the-eye.eu american 1993-2013... · t he pitohui sounds like what it is: something to spit out. the...

JANUARY 1993$3.95

Sticky sugars: carbohydrates mediate many cellular

interactions, such as infection and inßammation.

The turbulent birth of the Milky Way.

Lemurs: a glimpse at our evolutionary past.

From quantum dots to designer atoms.

Copyright 1992 Scientific American, Inc.

January 1993 Volume 268 Number 1

64

72

82

90

Coral BleachingBarbara E. Brown and John C. Ogden

How the Milky Way FormedSidney van den Bergh and James E. Hesser

Carbohydrates in Cell RecognitionNathan Sharon and Halina Lis

The Earliest History of the EarthDerek York

Extensive areas of the subtly colored coral reefs that gird tropical shores havebeen turning a dazzling white; some stretches of the aÝected coral have evendied. Bleaching may be a call of distress from these complex and highly produc-tive ecosystems, usually emitted when they experience abnormally high seawatertemperatures. Do bleached reefs signal global warming?

For more than a decade, astronomers have believed our galaxy and others like itformed from the rapid collapse of an enormous cloud of hydrogen and heliumgas. Observation no longer entirely supports this simple model. The Milky Waycame into being under the inßuence of exploding stars, its own rotation and per-haps a propensity to capture and gobble up other protogalaxies.

Carbohydrate molecules are the chemical braille that enables cells to recognizeand respond to one another. With them, bacteria identify their hosts, and thecells of the immune system single out diseased tissue. Carbohydrates also directcellular organization in embryos. Nature has selected them for such coding be-cause they form the largest number of combinations from a few components.

The earth is extremely good at destroying evidence of its past. The massive tecton-ic plates regularly plunge under one another, returning the ocean ßoor to moltenoblivion and causing continents to collide. Yet increasingly sophisticated radioac-tive dating techniques are enabling geologists to pry the history of the planetÕsÞrst billion and a half years from ancient, previously taciturn continental rock.

The proper study of humans is the lemur. Of all living creatures, none moreclosely resembles the ancestor from which humans and the great apes branched50 million years ago. But the lemursÕ diverse Madagascan habitats are disappear-ing fast, and so are they. Hundreds of species are already extinct; unless huntingand deforestation cease, the rest may meet the same fate.

4

110 MadagascarÕs LemursIan Tattersall

Copyright 1993 Scientific American, Inc.Copyright 1993 Scientific American, Inc.Copyright 1993 Scientific American, Inc.

118

124

130

The Mind and Donald O. HebbPeter M. Milner

Born in Nova Scotia early in this century, Hebb began his adult life as an aspiringnovelist and enrolled at McGill University on the theory that a writer of Þctionshould understand Freud. By the end of his life he was one of the most importantpsychologists of his time, laying the groundwork for contemporary neuroscience.

What do bacteria colonies and economies have in common? In trying to Þnd out,a group of multidisciplinary researchers at the Santa Fe Institute hope to derive a theory that explains why all such complex adaptive systems seem to evolve to-ward the boundary between order and chaos. Their ideas could result in a view of evolution that encompasses living and nonliving systems.

DEPARTMENTS

50 and 100 Years Ago1893: Convincing a kangarooto Þght by QueensberryÕs rules.

160

142

150

154

16 12

14

5

Letters to the EditorsOf a diÝerent mind... . Whenbiotech comes to dinner.

Science and the Citizen

Science and BusinessBook ReviewsStargazing.. . . A tome of ani-mals.. . . Stairs, a step at a time.

Essay: Howard M. Johnson

What it takes for a black to suc-ceed in a white science.

The Amateur ScientistEver wonder how manyspecies live on your lawn?

Rapid progressÑand big surprisesfrom the genome project.. . . Steppingup the search for dark matter.. . . Anend to lonely nights.. . . Verifying theaccuracy of huge proofs.. . . Poisonousplumage.. . . PROFILE: NeurobiologistRita Levi-Montalcini.

Drugmakers return to their roots.. . .Slices of life.. . . A new mission for the weapons labs.. . . Peeking insidecompetitorsÕ parts.. . . Programmer-friendly software.. . . THE ANALYTI-CAL ECONOMIST: Rationalizing invest-ments in infrastructure.

TRENDS IN NONLINEAR DYNAMICS

Adapting to ComplexityRussell Ruthen, staÝ writer

Quantum DotsMark A. Reed

By shrinking semiconductor devices to a billionth of a meter, nanotechnologistsare able to conÞne electrons to a mathematical point. These quantum dots haveopened a new realm of physics and chemistry. They may Þnd important electron-ic and optical applications, including computers of unprecedented power.

Scientific American (ISSN 0036-8733), published monthly by Scientific American, Inc., 415 Madison Avenue, New York, N.Y. 10017-1111. Copyright © 1993 by Scientific American, Inc. All rightsreserved. Printed in the U.S.A. No part of this issue may be reproduced by any mechanical, photographic or electronic process, or in the form of a phonographic recording, nor may it be storedin a retrieval system, transmitted or otherwise copied for public or private use without written permission of the publisher. Second-class postage paid at New York, N.Y., and at additional mail-ing offices. Authorized as second-class mail by the Post Office Department, Ottawa, Canada, and for payment of postage in cash. Canadian GST No. R 127387652. Subscription rates: one year $36 (outside U.S. and possessions add $11 per year for postage). Subscription inquiries: U.S. and Canada 800-333-1199; other 515-247-7631. Postmaster : Send address changes to Scien-tific American, Box 3187, Harlan, Iowa 51537. Reprints available: write Reprint Department, Scientific American, Inc., 415 Madison Avenue, New York, N.Y. 10017-1111, or fax: (212) 355-0408.

Copyright 1993 Scientific American, Inc.Copyright 1993 Scientific American, Inc.

¨

Established 1845

THE COVER painting depicts selective adhesion between two cells. This attach-ment is mediated by the carbohydratesin a branching molecule (pink ) that ex-tends from an endothelial cell. A comple-mentary molecule on a lymphocyte calledan L-selectin (blue ) binds speciÞcally to asubunit in the carbohydrate, thereby tether-ing the cells together. Carbohydrates deter-mine many interactions between cells, in-cluding infection (see ÒCarbohydrates inCell Recognition,Ó by Nathan Sharon andHalina Lis, page 82).

Page Source

64Ð65 Larry Lipsky/BruceColeman, Inc.

66Ð67 Joe LeMonnier (top),Jana Brenning (bottom)

68 Jana Brenning

69 Alan E. Strong,U.S. Naval Academy

70 Barbara E. Brown

72Ð73 Alfred Kamajian

74 Michael J. Bolte,Lick Observatory;Johnny Johnson

75 Johnny Johnson (top),Canada-France-HawaiiTelescope Corporation(bottom)

76 Sidney van den Bergh(top), Johnny Johnson(bottom)

77 Alfred Kamajian(top), Anglo-AustralianTelescope Board (bottom)

78 Anglo-AustralianTelescope Board

82Ð83 Tomo Narashima

84 Jared Schneidman/JSD

85 Courtesy of KazuhikoFujita, JuntendoUniversity Schoolof Medicine, Tokyo

86 Kathleen Katims/JSD

87 Courtesy of Steven Rosen,University of California,San Francisco

88 Kathleen Katims/JSD(top), courtesy of KazuhikoFujita, Juntendo UniversitySchool of Medicine,Tokyo (bottom)

89 Photo Researchers, Inc.

90Ð91 Wayne Fields

92 Johnny Johnson

93 Wayne Fields (left ),Samuel A. Bowring (right )

Page Source

94Ð96 Ian Worpole

110 Joe LeMonnier

111 David Haring, DukeUniversity Primate Center

112 Frans Lanting/MindenPictures (top and middle),David Haring (bottom)

113 David Haring (top leftand bottom), Frans Lanting(top right and middle)

114Ð115 Patricia J. Wynne

116 Joe LeMonnier

117 Frans Lanting(left and right )

119 Robert Prochnow

120Ð121 Michael Goodman

122 Mark A. Reed

123 Daniel E. Prober,Yale University

124Ð125 Courtesy of PeterM. Milner

126 Gabor Kiss

127 Eric Mose

128Ð129 Courtesy of SamuelM. Feldman

130Ð131 left to right : Robert A.Blanchette, Universityof Minnesota; Todd A.Burnes and Robert A.Blanchette, Universityof Minnesota; M. P. Kahl,Bruce Coleman, Inc.; R. J.Erwin, Photo Researchers,Inc.; Murrae Haynes

132Ð133 Ian Worpole

134 Jason Goltz

135 Murrae Haynes

138 Andrew Freeberg

140 Murrae Haynes

150 Andrew Christie

151Ð152 Johnny Johnson

THE ILLUSTRATIONSCover painting by Tomo Narashima

EDITOR: Jonathan Piel

BOARD OF EDITORS: Alan Hall , Executive Editor;Michelle Press, Managing Editor; Timothy M.Beardsley; Elizabeth Corcoran; Marguerite Hol-loway ; John Horgan, Senior Writer; Philip Morri-son, Book Editor; Corey S. Powell; John Rennie;Philip E. Ross; Ricki L. Rusting; Russell Ruthen;Gary Stix; Paul Wallich; Philip M. Yam

ART : Joan Starwood, Art Director; Edward Bell,Art Director, Graphics Systems; Jessie Nathans,Associate Art Director; Nisa Geller, PhotographyEditor; Johnny Johnson

COPY: Maria-Christina Keller, Copy Chief; NancyL. Freireich; Molly K. Frances; Daniel C. SchlenoÝ

PRODUCTION: Richard Sasso, Vice President Pro-duction; Managers : Carol Albert, Print Produc-tion; Tanya DeSilva , Prepress; Carol Hansen,Composition; Madelyn Keyes, Systems; Leo J.Petruzzi , Manufacturing & Makeup; WilliamSherman, Advertising Production; Carl Cherebin

CIRCULATION: Lorraine Leib Terlecki, CirculationDirector; Cary Zel, Circulation Manager; RosaDavis, FulÞllment Manager; Katherine Robold,Newsstand Manager

ADVERTISING: Robert F. Gregory, Advertising Di-rector. OFFICES: NEW YORK: Meryle Lowenthal,New York Advertising Manager; William Buchan-an, Manager, Corporate Advertising; Peter Fisch,Elizabeth Ryan. Michelle Larsen, Director, NewBusiness Development. CHICAGO: 333 N. Michi-gan Avenue, Chicago, IL 60601; Patrick Bachler,Advertising Manager. DETROIT: 3000 Town Cen-ter, Suite 1435, SouthÞeld, MI 48075; Edward A.Bartley, Detroit Manager; William F. Moore. WESTCOAST: 1554 S. Sepulveda Blvd., Suite 212, LosAngeles, CA 90025; Kate Dobson, AdvertisingManager; Lisa K. Carden, Lianne Bloomer, SanFrancisco. CANADA: Fenn Company, Inc. DAL-LAS: GriÛth Group

MARKETING SERVICES: Laura Salant, MarketingDirector; Diane Schube, Promotion Manager;Mary Sadlier, Research Manager; Ethel D. Little,Advertising Coordinator

INTERNATIONAL: EUROPE: GWP International,D�sseldorf; Roy Edwards, GWP International,London; Kurt B�hlmann, GWP International, Z�-rich; Vivienne Davidson, Linda Kaufman, Inter-media Ltd., Paris. SEOUL: Biscom, Inc. TOKYO:Nikkei International Ltd. SPECIAL PROJECTS: BarthDavid Schwartz, Director

ADMINISTRATION: John J. Moeling, Jr., Publisher;Marie M. Beaumonte, Business Manager

SCIENTIFIC AMERICAN, INC.415 Madison AvenueNew York, NY 10017

(212) 754-0550

PRESIDENT AND CHIEF EXECUTIVE OFFICER :John J. Hanley

CHAIRMAN OF THE BOARD:Dr. Pierre Gerckens

CHAIRMAN EMERITUS : Gerard Piel

CORPORATE OFFICERS: Executive Vice President

and Chief Financial OÛcer, R. Vincent Barger;Senior Vice President, Linda Chaput; Vice Presi-

dents: Jonathan Piel, John J. Moeling, Jr.

10 SCIENTIFIC AMERICAN January 1993 Copyright 1992 Scientific American, Inc.

Thinking about Mind

In your special issue on ÒMind andBrainÓ [SCIENTIFIC AMERICAN, Septem-ber 1992], each article provided morefood for thought than my neural net-works could process in a score of read-ings. A veritable mental feast!

Indeed, if the issue was a banquet,then Jonathan MillerÕs stimulating es-say ÒTrouble in MindÓ was a Þne bran-dy at the end of a good meal. To me,Miller has always embodied the ques-tioning mind turned inward on itself.

THOMAS SALES

Somerset, N.J.

The splendid ÒMind and BrainÓ issueseems to end on an unduly negativenote. Miller forecasts that we will nev-er fully understand the connection be-tween brain and consciousness. Thatassumption appears to overlook thatconsciousness is routinely interruptedby general anesthetics. The loss of con-sciousness under anesthesia and thelater recovery of it can, in principle,surely be elucidated as thoroughly asany other drug-induced changes.

B. RAYMOND FINK

Department of AnesthesiologyUniversity of Washington School

of Medicine

Sex on the Brain

In her otherwise well-balanced review[ÒSex DiÝerences in the Brain,Ó SCIEN-TIFIC AMERICAN, September 1992], Dor-een Kimura perpetuates some long-standing myths. The traditional viewthat the female pattern of neural orga-nization occurs by default from the lackof exposure to masculinizing levels oftestosterone and estradiol should Þnal-ly be put to rest.

Considerable evidence has accumulat-ed during the past 20 years that femi-nization of neural structure and func-tion is an active process. Numerousstudies in rodents have demonstratedthat feminization depends on levels of estrogen that are too low to elicitmasculinization. My own studies haveshown that in the absence of testos-terone, the removal of endogenous es-trogen dramatically reduces the out-growth of neuronal processes.

The dogma that estrogen-bindingplasma proteins, such as alpha-fetopro-tein (AFP), ÒprotectÓ the female brainfrom masculinization is erroneous. Inrodents, AFP is far more likely to act asa reservoir for estrogen, which may beused to initiate the growth of axonsand dendrites. Estrogen may thereforeregulate sexual diÝerentiation in bothmale and female brains.

C. DOMINIQUE TORAN-ALLERAND

Department of Anatomyand Cell Biology

Columbia University

Kimura contends that many of theskill diÝerences between men and wom-en are mediated by brain organization.Yet two of her examples can be ex-plained by simple physical distinctions.Some experiments have shown that per-formance diÝerences that favor wom-en in pegboard tasks disappear whenthe larger Þnger size of a man is fac-tored out.

Men are reported to be better thanwomen at dart throwing and other tar-get-directed motor skills. It has beenconsistently demonstrated that bothtiming and spatial errors decrease inballistic motor tasks as force approach-es maximum. The greater strength ofmen should grant them an advantagein such tasks. Perhaps sex diÝerencesin ballistic motor tasks found in prepu-bertal children, where strength is simi-lar between the sexes, are inßuencedby socialization.

JOHN S. RAGLIN

Department of KinesiologyIndiana University

Kimura replies:Toran-Allerand makes a valid point,

which for brevity I had to omit frommy article. Moreover, it is still problem-atic whether the evidence for such aninßuence on the organization of repro-ductive behavior is compelling.

Raglin suggests that the sex diÝerenc-es in motor behavior are reducible tophysical diÝerences. Even if the physi-cal diÝerences were decisive, one wouldexpect neural parallels to them. Thestrength diÝerences among three-year-old children are minimal, yet boys haveshown superior accuracy in a targetingtask. Other data also demonstrate per-

formance diÝerences between the sexesand between homosexual and hetero-sexual men that cannot be attributed todiÝerences in size. It seems reasonableto conclude that over and above con-siderations of size, speed and strength,womenÕs brains are endowed with bet-ter digital control and that menÕs brainsare better endowed for targeting exter-nal stimuli.

Genes on the Menu

The Þrst half of Deborah EricksonÕsarticle ÒHot PotatoÓ [ÒScience and Busi-ness,Ó SCIENTIFIC AMERICAN, September1992], about new biotech-derived food,is overly negative.

Yeasts have been used to brew beerfor 8,000 years, and farmers were cross-breeding livestock long before GregorMendel and his experiments. For de-cades, genes have been transferred fromone species to another and even fromone genus to another. These Ògenetical-ly engineeredÓ plants are the very sameoats, rice, currants, potatoes, tomatoes,wheat and corn that we now buy at the local supermarket or farm stand. The techniques of Ònew biotechnologyÓspeed up the process and target withgreater precision the kinds of geneticimprovement we have long conductedwith other methods.

Contrary to the assertions of the neo-Luddites, the recently announced policyof the Food and Drug Administrationfor the regulation of new plant variet-ies is based on solid scientiÞc princi-ples. The bottom line is that the FDAwill not tolerate unsafe foods, and ourpolicy reßects this commitment.

HENRY I. MILLER

Director, OÛce of BiotechnologyFood and Drug Administration

Because of the volume of mail, letters

to the editor cannot be acknowledged.Letters selected for publication may be

edited for length and clarity.

LETTERS TO THE EDITORS

12 SCIENTIFIC AMERICAN January 1993

ERRATUMThe graph on page 122 of the Novem-

ber 1992 issue illustrates the budget of Sematech between 1988 (not 1982)and 1992.


JANUARY 1943

ÒFormerly, if an enemy submarine layquietly on the bottom of the sea to avoiddetection, the business of Ôputting theÞngerÕ on a sub became more diÛcultand less accurate in its results. In thepresent conflict, the principle of soundreßection under water, long applied tolarger merchant and war ships to main-tain a continuous graphical record ofthe oceanÕs ßoor beneath the cruisingship, is being adapted to search out si-lent submersibles that endeavor to ÔplaypossumÕ far beneath the waves. The ex-act extent to which echo-sounding de-vices are utilized and their scientiÞcand mechanical constituency are amongthose things which cannot now be told.Ó

ÒIn a degenerate mass of gas, when thevelocities of the moving electrons beginto become comparable with that of light,the law connecting pressure and density

changes. Chandrasekhar has shown that,when this is taken into account, a starof small mass (less than twice the SunÕs)will settle down into a permanent statewith a degenerate core, as a white dwarf,and Þnally as a Ôblack dwarf,Õ cold onthe surface; but a large mass (ten timesthe SunÕs or more) should continue tocontract without limit. It is natural tosuppose that something would ultimate-ly happen to end this process, and itmay well be that the contracting starblows up, ejects enough matter to leavea residue small enough to form a de-generate core, and then develops suc-cessively into a blue, a white, and a blackdwarf. At the Paris Conference of 1939,Chandrasekhar suggested that some cat-astrophic change of this sort might beresponsible for a super-nova.Ó

ÒThe requirements for carotene (pro-vitamin A), ascorbic acid (vitamin C),and iron can readily be met by eating

moderate quantities of dried grass. Inthe case of calcium and the vitamin Bcomplex factors, between four and sixounces need be eaten, amounts so largeas to be undertaken only by an enthusi-ast. Undoubtedly the wisest and safestrecommendation is to use dried grass, if at all, in small amounts and Þnelyground, either as an added ingredientin common foods such as bread, or asa supplement to the diet in the form oftablets, which should be prescribed onlyon advice of a physician.Ó

JANUARY 1893

ÒTo set oÝ this piece of Þreworks it isnot necessary to be a pyrotechnist. Pro-vide yourself simply with a blowpipe oreven a clay tobacco pipe. Take a fewsheets of thin tinfoil, such as is used asa wrapping for chocolate, and cut theminto strips of a width of about an inch.Then present each slip to the ßame ofthe blowpipe, when the metal will igniteand fall in incandescent globules, whichwill rebound and run over the table onwhich you operate and travel a con-siderable distance. When the ßame isstrong and the tinfoil burns briskly, theglobules are very abundant and thenpresent the aspect of a bouquet of Þre-works in miniature. By such combina-tion of a metal with the oxygen of theair, the tinfoil is converted into a whiteoxide. It was by studying the increasein weight exhibited by tin heated incontact with air that John Rey, a chemistof the seventeenth century, succeededin understanding the Þxation of the airupon metals.ÑLa Nature.Ó

ÒThe way in which the natural kanga-roo spars in the bush, his birthplace, ispeculiar. He places his front paws gent-lyÑalmost lovinglyÑupon the shoul-ders of his antagonist, and then pro-ceeds to disembowel him with a suddenand energetic movement of one of hishind feet. From this ingenious methodof practicing the noble art of self-de-fense the kangaroo at the Royal Aquar-ium has been weaned. The clever in-structor of this ingenious marsupial hastrained it to conduct a contest underthe conditions known as the Marquisof QueensberryÕs rules.Ó

50 AND 100 YEARS AGO


The kangaroo as a prizefighter


When the idea of mapping andsequencing all the genes thatmake up a human being was

Þrst proposed, it seemed an undertak-ing tantamount to putting a man onthe moon. The massive internationaleÝort was expected by some to contin-ue for 15 years or more. But after onlytwo years, the Human Genome Projectis proceeding more rapidly than mostbiologists had dared predict. ÒWe aretwo or three years ahead of schedule,Ósays Daniel Cohen of the Center for theStudy of Human Polymorphism (CEPH)in Paris. ÒI believe it will be possible tohave a very good map of the genomeby the end of 1993. Probably the se-quence of the genome will be Þnishedby the end of the century.Ó

Sketchy though they are, the latestgenetic road maps are already obliginggeneticists to reappraise their theoriesabout the functions of some humanchromosomes. Meanwhile parallel workon simpler organisms, such as the muchstudied roundworm Caenorhabditis el-

egans, is revealing that they have unex-pectedly large numbers of genes. As aresult, some investigators are speculat-ing that the human genome may turnout to be far larger than the 100,000 orso genes it is believed to contain.

Norton D. Zinder of the RockefellerUniversity, a former co-leader of theproject who now advises the NationalInstitutes of Health on its eÝort, be-lieves many of the recent discoveriescould not have been made without acomprehensive gene-sequencing eÝort.ÒThere are real data coming in, and itproves that we are going in the direc-tion we should be,Ó he says.

The genome project involves devel-oping three increasingly detailed mapsof the DNA in cells. The Þrst is a genet-ic linkage map, which shows the rela-tive distances between markers on achromosome. The second is a physicalmap, which locates similar genetic land-marks but speciÞes the actual numberof nucleotide bases, or DNA subunits,between them. The ultimate map is theordered sequence of bases in a chro-mosome that describes the genes andthe proteins they make.

In early October, through a colossalcombined eÝort by the NIH and CEPH,genetic linkage maps for 23 of the 24 types of human chromosomes werecompiled and published. Simultaneous-ly, physical maps for two of the chromo-somes were released: chromosome 21,which was mapped by Cohen and hiscolleagues, and the Y chromosomeÑforwhich there was not a linkage mapÑbyDavid C. Page, Simon Foote, DouglasVollrath and Adrienne Hilton of theWhitehead Institute at the Massachu-setts Institute of Technology.

Cohen and Page both used essential-ly the same techniques to map the chromosomes. Through a process calledsequence-tagged site mapping, they es-tablished the order of small marker sequences on the chromosomes. Theythen chopped the chromosomes intopieces of DNA about a million bases longand spliced the pieces into yeast DNA toproduce artiÞcial chromosomes, whichcould be measured conveniently. By

looking for the markers on the artiÞcialchromosomes, the researchers deducedhow to Þt them together, like pieces ofa puzzle. Some segments of the humanchromosomes are missing from thesemaps, but they are not believed to con-tain any genes.

Because chromosome 21 has been as-sociated with DownÕs syndrome, someforms of AlzheimerÕs disease and otherdisorders, the clearer picture of its ge-netic contents is expected to have greatmedical relevance. In the short run, how-ever, the Y chromosome may beneÞtmost from the new map because it isthe least typical of the human chromo-somes and in many ways the least un-derstood. ÒWeÕre trying to make thischromosome respectable,Ó Page says.

The Y chromosome, according to Page,has often been regarded as Òbasically ajunkyardÓ containing no more than afew genes related to spermatogenesisand other functions peculiar to males.ÒMany people refer to the Y as a male-

How Many Genes and YGene mappers Þnd plenty, even in ÒjunkÓ chromosomes

SCIENCE AND THE CITIZEN

DRAWING A MAP of the Y chromosome was the task undertaken by Adrienne Hiltonand her molecular geneticist colleagues at the Whitehead Institute at M.I.T.


STA

NLE

Y R

OW

IN


ness chromosome,Ó he says. ÒI think thatis much too narrow a cubbyhole to Þtthis chromosome into.Ó

One piece of evidence on his side isthe discovery by his mapping team that25 percent of the studied Y regions arehomologous, or highly similar, to partsof the X chromosome. On the X, severalgenes essential to both sexes are foundin these areas. Other studies have alsofound similarities in gene sequence onthe two chromosomes. ÒIÕm sure thatÕsjust the tip of the iceberg,Ó Page addsenthusiastically.

One important implication of thosesimilarities is that a classic tenet of ge-neticsÑthat males have only one copyof all the genes on the XÑis wrong. Con-sequently, Page argues, the work on theY chromosome sequence Òforces us to

rethink not only the functions of the Ybut also of the X.Ó How important thegenes on the Y chromosome are remainsto be seen, but Page contends that Òhis-tory is on the side of people predictingan ever widening array of functions forthis chromosome.Ó

The Þnal stage of the project, sequenc-ing the individual genes, has not yetbegun. But related eÝorts in other spe-cies are well under way. In the spring of 1992 Robert Waterston of the Washing-ton University School of Medicine andJohn Sulston of the Medical ResearchCouncil in Cambridge, England, andtheir colleagues published the sequenceof more than 120,000 bases in the DNAof C. elegans. That represented only atenth of a percent of the total genome,but the pace of sequencing is accelerat-

ing: Waterston reports that they havenow sequenced about one million basesand expect to Þnish another two millionbases within a year.

Meanwhile a European consortium of145 scientists has been sequencing chro-mosomes of the common yeast Saccha-

romyces cerevisiae. Last May the grouppublished the complete sequence ofchromosome III. According to StephenG. Oliver of the University of Manches-ter Institute for Science and Technol-ogy, who served as DNA coordinatoron the project, yeast chromosome XI isnow about two thirds Þnished, and chro-mosome II is about half done; extensivework has also been done on chromo-somes I and VI.

Perhaps the most surprising observa-tion about the newly sequenced genes is

SCIENTIFIC AMERICAN January 1993 17

Endangered Genesan you name the male and female leads of the Hu-man Genome Project? They star in Gray’s Anatomyand have white skin, urban homes and composite

ancestry. Still can’t place them? They are John and Jane Doe.So much for ethnic diversity. The ethnocentric bias of

the genome project has riled an international group of an-thropologists who hope a more extensive catalogue of hu-man genes will allow them to reconstruct human evolu-tion. For the past two years, they have been planning aparallel initiative called the Human Genome Diversity Proj-ect. Their goal is to sample the genes of aboriginal peo-ples before these peoples die out or assimilate.

A quick survey of the most endangered groups shouldtake about five years and cost about $23 million, says Lu-igi L. Cavalli-Sforza of Stanford University. Those who mapthe genes of John and Jane Doe will never miss that paltrysum, although they may gain a substantial return on theinvestment. If the sample turns up genetic adaptations todisease, for example, workers may use the knowledge todevelop new therapies.

The project began two years ago, when five geneticistspublished a manifesto challenging the ethnocentricity ofthe genome project in the journal Genomics. Others quick-ly jumped on the bandwagon because two of the authorscommanded such respect in the field’s main camps: thosewho study populations and those who study individu-al gene lineages. The first approach was championed by Cavalli-Sforza, the second by the late Allan C. Wilson ofthe University of California at Berkeley.

Proponents of the two approaches worked out their differences at a workshop held last summer at Stanford. Cavalli-Sforza argued for intensive sampling, the only wayto get the statistical depth he needs to look at gene fre-quencies in different populations. But to obtain enoughspecimens in each sample, Cavalli-Sforza conceded, hewould have to make do with relatively few samples. Hetherefore wanted to study ethnic groups whose linguisticdistinctiveness suggests they are of ancient descent.

Wilson’s disciples favored a more extensive survey. Be-cause they study the lineages of individual genes, theycould broaden the coverage at the expense of sample size.In their most controversial work, they surveyed a few hun-

dred individuals to build a genealogy tracing all humansto an African matriarch who lived some 200,000 years ago.

The two schools clashed on a practical matter as well.Cavalli-Sforza’s group wanted to preserve specimens byimmortalizing cells, a procedure that requires rushing freshblood to the laboratory before the white cells die. Wilsonwanted to facilitate a broad survey by letting ethnograph-ers put the blood on ice, so that they could go on collect-ing for weeks. They could then deposit their trove in re-positories from which future generations could draw re-peatedly, using the new techniques of DNA amplification.

The workshop compromised: ethnographers would con-centrate on distinct ethnic groups, as Cavalli-Sforza want-ed, but they would spread their resources over a great-er number of groups, as Wilson’s team wanted. They alsoagreed to immortalize only a fraction of the specimens.They projected a sample of about 400 groups.

A second workshop chose the groups at PennsylvaniaState University over the Halloween weekend. Anthropol-ogists, linguists and geneticists divided into teams spe-cializing in each region save Europe, which has its ownproject under way. Eyes glazed as specialists struggled tofill out forms assigning priorities to tribes and pointingout problems ethnographers might face. Watch out for guer-rillas and coca smugglers, said the South America group.Survey the hundreds of Polynesian populations at a fewcentral labor exchanges, suggested the Pacific group. Re-fuse to report HIV-positive cases to governments on groundsof medical confidentiality, counseled the Africa group.

All were concerned about the language they—and re-porters—might use to describe their work. “You can talk of‘tribes’ in Africa but not in this country,” said one parti-cipant. Others worried that labeling a group as “endan-gered” would offend the majority group in their country.

The third workshop, to be held in Washington early thisyear, and the fourth, to be held in Sardinia next fall, willdiscuss the logistics of reaching all points on the globe,the techniques for collecting and analyzing materials andthe ethical problems in exploiting native peoples for theirgenes. Some groups find anthropomorphic sampling sorepugnant that they refuse access to the dead as well asto the living. —Philip E. Ross

C


The pitohui sounds like what it is:something to spit out. The skin,feathers and some organs of this

orange-and-black New Guinean birdcontain a potent poison. Although oth-er speciesÑincluding certain snakes,insects and frogsÑwere known to pro-duce toxins as deterrents, it was gener-ally thought that birds did not. So thediscovery of the same device in the pit-ohui has ruÜed some notions of aviandefensive strategies and coloration.

The pitohuiÕs defense mechanism wasnoticed two years ago by John P. Dum-bacher, a graduate student in ecologyand evolution at the University of Chi-cago. He felt numbness and burning in his mouth when he licked his handsafter handling the hooded pitohui, re-ferred to in New Guinea as a ÒrubbishÓbird because of the taste of its skin.Dumbacher and his colleagues recent-ly reported in Science that three spe-cies of the genus PitohuiÑthe hooded,the variable and the rustyÑproduce anoxious chemical, which they identiÞedin 1992.

The poison, homobatrachotoxin, turnsout to be identical to that of a SouthAmerican poison-dart frog, which alsohas aposematic, or warning, coloring of orange and black. ÒI was very sur-prised,Ó says John W. Daly, a chemist at the National Institutes of Health who analyzed the frog toxin in the 1970s and that of the pitohuis last year. ÒThere certainly has been a specif-ic evolutionary ability to accumulate thistoxin. I would like to say ÔmakeÕ it, but

we do not know if it is from the diet.ÓAlthough the pitohui is the Þrst poi-

sonous bird to be reported in the litera-ture, there have been anecdotal reportsof bad-tasting birds. Some experts an-ticipated the Þnding. ÒI am not at allsurprised,Ó comments Lincoln P. Brow-er, an ornithologist at the University ofFlorida, Òespecially given the fact thatsome insects are poisonous and thatbirds behave like those insects: they areconspicuous and brightly colored.Ó

Conventional ornithological wisdomholds that bright plumage among birdsexists to facilitate courtship and mat-ing. But the colorful feathers of the pit-ohui could serve as a warning to preda-tors. The fact that male and female pit-ohuis share the same palette reinforcesthis assumption. And just as there arenonpoisonous mimics of the poison-ous monarch butterßy, there are non-poisonous mimics of the pitohui.

Taste tests of birds conducted in the1940s and 1950s support the possibili-ty that color and palatableness are in-versely linked. Hugh B. Cott, a zoolo-gist at the University of Cambridge, ob-served that hornets in Africa avoidedcertain bird carcasses yet ate others.Cott then did the Òhornet testÓ on a se-ries of birds. The results encouragedhim to conduct his own gourmet, dou-ble-blind trials. To carry them out, a col-leagueÕs wife prepared repasts of 200bird species. After CottÕs feasts, dinersagreed with what might be called CottÕsrule: the blander the bird looks, thebetter it tastes. Birds that had crypticcoloringÑthat is, those that blended inwith the backgroundÑtasted best.

ÒConversely, there is some evidencethat numbers of highly conspicuousbirds belonging to many diÝerent or-ders are deÞnitely unÞt for the table:


their staggering number. If the sequenceanalyzed by Sulston and Waterston isrepresentative, C. elegans may have 15,-000 genesÑthree times more than wasonce believed. Researchers had thoughtyeast chromosome III contained onlyabout 34 genes, but the Europeans foundevidence for 182.

Most of the genes seem to have es-caped detection previously because mu-tations in them did not have noticeableeÝects. Some biologists have thereforespeculated that many of the genes areredundant or unnecessary. That notionhas its critics, however. As Page asserts,ÒWe donÕt have any idea of how manygenes it ought to take to perform func-tions.Ó He points out that nobody has yetshown what happens if combinations ofthese seemingly redundant genes areknocked out. ÒHow deep is the redun-dancy?Ó he asks.

Oliver suggests that the seemingly re-dundant genes may be important onlyduring brief periods of an organismÕslifeÑand possibly not at all under stan-dard laboratory conditions. ÒOne mayneed to make the organism jump throughrather speciÞc physiological hoops be-fore only gene X and not gene X ′ willwork,Ó he says.

When researchers sequence the hu-man genome, it is diÛcult to predictwhether they will Þnd more than the100,000 genes they now expect. Zinder,for one, thinks they may. Cohen, who be-lieves the current Þgure is roughly cor-rect, says it is far harder to recognizegenes in humans because human genesare more extensively subdivided andseparated than are those of yeast androundworms. Page maintains that geneestimates in all organisms have beencreeping up for years. Just a few yearsago, he notes, most geneticists claimedthat humans had only 30,000 genes.

Investigators are also discovering thatmany of the genes sequenced so far inC. elegans and yeast are extremely simi-lar to ones found in other organisms,from mammals to bacteria. Waterstonbelieves the strong similarities betweenroundworm enzymes and mammalianenzymes show they are serving almostthe same function. Nevertheless, he adds,Òwhether theyÕre working on the samesubstrate or not is another matter.Ó

Some of the shared genes are incon-gruous: yeast, for example, carries a genefor a protein that enables bacteria to Þxnitrogen into biological compounds, eventhough yeast does not have that ability.To Oliver, the presence of that gene inyeast suggests Òwe donÕt understand inany deep way the function of the pro-tein in the nitrogen-Þxing bacteria.Ó It islikely, he thinks, that all organisms usethat gene somehow; its application to

the Þxation of nitrogen is just particu-larly noticeable.

Such a gene may therefore turn up inhumans as well, once the sequencers areready. That time may come soon, be-cause the mapping stage may not lastmuch longer. While Page and others arecontinuing to make physical maps ofindividual chromosomes, Cohen is bold-ly pursuing a complementary approach:he is mapping all the chromosomes atonce. ÒWe believe the approach used forchromosomes 21 and Y is far too tedi-ous and expensive,Ó he explains.

Instead of using the sequence-taggedsite markers, CohenÕs group is usingolder DNA ÒÞngerprintingÓ technologyto ßag distinctive sequences on all thechromosomes simultaneously. Cohenreports that they have currently mapped

about 70 percent of the entire humangenome this way, and he expects to havethe complete genome mapped at lowresolution by this February. That physi-cal map will have fairly little detail butcan serve as a Òbackbone,Ó Cohen says,for further physical maps based on thenew technology. ÒWith the two, you geta synergy,Ó he remarks.

That synergy should greatly speedup the mapping. But because no one hasexperience with sequencing large chunksof human DNA, Page is more cautiousthan Cohen about projecting an end tothe genome project by 2001. ÒI donÕtthink thatÕs unreasonable, but mostly IdonÕt think itÕs unreasonable becauseitÕs still eight years awayÑI mean, ineight years, we could do almost any-thing, right?Ó ÑJohn Rennie

Pitohui!The colorful bird looks better than it tastes


It began with a method for keepingspies honest and may end up veri-fying the notorious four-color map

theorem. The technique, known as aholographic proof, makes it possible to achieve a high degree of conÞdencethat a set of logical assertions (such asa theorem and the reasoning involvedin its proof) is internally consistent bychecking only a tiny fraction of thesetÕs statements.

Testing a mere 300 lines of a 100,000-line proof could reduce the probabil-ity of an undetected error to less than one divided by the number of particles in the universe, asserts mathematicianLeonid Levin of Boston University. Somemathematical proofs already run up-ward of 10,000 pages, and no one canpossibly comprehend them in their en-tirety, much less certify their reasoning.

Furthermore, the same techniquecould in theory be used to check the out-put of complex computer programs. Rig-

orous proof that a program does whatits designers intended is infeasible bymeans of conventional methods unlessthe program is only a few pages long.In addition, even if the software is cor-rect, there is no guarantee that the hard-ware has not suÝered a random glitch.A holographic check would test the pro-gram and its execution simultaneously.

The Òzero knowledgeÓ proof, whichhelped to set the stage for the holo-graphic variety, was developed as partof cryptographic protocols for verify-ing facts without revealing them. Cryp-tographers have shown that one cananswer a series of random mathemat-ical queries about some hidden fact insuch a way that its secret remains hid-den, but anyone who does not actuallyhave the knowledge has only an inÞni-tesimal chance of answering the quer-ies consistently.

In a holographic proof, instead of an-swering a series of random queries, theprover in eÝect writes down answers toall possible queries in a book, and theveriÞer samples the answers randomly,looking for inconsistencies, says LanceFortnow, a mathematician at the Univer-sity of Chicago. The eÝect is the same.

The key to the trick is a techniquecalled arithmetization. According toFortnow, once a proof has been stat-ed in strict logical form, one turns itinto a polynomial expression of manyvariables by (more or less) substitutingaddition operations for every ÒorÓ inthe proof and multiplication for everyÒand.Ó The holographic proof then con-sists of a series of equations givingboth the polynomial and its value fordiÝerent combinations of the values ofits variables.

Checking is simply a matter of mak-ing sure the calculated value of the poly-nomial at any point matches that as-serted in the proof. Only a small num-ber of points need be checked, Levinexplains, because it is very diÛcult toconstruct two polynomials of low de-gree that are equal at some points yetdiÝerent at others.

Indeed, for a single-variable polyno-mial of degree 10 (of the form Ax10 +Bx9 + Cx8. . .), a mere 11 values are suf-Þcient to specify its shape precisely. Aslong as the proof does not contain toomany logical ÒandÓ statements strung to-gether, its polynomial degree will be low.The number of tests required to check


sheld-duck, crocodile bird, magpie, andswallows being examples,Ó Cott wrote inNature in 1945. Nevertheless, Jared M.Diamond of the University of Californiaat Los Angeles recently reported in Na-

ture that his Þeld assistantsÕ meal ofpitohui produced no Òuntoward eÝects.Ó

Gustatory recommendations aside,the issue of coloration may be twofold:vivid markings could be selected for by natural or sexual selection, or both.ÒThe two forces are not mutually ex-clusive,Ó Brower suggests. Sexual selec-tion could have favored brightly col-ored males and cryptic females at Þrst.

But if the birds then acquired a toxinfrom their diet and if it successfully pro-tected against predators, natural selec-tion for bright coloration in both sexeswould occur.

In addition, Stanley A. Temple, a wild-life ecologist at the University of Wis-consin, has found a correlation betweendiets rich in fruit and aposematic col-oring in birds. He suggests that suchmeals may allow the birds to sequesterchemicals that could form toxins. (Cottdid not relate birdsÕ diets to their tasti-ness.) Temple says he, too, has found apoisonous bird, the pink pigeon of Mau-

ritius, but has not yet published his re-search. The pigeon apparently derivesa toxic alkaloid from its diet. ÒThe spe-cies is probably in existence today be-cause of its defense mechanism,Ó Tem-ple notes. ÒThe dodo and others wereexterminated.Ó

Dumbacher and Daly and their col-leagues plan to study the pitohuiÕs poten-tial predators to see if they are repelledby the toxin. They will also examine pit-ohuis to determine how they avoid poi-soning themselves. Homobatrachotox-in works by opening up ion channels,causing cells to be infused with sodium.But Òthere are a number of creaturesthat are resistant to their own poison,ÓDaly says. In such species, ion channelsdo not respond to their toxin. Daly ex-pects to see the same kind of mecha-nism in pitohuis.

