, 7889. The thoughts of Darwin's contemporary. Good examples of how a great provocative theory takes root in widespreadterritories. MeyR, E. The Growth of BiologicalThought:Dirsersity,Eoolution,and Inheritance, Cambridge,Mass.: Harvard University Press,1982. A modern masteqpieceby a great evolutionist. One of the best accountsof Darwin, Darwinism, and the "subtheories" that make up a complex theory like the modern theory of evolution. of EoolutionaryTheoriesof Mind and Rrcnenps, R. J. Darain and the Emergence Behaoior,Chicago: University of Chicago Press,1987. A very comprehensiveexposition, rich in scholarship.The best up-to-date account of this subject

CHAPTER6 THoupsoN,D. W. On Growthand Form.Cambridge:CambridgeUniversity Press, 1942. One of the great classicson the subject of animal form, by a talented nonbeliever in Darwin. While its direct examplesare not very relevant to the nervous system,it is nonethelessfascinating. An Introductionto MolecularEmbryology,NewYork: EpnrunN,G. M. Topobiology, 1988. BasicBooks, A more extended and fundamentalaccount of the subject of this chapter.The last chapter of Topobiology,whichis by virtue of its subject matter the first

2s6

SelectedReadings volume of the trilogy on moryhology and mind (although not the first to be published), describes the connection between topobiology and selectionist theories of the brain.

CHAPTER8 Bunnrr, F. M. The ClonalSelectionTheoryof Acquiredlmmunify. Nashvillq Vanderbilt University Press,1959. This is the original extended account of Bumet's selectionistviews. For the evolutionary background, see Mayr's book, cited above.

CHAPTER9 New EpnrueN, G. M. lVeuralDarwinism: The Theoryof NeuronalGroup Selection. York BasicBooks, 1987. This book lays out the theory of neuronal group selection in ertensowith its defenseswell up. Much more scholarly than the present account.The original skeletal exposition of the theory may be found in G. M. Edelman and V. B. Mountcastle, The Mindful Brain, Cambridge,Mass.:MIT Press,1978. Benrow, H. B. "Neuroscience:A New Eril" Nature331, no. 1S (February1988): '1,2, 571- Cnrcr, F. "Neural Edelmanism." Trendsin Neurosciences no. 7 fluly 1989): 24V48. PunvEs,D. Body and Brain:A TrophicTheoryof Neural Connections.Cambridgu, Mass.:Harvard University Press,1988. These three authors attack aspectsof the theory of neuronal group selection. The abbreviated counterattacksare presentedin this chapter. Vive le sport! Mlcnop, R. E. "Darwinian Selection in the Brain." Eoolution 43, no. 3 (19S9): 694-96. A favorable review agreeing that the theory of neuronal group selection is a selectionistaccount in the spirit of population thinking. See also his reply to Crick's criticism of the TNGS, along with the accompanying reply of G. N. 13, no. 1 (1990): 1I-I4. Reeke,|r., in Trendsin Neurosciences For more information on the work of Eckhorn and his colleaguesand Gray and his colleagues,seethe referencesin Sporns,O., J. A. Gally, G. N. Reeke,fr., and G. M. Edelman, "Reentrant Signaling Among Simulated Neuronal Groups of the National Leads to Coherency in Their Oscillatory Activity." Proceedings 86 (1989): 726549. AcademVof Science

IO THROUGH13 CHAPTERS EpnuunN,G. M. TheRemembered Present: A BiologicalTheoryof Consciousness. New York BasicBooks, 1989. The last of the trilogy on moryhology and mind. An attempt to provide a 257

Srrscrnp REnorNcs Althoughit is thelast principledscientificaccountof thebasesof consciousness. volumeof the trilogy, I am told by somethat it is best readfirst. Mencrl, A. J.,and E. Brsncn, eds.Consciousness in Conternporary Science. Oxford: Clarendon.1966. A collectionof valuablepaperson the subject,rangingover a wide area. Benrrrrr, F. C. Remembering: A Studyin Erpeimntal and SocialPsychology. Cambridge:CambridgeUniversityPress,1964.Thinking:A Studyof Huttutts.New York BasicBooks,1958. Two classics, especiallythe fust,which givesa profoundanalysisof the act of remembering. Hras,D. O. TheOrganization of Behaoior: A Neuropsychological Theory,New York: Wiley,7949. 19E0. Hsss,D. O. Essayon Mind. Hillsdale,N.f.: LawrenceErhaumAssociates, phenomena Theftrstis oneof the earliestattemptsto accountfor psychological in termsof neuronaland cellularinteractions.The secondcontainsthe later reflectionsof a modemmaster. Fnrup,S. "Projectfor a ScientificPsychology."ln TheStandardMition of the Complete Psychalogical Worksof Sigmund Freud,vol.1, ed.J. Strachey.London: Hogarth, L976, 283-4L7. IntroductoryLectureson Psychoanalysis. New Yorlc Liveright, 1919.NewIntroductory Lectures on Psychoanalysis, New York Norton, L933.On Dreqrts,ed. J. Strachey.Reprint.New York Norton, 1963. "Projectfor a ScientificPsychology"elatedFreudandthenrepelledhim.It was savedfor posterityby his friend,Marie Bonaparte.The rest of theseworks representthe best kemelsof the maste/swork at a popularlevel.The chef d'oeuvre,On Dreqms,was the work he consideredto be his greatest. Enprrvt,M. H. Psychoanalysis: Freud'sCognitioePsychology. New York Freeman, 1985. An excellentaccountof key Freudianconcepts,particularlyFreud'sideason memoryand the unconscious. In koceedings of Bnotrwr&L. E.I. "Consciousness, Philosophy,andMathematics." of Philonphy,vol. 2, ed.E. W. Beth,H. J. Pos, the Tmth Internstional Congress andJ. H. A. Hollak.Amsterdam: North Holland,t91s,1.235-49. pieceof imaginationby a topologistand philosopherof matheA remarkable matics.Post-Kantian, hard to grasp,but very stimulating. Hrrcnnp,E. R. DividedConsciousness: Multiple Conholsin Hunun Thoughtand Action.Expandededition.New York lNiley, 1977. A different view decidedly post-Freudian.Full of fascinatingexamplesand For and hypnoticphenomena. of split and multipleconsciousness discussions seethemarvelousbookby a startlingexplorationof visualGestaltphenomena, GaetanoKanizsathat is listedin the readingsfor chapter4. andLearning. CambridgeCambridgeUniverSreppoN,I. E. R. AdaptioeBehavior sity Press,1983. A good accountof leamingseenfrom a wide biologicalbase. eds. Artxeuprn,R.D. "Evolutionof the HumanPsyche."ln TheHumanReoolution, P. Mellarsand C. Stringer.PrincetoryNJ.: PrincetonUniversityPress,1989. 258

Selece t d Readings Self-deceitas an adaptive phenomenon in the deceiving of others during the struggle for survival-an unlikely but fascinatinghypothesis.

CHAPTER14 Rorn, G. C. "Mathematics and Philosophy, The Story of a Misunderstandi^g." Reaiewof Metaphysicsaa (December 1990): 259-71.. A p*gently phrasedand thoughtful article by u distinguishedmathematician on the distrust one should have of excessivereliance on axiomatics: "The snobbishsymbol dropping one finds nowadays in philosophicalpapersraises eyebrows among mathematicians.It is as if you were at the grocery store and you watched someonetrying to pay his bill with Monopoly money."