Daly also hopes to determine if pito-huis can synthesize the chemical or iftheir metabolism produces it as a by-product. Poison-dart frogs living in cap-tivity do not make the stuÝ of theirdeadly bolus, apparently because theylack something that their rain-forestdiet of leaf litter normally supplies.

The search is also on for more poi-soned plumage. Although Dumbachersays he will conÞne himself to pitohuisfor now, he does admit a numbing, burn-ing curiosity: ÒIÕve thought of lickingother birds.Ó ÑMarguerite Holloway

Crunching EpsilonCryptography may be the keyto checking enormous proofs

POISON PLUMAGE recently discovered in the hooded pitohui suggests that otherbrightly colored birds may also use toxins to repel predators.

W. P

EC

KO

VE

R V

IRE

O


it will be trivial compared with the sizeof the proof, Levin says.

Unfortunately, one reason the num-ber of tests will be so small is that themethod causes the proof itself to be-come enormous. First, instead of theshorthand mathematicians usually use,Levin notes, every step and every ruleof inference must be rigorously de-fined. Mario Szegedy of AT&T Bell Lab-oratories estimated it might take 500lines just to prove the Pythagorean the-orem, in contrast to the dozen or sonow considered adequate.

Worse yet, the arithmetized version ofthe proof will be even longer. If the for-mal version contains N lines, the arith-metized one will contain K times N

raised to the power of one plus epsilon.K is very large, and the various math-ematicians disagree on the size of ep-silon. Szegedy pegs it conservativelynear one, in which case the arithmetizedversion of a 10,000-line proof could run

upward of 100 million lines. Fortnowmarks epsilon at one half, Levin nearerone third. Indeed, Levin says, epsilon canbe reduced as close to zero as desiredbut only at the cost of increasing KÑperhaps to a point that would swampany improvements in the exponent.

Bringing K and epsilon down to lev-els that might make the holographictechnique practicalÑeither for mathe-matical proofs or for checking comput-er programsÑwill take Òa lot of hardwork,Ó Fortnow says. Indeed, he sug-gests, it will probably require the inven-tion of one or two mathematical tricksfor doing holographic transformationand perhaps the same number of fun-damental insights. ÒItÕs not clear that itwill be possible,Ó he asserts.

Even in its present unwieldy form,however, the holographic proof tech-nique has its uses. Szegedy and his col-league Carsten Lund and others haveemployed a variant of its principles to

prove lower bounds on the diÛculty ofsolving certain hard problems in com-puter science. And Szegedy is hopingto perform a holographic veriÞcationof the four-color theorem, whose brute-force computer enumeration of all rele-vant maps has troubled certainty-mind-ed mathematicians for most of the twodecades since it was completed. (To thisday, no one knows if there might havebeen some mistake in the algorithm.)

Of course, for purists the idea of aproof that is only probably veriÞed pre-sents its own problem. But Szegedy isnot one of those. ÒPeople are becomingsatisÞed with probabilistic methods,Óhe says, noting that probabilistic tech-niques are also being used to factorlarge prime numbers. ÒYou might be-lieve that mathematics is part of natureor that it is what humans use to theiradvantage. If you take the practical ra-ther than the idealistic approach, yousave a lot of headaches.ÓÑPaul Wallich


hen a greasy burger or a handful of salted peanutssends someone’s blood pressure soaring, theremay be more to the clinical picture than the haz-

ards of gobbling on the run. Research teams in the U.S.and France have found that for some people, a particulargene seems to increase the likelihood of acquiring a formof hypertension—specifically, the kind involved in salt re-tention. Although physicians have known for some timethat heredity plays a strong role in the illness, the gene,one of perhaps several that are thought to be implicated,represents the first direct, supporting evidence.

The studies, conducted by the University of Utah’s How-ard Hughes Medical Center and the French research insti-tute INSERM, looked at hypertensive siblings and com-pared them with unrelated people who had normal bloodpressure. The researchers found that the hypertensive sib-lings (those with blood pressure exceeding 140 over 90)tended to have the same kind of variation in the gene thatencodes a protein called angiotensinogen.

Angiotensinogen works in conjunction with renin, anenzyme produced by the kidneys, to form angiotensin,which raises blood pressure by constricting blood vesselsand, perhaps more important, changing the body’s balanceof sodium and water. Jean-Marc Lalouel, who headed theUtah group, speculates that the gene variation could leadto a small increase in circulating angiotensinogen. By thetime an individual reaches middle age, the overproductionmakes the body sensitive to sodium. The retention of so-dium causes the volume of blood to expand. To compen-sate for the additional fluid fed to the body’s tissues, thearterioles constrict. As a result, blood pressure rises. “Asmall increase in angiotensinogen may act over the longterm,” he says.

Like many discoveries concerning the genetics of dis-ease, the ability to identify the angiotensinogen gene wouldbe useful in screening for susceptible individuals. Thenew finding could be especially relevant to black Ameri-cans, in whom the condition develops two to six times

more frequently than in white Americans, according tovarious estimates. Furthermore, most hypertensive Ameri-can blacks have the salt-sensitive form of the disease. Pre-liminary findings from other studies indicate that such in-dividuals also display the related variations in the angio-tensinogen gene.

The finding could breathe new life into a hypothesis of-fered a few years ago by Clarence Grim of the Charles R.Drew University of Medicine and Science in Los Angeles.Grim has proposed a highly controversial idea of why blackAmericans have high blood pressures. Grim noted that indigenous populations in sub-Saharan Africa have a verylow incidence of hypertension, unlike those in the WesternHemisphere.

Grim used historical accounts of the slave trade to ex-plain the difference. The cause of death for most slaveswas diarrhea, and the sweating, vomiting and lack of drink-ing water on the grueling trans-Atlantic journey in thepoorly ventilated cargo holds contributed to dehydration.Those who could retain water—with body salt—survived.“Death during the slave trade may well have focused onthe ability to conserve salt,” he says.

According to Grim, the death rates, which ranged from30 to 50 percent, were sufficient to select for a gene. Hecites several studies of black populations that, while notexpressly proving the theory, are at least consistent withit. Given a candidate gene, he will look for variations inAfrican and American populations.

Although Lalouel thinks “clearly there is something thathas to do with salt handling in blacks,” he does not buyGrim’s hypothesis. He believes most of the observed bloodpressure differences seen in African and American popula-tions arise from environmental factors. The food of mostAfricans living in aboriginal conditions, he points out, islow in sodium; he predicts that changing their diets to re-semble those of the more industrialized nations wouldraise their blood pressures. Avoiding fast foods is stillgood advice for anyone. —Philip Yam

A Gene for Hypertension

W


Suppose a United Nations peace-keeping force disarms one of twoopposing sides in an internation-

al hot spot and then directs the still bat-tle-ready adversary to annihilate its nowhelpless enemies. Although this may be a draconian solution, a total war inwhich both sides could potentially bewiped out is avoided.

University of Oxford researcher Lau-rence D. Hurst oÝers this metaphor toanswer a question that has intrigued ge-neticists for many decades. The gametesfrom sexually reproducing organismsfuse into a cell that contains one set ofchromosomes from the nucleus of eachparent. Yet there are other parties tothe event: the strands or bundles of ex-tranuclear DNA carried in the mitochon-dria of animals or the chloroplasts andmitochondria of plants. Unless one setof this material is eliminated, the grow-ing daughter cell will be imperiled bywithering exchanges of enzymes andother chemical ordnance as the twosides struggle for dominance.

How is the prospect of conßict to beprevented? Hurst thinks the nuclearDNA plays the role of the U.N. force. Iteliminates one set of contestants. This,Hurst believes, may go a long way to-ward explaining why there are separatesexes and why it is often so diÛcult toÞnd a date on a Saturday night. Hurstcontends that it is easier to keep thecytoplasmic peace when there are onlytwo sexes. Imagine, he says, sorting outsuch a conßict if any pairing amongfour, eight or 13 sexes could produce a new individual.

Hurst is a leading proponent of aschool of biologists who conceive of evo-lution as more than just a competitionto determine which organism adaptsbest to its environment, the classic Dar-winian interpretation. Rather these ad-vocates of what is called intragenomicconßict assert that much of evolution-ary history may be explained by a kindof genetic Hegelian dialectic in whichgroups of genes within an organism en-gage in a constant game of one-upman-ship with each other.

Early advocates of these theories,such as William D. Hamilton, a profes-sor of evolutionary biology at Oxford,believe intracellular conßictsÑand theway they get resolvedÑmay help an-

swer other major evolutionary ques-tions. One of these conundrums maybe the very beginnings of sex itself.

In a recent article in the Proceedings

of the Royal Society of London, Hurstand Hamilton (who was HurstÕs formergraduate adviser at Oxford) focus onthe war zone of the cytoplasm. Where-as the nuclear genes have created acomfortable demarche, the mixing ofcytoplasmic genes when gametes cometogether is potentially a microscopicBosniaÑÒa tragedy of the common cy-toplasm,Ó Hurst calls it.

A war in which mitochondria fromdiÝerent gametes battle to the deathendangers the very existence of thecombined cell, or zygote. Successful sexrequires that the DNA in the nucleusÞnd some means of suppressing con-ßict in order to preserve the organism.The ultimate evolutionary winners arecellular unions in which the nuclearcommand post in one of the two ga-metes issues strict orders on the fateof mitochondria or chloroplasts.

A sperm, for example, sheds its mito-chondria before entering the egg. Mush-rooms avoid potentially lethal cytoplas-mic skirmishes altogether by forgoingcell fusionÑwhich, Hurst says, distin-guishes them from organisms that havedeveloped discrete sexes.

Instead of merging two gametes, theysimply transfer nuclei to one anoth-er through the process of conjugation,creating a small opening between thetwo cells. Thus, they can mate with anyother member of their species, except afew individuals that have incompatibili-ty markers, indicating that they are ge-netically too close.

To provide evidence for the existenceof cytoplasmic dueling, Hurst and Ham-ilton sorted through the literature andidentiÞed a strange miscellany of cili-ates, fungi and slime molds. In Chlamy-

domonas algae, for example, one chloro-plast is inherited from the female andone from the maleÑor more preciselya Ò+Ó and ÒÐÓ mating type, since thisalga has not developed the size diÝer-ences (a small sperm and a large egg)that are characteristic of other sex cells.In the merged cells the two cytoplasmsbegin to attack each other with deadlyenzymes. But the + type, roughly anal-ogous to the female, gets eaten awaymore slowly than does the Ð type (thequasi-male) and so prevails.

HurstÕs prize organism is a primitiveslime mold that a group of Japanese re-searchers reported on in 1987 in theJournal of General Microbiology. Theslime mold, Physarum polycephalum,appears to have 13 sexes, each of whichcan mate, or permanently fuse cells,with any other sex except its own. If hu-

Anything GoesWhy two sexes are better than 13


Almost 60 years ago pioneering cos-

mologist Fritz Zwicky made theshocking claim that much of the

mass of the universe is Òmissing.Ó With-in the past decade, improved observa-tions have transformed ZwickyÕs asser-tion into the accepted wisdom. The rateat which galaxies rotate and the man-ner in which they sail about in clustersand superclusters indicate that as muchas 99 percent of the cosmos consists ofan invisible component, known as darkmatter. Theorists have identiÞed threemajor possible dark components of theuniverse: MACHOs, WIMPs and the moreprosaic neutrinos. A bevy of new exper-iments using computerized telescopes,particle accelerators and neutrino de-tectors may Þnally pin down the truenature of this mysterious matter. AsKim Griest of the University of Califor-nia at Berkeley puts it: ÒDark matterÕstime has come.Ó

One of the most fruitful places tosearch for dark matter is in the outerhalos of galaxies. Studies of disk galax-ies show that their outer regions rotatemuch faster than would be expectedfrom just the visible stars and gas theycontain. Large amounts of unseen mat-ter must be present to create an extragravitational tug.

Many astronomers have speculatedthat the mass in the outer parts of galax-

ies may be hidden in such nonluminousbodies as free-ranging planets, burned-out stars, brown dwarfs (starlike ob-jects too small to shine) and black holes.Griest has whimsically coined the termÒMACHOsÓÑmassive compact halo ob-jectsÑfor this class of dark matter can-didates. Charles Alcock of Lawrence Liv-ermore National Laboratory, workingwith Griest and several others, has re-cently embarked on an ambitious searchfor MACHOs, as have groups of Frenchand Polish astronomers.

MACHOs cannot be perceived direct-ly, but if one were to pass between theearth and a more distant star, its grav-ity would slightly bend and amplify thestarÕs light. Alcock and his collabora-tors are monitoring three million starsin the Magellanic Clouds for telltalesigns of previously unperceived cos-mic vagrants. The rate at which a starÕsbrightness changes would reveal the in-tervening objectÕs mass. A Jupiter-massbody would cause a star to brighten anddim over the course of a few days,whereas events associated with blackholes could last well over a year.

AlcockÕs MACHO investigation willrun for four years, but positive resultscould show up much sooner. ÒIf we ÞndMACHOs, we will have solved the ob-servationally secure dark matter prob-lem,Ó Griest says. If, on the other hand,the various surveys come up empty-handed, astronomers will be forced toconsider some of the more bizarre ex-planations for dark matter.

In fact, most cosmologists have al-ready come to believe that unfamiliar

MACHOs or WIMPs?Astronomers stalk the invisible cosmic majority

LAWRENCE KRAUSS of Yale University has abandoned the telescope in favor of theparticle accelerator to search for the missing mass in the universe.


JAS

ON

GO

LTZ

mans had 13 sexes, and a person couldmate with anyone but oneÕs own sex,there would be no more lonely nights.ÒIt would be gorgeous beyond beliefÑthe Þrst person you meet you couldmate with,Ó Hurst says.

But, Hurst notes, the slime moldpays an intracellular price for its bliss.To curb cytoplasmic conßict, the slimemolds have a rigid hierarchy that setsout which sexes can inherit the mito-chondria of others. Moreover, this elab-orate bookkeeping system is subject tocheating. In his own journeys throughthe literature, Hurst found documenta-tion of a renegade mitochondrion thatrefused to respect the pecking order.Avoiding this complexity is probablywhy evolution favors two sexes, the Oxford researchers argue. Indeed, theslime molds may Þnd themselves hard-pressed in centuries to come as sexualchaos reigns. ÒMultiple sex types mightbe expected to collapse to binary types,Óthey write.

Although amused, not everyone isconvinced of this explanation. ÒItÕs whatI would call advocacy science,Ó com-ments Brian Charlesworth, a researcherin population genetics at the Universityof Chicago, who has proposed a diÝer-ent theory for how cytoplasmic genesare inherited. ÒYou try to make a mod-el and then Þnd something later thatsupports it.Ó Charlesworth also chargesthat Hurst and Hamilton fail to explainadequately in their paper the origins ofsexual dimorphism: why the male andfemale sex cells take on diÝerent sizesand shapes, something that Hurst as-serts is a distinct evolutionary issue.

Charlesworth also raises questionsabout whether Hurst and HamiltonÕsseveral dozen citations constitute abroad enough inspection of the litera-ture to make sweeping claims about theevolution of separate sexes. ÒNeither ofthese guys works in experimental ge-netics, which is involved with these phe-nomena,Ó Charlesworth remarks. ÒTheyare very much armchair theorists. It isdiÛcult to evaluate these data withouthands-on experience.Ó

Hurst acknowledges the need for ex-periments to conÞrm his ideas. He isworking with Rolf Hoekstra, a biologistat the Agricultural University of Wagen-ingen in the Netherlands, who suppliedthe evolutionary model used by Hurstand Hamilton in making their predic-tions. Hoekstra is trying to determinewhether a highly inbred species of fun-gus, Aspergillus nidulans, whose cyto-plasmic genes would be identical fromcell to cell, has any need for separatesexes. That work may provide a clue asto why opposites attractÑor pluses andminuses, if you prefer. ÑGary Stix


forms of matter must be an importantpart of the puzzle. Theoretical modelsof the big bang imply that the densityof ordinary, baryonic matter (protonsand neutrons) cannot exceed one tenthof the critical density needed to haltthe present cosmic expansion, other-wise the composition of the universewould be far diÝerent. But some stud-ies of large-scale motions of galaxies,as well as the currently favored versionof the big bang, require the universe to have the full critical density. Ninetypercent of the universe must consist ofexotic, as yet undetected, particles.

Lawrence Krauss of Yale Universityadmits that Òit sounds strange, but it ismore conservative to assume nonbary-

onic dark matterÓ than to abandon pres-ent models of cosmic genesis. Cosmol-ogists have proposed two general kindsof exotic dark matter: cold dark mat-ter, which would clump together read-ily, and hot dark matter, which wouldgather on far larger scales. Most cos-mologists prefer cold dark matter be-cause it seems better able to explainthe known distribution of galaxies.

Current uniÞed physics theories allowfor the existence of a bewildering arrayof potential ÒcoldÓ particles, includingaxions, magnetic monopoles and weaklyinteracting massive particles, or WIMPs.Krauss has done some housecleaningby analyzing recent particle physics ex-periments at the LEP collider in Switzer-

land and the Stanford Linear Colliderin California. His work has eliminated anumber of possible WIMPs and tightlyconstrained the potential properties ofthe others. Krauss is optimistic that thenegative results will help guide the nextround of searches for cold dark matter.Griest concurs. ÒI bet dark matter willbe found in a particle accelerator, eitherat LEP or in the Superconducting SuperColliderÑif it is built,Ó he says.

More direct searches for WIMPs arealso under way. Assuming dark matterparticles exist all around, they shouldoccasionally collide with the nuclei ofordinary atoms, leaving a detectable trailof ionization. One major experiment tosearch for such ionization signals will


Booby Prizesmid cries of “Excelsior!” and strains from the Close En-counters of the Third Kind theme, the oxymoronicSecond First Annual Ig Nobel Prize Ceremony be-

gan. There was one problem, though: because the stagedoors were locked, the presiding Swedish Meatball Kingand Queen had to knock for someone to let them in. Atleast a few of the evening’s prize recipients probably wishno one had. The Ig Nobels, unlike their more prestigiouscounterparts, honor individuals whose achievements can-not or should not be reproduced.

The Ig Nobel Prize Ceremony is a new October traditionin bad taste and indifferent science co-sponsored by theMassachusetts Institute of Technology Museum and theJournal of Irreproducible Results, a compendium of ersatzexperiments. The Ig Nobels are the brainchild of the jour-nal’s editor, Marc Abrahams, who hosted the festivities withdeadpan earnestness.

Among the winners were half a dozen scientists fromthe Shiseido Research Center in Yokohama, who took theprize in medicine for their studies of the chemicals re-sponsible for foot odor. Abrahams hailed their conclusionthat “people who think they have foot odor do, and thosewho don’t, don’t.”

Cecil Jacobson, a formerphysician described by Abra-hams as a “relentlessly gen-erous sperm donor,” won theIg Nobel Prize in Biology. Thispast March a Virginia juryconvicted Jacobson of fraudin a case in which prosecu-tors claimed he had secretlyimpregnated female patientsat his clinic with his own se-men. Jacobson, who was sen-tenced to five years in prisonbut is free on bond pendingan appeal, did not attend theceremony.

Prolificacy of a differentkind brought the literatureaward to Yuri Struchkov ofthe Institute of Organoele-mental Compounds in Mos-

cow. As Abrahams explained, Struchkov published 948scientific papers between 1981 and 1990—an average ofone every 3.9 days.

For identifying the cause of the mysterious circles offlattened crops that appeared in many British fields, DavidChorley and Doug Bower of the U.K. won the physics prize.Many hypotheses, ranging from odd meteorologic phe-nomena to UFOs, have been advanced, but their explana-tion has simplicity on its side: they claim to have made thecircles themselves with boards and pieces of string.

The French youth group Eclaireurs de France, a namemeaning “those who light the way,” won special recognitionfor its singular contributions to archaeology. While on anantigraffiti campaign near the village of Bruniquel, the eageryouths erased 15,000-year-old paintings from the walls ofthe Mayrieres Cave.

Daryl Gates, the former Los Angeles police chief, gar-nered the peace prize “for his uniquely compelling meth-ods of bringing people together.” Accepting the award forGates was Stan Goldberg of a Harvard Square camerastore, who confirmed that “Daryl Gates has done more forthe video camera industry than any other individual.”

The final prize of the eve-ning, for Ig Nobel accomplish-ments in art, went to JimKnowlton, whose poster “Pe-nises of the Animal Kingdom”shows the relative sizes andshapes of phalluses from hu-mans, pigs, whales and oth-er species. The National En-dowment for the Arts wasnamed as a co-recipient forallegedly encouraging Knowl-ton “to extend his work in theform of a pop-up book.”

At the end of the evening,with the stage doors stilllocked, the king and queenhad to shuffle out the sideexit. Next year, along with anew list of “ignitaries,” per-haps they will bring a key.—Shawna Vogel and J. RennieWEIRD SCIENCE prevails at the Ig Nobels.

A

STA

NLE

Y R

OW

IN


begin later this year in an unused tun-nel at the Stanford High Energy Phys-ics Laboratory. The experiment will usechunks of germanium cooled nearly toabsolute zero, which act as hypersensi-tive energy detectors.

David O. Caldwell of the University of California at Santa Barbara, a partici-pant in the project, notes that such di-rect dark matter searches Òare not sen-sitive to a particular candidate.Ó PositiveÞndings can be tested by looking forthe expected annual variation in ioniza-tion energy as the solar system movesthrough the sea of dark matter.

Even WIMPs may not be able com-pletely to solve the dark matter prob-lem. During the past year, the Cosmic

Background Explorer satellite has madesensitive microwave maps of the skythat underscore a serious defect in colddark matter models: they can accurate-ly simulate the structure of the universeon very large scales or on the scales ofindividual galaxies, but not both.

One way to Þx the models is to mixin a smidgen of hot dark matter withthe cold dark matter. Such cosmic com-bination can explain the structure ofthe universe on all scales, but it re-quires the existence of a hot dark mat-ter particle. Even Marc Davis of Berkeley,who recently published such a mixedmodel, admits that Òa few years ago Iwould have called this abhorrentÑinfact, I did call it abhorrent.Ó

Cosmologists looking for a hot darkmatter particle can at least point to acandidate that actually exists: the neu-trino. Physicists had long assumed neu-trinos to be massless. Unexpected re-sults from experiments designed to de-tect neutrinos emitted by the sun havebegun to suggest otherwise. Prelimi-nary Þndings from two new neutrinodetectors, SAGE in Russia and GALLEXin Italy, seem to bolster theories thatneutrinos do indeed possess a smallmass. But nobody yet knows if neutri-nos are massive enough to have playeda signiÞcant role in the evolution ofgalaxies.

Finally, John A. Bahcall of the Insti-tute for Advanced Study in Princeton,N.J., cautions that the dark matter prob-lem may be a sign that some fundamen-tal aspect of physics, such as the theoryof gravity, demands revision. And manyassumptions about dark matter dependon the essential validity of big bang cos-mology. ÒDark matter is the fundamen-tal problem that astronomers and physi-cists share,Ó Bahcall says. The outcomeof the current searches will test not onlythe cosmological orthodoxy but scien-tistsÕ ability to deduce the nature of auniverse that is mostly inaccessible totheir gaze. ÑCorey S. Powell

SCIENTIFIC AMERICAN January 1993 29Copyright 1992 Scientific American, Inc.

s a feminist in a family with Victo-rian mores and as a Jew and free-thinker in MussoliniÕs Italy, Rita

Levi-Montalcini has encountered variousforms of oppression many times in herlife. Yet the neurobiologist, whose tenac-ity and preciseness are immediately ap-parent in her light, steel-blue eyes andelegant black-and-white attire, embracesthe forces that shaped her.ÒIf I had not been discrim-inated against or had not suÝered persecution, I wouldnever have received the No-bel Prize,Ó she declares.

Poised on the edge of acouch in her apartment inRome that she shares withher twin sister, Paola, Levi-Montalcini recalls the long,determined struggle that cul-minated in joining the smallgroup of women Nobelistsin 1986. She won the prizefor elucidating a substanceessential to the survival ofnerve cells. Her discovery ofnerve growth factor led to anew understanding of thedevelopment and diÝerenti-ation of the nervous system.Today it and other similarfactors are the subject of in-tense investigation becauseof their potential to revivedamaged neurons, especial-ly those harmed in such dis-eases as AlzheimerÕs.

The journey from Turin,where she was born in 1909,to this serene and impecca-ble Roman living room ladenwith plants and with the etch-ings and sculptures of Pao-la, a well-known artist, test-ed Levi-MontalciniÕs mettlefrom her earliest years. ÒItwas a very patriarchal society, and I sim-ply resented, from early childhood, thatwomen were reared in such a way thateverything was decided by the man,Óshe proclaims. Initially, she wanted to bea philosopher but soon decided she wasnot logically minded enough. When hergoverness, to whom she was devoted,died of cancer, she chose to become adoctor. There only remained the smallmatter of getting her father, an engi-

neer, to grant permission and of mak-ing up for the time she had lost in agirlsÕ high school, where graduation ledto marriage, not to the university. ThatÒannoyed me so much that I decided tonever do as my mother did. And it wasa very good decisionÑat that time, Icould never have done anything in par-ticular if I had married.Ó Levi-Montalci-

ni pauses, leans forward and asks in-tensely, ÒAre you married?Ó She sighswith relief at the answer. ÒGood,Ó shesays, smiling.

After she received her fatherÕs grudg-ing consent, Levi-Montalcini studied forthe entrance examination and then en-rolled in the Turin School of Medicineat the age of 21. Drawn to a famous, ec-centric teacher, Giuseppe Levi, she de-cided to become an intern at the Insti-

tute of Anatomy. There Levi-Montalcinibecame adept at histology, in particu-lar at staining nerve cells.

Since Levi was curious about aspectsof the nervous system, he assigned hisstudent a Herculean labor: to Þgureout how the convolutions of the humanbrain are formed. In addition to theoverwhelming undertaking of Þndinghuman fetuses in a country where abor-tion was illegal, Òthe assignment was an impossible task to give your student

or an established scientist,ÓLevi-Montalcini explains, hervoice hardening. ÒIt was a re-ally stupid question, which IcouldnÕt solve and no onecould solve.Ó

She abandoned the proj-ectÑafter a series of un-pleasant forays for subjectmatterÑand with LeviÕs per-mission began to study thedevelopment of the nervoussystem in chick embryos.Several years later she wasforced to stop that work aswell. Mussolini had declaredhis dictatorship by 1925 andsince then anti-Semitism hadgrown in Italy. By 1936, hos-tility was openly apparent,and in 1939, Levi-Montalciniwithdrew from the universi-ty, worried about the safetyof her non-Jewish colleagueswho would be taking a riskby letting her study.

Levi-Montalcini acceptedan invitation to conduct herresearch at a neurological in-stitute in Belgium. But, fear-ing for her family, she soonreturned to TurinÑjust be-fore Mussolini and Hitlerforged their alliance. Unde-terred, Levi-Montalcini con-tinued her research: ÒI im-mediately found a way toestablish a laboratory in my

bedroom.Ó In the years that followed,bombs fell repeatedly, and again andagain she would lug her microscopeand slides to safety in the basement.

In spite of the hardshipÑor perhaps,as Levi-Montalcini sees it, because ofthe adversityÑit was during this timethat she laid the groundwork for her lat-er investigation of nerve growth factor.ÒYou never know what is good, what isbad in life,Ó she muses. ÒI mean, in my

Finding the Good in the Bad

PROFILE: RITALEVI-MONTALCINI

NOBEL LAUREATE Rita Levi-Montalcini conducted neurobiolog-ical research as bombs fell on her town during World War II.


AP

Wor

ldw

ide

Pho

tos


case, it was my good chance.Ó Levi-Mon-talcini and her family left Turin in 1942for the surrounding hills and success-fully survived the war in hiding. By con-vincing farmers that she needed eggs forher children (whom she did not have),Levi-Montalcini studied how embryon-ic nerve tissue diÝerentiates into spe-cialized types. The prevailing theory, de-veloped by renowned biologist ViktorHamburger of Washington University,held that the diÝerentiation, or special-ization, of nerve cells depends in largepart on their destination. In his experi-ments, Hamburger removed developinglimbs in chick embryos to see how suchexcision would aÝect the later growthand diÝerentiation of the nerve cellsdestined for that region of the embryo.

Hamburger observed that the centersof embryonic nerve cells near and in the developing spinal columnÑwherethe cells start their journey out to othertissuesÑwere much smaller when he ex-cised the limb buds. He suggested thatsome inductive or organizing factor,probably contained in the limb, couldno longer call out to the nerve cells.Therefore, they neither specialized norgrew away from the developing spinalcord into the region of the absent limb.

After conducting experiments direct-ed at the same question, Levi-Montal-cini reached a diÝerent conclusion. Shefound that fewer nerve cells grew intothe area where the limb bud had beeneliminated, but she proposed that somekind of nutrient, important for the sur-vival of nerve cells and normally pro-duced by the limb, was missing. Hertheory diÝered from HamburgerÕs viewbecause Levi-Montalcini proposed thatnerve cell diÝerentiation did take placedespite the removal of the limb but thatthe cells soon died because they did notreceive some sustaining, trophic factor.The limb did not contribute to diÝeren-tiation, that is, it did not contain an organizing factor; rather it producedsomething that nourished already spe-cialized nerve cells.

A paper of hers on this topic waspublished in a Belgian journal and wasread by Hamburger, who invited her toSt. Louis in 1946. Hamburger wantedto work with Levi-Montalcini on theproblem of nerve cell diÝerentiationÑand, indeed, later came to agree withher interpretation. Although she initial-ly accepted a semester-long researchposition at Washington, Levi-Montalciniremained until 1961. She is now profes-sor emeritus at Washington but spendsmost of her time in her native country.

Levi-Montalcini recalls being unsureof the future of her research after shearrived in the U.S. One afternoon, a series of observations, as well as the

presentation of a challenge, gave her a renewed sense of purpose. At thattime, neurobiologists thought diÝerenc-es in the number and function of vari-ous nerve cells were mostly the conse-quence of proliferative processes.

But Levi-Montalcini was about to dis-cover that the developing nervous sys-tem, at least in parts, uses a strategy dif-ferent from the one previously assumed.She had prepared a series of tissue slides

of chick embryo spinal cords in diÝer-ent stages of development. By looking at the succession of slides, she was able to observe the migration of nerve cells early in development to their Þnal po-sitions alongside the spinal column.There, for the Þrst time, she saw thelater elimination, or pruning back, ofsome of them. ÒI put on a Bach cantatabecause I was so terribly happy. I hadrealized that there was still so much tobe discovered,Ó says Levi-Montalcini, herdelight vividly clear.

Over the next several years, Levi-Mon-talcini focused on searching for the mys-terious trophic factor that she had intu-ited during the war. A former studentof HamburgerÕs had fortuitously noticedthat a certain mouse tumor cell lineÑcalled sarcoma 180Ñcaused more nervecells to grow. When Levi-Montalcini in-corporated the tumor cells into devel-oping chicks, she observed the same ef-fect. Something in the tumor caused thediÝerentiation of the nerve cells to ac-celerate; it also caused the creation ofexcessive numbers of nerve Þbers.

Levi-Montalcini started trying to iso-late the trophic factor and began to col-laborate with biochemist Stanley Cohen,then at Washington and now at the Van-derbilt University School of Medicine.They found that the partially puriÞedfactor contained both protein and nucle-ic acid. By adding enzymes from snakevenomÑwhich breaks down these com-poundsÑin hopes of determining whichcomponent contained the biological ac-tivity, the two discovered that the ven-om itself contained the factor.

This Þnding (described in detail in herautobiography, In Praise of Imperfection)led to the realization that nerve growthfactor is produced in salivary glands inmice, providing a new, easy source forstudies of the material. By designing anantiserum, Levi-Montalcini and Cohenwere able to chart the role of the factor.

It became clear that it is essential to thediÝerentiation and health of nerve cells.

In 1986 Levi-Montalcini and Cohenshared the Nobel Prize for this achieve-ment. When the phone rang in Romewith the news, she was pages from theend of Agatha ChristieÕs Evil under the

Sun. ÒAt the moment that I was Þndingout about the criminal, they told methat I was awarded the Nobel,Ó shelaughs, getting up to retrieve the bookfrom the hallway. She points to a hand-written note on the second-to-lastpageÑbeÞtting a neuroscientist, her ed-ition has a skull on the coverÑwhereshe had marked Òcall from StockholmÓand the time. ÒSo I was very happy aboutit, but I wanted much more to knowthe end of the story,Ó she admits.

Although she says her popularity in-terferes with her life, Levi-Montalcini hasused the Nobel to extend her work intoareas that concern her. She is presidentof the Italian Multiple Sclerosis Associ-ation and is a member of the PontiÞcalAcademy of Sciences; she was the Þrstwoman to be elected to the academy. ÒIcan do things that are very, very impor-tant, which I would never have been ableto do if I did not receive it,Ó she says. ÒIthas given me the possibility of helpinga lot of people.Ó And she helps whomev-er she can. The phone rings incessantlyin her apartment. ÒPeople ask for medi-cal help,Ó she explains, after answeringeach call and graciously talking with theparents or other relatives of someone ill.ÒBut sometimes there is nothing to do.Ó

In addition, Levi-Montalcini and hersister recently started their own project:a foundation that will provide mentors,counseling and grants to teenagers de-ciding what Þeld, whether it be art orscience, to enter. For several hours ev-ery week, she receives young students inher laboratory at the Institute of Neuro-biology at the National Research Councilin Rome and talks with them about theirinterests and her experiments. ÒTheonly way to help is to give young peoplea chance for the future. Because we can-not Þght the MaÞa, we cannot Þght cor-ruption without giving an alternative toyoung people,Ó she says.

Levi-MontalciniÕs research at the insti-tute, which she founded in the 1960s,has also taken a new turn. She is study-ing the role of nerve growth factor inthe immune and endocrine systems.ÒThe neotrophic factor was just the tipof the iceberg,Ó she notes. ÒSo even nowI am doing something entirely diÝer-ent. Just in the same spirit as when Iwas a young person. And this is verypleasing to me,Ó she says, laughing. ÒImean, at my old age, I could have nomore capacity. And I believe I still haveplenty.Ó ÑMarguerite Holloway

ÒI simply resented. . . thatwomen were reared insuch a way that everythingwas decided by the man.Ó


ate summer of 1987 seemed typi-cal for that time of year in theVirgin Islands. Huge, ßat-bot-

tomed cumulus clouds moved west-ward on light trade winds. Calm seaswere rarely disturbed by squalls sweep-ing into the northeastern Caribbean Seafrom the Atlantic Ocean. The only sug-gestion that something might be amisswas the water, which, though not sys-tematically measured, seemed unusual-ly warm to people swimming near theshallow coral reefs.

Something atypical had indeed oc-curred. The normally golden-brown,green, pink and gray corals, sea whipsand sponges had become pure white.In some cases, entire reefs were so daz-zlingly white that they could be seenfrom a considerable distance. In otherareas, pale corals punctuated the reefsurface while unbleached corals of thesame species grew as neighbors.

The phenomenon, which can be le-thal to coral, was not conÞned to theVirgin Islands. Observers at numerousmarine laboratories in the Caribbean

noted the same whitening. Nor was it the Þrst occurrence of such bleaching.In 1982 and 1983, after the atmosphericand oceanographic disturbance called ElNi�o/Southern Oscillation (ENSO), cor-als in certain areas of the Florida Keyswhitened and died, and oÝ the coast ofPanama mortality reached 50 percent.But it was only between 1987 and 1988,also an ENSO year, that reports of exten-sive bleaching became widespread. Theyhave increased in frequency ever since.

The association of coral bleachingwith ENSO, which ushers in warm wa-ter, and with water temperatures twoto three degrees Celsius above normalhas led some scientists to suggest thatthe bleaching is a manifestation of glob-al warming. Others point out that coralreefs have been studied only for a fewdecadesÑtoo short a time to permitgeneralized conclusions about a poorlyunderstood event.

Nevertheless, coral reefs around theworld are suÝering bouts of bleach-ing from which many do not recov-er. Although several factors can causethe processÑincluding disease, excessshade, increased ultraviolet radiation,sedimentation, pollution and changes insalinityÑthe episodes of the past de-cade have consistently been correlatedwith abnormally high seawater temper-atures. Understanding the complex pro-

BARBARA E. BROWN and JOHN C. OG-DEN work on international issues of coralreef ecology and management. Brown isdirector of the Centre for Tropical Coast-al Management and is reader in tropicalmarine biology at the University of New-castle upon Tyne in the U.K. She is alsoa founder of the International Societyfor Reef Studies. Ogden is the director ofthe Florida Institute of Oceanography andprofessor of biology at the University ofSouth Florida. He has served as a mem-ber of the faculty and as the director ofthe West Indies Laboratory in St. Croix,where he began his work on coral reefs.


Coral BleachingEnvironmental stresses can cause irreparableharm to coral reefs. Unusually high seawater

temperatures may be a principal culprit

by Barbara E. Brown and John C. Ogden

BOULDER CORAL has bleached only inpartsÑthe rest remains healthy for now.Although several factors cause potential-ly lethal whitening, recent bouts havebeen consistently correlated with higherthan average seawater temperatures.


cess of bleaching can help pinpoint, andperhaps eventually deter, this threat tothe ecology of the reefs.

Tropical, shallow-water ecosystems,coral reefs are found around theworld in the latitudes that general-

ly fall between the southern tip of Flori-da and mid-Australia. They rank amongthe most biologically productive of allmarine ecosystems. Because they harbora vast array of animals and plants, coralreefs are often compared to tropical rainforests. Reefs also support life on landin several ways. They form and maintainthe physical foundation for thousandsof islands. By building a wall along thecoast, they serve as a barrier againstoceanic waves. And they sustain the Þsh-eries and tourist diving industries thathelp to maintain the economies of manycountries in the Caribbean and PaciÞc.

Although corals seem almost architec-tural in structureÑsome weigh manytons and stand between Þve and 10 meters highÑthey are composed ofanimals. Thousands of tiny creaturesform enormous colonies: indeed, nearly60 percent of the 220 living genera ofcorals do so. Each colony is made up ofmany individual coral animals, calledpolyps. Each polyp is essentially a hol-low cylinder, closed at the base and in-terconnected to its neighbors by the gutcavity. The polyps have one or morerings of tentacles surrounding a centralmouth. In this way, corals resemble seaanemones with skeletons. The soft ex-ternal tissues of the polyps overlie ahard structure of calcium carbonate.

Many of the splendid colors of coralscome from their symbionts, creaturesthat live in a mutually dependent re-lation with the coral. Symbiotic algae

called zooxanthellae reside in the oftentransparent cells of the polyps. Thereare between one and two million algaecells per square centimeter of coral tis-sue. Through photosynthesis the algaeproduce carbon compounds, which helpto nourish the coralÑsome species re-ceive 60 percent of their food from theiralgae. Algal photosynthesis also acceler-ates the growth of the coral skeleton bycausing more calcium carbonate to beproduced. The corals provide algae withnutrients, such as nitrogen and phos-phorus, essential for growth, as well aswith housing. The association enables al-gae to obtain compounds that are scarcein the nutrient-poor waters of the trop-ics (where warm surface waters overlieand lock in cold, nutrient-rich watersÑexcept in restricted areas of upwelling).

When corals bleach, the delicate bal-ance among symbionts is destroyed.



The corals lose algae, leaving their tis-sues so colorless that only the white,calcium carbonate skeleton is apparent.Other organisms such as anemones, seawhips and spongesÑall of which harboralgae in their tissueÑcan also whiten inthis fashion. Some of this loss is routine.A healthy coral or anemone continuous-ly releases algae, but in very low num-bers. Under natural conditions, less than0.1 percent of the algae in a coral islost during processes of regulation andreplacement. When subject to adverse

changes, such as temperature increas-es, however, the corals release increasednumbers of algae. For example, trans-ferring coral from a reef to a laboratorycan cause a Þvefold elevation in thenumbers of algae expelled.

The mechanism of algal release is notfully understood. Even deÞning bleach-ing remains tricky. The current deÞni-tion has its basis in laboratory mea-surements of the loss of algae and thereduction in algal pigments. The labo-ratory approach, however, is rarely, if

ever, applied in the Þeld. There judg-ment must rely on the naked eyeÕs abil-ity to detect loss of coloration. Althoughsuch methods may be reliable for in-stances of severe bleaching, a determi-nation that pale colonies are bleachedcan be extremely arbitrary, given thenatural variability of pigmentation.