15AND 16 CI1APTERS of Philonphy,vols. 1-4. New York The Free Epwenos,P., ed. TheEncyclopedia to theWorld:TheBssicConcepts of Philosophy. Press,1973.D*vro, A. Connections L972. of Philosophy. New York Harper& Row, 1989.Russnr,B. TheProblems Reprint.Oxford: Oxford UniversityPress,1959. A referencework and two lucid inhoductions,one new, one old. Science. Hillsdale,N.J.: of Mitrd:An Oventiew BscnreLW. Philosophy for Cognitioe 1988.CnuncxreNp,P. M. MatterandConsciousLawrenceErhaumAssociates, ness.Cambridge:MIT Press,1984. Two short introductionsto the philosophyof mind. An Intermittently Philosophical Dictionary.Cambridge: QurNE,W. Y. Quiddities: Press, 1987. HarvardUniversity/Belknap Amusing idioslmcraticnotes by an outstandingAmericanphilosopherand Iogician. WrrrcENsrrrN,L. Philosophical Inaestigations. The English text translationby G. E. M. Anscombe.New York Macmillan,1953. The posthumouslypublishedrevisionistviews of one of the most interesting discussed philosophicmindsof thiscentury.Bearson the issueof categorization throughout the presentwork. Wnrrrnrep, A. N. Modesof Thought.New York The FreePress,1938. Metaphysicalreflections,toutcourt,by a modemthinkernow temporarilyfallen from gracein most university circles.Well worthwhile for its imaginativeand suggestiveinsights. Necn, T. 'lVhat Is It Like to Be a BaB" Philosophical Reoieut 83 (7974):435-50. TheViewFromNowhere, New York CambridgeUniversityPress,1986. Penetratingand clear analysesof the dilemmas of epistemology and metaphysics. Ryru,G. TheConcept of Mind. Chicago:Universityof ChicagoPress,1949. A searingcritiqueof categoryerrorsin the philosophyof mindby the originator of the phrase"the ghost in the machine." 259

Srrrcrro

RrlprNcs

Russrrr,B. A Historyof Western Philosophy. New York Simon& Schuster,1945. Already mentioned.Lucid,sui geneis;an accountby one of the pioneersof mathematical logic and one of the most courageously opinionatedof modem philosophers. AvBR"A. J. TheProblem of Knowledge. Middlesex,N.J.:Penguin,1956. A view of the whole issuein epistemologyfrom a formerlogicalempiricist. Pl,c,crr,J.BiologyandKnowledge: An FssayontheRelatiotrs BetwemOrganicRelatiorc and Cognitioe kocesse*Chicago:Universityof ChicagoPress,1971. The views,presented herefor contrast,of a greatdevelopmental psychologist. Not only idiosyncraticand original but revealingof the gulf betweenthe attitudesof scientists andphilosophers. in its compariSomewhatmetaphorical sonsof embryologyand psychology. DevIs,P. I., and R. Hrnsn.Descqrtes's Dream:TheWorldAccordingto Mathematics. Boston:HoughtonMifflin, 1966. A beautifulaccountby two mathematicians of the natureand limits of mathematics.Unsympathetic to the Platonicview of mathematics. Moncrl, M.l. Molyneur'sQuestion: Vision,Touch,andthePhilosophy of Perception. CambridgeCambridgeUniversityPress,'1.977. An elegantessayon somehistoricalaspectsof the psychologyof spatial perception.Never mind the view of a bat-what would happenif you were alwaysblindandthensuddenlyregainedyour sight?Wouldyour "touchspace" and "visualspace"correspond? Hurr, J. M. Touchingthe Rock:An Ftpeienceof Blindness. New York Pantheon, 7990. A moving accountof how an individual'sconsciousness is alteredby the loss of vision. Bovo, R., and P. J. Rrcnnsow.Cultureand the Eoolutionary Process. Chicago: Universityof ChicagoPress,1985. A remarkablybalancedaccountof how humansocialbehaviorand evolution may interact.One of the bestforaysinto this dangerousthicket. Bennow,J. D., and F. f. Trrrnn. TheAnthropicCosmological kinciple, Oxford, Oxford UniversityPress,1988. Perhaps Sourceof my quoteon wastepaper baskets. the relativelylargenumber of references on philosophycontainedin this sectionprovesthe contentionof that quote(seepage 159). Frrw, A. An Inhoductionto WesternPhilosophy: IdessandArgumentsfrom Platoto Popper. New York Thamesand Hudson 1989. of the ideaof the soul. Containsa good discussion to chapter4. Seealsothe references

17 CHAPTER AneNpr,H. TheLife of the Mind, vol. 1, Thinking;vol. 2, WilW.San Diego: HarcourtBraceJovanovidr,1978.

S e l e c t e dR e a d i n g s A philosophicalr6sum6.It is revealingto contrastits views with those of readings to Bartlett,seetheselected experimentalists like Bartlett(for references for chapters10-L3). vols. 1-3. Baltimore Johns Lrucnn, S. K. Mittd: An ksay on lluman Feeling, HopkinsUniversityPress,L967,7972,1973. A masterftrlsurveyby a philosopher,historianof ideas,and scholarof artistic symbolism.Poignantly cut short in volume 3 by the author's advancing blindness. New York Norton, of Emotionand Stress. MeNplE&G. Mind andBody:Psychology 1984.SorouoN, R. C. 1978.ThePassions. Notre Dame Universityof Notre DamePress,1983.Dr Sousr"R. TheRationalityof Emotion.Cambridge:MIT Press.1987. Three accountsof emotion, the first scientific,the last two philosophical. Together, they bring out the extraordinarily complex, multilevel nature of emotions.

CHAPTER 18 Behaoior, andtheMind. New York Wiley, 1979.Kots, Wrrutus, M. BrainDamage, 3rd ed. San of HumanNeuropsychology, 8., and I. Q. Wnrsnlw. Fundammtals Freeman,1990.McCmrnY, R. A., andE. K. W.*nncrou. Cognitioe Francisco: A ClinicalIntroduction.New York Academic,1990. Neuropsychology: Three books on the effectsof brain damage(seealso A. Luria The Working to chapter1). Brain,in the references vols.7-2. Textbook of Psychiatry/IV, Kerrev H. L, and B. Sepocx.Comyehmsioe Baltimore:Williams& Wilkins, 1989. A largepsychiatrytext. For the brave. Moprrr, A. H. Othn Times,Other Renlities:Towardsa Theoryof Psychoanalytic E.M.Philoso' HarvardUniversityPress,1990.HuNpsRr, Cambridge: Treatment. Mind. ThreeApproaches to the Oxford: Clarenphy,Psychology, and Neuroscience: don, 1969. Two psychiatristsapply the theory of neuronalgroup selectionto aspectsof their subject. Scnacrr& D. L., M. P. McANpnrws,and M. Moscovlrcn. "Accessto ConDissociations BetweenImplicit and ExplicitKnowledgein Neurosciousness: ed. L. Weiskrantz. psydrologicalSyndromes."ln ThoughtWithoutLanguage, '1988,242-78. BtsncH,E. "LanguageWithout Thought." Oxford: Clarendon, ed. L. Weiskrantz,Oxford: Clarendon,1968, ln ThoughtWithout Language, 465+1. Two revealinga*icles on somedissociativesyndromesof consciousness. Secrs,O. TheMan WhoMistook His Wifefor a IIat and OtherClinicalTales.New York Harper& Row, 1987. A fascinatingset of accountsby a humanemind, a clinician,and a magniftcent storyteller. 267

S E T E c T E DR s e D T N G S CHAPTER19 RnErcE, G. N., Jn.,and G. M. EpErueN."Real Brains and Artificial Intelligence." Daedalus1I7 , no. 1,,(Winter ISSS); 1.43-73. See also Reeke,G. N., Jr., L. H. Finkel, O. Sporns,and G. M. Edelman."synthetic Neural Modeling: A MulLocal tilevel Approach to the Analysis of Brain Complexity." ln Signaland Sense: and Global Order in PerceptualMaps, eds. G. M. Edelman, W. E. Gall, and W. M. Cowan, New York: Wiley-Liss, I99O:607-707. Edelman,G. M., and G. of the N. Reeke,Ir. "ls It Possibleto Construct a PerceptionMachine?" Proceeds (1990): Philosophical American 3G73. Society134, no. T These articles reflect the philosophy underlying a major researchprogram at The NeurosciencesInstitute. The secondis quite technicaland extensive.The third offers some reflectionson the relation of this work to other work in the "real thing"-is yet in print, but undoubtfield. No article on NOMAD-the edly one will be publishedby the time this book appears.

1',

CHAPTER20 Aprun, R. K. The Great Design:Particles,Fields,and Creation,New York Oxford University Press,1987. A beautiful summary of modem physics with a smidgeon of cosmology. Technical,but worth the effort. Znn, A. Fearful Symmetry: The Searchfor Beauty in Modern Physics.New York: Macmillan, 1956. WEvt, H. Symmetry,Princeton:Princeton University Press, 1952. Tnnesov, L. This Amazingly SymmetricalWorld. Moscow: Mir, 1986. Three books on the significance of symmetry in the field of physics and elsewhere.Zee'sbook is the easiest.Weyl , a great mathematician,wrote his book quite early in the game. No referencesare given here for memory as a principle of nature. The references cited earlier should serve to capture many of the details.

MIND WITHOUT BIOLOGY:A CRITICALPOSTSCRIPT Here I must give a rather long (but still incomplete)list arrangedaccordingto the order of the subheadingsof the Critical Postscript.

THE SURROGATE SPOOK PHYSICS: PENnosE, R. The Emperor'sNeutMind, Oxford: Oxford University Press,1989. book in the senseof the sizeof its salesto the laity. Charming A very successful 262

Selece t d Readings and lucid accountsof strangephysics,quantum measurement,and so forth. But most of the book is practically irrelevant to its goals and claimsabout the mind, as I discussin the text. LocrwooD, M. Mind, Brain,and the Quantum:TheCompound"L" Cilrrbridge: Basil Blackwell, T9s9. A philosopher'sdiscussionof many of the same issuescovered by Penrose. Inconclusive. Definedby the New Zonen, D. The QuantumSeIf:Human Natureand Consciousness Physics,New York: William Morrow, 1990. Quantum this, quantum that, quantum everything. A book that goes as far out in the domain of physics as the surrogatespook as anything well-intentioned can go. Compared to Penrose,a very soft example of the genre.

THE FAISEANALOGUE DIGITALCOMPUTERS: Hopcns, A. Alan Turing: The Enigma.New York Simon & Schuster,1983. Biography of a remarkablemind and a life with a sad ending. Turing is the key theoretical figure in computing, outside of von Neumann. Of course, many other logicians and mathematiciansset the stage,as I mention in the text. JonusoN-Lenp, P. N. TheComputerand theMind. Cambridge:Harvard University Press,1988. The best account of the mind-as-machineview. GneusARD,S. R., ed. "Artificial Intelligence."DaedalusI17, no. 1 (198S). A seriesof essayson the subject,both supportive and critical. and Reality.Cambridge: MIT Press, 1988. PuruaM, H. Representation machine functionalism-by one of A refutation of his own doctrin*Turing the most distinguishedliving philosophers. Foundationsof Research, eds. Neurocomputing: ANpensoN,J. A., and E. RosENFELD, AxpnnsoN, A. and E. RosnNFELD, MIT Press, 1988. A., Pnrr.IoNlsz, Cambridge: J. MIT Press, 7990. Direction Research. Cambridge: eds. Neurocomputing: for aspects and connectionism. of neural modeling Two collectionsof paperson

UCIOUS CIRCLES IN THE COGNITIVEIANDSCAPE H. The Mind's New Science, GA,npNER, New York Basic Books, 1985. An excellentgeneral survey of cognitive science. WrrrcENsrErN,L. Philosophical Inoestigations. The English text of the 3rd ed. New York Macmillan, 1953. Already cited-pioneering in its early dissectionsof the problemsof categorization and family resemblance. The article by G. C. Rota listed in the readings for chapter 14 pertainshere too. The second part of the figure on categorizationand polymorphous sets (P4, 263

SsrEcrED REaDTNGs right) is from Dennis et al., "New problem in concept formation." Nature 243 (1973): 107-2. RoscH,E. "Human Categorization." In Studiesin Cross-CulturalPsychology, ed. N. Warren. New York Academic, 1977, 1-49. An accountby one of the most important psychologistsin the areaof categorization. BEnuN,8., and P. Kev. BasicColor Terms:Their Unioersalityand Euolution.Berkeley: "ProbabilUniversity of Califomia Press,1969.Tvensry,A., and D. KnHNEMAN. ity, Representativeness, and the Conjunction Fallaqt." Psychological Reoiew90, no. 4 (1,990):293-31,5. Pionee.ing studies on color categorization and on inference,inductive and otherwise. The referencesto L. Rips and to L. Barsalouin the text may be found in their articles,which are included tn Similarity and AnalogicalReasoning, eds. S. Vosniadou and A. Ortony. Cambridge:Cambridgu University Press,1959. Fooon, |. A. Representations: PhilosophicalEssayson the Foundationsof Cognitioe Cambridge:MIT Press,1981. Science. By the prolific philosopher and defenderof "mentalese." Menn, D. Vision:A Computationallnoestigationinto the Human Representation and Processing of Visual Information,San Francisco:Freeman,7982. The last work by this late influential figure in psychophysicsand neuroscience. Espousesthe computationalview but gives a good summary of "early" visual processes.Gets worse in later chaptersas it addressesproblems of categorization. MrrureN, R. G. Language,Thought,and OtherBiologicalCategories: New Foundations "Thoughts Realism, Cambridge: MIT Press, 1984. Without Laws; Cognitive for Sciencewith Content." Philosophical ReoiewXCV, no. 1 (Jan.7986): 47-80. A major figure in what I have called the Realists'Club, Millikan has put forth a powerful and original critique of what she calls meaning rationalism (roughly equivalent to what I have inveighed against in the Postscript). Gnurp, A. "Cognitive Psychology,Entrapment,and the Philosophy of Mind." In TheCqsefar Dualism,eds.I. R. Smythiesand J.Beloff.Charlottesville:University Pressof Virginia, 7989, 757-253. One does not have to agreewith dualismto appreciateGauld'sscaldingattack. B. "semantic Representationof Meaning: A Critique." Psychological SHeruoru, Bulletin 104, no. 1 (1988):7V83. A fine surrrmaryof the difficulties of functionalism, objectivism, and the ideas of mental representation. PurNau, H. Representation and Reality.Cambridge:MIT Press,1988. Already mentioned, a chip from the master'sworkbench. Bnuun& l. Acts of Meaning. Cambridge: Harvard University Press, 1990. A beautiful essayby one of the foundersof modem cognitive sciencepleading for the recognition of narrative as an important aspect of our mental life. and Action, ed. H. Heuer VoN HorsrsN, C. "Catchi.g." ln Perspectioes on Perception and A. F. Sanders.Hillsdale,N.J.t LawrenceErlbaumAssociates,1987, 33-46. An attack on the information processingview from the motor side. 264