In some cases, normal coral under-going an adaptive behavioral responsecan look bleached. In 1989, while work-ing at the Phuket Marine Biological Cen-ter in Thailand, one of us (Brown) ob-served that some intertidal coral spe-ciesÑthose that thrive in shallow waterand are exposed to the air at low tideÑ

CORAL REEFS (red) thrive around the world in tropical, shal-low watersÑthose areas falling between the lines. They are

the most biologically productive of all marine ecosystems.Coral reefs also support life on land by providing a barrier

Some of the SpeciesThat Thrive on Coral Reefs

SPINY LOBSTER QUEEN CONCH LOGGERHEAD SPONGE SEA ANEMONE BUTTERFLY FISH

PACIF IC OCEAN ATLANTIC OCEAN

CARIBBEANSEA

EQUATOR


appear completely white during lowspring tides. It became clear that thesecorals are able to pull back their exter-nal tissues, leaving their skeletons ex-posed; they do not lose their algae.This behavior should perhaps be moreaccurately described as blanching, a re-sponse that may reduce desiccation dur-ing exposure to air.

Despite the absence of an unassail-able deÞnition of the bleaching pro-cess, several mechanisms have beenproposed that may be at work. In 1928Sir Maurice Yonge and A. G. Nicholls,who participated in an expedition to theGreat Barrier Reef, were among the Þrst

to describe coral bleaching. They sug-gested that algae migrated through cor-al tissue in response to environmentalstress, before being released into thegut and ultimately expelled through themouth. The precise trigger for the re-lease and the stimulus causing the algaeto be so conveyed were unknown thenand remain largely unknown today.

One of the several theories proposedby Leonard Muscatine of the Universi-ty of California at Los Angeles is thatstressed coral polyps provide fewer nu-trients to the algae. According to thistheory, the algae would not necessar-ily be directly aÝected by, say, hightemperature, but the metabolism of thecoral would be lowered. Supplies ofcarbon dioxide, nitrogen and phospho-rus would become insuÛcient, and thisrationing would in turn cause the algaeto abandon their residence.

In addition, Muscatine, R. GrantSteen and Ove Hoegh-Guldberg, also atU.C.L.A., studied the response of anem-ones and corals to changes in temper-ature, light and salinity. They describedthe release of algae from the tissues intothe gut cavity and hypothesized that the coral was actually losing animal tissue along with the algal cells. Work by Suharsono at the University of New-castle upon Tyne supported this idea.He showed that anemones exposed towarmer temperatures in the laborato-ry lose their own cells and algal cells during bleaching. Thus, host tissuethinned signiÞcantly, perhaps reducingthe space available for the algae.

The direct release of algae into thegut, however, may be a mecha-nism of algal loss that results

only from the extreme shocks invokedin a laboratory. It is not yet clear thatalgae behave the same way in corals in situ. Under natural conditions, it isquite likely that algae are released by a variety of mechanisms. All experi-mental work carried out on bleachinghas involved exposure to extreme tem-perature changesÑthat is, increases ofsix degrees C or more over a period of16 to 72 hours. In nature the tempera-ture increases that induce bleachingare much smaller, about two degrees C,and may occur over several months.

Another hypothesis suggests that al-gae emit poisonous substances whenthey experience adverse conditions and

that these toxins may deleteriously af-fect the host. Algae produce oxygencompounds, called superoxide radicals,in concentrations that can damage thecoral. (Molecular oxygen is relativelyunreactive, but it can be chemically al-tered to form the superoxide radical.)An enzyme, superoxide dismutase, inthe coral detoxiÞes the radicals.

But Michael P. Lesser and his col-leagues at the University of Maine not-ed that in certain cases oxygen toxicitycould lead to bleaching. Although Les-ser and his team were unable to mea-sure the oxygen radicals directly, theyfollowed the production of superoxidedismutase. They found that exposureto elevated temperatures and to in-creased ultraviolet radiation indepen-dently spurred enzymatic activity. Theresearchers concluded that oxygen tox-icity could be responsible for bleachingbecause harmful oxygen radicals wereexported from the damaged algae tothe animal host.

Other biochemical changes may takeplace as well. David Miller of the Univer-sity of Leeds and students from BrownÕslaboratory suggest that alterations ingene expression occur as a response to deleterious environmental changes.These changes may involve the synthe-sis of heat-shock proteinsÑcompoundsfound in all living systems subject toadversity that serve to protect cells tem-porarily from heat damage. Miller deter-mined that these proteins are enhancedin anemones undergoing heat shock.Furthermore, in anemones that toleratetemperature increases, the presence ofproteins appears to be correlated withreduced bleaching during heat shock.

Genetic variability also plays an im-portant role in bleaching. Environmen-tal factors may aÝect species of algae orcoral in diÝerent ways. Of course, pre-dicting the ability of corals and their algae to adapt to increases in seawatertemperature or global climatic changemay be possible by identifying the typesof corals or algae at highest risk.

When working together at the Uni-versity of California at Santa Barbara,Robert K. Trench and Rudolf J. Blankshowed that diÝerent corals act as hoststo varied strains of algae. Subsequent-ly, Rob Rowan and Dennis A. Powers ofStanford University have found that al-gae living in a single species of coral aregenetically similar in composition but


against oceanic waves and by formingthe foundation for thousands of islands.

SEA CUCUMBER CROWN-OF-THORNS STARFISH PARROT FISH SQUID SEA URCHIN COMB JELLY TUBE WORM

ARABIANSEA

INDIANOCEAN


genetically diÝerent from algae in othercoral species. Certain algae may proveto be particularly sensitive to tempera-ture and may have varying temperaturetolerances. If so, Rowan and PowersÕsÞndings would help explain why relat-ed, but not identical, corals exposed towarmer temperatures frequently showdiÝerent susceptibilities to bleaching.

Alternatively, the variability may liewith the coral instead of the algae. Stud-ies of several species of coral have indi-cated that genetically dissimilar strainsexist within a species of coral. Suchstrains may have diÝerent environmen-tal tolerances, which could account forthe observation that one colony of aparticular species appears to have beenbleached while a nearby member of thesame species has not been.

The reports of bleaching in the Ca-ribbean in the 1980s seem to be relatedmost consistently to elevated sea tem-peratures. Coral reefs normally thrivebetween 25 and 29 degrees C, depend-ing on their location. When patternsdepicting coral diversity are plotted ona globe, it is apparent that diversity declines as the reefs get farther away

from two centersÑone in the Indo-Pa-ciÞc and the other in the Caribbean.The outlines of a map marking plum-meting diversity coincide with the con-tours of lower seawater temperatures.

The narrow temperature range forhealthy coral is very close to its upperlethal temperature: an increase of oneto two degrees above the usual sum-mer maximum can be deadly. Paul Joki-el and Stephen Coles of the Universityof Hawaii have shown that bleachingand coral mortality are not induced bythe shock of rapidly ßuctuating temper-atures but are a response to prevailinghigh temperatures and to signiÞcant de-viations above or below the mean.

Many times during a 10-monthperiod in 1982Ð1983, an unusu-ally severe ENSO warmed the

waters of the eastern PaciÞc three tofour degrees C over the seasonal aver-age. Peter W. Glynn and his colleaguesat the University of Miami tracked theevent and the subsequent developmentsin that region. As a result of elevatedtemperatures, coral reefs underwentbleaching. Between 70 and 90 percent

of the corals in Panama and Costa Ricaperished several weeks later; more than95 percent of the corals in the Gal�pa-gos were destroyed.

Glynn and Luis DÕCroz of the Univer-sity of Panama also linked coral mortal-ity with high temperatures in a series oflaboratory experiments that duplicatedÞeld conditions during ENSO. The ma-jor reef-building coral of the easternPaciÞc, Pocillopora damicornis, took thesame amount of time to die in the lab-oratory at 32 degrees C as it did in theÞeld, indicating that the experimentshad replicated the natural condition.Glynn and DÕCroz also suggested thatthe temperature disproportionately af-fected types of coral that normally expe-rienced seasonal upwelling of deep, coolwater in the Gulf of Panama.

Evidence for the 1987 warming of theseawater in the Caribbean is not as de-Þnitive. Donald K. Atwood and his col-leagues at the Atlantic Oceanographicand Meteorological Laboratory in Miamiexamined the National OceanographicData CenterÕs sea-surface temperaturerecords from 1932 to the present andfound no discernible long-term increasesin seawater temperature in the Caribbe-an. The monthly mean sea-surface tem-perature did not exceed 30.2 degrees Cin any of the regions examinedÑin otherwords, the water remained well below the32 degrees C required to induce bleach-ing in GlynnÕs laboratory experiments.

Atwood also examined the maps fromthe National Climate Data Center of theNational Oceanic and Atmospheric Ad-ministration (NOAA). These records pro-vide average monthly sea-surface tem-perature and track anomalies derivedfrom satellite data that are validated bymeasurements taken from ships. Themaps indicate that in 1987 the surfaceof the Caribbean was generally less than30 degrees C. Other groups examinedsimilar temperature records and con-cluded that the temperatures of somesectors of the Caribbean reached 31 de-grees C or more during 1990, anotheryear of bleaching.

The records, of course, are subject tointerpretation based on the geographicscale of the satellite measurements andthe integration of these data with in situmeasurements. Unfortunately, there areno long-term temperature records takenat the small geographic scale needed toclarify the cause of damage to corals.

The 1987 reports of coral bleachingcoincided with escalating concern aboutglobal warming. It was not surprising,therefore, that some scientists and oth-er observers reached the conclusionthat coral reefs served as the canary inthe coal mineÑthe Þrst indication ofan increase in global ocean tempera-


Anatomy of a Coral Polyp

he coral animal is essential-ly a digestive sac—including

related organelles, or mesenterialfilaments—overlying a skeleton

Tof calcium carbonate. Itstentacles pull in food fromthe seawater. Algae calledzooxanthellae thrive in asymbiotic relation with thecoral. The algae sustain thecoral with oxygen and foodand also stimulate produc-tion of the skeleton, whichcan grow 10 centimeters a year. The coral, in turn,provides a home for the al-gae in its tissues and makesavailable nutrients such asnitrogen and phosphorus.

STINGINGCELL

MESENTERIALFILAMENTS

SKELETON

ZOOXANTHELLAEGULLET

TENTACLES



tures. Although it appears that elevat-ed local seawater temperatures causedbleaching, linking this eÝect to globalwarming cannot be conclusive at thistime. With the support of the Nation-al Science Foundation, NOAA and theEnvironmental Protection Agency, reefscientists and climatologists convened inMiami in June 1991 to discuss coral reefsand global climatic change. The work-shop determined that reports of coralbleaching were indicative of threats tothe ecosystem and that bleaching did ap-pear to be associated with local tempera-ture increases. But the paucity of knowl-edge about the physiological responseof corals to stress and temperature, theinadequacy of seawater temperature rec-ords and the lack of standardized pro-tocol for Þeld studies made it impossi-ble to decide whether bleaching reßectsglobal climatic change in the ocean.

Several international monitoring ef-forts are now in progress or are plannedso that the appropriate data can begathered. For example, the CaribbeanCoastal Marine Productivity Program, acooperative research network of morethan 20 Caribbean marine research in-stitutions in 15 countries, was foundedin 1990 and began systematic observa-tions of coral reefs in 1992. Other net-works of marine laboratories have beenproposed in the central and westernPaciÞc Ocean.

Whatever its cause, bleaching hasimportant implications for thecommunity structure, growth

and accretion of coral reefs. Develop-ing countries are particularly dependenton coral reefs for food resources andhave made heavy investments in reef-related tourism. Bleaching, added tothe accumulated toll taken by pollutionand overÞshing, may seriously burdenthe future economies of many nations.

The death of coral over such a widegeographic range in the eastern Pacif-ic during the 1982Ð1983 ENSO had se-vere biological repercussions. Before thewidespread bleaching, Glynn and hiscolleagues noticed that large Þelds ofPocillopora served to protect more mas-sive coral species from the coral-feed-ing crown-of-thorns sea star (Acanthas-

ter planci ). The starÞsh did not ventureacross the dense coral stands, because

SEAWATER TEMPERATURE ßuctuationsin the Caribbean Sea in 1990 were trackedby satellite. Temperatures reached be-tween 31 and 32 degrees Celsius (red ) incertain areas during the months of Au-gust (top), September (middle) and Octo-ber (bottom). Such unusually warm wateris believed to cause coral reefs to bleach.



Pocillopora repelled it with the stingingcells of its tentacles. In addition, sever-al species of symbiotic shrimp and crabin the Pocillopora attacked the sea stars,driving them away. As a result of warm-er water, however, Pocillopora suÝeredhigher mortality and lower fecundi-ty, and large corals were consequentlyopen to attack by the sea star. The pred-atory crustaceans were also aÝected.Because they normally feed on the lip-id-rich mucus produced by the Pocillo-

pora coral, a decline in the quantityand lipid content of the mucus broughtabout by the thermal stress triggered adecrease in the crustacean population.

The massive reduction in coral coveron the reefs of Panama and the Gal�pa-gos in 1982 and 1983 also restrictedthe range of one species of hydrocoralcalled Millepora and caused the appar-ent extinction of another species of thesame genus. Glynn and W. H. de Weerdt,now at the University of Amsterdam Zoo-logical Museum, speculate that thesespecies of corals were most severely aÝected because of their limited rangeand extreme sensitivity to increases in temperature. The disturbance alsocaused a nearly complete interruptionin the long-term accumulation of calci-um carbonate on the reefs of the region.

The fact that healthy reefs ßourishedin the eastern PaciÞc Ocean before 1982indicates that an event of this magni-tude is rare. Glynn estimated the age oftwo species of corals that were killedor heavily damaged in the Gal�pagosby multiplying their radius by their an-

nual growth rate of approximately onecentimeter. He concluded that an ENSOlike that of 1982Ð1983 had not occurredin the Gal�pagos for at least 200 years,possibly 400. The estimate is similar forcorals in Panama. Interestingly, even attheir most healthy, the coral reefs of theeastern PaciÞc are less well developedthan those of the Caribbean. Their com-paratively meager development may bepartly explained by the relatively fre-quent high- and low-temperature dis-turbances over thousands of years.

The coral frameworks of the reefs of Panama and the Gal�pagos havechanged dramatically as a result of thebleaching. Large areas of dead coralhave become colonized by benthic al-gae, which in turn support increasedpopulations of herbivores, particularlysea urchins. Sea urchins are grazers;they scrape the coral rock surface of thereef as they feed, contributing to theerosion of the reef structure.

Glynn and Ian Macintyre of theSmithsonian Institution and Gerard M.Wellington of the University of Hous-ton have estimated the rates of cal-cium carbonate accretion and erosionon the reef. The rates of erosion af-ter the 1982Ð1983 ENSO attributable tosea urchins alone are greater than therates of accumulation on the healthyreefs before 1983. This Þnding sug-gests that without recovery of coral pop-ulations, these reefs will soon be re-duced to carbonate sediments. Becausethe grazers erode the reef surface, theymay also interfere with the recruitment

of new coral colonies, prolonging, oreven preventing, their recovery.

Coral became bleached at many oth-er places in the Indo-PaciÞc during the1982Ð1983 ENSO, including the Socie-ty Islands, the Great Barrier Reef, thewestern Indian Ocean and Indonesia.Brown and Suharsono noted widespreadbleaching and loss of as much as 80 or 90 percent of the coral cover on theshallow coral reefs of the Thousand Is-lands in the Java Sea. The corals mostaÝected were on the shallow reef tops.Five years later coral cover was only 50percent of its former level.

The extent of bleaching, environ-mental tolerances and the life-his-tory characteristics of the domi-

nant corals determine whether a reef re-covers from the loss of most of its livingcoral. So do the nature and timing of oth-er disturbances, such as predation andgrazing. When bleaching is severe or pro-longed, the coral may die. If the bleachingepisode is short, the coral can rebuild itsalgal population and continue to live,but biological processes such as growthand reproduction may be impaired.

Because we are only now forming net-works of sites that will conduct coop-erative observations, the extent of coralreef damage brought about by bleach-ing has not been globally assessed. In 1987 Ernest H. Williams, Jr., of the Uni-versity of Puerto Rico collected reports of bleaching from nearly every tropicalocean region. But until we have an ade-quate deÞnition of coral bleaching inthe Þeld and have standardized our ob-servations, the global impact of coralbleaching will remain a mystery.

If the temperature increase of one ortwo degrees C, predicted by the Inter-governmental Panel on Climate Change,does take place over the next 50 yearsin the tropical latitudes, the consequen-ces for coral reefs could be disastrous.Unlike the miners with the canary, wecannot yet link bleaching to a clearcause. But that does not mean we shouldignore the coralÕs message.

FURTHER READING

GLOBAL ECOLOGICAL CONSEQUENCES OFTHE EL NINO SOUTHERN OSCILLATION1982Ð83. Edited by Peter W. Glynn. El-sevier, Amsterdam, 1989.

CORAL BLEACHING. Edited by Barbara E.Brown. Special Issue of Coral Reefs, Vol.8, No. 4; April 1990.

WORKSHOP ON CORAL BLEACHING, COR-AL REEF ECOSYSTEMS AND GLOBALCHANGE: REPORT OF PROCEEDINGS. Or-ganized by Christopher F. DÕElia, Rob-ert W. Buddemeir and Stephen V. Smith.Maryland Sea Grant College, 1991.

INTERTIDAL CORALS in Thailand ÒblanchÓ when exposed to the air. Unlike bleach-ing, this phenomenon appears to be adaptive: the coral polyps retract their soft tis-sues during low tide, leaving the calcium carbonate skeleton exposed. When waterwashes over again, the polyps and tentacles expand to cover the skeleton.

~


ttempts to reconstruct how theMilky Way formed and began toevolve resemble an archaeolog-

ical investigation of an ancient civiliza-tion buried below the bustling centerof an ever changing modern city. Fromexcavations of foundations, some pot-tery shards and a few bones, we mustinfer how our ancestors were born,how they grew old and died and howthey may have helped create the livingculture above. Like archaeologists, as-tronomers, too, look at small, disparateclues to determine how our galaxy andothers like it were born about a billionyears after the big bang and took ontheir current shapes. The clues consistof the ages of stars and stellar clusters,their distribution and their chemistryÑall deduced by looking at such featuresas color and luminosity. The shapesand physical properties of other galax-ies can also provide insight concerningthe formation of our own.

The evidence suggests that our gal-axy, the Milky Way, came into being asa consequence of the collapse of a vastgas cloud. Yet that cannot be the wholestory. Recent observations have forcedworkers who support the hypothesis ofa simple, rapid collapse to modify theiridea in important ways. This new infor-

mation has led other researchers to pos-tulate that several gas cloud fragmentsmerged to create the protogalactic Mil-ky Way, which then collapsed. Other var-iations on these themes are vigorouslymaintained. Investigators of virtually allpersuasions recognize that the births ofstars and supernovae have helped shapethe Milky Way. Indeed, the formationand explosion of stars are at this mo-ment further altering the galaxyÕs struc-ture and inßuencing its ultimate fate.

Much of the stellar archaeologi-cal information that astrono-mers rely on to decipher the

evolution of our galaxy resides in tworegions of the Milky Way: the halo andthe disk. The halo is a slowly rotating,spherical region that surrounds all theother parts of the galaxy. The stars andstar clusters in it are old. The rapidlyrotating, equatorial region constitutesthe disk, which consists of young starsand stars of intermediate age, as wellas interstellar gas and dust. Embeddedin the disk are the sweepingly curvedarms that are characteristic of spiral gal-axies such as the Milky Way. Among themiddle-aged stars is our sun, which islocated about 25,000 light-years fromthe galactic center. (When you view thenight sky, the galactic center lies in thedirection of Sagittarius.) The sun com-pletes an orbit around the center in ap-proximately 200 million years.

That the sun is part of the Milky Waywas discovered less than 70 years ago.At the time, Bertil Lindblad of Swedenand the late Jan H. Oort of the Nether-


SIDNEY VAN DEN BERGH and JAMES E. HESSER both work at Dominion Astro-physical Observatory, National ResearchCouncil of Canada, in Victoria, BritishColumbia. Van den Bergh has a longtimeinterest in the classiÞcation and evolu-tion of galaxies and in problems relatedto the age and size of the universe. Hereceived his undergraduate degree fromPrinceton University and a doctorate inastronomy from the University of G�t-tingen. HesserÕs current interests focuson the ages and compositions of glob-ular star clusters, which are among theoldest constituents of the galaxy. He re-ceived his B.A. from the University ofKansas and his Ph.D. in atomic and mo-lecular physics from Princeton.

How the Milky Way FormedIts halo and disk suggest that the collapse

of a gas cloud, stellar explosions and the captureof galactic fragments may have all played a role

by Sidney van den Bergh and James E. Hesser

MILKY WAY COMPONENTS include thetenuous halo, the central bulge and ahighly ßattened disk that contains thespiral arms. The nucleus is obscured bythe stars and gas clouds of the centralbulge. Stars in the bulge and halo tendto be old; disk stars such as the sun areyoung or middle-aged.

HALO

SOLAR SYSTEM


lands hypothesized that the Milky Waysystem is a ßattened, diÝerentially ro-tating galaxy. A few years later John S.Plaskett and Joseph A. Pearce of Do-minion Astrophysical Observatory ac-cumulated three decadesÕ worth ofdata on stellar motions that conÞrmedthe Lindblad-Oort picture.

In addition to a disk and a halo, theMilky Way contains two other subsys-tems: a central bulge, which consistsprimarily of old stars, and, within thebulge, a nucleus. Little is known aboutthe nucleus because the dense gas

clouds in the central bulge obscure it.The nuclei of some spiral galaxies, in-cluding the Milky Way, may contain alarge black hole. A black hole in the nu-cleus of our galaxy, however, would notbe as massive as those that seem to actas the powerful cores of quasars.

All four components of the MilkyWay appear to be embedded in a large,dark corona of invisible material. Inmost spiral galaxies the mass of thisinvisible corona exceeds by an order ofmagnitude that of all the galaxyÕs visi-ble gas and stars. Investigators are in-

tensely debating what the constituentsof this dark matter might be.

The clues to how the Milky Way de-veloped lie in its components. Perhapsthe only widely accepted idea is thatthe central bulge formed Þrst, throughthe collapse of a gas cloud. The centralbulge, after all, contains mostly mas-sive, old stars. But determining whenand how the disk and halo formed ismore problematic.

In 1958 Oort proposed a model ac-cording to which the population of starsforming in the halo ßattened into a

DISK

GLOBULARCLUSTERS

CENTRALBULGE



thick disk, which then evolved into athin one. Meanwhile further condensa-tion of stars from the hydrogen leftover in the halo replenished that struc-ture. Other astronomers prefer a pic-ture in which these populations are dis-crete and do not fade into one another.In particular, V. G. Berman and A. A.Suchkov of the Rostov State Universityin Russia have indicated how the diskand halo could have developed as sepa-rate entities.

These workers suggest a hiatus be-tween star formation in the halo andthat in the disk. According to their mod-el, a strong wind propelled by superno-va explosions interrupted star forma-tion in the disk for a few billion years.In doing so, the wind would have eject-ed a signiÞcant fraction of the mass ofthe protogalaxy into intergalactic space.Such a process seems to have prevailedin the Large Magellanic Cloud, one ofthe Milky WayÕs small satellite galaxies.There an almost 10-billion-year inter-lude appears to separate the initial burstof creation of conglomerations of oldstars called globular clusters and themore recent epoch of star formation inthe disk. Other Þndings lend additionalweight to the notion of distinct galac-tic components. The nearby spiral M33contains a halo but no nuclear bulge.

This characteristic indicates that a halois not just an extension of the interiorfeature, as many thought until recently.

In 1962 a model emerged that servedas a paradigm for most investigators.According to its developersÑOlin J. Eg-gen, now at the National Optical Astro-nomical Observatories, Donald Lynden-Bell of the University of Cambridge andAllan R. Sandage of the Carnegie Insti-tutionÑthe Milky Way formed when alarge, rotating gas cloud collapsed rap-idly, in about a few hundred millionyears. As the cloud fell inward on itself,the protogalaxy began to rotate morequickly; the rotation created the spiralarms we see today. At Þrst, the cloudconsisted entirely of hydrogen and he-lium atoms, which were forged duringthe hot, dense initial stages of the bigbang. Over time the protogalaxy start-ed to form massive, short-lived stars.These stars modiÞed the compositionof galactic matter, so that the subse-quent generations of stars, includingour sun, contain signiÞcant amounts ofelements heavier than helium.

Although the model gained wide ac-ceptance, observations made during thepast three decades have uncovered anumber of problems with it. In the Þrstplace, investigators found that many of the oldest stars and star clusters in

the galactic halo move in retrograde or-bitsÑthat is, they revolve around thegalactic center in a direction oppositeto that of most other stars. Such orbitssuggest that the protogalaxy was quiteclumpy and turbulent or that it capturedsizable gaseous fragments whose mat-ter was moving in diÝerent directions.Second, more reÞned dynamic modelsshow that the protogalaxy would nothave collapsed as smoothly as predict-ed by the simple model; instead thedensest parts would have fallen inwardmuch faster than more rareÞed regions.

Third, the time scale of galaxy forma-tion may have been longer than thatdeduced by Eggen and his colleagues.Exploding supernovae, plasma windspouring from massive, short-lived starsand energy from an active galactic nu-cleus are all possible factors. The galaxymay also have subsequently rejuvenateditself by absorbing large inßows of pris-tine intergalactic gas and by capturingsmall, gas-rich satellite galaxies.

Several investigators have attempt-ed to develop scenarios consistent withthe Þndings. In 1977 Alar Toomre ofthe Massachusetts Institute of Technol-ogy postulated that most galaxies formfrom the merger of several large piecesrather than from the collapse of a sin-gle gas cloud. Once merged in this way,according to Toomre, the gas cloud col-lapsed and evolved into the Milky Waynow seen. Leonard Searle of the Car-negie Institution and Robert J. Zinn ofYale University have suggested a some-what diÝerent picture, in which manysmall bits and pieces coalesced. In thescenarios proposed by Toomre and bySearle and Zinn, the ancestral frag-ments may have evolved in chemical-ly unique ways. If stars began to shineand supernovae started to explode indifferent fragments at diÝerent times,then each ancestral fragment wouldhave its own chemical signature. Recentwork by one of us (van den Bergh) indi-cates that such diÝerences do indeedappear among the halo populations.

Discussion of the history of galac-tic evolution did not advancesigniÞcantly beyond this point

until the 1980s. At that time, workersbecame able to record more preciselythan ever before extremely faint imag-es. This ability is critically importantbecause the physical theories of stellarenergy productionÑand hence the life-times and ages of starsÑare most se-cure for so-called main-sequence stars.Such stars burn hydrogen in their cores;in general, the more massive the star,the more quickly it completes its main-sequence life. Unfortunately, this fact

COLOR-LUMINOSITY DIAGRAMS can be used to determine stellar ages. The oneabove compares the plots of stars in globular clusters NGC 288 and NGC 362 withage tracks (black lines) generated by stellar evolution models. The color index, ex-pressed in magnitude units, is a measure of the intensity of blue wavelengths mi-nus visual ones. In general, the brighter the star, the lower the color index; thetrend reverses for stars brighter than about visual magnitude 19. The plots suggestthe clusters differ in age by about three billion years. The temperature (inverselyrelated to the color index) and luminosity have been set to equal those of NGC 288.

COLOR INDEX (MAGNITUDE UNITS)0.6

VIS

UA

L M

AG

NIT

UD

E

0.4 0.8 1

16

18

20

22

21

19

17

23

LUM

INO

SIT

Y (

SO

LAR

UN

ITS

)

0.10

1.0

10

TEMPERATURE (DEGREES CELSIUS)4,9005,150 4,650 4,400

12-BILLION-YEAR-OLD AGE TRACK

15-BILLION-YEAR-OLD AGE TRACK

NGC 288

NGC 362


means that within the halo the only re-maining main-sequence stars are theextremely faint ones. The largest, mostluminous ones, which have burned pasttheir main-sequence stage, became in-visible long ago. Clusters are generallyused to determine age. They are crucialbecause their distances from the earthcan be determined much more accurate-ly than can those of individual stars.

The technology responsible for open-ing the study of extremely faint halostars is the charge-coupled device (CCD).This highly sensitive detector producesimages electronically by converting lightintensity into current. CCDs are far su-perior in most respects to photographicemulsions, although extremely sophisti-cated software, such as that developedby Peter B. Stetson of Dominion Astro-physical Observatory, is required totake full advantage of them. So used,the charge-coupled device has yielded atenfold increase in the precision of mea-surement of color and luminosity ofthe faint stars in globular clusters.

Among the most important resultsof the CCD work done so far are moreprecise age estimates. Relative age databased on these new techniques have re-vealed that clusters whose chemistriessuggest they were the Þrst to be creat-ed after the big bang have the same ageto within 500 million years of one an-other. The ages of other clusters, how-ever, exhibit a greater spread.

The ages measured have helped re-searchers determine how long it tookfor the galactic halo to form. For in-stance, Michael J. Bolte, now at Lick Ob-servatory, carefully measured the colorsand luminosities of individual stars inthe globular clusters NGC 288 and NGC362 [see illustration on opposite page].Comparison between these data andstellar evolutionary calculations showsthat NGC 288 is approximately 15 bil-lion years old and that NGC 362 is onlyabout 12 billion years in age. This dif-ference is greater than the uncertain-ties in the measurements. The observedage range indicates that the collapse ofthe outer halo is likely to have taken an order of magnitude longer than theamount of time Þrst envisaged in thesimple, rapid collapse model of Eggen,Lynden-Bell and Sandage.

Of course, it is possible that morethan one model for the forma-tion of the galaxy is correct.

The EggenÐLynden-BellÐSandage scenar-io may apply to the dense bulge and in-ner halo. The more rareÞed outer partsof the galaxy may have developed by themerger of fragments, along the linestheorized by Toomre or by Searle and

Zinn. If so, then the clusters in the innerhalo would have formed before thosein the more tenuous outer regions. Theprocess would account for some of the age diÝerences found for the glob-ular clusters. More precise modeling may have to await the improved imagequality that modiÞcations to the Hub-

ble Space Telescope cameras will aÝord.Knowing the age of the halo is, how-

ever, insuÛcient to ascertain a detailedformation scenario. Investigators need

to know the age of the disk as well andthen to compare that age with thehaloÕs age. Whereas globular clustersare useful in determining the age of the halo, another type of celestialbodyÑvery faint white dwarf starsÑcan be used to determine the age of thedisk. The absence of white dwarfs in thegalactic disk near the sun sets a lower limit on the diskÕs age. White dwarfs,which are no longer producing radiantenergy, take a long time to cool, so their

OXYGEN-TO-IRON RATIOS as a function of metallicity (abundance of iron) for haloand old disk stars indicate diÝerent formation histories. The high ratios in metal-deÞcient halo stars suggest that those stars incorporated the oxygen synthesizedin supernovae of types Ib, Ic and II. Type Ia supernovae seem to have contributedmaterial only to the disk stars. Beatriz Barbuy and Marcia Erdelyi-Mendes of theUniversity of S�o Paulo made the measurements.

GLOBULAR CLUSTERS, such as Messier 5 above, appear to be among the oldest ob-jects known. They oÝer invaluable insight into the haloÕs formation some 15 billionyears ago. The 100,000 or so stars exhibit similar abundances of heavy elements,implying that the gas cloud from which each arose was chemically homogeneous.


IRON ABUNDANCE (SOLAR UNITS)

OX

YG

EN

-TO

-IR

ON

RA

TIO

(S

OLA

R U

NIT

S)

0.010.001 0.1 1.0

4

3

2

1

HALO STARSOLD DISK STARS


absence means that the population inthe disk is fairly youngÑless thanabout 10 billion years. This value is sig-niÞcantly less than the ages of clustersin the halo and is thus consistent withthe notion that the bulk of the galacticdisk developed after the halo.

It is, however, not yet clear if there isa real gap between the time when for-mation of the galactic halo ended andwhen creation of the old thick disk be-gan. To estimate the duration of such a

transitional period between halo anddisk, investigators have compared theages of the oldest stars in the disk withthose of the youngest ones in the halo.The oldest known star clusters in thegalactic disk, NGC 188 and NGC 6791,have ages of nearly eight billion years,according to Pierre Demarque and Da-vid B. Guenther of Yale and ElizabethM. Green of the University of Arizona.Stetson and his colleagues and RobertoBuonanno of the Astronomical Obser-vatory in Rome and his co-workers ex-amined globular clusters in the halopopulation. They found the youngestglobularsÑPalomar 12 and Ruprecht106Ñto be about 11 billion years old. Ifthe few billion yearsÕ diÝerence betweenthe disk objects and the young globu-lars is real, then young globulars maybe the missing links between the diskand halo populations of the galaxy.

At present, unfortunately, the relativeages of only a few globular clusters havebeen precisely estimated. As long as thisis the case, one can argue that the MilkyWay could have tidally captured Palo-mar 12 and Ruprecht 106 from the Mag-ellanic Clouds. This scenario, proposedby Douglas N. C. Lin of the Universityof California at Santa Cruz and HarveyB. Richer of the University of British Co-lumbia, would obviate the need for along collapse time. Furthermore, the ap-parent age gap between disk and halomight be illusory. Undetected systemat-ic errors may lurk in the age-dating pro-cesses. Moreover, gravitational interac-tions with massive interstellar cloudsmay have disrupted the oldest disk clus-ters, leaving behind only younger ones.

Determining the relative ages of thehalo and disk reveals much about thesequence of the formation of the gal-axy. On the other hand, it leaves openthe question of how old the entire gal-axy actually is. The answer would pro-vide some absolute framework by whichthe sequence of formation events canbe discerned. Most astronomers whostudy star clusters favor an age ofsome 15 to 17 billion years for the old-est clusters (and hence the galaxy).

ConÞdence that those absolute agevalues are realistic comes from the mea-

sured abundance of radioactive isotopesin meteorites. The ratios of thorium 232to uranium 235, of uranium 235 to ura-nium 238 or of uranium 238 to plutoni-um 244 act as chronometers. Accordingto these isotopes, the galaxy is between10 and 20 billion years old. Althoughages determined by such isotope ratiosare believed to be less accurate thanthose achieved by comparing stellar ob-servations and models, the consistencyof the numbers is encouraging.

Looking at the shapes of other galax-

ies alleviates to some extent theuncertainty of interpreting the

galaxyÕs evolution. SpeciÞcally, the studyof other galaxies presents a perspectivethat is unavailable to us as residents of the Milky WayÑan external view. Wecan also compare information from oth-er galaxies to see if the processes thatcreated the Milky Way are unique.

The most immediate observation onecan make is that galaxies come in sev-eral shapes. In 1925 Edwin P. Hubblefound that luminous galaxies could bearranged in a linear sequence accord-ing to whether they are elliptical, spi-ral or irregular [see top illustration on

this page]. From an evolutionary pointof view, elliptical galaxies are the mostadvanced. They have used up all (or al-most all) of their gas to generate stars,which probably range in age from 10 to 15 billion years. Unlike spiral galax-ies, ellipticals lack disk structures. Themain diÝerences between spiral and ir-regular galaxies is that irregulars haveneither spiral arms nor compact nuclei.

The morphological types of galax-ies can be understood in terms of the speed with which gas was used to cre-ate stars. Determining the rate of gasdepletion would corroborate estimatesof the Milky WayÕs age and history. Starformation in elliptical galaxies appearsto have started oÝ rapidly and eÛcient-ly some 15 billion years ago and thendeclined sharply. In most irregular gal-axies the birth of stars has taken placemuch more slowly and at a more near-ly constant rate. Thus, a signiÞcant frac-tion of their primordial gas still remains.

The rate of star formation in spiralsseems to represent a compromise be-tween that in ellipticals and that in ir-regulars. Star formation in spirals be-gan less rapidly than it did in ellipticalsbut continues to the present day.

Spirals are further subdivided intocategories Sa, Sb and Sc. The subdivi-sions refer to the relative size of thenuclear bulges and the degree to whichthe spiral arms coil. Objects of type Sahave the largest nuclear bulges and themost tightly coiled arms. Such spiralsalso contain some neutral hydrogen gas


MORPHOLOGICAL CLASSIFICATION ofgalaxies (top) ranges from ellipticals (E)to spirals (subdivided into categories Sa,Sb and Sc) and irregulars ( Ir ). The histo-ry of star formation varies according tomorphology (bottom). In elliptical galax-ies, stars developed in an initial burst.Star formation in spirals was less vigor-ous but continues today. In most irreg-ular galaxies the birthrate of stars hasprobably remained constant.

TIME (BILLIONS OF YEARS)

RA

TE

OF

ST

AR

FO

RM

AT

ION

100 5 15

Sa

E

Sb

ScIr

E

Sa

Sc

Sb

Ir

EDGE-ON FACE-ON


and a sprinkling of young blue stars. Sbspirals have relatively large populationsof young blue stars in their spiral arms.The central bulge, containing old redstars, is less prominent than is the cen-tral bulge in spirals of type Sa. Finally, inSc spirals the light comes mainly fromthe young blue stars in the spiral arms;the bulge population is inconspicuousor absent. The Milky Way is probably in-termediate between types Sb and Sc.

Information from other spirals seemsconsistent with the data obtained forthe Milky Way. Like those in our galaxy,the stars in the central bulges of otherspirals arose early. The dense inner re-gions of gas must have collapsed Þrst.As a result, most of the primordial gasinitially present near the centers hasturned into stars or has been ejectedby supernova-driven winds.

There is an additional kind of evi-dence on which to build our un-derstanding of how the Milky Way

came into existence: the chemical com-position of stars. This information helpsto pinpoint the relative ages of stellarpopulations. According to stellar mod-els, the chemistry of a star depends onwhen it formed. The chemical diÝerenc-es exist because Þrst-generation starsbegan to ÒpolluteÓ the protogalaxy withelements heavier than helium. Such so-called heavy elements, or Òmetals,Ó asastronomers refer to them, were creat-ed in the interiors of stars or duringsupernova explosions. Examining themakeup of stars can provide stellar evo-lutionary histories that corroborate orchallenge age estimates.

DiÝerent types of stars and super-novae produce diÝerent relative abun-dances of these metals. Researchers be-lieve that most Òiron-peakÓ elements(those closest to iron in the periodictable) in the galaxy were made in su-pernovae of type Ia. The progenitors ofsuch supernovae are thought to be pairsof stars, each of which has a mass a fewtimes that of the sun. Other heavy ele-mentsÑthe bulk of oxygen, neon, mag-nesium, silicon and calcium, amongothersÑoriginated in supernovae thatevolved from single or binaries of mas-sive, short-lived stars. Such stars haveinitial masses of 10 to 100 solar mass-es and violently end their lives as su-pernovae of type Ib, Ic or II.

Stars that subsequently formed in-corporated some of these heavy ele-ments. For instance, approximately 1to 2 percent of the mass of the sunconsists of elements other than hydro-gen or helium. Stars in nuclear bulgesgenerally harbor proportionally moreheavy elements than do stars in theouter disks and halos. The abundance


MODELS OF GALAXY FORMATION fall into three general categories. In the EggenÐLynden-BellÐSandage model, the Milky Way formed by the rapid collapse of a sin-gle gaseous protogalaxy (a ). In the Toomre model, several large aggregates of gasmerged (b). The Searle-Zinn picture is similar to the Toomre model except that theancestral fragments consisted of much smaller but more numerous pieces (c ).

MESSIER 83 is a typical type Sc spiral galaxy. The Milky Way probably has a similarappearance, although its arms may be somewhat more tightly coiled.

a

b

c


of heavy elements decreases graduallyby a factor of 0.8 for every kiloparsec(3,300 light-years) from the center tothe edge of the Milky Way disk. Some70 percent of the 150 or so known glob-ular clusters in the Milky Way exhibit an average metal content of about onetwentieth that of the sun. The remain-der shows a mean of about one thirdthat of the sun.

Detailed studies of stellar abundanc-es reveal that the ratio of oxygen toiron-peak elements is larger in halostars than it is in metal-rich disk stars[see upper illustration on page 75 ]. ThisdiÝerence suggests the production ofheavy elements during the halo phaseof galactic evolution was dominated bysupernovae of types Ib, Ic and II. It ispuzzling that iron-producing type Ia su-pernovae, some of which are believed to have resulted from progenitor starswith lifetimes as short as a few hundredmillion years, did not contribute moreto the chemical mixture from whichhalo stars and some globular clustersformed. This failure would seem to im-ply that the halo collapsed very rapid-lyÑbefore supernovae of type Ia couldcontribute their iron to the halo gas.

That idea, however, conßicts with thefour-billion-year age spread observedamong galactic globular clusters, whichimplies that the halo collapsed slow-ly. Perhaps supernova-driven galacticwinds swept the iron-rich ejecta fromtype Ia supernovae into intergalacticspace. Such preferential removal of theejecta of type Ia supernovae might haveoccurred if supernovae of types Ib, Icand II exploded primarily in dense

gas clouds. Most of type Ia supernovaethen must have detonated in less denseregions, which are more easily sweptout by the galactic wind.

Despite the quantity of data, informa-tion about metal content has provedinsuÛcient to settle the controversyconcerning the time scale of disk andhalo formation. Sandage and his col-league Gary A. Fouts of Santa MonicaCollege Þnd evidence for a rather mono-lithic collapse. On the other hand, JohnE. Norris and his collaborators at theAustralian National Observatory, amongothers, argue for a signiÞcant decou-pling between the formation of haloand disk. They also posit a more chaot-ic creation of the galaxy, similar to thatenvisaged by Searle and Zinn.

Such diÝerences in interpretation of-ten reßect nearly unavoidable eÝectsarising from the way in which particu-lar samples of stars are selected forstudy. For example, some stars exhibitchemical compositions similar to thoseof ÒgenuineÓ halo stars, yet they havekinematics that would associate themwith one of the subcomponents of thedisk. As vital as it is, chemical informa-tion alone does not resolve ambigu-ities about the formation of the galac-tic halo and disk. ÒCats and dogs mayhave the same age and metallicity, butthey are still cats and dogsÓ is the wayBernard Pagel of the Nordic Institutefor Theoretical Physics in Copenhagenputs it.

As well as telling us about the pasthistory of our galaxy, the disk and haloalso provide insight into the Milky WayÕsprobable future evolution. One can easi-

ly calculate that almost all of the ex-isting gas will be consumed in a fewbillion years. This estimate is based onthe rate of star formation in the disksof other spirals and on the assumptionthat the birth of stars will continue at itspresent speed. Once the gas has beendepleted, no more stars will form, andthe disks of spirals will then fade. Even-tually the galaxy will consist of nothingmore than white dwarfs and black holesencapsulated by the hypothesized darkmatter corona.

Several sources of evidence existfor such an evolutionary scenar-io. In 1978 Harvey R. Butcher of

the Kapteyn Laboratory in the Nether-lands and Augustus Oemler, Jr., of Yalefound that dense clusters of galaxieslocated about six billion light-yearsaway still contained numerous spiralgalaxies. Such spirals are, however, rareor absent in nearby clusters of galaxies.This observation shows that the disksof most spirals in dense clusters musthave faded to invisibility during thepast six billion years. Even more directevidence for the swift evolution of gal-axies comes from the observation ofso-called blue galaxies. These galaxiesare rapidly generating large stars. Suchblue galaxies seem to be less commonnow than they were only a few billionyears ago.

Of course, the life of spiral galaxiescan be extended. Copious infall of hy-drogen from intergalactic space mightreplenish the gas supply. Such infall canoccur if a large gas cloud or anothergalaxy with a substantial gas reservoir isnearby. Indeed, the Magellanic Cloudswill eventually plummet into the MilkyWay, brießy rejuvenating our galaxy.Yet the Milky Way will not escape its ul-timate fate. Like people and civiliza-tions, stars and galaxies leave behindonly artifacts in an evolving, ever dy-namic universe.


FURTHER READINGGALACTIC ASTRONOMY: STRUCTURE ANDKINEMATICS. Dimitri Mihalas and JamesBinney. W. H. Freeman and Company,1981.

THE MILKY WAY AS A GALAXY. GerardGilmore, Ivan R. King and Pieter C. vander Kruit. University Science Books, MillValley, Calif., 1990.

THE FORMATION AND EVOLUTION OFSTAR CLUSTERS. Edited by KennethJanes. Astronomical Society of the Pa-ciÞc, 1991.

THE STELLAR POPULATIONS OF GALAXIES.Edited by B. Barbuy and A. Renzini. Klu-wer Academic Publishers, Dordrecht,Holland, 1992.

LARGE MAGELLANIC CLOUD is one of the Milky WayÕs two largest satellite galax-ies. Slowly spiraling into the Milky Way, the cloud will brießy rejuvenate our gal-axy at some time in the distant future.


n 1952 Aaron Moscona of the Uni-versity of Chicago separated thecells of a chick embryo by incubat-

ing them in an enzyme solution andswirling them gently. The cells did notremain apart; they coalesced into a newaggregate. Moreover, Moscona saw thatwhen retinal cells and liver cells were al-lowed to coalesce in this way, the reti-nal cells always migrated to the innerpart of the cellular mass. Three yearslater Philip L. Townes and JohannesHoltfreter of the University of Roches-ter performed a similar experiment withcells from amphibian embryos, whichre-sorted themselves into tissue layerslike those from which they had come.

Those experiments and countless oth-er observations testify to the keen abili-ty of cells to recognize one another andto respond accordingly. Sperm, for ex-ample, can distinguish eggs of their ownspecies from those of others, and theywill bind only with the former. Somebacteria settle preferentially in the in-testinal or urinary tract; others fancydiÝerent organs.

It is not surprising, then, that decod-ing the language of cellular interactionsholds profound interest for researchersin many areas of biology and medicine.Although we still do not understand

the chemical basis of most cell-recogni-tion phenomena, clear explanations forsome have emerged during the past de-cade. Proteins, which mediate most ofthe chemical reactions inside living or-ganisms, appear on the cell surface aswell, and they certainly play a part.

Yet the accumulating evidence alsosuggests that in many cases carbohy-drates (frequently referred to as sug-ars) are the primary markers for cellrecognition. Discoveries about the in-volvement of speciÞc sugars in recogni-tion will have practical applications tothe prevention and treatment of a vari-ety of ailments, including cancer.

Biologists generally accept that cellsrecognize one another throughpairs of complementary struc-

tures on their surfaces: a structure onone cell carries encoded biological in-formation that the structure on the oth-er cell can decipher. That idea repre-sents an extension of the lock-and-keyhypothesis formulated in 1897 by EmilFischer, the noted German chemist. Heused it to explain the speciÞcity of in-teractions between enzymes and theirsubstrates. Pioneering immunologistPaul Ehrlich extended it in 1900 to ac-count for the highly speciÞc reactionsof the immune system, and in 1914Frank Rattray Lillie of the University ofChicago invoked it to describe recogni-tion between sperm and eggs.

By the 1920s the lock-and-key hypoth-esis had become one of the central the-oretical assumptions of cellular biolo-gy. Yet for many years thereafter, thenature and identity of the molecules in-volved in cellular recognition remaineda complete mystery.

To most biologists, the idea that the molecules might be carbohydratesseemed farfetched. That large class ofcompounds consists of monosaccha-rides (simple sugars such as glucose and

fructose) and of oligosaccharides andpolysaccharides, which are composed oflinked monosaccharides. Until the late1960s, carbohydrates were thought toserve only as energy sources (in theforms of monosaccharides and storagemolecules such as the polysaccharidestarch) and as structural materials (thepolysaccharides cellulose in plants andchitin in the exoskeletons of insects).The two other major classes of biologi-cal materialsÑnucleic acids, which car-ry genetic information, and proteinsÑwere obviously far more versatile. Bycomparison, carbohydrates looked likedull, second-class citizens.

Interest in carbohydrates was furtherdiscouraged by the extraordinary com-plexity of their structures. In contrastto the nucleotides in nucleic acids andthe amino acids in proteins, which caninterconnect in only one way, the mono-saccharide units in oligosaccharides andpolysaccharides can attach to one an-other at multiple points. Two identi-cal monosaccharides can bond to form 11 diÝerent disaccharides, whereas twoamino acids can make only one dipep-tide. Even a small number of monosac-charides can create a staggering diver-sity of compounds, including many withbranching structures. Four diÝerent nu-cleotides can make only 24 distinct tet-


NATHAN SHARON and HALINA LIShave been members of the biophysicsdepartment of the Weizmann Institute ofScience in Rehovot, Israel, for more than30 years. During most of that time, theyhave collaborated closely on the studyof complex carbohydrates and lectins,proteins that bind selectively to carbo-hydrates. In addition to their many jointscientiÞc papers, they have written sev-eral widely cited review articles on thosetopics, as well as the recent book Lectins(Chapman & Hall, London). This is Shar-onÕs Þfth article for Scientific American.

Carbohydratesin Cell RecognitionTelltale surface sugars enable cells to identify

and interact with one another. New drugs aimed at thosecarbohydrates could stop infection and inflammation

by Nathan Sharon and Halina Lis

HORMONE

GLYCOPROTEIN


ranucleotides, but four diÝerent mono-saccharides can make 35,560 uniquetetrasaccharides.

This potential for structural diver-sity is the bane of the carbohy-drate chemist, but it is a boon to

cells: it makes sugar polymers superblyeÝective carriers of information. Carbo-hydrates can carry much more informa-tion per unit weight than do either nu-cleic acids or proteins. Monosaccharidescan therefore serve as letters in a vo-cabulary of biological speciÞcity; thecarbohydrate words are spelled out byvariations in the monosaccharides, dif-ferences in the links between them andthe presence or absence of branches.

Scattered reports that carbohydratescould deÞne speciÞcity began to appearquite early in the scientiÞc literature, al-though they often went unnoticed. Bythe 1950s, for example, it was well es-tablished that injected polysaccharidescould stimulate the production of an-tibodies in animals. Researchers alsoknew that the major ABO blood typesare determined by sugars on blood cellsand that the inßuenza virus binds to a red blood cell through a sugar, sialicacid. Yet not until the 1960s did sugarscome into their own.

Two major developments promptedthat change. The Þrst was the realiza-

tion that all cells carry a sugar coat. Thiscoat consists for the most part of glyco-proteins and glycolipids, two types ofcomplex carbohydrates in which sugarsare linked to proteins and lipids (fats),respectively. Several thousands of glyco-protein and glycolipid structures havebeen identiÞed, and their number growsalmost daily. This diversity is surely sig-niÞcant: the repertoire of surface struc-tures on a cell changes characteristicallyas it develops, diÝerentiates or sickens.The array of carbohydrates on cancercells is strikingly diÝerent from that onnormal ones.

An additional stimulus came fromthe study of lectinsÑa class of proteinsthat can combine with sugars rapidly,selectively and reversibly. Biologists oncethought lectins were found only inplants, but in fact they are ubiquitous innature. Lectins frequently appear on thesurfaces of cells, where they are strate-gically positioned to combine with car-bohydrates on neighboring cells. Theydemonstrate exquisite speciÞcity: lectinsdistinguish not only between diÝerentmonosaccharides but also between dif-ferent oligosaccharides.

A landmark discovery about the roleof lectin-carbohydrate interactions incell recognition came from the work ofG. Gilbert Ashwell of the National In-stitutes of Health and Anatol Morell ofthe Albert Einstein College of Medicine.In 1968 they enzymatically removed afew sialic acid molecules from certain


SURFACE CARBOHYDRATES on a cell serve as points of attachment for other cells,infectious bacteria, viruses, toxins, hormones and many other molecules. In thisway, carbohydrates mediate the migration of cells during embryo development,the process of infection and other phenomena. Compounds consisting of carbohy-drates that are chemically linked to proteins are called glycoproteins; those inwhich the carbohydrates are linked to fats are glycolipids.

TOXIN

VIRUS

BACTERIUM

CELL


blood plasma glycoproteins, then in-jected the glycoproteins into rabbits. Or-dinarily, such molecules would persistin the animalsÕ circulation for some time,but the sialic acidÐdeÞcient moleculesquickly disappeared.

Ashwell and Morell found that the gly-coproteins ended up in the liver. The re-moval of the sialic acids had unmaskedgalactose in the glycoproteins, and theexposed galactoses had attached to alectin on the liver cells. Subsequently,the researchers learned that if they re-moved both the sialic acids and the un-covered galactoses from the glycopro-teins, the rate at which the moleculeswere eliminated from the blood re-turned to normal. From those results,Ashwell and Morell concluded that car-bohydrate side chains on proteins mayserve as markers for identifying whichones should be removed from the circu-lation and eventually degraded.

Like the surface carbohydrates, thesurface lectins go through changes thatcoincide with a cellÕs physiological andpathological states. For instance, in 1981Reuben Lotan and Abraham Raz of theWeizmann Institute of Science showedthat tumor cells from mice and humanscarry a surface lectin not found on nor-mal cells. They and other researcherslater proved that this lectin is involvedin the development of metastases.

Astriking recent illustration of therole of surface sugars and themolecules that bind to them

comes from studies of embryo forma-tion by Senitiroh Hakomori of the FredHutchinson Cancer Research Center inSeattle and by Ten Feizi of the ClinicalResearch Center in Harrow, England.Working with mouse embryos, theyhave shown that as a fertilized egg di-vides, the carbohydrate structures on

the resulting embryonic cells change incharacteristic ways. One of the carbo-hydrates is a trisaccharide known bothas stage-speciÞc embryonic antigen 1(SSEA-1) and as Lewisx (Lex ). It appearsat the eight- to 16-cell stage, just as theembryo compacts from a group of loosecells into a smooth ball.

HakomoriÕs group has shown that asoluble compound carrying multipleunits of the same trisaccharide inhib-its the compaction process and dis-rupts embryogenesis. Closely related butstructurally diÝerent carbohydrates haveno eÝect. Thus, the Lex trisaccharideappears to play a part in compaction.

Adhesive carbohydrates are there-fore essential to embryonic develop-ment. As research continues, their rolein that process will become more de-tailed. Today the two best-understoodphenomena of that type are microbialadhesion to host cells and the adhesion


The Complexity of Carbohydrate Structuresarbohydrates, nucleic acids and proteins all carry bi-ological information in their structures. Yet carbohy-

drates offer the highest capacity for carrying informationbecause they have the greatest potential for structuralvariety. Their component molecules, monosaccharides,can interconnect at several points to form a wide varietyof branched or linear structures; in the example below,

the branching carbohydrate is only one of many possiblestructures that can be made from four identical glucosemolecules. The amino acids in proteins as well as the nu-cleotides in nucleic acids can form only linear assemblies,which restricts their diversity. The peptide (protein frag-ment) shown here is the only one possible made from fourmolecules of the amino acid glycine.

AMINO ACID (GLYCINE) PEPTIDE (TETRAGLYCINE)

MONOSACCHARIDE (GLUCOSE) OLIGOSACCHARIDE (BRANCHED TETRAGLUCOSE)

CARBON OXYGEN NITROGEN HYDROGEN

C


of white blood cells to blood vessels.The more thoroughly characterized ofthese interactions is microbial adhe-sion, which has been studied for nearlytwo decades and serves as a model forother forms of carbohydrate-mediatedcell recognition.

To cause disease, viruses, bacteria or protozoa must be able to stick to atleast one tissue surface in a suscepti-ble host. Infectious agents lacking thatability are swept away from potentialsites of infection by the bodyÕs normalcleansing mechanisms. Microorganismsin the upper respiratory tract, for ex-ample, may be swallowed and eventu-ally destroyed by stomach acid; thosein the urinary tract may be ßushed outin the urine.

The Þrst clues about the mechanismof bacterial adhesion sprang from a series of pioneering studies by J. P. Duguid of Ninewells Hospital MedicalSchool in Dundee that began in the1950s. Duguid demonstrated that manystrains of Escherichia coli (a bacterialdenizen of the intestines that can alsocolonize other tissues) and related bac-teria adhered to cells from the epitheliallining of tissues and to erythrocytes, orred blood cells. In the presence of stickybacteria, the erythrocytes would clumptogetherÑa phenomenon called hemag-glutination. (Researchers still routinelyuse hemagglutination as a simple testfor the adhesion of bacteria to animalcells.) To learn how the bacteria boundto the cells, Duguid exposed them to awide range of compounds. He foundthat only the monosaccharide mannoseand very similar sugars could inhibithemagglutination.

Duguid also made the important ob-servation that the bacterial strains re-sponsible for mannose-sensitive hemag-glutination had submicroscopic, hair-like appendages on their surfaces. Thesestructures were Þve to 10 nanometersin diameter and several hundreds ofnanometers long. He called them Þm-briae, from the Latin word for fringe.Almost simultaneously, Charles C. Brin-ton, Jr., of the University of Pittsburghdescribed the same structures andnamed them pili, from the Latin wordfor hairs. Both terms are still in use.

Later, starting around 1970, RonaldJ. Gibbons of the Forsyth Dental Centerin Boston and his colleagues began re-porting on the selective adhesion ofbacteria to niches within the oral cavi-ty. Gibbons observed that Actinomyces

naeslundii colonizes both the epithelialsurfaces of infants without teeth andthe teeth of children and adults. Con-versely, the related bacterium A. visco-

sus does not appear in the mouth untilthe teeth erupt from the gums; it exhib-

its a preference for teeth rather thanoral epithelial surfaces.

Today it is clear that the tissuespeciÞcity of bacterial adhesionis a general phenomenon. For ex-

ample, E. coli, the most common causeof urinary tract infections, is abundantin tissues surrounding the ducts thatconnect the kidneys and the bladder,yet it is seldom found in the upper res-piratory tract. In contrast, group Astreptococci, which colonize only theupper respiratory tract and skin, rarelycause urinary tract infections.

Bacterial adhesion varies not only be-tween tissues but also between speciesand sometimes between individuals ofthe same species, depending on theirage, genetic makeup and health. In theearly 1970s R. Sellwood and Richard A.Gibbons and their colleagues at the In-stitute for Research on Animal Diseasesin Compton, England, studied the infec-tivity of the K88 strain of E. coli. Be-cause those bacteria cause diarrhea inpiglets, they are a costly nuisance forfarmers. GibbonsÕs group found thatthe K88 bacteria adhered to the intesti-nal cells of susceptible piglets but notto those of adult pigs or of humans,which the bacteria cannot infect. Bacte-rial mutants that had lost the ability tobind to intestinal cells proved unableto infect the animals.

Moreover, as GibbonsÕs work showed,some piglets had a genetic resistanceto K88 bacteria: even potentially viru-lent bacteria could not bind to cellsfrom their intestines. By selecting ge-netically immune piglets for breeding,farmers were able to obtain K88-resis-tant progeny.

The gonorrhea organism, Neisseria

gonorrhoeae, serves as another exampleof species and tissue speciÞcity. It ad-heres to human cells of the genital andoral epithelia but not to cells from otherorgans or other animal species. That factexplains why humans are the exclusivehost for N. gonorrhoeae and why otheranimals do not contract gonorrhea.

A strong impetus to the study of bac-terial adhesion was a proposal made in1977 by Itzhak Ofek of Tel Aviv Uni-versity, David Mirelman of our depart-ment at the Weizmann Institute and oneof us (Sharon). We suggested that bac-terial adhesion is mediated by surfacelectins on bacteria that bind to comple-mentary sugars on host cells. That ideahas proved to be generally valid. Workin many laboratories has shown thatbacteria produce lectins speciÞc for cer-tain carbohydrates and that the bacte-ria depend on those lectins for adher-ing to a hostÕs tissue as the Þrst step inthe process of infection.

Bacterial lectins have already beenthe focus of much study, although far


BACTERIA ADHERE to tissues selectively. Hairlike protrusions called Þmbriae onthe bacteria bind exclusively to certain surface carbohydrates. These interactionsdetermine which tissues are susceptible to bacterial invasion. Rod-shaped Esche-richia coli bacteria are shown here on tissue from the urinary tract.


more remains to be done. The best-char-acterized lectins are the type 1 Þmbriaeof E. coli , which bind preferentially tosurface glycoproteins containing man-nose. Other research on E. coli during the past decade, primarily by CatharinaSvanborg-Ed�n and her colleagues whileworking at the University of G�teborg,has described in detail the P Þmbriae.Those Þmbriae interact speciÞcally withthe P blood-group substance, an ex-tremely common glycolipid containingthe disaccharide galabiose. Researchgroups led by Karl-Anders Karlsson ofthe University of G�teborg and VictorGinsburg of the NIH have mapped thespeciÞcities of lectins from a wide rangeof other bacterial species and strains.

These studies have shown that bacte-ria do not bind solely to the ends ofsurface carbohydratesÑthey can alsosometimes bind to sugars located with-in the structure. Furthermore, diÝerentbacteria may bind to diÝerent parts ofthe same carbohydrate. Occasionally,only one face of an oligosaccharide maybe exposed on a particular cell, and asa result the cell will bind bacteria of onekind and not other ones. The ability ofcell-surface sugars to serve as attach-ment sites therefore depends not onlyon the presence of these sugars but alsoon their accessibility and their mode ofpresentation.

Considerable experimental evi-dence now greatly strengthensthe conclusion that the binding

of bacteria to host cell-surface sugarsinitiates infection. For example, uroep-ithelial cells from those rare individu-als who lack the P blood-group sub-stance do not bind to P-Þmbriated E.

coli. Such individuals are much less sus-ceptible to infections from those bacte-ria than the rest of the population is.

Experiments have shown, however, thatthe bacteria will bind if the epithelialcells are Þrst coated with a synthetic gly-colipid containing galabiose.

Similarly, intestinal cells from pigletsthat are resistant to the diarrhea-caus-ing K88 E. coli lack the large carbohy-drate to which the bacteria bind. Al-though the exact structure of this car-bohydrate has not yet been elucidated,it is known to be present in susceptiblepiglets but absent in adult pigs. This ex-plains why the bacteria were unable toattach to and colonize the intestines ofthe adult pigs, while they caused infec-tion in the young piglets.

Another interesting case is that of theK99 strain of E. coli. Like the K88 strain,K99 causes diarrhea in farm animalsbut not in humans. It is less speciÞc thanK88, however, because it infects youngcalves and lambs as well as piglets. TheK99 bacteria bind speciÞcally to an un-usual glycolipid that contains N-glycol-oylneuraminic acid (a special type ofsialic acid) linked to lactosylceramide.This glycolipid, which is present in pig-lets, calves and lambs, is absent fromthe cells of adult pigs and humans,which instead contain N-acetylneura-minic acid, a nonbinding analogue ofsialic acid. Here a small diÝerence be-tween two highly similar sugarsÑthe re-placement of an acetyl group by a gly-coloyl groupÑis readily detected by thebacteria and explains the host range ofinfection by the organism.

Further conÞrmation of the above con-clusions was recently obtained in experi-ments on two Þmbrial lectins from E.

coli that infect the urinary tracts of ei-ther humans or dogs. Both lectins rec-

ognize galabiose, yet one binds only tothe human uroepithelial cells and theother only to canine cells; the galabiose-bearing glycolipids on the surface ofthe cells are presented in subtly differ-ent ways. Those lectin-binding patternsaccord with the host speciÞcities of theE. coli strains.

Because bacterial adhesion is socritical to infection, medical re-searchers are seriously consider-

ing the use of sugars for prevention andtreatment. Sugars that selectively inhib-ited adhesion could act as molecular de-coys, intercepting pathogenic bacteriabefore they reached their tissue targets.Urinary tract infections have been thefocus of particular attention becausethey are second only to respiratory in-fections in frequency.

In collaboration with Ofek, MosheAronson of Tel Aviv University and Mi-relman, we performed the Þrst studyalong those lines in 1979. We injected a mannose-speciÞc strain of E. coli intothe urinary bladder of mice. In some animals, we also injected methyl alpha-mannoside, a sugar that in the test tubeinhibited bacterial adhesion to epithelialcells. The presence of the sugar reducedthe colonization of the urinary tract bybacteria.

Svanborg-Ed�n has performed analo-gous experiments with P-Þmbriated E.

coli that infect the kidneys of mice. Sheincubated the bacteria in solutions ofglobotetraose, a sugar found in the gly-colipid of kidney cells. When she sub-sequently injected those bacteria into


WHITE BLOOD CELLS of the immune system, such as lymphocytes, defend the bodyagainst infection by leaving the general circulation and migrating into the tissues.The Þrst step involves selective adhesion between the white blood cells and thewalls of blood vessels called high endothelial venules (photograph). Adhesion de-pends on surface molecules called selectins, which bind to carbohydrates on other

LYMPHOCYTE

HOMINGRECEPTOR(L-SELECTIN)

LYMPHOCYTE MIGRATINGOUT OF VENULE

CARBOHYDRATES

LYMPHOCYTEWITH DIFFERENT

HOMING RECEPTORS

ENDOTHELIAL

CELL

PEYER’S PATCH

PERIPHERAL LYMPH NODE


mice, they persisted in the kidneys forless time than untreated bacteria did.James A. Roberts of Tulane Universityobtained similar results in experimentson monkeys: incubation of P-ÞmbriatedE. coli with a galabioselike sugar signif-icantly delayed the onset of urinary tractinfections.

Glycopeptides can also interfere withthe binding of bacteria to host tis-sues. In 1990 Michelle Mouricout of theUniversity of Limoges in France andher co-workers showed that injectionsof glycopeptides taken from the bloodplasma of cows can protect newborncalves from lethal doses of E. coli. Theglycopeptides, which contain sugars forwhich the bacteria have afÞnities, de-crease the adhesion of the bacteria tothe intestines of treated animals.

Indeed, to interfere with bacterial ad-hesion, one need not even use a carbo-hydrateÑany agent that competitivelybinds to either the bacterial lectin or thehost cellÕs surface carbohydrate will do.For example, Edwin H. Beachey and hiscolleagues at the Veterans Administra-tion Medical Center and the Universityof Tennessee at Memphis have used an-tibodies against mannose to preventcertain mannose-speciÞc E. coli frominfecting mice. The antibodies bind tomannose on the cells, thereby blockingthe sites of bacterial attachment.

Those successful experiments makea clear case for antiadhesive therapies

against microbial diseases. The applica-tion of this approach in humans is nowthe subject of intense research. Furtherstudies of the sugars on host cells andof bacterial lectins should lead to thedesign of better adhesion inhibitors.One point about the approach is certain:because diÝerent infectious agentsÑeven diÝerent bacteria within the samestrainÑcan have a wide variety of carbo-hydrate speciÞcities, a cocktail of inhib-itors will undoubtedly be necessary toprevent or treat the diseases.

Carbohydrate-directed interactionsbetween cells are not restrictedto pathological phenomena; they

are also crucially important to thehealthy operation of the immune sys-tem. The immune system has manyparts, but its most important soldiersare the cells called leukocytes. Thisgroup includes an array of diverse whiteblood cellsÑlymphocytes, monocytesand neutrophilsÑthat act jointly toeliminate bacteria and other intrudersand to mediate the inßammation re-sponse in injured tissues. All these cellscirculate in the blood, but they accom-plish their major functions in the ex-travascular spaces.

The picture emerging from researchis that the inner lining of blood vessels,called the endothelium, actively snareswhite blood cells and guides them towhere they are needed. This process re-

quires an exquisitely regulated recogni-tion between the circulating leukocytesand the endothelial cells.

Such recognition seems to be mediat-ed by a family of structurally relatedlectins. Because this Þeld of research isso new and because diÝerent laborato-ries often identify the same adhesionmolecules simultaneously, the nomen-clature is still in a somewhat chaoticstate. Most researchers refer to thesemolecules as selectins because they me-diate the selective contact between cells.Another name in vogue is LEC-CAMs, anacronym for leukocyte-cell (or lectin)adhesion molecules.

The selectins are highly asymmetriccomposite proteins, with an unusualmosaic architecture. They consist ofthree types of functional domains: onedomain anchors the selectin in the cellmembrane, and a second makes upmost of the body of the molecule. Thethird domain, located at the extracellu-lar tip of the molecule, structurally re-sembles animal lectins that work onlyin the presence of calcium ions. Thebinding of carbohydrate ligands to thatdomain is central to the function of se-lectins in interactions between cells.

About 10 years ago Eugene C. Butch-er and Irving L. Weissman of Stan-ford University laid the foundation forour current understanding of how se-lectins (which were then unknown) di-rect lymphocyte traÛc. Lymphocytesare unique among leukocytes in thatthey continuously patrol the body insearch of foreign antigens (immunolog-ically signiÞcant molecules) from bac-teria, viruses and the like. For that pur-pose, lymphocytes leave the blood ves-sels and migrate through the lymphnodes, the tonsils, the adenoids, thePeyerÕs patches in the intestines or oth-er secondary lymphoid organs. DiÝer-ent lymphocytes migrate selectively, orhome, toward particular organs. To exitfrom the bloodstream, lymphocytesmust Þrst bind to specialized submi-croscopic blood vessels less than 30microns in diameter, known as high en-dothelial venules.

Using an assay technique developedby Hugh B. Stamper, Jr., and Judith J.WoodruÝ of the State University of NewYork in Brooklyn, Butcher and Weiss-man observed that the homing speciÞc-ity of mouse lymphocytes is dictatedby their selective interaction with thehigh endothelial venules in their tar-geted organs. Butcher and Weissmanthen developed a monoclonal antibody,MEL-14, that bound only to mouse lym-phocytes that went to the peripherallymph nodes. On slices of tissue, theantibody blocked the attachment of thelymphocytes to high endothelial venules


cells. The L-selectins, or homing receptors, on lymphocytes determine the endothe-lial cells to which a lymphocyte will stick : for example, some adhere only in periph-eral lymph nodes or to the PeyerÕs patches in the intestines. After a lymphocyte hasattached to the endothelium, it can migrate out of the blood vessel.


from those tissues but not from oth-er lymph organs. When injected intomice, MEL-14 inhibited the migration oflymphocytes into the peripheral lymphnodes.

Butcher and Weissman went on toshow that their antibody bound on thelymphocyte membrane to a single gly-coprotein, now known as L-selectin. Be-cause that glycoprotein is responsiblefor the speciÞc binding of the lympho-cytes to the high endothelial venules, itis also known as the homing receptor.

If high endothelial venules from thelymph nodes are exposed to solutionsof L-selectin, lymphocytes cannot bindto them: the L-selectin molecules occu-py all the potential attachment sites onthe endothelial cells. Conversely, as Ste-ven D. Rosen of the University of Cali-

fornia at San Francisco has shown, cer-tain small sugars and larger polysac-charides can also block interactions be-tween lymphocytes and endothelialvenules. In those cases, the sugars arebinding to the L-selectin.

In 1989 separate experiments byWeissman and by Laurence A. Lasky ofGenentech in South San Francisco incollaboration with Rosen proved con-clusively that the homing receptor me-diates the adhesion of lymphocytes toendothelial cells. The structure of theendothelial carbohydrate to which itbinds is still unknown.

In contrast to the homing receptor,the two other known selectins are foundmainly on endothelial cells, and thenonly when they are actively attractingleukocytes. One of these, E-selectin

(ELAM-1), was discovered in 1987 byMichael P. Bevilacqua of Harvard Medi-cal School. The third member of thegroup, P-selectin (previously known asGMP-140 and PADGEM), was indepen-dently discovered a couple of years lat-er by Rodger P. McEver of the Oklaho-ma Medical Research Foundation andby Bruce and Barbara Furie of the TuftsUniversity School of Medicine.

Research has clariÞed how tissuesuse selectins to steer white blood cellswhere they are needed. When a tissueis infected, it defensively secretes pro-teins called cytokines, such as interleu-kin-1 and tumor necrosis factor. Thecytokines stimulate endothelial cells inthe venules to express P- and E-selectinson their surfaces. Passing white bloodcells adhere to these protruding mole-

BACTERIUM

LECTIN

SURFACECARBOHYDRATES

CARBOHYDRATEDRUG

a b c

LECTINDRUG

CELL


BLOCKING BACTERIAL ATTACHMENT is one strategy for com-bating infections. As a prelude to infection, bacterial surfaceproteins called lectins attach to surface carbohydrates onsusceptible host cells (a). Drugs containing similar carbohy-

drates could prevent the attachment by binding to the lectins(b). Alternatively, drugs consisting of lectinlike moleculescould have the same eÝect by innocuously occupying thebinding sites on the carbohydrates (c).

SELECTIVE EFFECTS of carbohydrates on bacteria are il-lustrated in these photographs. These E. coli have a lectin for the P glycolipid. Bacteria incubated in the sugar mannose

can still cling to epithelial tissue (left ). A constituent of the P glycolipid binds to the bacteriaÕs lectin and prevents ad-hesion (right ).


cules because their carbohydrate coatcontains complementary structures.Once attached to the wall of a venule, aleukocyte can leave the bloodstream bysqueezing between adjacent endothe-lial cells.

These two selectins appear on endo-thelial cells at diÝerent times and re-cruit diÝerent types of white blood cells.Endothelial cells have an internal stock-pile of P-selectin that they can mobilizeto their surface within minutes after aninfection begins. P-selectin can there-fore draw leukocytes that act duringthe earliest phases of the immunologicdefense. In contrast, endothelial cellssynthesize E-selectin only when it is required, so it takes longer to appear.That selectin seems to be most impor-tant about four hours after the start ofan infection, after which it graduallyfades away.

The mechanism that helps leuko-cytes to breech the endothelial barrieris indispensable to the fulÞllment oftheir infection-Þghting duties. Yet act-ing inappropriately, that same mecha-nism also allows leukocytes to accumu-late in tissues where they do not be-long, thereby causing tissue damage,swelling and pain.

The inßammation of rheumatoid ar-thritis, for instance, occurs when whiteblood cells enter the joints and releaseprotein-chopping enzymes, oxygen rad-icals and other toxic factors. Anotherexample is reperfusion injury, a disor-der that occurs after the ßow of bloodis temporarily cut oÝ from a tissue,such as during a heart attack. When theblood ßow resumes, the white bloodcells destroy tissues damaged by lackof oxygen.

The development of pharmaceuticalreagents that would inhibit adverse in-ßammatory reactions holds major in-terest for the academic, clinical and in-dustrial sectors. In theory, any drug thatinterferes with the adhesion of whiteblood cells to the endothelium, and con-sequently with their exit from the bloodvessel, should be anti-inßammatory. Thekey to developing such drugs is theshape of the binding regions of the se-lectin molecules and the shape of thecarbohydrates that Þt into them. Workon determining those shapes is pro-ceeding at a ferocious pace. In parallelresearch, intensive attempts are beingmade to synthesize carbohydrate in-hibitors of the P- and E-selectins.

For an antiadhesive therapy to be suc-cessful, the drugs must simultaneous-ly accomplish two seemingly incompat-ible ends. On the one hand, they muststop white blood cells from leaving thebloodstream inappropriately; on theother hand, they must still allow the

cells to go where they are needed. Thosegoals may be achievable because thespeciÞcities of adhesion molecules varyin diÝerent tissues. One can envision,for example, a drug that keeps whitecells from entering the joints but notother parts of the body.

Aside from their involvement in in-ßammation, cell-adhesion molecules may play a role in other

diseases, such as the spread of cancercells from the main tumor throughoutthe body. For example, the carbohydraterecognized by E-selectin is expressedon cells from diverse tumors, includingsome cancers. Bevilacqua has recentlyreported that at least one type of hu-man cancer cell binds speciÞcally to E-selectin expressed on activated endo-thelium. Perhaps to promote their ownmetastasis, some malignant cells re-cruit the adhesion molecules that arepart of the bodyÕs defenses.

If so, antiadhesive drugs may also turnout to be antimetastatic. One hopefulsign in that direction recently came fromHakomoriÕs group, which was studyinghighly metastatic mouse melanoma cellsthat carry a lectin for lactose, the sugarin milk. The researchers found that byexposing the melanoma cells to com-pounds containing lactose before in-jecting them into mice, they could re-duce the metastatic spread of the cellsalmost by half.

Although the importance of carbohy-drates in cell recognition is immense,other modes of recognition that rely ona peptide language do exist. Some forms

of attachment, for instance, involve sur-face proteins called integrins and com-plementary peptides. The existence ofmore than one system for binding activ-ities lends greater ßexibility to a cellÕsrepertoire of interactions.

Biomedical researchers are still striv-ing for a better understanding of thesugar structures on cell surfaces and ofthe speciÞcities that lectins have forthose structures. As they learn more,they will be in a better position to de-sign highly selective, extremely power-ful inhibitors of cell interactions. Theday may not be far oÝ when antiadhe-sive drugs, possibly in the form of pillsthat are both sugar-coated and sugar-loaded, will be used to prevent and treatinfections, inßammations, the conse-quences of heart attacks and perhapseven cancer.

CANCER CELLS have unusual carbohydrates on their surface, which may accountfor many of their invasive properties. Drugs that interfere with the adhesiveness ofabnormal cells may someday be used in cancer therapies.


FURTHER READINGLECTINS AS CELL RECOGNITION MOLE-CULES. N. Sharon and H. Lis in Science,Vol. 246, pages 227Ð234; October 13,1989.

GLYCOBIOLOGY: A GROWING FIELD FORDRUG DESIGN. Karl-Anders Karlsson inTrends in Pharmacological Sciences,Vol. 12, No. 7, pages 265Ð272; July1991.

CARBOHYDRATES AND GLYCOCONJU-GATES: UPWARDLY MOBILE SUGARS GAINSTATUS AS INFORMATION-BEARING MOL-ECULES. K. Drickamer and J. Carver in Current Opinion in Structural Biolo-gy, Vol. 2, No. 5, pages 653Ð654; Octo-ber 1992.


adioactive dating provides a pow-erful way to measure geologictime. It has revealed the overall

pace of the earthÕs evolution and has en-abled researchers to calculate that ourplanet is about four and a half billionyears old. Yet the earliest stages of ter-restrial historyÑduring which the earthacquired its large iron core and light,mobile continentsÑhave eluded easy in-vestigation because of the many process-es that act to reset the radioactive clock.As continents drift across the surface,the ocean ßoor between them recyclesinto the hot interior. Where continentsclash, they raise folded mountains. Hotunderlying material invades the conti-nental rocks and can break through, un-leashing lava that covers the surface.Erosion planes oÝ the mountains andsweeps sediments into ocean trencheswhere they, too, return to the mantle.

Earth scientists are now using increas-ingly sophisticated techniques to prymeaningful stories about the earliestevents in the earthÕs history from previ-ously taciturn rocks. Examinations ofancient minerals are revealing when theÞrst continents appeared and how ex-tensive they were. Workers are also un-covering evidence that plate tectonics

has operated throughout most of theearthÕs history much the same as it doesnow, contrary to some theoretical expec-tations. The recent discoveries are Þllingin the long-mysterious details of the for-mative years when our planet acquiredmany of its most characteristic traits.

To search for clues about the earthÕsyouthful nature, geophysicists make useof an assortment of radioactive datingmethods. These methods vary in theirstrengths and weaknesses, but they allrely on determining the relative abun-dances of a radioactive isotope and thesubsequent isotope, or daughter nucle-us, into which it decays. Every radioac-tive isotope eventually produces a Þnal,stable decay product. Knowing the rateat which the nuclear transformation oc-curs (which can be measured to highprecision in the laboratory) allows oneto infer how long the decay productshave been collecting in a rock. That in-formation, taken with other evidence,reveals much about geologic history.

In the ongoing search for the oldestcontinental remnants, researchersprimarily examine isotopes of ura-

nium. Uranium ultimately decays intolead, so the relevant dating technique iscalled the uranium-lead method. Thatapproach greatly beneÞts from the factthat samples of uranium and lead largeenough to analyze can usually be ex-tracted from zircon crystals. Such crys-tals are very commonly found in gran-itic and metamorphic rocks, as well as in some volcanic rocks and in sedimen-tary material derived from any of thoserocks. Zircons also resist heat andweathering strongly, so they may surviveintact in rocks that have experiencedone or more metamorphic episodes.

A potential problem with the urani-um-lead dating method is that rocks ex-posed to tremendous heat and pressuremay lose a signiÞcant amount of their


DEREK YORK has spent more thanthree decades reÞning the tools of geo-chronology in order to investigate theearthÕs distant past. York received hisD.Phil. from the University of Oxford in 1960, at which point he joined the department of physics at the University of Toronto. Last year he was appointedchairman of the department. York hasalso been extensively involved in thepopularization of science. He has fre-quently contributed news stories to TheGlobe and Mail and was a guiding forcebehind the recent Þlm Chaos, Scienceand the Unexpected, produced by the Ca-nadian Broadcasting Corporation.

The Earliest Historyof the Earth

Radioactive dating techniques have illuminated vaststretches of geologic history, bringing the mostancient eras of the earth’s evolution into view

by Derek York

ACASTA GNEISS, a group of metamor-phic rocks in northern Canada, is the


lead, thereby resetting the radioactiveclock. In 1956 George W. Wetherill ofthe Carnegie Institution of Washingtonshowed a way to circumvent the dif-Þculty. His procedure depends on thefact that there exist two radioactive iso-topes of uranium, uranium 238 anduranium 235. Each form of uraniumfollows its own decay path: uranium238 breaks down into lead 206, ura-nium 235 into lead 207. Therefore, forany uranium-bearing mineral, research-ers can derive an age estimate fromtwo sources.