S e l e c t e dR e a d i n g s LeNcrcxrn,R.rN.Foundations of Cognitioe Grammar, vol.7, Theoretical Prerequisites. Stanford:StanfordUniversityPress,'I.,987. An accountby one of the earlyworkersin this importantfield,work that has been extendedby Lakoff (seethe next reference). Lexorr, G. Women,Fire,and DargerousThings:What CategoiesReoealAbout the Mind. Chicago:Universityof ChicagoPress,1987.JoHNsoN, M. TheBodyin the Mitd: TheBodilyBasisof Meaning,lmagination, and Reason. Chicago:University of ChicagoPress,1987. Two important booksby authorswho have collaborated.They contain referencesto the other authorscited in this section.Johnsonmakesthe casefor metaphorasa powerful resultof embodiment.Lakoffs more extensiveaccount includesa historyof ideas,a critiqueof thenotionof mindwithoutbiology,and a pleafor the recognitionof embodimentas the basisfor meaningand mind. It providesa basisfor muchof the last part of the CriticalPostscript.It also describesat length the cognitivegrammarI havesummarized here.The present book and my trilogy may be considered a theoreticalanswerto the question "How is the mind embodied?" This questionis raisedby the aforementioned works in sucha way that it cannotbe ignored. wHY THEFORMAL IANGUAGE: APPROACH FAITS Csotusrt N. CartesianLinguistics. New York Harper & Row, 7966. Rulesand Representations. New York ColumbiaUniversityPress,1980. Two works by the most influentiallinguistof recenttimes,a defenderof the formal approachand its most powerful proponent. Llcrnroor, D. TheInnguageLottery:Towarda Biologyof Grammars, Cambridge: MIT Press,1982. An informativeaccountby one of the epigones. and theComputationsl Mind, CambridgeMIT Press, J^rcrrwoorr,R. Cotrsciousness 1987. As good a surnmaryof the combinedview of languageas syntaxand mind as machineas one can get. The result: consciousness as an epiphenomenon. Obviously,I rejectthis view. Bnrsxeu,I. ed. TheMental Representation of GrammaticalRelatiota.Cambridge: MIT Press,1982. An extensiveaccountof lexicalfunctionalgrarnmaqone basisfor the analysis by Pinkerof languageacquisition.Heavy going for neophytes. Pwrr& S. Language Learnabilityand Language Deoelopment, Cambridge Harvard UniversityPress,1.984. An intelligentand penetratinganalysis.Technical. Doxrrpson, M. Childrm'sMinfu. New York Norton, 1.928. Lesstechnical.Charmingand yet nearly lethal in its attack on the idea of a languageacquisitiondevice. Lnwrr, W. I. M. Speaking: FromIntentionto Articulation.Catnbidge MIT press, 1989.Gnrcr,H.P. Itgic andCononsation.ln Studies in Syntat vol. 3, ed.p. Cole and I. L. Morgan.New York Academic1967, AL-SE. 265

Srrucrrp Rreptxcs Levelt'saccountdealswith the actualproblemof speakingin a comprehensive way. The frameworkis still mainsheamcognitive science,however.Probably the best singlevolume on the subject.Grice is the highly original analyzerof the requirementsfor effectiveexchangein speech. CambridgeMIT Press,1962. Vvcorsrt L. S. ThoughtandLanguage. The Soviet thinker (a colleagueof A. Luria)who emphasizedinterpersonaland socialexchangesand the "interiorization" of speechfor purposesof thought. Psncy,W. TheMessagein the Bottle:How QueerMan Is,How Qruerl-atguagels, and Whst One IIas to Do with the Other.New York Farrar,Straus& Giroux, r976. Amateur,moving, and at the sametime, deep.Proof that thought requiresno advanceddegreeandthat havingan M.D.is not alwaysruinous.Percywasa ffne minor novelist. Cambridge,Harvard University Kru'ar,8., and U. BEuucL The Signsof Lnnguage. Press,1976. A discussionof Bellugi'spioneeringstudiesshowing that sign languagehasa of spokenlanguage. syntax, dialects,and other characteristics Chicago:University of ChicagoPress,1990. BrcreRroN,D. Inngwge atd Species. A valiant attempt to accountfor the evolution of speechby meansof an Provocative. intermediate"pidgin" or a protolanguage. Beluo' Thought,and Selfless LTnIERMAN, P.lJniquelyHuttun: TheEoolutionof Speech, ior, Cambridge:HarvardUniversityPress,1991. accountby an authorityon the evolutionof the speech A good semipopular apparafus. H. TheStoryof My Lit'e.1902.Reprint.New York Doubleday,gSL KeLrER, Poignant. A fitting last reference-the achievementof linguistic behavior againstall odds.

266

Credits

QUOTATIONS Philosophers, vol. 2 (New Frontispiecefrom Empedocles,in J. Barnes, ThePresocratic York' Routledg., Chapman and Hall, 1982), 180. Feynman, R., The Characterof PhysicalLaw (Cambridge:MIT Press, 1965), 125. Dedicationfrom Ecclesiastes, 6:78, circa 250 s.c. Works of Descartes, Chapter 1 from Descartes,R., Meditations,ln The Philosophical vols. 1 and 2, ed. E. Haldane and G. Ross (Cambridge CambridgeUniversity Press,1975). De Unamuno, M., TragicSenseof Life, trans. C. J. Flitch (New York Macmillan, 1927), 34. Chapter 2 from Whitehead, A. Macmillan, L925), 2-3.

Scienceand the Modern World (New York:

Chapter 3 from Maxwell, J. C., in TheAnthropic Cosmologicalkinciple, J. D. Banow and F. J. Tipler (Oxford: Oxford University Press,1986), 545. Diderot, D., Rameau'sNephew/D'Alembe{s Dteam (London: Penguin, 1.966 1r769D,270-72. Chapter 4 from Adams, H., The Hucation of Henry Adams (New York, Houghton Mifflin, 1961). Chapter 5 from Darwin, C., in H. Zinsser, tk I RememberHim: The Biography of R. S. (Magnolia, Mass.: Peter Smith, 7970lt939l). 267