Wetherill measured the two uranium-lead abundance ratios in a large numberof samples and plotted them againsteach other. On such a plot, samples that

have never been disturbed, and hencethat are perfect clocks, would lie alonga continuous curve that Wetherill calleda concordia curve. (The curve simplyreßects the fact that both uranium 235and uranium 238 decay at a steady,predictable rate.)

Wetherill then made the remarkablediscovery that by plotting abundanceratios he could determine the age of agroup of rock samples (all of the sameage), even if much of their lead hadleaked out during a metamorphic epi-sode. His method works because lead206 and lead 207 are chemically iden-tical, and equal fractions of the two isotopes would have escaped from therocks. When the abundance ratios of

uranium to lead in the rocks are mea-sured and plotted, the data points asso-ciated with the various samples fall on a straight line lying below the concor-dia curve. The end points of that line in-tersect the concordia curve at locationscorresponding to the time of crystalliza-tion and to the time of metamorphism[see box on next page ].

Applying the uranium-lead methodto zircons can be diÛcult because zir-con crystals frequently have a layeredstructure in which the original core iswrapped in subsequent mineral coat-ings. In the 1970s Thomas E. Krogh ofthe Royal Ontario Museum in Torontodemonstrated how to abrade zircons toisolate their cores; he also showed that


oldest known intact piece of the earthÕs surface. Radioactivedating indicates that the Acasta gneiss is nearly four billion

years old, proving that some continental material existed onlya few hundred million years after the earthÕs formation.



the uranium-lead ratios of the coresfrequently fall on WetherillÕs concordiacurve. Krogh concluded that the innerparts of the zircons had not been chem-ically altered and so still recorded thetime when the mineral Þrst crystallized.

During the 1980s, William Compstonand Steven Clement of the AustralianNational University in Canberra tookzircon dating one step further. Insteadof analyzing the entire zircon core allat once, as researchers had previouslydone, Compston and Clement soughtto study the compositionÑand hencethe ageÑof the zircon at many spots.The researchers constructed a devicethat could blast a sample with a tightlyfocused (only 25 microns wide) beam ofionized, or electrically charged, oxygenatoms. Compston and Clement calledtheir instrument SHRIMP, an acronymfor Super High-Resolution Ion Micro-Probe. They trained SHRIMP at one spoton the inside of a zircon crystal thathad been cut in half. The ions vaporizeatoms of uranium and lead from thatspot; the atoms then pass through amass spectrometer that separates themby mass and counts them.

In 1983 SHRIMP began to supply note-worthy new information about the ageof the earthÕs crust. Derek O. Froude of the Australian National University,working with Compston and others,began probing single crystals of zirconin quartzite, or metamorphosed sand-stone, at Mount Narryer in WesternAustralia. Earlier work had shown thatthis region contains rocks about 3.6billion years old. FroudeÕs group per-formed detailed analysis on 20 zirconcrystals from one rock specimen. Fourof the crystals yielded lead-to-uraniumdata points indicative of an age of 4.1to 4.2 billion years. Previously, the old-est known pieces of terrestrial materialwere rocks from southwest Greenland,which Stephen Moorbath of the Univer-sity of Oxford and his co-workers hadfound to be 3.8 billion years old. Theother 16 zircons furnished isotopic ra-tios that clustered around three linesthat intersected the concordia curve atages of about 3.75, 3.3 and 3.1 billionyears, respectively.

Froude inferred that the 4.1- to 4.2-billion-year-old zircons and the 3.75-billion-year-old specimens formed longbefore they were incorporated into thesurrounding sedimentary rocks. Some-how the zircons eroded from their par-ent rocks and found their way into sed-iments that later, under enormous heatand pressure, became the Mount Narry-er quartzite. The younger, 3.3- and 3.1-billion-year-old zircons probably start-ed growing during that period of meta-morphism. Because zircons are found

How Uranium-Lead Dating Works

atural minerals such as zircons contain two isotopes of radioactive ura-nium: uranium 235, which decays into lead 207, and uranium 238,

which decays into slightly lighter lead 206. In undisturbed zircons, plotscomparing the two uranium-lead abundance ratios fall on a curve known asthe concordia curve; their locations on the curve indicate the age of eachsample. The uranium-lead data for zircons that underwent a metamorphicepisode, thereby losing some of their lead, will fall along a line intersectingthe concordia curve at two points (top). The upper point represents the orig-inal time when the rocks crystallized; the lower point indicates the time ofmetamorphism. Zircons from Mount Narryer in Australia all show somesigns of disturbance (middle). Judging from how they fall on the concordiadiagram, the oldest Mount Narryer zircons seem to be 4.1 to 4.2 billionyears old; the others form three families 3.1, 3.3 and 3.75 billion years old.Based on similar reasoning, zircons from the Acasta gneiss appear to havecrystallized about 3.96 billion years ago (bottom).

LEA

D 2

06/U

RA

NIU

M 2

38LE

AD

206

/UR

AN

IUM

238

500 10 20 30 40

1.0

0.8

0.6

0.4

0.2

060 70

1.0

0.8

0.6

0.4

0.2

0500 10 20 30 40 60 70

LEAD 207/URANIUM 235

SAMPLE CONCORDIA CURVE

TIME OF CRYSTALLIZATION

TIME OF METAMORPHISM

ZIRCON CORES

1.5

2.0

3.0

LEA

D 2

06/U

RA

NIU

M 2

38

500 10 20 30 40

1.0

0.8

0.6

0.4

0.2

060 70

AGE (BILLIONS OF YEARS)

MOUNT NARRYER ZIRCONS

2.5

3.5

4.0

4.5

2.0

3.0

4.0

ALTERED ZIRCONS

OLDEST ZIRCONS

2.0

3.0

4.0ACASTA ZIRCONS

N


overwhelmingly in continental ratherthan oceanic rocks, the AustraliansÕ Þnd-ing strongly suggested that at least somecontinental material existed more thanfour billion years ago. Unfortunately, thezircons seem to be the only survivingrelics of those ancient rocks.

In 1989, Samuel A. Bowring, then atWashington University, along withIan S. Williams of the Australian Na-

tional University and Compston, provedthe existence of intact rocks nearly asold as the Australian zircons. The re-searchers trained SHRIMP on zirconsfrom the Acasta gneiss, a small patch ofmetamorphic rock southeast of GreatBear Lake in the Northwest Territories.Bowring and his colleagues had earlierused KroghÕs abrasion technique to de-termine that some zircons in the Acastagneiss are at least 3.8 billion years old.Bowring suspected that buried insidethe individual zircons lay evidence ofeven older ages; such information pos-sibly could be extracted using SHRIMP.He therefore ßew to Canberra carryingzircons from two Acasta rock samples.

The researchers probed a total of 82spots on 53 of the zircons. When dis-played on a Wetherill chart, the urani-um-lead ratios from the two samplesformed fan-shaped plots lying near theconcordia curve. One plot fell between3.6 and 3.96 billion years, the other be-tween 3.8 and 3.96 billion years. Bow-ring and his co-workers concluded thatthe oldest zircons recorded the origi-

nal crystallization age of the rocks. Thespread in the data probably indicatesthat the zircons underwent at least twometamorphic episodes, one during theÞrst few hundred million years aftercrystallization and another about twobillion years ago.

If that interpretation is correct, thenthe Acasta gneiss is the metamorphosedremains of the oldest known intact, sol-id rock on the earthÕs surface. Geolo-gists have identiÞed other rocks nearlyas old in Greenland, Labrador and West-ern Australia. Lance P. Black of the Bu-reau of Mineral Resources in Canberraand others, also using SHRIMP, recent-ly reported Þnding 3.87-billion-year-oldzircons in Antarctica as well.

These results leave little doubt thatat least small patches of continentalrock existed on the earthÕs surface dur-ing the Þrst 700 million years of its his-tory. They also underscore the incredi-ble scarcity of crustal rocks more thanfour billion years old. Plate tectonicsalone may not suÛce to explain whysuch rocks are so rare. Perhaps theearth had an extensive primeval crustthat was destroyed and remixed intothe interior by the impact of giant me-teorites, the leftovers from planetaryformation. Vigorous convection, drivenby the great internal heat of the new-born planet, might have aided that dis-ruption by tearing apart blocks of con-tinental rocks and by pulling continen-tal sediments into the hot depths.

On the other hand, evidence is mount-

ing, based on geochemical arguments,that the total amount of continentalcrust before about four billion yearsago was minuscule. Studies of the rela-tive abundances of isotopes of neody-mium, strontium and lead in continen-tal and oceanic crust imply that no morethan trivial quantities of continentalcrust existed before then. About 3.8billion years ago the earthÕs mantle be-gan to separate into lighter and densercomponents, releasing the raw materialfrom which continental blocks formed.The continents seem to have continuedgrowing rapidly until roughly 2.5 bil-lion years ago.

What were the earthÕs internaldynamics like during that eraof swiftly expanding conti-

nents? My group in Toronto, in collab-oration with Alfred Kr�ner of Guten-berg University in Mainz and MichaelO. McWilliams of Stanford University,has been trying to address that ques-tion. We do so by assessing the extentof early continental drift. Geophysicistscan trace recent continental motions bymeans of the magnetic record preservedin the ocean ßoor. Ocean ßoors surviveonly about 200 million years beforethey sink back into the earthÕs mantleat oceanic margins, such as along thePaciÞc trench oÝ Asia.

Determining the motions of the con-tinents more than two billion years agodemands extending the tools of geo-chronology to include measurements


SEARCH FOR THE OLDEST ROCKS led Samuel A. Bowring(left ), now at the Massachusetts Institute of Technology, tothe Acasta gneiss. Bowring and his collaborators collectedzircons from those rocks (right ) and polished them to reveal

their inner structure. The researchers then measured the lead-to-uranium ratios at various spots in the zircons to Þnd theoldest parts. The shallow pits were produced by beams ofions used to vaporize bits of the zircon crystals for analysis.


of a rockÕs internal magnetism. Whenlavas erupt or when granites form inthe outer layers of the earth, iron ox-ides in the rock become magnetized in the direction of the earthÕs magneticÞeld at that site (the iron oxides act astiny bar magnets that point toward thenorth pole). Measuring the direction ofthe Þeld frozen into the rock revealshow far it was from the magnetic polewhen it cooled. By studying the magnet-ism of rocks of varying ages, all fromthe same general site, one can, in princi-ple, learn how much the continent hasmoved toward or away from the poleover time.

Unfortunately, if a rock is heatedabove a critical temperature, it loses its original magnetic direction and be-comes remagnetized the next time itcools. The new direction may be total-ly diÝerent from the original if the con-tinent has drifted signiÞcantly in lati-tude during the intervening years. If therock was not severely reheated, how-ever, some of the original magnetic ori-entation remains. In that case, geophys-icists can extract from the rock a rec-ord of two ancient pole positions: oneat the time of crystallization or initialcooling, the other at the time of meta-morphic heating. Because all knownPrecambrian rocks have suÝered someheating episode, it is critical to uncovera rockÕs thermal history to decode itstrue magnetic record.

A valuable radioactive dating methodknown as the potassium-argon tech-nique can sometimes permit research-ers both to determine the age of mag-netized rocks and to assess whether

(and how much) they have been heatedduring their lifetime. Potassium 40, arare isotope of potassium, decays toproduce argon 40, a heavy version ofthe unreactive gas argon; the half-life ofpotassium 40 is 1.3 billion years. Fromthe accumulation of argon 40 in potassi-um-bearing minerals, one can determinehow long ago the mineral solidiÞed.

Potassium-argon dating has assistedgeochronologists in outlining the timescale of biological evolution over thepast 500 million years. It has provedless appropriate for probing the earthÕsmore distant history because argontends to leak from minerals duringtimes of metamorphic heating. The ura-nium-lead method, along with a some-what less widely used approach involv-ing isotopes of rubidium and strontium,is more eÝective for revealing crystal-lization ages of the oldest rocks.

The ease with which the potassium-argon clock can be disturbed oÝers acompensating advantage: it enables re-searchers to unravel a rockÕs thermalhistory. That information in turn makesit possible to make sense of the rockÕsmagnetic history. Geochronologistshave studied many diÝerent kinds ofminerals to determine how readily theyrelease entrapped argon when heated.The common mineral hornblende hasbeen shown to be quite resistant to lossof argon. Usually, only severe heating(above about 500 degrees Celsius) suf-Þces to let some of the argon escape.The minerals muscovite and biotite, twoforms of mica, are somewhat less resis-tant to heat; they can be disturbed bytemperatures in the range of 250 to 350

degrees C. At the other extreme, themineral feldspar shows signs of argonloss below 200 degrees C.

Workers searching for evidence ofcontinental drift during the Þrst half ofthe earthÕs history have concentratedmuch of their attention on a series ofremarkably well preserved rocks in theBarberton Mountain Land greenstonebelt, located on the border betweenSouth Africa and Swaziland. The rocksare part of the Kaapvaal Craton, a sec-tion of stable, extremely old continentalcrust. The greenstone belt consists ofnumerous volcanic rocks buried under-neath later sediments. Younger, granit-ic rocks have forced their way into theolder greenstone layers.

My co-workers and I have lookedfor signs of tectonic drift morethan two billion years ago by

carrying out detailed magnetic analysesand by conducting both potassium-ar-gon and uranium-lead zircon dating oftwo samples of granite from the Kaap-vaal Craton. In Mainz, Kr�ner appliedthe uranium-lead method to single crys-tals of zircon from lavas in the cratonand showed that they crystallized about3.5 billion years ago, in broad agree-ment with earlier measurements madeat the University of Cambridge. He alsofound that granites from the Nelshoot-ge region of the Kaapvaal Craton crys-tallized about 3.2 billion years ago; gran-ites from the nearby Mbabane regionÞrst cooled 2.69 billion years ago.

Paul W. Layer, Margarita Lopez-Mar-tinez and I performed potassium-argondating in my laboratory at the Universi-


Using a Laser to Illuminate the Past

aser step-heating offers an efficient way to measure both the age and thethermal history of a mineral sample. In this approach, the age is measured by

the degree to which radioactive potassium has decayed into argon 40. In thefirst step, the sample is bombarded by neutrons, which transform some of thepotassium into argon 39 (left). An intense laser beam heats the sample and re-leases both isotopes of argon; a mass spectrometer measures the exact abun-dances of the isotopes. As the laser heat increases, more and more argon is

L

NEUTRON BOMBARDMENTOF MINERAL

ARGON39

MASS SPECTROMETER

LASERBEAM

POTASSIUM

ARGON39

ARGON40

ARGON40

AG

E

FRACTION OF ARGON RELEASED

0 1

1 2 3 4

PLATEAU AGE


ty of Toronto to examine the thermalhistory of the rocks as the Þrst step ininterpreting their magnetic record. Weused a clever variant of the potassium-argon method, known as argon-argondating, that Craig M. Merrihue of theUniversity of California at Berkeley hadoriginally proposed. In MerrihueÕs ap-proach, one irradiates a rock samplewith neutrons from a nuclear reactor,transmuting some of the potassium intoargon 39. The sample is then melted ina vacuum chamber, freeing both the ar-gon 39 and argon 40. Instead of makingseparate measurements of the potassi-um and argon abundances in the sam-ple, a researcher need make only a sin-gle observation of the two argon iso-topes to derive the age of a sample.

Merrihue realized that his techniquecould also provide vital informationabout the thermal history of the rockbeing studied. He suggested that rath-er than melting the irradiated mineral all at once, one could heat the samplethrough a series of steps and measurethe ratio of argon 40 to argon 39 (anddeduce an age corresponding to thatratio) at each temperature. If every stepgives the same age, one can reasonablyassume that the sample has suÝeredlittle thermal disturbance during its his-tory. If the measured ages begin low atcool laboratory temperatures and climbto a plateau of greater ages at hotterones, then some of the argon 40 musthave been lost from the mineral duringone or more heating events. The age ofthe plateau takes into account all theargon 40 that is released only underÞerce heat and so should render a good

indication of the time when the rockÞrst cooled and the potassium-argonclock switched on.

At Toronto, my collaborators and Iapplied the step-heating method to theNelshootge and Mbabane granites, aswell as to lavas related to the volcan-ic rock whose age had been measured by Kr�ner using the zircon method. Wealso used another innovation: a laser-heating technique I developed at To-ronto in collaboration with Chris M.Hall, now at the University of Michigan,and Yotaro Yanase. A similar techniquehad been pioneered in the 1970s byGeorge H. Megrue of the SmithsonianInstitution. He showed that it is possi-ble to measure the argon-argon age ofa mineral by vaporizing part of a sam-ple with a pulsed laser. His method didnot gain wide acceptance at the time,because he could not modulate the la-serÕs power to heat the specimen inneat, steplike increments.

By modifying MegrueÕs technique,my colleagues and I succeeded inusing a continuously beamed la-

ser to produce a precise age spectrumfrom a single grain of rock. We beginby bathing a bit of mineral in a low-power laser beam for 30 seconds andanalyzing the released argon by meansof a mass spectrometer. Then we in-crease the power of the laser beam in aseries of increments and derive an ageat each step. When the mineral fuses, ithas reached its crystallization tempera-ture, and the age spectrum is complete[see box above].

The combination of laser heating and

argon-argon analysis is supplying im-pressively accurate information aboutthe age and the thermal histories of an-cient rocks. That information has en-abled us to reconstruct an unprecedent-edly well documented account of conti-nental drift as it occurred billions ofyears ago. We began by considering thehistory of the Mbabane granite. Layerand McWilliams had discovered thatthis rock contains a magnetic image ofthe pole, indicating that the rock solidi-Þed somewhere near the earthÕs equa-tor. Further analysis showed that theMbabane granite acquired its magneticorientation when the rock cooled fromaround 600 to 500 degrees C.

Laser argon-argon dating of fourhornblende crystals in the granite, con-ducted by my group at Toronto, yield-ed an age spectrum displaying a broadplateau of 2.69 billion years, essentiallyidentical with the uranium-lead age ofthe surrounding zircons derived by Kr�-ner. The close agreement of the two dat-ing techniques implies that the horn-blendes, and hence the whole granite,had never been heated to temperaturesabove 500 degrees C after they Þrstcooled. We concluded that the magnet-ic pole position detected in the granitewas recorded when the rocks initiallycooled and solidiÞed some 2.69 billionyears ago, at which time the Mbabanegranite lay at the equator.

Layer and his colleagues had alreadydeduced that another group of igneousrocks now located 12 kilometers fromthe Mbabane granite sat a little morethan 30 degrees from the equator 2.875billion years ago. If those two forma-tions have always been close neighbors,then that area of the Kaapvaal Cratonmust have drifted through 30 degreesof latitude between 2.875 and 2.69 bil-lion years ago. Such movement impliesa drift rate of about 1.5 centimeters ayear, comparable to the speed at whichNorth America has drifted away fromthe Mid-Atlantic Ridge during the past100 million years. We naturally won-dered if plate tectonics occurred at thesame pace during even earlier times.

Kr�ner had shown that the Nelshoot-ge granite is 3.21 billion years old. La-ser analysis of individual hornblendecrystals in the granite suggested thatthe rock acquired the magnetizationearly in its history, at least 3.18 bil-lion years ago. When we examined the orientation of the magnetic Þeld in thegranite, we found that it, too, seemed tohave formed 90 degrees away from thepole, that is, on the equator. But the di-rection of the North Pole recorded in theNelshootge granite lay many degreesaway from the pole position frozen intothe Mbabane granite, so the Kaapvaal


forced out (steps 1–4, center ); at each step, an age is derived from the ratio of argon 40 to argon 39. Low ages at cool temperatures indicate that some ar-gon leaked out during the sample’s history; the plateau age at hotter tempera-tures should reflect the sample’s true age. At the right, laser step-heating of fourhornblende grains from the Mbabane granite in Swaziland shows they are 2.7billion years old, in close agreement with the age derived from uranium-leadmeasurements. Analysis of komatiite lavas from the same region reveals thatthey first cooled about 3.5 billion years ago.

AG

E IN

DIC

AT

ED

(BIL

LIO

NS

OF

YE

AR

S)

FRACTION OF ARGON RELEASED0.2 0.4 0.6 0.8 100

2.0

2.6

3.0

3.6

2.6

3.0

3.6

FRACTION OF ARGON RELEASED0.2 0.4 0.6 0.8 10

KOMATIITE LAVASMBABANE GRANITE

URANIUM-LEAD AGE


Craton must have rotated considerablyduring the intervening years.

Finally, we moved on to examine the oldest Barberton rock in our col-lection, fragments of magnesium-richlavas known as komatiites. Zircons involcanic rocks related to those lavasdate from almost 3.5 billion years ago.David J. Dunlop and Chris J. Hale ofthe University of Toronto recovered aweak magnetic Þeld frozen into the ko-matiites, indicating that when they Þrstcooled, the lavas were only about 18 de-grees from the poleÑnowhere near theequator. The signiÞcance of this Þnd-ing was unclear, however. The lavas ev-idently have been exposed to intenseheat and pressure subsequent to theirformation; they preserve few if any oftheir original minerals, so the magneti-zation observed by Dunlop and Halecould be much younger than 3.5 bil-lion years.

To clarify the matter, Lopez-Martinez,then a graduate student in my laborato-ry, carried out an argon-argon measure-ment to determine the age of the ko-matiite samples. Much to her surpriseÑand to mineÑher measured ages wereindistinguishable from those that Kr�-ner had found by studying zircons inthe lavas. Zircons date the initial crys-tallization of the lavas, whereas argon-argon analysis measures the age oftremolite, a mineral that forms duringmetamorphism. The only way to makesense of Lopez-MartinezÕs result is toconclude that the metamorphic eventthat created the tremolite must haveoccurred almost immediately after thelavas Þrst erupted. If so, the pole loca-tion in the Barberton komatiites is agenuine recording of the location ofthe rock nearly 3.5 billion years ago.

Pulling together the above evidenceyields an 800-million-year history ofthe drift of the Kaapvaal Craton, begin-ning 3.5 billion years ago. At that time,the craton was near the pole. By about3.18 billion years ago it had wanderedto the equator. The craton then movedmore than 3,000 kilometers poleward,so that 2.875 billion years ago it lay atleast 30 degrees from the equator. By2.69 billion years ago the craton haddrifted back to the equator, but its ori-entation was signiÞcantly diÝerent thanit had been 490 million years earlier.

Our work implies that the Kaap-vaal Craton was drifting aboutas fast as do modern continents

since at least 3.5 billion years ago. Pre-sumably, other continental fragmentswere behaving the same way. I shouldemphasize that despite the many ad-vances in methodology and instrumen-tation, studies of magnetic traces stillleave considerable room for error. Nev-ertheless, uranium-lead and argon-ar-gon dating, when combined with paleo-magnetic studies, have very much bol-stered understanding of the earthÕsdynamic early history.

The success of our investigation ofthe Barberton komatiites led us to won-der if the very oldest known rocks, theAcasta gneiss, carry a snapshot of theirposition relative to the magnetic pole3.96 billion years ago. Much to our re-gret, they seem not to. In my labora-tory, Hall performed preliminary ar-gon-argon dating of hornblende crys-tals from one of BowringÕs samples.His results show that the rocksÕ ar-gon-argon clocks were reset by a pulseof tectonic activity 1.8 billion years ago. Any 3.96-billion-year-old magnetic

traces evidently would have been com-pletely erased.

Yet hope persists. Perhaps furthermeasurements of rocks from the sameregion in northern Canada may yield a small fragment that escaped severereheating during mountain-buildingevents. The magnetic record in such arock would be invaluable for decodingthe earthÕs evolution. Evidence that theearth possessed a signiÞcant magneticÞeld at that early date would be a tellingsign that it had already developed asizable metallic coreÑa key step in thetransformation of an infant planet intoa well-ordered, complex world.

0°

NELSHOOGTEGRANITE

MBABANEGRANITE

80°N

KOMATIITELAVAS

USUSHWANACOMPLEX

60°N

30°N

30°S

72°NLA

TIT

UD

E

3.5 BILLION 3.2 BILLION 2.9 BILLION 2.7 BILLIONAGE (YEARS BEFORE PRESENT)

MODERNAFRICA

KAAPVAALCRATON(ANCIENTLAND MASS)

ANCIENT CONTINENTAL DRIFT is recorded by magnetic Þeldsfrozen into four kinds of rocks found in the Kaapvaal Craton,a region of stable, extremely old continental crust (the Africancontinent, shown for reference, did not then exist in its mod-

ern form). The magnetized rocks act as compasses that revealthe cratonÕs orientation and latitude at the time when the rockscrystallized. The rate at which the Kaapvaal Craton wanderedbroadly resembles the present pace of plate tectonics.


FURTHER READINGCOOLING HISTORIES FROM 40AR/39ARAGE SPECTRA: IMPLICATIONS FOR PRE-CAMBRIAN PLATE TECTONICS. D. York in Annual Review of Earth and Plane-tary Sciences, Vol. 12, pages 383Ð409,1984.

THE DERIVATION OF 40AR/39AR AGESPECTRA OF SINGLE GRAINS OF HORN-BLENDE AND BIOTITE BY LASER STEP-HEATING. P. W. Layer, C. M. Hall and D. York in Geophysical Research Let-ters, Vol. 14, No. 7, pages 757Ð760; July1987.

ELEMENTS OF THE ARCHEAN THERMALHISTORY AND APPARENT POLAR WAN-DER OF THE EASTERN KAAPVAAL CRA-TON, SWAZILAND, FROM SINGLE GRAINDATING AND PALEOMAGNETISM. P. W.Layer, A. Kr�ner, M. McWilliams and D.York in Earth and Planetary Science Let-ters, Vol. 93, No. 1, pages 23Ð34; May1989.

CRUST FORMATION AND PLATE MOTIONIN THE EARLY ARCHEAN. A. Kr�ner andP. W. Layer in Science, Vol. 256, pages1405Ð1411; June 5, 1992.


rom dense rain forests to broadcoastal plains, from deciduousthickets to desert, Madagascar of-

fers an extraordinary range of environ-ments. These habitats harbor an equallyextraordinary primate fauna that showsmore clearly than any other what ourown primate ancestors were like earlyin the Age of Mammals, approximately50 million years ago.

These animals are the lemurs, Mada-gascarÕs dominant mammals. Why theyare there is not entirely clear. The sto-ry used to be that the 1,000-mile-longMadagascan minicontinent had simplypreserved (in somewhat impoverishedfashion) an archaic fauna that was ma-rooned on it when Madagascar sepa-rated from the African mainland anddrifted out to sea. In many ways the islandÕs living ÒlowerÓ primates moreclosely resemble the primates of theEocene epoch (about 57 to 35 millionyears ago) than they do the ÒhigherÓprimates dominating the tropical conti-nents today. That fact was taken to im-ply that the separation had occurred inthe Eocene or thereabouts.

But we now know that Madagascarbegan its journey away from Africa asmuch as 165 million years ago, whenthe dinosaurs ruled and the only mam-mals were tiny and vaguely shrewlike.Moreover, the island appears to havereached its present separation of rough-

ly 250 miles from Africa some tens ofmillions of years before the great diver-siÞcation of the mammals and thus wellbefore todayÕs familiar groups such asthe primates, bats and rodents cameinto existence.

For land-bound latecomers such asprimates, then, the only possible meansof access to Madagascar was by Òraft-ingÓ: ßoating across the MozambiqueChannel from Africa on matted tanglesof vegetation. Arriving thus, the ances-tral lemurs would have found an in-credible wealth of ecological opportu-nities on an island that is only slightlysmaller than Texas and topographical-ly, climatically and ecologically muchmore diverse.

The island of Madagascar looksrather like a giant left footprintin the sea, its long axis oriented

more or less north-south. It extendsfrom within 12 degrees of the equatorall the way to the southern subtropicalzone. Its eastern side presents a steepescarpment to the prevailing easterlywinds; here heavy rainfall year-roundsupports a dense growth of rain forest.To the west, a rugged central plateauslopes more gently to broad coastalplains that become drier toward thesouth. Moist forests in the northwestgive way to deciduous forests and thick-ets. These in turn yield in the far southto an extraordinary desert-adapted ßorain which as many as 98 percent of thespecies are unique. Add to these majorregions a host of local microclimatesand secondary physiographic features,and you have an unparalleled assort-ment of environments for forest-livingmammals to exploit.

No one knows what the range ofhabitats was in Madagascar at the re-mote time when primates Þrst arrivedor what lived in those habitats; the fos-sil record is simply lacking. What is cer-tain, though, is that the primates ßour-ished and that by the time human be-ings arrived under 2,000 years ago theisland was home to at least 45 lemur

species. These primates ranged in bodysize from the two-ounce mouse lemur,Microcebus, to the 400-pound Archaeo-

indris, a match for a large gorilla.Each of the lemur species was a low-

er primate, belonging (with the bush ba-bies, pottos and lorises) to the primatesuborder Strepsirhini. We, on the oth-er hand, are higher primates, belonging(with the monkeys and apes) to the sub-order Anthropoidea. (There is no con-sensus about which group the tiny andenigmatic tarsier of Southeast Asia be-longs to.) The distinction between lowerand higher primates is in fact a ratherarchaic concept that is now going outof fashion, but it is a convenient one touse here.

The higher primates appeared on theevolutionary scene much later than didthe lower primates, from one of whichthey presumably evolved toward theend of the Eocene. MadagascarÕs lemurs


IAN TATTERSALL is a paleontologistand primatologist who has worked ex-tensively on the living and subfossil le-murs of Madagascar as well as on a rangeof problems in early primate and humanevolution. Curator and chairman in thedepartment of anthropology of the Amer-ican Museum of Natural History, Tatter-sall is especially interested in the inte-gration of evolutionary theory with thefossil record. He has also served as cura-tor for several major exhibitions at themuseum, the most recent of these beingthe Hall of Human Biology and Evolu-tion, scheduled to open in early 1993.

MadagascarÕs LemursThese primates can tell us a great deal about our own

evolutionary past. But many species are already extinct,and the habitats of those that remain are shrinking fast

by Ian Tattersall

TROPIC OF CANCER

EQUATOR

TROPIC OF CAPRICORN

A F R I C A

MADAGASCAR

COQUERELÕS SIFAKAS, a variety of le-mur, rest in a forest in northwest Mada-gascar (opposite page). These lower pri-mates evolved from ancestral lemursthat may have traveled on rafts of veg-etation across the Mozambique Channellong after Madagascar separated fromAfrica. The 1,000-mile-long island liesabout 250 miles oÝ the coast (above).


VERREAUX’S SIFAKA Propithecus verreauxi verreauxi

RUFOUS MOUSE LEMUR Microcebus rufus

GOLDEN CROWNED SIFAKAPropithecus tattersalli

RED RUFFED LEMURSVarecia variegata rubra

RINGTAILED LEMURS Lemur catta GOLDEN BAMBOO LEMUR Hapalemur aureus

and their continental cousins have muchin common with Eocene forms, retaininga suite of physical characteristics thathave been lost among higher primates.But only in Madagascar do we still Þndlower primates that are diurnal, or ac-tive during the day. Virtually all mod-ern higher primates are diurnal, and ifanything in paleontology is certain, it is that all anthropoids are descendedfrom a diurnal common ancestor. So ifwe wish to Þnd analogies to our remoteEocene ancestors, we must turn to theprimates of Madagascar.

Modern lower and higher pri-mates are distinguished fromone another by a number of

structural features, most signiÞcantlyin their nervous systems and sense organs. The lower primates have muchsmaller brains relative to body sizethan do higher primates. They also dif-fer in the development of the associ-ation areas, which govern the transferof information between the variousbrain centers.

The balance between the senses of vi-sion and smell also diÝers. Although

the eyes of lower primates are quiteforward-facing, the left and right visualÞelds do not overlap as much as theydo in higher primates. This arrangementlimits depth perception to the centralpart of the Þeld of view. And whereas itis natural that the nocturnal lower pri-mates lack color-sensitive cone cells intheir retinas, what little is known aboutvisual discrimination in the diurnal le-murs suggests that their color vision isat best limited.

As for the sense of smell, the lowerprimates have roomier nasal cavitiesthan do higher primates, with morecomplex internal structures. Living low-er primates retain the primitive mam-malian rhinarium, or Òwet nose.Ó It ispart of a system for the transfer of par-ticles to the nasal cavity, where they areanalyzed by an organ that is, at most,vestigial in most higher primates.

Many lower primates (but only a fewSouth American monkeys among thehigher primates) have scent glands thatexude substances used to ÒmarkÓ theenvironment. This process is importantin communication between individuals.Visual cues are less important; the fac-


BLACK LEMUR, FEMALE Eulemur macaco macaco

BLACK LEMUR MALEEulemur macaco macaco

MONGOOSE LEMUR, FEMALE ANDINFANT Eulemur mongoz

BLUE-EYED LEMUR, FEMALE ANDINFANT Eulemur macaco flavifrons

es of lower primates lack the muscula-ture needed to produce the complexexpressions through which higher pri-mates convey their states of mind.

Living primates in general have lostthe primitive claws that enabled theÞrst pre-Eocene primates to scale treeswithout the ability to grasp. In theirstead is a thumb, shifted away from the other digits, that is at least to someextent opposable to them, and sensitivetactile pads on the Þngertips, which arebacked by ßat nails. This fundamentalchange has consequences that extendfar beyond locomotion as well as intomanipulative abilities. But whereas high-er primates generally manipulate ob-jects using the thumb in opposition tothe other digits, lemurs tend to pickthings up with the whole hand. An itemheld in this way is then more likely tobe sniÝed rather than inspected visual-ly and turned in the Þngers for furtherexamination.

Both the lower and the higher pri-mates are, of course, extremely diversegroups, whose members represent myr-iad variations on the themes outlinedabove. Nevertheless, the lower primates

are obviously more primitive than thehigher primates (with respect to thebody systems discussed) in the sensethat they more closely resemble the an-cestor from which both modern groupsemerged. Although fossils are limitedto bones and teeth, they reveal clearlythat in the Eocene primate brains wereeven smaller in relation to body sizethan are those typical of modern le-murs and that the visual sense had notachieved the total domination over ol-faction we see in todayÕs higher pri-mates. Eocene primate hands and feetwere certainly capable of grasping butcould probably manipulate objects nomore precisely than lemurs do at pres-ent. In other words, as functioning or-ganisms, Eocene primates were proba-bly not too diÝerent from todayÕs le-murs. They were certainly close enoughso that by studying the lives of lemurswe can glimpse something of the Eo-cene behavioral potential from whichour vaunted human capacities ultimate-ly arose.

Not until the 1960s did informationabout the behavior of lemurs begin to beavailable from Þeld studies in Madagas-

VERREAUX SIFAKA, FORSYTHMAJOR’S VARIANT Propithecusverreauxi verreauxi

CROWNED LEMUR, FEMALE* Eulemur coronatus*Sex is indicated if male and female coloring differs


car. The results of those studies havebeen increasingly at odds with earlieranticipations. For example, as long as itwas possible to think of Òthe lemurÓ asa general ancestral model, the expecta-tion, if any, was simple: lemurs wouldshow a set of relatively stereotypical be-havior patterns, similar to those exhib-ited by early primates and from whichthe higher primates had managed toemancipate themselves. Yet perhaps themost remarkable revelation of the Þeldstudies is how diverse the lemurs arein their ways of life.

Of course, certain predictions fromanatomy were indeed borne out. Olfac-tory marking, using urine and feces as well as the secretions of specializedglands, has been shown to be a signiÞ-cant component of lemur behavior. Tojudge by its wide occurrence amongmammals, such marking is an ancientbehavior, and few would question itsimportance in communication amongEocene primates.

Lemurs also turn out, as expected,

to have a tendency to exploretheir environments with their nos-

esÑto choose ripe fruit, for example, by smell rather than by appearance.Undoubtedly, then, olfaction is of thegreatest importance to members of allÞve living families of lemurs, as it wasto their Eocene precursors, and muchmore so than it is to most higher pri-mates. But does the reverse apply to vision? Anyone who has watched a si-faka lemur hurtle through the forestwould conclude that the animal is hard-ly handicapped by its lack of a higherprimateÕs visual system.

Lemurs, in fact, turn out to be as di-verse as or even more so than the high-er primates in most aspects of their be-havior and ecology. Their diets, for in-stance, consist of much the same itemsas those consumed by the higher pri-mates: fruit, ßowers, leaves, buds andinsects. But feeding on nectar, which isunusual in higher primates, is being increasingly noted. As a whole, lemurstend to be dietary generalists, but someat least are highly specializedÑfor ex-ample, the bamboo lemurs. Many ofthem concentrate not just on bam-boo but on particular parts of bambooplants. Indeed, individuals of the speciesHapalemur aureus, weighing only a cou-ple of pounds each, daily eat shootscontaining enough cyanide to kill half a dozen humans.

Similarly, lemurs occupy nearly all ofthe vegetational habitats exploited byhigher primates, from rain forest to aridbrush. Again, whereas almost all higherprimates are diurnal, some lemurs arenocturnal, others diurnal and yet oth-

ers Òcathemeral,Ó spreading their activi-ty fairly evenly between daylight anddarkness.

The range of types of social organi-zation among lemurs is enormous.Some lead more or less solitary lives, inwhich small female ranges are over-lapped by larger male ones. In certainspecies, adult pairs rear immature oÝ-spring, whereas in others, small groupsconsist of just a few adults of each sex.And still other species live in ßuidgroups or in more stable larger unitsthat contain a couple of dozen individ-uals or more.

Within these broad categories, thereare still further variations. What is per-haps most surprising is that even with-

in the same species, substantial varia-tion in social organization may be foundfrom place to place. What seems to bemost signiÞcant is that even thoughlemur brains tend to be smaller thanthose of higher primates, at least somelemurs display the complex socialitywe usually associate with the higherprimates.

The picture that is therefore begin-ning to emerge of the behavioral vari-ety of the lemurs has important im-plications for the understanding of ourown Eocene ancestors. Despite the re-tention of such primitive behavioraltraits as olfactory marking, the widevariety suggests that from the begin-ning, well before the increase in brain


EXTINCT LEMURS include Archaeolemur (two are shown at bottom), Megalada-pis (upper left ) and Palaeopropithecus (upper right ). Archaeolemur was about thesize of a female baboon and was adapted to life on the ground. In contrast, Megal-adapis, weighing as much as 170 pounds, was arboreal and seems to have hadadaptations similar to those of the Australian koala. Palaeopropithecus, which


size that we associate with the higherprimates today, primates showed a be-havioral ßexibility and adaptability thatbelie the inference most people woulddraw from the description of these ear-ly forms as Òprimitive.Ó It seems to methat this is the essential evolutionaryheritage of our order, something farmore important than any of the ana-tomic characteristics to which we arewont to draw attention. As higher pri-mates ourselves, we tend to look at thelemurs and ask why they did not evolvefurther in our direction. But in doingthis, we are missing the critical pointthat, as a whole, the lemurs have actu-ally inherited from their Eocene ances-tors, as we have from ours, the most

signiÞcant primate characteristic of all.Looking at this picture from another

perspective, it is certainly clear that thelemurs had no need of higher primatephysical characteristics to make fulluse of the varied ecological opportuni-ties that Madagascar oÝered them. Per-haps this point is emphasized most dra-matically by considering the full rangeof lemur species that existed on Mada-gascar when humans Þrst arrived there.The present range is impressive enough:at the small end is the tiny two-ouncemouse lemur and at the large end, the15-pound babakoto. But in its pristinestate, the island was home to a spec-trum of primate types that equals, if itdoes not surpass, the entire variety

achieved by anthropoids in the rest ofthe world.

Late in the 19th century, excavationsat sites in the central plateau of Mada-gascar began bringing to light the sub-fossil, or partially fossilized, remains oflarge-bodied extinct lemurs that wereclearly not of great age. We know of atleast 15 species of subfossil lemurs, be-longing to eight or possibly more diÝer-ent genera. All of them are larger thanany surviving lemurs. The same patternof large body size is found among thenonprimate members of the subfossilfauna, including the Òelephant bird,ÓAepyornis maximus (the largest birdthat ever lived, possibly weighing almosthalf a ton), the pygmy hippopotamusand a giant tortoise.

A long series of studies has revealedan astonishing array of locomotor andpositional behaviors among the subfos-sil lemurs. Even by themselves, the liv-ing lemurs show a great variety of suchbehaviors, ranging from the rapid, scur-rying quadrupedal motion of the tinyMicrocebus to the spectacular leapingof the long-legged indrid lemurs (sifa-kas, babakotos and the like). It cannotreally be said that any of the living le-murs is excluded from any part of theforest environment by its anatomic lo-comotor specializations, but it is gener-ally true that most species avoid spend-ing much time on the ground. The onlynotable exception is the ringtailed le-mur, Lemur catta.

Among the extinct lemurs, on the oth-er hand, was a group that was clearlyadapted for life on the ground. This isthe family Archaeolemuridae, contain-ing the two medium-sized genera Ar-

chaeolemur and Hadropithecus. LaurieR. Godfrey of the University of Massa-chusetts at Amherst (the source for allthe estimates of subfossil body weightpresented here) has estimated that thevarious archaeolemurid species weighedbetween 35 and 55 pounds.

These lemurs were rather powerful-ly built, short-legged relatives of the in-drids, with highly specialized teeth. Clif-ford J. Jolly of New York University hascompared them, respectively, with twoAfrican higher primates, the commonand gelada baboons. The common ba-boon is an extremely adaptable form, athome in woodlands and deciduous for-ests as well as in the savanna habitats in which it is so familiar. The gelada ba-boon is speciÞcally adapted to the tree-less Ethiopian highlands, and virtually allits food is derived from terrestrial sourc-es. Both the dentition and what is knownof the body skeleton of Hadropithecussuggest strongly that this lemur hadsimilar dietary and habitat preferences.

Another extinct indrid relative showed


weighed at most 130 pounds, was a slothlike arboreal dweller. These reconstruc-tionsÑdone with the advice of Laurie R. Godfrey of the University of Massa-chusetts at AmherstÑare based on fossils that have been unearthed since the late19th century. Excavations have provided mostly incomplete skeletons of at least15 species of extinct lemurs, belonging to some eight genera.


a totally diÝerent set of adaptations. Foranalogues, one must look well beyondthe primates. Palaeopropithecus com-prised at least two species with a weightrange of perhaps 90 to 130 pounds.Ross MacPhee, my colleague at theAmerican Museum of Natural History,has analyzed a substantially completeskeleton recovered in northern Mada-gascar a few years ago and concluded itwas a generally slow-moving and some-what slothlike arboreal hanger, built forßexibility rather than strength. Its evenlarger relative Archaeoindris (proba-bly more than 400 pounds) is poorlyknown, but Martine Vuillaume-Randria-manantena of the University of Mada-gascar believes it was probably a terres-trial quadruped somewhat like the ex-tinct ground sloths of the New World.Both of these forms had specializa-

tions of the skull, particularly of thenose area, that are unmatched amongliving primates.

We also have to look outside ourown order to Þnd an analogue to thebest-known subfossil lemur, Megaladap-

is. The three species, ranging in weightfrom about 90 to 170 pounds, havebeen described by Alan C. Walker of theJohns Hopkins School of Medicine asclosest in locomotion to the marsupialkoala of Australia. Like the koala, theselemurs would have climbed slowly, pre-sumably preferring vertical supports,and they had limited leaping capabili-ties. A number of specializations of theskull may have compensated for suchlocomotor limitations. They would haveallowed the animal to feed in a large ra-dius from a single sitting position.

The best known of the sites that

have yielded extinct lemurs is Ampa-sambazimba in MadagascarÕs centralhighlands. The 14 primate species (in-cluding both extinct and surviving le-murs) whose bones have been recov-ered there compare favorably in abun-dance with the species at any otherplace in the world where primates arefound. Yet Ampasambazimba is in themiddle of what is now an essentiallytreeless plateau. How does it boast thisrich and heterogeneous forest-livingfauna?

Studies dating from early in thiscentury seemed to have the an-swer. Before the advent of hu-

mans, the argument goes, Madagascarwas essentially fully forested. The factthat forest had survived only in patch-es was attributed to the propensity ofthe early settlers to burn oÝ vast areasof forest to provide grazing for cattleand land for agriculture. Indeed, theprocess is still only too evident.

Following this argument, loss of hab-itat must have been at least a major in-ßuence in the disappearance of Mad-agascarÕs large birds and mammals.There was also a selective aspect of thisextinction: the lemurs known to havebecome extinct were large bodied andthus of the greatest interest and vul-nerability to hunters. Presumably, theyalso reproduced more slowly than thesmaller forms that have survived. Thecase for a combination of direct and in-direct human activity as the agent of ex-tinction seems compelling.

Climatic change has also been pro-moted as an agent of extinction, initial-ly because many subfossil sites consistof dried-up lakes or marshes. As a totalexplanation of the extinctions, dryingwas never completely convincing. Still,interest in a possible climatic eÝect hasbeen revived recently by the demon-stration that some of the grasslands ofMadagascarÕs center are of long (andcertainly prehuman) standing.

Analyses of lake cores by David A.Burney of Fordham University have un-derscored the fact that, like every otherregion of the earth, Madagascar has un-dergone climatic ßuctuations over thepast few thousand years. At the end ofthe last ice age some 10,000 years ago,MadagascarÕs forests were apparentlyjust beginning to recover from a peri-od of contraction. It is therefore hard-ly surprising that the central highlandswere not fully forested by the timeMadagascarÕs Þrst people arrived.

Climatic stress and the consequent reduction and redistribution of foreststhus constitute one potential factor in the disappearance of the subfossilfauna of Madagascar. But it is clear


MadagascarÕs Habitats

he island of Madagascar offersan extensive range of environ-

ments—from lush rain forests anddeciduous forests to deserts andcoastal plains. Population pressuresand desperate poverty have contrib-uted to deforestation. For example,the forest that runs along the east-ern coast was reduced by 66 per-cent between 1900 and 1985.

T

SOURCES: Glen M. Green, Robert W. Sussman, H. Humbert and G. Cours Darne

HUMIDSECONDARY FOREST

DENSERAIN FOREST

GRASSLAND

EXTENT OF EASTERNFOREST IN 1900

MOUNTAIN FOREST

DECIDUOUS FOREST

DRY BRUSHAND THICKET


that in that land as elsewhere, period-ic disturbances of this kind have oc-curred throughout time. The ancestrallineages of the extinct lemurs obvious-ly survived these earlier vicissitudes,and there is no reason to believe that ontheir own the most recent (rather mild)round of climate changes should havehad a fatal eÝect on several lineagessimultaneously. This is especially truewhen we consider that most subfos-sil lemurs were probably ecological gen-eralists. Megaladapis, Palaeopropithe-

cus and Archaeolemur, among others,lived in environments ranging from hu-mid to arid and were clearly highlyadaptable in terms of choice of habitat.Hence, something other than simply an-other cycle of disturbance of naturalhabitat must be invoked to explain thedisappearance of the giant lemurs. Onlyone truly novel factor presents itself :Homo sapiens.

To talk of the Òextinction eventÓ ofthe large-bodied lemurs implies a process that has ended. That is

not so. The extinct and living lemursform part of a single process that iscontinuing. The smaller, ßeeter lemurshave survived so far, but they are them-selves under increasing pressure froman expanding human population. Thetoll of hunting increases as humannumbers and the use of more sophisti-cated weapons do. Another importantconsideration is the population move-ments that tend to erode the local be-liefs that, in certain places, have tradi-

tionally protected various lemur species.More worrying yet is destruction of

habitat, principally by slash-and-burn ag-riculture but also by the cutting of treesfor fuel and by commercial logging. Thelargest continuous forest tract in Mad-agascar lies along the islandÕs humideastern escarpment. By analyzing his-torical documents and satellite images,Glen M. Green and Robert W. Sussmanof Washington University have shownthat 66 percent of the area covered bythis forest at the turn of the centuryhad been denuded by 1985. They esti-mate that within another 35 years onlythe steepest slopes of the escarpmentwill still bear trees. In the ßatter west-ern and southern regions of the island,the rate of disappearance of forest isprobably faster.

Such pressures have existed for manydecades. In the 1920s the colonial au-thorities in Madagascar set up one ofthe worldÕs Þrst systems of natural re-serves. But as one of the worldÕs poor-est countries, Madagascar, despite itsgovernmentÕs genuine concern, cannotaÝord to police these reserves adequate-ly or, in some cases, at all. Fortunately,in recent years the island has attractedthe attention of the international con-servation community. The country Þg-ured recently in one of the Þrst Òdebtfor natureÓ swaps, in which foreign debtis reduced in exchange for the protec-tion of natural areas.

It is, of course, the real and im-mediate needs of desperately poor lo-cal communities that are being met by

most of the forest destruction. In manycases, large-scale agreements on con-servation have yet to devolve from thehigher governmental sphere to actu-al projects on the ground. But as they do, one can hope for the stabilizationand perhaps even for the long-term im-provement of MadagascarÕs environ-mental situation.

Meanwhile the lemur populationscontinue to dwindle. It is tragic to seeany part of the worldÕs biodiversity disappear, but the tragedy is particular-ly acute in the case of MadagascarÕs le-murs, which still have so much to teachus about our own past.


FURTHER READING

A WORLD LIKE OUR OWN: MAN AND NA-TURE IN MADAGASCAR. Alison Jolly. YaleUniversity Press, 1980.

THE PRIMATES OF MADAGASCAR. Ian Tattersall. Columbia University Press,1982.

ENVIRONMENT, EXTINCTION, AND HOLO-CENE VERTEBRATE LOCALITIES IN SOUTH-ERN MADAGASCAR. R.D.E. MacPhee inNational Geographic Research, Vol. 2,No. 4, pages 441Ð455; Autumn 1986.

LEMURS LOST AND FOUND. Patricia C.Wright in Natural History, Vol. 97, No.7, pages 56Ð60; July 1988.

MADAGASCARÕS LEMURS: ON THE EDGEOF SURVIVAL. Alison Jolly in NationalGeographic, Vol. 174, No. 2, pages 132Ð161; August 1988.

MADAGASCAR: A NATURAL HISTORY. KenPreston-Mafham. Facts on File, 1991.

BURNED LAND is seeded for agriculture by Malagasy farmers(left ). The slash-and-burn technique began when settlers Þrst ar-rived on Madagascar under 2,000 years ago. Extensive defores-

tation for logging has also led to erosion on many slopes (right).The loss of habitat resulting from these two practices has con-tributed to the disappearance of large birds and mammals.


uring the past few years, re-search in semiconductors hastaken on, quite literally, new

dimensions. Their numbers are two,one and zero. Electrons in recently de-veloped devices can be conÞned toplanes, lines or mathematical pointsÑquantum dots.

Microchip manufacturers have devel-oped a toolbox of nanofabrication tech-nologies capable of creating structuresalmost atom by atom. These techniqueshave opened up a new realm of funda-mental physics and chemistry as work-ers make and study artiÞcial analoguesof atoms, molecules and crystals. Exper-imenters are no longer limited by theatomic shapes, sizes and charge distri-butions available in nature.

In addition to the exciting sciencethey portend, quantum dots promiseproperties that could be harnessed fora range of electronic and optical appli-cations. Arrays of densely packed dotscould form a substrate for computersof unprecedented power; indeed, Nor-man Margolus of the Massachusetts In-stitute of Technology has coined theterm Òcrayonium.Ó Dots could also con-stitute materials capable of absorbingand emitting light at whatever set ofwavelengths their designers specify orcould even serve as the basis of semi-conductor lasers more eÛcient and pre-cisely tuned than any now in existence.

Planes, lines and dots are mathemat-ical constructs. They have no physicalextent. How is it possible to make themin a real, three-dimensional material?

The answer lies in quantum mechan-ics and HeisenbergÕs uncertainty princi-ple. The position of an object (an elec-tron, for instance) and its momentumcannot both be known to arbitrary pre-cision. As an electron is more closelyconÞned, its momentum must be moreuncertain. This wider range of momen-ta translates to a higher average ener-gy. If an electron were conÞned in aninÞnitely thin layer, its energy wouldalso be inÞnite.

In general, the energy of electrons in asemiconductor is limited by their tem-perature and by the properties of thematerial. When the electrons are con-Þned in a thin enough layer, howev-er, the requirements of the uncertaintyprinciple in eÝect override other consid-erations. As long as the electrons do nothave enough energy to break out of con-Þnement, they become eÝectively two-dimensional.

This locution is not just an approxi-mation. Electrons conÞned in a planehave no freedom of motion in the thirddimension. Those conÞned in a quan-tum wire are free in only one dimen-sion, and those conÞned in a quantumdot are not free in any dimension. Forcommon semiconductors, the lengthscale for a free conduction electron isabout 100 angstroms. (One angstrom is10Ð10 meter, approximately the radiusof a hydrogen atom.) An electron in-side a cube of semiconducting material100 angstroms on a side is essentiallyconÞned to a point.

Engineering of less than three-di-mensional semiconductors beganin earnest during the early 1970s,

when groups at AT&T Bell Laborato-ries and IBM made the Þrst two-dimen-sional Òquantum wells.ÕÕ These struc-tures, made by thin-Þlm epitaxial tech-niques that build up a semiconductorone atomic layer at a time, are thin re-gions of semiconducting material (usu-ally gallium arsenide and related com-pounds) that attract electrons. The en-ergy of electrons residing within the

well is lower than the energy of thoseresiding elsewhere, and so the electronsßow in, just as water runs downhill toÞll up a deep well.

It is possible to create not only quan-tum wells but also quantum barriersÑtwo-dimensional ÒhillsÓ of material thatrepel electrons. In combination, thewells and barriers can be used to buildcomplex structures that previously ex-isted only as examples in quantum me-chanics textbooks [see ÒDiminishing Di-mensions,ÕÕ by Elizabeth Corcoran; SCI-ENTIFIC AMERICAN, November 1990].

Quantum wells have now becomecommonplace. They are the basis ofthe laser diodes found in compact-discplayers and the sensitive microwave re-ceivers that pull in signals from a satel-lite dish. In the meantime, researchershave learned how to conÞne electronsnot simply in a plane but in a point.

The Þrst hints that zero-dimension-al quantum conÞnement was possiblecame in the early 1980s, when A. I. Eki-mov and his colleagues at the IoÝePhysical-Technical Institute in St. Pe-tersburg noticed unusual optical spec-tra from samples of glass containingthe semiconductors cadmium sulÞdeor cadmium selenide. The samples hadbeen subjected to high temperature;Ekimov suggested tentatively that theheating had caused nanocrystallites ofthe semiconductor to precipitate in theglass and that quantum conÞnement ofelectrons in these crystallites causedthe unusual optical behavior.

To understand this chain of reason-ing, imagine an electron trapped in abox. Quantum mechanics states thatthe electron has wave properties, likethe ripples on water or the vibrationsof a violin string. Just as a violin stringis tied down at both ends, so the elec-tron wave is bounded by the walls ofthe box. The wavelength of the stringÕsvibrations (or the electronÕs) must Þtwithin those conÞnes [see illustrationon page 121].

In the case of the violin string, thepoint at which it is tied down changes


MARK A. REED studies mesoscopicsemiconductor systemsÑthose whosebehavior shows a mixture of classical andquantum mechanical traits. After receiv-ing his doctorate in solid-state physicsfrom Syracuse University in 1983, Reedjoined Texas Instruments. In 1987 histeam there made the Þrst lithographical-ly deÞned quantum dots. Since 1990 hehas been a professor of electrical engi-neering at Yale University.

Quantum DotsNanotechnologists can now confine electrons

to pointlike structures. Such “designer atoms” may lead to new electronic and optical devices

by Mark A. Reed


as the violinistÕs Þnger slides up theÞngerboard. The length of the allowedwaveform shortens, and the frequencyof the stringÕs vibrations increases, asdoes that of all its harmonic overtones.If the size of an electronÕs conÞningbox is made smaller, the electronÕs low-est energy level (the analogue of thefundamental pitch of the violin) will in-crease. For semiconductor nanocrys-tallites, the fundamental ÒpitchÓ is thethreshold energy for optical absorp-tion, and the harmonic overtones cor-respond to new absorption features athigher energies.

How small must a nanocrystallite befor this phenomenon to be visible? In a vacuum the eÝects of conÞnementwould begin to appear when the elec-tron was trapped in a volume about 10angstroms across. That size implies anelectron wavelength of 20 angstromsand therefore an energy of about onefortieth of an electron volt.

Here semiconductor physics comesto the aid of the nanotechnologist. Thewavelength of an electron depends onits energy and its mass. For a givenwavelength, the smaller the mass, thelarger the energy and the easier it is toobserve the energy shift that conÞne-ment causes. The electrostatic poten-tials of the atoms in the crystalline lat-tice superimpose to provide a mediumin which electron waves propagate withless inertia than they do in free space.The ÒeÝective massÓ of the electron isthus less than its actual mass. In galli-um arsenide the eÝective mass is about7 percent of what it would be in a vacu-um, and in silicon it is 14 percent. As aresult, quantum conÞnement in semi-conductors occurs in volumes roughly100 angstroms across.

The optical absorption threshold fornanocrystallites of this size shifts tohigher energiesÑaway from the red endof the spectrumÑas the crystallite be-comes smaller. This eÝect appears mostelegantly in cadmium selenide clusters;the progression from deep red to or-ange to yellow as the diameter of the


QUANTUM CONFINEMENT is respon-sible for the colors of cadmium selen-ide crystallites, each a few nanometersacross, synthesized by Michael L.Steigerwald of AT&T Bell Laboratories.Electrons within the tiny specks ofsemiconductor scatter photons whoseenergy is less than a minimum deter-mined by the size of the crystallite andabsorb those whose energy is higher.The largest crystallites can absorb low-er-energy photons and so appear red,whereas the smallest absorb only high-er-energy quanta and so appear yellow.



cluster declines can be clearly seen bythe naked eye. (An intriguing and as yetunresolved question is what happenswhen the crystallite is so smallÑlessthan 10 angstroms acrossÑthat the ef-fective mass concept, derived from bulksolids, no longer makes sense. Quantumdots that small have yet to be made.)

EkimovÕs hypothesis turned out tobe true, but it took years of workby groups at Corning Glass, IBM,

City College of New York and elsewhereto sort out the correct glass preparationtechniques and convincingly demon-strate quantum conÞnement. MeanwhileLouis E. Brus and his co-workers at BellLabs were making colloidal suspensionsof nanocrystallites by precipitation fromsolutions containing the elements thatmake up semiconductors.

Such crystallites grow by the addi-tion of individual ions until the supplyis either depleted or removed. Conse-quently, by arresting the precipitationafter a certain time, Brus and his col-leagues could control the size of theprecipitate in a range between 15 andabout 500 angstroms. Sizes within asingle batch varied by no more than 15percent. Just as in the case of the glass-encased nanocrystallites, a dramaticshift of the fundamental absorption en-ergy to higher energies suggested quan-tum conÞnement.

Workers in many laboratories world-wide have built on this approach. Forexample, A. Paul Alivisatos and his col-leagues at the University of Californiaat Berkeley have extended the range ofelements from which crystallites canbe made. In addition to the II-VI com-

pounds (such as cadmium from the sec-ond column of the periodic table and se-lenium from the sixth), they have alsoprecipitated III-V compounds such asgallium arsenide. Michael L. Steigerwaldof Bell Labs and many others have em-ployed an organic Òsoap bubbleÕÕ wrap-ping known as a reverse micelle to stabi-lize the surface of the tiny semiconduc-tor crystals. Groups at the University ofCalifornia at Santa Barbara, the Univer-sity of Toronto and elsewhere are stuÝ-ing clusters of atoms into the nanome-ter-scale cavities of zeolites, a techniquethat confers the advantage of precise di-mensional control.

Encasing nanocrystals inside anoth-er material could signiÞcantly improvetheir quantum performance. The tinyspecks of semiconductor have a verylarge surface-to-volume ratio, and sur-

abrication of quantum dots pro-ceeds through a series of masking

and etching steps. First, an electronbeam scans the surface of a semicon-ductor containing a buried layer ofquantum-well material (1). Resist is re-moved where the beam has drawn apattern (2 ). A metal layer is depositedon the resulting surface (3 ), and then asolvent removes the remaining resist,leaving metal only where the electronbeam exposed the resist (4 ). Reactiveions etch away the chip except where it is protected by metal (5 ), leaving aquantum dot (6 ). An alternative fabri-cation method lays down a pattern ofelectrodes above a buried quantum-well layer. When a voltage is applied tothe electrodes, the resulting field expelselectrons from the layer except in cer-tain small regions (top right). The de-gree of quantum confinement in thoseregions can be manipulated by chang-ing the electrode voltages.

Building in Zero Dimensions

F

QUANTUM DOT

ELECTRON BEAM

RESIST

QUANTUM- WELLMATERIAL

METAL

1

2

3 4 5

6

QUANTUM DOT ELECTRIC

FIELD

ELECTRODES

REACTIVE IONS


faces in general are marked by atomswith dangling chemical bonds. Theseimproperly terminated bonds can act asdampers, absorbing the energy of elec-trons vibrating in higher-energy (short-er-wavelength) modes. As a result, manynanocrystallites do not show the dra-matic harmonic series of energy levelsthat would be expected from a quan-tum dot.

The inherent diÛculties of mak-ing quantum dots from clustersof atoms led researchers in the

mid-1980s to look for other fabricationschemes. My co-workers and I at TexasInstruments in Dallas made the Þrstlithographic quantum dots in 1987. Wecut slabs of quantum-well material intopillars by means of advanced etchingtechniques similar to those used in thefabrication of state-of-the-art integrat-ed circuits.

Making pillars 100 angstroms widerequires electron-beam lithography in-stead of the optical techniques used tomake most chips. An electron beamscans the semiconductor surface, whichhas been coated with a thin polymerlayer called a resist. (Similar eÝects canalso be achieved by means of x-rays orion beams.) A series of process stepsreplaces the resist with a thin layer ofmetal in areas where the beam wasscanned at high intensity. A shower ofreactive gas then etches away the un-protected quantum-well material, leav-ing the pillars behind. Using this tech-nique, pillars or other features as smallas 1,000 angstroms across can be quiteeasily constructed. But the process be-comes increasingly diÛcult as the scalefalls to about 100 angstroms, the limitof the best-known resist.

Above and below the quantum-wellmaterial in these pillars lie ultrathin in-sulating layers called tunnel barriers,followed by conductive contacts. Theinsulators conÞne electrons in the wellfor a very long time, but eventually theelectrons can quantum-mechanicallytunnel in and out, carrying a small cur-rent that can serve to probe the inter-nal energy states of the well. Wheneverthe voltage across the well matches theenergy of one of its resonant states, cur-rent ßow increases. If the diameter ofthe pillar is very small, its current-volt-age spectrum displays the harmonic series of peaks that marks quantumconÞnement. Indeed, by making only asingle pillar that is isolated from its sur-roundings, one can calculate the prop-erties of a single quantum dot, a taskthat is hard to imagine carrying out withnanocrystallites.

Moreover, the lithographic fabricationprocess naturally clads and protects

the quantum dot from surface eÝects,at least on two faces. The top and bot-tom of the dot are single crystal inter-faces made by advanced epitaxy andare essentially perfect. Because the pil-lar is electrically conductive, the surfacebonds of the semiconductor we usedcreate a positive charge with respect to the internal core of the pillar. Thischarge repels electrons from the surfaceinto the quantum-conÞned interior; theregion from which the electrons havedeparted forms an insulating sheatharound the pillar, protecting the sides ofthe dot. A 1,000-angstrom pillar wouldthus contain a 100-angstrom dot.

The realization of a quantum dot de-pends on an insulating sheath of thecorrect thickness, which depends inturn on the size of the etched pillar.When we Þrst tried to make quantumdots, no one knew what the correct siz-es might be, and many attempts endedin failure. But early on the morning ofAugust 20, 1987, as I was preparing togive a talk at a conference on quantum-well devices, my colleagues called tosay they had successfully measured aquantum dot. I rushed to the hotel faxmachine just in time to see the dataprint out, showing a rich harmonic se-ries of electron energy levels. An hourlater I had rewritten the Þnale of mytalk, the hotel staÝ had transformedthe fax into a viewgraph, and I deliv-ered the news.

Subsequent measurements conÞrmedthat dots of diÝerent sizes produceddiÝerent harmonic spectra, a clear sig-nature of quantum conÞnement. Sincethen, groups at CNET in France, NTT inJapan, the University of Cambridge, theState University of New York at StonyBrook and Princeton University havealso employed this fabrication tech-nique. Pierre Gueret and his co-work-ers at IBM Z�rich have even made aÒsqueezableÓ dot by placing an elec-tronic gate around the dot in a tour deforce of fabrication technology. Increas-ing the electric potential on the gate re-duces the size of the dot and increasesthe fundamental energy and harmonicsin its spectrum.

The success of these electrical mea-

surements on lithographically deÞnedquantum dots, in contrast to the rela-tive diÛculty of optical measurementson dots made from atomic clusters, hasunderscored the importance of control-ling damaging surface eÝects. Groupsat IBM, AT&T, the universities of Ham-burg and Munich, Delft University ofTechnology in the Netherlands, Philips,Cambridge, the Max Planck Institutefor Solid State Physics in Stuttgart andM.I.T. have managed to eliminate sur-face eÝects entirely. They make quan-tum dots by placing tiny gate electrodeson top of a buried layer that conÞneselectrons in two dimensions. The topelectrodes squeeze the electrons intoquantum-conÞned Òislands.Ó

One advantage of this approachis that it is possible to put asmany or as few electrons in the

dot as desired, simply by varying thesqueezing voltage. The result is whatmight be called a designer atom: theconÞning potential acts as an attractivenucleus, and the valency (the numberof electrons) is determined by the ex-ternal gate voltage.

In natural atoms, conÞnement of elec-trons is caused by the radially directedelectrostatic force of the nucleus, andthe electron wave functions are radially


EN

ER

GY

ELECTRON IN A BOX is constrained tohave a quantum wave function that Þtsevenly within its borders. The lowestenergy level corresponds to a standingwave with a single antinode, the nextlowest to a wave with two, and so on.Electron energy is inversely proportion-al to the square of the wavelength, andso the energy levels rise rapidly. Thisharmonic series of energy levels is thesignature of a quantum dot.


symmetric. In these quantum dots theshape of the gate electrodes controlsthe size, shape and symmetry of theconÞning potential, and so Òwave-func-tion engineersÓ may eventually be ableto study atomic physics previously in-accessible in nature, such as the wavefunctions or electrons in square or rec-tangular atoms.

A group at the Stuttgart Max PlanckInstitute, as well as teams at IBM andAT&T, has made large, periodic arraysof dots by fabricating a gridlike gateelectrodeÑthe nanostructure equiva-lent of a window screen. Voltage ap-plied to the grid forms a regular latticeof quantum conÞnement in the under-lying material. The size and the num-ber of electrons in each dot can be con-trolled, as can the height and thicknessof the barrier between the dots. Regu-lar peaks appear in the optical absorp-tion spectra of these structures. Thissurprising phenomenon testiÞes to theprecision with which the arraysÑsomecontaining more than a million dotsÑhave been made, as any variation insize would smear out the harmonicspectra. Ray C. Ashoori and Horst L.St�rmer of AT&T recently measuredthe capacitance of individual dots anddemonstrated that it is possible to cap-ture a single electron in each one. Elec-

trons can then be added one at a time,in a digital fashion.

These results open up the possibilityof making a planar artiÞcial lattice inwhich virtually all the properties of theconstituent ÒatomsÓ can be controlled.Just as individual quantum dots dis-play energy levels analogous to thoseof atoms, an artiÞcial lattice would pos-sess an energy band structure analo-gous to that of a crystalline semicon-ductor. It could be used to study manyquestions in quantum physics andmight also form the basis for a super-fast electronic oscillator.

No one, however, has yet made a pla-nar artiÞcial lattice and unambiguouslydemonstrated its band structure. Suc-cess will require not only exacting pre-cision in fabricating the electrode gridbut also heroic control of defects in theunderlying quantum-well material. Innatural semiconductor lattices, engi-neers can rely on the fact that all sili-con atoms, for instance, are identical,but in an artiÞcial lattice, they will haveto impose this uniformity by craft.

An intriguing twist in this genre isthe ÒantidotÓ lattice. If the voltage onthe grid is reversed, the islands that at-tracted electrons now repel them. Elec-trons are forced to reside in the inter-vening space, bouncing oÝ the antidots

as they move through the array in whatis probably the smallest Òpinball ma-chineÓ ever made.

Another variant of the grid tech-nique, demonstrated by Kathleen Kashand her co-workers at Bell Communica-tions Research (Bellcore), uses compres-sive stress in place of electrodes to im-pose quantum conÞnement. The Bell-core team lays down a strained layer (alayer of material whose atomic latticespacing diÝers from that of the sub-strate below it) on top of the quantum-well material, compressing it laterally.The workers then etch a pattern intothe strained layer; wherever the layer isetched away, the compressive stress isrelieved. The resulting tiny variationsin atomic spacing within the quantum-well layer cause changes in electron en-ergy levels that can form quantum dots.

Electrostatic squeezing producesdots whose quantum conÞnementcan be controlled more easily than

can that of dots produced by othermethods. Until recently, it was impossi-ble to make electrodes small enoughfor tunnel-barrier contacts to a singleelectrostatically squeezed dot. Dur-ing the past three years, a number ofgroups have succeeded in this task,producing lateral contacts to the dotthat consist of electrostatically control-lable tunnel barriers.

This structure gives the researchercontrol of many of the variables thatdeÞne a dot, including size, number ofelectrons and transparency of the con-Þning barriers. Such systems are idealfor testing textbook quantum mechan-ics problems, such as the properties ofzero-dimensional states or the probabil-ity of electrons tunneling through bar-riers. By stringing two dots together toform an artiÞcial molecule, one can in-vestigate coupling between the statesof adjoining quantum dots. And, as LeoP. Kouwenhoven of Delft University hasdemonstrated, it is even possible tostring many dots together in a pearl-necklace fashion to generate an arti-Þcial one-dimensional crystal and towatch how the energy band structureof a crystal forms.

The Delft and M.I.T. groups have dis-covered that the energy levels of thesesmall dots are determined not only by quantum mechanical rules based on size but also by the quantization of the electron charge [see ÒSingle Elec-tronics,Ó by Konstantin K. Likharev andTord Claeson; SCIENTIFIC AMERICAN,June 1992]. The energy level of a dotdepends in part on its capacitance andthe amount of charge contained withinit, and the amount of charge must ofcourse be a multiple of e.


ADJUSTABLE QUANTUM DOT lies buried at the intersection of electrodes in theabove micrograph. The four interior electrodes ÒsqueezeÓ electrons in the buriedquantum-well layer into the dot. The outer electrodes serve as contacts for elec-trons to tunnel in or out of the dot; tunneling rates increase when the electron en-ergies match the dotÕs energy levels. Those levels, in turn, can be controlled bychanging the voltage on the inner electrodes.


The coexistence of these two kindsof quantization causes a complex in-terplay of eÝects. To understand whichwill be most important, one needs toknow not just the wavelength and ef-fective mass of the electron in the dotbut also its electrical capacitance. If adot is made from a metallic particle, ithas many more conduction electronsthan does a semiconductor; further-more, the wavelength of the conduc-tion electrons is only a few angstroms.As a result, in a 100-angstrom metal-lic dot, charge quantization exerts amuch stronger eÝect, relatively speak-ing, than size quantization. The ca-pacitance of the metal dot, however, isnot so diÝerent from that of a semi-conductor dot of the same size, and in the semiconductor the energies ofthe two eÝects may be approximatelythe same.

The development of quantum dotsis a culmination of 20 years ofwork, during which researchers

have learned how to tailor electronicmaterials. Before the 1970s, research in solid-state science was conÞned tomaterials provided by nature. The re-Þnement of ultrathin-layer epitaxy dur-ing that decade gave researchers thetools to fabricate the two-dimension-al structures that dominate technologytoday. Extensions of that technologyhave now led to exploration of the one-and zero-dimensional domains. Beforethese discoveries can be applied on acommercial scale, however, a new gen-eration of fabrication techniques mustbe developed.

The most challenging obstacle is toachieve essentially perfect control overthe size and purity of these nanostruc-tures. The Òtop-downÓ approach to fab-ricationÑcarving, dicing or squeezingsemiconductorsÑmay not be suÛcientwithout revolutionary advances in ma-terials and nanofabrication. Current pro-totype devices are large (although theactive region of the device is quantum-sized, the electrodes and contact padstake up enormous space), and they op-erate only at very low temperature.

Moreover, these devices are made by electron-beam lithography, a fab-rication technology that cannot beused to make large numbers of thecomplex circuits crucial for economicsuccess. New lithographic tools thatpermit three-dimensional atomic-scalecontrol, such as structured epitaxialgrowth or self-organizing molecular as-sembly, are needed. Indeed, it may benecessary to develop novel materialsand synthesis techniques that blendconventional semiconductor technolo-gy with alternative approaches. Work-

ers at the Fujitsu Laboratories in Ja-pan, for example, have made quantumwires and dots from organic polymers.The location of conducting atoms with-in the polymer molecules is Þxed, sothis method oÝers much Þner controlthan is possible with electron-beamlithography. If Òbottom-upÓ assemblyof quantum devices proves feasible,the methods now used to producequantum dots will seem akin to mak-ing the books in the Library of Con-gress by whittling away at a large blockof wood.

The most important challenge thatresearchers face, however, is not learn-ing how to build quantum-conÞne-ment devices in quantity; instead it isto design useful circuits that exploittheir potential. Although the techno-logical size limits to quantum devicesare in theory signiÞcantly smaller thanthose projected for silicon, the viabilityof quantum circuits will be determinedby how well they compete in the mar-ketplace against the coming decadeÕsdevelopments in conventional silicontechnology.

Just as transistors found uses far be-yond their initial role in radio receivers,so the ultimate application of quantumdevices could be quite tangential to thetasks of digital computation and com-munications for which they have thusfar been developed. If engineers can fab-ricate lattices containing millions or bil-lions of quantum dots, specifying theshape and size of each one, they will be able to make any electronic or opti-

cal material of which they can conceive.Emission, absorption and lasing spectracould be precisely tailored, and a singleslab of material could even be designedto contain a myriad tiny computerswhose interconnections and internal ar-chitecture would change to match eachnew problem posed to them.

Even beyond the practical applica-tions of quantum devices and the newintellectual territory they oÝer exper-imental physicists, quantum dots areexciting to researchers. The ability tomanipulate matter on an atomic scaleand create unique materials and de-vices with custom-designed propertieshas universal appeal. It marks a triumphof human ingenuity and imaginationover the natural rules by which materi-als are formed.


FURTHER READINGTRENDS IN MATERIALS: DIMINISHING DI-MENSIONS. Elizabeth Corcoran in Scien-tific American, Vol. 263, No. 5, pages122Ð131; November 1990.

ENGINEERING A SMALL WORLD: FROMATOMIC MANIPULATION TO MICROFAB-RICATION. Special Section in Science,Vol. 254, pages 1300Ð1342; November29, 1991.

QUANTUM SEMICONDUCTOR STRUCTURES:FUNDAMENTALS AND APPLICATIONS.Claude Weisbuch and Borge Vinter. Aca-demic Press, 1991.

NANOSTRUCTURES AND MESOSCOPIC SYS-TEMS. Edited by Wiley P. Kirk and MarkA. Reed. Academic Press, 1992.

ELECTRODE GRID creates a lattice of quantum dots in the material underneath it. Sucha lattice is, in eÝect, a crystalline layer made of artificial atoms whose energy levelscan be controlled precisely. Arrays of quantum dots aid in the study of fundamentalphysics and also may eventually be made into novel electronic or optical devices.


onald O. Hebb, one of the mostinßuential psychologists of histime, began his adult life in-

tending to be a novelist. Deciding thathis calling required an understandingof psychology, he embarked on a coursethat led him into two decades of re-search. His studies culminated in 1949with the publication of The Organiza-

tion of Behavior, a keystone of modernneuroscience.

The monograph broke new ground bypositing neural structures, called cell as-semblies, which were formed throughthe action of what is now called theHebb synapse. The cell-assembly theoryguided HebbÕs landmark experimentson the inßuence of early environmenton adult intelligence. It foreshadowedneural network theory, an active line ofresearch in artiÞcial intelligence.

HebbÕs book came at the right timebecause it ßew in the face of behavior-ism just as that school was losing itsdominance. The behaviorists denouncedexplanations of behavior by associationof ideas (which they called mentalism)and by the action of neurons (which

they called physiologizing). But manypsychologists had grown weary of theartiÞcial theories these strictures hadengendered, and they were captivatedby HebbÕs project and his engaging lit-erary style. The book became a classic,and Hebb became a household word (atleast in psychologistsÕ households).

Hebb never claimed that his 1949 the-ory was Þrmly grounded in physiology.His model gave workers something tolook for, and later, as knowledge of thebrain grew, it became possible to framehis ideas in more realistic neural terms.None of this subsequent research hasinvalidated HebbÕs basic hypothesis. In-deed, its inßuence appears in many ar-eas of current research.

Hebb was born in Chester, asmall Þshing and boat-buildingtown in Nova Scotia. His parents

were physicians, and his two brothersand his sister followed in their parentsÕfootsteps. But Donald demonstratedhis independence early by studying En-glish in preparation for a career as awriter, graduating in 1925 from Dal-housie University in Halifax. To earnhis living while gestating his Þrst nov-el, he taught school in his hometown. A year later he set out to see life, goingwest to work an eight-horse team onprairie farms. Then, failing to get a jobas a deckhand on a freighter to China,he returned east and got a job as a la-borer in Quebec.

In 1927 an aspiring novelist not onlyhad to know life but also the works ofSigmund Freud. This was HebbÕs intro-duction to psychology. He was suÛ-ciently intrigued to apply to the psy-chology department of McGill Univer-sity, where he was accepted in 1928 as a part-time graduate student. Again he supported himself by teaching and,again, what started out as a temporaryinterest verged on becoming a career.After one year he was made principalof an elementary school in a working-class district of Montreal. He was deter-mined to make learning enjoyable, tak-

ing care to prevent schoolwork from be-ing used as a punishment, instead send-ing miscreants out of class to play in theschool yard. Hebb became absorbed inhis educational experiments and seri-ously considered remaining in the pro-fession. Two developments dissuadedhim. He came down with a tubercularhip that conÞned him to bed for a yearand left him with a slight limp. Then hisbride of 18 months was killed in an au-tomobile accident. He therefore decid-ed to leave Montreal.

While conÞned to bed, Hebb wrote a masterÕs thesis that involved him in the nature-nurture controversy. The the-sis attempted to explain spinal reßexesas the result of Pavlovian conditioningin the fetus. He subsequently buried allreferences to this essay both becausehe changed his mind about its contentand because he came to oppose psy-chological research that lacked an ex-perimental foundation.

One of his examiners was Boris P.Babkin, a physiologist who had worked with Pavlov in St. Petersburg. He recom-mended that Hebb get some experience in the laboratory and arranged for him


PETER M. MILNER is professor emeri-tus of psychology at McGill University. Hereceived a B.Sc. in engineering from theUniversity of Leeds in the U.K. in 1942and was immediately recruited for radarresearch by the novelist C. P. Snow. ÒMyÞrst serious encounter with psychologyfollowed soon afterward,Ó he says, Òwhileworking on the interface between radardisplays and the human visuomotor sys-tem.Ó He later went to Canada to work onnuclear energy. In 1948 he left physicsto take a Ph.D. in physiological psychol-ogy at McGill under the supervision ofDonald O. Hebb. Milner is co-discoverer,with the late James Olds, of the rewardmechanism in the brain and has writtenextensively on the system and its impli-cations for theories of reinforcement andlearning. He has also written a textbookon physiological psychology as well aspapers on attention, visual recognition,memory and the mind-body problem.

The Mind and Donald O. HebbBy rooting behavior in ideas, and ideas in the brain,

Hebb laid the groundwork for modern neuroscience. Histheory prefigured computer models of neural networks

by Peter M. Milner

DONALD O. HEBB made his name as a theoretician but was equally distin-guished as a teacher. Here he appearsin a seminar held in the late 1960s.


to work with another Russian emigr�,Leonid Andreyev. Hebb conditioneddogs and became less impressed withPavlovian techniques. After much soul-searching as to whether he should con-tinue in psychology, he decided in 1934to burn his boats, borrow money and goto Chicago to continue his doctoral re-search under Karl S. Lashley.

The elder scientist was to exert a pro-found inßuence on HebbÕs approach,above all in his emphasis on physiolo-gy. Lashley had never doubted that tounderstand behavior one must Þrst un-derstand the brain. As a lab boy in 1910,he had salvaged slides of a frog brainfrom the trash heap and tried to Þnd in the neural connections some clue tofrog behavior. Lashley performed exper-iments to detect memory traces in thebrain, inventing techniques for makingbrain lesions and measuring their loca-tion and extent. By around 1930 he hadbecome convinced that memories could

not be stored in a single region of thebrain but must be spread throughout.In 1934, when Hebb went to Chicago,Lashley was concentrating on the studyof vision.

Ayear later Lashley was oÝered aprofessorship at Harvard Univer-sity and managed to take Hebb

along. Hebb had to start his researchfrom scratch, and having only enoughmoney for one more year, he sought anexperiment that could support a thesisno matter how it came out. He con-trived to adapt his interest in the na-ture-nurture question to LashleyÕs vi-sion project by investigating the eÝectsof early experience on the developmentof vision in the rat.