Cnrolrs Transmutation 1836-1844: Geology, Datwin'sNotebqoks Darwin C.,in Charles of Species, MetaphysicalEnquiries,ed. P. H. Banett, P. J. Gautrey,S. Herbert, D. Kohn,and S. Smith(lthaca:Comell UniversityPress,1.987),539. Chapter6 from SpitzeaN., "The ChickenandtheEgg,Togetherat Last"(areview TheNan York Times An Inhoductionto MolecularEmbryology), of Topobiology: BookReoiew,22lanuary 1989,p. 1.2. trans.A. J. Krailsheimer(New York Penguin, Chapter7 ftom Pascal,8, Pensies, 1966). Chapter6. Anonymous. 12,no. Trends in Neuroscimce Chapter9 from Crick,F.H.C.,"NeuralEdelmanism," 7 (fuly 1989\ 247. lnoestigations, ThePhilosophkal Chapter10 from Wittgenstein,L., in Wittgmstein: ed. G. Pitcher(London:Macmillan,1968),465. Voltaire,in Philonphical Dictionary,vol. 1, ed.PeterGay (New York Basic Books,7962),308. BrieferCourse(CambridgeHarvard UniChapter11 from fames,W., Psychology: versity Press,1964),401. '12 Chapter fromFocillon,H., in K. Atchity, AWiter'sTime, (New York Norton, 1986),78O. Works Psychological Chapter13 from Freud,S.,TheStandardHition of theComplete Freud,vol.14,trans.anded.James Strachey(London:HogarthPress, of Sigmund L976),280. Yal6ry,P.,in W. H. Auden,A CertainWorld(NewYork Viking,1970),26f,. (Winter,7958):93. Chapter14 from Bridgman,P.W., in "Quo Vadis"in Daedafus Chapter15 from Holmes,O. W., TheCompleteWorlcsof Olioer WendellHolmes' (St.Clair Shores,Mich.: ScholarlyPress,1972). EinsteinA., in K. Atchity, AWriter's Tirze(New York Norton,1986),780' Chapter16 from Planck M., in J. D. Barrow and F. f. Tipler, TheAnthropic hinciple(Oxford:Oxford UniversityPress,1986),123. Cosmological Allen, W. By permissionof Mr. Allen. A., Counsels andMsrims (St.ClairShores,Mich.: Chapter17 fromSchopenhauer, ScholarlyPress,1981).

Credits ed. M. Bonaparte, Chapter 18 from Freud, S., The Origitrs of Psycho-Analysis, A. Freud,and E. Kris (New York BasicBooks, 1.954),I2O. Chapter1.9from de la Mettrie,I. O., L'HommeMachine.Reprinted.(Peru,Ill.: Open '1.961), I4F4I. Court Publishing, Chapter 20 from Einstein,A., in R. H. March, Physics for Poefs(Chicago:Contemporary Books, 1978), 735. Val6ry,P., in The PracticalCogitator,ed. C. P. Curtis, lr., and F. Greenslet (Boston:Houghton Mifflin, t962), 597. Critical Postscript quote in text from Quine, W. V., The Ways of Parador (New York Random House, 7966), 66.

ILLUSTMTIONS Figures 7-7, T-2, 4-1., and 5-I from the Mary Evans Picture Libr?ri!, London. Figures 2-I and 2rI permission.

Copyright @ American Institute of Physics.Reprintedby

Figure 2-_2from H. Banow, C. Blakemore,and M. Weston-Smith,ed., Imagesand (New York: Cambridge University Press,1990), 261. Copyright Understanding @ 1990 by Cambridge University Press.Reprinted by permission. Figure 3-1, from C. Blakemore and S. Greenfield, ed., Mindwaoes (New York: Basil Blackwell, 1987), 4. Copyright @ 1987 by Basil Blackwell. Reprinted by permission. Figures 3-2 and 3-3 from Gerald M. Edelman, Topobiology'An Introductionto Molecular Embryology,Copyright @ 1988 by Basic Books, Inc. Reprinted by permissionof HaqperCollinsPublishers. Figure 3-5 (top left) from K. G. Pearsonand C. S. Goodman, "Correlation of variability in structure with variability in synaptic connection of an identified interneuron in locusts,"Journalof ComparatioeNeurology184 (1.979):741-65. Reprinted by permissionof Corey S. Goodman. Figure 3-5 (bottom) from Michael M. Merzenich et al.,"Topographic reorganization of somatosensorycortical areas 3b and 7 in adult monkeys following restricted deafferentation"and "Progressionof changefollowing median nerye sectionin the cortical representationof the hand in areas3b and 1 in adult owl (19S3): and squirrel monkeys," Neuroscience I (19S3): 33-55 and Neuroscience !O:63945. Reprintedby permissionof Michael M. Merzenich.

Cnrplrs Figure3-5 (top right) from EduardoR. Macagno,V. Lopresti,and C. Levinthal, "Strucfureand developmentof neuronalconnectionsin isogenicorganisms: variationsandsimilaritiesin the optic systemof Daphniamagns,"kouedingsof 70 (L973):5741. Reprintedby permissionof the NationalAcadmy of Science EduardoR. Macagno. andthe Rousseau Figure4-1 (left) from W. andA. Durant, TheStoryof Cioiliz,ation: 1967).Reprintedby permisReoolution, vol. x (NewYork SimonandSchuster, sion of the estateof EthelDurant. Figure4-2 Reprintedby permissionof GreenwoodPublishingGroup,Inc.,Westby Gaport, Conn., from Organizationin Vision:Essayson GeshtaltPerception etanoKanizsa(New York Praeger,1979),78, and74. Copynght@ 1979by GaetanoKanizsa. Figure 5-3 adaptedfrom Richard Lewontin, The GeneticBasisof Eoolutionary (New York ColumbiaUniversityPress,1974).Copyright @ 97aby Change RichardLewontin.Reprintedby permission. Figure5-4 (left) from Hugo lltis, TheLifeof Mendel(New York HafnerPublishing Co., 1966).Reprintedby Unwin Hyman,of HarperCollinsPublishersLtd. 1836-1844:Geology,TrattsFigure5-5 reprinted trom CharlesDarutin'sNoteboolcs mutqtionof Species, MetaphysicalEnquiries,transcribedand edited by Paul H. Barrett,PeterJ. Gautrey,SandraHerbert,David Kohn, and SydneySmith. Originally publishedby the BritishMuseum(NaturalHistory).Copyright @ 7967by Paul H. Barrett,PeterGautry, SandraHerbert,David Kohn, Sydney Smith.Usedby permissionof the publisher,Comell UniversityPress. to Humanity Figure5-6 from G. LedyardStebbins,Darutinto DNA, Molecules (New York W. H. Freemanand Company,7982).Copyright @ 1932 by W. H. Freemanand Company.Reprintedby permission. Figure6-2 from C. M. Anderson,F. H. Zucker,and T. A. Steitz "Space-Filling 204 (7979): Models of KinaseClefts and ConformationChanges,"Science on p. 376.Copyright@ L979bytheAmericanAssocia375-80.Figureappears Reprintedby permission. tion for the Advancementof Science. "CellAdhesionMolecules: A Molecular Figure6-3 (top)from GeraldM. Edelman, Amnican250,no. a G984):118-29.Figure Basisfor Animal Form,"Scientific appearson pp. 12O-Zt. Copyright @ 1984 by ScientiftcAmerican,Inc. All rights reserved.Reprintedby permission. Body,Sth ed' (Orlando, Figure 6-3 (bottom) from Alfred Romer, TheVertebrate Copyright@ 1977by Saunders CollegePublishing,7963),1'1.9. Fla.:Saunders CollegePublishing.Reprintedby permissionof the publishers. 270

Cr e d i t s Figure6-5 reprintedby permissionof Dr. WalterJ.Gehring,Universityof Basel, Basel,Switzerland. Figwes9-1.,9-2,9-4,9-5, 77-1, 72-4, and,tZ-S from GeraldM. Edelman,The Copyright @ 19s9 by kesent:A BiologicalTheoryof Consciousness. Remembered BasicBooks,Inc. Reprintedby permissionof HarperCollinsPublishers. Figure 9-2 reprintedby permissionof Dr. Semir Zeki, University College, London. The Theoryof Figure 9-3 (top) from Gerald M. Edelman,NeuralDaruninism: Basic Books, Inc. Reprintedby L987 by Copyright @ NeurorwlGroupSelection. Publishers. permissionof Haqpe€ollins O. Spoms,andG. N. Reeke, Figures}{ (top andlowerleft) from G. M. Edelman, of PopulationandConnectionist Jr.,"syntheticNeuralModeling:Comparisons F. Fogelin Perspectioe, ed.R. Pfeifer,Z. Schreter, Approaches"in Connectionism and L. Steels(New York ElsevierSciencePublishers,L9S9),12O man-Soulie, and tz6. Copyright @ 1989 by ElsevierSciencePublishers.Reprintedby permission. Figurets6 (lowerright)from G. M. Edelmanet al.,"SyntheticNeuralModeling," ed. G. M. Edelman,W. E. Gall, and W. M. Cowan(New in Signaland Sense, Reprintedby permisResearch Foundation,L99O),695. York TheNeuroscience sion. (CanFigure 12-1.from Philip Lieberman,TheBiologyand Eaolutionof Lnnguage bridge:HarvardUniversityPress,1984).Copyright @ 1984by the President and Fellowsof HarvardCollege.Reprintedby permissionof the author. Fig;rc 72-2 (top) from W. Penfieldand L. Roberts,SpuchandBrainMeclunisms. Copyright @ 1959by PrincetonUniversityPress,renewed1957.FigureX-4, p. 201,reprintedby permissionof the literaryexecutorsof the PenfteldPapers and PrincetonUniversityPress. of French Figure12-2 (bottom)from FrancisSchiller,PaulBroca:Founder Anthropology,Erplorerof theBrain(Berkeley, Calif.:Universityof CalifomiaPress,'1,979), 189. Courtesyof Dr. Juster,Institut de Parasitologie, EcolePratique,Paris. Reprintedby permissionof the author. Figure 13-1 copyright @ 1983 by SigmundFreudCopyrights.Reprintedby permissionof A. W. Freudet al., by arrangementwith Mark Patterson& Associates. Figure P-3 (top) reproducedby permissionof the Computer Museum, Boston. 277

CnEDrrs FigureP-3 (bottom)courtesyof N-Cube,Belmont,Calif. FigureP-6 (right)from I. Dennis,J.A. Hampton,andS.E.G.Lea,"New Problem in ConceptFormation," Nature243(1973):101.Copyright@ 1973Macmillan Magazines,Ltd. Reprintedby permissionof Ian Dennisand Nature,

272

Altruism, 47-48 Amino acids, 53 Anosognosia, 186 Antibodies, 75-77 Arborization, 25, 27 Artificial intelligence, 22F227 '1,4I-L44; basal ganglia and, Attention, 1,42-143;learning, I42; motor theory, I42; naffownessof, L43; selective, L4I--1,42 Augustine, St., 168 Automatic activity, intemrpted by novelty, r43 Avery, Oswald, 53,55 Axiomatic systems, 153

Barlow, Horace, 94, 257 Bartlett, Sir Frederic,38, 258 Basal ganglia: in attention, 142-'1,43; memory, 105-1,06 Behaviorism, II-12, 38 Berkeley,George, i5 Biological organization, layers, 746 Biology, 23F24I; linking to psychology, 34 Bisiach,Eduardo,1.86,262

Blastoderm,58 '1.22, 183 Blindsight, Bohr, Niels, 216 Brain: anatomical alrangement, 22-23; as computer, 27, 228; functionalism, 220, 222-224; networh 25; nuclei, 18; organizing principles, 2I-22; size, complexity, and behavior, 5L; structures,7; as Turing machine,227 Brain cells, functioning pattems, 17 Brain stem, 11,8-1,1,9 Brentano,Franz,5, 253 Broca, Paul, 38 Broca's atea, 127-128 Brouwer, L. E. ]., L68 Bnrner, Ierome, 175, 265 Bumet, Sir Frank MacFarlane,77, 257

Cartesianrationalism, 34-35 Categorization, 233-237; alterationsin mental disease,I80; mechanismsin global mapping,'/..52;polymoqphous sets, 235; selectivesystem, 169 '/.,57 Causality, Cell adhesion,60-61 Cell adhesion molectrles,6H.2

273

INpnx Cell differentiation,homeotic gen es,62 Cell junctional molecules,60 Cells, processes,58 Cerebellum,memory, lls Cerebral cortex, I7-I8 Chomsky, Noirl, 242-243, 265 Church's thesis, 220 Classification couple, 87, 90; valuecategory memory, 207 Clonal selection,77 Cognitive grammar, 244-246, 24825'1, Cognitive models, I52, 24F247; idealized, 249; symbolic idealized, 24925r Cognitive science,I3-L4 Cognitivism, 14-15; critical arguments against,L5 Computational view, 233 Computers,218-227; algorithm, 22F 222; artificial intelligence, 22F227; physical symbol system hypothesis, 222-223; Turing machine, 2IF220 Computer simulation, 1,89-190 Conceptual abilities, 108-110; in animals, 108; evolutionary development, 108-109; frontal cortex, 10971.0;global mappings, IO9 Conceptual categorization, T25, 233236; concurrent perceptions, I19 Conceptual embodiment, 246 'LIg Conceptualmemory system, Connectionist systerns,22G227 Connections: cortical sheet, !7; neurons, I*20 Conscious artifact, L88-I96; future ability to construct, 194-I9S; NOMAD, 192; seealso Darwin III Consciousness,3 7, 67, II7-I23; adaptive advantages,!33, I35; definition, '1,33-134; tIt-M, 168; evolution, evolution and temperature,L53;me'1,!6; diation, I22; models, mystique, '1,37-1.38; I39; properties, qualia, LTA;qualia assumption, 113--1.16;as

result of naturalselection,I49; scientific explanation, 138-13g; theoretical assumptions, LI3-II4; theories, II2-II3; seealsoPrimary consciousness,model Conservation laws of physics, connection with symmetry, 2OO,202 Cortex, memory, 1,04-T06 Cortical appendages, mem oty, I04107 Cortical sheet,connections,17 Cosmologicaltheories, 203 Cosmology, 797 Craik,K. I.W., 230 Cranial capacity,evolution, 48, 50 Crick, Francis, 94-95, 257

Danto, Arthur, IST Darwin, Charles, 42-43, 256; journals, 4g-49 Darwin III, 97-94, I9I; behavior, 792; global mapping, 93; memory, IO2IO3; output through reentrant maps, 93; value signal, I92 Darwin's program, 42-46, 48 Deixis, 242 Dendrites,20, 25, 27 Descartes,Ren6, 4, 17, T46, 255; on thought, 34-35 Determinism, 1,70 de Vaucanson,Jacques,1,8F789 Development:evolution and, 5l; place dependencies,22, 24 Developmentalstudies,40, 69 Diderot, Denis, 19, 2!, 167, 255 Dilthey, Wilhelm, 1.77 DNA, 52-55; replication, 53-54, 206