Contrary to the empiricist ideas of hismasterÕs thesis, Hebb found that ratsreared in complete darkness could dis-tinguish the size and brightness of pat-terns as accurately as rats reared nor-

mally. This Þnding indicated that theorganization of the visual system wasinnate and independent of environmen-tal cues, a view coinciding with that ofthe Gestalt school, to which Lashley wassympathetic [see ÒThe Legacy of Ges-talt Psychology,Ó by Irvin Rock andStephen Palmer; SCIENTIFIC AMERICAN,December 1990]. What Hebb did notnotice, although the results were includ-ed in a paper he published at the time,was that the dark-reared rats took muchlonger than normal rats to learn to dis-tinguish vertical from horizontal lines.Only many years later, after he hadagain changed his ideas about the rela-tive importance of innate and learnedmechanisms, did he appreciate the sig-niÞcance of this result.

Hebb received his Ph.D. from Har-vard in the middle of the Depression,when there were no jobs in physiologi-cal psychology to be had. He thereforestayed on for a year as a teaching assis-tant, a post that enabled him to contin-ue his work with Lashley. In 1937 therewas still no improvement in the jobmarket, but HebbÕs luck held out. Hissister was taking her Ph.D. in physiolo-gy at McGill and heard that Wilder Pen-Þeld, a surgeon who had just estab-lished the Montreal Neurological Insti-tute there, was looking for someone tostudy the consequences of brain surgeryon the behavior of patients. She passedon the information to her brother, andhis application for the two-year fellow-ship was successful. He married againand returned to Montreal. The youngman who thought he could run awayfrom his family destiny and become anovelist found himself one of a medi-cal group pioneering the treatment ofneurological disorders.

PenÞeldÕs specialty was the treat-ment of focal epilepsy by surgically re-



moving scarred areas of the cerebralcortex. He was acutely aware that hewas operating on the organ of the mindand that a false move could deprive hispatient of speech, intelligent behavioror even consciousness. Although Pen-Þeld was not a psychologist, his workexposed him to the relation betweenthe mind and the nervous system. Thisexperience no doubt inßuenced his de-cision to appoint psychologists to histeam and explained the close interesthe took in their Þndings.

HebbÕs main responsibility was tostudy the nature and extent of any intel-lectual changes in patients consequentto cortical excisions. Such research wasnot new: it began after World War I withthe psychometric testing of soldierswho had suÝered penetrating headwounds and continued later in patientswith brain tumors. In many cases, thelesions produced signiÞcant intellectu-al loss, but their locus and extent werediÛcult to determine. In contrast, sur-gical removals are more precisely de-Þned, and epileptic scars do not causethe widespread damage that bullets ortumors do.

Hebb soon faced a peculiar problem.Psychologists then regarded the frontallobes of the cerebral cortex as the seatof human intelligence, on the groundsthat this region is relatively much larg-

er than the corresponding areas in lessintelligent animals. Yet Hebb was notable to detect intellectual loss in pa-tients whose frontal lobes had been destroyed by accident or surgical ne-cessity. This seeming lack of eÝect im-pressed Hebb deeply and inspired hisquest for a theory of the brain and in-telligent behavior.

Although his observations set himoÝ on fruitful lines of inquiry,later work showed that Hebb

had relied too heavily on standard in-telligence tests. Brenda Milner, one ofhis students, who continued the workhe had begun on PenÞeldÕs patients,found that frontal-lobe lesions oftenmake it diÛcult for the patient to relin-quish a behavior that has ceased to beappropriate. Although they may not bedetected by intelligence tests, personal-ity changes after frontal-lobe damagecan profoundly aÝect the patientÕs life.

At the end of his fellowship at theneurological institute, Hebb Þnallyfound a permanent job at QueenÕs Uni-versity in Kingston, Ontario. There, de-spite his heavy teaching load, he keptup work on the problem of intelligence.Together with a student, Kenneth Wil-liams, he developed a variable-path ratmaze as an analogue to human intelli-gence tests. The Hebb-Williams maze

was widely used for the next quartercentury. But Hebb was proudest of atheoretical paper in which he proposedthat adult intelligence was crucially in-ßuenced by experience during infancy,basing his argument on the results ofhis research at the Montreal Neurolog-ical Institute. The paper was virtuallyignored at the time, although it is nowaccepted almost as a commonplace,having been embodied in such pre-school enrichment programs as HeadStart. But the concept was too advancedfor its time: in 1940 most psychologistspractically deÞned intelligence as an in-nate characteristic.

To reconcile his studies of childhoodinßuences with the apparent harmless-ness of frontal-lobe lesions, Hebb hy-pothesized that the regionÕs main func-tion was not to think but rather to fa-cilitate the tremendous acquisition ofknowledge during the Þrst few years oflife. Experiments to determine the rela-tive eÝects of early and late brain le-sions did not always support this idea,but it provided a stepping-stone toHebbÕs later theories.

In 1942 Lashley became the directorof the Yerkes Laboratories of PrimateBiology in Florida, and he invited Hebbto join his research team to study chim-panzee behavior. Hebb jumped at thechance of doing full-time research with

��

INACTIVE NEURON

FIRING NEURON

FIRING NEURON, PART OFCELL ASSEMBLY

CONTACTLENS

LENS

SLIDING TUBES

DIFFUSER

LIGHT

TARGET

POSTMOUNTED ON CONTACTLENS

HYPOTHETICAL CELL ASSEMBLY begins with parallel Þbersconnecting input from the retina to corresponding points inthe primary visual cortex. These neurons, in turn, connect tothe ÒassociationÓ cortex. Converging input Þres cells and acti-vates closed loops (dark red ). Synaptic changes ensue that en-able the loop to Þre with little input, producing output that rep-resents to the brain what the eye has seen.

RETINAL FATIGUE supports the cell-assembly theory by caus-ing images to fade in a peculiar fashion. The apparatus Þxesan image on receptors until their signal decays. Then linesdrop out, one or two at a time, until the Þgure is gone. Hebbargued that each line was represented by a neuronal feed-back loop. When the retinal signal falls below the critical val-ue, the loop stops oscillating, and the line disappears.


Lashley again, although he was not at Þrst very enthusiastic about work-ing with chimpanzees. LashleyÕs inten-tion was to develop tests of learningand problem solving for the animals,while Hebb would study their person-alities and emotional characteristics.Then they would start a program to de-termine how brain lesions aÝected arange of variables.

The chimpanzees proved more diÛ-cult to train than Lashley had imagined.The delays meant that no brain opera-tions were carried out during HebbÕstenure at Yerkes. Nevertheless, he wasfascinated by his observations of chim-panzees and said he learned more abouthuman personality in his Þve years ofwatching chimpanzees than at any oth-er time since his own Þrst Þve years of life. The apes manifested distinctpersonalities and a sense of fun thattended toward slapstick. Hebb and theother members of the staÝ derived amore cerebral amusement from the ver-bal contortions of orthodox behavioristvisitors as they attempted to describethe animalsÕ practical jokes and broadclowning without resorting to Òmental-isticÓ language.

HebbÕs long and close observationof the many chimpanzees in theprimate laboratory taught him

that experience was not the only fac-tor in the development of personality, including pathological manifestationssuch as phobias. He showed, for exam-ple, that young chimpanzees, born inthe laboratory and known never to haveseen a snake before, are frightened theÞrst time they are shown one. Chim-panzees are also frightened of modelsof chimpanzee or human heads or oth-er isolated body parts or of familiarcaretakers wearing unusual clothing.Moreover, Hebb was one of the Þrst toobserve the social behavior of captiveporpoises and to suggest that it im-plied a level of intelligence comparableto that of the apes. His observationsmay have inßuenced his later conclu-sion that level of play provides a goodindex of intelligence.

LashleyÕs interest in the ways thebrain categorizes perceptions intoknowledge about the world rekindledHebbÕs curiosity about concepts andthinking. The problem can be rephrasedas a question: How does the brain learnto lump one triangle, car or dog withanother even though no two triangles,cars or dogs produce the same patternof stimulation on sensory receptors?

The turning point came when Hebbread about the work of Rafael Lorentede N�, a neurophysiologist at the Rocke-feller Institute for Medical Research,

who had discovered neural loops, orfeedback paths, in the brain. Up to thatpoint, all psychological theories, wheth-er physiological or not, assumed that in-formation passed through the organismalong a one-way track, like food throughthe digestive system. Hebb recognizedthat LorenteÕs looping paths were justwhat he needed to develop a more real-istic theory of the mind.

Feedback was not entirely new inlearning theory. Almost all models as-sumed that the output of the organisminßuences the input in some way, forinstance, by enabling the animal to re-ceive a reinforcing stimulus. Unfortu-nately, feedback proceeding in this way,through a single path, would operateslowly and often unreliably. But withmillions of internally connected feed-back paths, it would clearly be possibleto establish internal models of the en-vironment that might predict the effectsof possible responses without having tomove a muscle.

HebbÕs specialization in vision ledhim to concentrate his early neural the-ories on that system. Knowing that thepoint-to-point projection from the reti-na to the cortex does not extend be-yond the primary visual cortex, he as-sumed that the neural relays projectedinto the surrounding cortex in randomdirections, thus scrambling the retinalpattern [see ÒThe Visual Image in Mindand Brain,Ó by Semir Zeki; SCIENTIFIC

AMERICAN, September 1992]. Such anarrangement could recombine signalsfrom diÝerent parts of the imageÑthatis, they could converge on the sametarget neuron, causing it to Þre. The re-sulting impulses could then return tothe earlier neurons in the path, closingthe feedback loops.

Repeated activation of any givenloop might then strengthen that loopin the following way. If the axon of anÒinputÓ neuron is near enough to excitea target neuron, and if it persistentlytakes part in Þring the target neuron,some growth process takes place in oneor both cells to increase the eÛciencyof the input neuronÕs stimulation. Syn-apses that behave according to thispostulate became known as Hebb syn-apsesÑsomewhat to HebbÕs amuse-ment, it may be said, because this pos-tulate is one of the few aspects of thetheory he did not consider completelyoriginal. Something like it had beenproposed by many psychologists, in-cluding Freud in his early years as aneurobiologist.

Nevertheless, HebbÕs postulate wasthe most clear and formal statement,although in 1949 it was pure specula-tion. Since then, however, studies ofsingle neurons have conÞrmed thatsynaptic strengths do change in someneurons in accordance with the postu-late. Hebb may also have been correctabout the mechanism of permanentchange. A former student of his, AryehRouttenberg of Northwestern Universi-ty, has recently pointed out that a pro-tein associated with neuronal growth isproduced when neurons are stimulatedin ways that increase synaptic strength.

Hebb assumed that most of the syn-apses in the cortical lattice are initiallytoo weak to Þre spontaneously. To Þre,they would require the converging ofstimulation from a number of activeneurons. Some neurons in the latticereceive converging inputs and thus Þrewhen a particular pattern of neurons inthe sensory cortex is Þred by a stimulus.Some of the activated neurons have


ISOLATION EXPERIMENT carried the study of sensory deprivation beyond therealm of individual cell assemblies. CuÝs prevented touch, a plastic shield disrupt-ed pattern vision and a U-shaped foam cushion attenuated sounds not masked by the air conditioner in the ceiling. EEG electrodes recorded the subjectÕs brainwaves, and a microphone enabled him to report his experiences. The volunteersÕability to think deteriorated, and some of them even started to hallucinate.


synaptic connections with one another,which are also strengthened wheneverthe stimulus is presented. Eventuallythe connections between the simulta-neously Þring neurons in the lattice be-come strong enough for them to con-tinue Þring one another in the absenceof input from the stimulus, creating aninternal representation of the stimulus,called a Òcell assemblyÓ by Hebb.

The concept of the cell assembly,in my view, was HebbÕs greatestcontribution to psychological the-

ory, not to mention philosophy. It re-vived the 19th-century psychologistsÕattempt to explain behavior in terms ofthe association of ideas, a project thatthe behaviorists had derailed by argu-ing that ÒideasÓ were no more real thanthe notion of little men inside the head.By so arguing, the behaviorists main-tained that ideas, and thus mentalism,had no place in scientiÞc psychology.

Unfortunately, few seemed to noticethat the behaviorists replaced ideaswith equally insubstantial constructswith misleading names, such as Òstim-uliÓ and Òresponses.Ó These were notreal events or chains of events but at-tributes that became associated withone another in some imaginary blackbox that scientists were forbidden to re-fer to as the brain. Hebb put a stop tothis charade by showing, in principle atleast, that ideas could have just as Þrma physical basis as muscle movements.They could consist of learned patternsof neuronal Þring in the brain, initiallydriven by sensory input but eventuallyacquiring autonomous status.

In its original form the neural theorywas undoubtedly too simple to haveworked. A major problem was that thecell assembly did not incorporate inhi-bition, because contemporary sciencedid not recognize it. Sir John C. Eccles, avery inßuential neurophysiologist at theAustralian National University in Can-berra, was still vigorously denying theexistence of inhibitory synapses. More-over, many important connections ofthe neocortex had not yet been discov-ered, and the functional signiÞcance ofthe diversity of cortical neurons wasonly hinted at.

Without inhibiting factors, however,learning would strengthen synaptic con-nections until all neurons Þred continu-

ously, making the system useless. ThiseÝect was observed in computer mod-els of the cell assembly, called concep-tors, constructed in the 1950s by Na-thaniel Rochester and his colleagues atthe IBM research laboratory in Pough-keepsie, N.Y. Hebb himself seems neverto have set Þnger to a computer to testhis idea that random nerve nets couldorganize themselves to store and re-trieve information. But such so-calledneural nets later inspired many comput-er models, from the perceptron to par-allel distributed processing, and haveeven found applications in industry.

By the time The Organization ofBehavior reached publication,Hebb was back in Montreal as

chairman of McGillÕs psychology depart-ment. Ten years later, when he steppeddown as chairman, he had forged one ofthe strongest departments in NorthAmerica. He found it easier to buildwhat he wanted because the depart-ment was almost nonexistent when hebegan, and he turned out to be adeptat campus politics and soon discoveredhow to use his growing reputation toapply pressure where it would do themost good. It is perhaps signiÞcant thathe was also one of the best chess play-ers at the university.

Most of HebbÕs research at McGill

was related to his cell-assembly theo-ry. Experiments to obtain direct phys-iological evidence for the theory werefar beyond the scope of contempo-rary methodology. (They still are.) In-stead he tested behavioral predictionsof the theory. He tried, for instance, tostrengthen his earlier conclusions onthe inßuence of rearing on adult intelli-gence. Most of the results supportedhis theory that animals raised in an en-riched, or more complex, environmentwould, in later life, outperform animalsraised in bare cages.

There was one embarrassing excep-tion. Litters of pure-bred Scotties weresplit, and half the pups were reared aspets in the homes of members of thestaÝ and half were reared in cages inthe laboratory. Hebb was not fortunatein the choice of his puppy, Henry. Itwas congenitally incapable of Þndingits way around, invariably got lost assoon as it was out of sight of the houseand had to be recovered from the dogpound on several occasions. Naturally,Henry turned out to be near the bot-tom of the class when, as a full-growndog, it was tested in a maze.

In a related series of experiments,Hebb investigated the eÝect of impov-erished sensory input on the behaviorof adults, including human volunteers[see ÒThe Pathology of Boredom,Ó by


HEBBÕS INFLUENCE propagated as muchthrough his disciples as through his publi-cations. Here, in a graduate seminar fromthe early 1950s, Hebb appears at the farright, the author in the foreground. Theparticipants went on to pioneer the newÞeld of physiological psychology.

BEN DOANDalhousie University

PETER MILNERMcGill University

HELEN MAHUTNortheastern University

JIM OLDS Caltech (DECEASED)


Woodburn Heron; SCIENTIFIC AMERICAN,January 1957]. Students were paid gen-erously to undergo severe sensory dep-rivation for as long as they could standit (none lasted even a week). Their abilityto think began to deteriorate, and someof them even started to hallucinate. TheKorean War was then in progress, andmany workers attempted to use suchisolation experiments to understandand combat the ÒbrainwashingÓ tech-niques employed by the Chinese.

Hebb also pursued his old idea thatearly brain injury should be more dam-aging than injury in an adult. But the results were rendered uncertain by sev-eral factors, the most important beingthe capacity of the young brain to reor-ganize itself. For example, if an infantsustains an injury in an area of the lefthemisphere that is important for speechin the adult, the right hemisphere takesover this function, and speech is not se-riously impaired. But if an adult sustainsdamage in the same area, the result maybe a permanent loss of language skills.

Because of such problems with thestudy of cognition, Hebb came to be-lieve that the best evidence for the cellassembly came from experiments onretinal fading. Images of simple Þgureswere projected onto the eye by a verysmall lens system attached to a contactlens, ensuring that the image always

fell on the same place. As the receptorcells become fatigued, the image fadesand disappears, but not all at once.Usually entire lines disappear sudden-ly, one or two at a time, until the entireÞgure is gone. Hebb explained the phe-nomenon by saying that each line isrepresented by neuronal activity cir-culating in a closed loop. The activity,once started, continues even after theinput from the retina has decayed to alow value because of feedback aroundthe loop. But at some critical value thereverberation stops abruptly, and theline disappears. These experiments donot provide conclusive evidence for thecell assembly as Hebb envisaged it. Yeteven if HebbÕs version should turn outto be incorrect, it would not diminishthe value of his idea that some neuralactivity continues to symbolize an ob-ject even after the object has stoppedstimulating the sense organs.

Had The Organization of Behavior

consisted only of the chapters in which Hebb criticizes current

approaches and elaborates his cell-as-sembly theory, it is likely that few peo-ple would have read it. The bookÕs ap-peal lies in its second half, in whichHebb discusses emotion, motivation,mental illness and the intelligence ofhumans and other species in the light

of his theory. These essays are refresh-ingly forthright. On mental health, forexample, Hebb wrote: ÒWe still need anAjax to stand up and defy the lightningand ask, What is the evidence? whensome authority informs the public thatbelieving in Santa Claus is bad for chil-dren, that comic books lead to psycho-logical degeneracy, that asthma is dueto a hidden mental illness.Ó

Hebb built his department and hisÞeld by capturing the interest and imag-ination of the best students at an earlystage. He taught the introductory coursehimself, making it immensely popularÑat one point it numbered 1,500 stu-dents, about half the yearly undergradu-ate enrollment. Many future professorsof psychology found their calling inthese lectures. Like most of what Hebbdid, his course was unique; no textbookat the time came close to including thematerial and ideas he dealt with, so he wrote his own. The Þrst edition of A Textbook of Psychology appeared in1958. In contrast to the majority of in-troductory texts of the day, it had moreideas than pictures.

Hebb also gave a graduate seminarthat was attended by every psychologygraduate student at McGill over a peri-od of 30 years. It was famous not onlyfor its stimulating discourse but alsofor HebbÕs ever-present stopwatch andthe slips of paper on which he notedincorrect pronunciations and other er-rors of presentation. It was HebbÕs am-bition never to have a McGill studentoverrun his or her allotted time at ameeting, and on the whole he was suc-cessful. McGill honored Hebb in 1970by naming him chancellor; he becamethe only faculty member ever appoint-ed to that position.

In 1977 Hebb retired to his birthplacein Nova Scotia, where he completed hislast book, Essay on Mind. He was ap-pointed an honorary professor of psy-chology at his alma mater, Dalhousie,and regularly participated in colloquiathere until his death, at 81, in 1985.


FURTHER READINGTHE ORGANIZATION OF BEHAVIOR: A NEU-ROPSYCHOLOGICAL THEORY. D. O. Hebb.John Wiley, 1949.

ESSAY ON MIND. D. O. Hebb. LawrenceErlbaum Associates, 1980.

PARALLEL LEARNING IN BRAINS AND MA-CHINES. G. Ferry in New Scientist, Vol. 109,No. 1499, pages 36Ð38; March 13, 1986.

TEXTBOOK OF PSYCHOLOGY. Fourth edi-tion. D. O. Hebb. Lawrence Erlbaum As-sociates, 1987.

MIND AND BRAIN. Special Issue of Scien-tific American, Vol. 267, No. 3; Septem-ber 1992.

G. ROLFE MORRISONMcMaster University

SETH SHARPLESSUniversity of Colorado, Boulder

DONALD O. HEBBMcGill University(DECEASED)

UNIDENTIFIED


TRENDS IN NONLINEAR DYNAMICS

ADAPTING TO COMPLEXITYby Russell Ruthen, staÝ writer

n an autumn afternoon John Holland strollsamong the pines and aspens that blanket,like a green and golden patchwork quilt, the

mountains just north of Santa Fe, N.M. What im-presses Holland about the scene is not just itsbeauty but also its complexity. The aspens com-pete with the pines for light as well as for waterand nutrients from the rocky soil. As each treegrows and adapts, it alters the supply of resourcesavailable to its neighbors, and in doing so, it chang-es its own chances for survival.

Holland and his colleagues at the Santa Fe Insti-tute in the valley below are searching for a uniÞedtheory that would explain the dynamics of all liv-

ing systems, be they groves of trees, colonies ofbacteria, communities of animals or societies ofpeople. All are systems of many agents, each ofwhich interacts with its neighbors and, most im-portant, can adapt to change.

The loose group of economists, biologists,physicists, computer scientists and researchersfrom many other disciplines who have assem-bled at the institute hope to discern the principlesthat underlie all such complex adaptive systems.In the process, they have devised new theoreticalframeworks, mathematical tools and computersimulations that attempt to capture the essenceof complex adaptive systems. And they are gain-


ing a better understanding of what complexity is,why it emerges naturally and why living systemsseem to evolve toward a boundary between or-der and randomness.

Indeed, the Santa Fe group is trying to answerthe question of evolution in the broadest sense of the word. ÒAs these various systems organizethemselves and learn and remember, evolve andadapt, persist and eventually disintegrate and dis-appear, what common patterns and fundamentalprinciples, if any, shape their remarkable behav-ior?Ó wonders George A. Cowan, one of the found-ers of the Santa Fe Institute. Holland, a computerscientist at the University of Michigan who helped

pioneer the field, has no illusions about the specu-lative nature of the enterprise. When asked whatkind of complex adaptive systems scientists real-ly understand, he just chuckles. ÒThe really safeanswer is none at all.Ó

From a primeval sea of organicmolecules arose plants, animals, globalecosystems, intelligent beings, interna-

tional organizations. What drives thenatural world toward complexity?

QUAKING ASPENSÑand all living thingsÑdisplay com-plexity at many levels of structure as the numerous com-ponents of the system grow and adapt. Leaves, branchesand roots arise from the interaction of cells and their ge-netic material. Forests and ecosystems emerge from theinterplay between the trees and their environment. Topredict the behavior of such adaptive systems, research-ers are consolidating the theories of evolution, the lawsof dynamics and the power of computer simulation.



To be sure, this generation of scien-tists is not the Þrst to attempt to un-derstand the behavior of these sys-tems. In the 18th century Newton andhis contemporaries hoped they couldpredict the behavior of any complexsystem by identifying all the parts andstudying all the interactions. But thathope soon vanished.

In the 19th century Sadi Carnot andhis peers realized it was impractical, ifnot impossible, to describe every inter-action, because most physical systemscontained zillions of parts. They discov-ered, however, that they could predictthe statistical behavior of a system aslong as the parts were identical and theinteractions were weak. These statisticalpredictions became the laws of ther-modynamics; they could explain, for ex-ample, the increase in temperature andpressure when gas molecules are heat-ed in a container. Yet thermodynamics,Carnot knew, did not provide a com-plete description of the most complexsystems, particularly those in which theinteractions are not weak, as they wouldbe if the gas molecules were strongly at-tracted to one another.

From Order to Chaos

During the past 100 years, HenriPoincar� and his intellectual disciplesrealized that if a system consisted of a few parts that interacted strongly, itcould exhibit unpredictable behavior.They invented chaos theory. For exam-ple, if three planets orbit around oneanother inßuenced only by the force ofgravity, it is impossible to predict themotions for a long period of time, even

if the positions and velocities of theplanets are known with great accuracy.

Beginning in the 1970s, physicistsßirted with the idea that chaos theorycould account for the behavior of com-plex systems. Although chaos theoryprovided many mathematical tools use-ful for the study of complexity, it didnot capture the wide range of dynamicsexhibited by complex systems. ÒThe ex-plosive development of computer hard-ware and the development of new math-ematical concepts,Ó Cowan says, Òhelpedraise the level of interest in complexitybeyond explorations of chaos.Ó

In fact, many years before the theoryof chaos became popular, a small groupof scientists, including Herbert Simon,Ilya Prigogine, Herman Haaken and oth-ers, had been searching for a theory ofcomplexity, the general principles for asystem whose many parts interact toproduce complex behaviors. These work-ers had some success in describingcomplexity in physical systems.

When Cowan began directing the San-ta Fe Institute eight years ago Òto nur-ture research on complexity,Ó investiga-tors there began to wonder if they couldapply the emerging theory of complex-ity to adaptive systems, such as cells, or-ganisms and economies. Adaptive sys-tems, like other complex systems, con-sist Òof many relatively independentparts that are highly interconnectedand interactive,Ó Cowan explains. ÒThesystems of interest are not in thermo-dynamic equilibriumÑthey metabolize,absorbing energy from an externalsource and dumping back waste.Ó

An adaptive agent can be anythingfrom a single-cell organism to human

society; it must be capable of formingand changing strategies. For instance,an aspen will orient its leaves to cap-ture the most sunlight, a strategy thatis encoded in its genes, and can changethrough mutation or recombination ofits genetic material. Complex systemscan Òlearn and adapt to changing con-ditions in the environment with whichthey interface,Ó Cowan says.

An adaptive system, then, consistsof many agents, each interacting withothers that have the same or diÝerentstrategies. The systemÕs behavior ob-viously depends on the sophisticationof the strategies and the mechanism of change. Yet adaptive agents can de-velop good solutions to extraordinarilydiÛcult problems. Holland is intriguedat the way a group of independent com-panies somehow manage to supply foodfor the seven million people in NewYork City. The companies constantlyreplenish the restaurants and super-markets, which do not keep more thana few daysÕ worth of food in reserve,without creating shortages or surplus-es. ÒFrom the point of view of physics,it is a miracle that happens without anycontrol mechanism other than sheercapitalism,Ó Holland exclaims.

To analyze the behavior of such sys-tems, researchers must rely heavily oncomputer simulation because the equa-tions that describe the forces underly-ing complex processes prove too dif-Þcult to solve. Holland, for one, simu-lates the process by which economicand biological agents adapt [see ÒGe-netic Algorithms,Ó by John H. Holland;SCIENTIFIC AMERICAN, July 1992]. ÒI cancaricature a process and get a model

��

@@@@@@@@@@@@@@@@@@@@@

��

ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀ

��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

yyyyyyyyyyyyyyyyyyyyy

��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��@@@@@@��ÀÀÀÀÀÀ��yyyyyy

21


SAND PILE illustrates the theory of self-organized criticality.When grains of sand fall on a ßat pile, they quickly Þnd a

resting spot (1). As grains continue to fall, the pile grows to acritical state in which a falling grain can trigger an avalanche


that is almost certainly overly simplebut captures features in the same sensethat a political cartoon captures some-thing,Ó he comments.

Echoing Evolution

Four years ago Holland began workon a simulation of an ecosystem inwhich digital organisms try to surviveand reproduce. He calls it ÒEcho,Ó a playon the words ÒecosystemÓ and Òecho.ÓAll organisms are endowed with Òchro-mosomesÓ that encode oÝensive and de-fensive strategies. Each organism wan-ders through an artiÞcial environment,searching for resources and absorb-ing them into an internal reservoir. Ifone organism encounters another, thetwo Þght, and the winner consumes the loser, acquiring all the resourcescontained in the loser. If an organism obtains enough resources, it produc-es oÝspring whose chromosomes maycontain mutations. New species arisewith ever more elaborate strategies foroÝense and defense.

Holland was surprised that such asimple simulation could generate speci-ation and arms races. Yet it was missinga key feature of evolutionary processes:cooperation. Organisms, or for that mat-ter companies and nations, learn to co-operate even as they may compete ag-gressively with one another. Hollandthinks he has a possible answer, whichwas inspired by the work of Robert Ax-elrod on a problem known as the Pris-onerÕs Dilemma. In this problem, the po-lice take two criminals into custody toextract a confession. If both prisonersconfess, they both go to jail. If one con-

fesses and the other does not, the con-fessor is granted immunity and is setfree; the one who keeps silent goes tojail. If both prisoners keep silent, theyare set free, although they risk pros-ecution if new incriminating evidence is found.

Clearly, the two prisoners would ben-eÞt by cooperating. Yet both will betempted to seek immunity and fear thatthe other will confess. This logic leadsthem both to confess. Similarly, two or-ganisms may beneÞt from cooperation,but one may receive the greatest payoÝif it can obtain the resources of anotherwithout sacriÞcing its own. The upshotis the organisms may Þght.

If adaptive agents are confrontedmany times with a situation similar tothe PrisonerÕs Dilemma, they developstrategies that include both combat andcooperation. One of the best strategiesis known as tit for tat: an agent will co-operate the Þrst time it encounters anopponent and thereafter mimics theprevious response of its opponent. Ifagent A tries to cooperate and agent Battacks, agent A will Þght on the nextround. But if agents A and C cooperateinitially, they will continue to do so.

This dynamic was just what Hollandneeded for his digital ecosystem. Hemade three major modiÞcations. First,each organism can choose to Þght forresources or trade for them. Second,each is assigned a tag, which is analo-gous to a molecular marker on themembrane of a cell. Last, each organ-ism is allowed to develop rules such asÒÞght only those organisms that carrya red tag.Ó If strategies and tags are ini-tially assigned at random, some organ-

isms, by luck, are more inclined thanothers to use a strategy resembling titfor tat. Those that play such strategiesare then more likely to survive and re-produce, thereby perpetuating the exis-tence of particular tags.

The changes had an immediate ef-fect. ÒWhat started out as a completelyrandom juxtaposition of tags and strat-egies begins to develop in a very orga-nized way,Ó Holland declares. Eventu-ally, the organisms ÒlearnÓ to associatetags with certain strategies. In the end,the digital ecosystem displays both spe-ciation and cooperationÑand such de-viations as mimicry and lying.

The apparent ability of HollandÕs sim-ulations to reßect nature have piquedthe interest of a small group of econ-omists, including W. Brian Arthur ofStanford University and Nobel laureateKenneth J. Arrow, who believe econo-mies can also be regarded as complexadaptive systems. ÒWe must view theeconomy as an evolving, changing sys-tem,Ó Arthur remarks.

Such activities as international trade,high-technology business and the emer-gence of new companies exhibit trulycomplex dynamics, Arthur argues, be-cause they involve both negative andpositive feedback mechanisms. Negativefeedback maintains the balance betweensupply and demand. Positive feedback,on the other hand, is destabilizing. If aÞrm gets ahead, it gains further advan-tage so that a clever strategy or luck candecide what company becomes a leaderin a Þeld [see ÒPositive Feedbacks in theEconomy,Ó by W. Brian Arthur; SCIENTIF-IC AMERICAN, February 1990].

Yet a mixture of positive and negative


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��

ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀ

��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

yyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy

��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

��


��

yyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy

��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��



��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��

@@@@@@@@@@@@@@@@@@@@@

��


��



3 4

of any size (2 ). Because of the avalanches, the slope of thepile remains constant. If some external force alters the shape

of the pile but is then removedÑa box, in this case (3 )Ñthepile will return to the critical state (4 ).


feedback presents two challenging prob-lems for economists. First, it means thatany given economic system can evolvedown many possible paths. Economistsmust therefore Þgure out how a systemends up on one particular path. Second,a Þrm that is choosing a product strat-egy must try to guess which strategiesother Þrms might use, knowing thatsmall changes in its own strategy mightalter the direction of others. Thus, econ-omists must decide how Þrms behave insuch complicated environments.

To solve these problems, Arthur andhis cohorts are using ÒHolland-likeÓcomputer simulations. From their per-spective, the economy consists of manyÞrms (or consumers), each continually

forming hypotheses about the environ-ment they face and the strategies oftheir rivals. As a Þrm adds to its knowl-edge, it strengthens its belief in somehypotheses while discarding others. Itchooses a strategy from these hypoth-eses, thereby inßuencing the environ-ment in which other Þrms form theirhypotheses. ÒThe resulting dynamicsare so complicated,Ó Arthur says, Òtheycan only be studied on a computer.Ó

Arthur is bemused by the idea in con-ventional economic theory that eachagent arrives at the absolute best strat-egy given the circumstances. The theo-ry means, he quips, that Òan economicagent can solve in an instant a problemthat a theoretical economist might have

to think hard about to formulate over aperiod of two years.Ó Arthur hopes thestudy of complex adaptive systems willlead to a much better understanding ofhow economic agents form strategies.

Yet computer simulations go only sofar. Although they mimic some of theattributes of complex adaptive systems,they provide little help to researcherswho want to examine the complexity ofexisting systems. Some of the best ideasfor deÞning and measuring complex-ity emerged from the work of Claude E. Shannon, Andrei N. Kolmogorov andGregory J. Chaitin during the 1950s and1960s. They proposed that the degreeof complexity is related to the size ofthe smallest description of a systemÕsbehavior. This theory led to a proposedtechnique for measuring the so-calledalgorithmic complexity of a process.

Computing Complexity

In theory, the complexity of two sys-tems could be compared by writing twocomputer programs that are the short-est capable of reproducing the originaldata. One data set, for example, mightbe a daily count of the number of leaveson an aspen during many seasons, andthe other set might be the number ofneedles on a pine. The algorithmic com-plexity of the data is then related to thenumber of instructions in the comput-er program. The program with the few-est instructions would describe the lesscomplex system.

The hitch is that algorithmic complex-ity is usually impossible to compute,notes James P. CrutchÞeld of the Univer-sity of California at Berkeley. To mea-sure algorithmic complexity accurately,a programmer would need to use a the-oretical type of computer known as auniversal Turing machine. Proposed byAlan M. Turing in the 1930s, this ma-chine could perform any computation,given enough time. A universal Turingmachine would have inÞnite memory,and a program could access or changeany part of its memory. Such a machinewould preserve every detail of the dataon aspen leaves or pine needles. Indeed,the computer could simulate any physi-cal process, although the programmermight literally have to wait from nowto eternity to Þnd out the results.

Other problems stand in the way ofalgorithmic complexityÕs becoming apractical tool for studying complex sys-tems. For one thing, the technique ismuch more sensitive to randomnessthan order. A universal Turing machinewould require many more instructionsto reproduce a series of random eventsthan it would take to simulate a periodicprocess. ÒIf you are working with a com-


NEW YORK CITY PARADOX: How is it possible for a group of independent compa-nies to supply food for the millions of people in the city when most restaurants andsupermarkets keep only a few daysÕ worth of food in reserve? More generally, howis it possible for adaptive agents to Þnd good solutions to complicated problems?


pletely deterministic Turing machine,ÓCrutchÞeld says, Òyou have to introducemany computational steps to generaterandom numbers.Ó Moreover, there is nogeneral method for Þnding the minimalprogram that simulates a given set ofdata. Even though a program reproducesthe data, a researcher still cannot be sureit is the shortest oneÑor, if the data aresuÛciently complex, whether the pro-gram is even close to the shortest one.

As a result, workers continue tosearch for new methods. In the late1980s CrutchÞeld and his colleaguesproposed a scheme based on the workof the linguist Noam Chomsky and oth-ers who developed a system for clas-sifying diÝerent kinds of computers.The universal Turing machines fall inthe most powerful ÒmodelÓ class. Otherkinds of computers are categorized ac-cording to the size and structure oftheir memories as well as other param-eters. For example, one class includesthose computers whose memory is inÞ-nite but is organized as a stack.

To measure complexity with this hi-

erarchy of model classes, a researcherwould Þrst choose a model class, thenattempt to Þnd the shortest programthat reproduces the data. Given the con-straints of the class, the investigator isforced to approximate the data, therebypreserving some essential features anddiscarding some random detail. If theprogram cannot reproduce the data, theexercise is repeated on a model classwith more power. If the program doesreproduce the data, the complexity isthen related to the number of instruc-tions and the number of constraints as-sociated with the model class.

CrutchÞeldÕs ideas may also be use-ful to investigators searching for waysto simulate complex processes. If aprogrammer attempts to reproduce thedata set from a complex process butchooses a model class that is too sim-ple, the simulation will generate only a rough approximation of the originaldata. To improve the simulation, theprogrammer must select a new modelclass from an inÞnite number of choic-es. CrutchÞeld has found that by com-

paring a series of increasingly better ap-proximations to the original data, scien-tists can Þnd clues as to what choice ofmodel class would be an improvement.It can help discover what kind of com-puter model is necessary to reproducesome process. ÒThere is a way to look atthe data,Ó CrutchÞeld remarks, Òand inthe data themselves is information abouthow to innovate to a new model class.Ó

Yet all the attempts to deÞne, mea-sure and model complexity do not ad-dress the fundamental question: Why isthere so much complexity in the world?Why do some systems, both adaptiveand nonadaptive, seem to evolve awayfrom the extremes of complete orderand complete randomness?

AÛnity for Criticality

Some researchers, most notably PerBak, a senior researcher at BrookhavenNational Laboratory, believe that one of those principles is a phenomenonknown as self-organized criticality. Dur-ing the late 1980s, Bak and his collabo-


JOHN H. HOLLAND, a computer guru at the University of Mich-igan, created a digital ecosystem Òto facilitate gedanken ex-periments on complex adaptive systems. It oÝers a way,Ó he

says, Òto build intuitions about typical interactions and theemergence of structure in such systems.Ó The ecosystem ex-hibits such lifelike properties as speciation and cooperation.


rators found a class of systems that ap-pear to evolve toward complexity. Theclearest example of these systems is apile of sand [see ÒSelf-Organized Criti-cality,Ó by Per Bak and Kan Chen; SCI-ENTIFIC AMERICAN, January 1991].

If grains of sand are dropped ontothe center of a pile, the system can ex-hibit three types of behavior: subcriti-cal, critical or supercritical. If the sandpile is ßatÑthe subcritical stateÑfallinggrains quickly come to rest and maytrigger small avalanches of sand on thesurface. As grains continue to fall, thepile grows and eventually achieves acritical state. In this state, adding grainsof sand triggers avalanches of any size,from one grain to the entire surface ofthe pile. As more grains are added and

more avalanches occur, the pile will con-tinue to grow, even as its slope remainsconstant. If the sand pile is then shapedso that the slope is steeper than thecritical valueÑthe supercritical stateÑadding a few grains of sand will triggeran enormous avalanche. The pile, onceagain, returns to the critical state.

The sand pile increases in complex-ity, growing out of the subcritical state,or orderly phase, and collapsing whenplaced in the supercritical state, or random phase. As long as energy isadded to the systemÑthat is, grains ofsand are droppedÑthe system will re-main in the critical state, and its dynam-ics will never settle. ÒNo matter whatyou do to the systemÑyou can perturbas much as you likeÑit will always

return to the critical state,Ó Bak asserts.In recent years scientists have gath-

ered evidence that suggests self-orga-nized criticality plays a role in manycomplex systems, from chemistry to ge-ology and perhaps biology. ÒSince Dar-win, we have come to think of organ-isms as tinkered-together contraptionsand selection as the sole source of or-der,Ó explains Stuart A. KauÝman, a the-oretical biologist at the University ofPennsylvania. ÒYet Darwin could nothave begun to suspect the power of self-organization. We must seek our princi-ples of adaptation in complex systemsanewÓ [see ÒAntichaos and Adapta-tion,Ó by Stuart A. KauÝman; SCIENTIF-IC AMERICAN, August 1991].

The Edge of Chaos

The theory of self-organization res-onated strongly with the work of KauÝ-man, Bernard Derrida of the Saclay Re-search Centre, Christopher G. Langton of Los Alamos National Laboratory andNorman Packard of the University of Illi-nois. In 1985 Derrida showed analytical-ly that a system goes through a phasetransition from order to randomness ifthe strength of the interaction betweeninterconnected agents is gradually in-creased. A few years later Langton dem-onstrated that a system can perform the most sophisticated computationsat the boundary between order and randomness. Meanwhile Packard began creating computer simulations showingthat complex adaptive systems tend toevolve toward the boundary throughthe process of natural selection.

Working from these ideas, KauÝmanand others have proposed that organ-isms change how strongly they interactwith others in such a way that theyreach the boundary between order andrandomness, thereby maximizing theaverage Þtness of the organisms. ÒThiscandidate principle,Ó KauÝman warns,Òis just that, a candidate principle.Ó

To illustrate why the principle is plau-sible, KauÝman invokes a variety of bio-logical systemsÑa pond harboring frogsand ßys, for example. In the putativepond the frogs and ßies have access tolimited resources, and they have sur-vival strategies that are encoded in theirgenetic material and nervous systems.Each frog or ßy can change its strategyin an attempt to maximize its Þtness,which can arguably be deÞned as thechance that it will survive long enoughto reproduce.