Ebbinghaus,Hermann,37 Einstein,Albert, 9, LL, 1,59

274

lndex Elliot, Stuart, 53 Embryo: formation, 574I; induction, 58, 60 Emotions,176 ENIAC, 227 '1,65-166 Enlightenment, Entorhinal cortex, 106 Entropy, 204 Epigenetic events, 23, 25; involving interactions with same species,4G 47 Epistemology, 167-162 2I2, 239 Essentialism, Ethology, 46_.47 Evolution: development and, 51; natural selection, 43-44; recognition, 78.79 Evolutionary assumption,I73

Gestalt phenomena,38-39 Global mapping, 89-91, I25; altered attention and priorities, I43;' categorization mechanisms, 752; concePtual abilities,109; Darwin IlI, 93; distinguishing classes,'J.09 Global symmetry, 203 Godel, Kurt, 1,52-153 Grammar, 242; cognitive, 244-246, 248-251; generative, 243-244; lexical functional, I29 Grand unification theory, 202 Group selection,48 Group theory, 2O0

Hebb, Donald, 39, 258 Hegel, Georg, 777 '10, Heisenberg uncertainty principle, 215 Hereditary principle, 204, 206 Faraday,Michael, 13 Hereditary process, memory, zOG Feelings,176 207 Fermi, Enrico, 79 Ewald, 36 Hering, 228 Fodor, Ierry, 'J,70 100 brain'functions, Higher Free will, IL2, II5, consciousness, Higher-order 1.79, Freud, Sigmund, 13,744-145, evolution, 13I-T36,'l'98; L24-I26, 182, 258 -1,49-150; evolutionary time, !33; Fritsch,Gustav,38 language, I25, 245; primary conFrontal cortex, conceptual abilities, requirement,T22;requiresciousness 709-rr0 ments, I25; scheme,T32; semantic Functionalism,220, 222-224, 23I, bootstrapping, I5O; speech, \2G 239 73I; symbolic memoU, 125 Functionalistviews, 229-230 Hilbert, David, I53 Hippocampus,memory, IOGL)7 Hitzig, fulius, 38 Homeostats,94 Galilei, Galileo, 9-Io, 254 Homeotic gene, 6243 Galois, Evariste, 200 Hominids,behavior,49 Gedankenexperiments,7I3 Homunculi, 80-81 General theory of relativity, I99 Hull, Clark Leonard,38 Generativegrammar, 243-244 Genes:changesin frequency of, 44-45; Hume, David, 35 form alteration, 48; homeotic, 62- Hundert, Edw ard, 1.82,261 Husserl,Edmund, 759 63; sequentialexpression,60 275

I Npnx Idealism,2I2 Immune memoty, 207 Immune system, 74-78 Intentionality,S, 68, II2, 230;in neurological disease,I87

Locke,]ohn, 35 Lorenz, Konrad, 40 Lymphocytes,75-77

Malthusian population dynamics, 94 Mapping, 19;semantics,6T;of types of maps, L09; seealso Global mapping |ames,William, 6, 3G37, 253 Maps, 69; anatomical,28; connectedby |ohnson, Marh 239, 247, 265 reentrant fibers, 8G87; during development, 6344; fluctuations in borders, 27-28; formation, 22-24; functioning, 28; local, 9F9I; par,allel Kanizsa,Gaetano, 39, 256 and reciprocal connections, 85; viKant, Immanuel, 35-36 sual area,86 Kin selection,4E Knowledge' heterogeneity,777; loops, Marr, David, 230, 264 Mayr, Emst, 73, 256 r48 Meaning,'J.70, I75, 224-225, 228; Koffka, Kurt, 38 semantic representation, 229-230; Kohler, Wolfg ang, 38 slmrbols, 226 Kreisler,Fntz, I04 Medawar, Sir Peter, 12 Memory, 98, lOO-108,L6Ft'69; basal ganglia,'1.05-!06; cerebellum,ro5; '1,04-L06; cortical appendcortex, Lakoff, G., t3o, 239, 24F249 III, 102-103; Darwin 104-1,07 ages, 265 !30, 245-246, Ronald, Langacker, ; prohereditary 207; development, 776, 26'1. Langer, Suzanne, 106hippoc:unpus, 20F207; cess, Language:acQuisition,242-244; cogniI07; languageand, 237-238; neurotive grarnmar, 244-246; definition, nal group selection, I0T; orderinI of 243; evolution, 125-126; formal successivechanges,I04; recategoriviews, 241-245; grammar, 242; zation, 1.02, I04; symbolic, T25; higher-order consciousness, I25, symmetry and, 203-208; temporal 245;memory and, 237-238; relation element, 167; theory of neuronal between form and meaning, 242; regroup selection, I02-I03; types, lation with thought, 242*243; soul 205;seealsoValue-categorymemory based on, 40; syntax and semantics, Mendel, Gregor, 46-.47 239 Mental disease,!78-187; alterationsin Lashley,Karl, 39 reentrant pathways and categonzaLearning, 28, 10f101; attention, 742 tion, 180; anosognosia,1'86;dissociLexical functional grammar, I29 ation, I83; functionaldisorders,I79; Lie, Sophus, 200 lesionsunderlying, I83; neurological Limbic system, 7-l,8-lIg disorders,178, 1.82-'l'84;psychologLipmann,Fritz, 199 ical development,I79; reintegration, Local maps, 9C-l91 1,67;schizophrenia,'l'84-186;theory Local symmetry, 203 276

Inder Neurally OrganizedMultiply Adaptive Device (NOMAD), 792-193 Neural network, architecture,227 Neural patterns, variability, 25-26 Mentalese, 229-230 Neural tube, 5940 Mental representation,228, 233-234 Neurological disease,18F'1.8I; intenMerzenich, Michael, 96 tionality, 787 Metaphor, 237, 24F247 Neurological disorders,178, 182--1.84 Metonymy, 237, 24F247 Neuromodulators,22 Michod, Richard, 96 Neuronal groups, demonstration of Millikan, Ruth Garret, 240, 264 existence, 95; heritability and Mind, 3-8; analogy with computer, selection,gGgT; m€mory and selec176; concept bases, I52; biological nerIO'j,; somatic selection, tion, in of, 5; connectionwith events T3; II, 94 vous system, 7; doctrine, 4, mortality of, I7I; physical matter Neurons, IG1.7; connections, 19-20; numberof, 17; sensitivityto stimulaunderlying, T6; as process,G7; Puttion, 22; shapes,\9 ting back into nature,11; things easy Neurophysiology, 3S-39 and hard to imagine about, L40 Neuroses,I79 Modell, Arnold, !82, 26r Noether, Emmy, 202 Modern synthesis,47 Noetic devices,I92 Monod, |acques,745 NOMAD, 792 Morality, 1.7T-r72 'j,8 Nuclei, Morgan, C. Lloyd, 40 49-50; evolutionary, Nucleotidebases,53 Moryhology, Numbers, idea of and time, 168 29 Motor program, 105-106 Motor theory, attention, I42 of neuronal group selection, L79180; value-categorymemory, I83'1,84

Natural selection, 42-44, L48; consciousnessas result of, I49; Darwin's program, 44-46; differential reproduction, 63 N-CUBE, 227 Nerve cells, 2I; "neighborhood," 64 Nerve signals,22 Nervous system: anatomical arrangement, 22; responsepattems, 226 Neural Darwinism, 8I-98; Darwin III, 9I-94; neuronal group, 8G88; perceptual categorization, 87, 89-9I; premises,160;r€€ntrantcorticalintegration model, 89;seealsoTheory of neuronal group selection

Objectivism, 14, 67-68, 230-2fi; inition, 230 Objects,partitioning, 28 Onsager, Lars, L86 Optic tectum, 22-24 Orientation tuning, 95

def-

Panpsychism,2I2-2I3 Parapraxes,T43 Pavlov, Ivan, 37-38 Penrose,Roger, 21.6,263 Perceptual categorization, 87, 89-9I, !OO,1"25;manifestation,9C-9I Percy, Walker, 245 Phenotype,57

INpnx Philosophy, 157-L64; relation to science, I58; selfhood, 1,67 Physical symbol system hypothesis, 222-223 Physics, 2I2-2L8; revolution, 273; scalesof nature, 2I3-ZL4 Physicsassumption,TI3 Piaget,Iean, 40, 260 Planck,Max, 65, I99, 2L3 Plato, 239 Polymoqphoussets, 235 Polypeptide, 53 Population thinking, 44, 73, 239 Postsynapticneuron, 79 Primary consciousness,II2-113; evolution, 149-1,50; Iamesian properties, L27, 150; required for higherorder consciousness, !22; theory of neuronal group selection, 721, Primary consciousness model, II7I23; conceptual memory system, II9; diagram, I2O; evolutionary development, 1,1,F1,I9; evolution of functions,116-1 L9; nervoussystem organization, TL7; qualia assump'l..I9; tion, 135-136; reentrantcircuit, theoreticalassumptions,135 Primary repertoire, 83, 85 Principle of complementarity, 216 Property dualists,12 Prosopagnosia, I22, 1,39 Protein, folding and function, 55-57 Pseudo-randomnurnber generator, 790 Psychiatricdisease,180-181 Psychology' as branch of biology, 240; linking to biology, 34; schools of thought, 36; social,40 Purves,Dale, 96, 257 Putnam,Hilary, 223-224, 237, 263

Qualia assumption,113-116, 135-136 Qualified realism, 16I-T62 Quantum gravity theory, 2I7 Quantum measurementproblem, 214216 Quantum mechanics,70, 799 Quantum theory, 215 Quine, W. V., 233, 259

Rabi, Isidor, 216 Rachmaninoff,Sergei, I04 Radial categories,247 Ram6n y Cajal, Santiago,39 Realism,qualified, 16r-162 Recategorization, 102, I04 Receptor sheets,L9 Recognition, 73-80; anatomicaldiversity, ll;brain, 79-80; definition, 7475; evolution, 78-79; immune system, 74-78; mechanisms,97 Recursivesynthesis,S9 Reductionism,!66, I70 Reentrant circuit, primary consciousness,model, rr9 Reentrant connections,between valuecategory memory and perception, 1,33 Reentrant cortical integration model, 89 Reentrant pathways, alterations in mental disease,180 Reentrantsignaling, 85, 87, 89-90 Reintegration,mental disease,187 Replicating system,aperiodic structure, 206 Representationalbeitrgs, I57 Repression,745 RNA, 55; replication, 206 Romanes,George,40 Rosch,Eleanor, 236 M; as higher-order Rous,Peyton,55 categorizaQualia, Russell,Bertrand, 232-233 tion, 11,6

lndex Schacter,Daniel, I83 Schizophrenia,!84-186 Science,relation to philosophy, 158 Scientific thought, revolution, 799 Scientific view, assumptions,759 Searle,Iohn, 224, 238 Secondary repertoire, 85; modeling plastic changes in map boundaries, 96 Second law of thermodynamics, 204 Selectional events, memorial events, T69 Selectionisttheory, T63 Selective system: categorization, 169; simulation, 1,90 Self-categorization,719-120 Selfhood, 167 Semanticbootstrapping,I27,'l'.29-130, 150 Sensoryproperties,levels, 151 Sensory transducers,18 Sherrington, Sir Charles,38 Silent synapse, 27 Skinner,B. F., 38 Social psychology, 40 Sociobiology, 47-48 Somatic selection, I48 Space,202 Spatializationof form hypothesis (Lakoff), 249 Speech: ocQuisition and conceptual power, L3O;epigenetictheory, 126I3I; semantic bootstrapping, I27, L29-r30 Spencer,Herb ert, 40 Split-brained individual, r8T Stimulus-responseparadigrl, 3 7-35 Substancedualism,17 Substrateadhesion molecules, 6A42 Supralaryngealtract, 1,2FI2T Symbolic memory, I25 Symbols, meaning, 226 Symmetry, L99-204; breaking, 203, 207; connection with conseryation

laws of physics, 2OO, 202; global, 203; locai, 203; memory *d, zo3208; types, 2Al Synapse,17, 19-20; silent, 27 Syntactic systems,153 Syntax, !3, 239; development, I29730;evolution, 150; idealizedcognitive model, 250 Szilard,Leo, 79

Teleonomy, 16I Thalamocortical systems, II7-II8 Theory of clonal selection, 77 Theory of instruction, 75 Theory of neuronal group selection, 82-85, 99-700; criticisms, 94-95, 97; extended,151-152, 16T;homeostats, 94; memor|, LO}-IO3; mental disease, 179-180; perceptual categonzation, 89; premise, 85; primary consciousness,I2I; in psychiatry, T8I-!82; tenets, E3-84; unit of selection, 85-86 Theory of relativity, I99 Thomdike, Edward, 38 Thought, 33, I73-|TS; experiments, II3; neural correlations with meaning, 176; neuroscientificexplanation, 174-775; purethought, 174; relation with language, 242-243 Time, 202; connection with idea of numbers, 765 Tinbergen, Nikolaas, 40 Titchener,Edw ard, 37 Topobiology, 52-64; CAM and SAM roles,62; celldivision, 58, 60; characteristic shape, 60; individual diversity, 64; seeuentialgene expression, 60 Transferencerelation, 182 Truth, IST Turing, Alan, 2I8

279

INpnx Turing machine, 218-220; functional- Vesalius,Andreas, 18 ism, 220, 222-224; intemal states, Visual area,maps, 86 225 Visual cortex, orientation columns, 'l',68 Two-icity, 95 von Helmholtz, Hermann,36 von Neumann, ]ohn, 2I4 Uhlenbeck,George, !98 Unconscious, 144-146; Freud's psychological theories, I45; repression, Wallace, Alfred, 42 145 Warhol, Andy, I95 Watson, Iohn, 38 Wave function, collapse, 215-216 Wernicke's area, L27--1.25 Value-category memory, 1.L9, I2C_ Wertheimer, Max, 38 I2I, 132-133; classificationcouples, I,Vhitehead,Alfred North, *!r, ziz207; formation, I2C-'l.,21;interaction 233, 254 with conceptual and speech areas, Wigner, Eugene, 2I4 '1,30; mental disease,183-'1,84;reen- Wittgenstein, Ludwig, 234-236, 259, trant connection with perception, 263 133 Wundt, Wilhelm, 36

280