At any instant in time, a frog or a ßywill have a strategy that yields a partic-ular Þtness level. For each organism,the set of all possible strategies can berepresented by a surface. Each point on


W. BRIAN ARTHUR of Stanford University views the economy as a complex adap-tive system. ÒA satisfactory theorizing in economics cannot proceed,Ó he says,Òwithout facing and resolving the key issues of learning, adaptation and cognition.Ó


the surface represents a diÝerent strat-egy; the height of each point indicatesthe Þtness level of a particular strategy.As the organism changes its strategy, it moves to new points on the surface.Yet since an organism can only changeits strategy to a certain degree, it doesnot make large jumps across the sur-face. It might be advantageous, for in-stance, for a frog to learn to ßy, but itis not going to suddenly sprout wings.

The dynamics of the frog-ßy systemdepends greatly on how strongly thetwo species interact. If the frogs do notÞnd the ßies tasty at all, then the Þtnessof the frog population will be indepen-dent of the Þtness of the ßies. The ßiescan attempt to maximize their Þtnesswithout worrying about the frogs, andvice versa. For each organism, there area few strategies that would improve Þt-ness, and there are many strategies that

would make things worse. The Þtnesssurface of each organism, therefore, hasa few well-deÞned bumps (high Þtness)and a large, ßat valley (low Þtness). Thefrog-ßy system is in an ordered, static,subcritical state.

The dynamics of the system becomemuch more complicated, however, if theÞtness surfaces of the organisms arecoupled, that is, if the frogs depend onthe ßies as a food source. As the ßies de-velop new strategies, they will alter theresources available to the frogs, there-by inßuencing the Þtness of the frogs. Achange in strategy thus means a changein the contours of the surface. Conse-quently, the dynamics depend on justhow strongly the ßies and frogs interact.

If the coupling is very strong, on theone hand, any slight change in strategyis likely to change the character of thewhole Þtness surface. No frog or ßy

can adjust to the rapidly changing sys-tem. So the Þtness surface of the aver-age organism will have many short hillsand will always be changing. The frog-ßy system is in a random, dynamic, su-percritical state.

If the coupling is neither too strongnor too weak, on the other hand, the dy-namics become truly complex. In thiscase, the frogs may develop a success-ful strategy that works for a long timebefore the ßies Þnd a new strategy caus-ing a major change in the Þtness sur-face. The Þtness surface of each organ-ism will change over time, displayingmany bumps that vary greatly in size.The frog-ßy system is in a complex, dy-namic, critical state.

ÒIt turns out,Ó KauÝman claims, Òthatin a wide variety of coupled systems thehighest mean Þtness is at the phasetransition between order and chaos.Ó Byselecting an appropriate strategy, organ-isms tune their coupling to their envi-ronment to whatever value suits thembest, KauÝman asserts. And if they ad-just the coupling to their own advantage,he believes, they will reach the boundarybetween order and randomnessÑtheregime of peak average Þtness. ÒThebold hypothesis is that complex adap-tive systems adapt to and on the edge ofchaos,Ó KauÝman declares. ÒIt now be-gins to appear that systems in the com-plex regime can carry out and coordinatethe most complex behavior, can adaptmost readily and can build the mostuseful models of their environments.Ó

Researchers at Santa Fe and else-where are just beginning to think aboutways in which this framework and oth-er new insights into complex adaptivesystems can be proved. But KauÝmanis conÞdent that more robust modelsand further experiments will support aview of evolution that bridges livingand nonliving systems. ÒEvery attemptto Þnd something that is being maxi-mized in evolution has always met withfailure,Ó KauÝman observes. ÒYet I havethis feeling that there is something verygeneral going on about how far fromequilibrium systems have organizedthemselves. I donÕt know what thatsomething is yet. But I can taste it.Ó


STUART A. KAUFFMAN is searching for the principles of adaptation in complexsystems. Do all living systemsÑbacteria, humans, societiesÑevolve toward aboundary between order and randomness?

FURTHER READINGEXPLORING COMPLEXITY: AN INTRODUC-TION. Gr�goire Nicolis and Ilya Prigogine.W. H. Freeman and Company, 1989.

COMPLEXITY: THE EMERGING SCIENCE ATTHE EDGE OF ORDER AND CHAOS. M.Mitchell Waldrop. Simon & Schuster,1992.

COMPLEXITY: SCIENCE ON THE EDGE OFCHAOS. Roger Lewin. Macmillan, 1992.


From cocaine to quinine, about onequarter of U.S. prescription drugscontain at least one compound

derived from plants. Yet in recent yearsplants lost their cachet at the big phar-maceutical Þrms as new ideas camefrom microbes or variants gleaned fromhuge data bases of synthetic chemicals.Now drug companies are emphasizingtheir roots. Botanists are once againavidly scouring the worldÕs Þelds andforests to locate plant sources for newdrugs. ÒSynthetics havenÕt proven to bethe panacea,Ó observes Mark J. Plotkinof Conservation International, a Wash-ington, D.C.Ðbased group that is work-ing to stop destruction of tropical for-ests. ÒThereÕs still no cure for AIDS orthe common cold.Ó

The lure is twofold: a ÒgreenÓ publicrelations boon and the prospect of dis-covering the next taxol, a treatment forovarian cancer that was originally ex-tracted from the bark of the relativelyrare PaciÞc yew tree. Botanists estimatethat 10 percent or less of the more than250,000 ßowering plant species havebeen surveyed for pharmacological ac-tivity. But the task of Þnding an activemolecule in a haystack is being eased bydevices that can quickly search throughtens of thousands of samples.

One molecule in 10,000 may get tomarket through random screening. Au-tomated methods that rely on enzymesand other chemical tags can enable acompany to test as many as 150,000samples annually, hundreds of timesmore than was possible by testing chem-icals in laboratory animals. ÒOnce youget to that level, youÕre bound to Þndsomething,Ó says Alan L. Harvey, direc-tor of the Strathclyde Institute for DrugResearch at the University of Strath-clyde in Scotland.

Among those looking is Monsanto,which has contracted with the MissouriBotanical Garden to supply the compa-ny with several thousand plants fromboth the U.S. and tropical countries.Merck & Co. has an arrangement withthe New York Botanical GardenÕs Insti-tute of Economic Botany for receivingplants from around the world. Biotics,a small start-up company in England

based at the University of Sussex, em-bodies another approachÑthat of aÒscience broker.Ó It has supplied Glaxoand SmithKline Beecham and otherswith samples and chemical extractsfrom plants gathered from the tropics.

Many of the plant-screening programsare so recent that the leads they havegathered remain shrouded in the secrecythat cloaks drug development programs.But the wraps have come off a few ofthe results. SmithKline, for example, ispushing a cancer treatment called to-potecan through clinical trials. An en-zymatic assay revealed this compound,which comes from a Chinese tree, Camp-

totheca accuminata. The active principalis derived from a chemical extractedfrom a tree that the National Cancer In-stitute (NCI) tagged for potential anti-cancer activity about 20 years ago.

The NCI has been a locus for search-ing for new drugs from plants. It mount-ed what may be the worldÕs most ex-tensive plant-testing program in 1986, replacing a 20-year screening projectlaunched in 1960. Although the earlier

eÝort (which relied on mice as screeningagents) produced taxol, it was scrappedin 1980 because it had not turned upenough leads. The eÝectiveness of tax-ol against ovarian cancer was not de-tected until the late 1980s.

So far the NCIÕs revived eÝort has col-lected 23,000 samples from 7,000 spe-cies in tropical areas and has identiÞedthree compounds that seem to workagainst AIDS in a laboratory dish. Allthree compounds have entered preclin-ical testing for toxicity.

One ÒhitÓ is an extract from a creep-ing vine that inhabits the canopies ofrain forests in Cameroon. This memberof the Ancistrocladus genus inhibits thereplication of the AIDS virus. The plantwas found in an attempt to searchacross the broadest possible taxonomicspectrum. It had no known medical useand still lacks a species name. ÒThereare very few people on the planet whohave ever seen this plant,Ó says JamesS. Miller, a botanist from the MissouriBotanical Garden. More recently a Uni-versity of Illinois team working for the


SCIENCE AND BUSINESS

TRADITIONAL MEDICINALS bring new ideas for Shaman Pharmaceuticals, a SanCarlos, Calif. , firm that interviews healers in Latin America, Africa and Asia.

Back to RootsDrug companies foragefor new treatments


the Healing Forest Conservancy, a foun-dation the company set up to promoterain-forest conservation.

More unusual is MerckÕs agreementto pay $1 million over a two-year peri-od directly to Costa RicaÕs National In-stitute of Biodiversity (INBio) for col-lecting plants, insects and microbes. Italso pledged to pay royalties for anyleads that turn into pharmaceuticals. Aportion of both the initial expenditureand any eventual royalties will help pre-serve wild areas in Costa Rica. MerckÕscollaboration with a nonproÞt institutein a country whose gross national prod-uct is less than the drugmakerÕs annualrevenues has generated intense inter-est in other nations between the trop-ics of Cancer and Capricorn.

INBio, which signed the accord withMerck in 1991, is now helping Indone-sia to set up a similar institute devot-ed to biodiversity. Some public-interestgroups have criticized the Merck-INBioagreement because of what they per-ceive to be a lack of accountability. ÒAnongovernmental organization doesnÕthave the right to sell genetic materialto the rest of the world,Ó says Jason W.Clay of Cultural Survival, an organiza-tion in Cambridge, Mass., that tries toÞnd uses for indigenous materials, suchas nuts for Ben & JerryÕs RainforestCrunch ice cream. INBio emphasizes thatit was established three years ago by agovernmental decree. A Merck oÛcialpointed out that others are not pre-cluded from looking for natural sam-ples in Costa Rica.

A Þve-year program set up last Juneby the National Institutes of Health, theNational Science Foundation and theU.S. Agency for International Develop-ment aims to promote biodiversity andestablish new economic ties with the de-veloping world. The program, with $1.5million in annual funding, will set upconsortia of universities, nonproÞt in-stitutes and industry to catalogue plantsand other organisms, with the goal ofisolating compounds that have phar-maceutical value.

Some environmentalists are rootingfor the drug companies. ÒIt is my hopeand my expectation that people are go-ing to Þnd something pretty soon thatis marketable,Ó says Michael J. Balick, di-rector of the Institute of Economic Bot-any at the New York Botanical Garden.This, he thinks, may lead to a Ògreaterappreciation of the true value of theseresources.Ó Even if that happens, timemay be running out for the rain forest.According to Cornell University profes-sor of biology Thomas Eisner, Òthe look-ing for new chemicals is going slowerthan the rate at which species are be-coming extinct.Ó ÑGary Stix

NCI discovered a Malaysian tree, Calo-

phyllum lanigerum, with potent anti-AIDS properties. Calanolide A, a com-pound from the tree, seems to workagainst an AZT-resistant form of theAIDS virus.

Another potential anti-AIDS drug wasdiscovered in a plant in Samoa. Paul A.Cox, a Brigham Young University pro-fessor of botany, collected Homalanthus

nutans on the recommendation of wom-en healers there. The NCI extracted pro-stratin from the plant, a chemical thatseems to protect immune cells frombeing destroyed by the AIDS virus. InSamoa, H. nutans was used as a treat-ment for yellow fever and other viralillnesses. ÒYou have accumulation ofknowledge in the culture that resem-bles human bioassay data,Ó Cox says.

Indeed, several entrepreneurial com-panies are turning to traditional med-icine to narrow their search for newdrugs. Their idea, like CoxÕs, is to testmaterials that have shown pharmacolog-ical activity for hundreds of years andthat are probably relatively nontoxic.Chemex Pharmaceuticals in Fort Lee,N.J., recently won approval from theFood and Drug Administration to mar-ket a treatment for precancerous skin

lesions. Called Actinex, the drugis derived from the creosote bush,which has a long history of medic-inal use. Two companies foundedwithin the past Þve yearsÑXeno-va in Slough, England, and Phar-magenesis in Palo Alto, Calif.Ñare combing through the tradi-tional Chinese pharmacopoeia forpromising leads.

Another company, Shaman Phar-maceuticals, was founded by aformer venture capitalist, Lisa A.Conte, who wanted to combinedrug development with an eÝortto preserve rain-forest ßora. Con-teÕs Þrm, based in San Carlos,Calif., sends teams of botanistsand physicians to Latin Ameri-ca, Africa and Asia to search forplants that work against viruses,fungi or diabetes or that may be-come sedatives or analgesics.

Plants sent back to Shaman fortesting must have been used inthree geographically distinct ar-eas. Conte claims that of the 200that have passed this preliminaryprobe, more than half show activ-ity against a targeted malady, ascompared with less than 1 percentfor mass screening.

In only 16 months Shaman hasmoved an antiviral drug for child-hood ßu caused by the respirato-ry syncytial virus from rural heal-ers and urban botanicas in South

and Central America into FDA testing inhumans. By showing medical photo-graphs to healers in Peru and Ecuador,ShamanÕs botanists also turned up anantiviral agent that works against drug-resistant herpes infections. ÒWhat Sha-man is doing is using thousands of yearsof human clinical trials,Ó says William L.Current, a senior research scientist withEli LillyÕs infectious disease group. LastOctober Lilly made a $4-million equityinvestment in Shaman and also struckan agreement to collaborate with theÞrm for four years in developing drugsthat work against fungal diseases.

Gone are the days when a companycould create drugs from plants thatoriginate in developing countries with-out negotiating to pay royalties, as Lillydid during the 1960s with the rosy peri-winkle. Extracts from the periwinkle pro-duce vincristine and vinblastine, drugsused primarily against childhood leuke-mia and HodgkinÕs disease, respectively.Shaman has pledged to pass up endan-gered plants and is committed to fur-nishing royalties from drug revenues toboth the government and the nativecommunities from which a plant is har-vested, a policy that Lilly supports. Sha-man will make these payments through

COSTA RICAN FLORA attracts botanists whoconduct basic research and who hope to Þndplants that may yield new drugs.


ST

EV

EN

FE

RR

YM

atrix


The new administration has vowedto divert about $7 billion a yearfrom armaments research to civil-

ian technologies. To do so, it must con-front an urgent issue: What to do withthe national laboratories?

Now that the cold war has thawed,AmericaÕs premier weapons laborato-riesÑnamely, Los Alamos, Lawrence Liv-ermore and SandiaÑseem to have losttheir principal mission. Those three lab-oratories, which are part of the Depart-ment of Energy (DOE), command a col-lective annual budget in excess of $3.5billion and employ more than 23,000people. ÒWeÕve got to change,Ó declaresSiegfried S. Hecker, director of Los Ala-mos. ÒIf we donÕt have a legitimate mis-sion in the world, we should die.Ó

The plight of the weapons laborato-ries is spelled out most clearly in dol-lars and political pledges. Almost half ofthe labsÕ budgets are devoted to nucle-ar weapons development and mainten-ance. In many cases, the organizationshave parallel programs. Yet internation-al accords aim to quash the evolutionof new nuclear arms. Last year Con-gress called for a nine-month moratori-um on nuclear testing, whichbegan in October. After thatban expires, the U.S. will bepermitted 15 tests (to evalu-ate ÒsafetyÓ issues); an unlim-ited test ban is slated to gointo eÝect as of October 1996.

Few believe the laboratorieswill wrap up their work on nu-clear weapons altogether. Na-tional security will always re-main a primary objective, con-Þrms Sidney D. Drell, deputydirector of the Stanford LinearAccelerator Center and recent-ly appointed chairman of acouncil charged with articulat-ing a strategic role for the sci-ence and technology programsof Livermore and Los Alamos.But rather than emphasizingdesign and testing, the researchgroups will step up work ontechnologies for monitoringnuclear proliferation, ensuringthe safety of stockpiled weap-ons and disposing of obso-lete arms. ÒThere are still manythousands of nuclear weaponsaround,Ó Hecker adds. ÒYoudonÕt just throw them away.Ó

At the same time, as thebattle cry of Òinternational com-

petitivenessÓ echoes throughout U.S. in-dustry, Drell and others see the labs de-voting more resources to civilian tech-nologies. A recent report from the pri-vate-sector Council on Competitivenessrecommended that between 10 and 20percent of the budgets of the DOE (andother) labs should be devoted to tech-nology transfer. ÒThe federal labs havea new customerÑU.S. industryÑandneed to develop customer-driven tech-nology transfer programs to service itsneeds,Ó the council stated.

On the face of it, the laboratories seemwell equipped to lend a hand on thecompetitiveness front. Traditionally, theU.S. nuclear weapons labs have em-ployed some of AmericaÕs sharpest sci-entiÞc elite. These researchers have donegroundbreaking work across a range ofÞelds, including semiconductors andlasers, advanced materials, high-speedcomputation and networking. ÒYou cantick oÝ Þve or six areas where the labsare superb,Ó Hecker asserts, Òand thenyou just have to Þnd the right way oflinking them with industry.Ó

Finding those kinds of links hasproved to be a headache. Beginning withthe Stevenson-Wydler Technology Inno-vation Act of 1980, Congress tried tounlock the doors of federal labs for in-dustry. Nine years and several acts laterthe government gave the DOE labs an-other push toward the private sector

through the National CompetitivenessTechnology Transfer Act of 1989.

Some progress has been made at thefederal laboratories outside the DOE.The National Institute of Standards andTechnology, for instance, is barely onethird the size of any one of the weap-ons groups but has more than 240 cooperative research and developmentagreements (CRADAs) with industry.Even divisions of the Department of De-fense, such as the Army Research Lab-oratory in Fort Monmouth, N.J., havewon praise for securing deals with in-dustry. But as of November, the threeDOE labs had signed only 118 CRADAs.ÒThe DOE and the weapons labs havebeen very slow on CRADAs,Ó observesJulie Fox Gorte, a senior associate withthe OÛce of Technology Assessment.

Instead the three weapons groupsÑalong with the DOEÕs other six Òmulti-purposeÓ labsÑhave become entangledin a mesh of bureaucratic trip wires.ÒDo you want to hear about the legal orthe cultural ÔdemotivatorsÕ Þrst?Ó asksJohn R. DeMember, the federal labora-tory liaison for Digital Equipment Cor-poration, who participated in creating amodel CRADA between an associationof computer makers and the DOE.

DeMember insists he is optimisticabout the possibilities of working withthe laboratories. But he is blunt aboutthe stumbling blocks. The government

researchers have seldom hadto meet the strict time andcost budgets faced by theirprivate-sector colleagues, De-Member notes. Lab scientistshave traditionally shied awayfrom product-oriented workeven as company researchershave nurtured a Ònot inventedhereÓ bias. As a result, Òthe pro-cess of getting the lab research-ers and my researchers togeth-er has been painful,Ó DeMem-ber says.

Legal issues have been evenmore vexing. Rather than al-lowing laboratory directors tosign oÝ on CRADAs, all suchcontracts have required theDOEÕs approval. And there is acollection of regulations thatcan be CRADA Òshowstop-pers,Ó DeMember notes. Forinstance, the DOE has been reluctant to exempt from theFreedom of Information Actany information the labora-tories receive from industri-al partnersÑmeaning that thematerial is publicly availableeven if considered proprietaryby the company. ÒBecause theDOE labs are lawyer-driven,

CAMPAIGN PLEDGE: Bill Clinton vowed to help the weap-ons labs turn to civilian technologies such as the environ-mentally benign process for making printed-circuit boardsshowcased by Sandia National Laboratory head Al Narath.

National ConundrumsFinding new work for thenational weapons labs


MA

RK

PO

ULS

EN

San

dia

Nat

iona

lLab

orat

orie

s


In the laboratory where about 140Microsoft engineers are laboring ona project code-named Cairo, their

mascotÑa plywood cutout of a cam-elÑleans patiently against the corridorwall. Strapped to its side is a packageof Tylenol. ÒThe camel was sick for awhile,Ó quips James Allchin, MicrosoftÕsvice president of advanced systemsand manager of Cairo.

Along with a pack of other compa-nies, Microsoft is racing to create a verydiÝerent kind of software, an operatingsystem based on Òobject-oriented pro-gramming.Ó The camel is ready testimo-ny to the headaches and long hours ofthe eÝort. But the payoÝ promises to beenormous. ÒWe have to do somethingto get to where every person can con-struct his or her own program,Ó Allchinsays. ÒFor people to become more pro-ductive in the information age, therecan be no canned programs.Ó

At present, writing computer pro-grams is mostly a handicraft in whicheven mundane sections demand timeand painstaking work. Software manu-facturers such as Microsoft legitimate-ly worry about how to boost their ownproductivity. But making software iseven more vexing for those who simplywant to use it. At the top of that list arehardware manufacturers, who must in-creasingly lace their products with so-phisticated programs and users, whoÞnd themselves tailoring code to ac-complish speciÞc tasks.

The solution is obvious: design soft-ware so that sections of code can be re-used. Doing that, however, is fantasti-cally hard. Reuse has been on researchagendas since the 1960s, but only halt-ing progress has been made. Now Òthetime is right, the basic technology is inplace,Ó asserts Daniel Sabbah, who man-ages software technology at the IBMThomas J. Watson Research Center. ÒForthe Þrst time, we can look ahead andsee a fundamental paradigm shift com-ing.Ó Sheer momentum may help. ÒItÕshard to Þnd someone not doing reuse,Ósays Norman M. Delisle, director of thesoftware technology research laborato-ry at Tektronix in Beaverton, Ore. ÒItÕspart of survival in the Õ90s.Ó

Object-oriented programming, a con-cept currently as trendy as artiÞcial intelligence was in the 1980s, is a keytool being used to construct the bravenew world of reusable software. Ob-jects are chunks of data and relatedprocedures and methods (which de-

scribe some way of manipulating thedata). Like Lego blocks, such encapsulat-ed components can be plugged togetheror pulled apart to make diÝerent sys-tems, provided they all use the sameÒinterfaces,Ó or plugs. Objects boast afull complement of other featuresÑsome Þt into ÒclassesÓ with similar ob-jects; others ÒinheritÓ characteristics.Engineers hotly debate the merits of allsuch traits.

In an operating system such as Cairo,objects promise to give users a con-nect-the-dots ability to get a job done.For instance, a person might see oneÒbuttonÓ on a screen that represents adocumentÑa sales reportÑand anoth-er button that represents a mailing list,Allchin explains. To send out the re-port, an individual need only draw aline between the two buttons with amouse. But the user might also add acaveat, such as Òif the sales Þgures arefalling, send the report only to seniorexecutives.Ó The objects will check thenumbers and route the document ac-cordingly, Allchin says. What is more,the system will make use of objectswherever they are storedÑon the userÕsdesktop system or elsewhere in a net-work of computers.

Cairo is not scheduled to make a pub-lic debut until 1994. And engineers atMicrosoft and another dozen or so com-panies working on competing projectscontinue to wrestle with weighty prob-lems. The Òsemantics,Ó or the precisedescriptions of the nature of a speciÞcobject, are still as undeveloped as thespeech of a baby. As a result, the objectsdesigned by one company, say, Micro-soft, may not be able to link togetherwith objects from others, such as Bor-land International, NeXT Computer orTaligent (an IBM-Apple joint venture).

There are nonetheless a handful oftools in addition to objects that re-searchers hope will smooth the path toreusable software. IBMÕs Sabbah pointsto the usefulness of compilers (whichtranslate software from one language toanother) and to higher-level, more ab-stract, languages that let users expresstheir tasks in the language of their busi-ness rather than in computerese.

In one project, IBM researchers wrotea Òconstraint language,Ó which could optimize any set of equations to giv-en parameters. With that language, the workers then devised an experimental stock-option trading program that rec-ommended buy and sell options, given constraints such as a balanced portfolioof risks. Traders would need no specialprogramming skills to run the system; itcould, moreover, be applied with equalease to optimizing the constraints thatgovern electrical circuit design.

Soft LegoHow software designers hopeto make programs reusable

theyÕre more focused on what could gowrong than on taking chances,Ó he adds.

A December 1990 report from theGeneral Accounting OÛce reinforcedthose tendencies. The audit criticizedthe laboratories for excessive use ofdiscretionary funding. ÒThat report wasreally not fair !Ó Hecker declares. LosAlamos endured the brunt of the at-tacks. Hecker maintains that both theformer Atomic Energy Commission andthe DOE had long encouraged Los Alam-os to fund a wide range of scientiÞcprojects so that the weapons labs couldattract top scientiÞc talent. ÒWe have noexcuses, no apologies. The work fund-ed was Þrst-rate stuÝ,Ó Hecker insists.

Yet the media attention spurred theDOE to increase the rules and regulationsgoverning the laboratories. In addition,the contracts between the DOE and in-dependent Òmanagement and operat-ingÓ organizations that run the facili-ties have become far more detailed.

For instance, the University of Califor-nia has overseen Livermore and Los Ala-mos since they were created. When thecontracts came up for renewal in 1992,the documentation ballooned from agentlemanÕs agreement of a few tens ofpages to an extensively negotiated con-tract of several hundred. ÒIn the past,the standard the labs were measuredagainst was technical excellence,Ó saysRobert W. Kuckuck, who has recentlybeen appointed special assistant for lab-oratory administration in the OÛce ofthe President at the University of Cali-fornia. ÒToday itÕs very much about thedetails of how you get the job done.Ó

Experts nonetheless point to signsthat the laboratories will soon Þnd asmoother path to industrial partner-ships. Kuckuck, who worked as a physi-cist at Livermore for 29 years, is deter-mined to use his new oÛce to comb outthe snarl of requirements facing thelabs. DeMember feels some of the mostrecent CRADAs signed by the DOE re-solve many of the legal problems andwill serve as pivotal models for public-and private-sector cooperation. ÒIÕveseen an enormous change in how bothindustry and the labs feel about coop-eration,Ó DeMember adds. As a result,ÒIÕm very optimistic about future jointworkÑbut itÕs still a hard process.Ó

Many are looking to the Clinton ad-ministration to foster more collabora-tion with industry. ÒWe are living in aworld where brain power will deter-mine the lives of all our people,Ó Clin-ton told weapons workers during acampaign visit to Sandia in September.ÒI want to make sure that each of youhas a chance to contribute your full po-tential to the welfare and to the futureof your country.Ó ÑElizabeth Corcoran


Agood man is hard to Þnd. So, too,a good woman. Unlike the Ma-rine Corps, the National Library

of Medicine (NLM) does not need a fewgood specimens. Just one of each sexwill do. The ideal candidates should bebetween the ages of 20 and 60 years, ofmedium height, not too thin, not tooplump. Race or ethnic background isnot an issue. But a prospect must meetmore problematic criteria: he or shemust have expired and been put into a cooler within the last 12 hours inpreparation for being cut up by a teamof medical specialists into more than1,000 thin slices.

The NLM has embarked on an 18-month search for contestants to becomethe ideal male and female cadavers.Those chosen may become known asAdam and Eve. This search may markthe beginning of an era in which com-puters may rival cadavers as primarysources of information on anatomy.

By the end of 1993 the NLM wants toenter into a computer every millimeterof a male and a female cadaver via com-puted tomography (CT), magnetic res-onance imaging (MRI) and standardphotography. The result of this eÝortÑ

the Visible Human ProjectÑwill be themost comprehensive digital record ofan entire human body. ÒBefore this, oneuniversity was doing the right knee whileanother was doing the left ear,Ó says NLM

director Donald A. B. Lindberg.The main use for this computer-age

GrayÕs Anatomy will, of course, be thestudy of anatomy. For the Þrst time, amedical student will be able to ÒreversedissectÓ a cadaver, reassembling it af-ter having cut it apart on the screen. Aresearcher might also inject a mathe-matical model of abnormal cell growthinto a computerized liver to grow a tu-mor. The data base may be used in de-signing prostheses or in assisting a sur-geon to plan an operation better. A pa-tientÕs CT image can be compared withAdamÕs and EveÕs standard images fora ÒnormalÓ human.

Before proceeding, NLM oÛcials andoutside experts debated for years theexact deÞnition of normal but failed toachieve any consensus. Their loose clas-siÞcation now encompasses a male anda female of any race between the agesof 20 and 60 years, less than six feet inheight and of average weight. ÒWe decid-ed that nothing can be considered aver-age or normal, but we had to start some-where,Ó says Michael Ackerman, theNLMÕs associate director for specializedinformation services, which oversees theproject. But even within this open deÞ-nition, it has been diÛcult to Þnd a body

Also promising are Òframeworks,Ó orwhat David Garlan of Carnegie MellonUniversity calls Òsoftware architectures.ÓRather than plugging components to-gether, designers Þrst try to uncoverÑand then reuseÑthe skeleton shared bya collection of related products. Spread-sheets are a textbook example, Gar-lan points out. Even as users custom-ize spreadsheet programs to suit theirneeds, they are ÒreusingÓ a huge por-tion of the prewritten code. Similarly,product manufacturers will Þnd enor-mous opportunities to reuse code if theyÞrst identify a common software archi-tecture for a family of products, Garlansuggests.

Such frameworks inevitably comple-ment libraries of software components.At Hewlett Packard, for instance, Mar-tin Griss has been forming a Òßexiblesoftware factoryÓ that aims to take ad-vantage of both frameworks and com-ponent libraries. Even so, the biggestimpediments to reusing software aremore sociological than technical, henotes. ÒItÕs hard for a lab to change thebalance of up-front design and down-stream assembly,Ó Griss says. HewlettPackard is consequently trying severalpilot projects that rely on multidisci-plinary teams of engineers, designers,managers and even anthropologists.

More far-reaching ideas for reuseloom on the horizon. In many ways, allthe techniques for reuse turn on invent-ing ways to talk about the work at hand.Objects describe a piece of code and pro-cedures; frameworks describe commonthemes. At Microsoft, chief architectCharles Simonyi is thinking about soft-ware design from an even loftier plane.He wants to create a notation that engi-neers can add to code to describe whatthey are trying to do.

In this way, Simonyi hopes to capturea larger portion of the most creative eÝorts of designersÑso that others canÒreuseÓ the cleverness of the program-mers, not just the code. To use a cook-ing metaphor, SimonyiÕs approach rep-resents the diÝerence between provid-ing a chef with a packet of powderedsauce that needs only milk or a videoof Jacques Pepin preparing the sauce.ÒIf software artifacts are frozen humancontributions, we want to look at thefreezing process,Ó he says. ÒItÕs the hu-man thinking we want to reuse.Ó

Simonyi expects he must put in an-other few years of eÝort before he canprove his theories of Òintentional pro-grammingÓ can be applied in practice.But like earlier proposals for reusingsoftware, Allchin of Microsoft observes,SimonyiÕs ideas are high riskÑyet if suc-cessful, they also promise more thanadequate returns. ÑElizabeth Corcoran

Habeas CorpusSeeking subjects to bea digital Adam and Eve

HEAD RECONSTRUCTION was created by combining 352 digitized photographs oftissue sections. The University of Colorado will now do the same for a whole body.


Working backward from a manu-factured part to a blueprint de-scription used to be a labori-

ous process. Engineers would slice partsinto pieces and take hundreds of mea-surements. Even so, they would oftenonly come close to the precise shape anddimensions of the original. Now an in-dustrial-strength version of the comput-ed tomography (CT) machine that fer-rets out brain tumors is making the jobalmost as easy as taking a medical x-ray.

Data on the interior and external di-mensions of a manufactured part serveas a guide for constructing two- or three-dimensional computerized blueprints.There is no damage to the part nor dis-tortions in spatial perspective that oc-cur when a part is cut into pieces. TheCT scan provides information about thecontours of the part that may not havebeen on the manufacturerÕs original de-sign drawing or for which there mayhave only been a wood or plaster mod-el. The image also allows a drawing tobe made of a part that was originallyproduced by a supplierÑor a competitor.

At $1 million and higher, the price ofmany CT machines is far too steep tobuy one just to spy on the competition.Indeed, the three main U.S.-based mak-ers of industrial CT scanners sell few-er than 10 machines a year. The mainreason for installing a machine is tomeasure dimensions of a Þnished part,to detect internal ßaws or to constructcomputer-aided design (CAD) drawingsfor parts whose dimensional data arenot yet in the computer.

General MotorsÕ Power Train Divisionfacility in DeÞance, Ohio, for example,has been using a CT machine from Scien-tiÞc Measurement Systems(SMS), an Austin, Tex.,manufacturer, for sever-al years to create designdrawings of the metal cast-ings for engine blocks,connecting rods and oth-er parts.

Boeing turned to re-verse engineeringÑor Òge-ometry acquisition,Ó as itis sometimes euphemisti-cally calledÑto re-create atail wheel for a 1940s Boe-ing 307 Stratoliner, whichwas being refurbished forthe Smithsonian Institu-tion. Engineers took sev-eral two-dimensional CTslices of the wheel. ÒItÕs as

if you cut the part open and had it ly-ing in front of you,Ó says Richard H.Bossi, a research engineer for BoeingDefense & Space Group.

Next a design engineer extracted therelevant dimensional points from thecross section and used a CAD systemto form a three-dimensional solid im-age. This drawing was used to producea half-size replica of the wheel using a process called stereolithography. Anultraviolet laser, guided by the coordi-nates in the drawing, solidiÞed layersof plastic to make the model.

Yet peeking at the competition stillholds an undeniable allure. ÒGM couldlook at a Ford product and know allabout it, or vice versa,Ó asserts GeorgeA. Edwards, vice president of adminis-tration for SMS. Indeed, SMS has scanneda Toyota engine block for Ford Motor,though not for competitive reasons.

John C. Cooper, a senior engineer inFord Manufacturing Development, saysSMS imaged ToyotaÕs aluminum cylin-der heads because they resembled anew Ford design. The scan, Cooper says,enabled Ford engineers to determinethe capabilities of the CT technology.ÒThis was a typical example of the kindof part you might want to image,Ó Coop-er observes. ÒCT is the only thing Iknow of that can deÞne a partÕs inter-nal features without cutting it apart.Ó

Cooper says the company might con-sider the purchase of a CT machine ifthe outlook for the automobile indus-try improves. ÒWeÕd be looking at ourown parts or our competitorsÕ,Ó he says.

Still, CT is not yet a standard tool inthe annual ritual of checking out thecompetition. But as these techniquesare further automated and the price ofthe technology drops, reverse engineer-ing is likely to Þnd a variety of applica-tions. And the temptation to acquire alittle comparative geometry will certainlybe one of them. ÑGary Stix

Geometry AcquisitionComputed tomography is aboon to reverse engineering

that has remained unscathed throughthe travails of daily living. ÒPeople diefor a reason,Ó Ackerman says.

The University of Colorado HealthSciences Center in Denver received themore than $700,000 contract (a wordthat is used advisedly among the proj-ectÕs staÝ) to Þnd several male and fe-male bodies. An NLM committee willchoose one Adam and one Eve from thecandidates the center screens.

Cutting was to have begun last July.But so far only Þve cadavers have beenfound, three males and two females. Allbut two of them still have shortcomings.ÒWeÕve got two good males, but weÕrestill not really happy with the femaleswe have now,Ó says Victor M. Spitzer, aprofessor of cellular and structural bi-ology who is supervising the project atthe university. The digital preservationof one of the two frozen males could be-gin within a few months, after the NLM

review panel makes a choice. Spitzer isstill searching in other states and ex-plaining the purpose of the project attechnical conferences.

A good candidate may be one whohas died of a drug overdose, carbonmonoxide poisoning or some other non-disÞguring cause of mortality and, ofcourse, has agreed to donate his or herbody to science. ColoradoÕs AnatomicalVisualization Laboratory will employ amilling machine that operates as if itwere an automated carpenterÕs plane,shaving oÝ a millimeter-thick slice froma section of the body. As each slice isremoved, the newly exposed surface is photographed. The digitized photo-graphs will be stored in a computerwith MRI and CT images that are madebefore the cutting process is begun. Af-terward, the real work will begin. An-other contractor will Þnish the job byidentifying and indexing each of thedigital slices, a project that will takeyears more and will cost an estimated$10 million.

The Þnal data base may consist of 50gigabytes of digital information, enough,if left uncompressed, to Þll more than80 data-storing compact discs. Insteadof distributing data on discs, the NLM

may have the option to transfer the in-formation over a proposed government-Þnanced network that could accommo-date transmission at the lightning speedof a gigabit per second, a project pro-posed by Vice PresidentÐelect Al Gore.

At some future date, Adam and Evemay have progeny, perhaps parents aswell. The NLM would like to add imagedcadavers of children, the elderly and thesick to its storehouse. Meanwhile thesearch continues for a select few menand women who will be the recipientsof 50 gigabytes of fame. ÑGary Stix

TAIL WHEEL for a vintage Boeing 307 Stratoliner wasre-created using computed tomography to help deter-mine the dimensions for a design drawing.


he legions of any freshman ad-ministration march into Wash-ington ßush with armfuls of eco-

nomic solutions. The previous genera-tion had supply-side economics; BillClintonÕs team has infrastructure.

Infrastructure oÝers an intriguingpolicy platform. During the past decade,newspaper headlines have decried thecollapse of bridges, the growth of pot-holes and the deterioration of publicschools. Robert B. Reich, a muse for Pres-ident-elect Clinton, lofted high the ban-ner of infrastructure in his 1991 book,The Work of Nations. Reich writes, ÒHere-in lies the new logic of economic na-tionalism: The skills of a nationÕs workforce and the quality of its infrastructureare what makes it unique, and uniquelyattractive, in the world economy.Ó

No economist would deny that infra-structure investments, like family val-ues, are important. But beyond that safeassertion rages one of the Þercest de-bates among economists today. At is-sue: Do investments in public infra-structure earn more, less or about thesame return as do private investments?The answer suggests clearly diÝerentpolicies. Although economists argue thatthey lack enough evidence to clinch aconsensus, politicians have been lessshy. As a result, infrastructure has be-come Òone of the more emotional is-sues economists get involved in,Ó saysRandall W. Eberts, an economist at theFederal Reserve Bank of Cleveland.

The Clinton clique has much in mindwhen they talk about infrastructure:not just improved roads but telecom-munications and high-speed data net-works and better schools. Yet as recent-ly as a few years ago infrastructure wasseldom mentioned in economic circles.

Economists might crunch throughcost-and-beneÞt analyses of a construc-tion project. But no one tried calculat-ing the value of such investments forthe economy at large. Even the classicÒproduction functionÓ describes growthas a function of changes in the amountof capital (contributed by industry) andlabor used, with an added kick fromnew technology. Better roads and sew-ers were believed to contribute broadlyto the Òquality of life.Ó

In 1989 David Alan Aschauer, whowas then with the Federal Reserve Bank

of Chicago, opened up a diÝerent per-spective by arguing that infrastruc-ture strongly bolstered private indus-try. He observed that U.S. productivi-ty had declined in tandem with fallinginvestments in what he called Òcoreinfrastructure,Ó namely, transportation(roads, mass transit and airports), waterand sewer systems and other publiclyowned facilities. His studies suggestedas much as 40 percent of the slowdownin productivity could be explained bydeclining public capital investments.Aschauer further estimated that everydollar invested in public infrastructureyielded four dollars in return.

The work stunned economists. ÒAsch-auer did a great serviceÓ by identify-ing a role for infrastructure, declaresAlicia H. Munnell, a senior vice presi-dent at the Federal Reserve Bank of Bos-ton. ÒWe should be embarrassed thatwe overlooked it.Ó That said, many econ-

omists quarrel with the speciÞc returns.ÒThereÕs been a lot of controversy aboutthose results,Ó Aschauer concedes. ÒPeo-ple say they are implausible.Ó

Munnell has done several studies thatsupport the proposition that infrastruc-ture spending yields some kind of posi-tive return. ÒBut infrastructure isnÕt builtto enhance the private-sector output,Óshe notes. Logically, since the privatesector purposefully invests to boost itsoutput, its capital spending should earnat least as high a payoÝ as public infra-structure, she suggests.

Aschauer, now a professor at BatesCollege, nonetheless believes the econ-omy reaps signiÞcant ÒnetworkÓ eÝectsfrom investing in infrastructure. A bet-ter bridge, for example, improves thebusiness of private producers as diverseas computer makers, farmers and truck-ing ßeets. By complementing privateinvestment, such public infrastructureraises proÞts and stimulates private in-vestment, Aschauer asserts.

Others respond that no economic the-

ory explains why infrastructure aloneshould stimulate growth. ÒItÕs the Field

of Dreams syndrome,Ó Eberts says. ÒIf webuild it, consumers will come.Ó

Statistical evidence is also muddled.In studies of regions and cities, Mun-nell has found positive correlations be-tween public infrastructure and growth,but she hesitates to say such public in-vestments contribute more than pri-vate ones might. Eberts believes he hasfound indications that during most ofthe 1900s, the mature cities of theNorth beneÞted more from public in-vestments than did the newer, faster-growing urban areas of the South. ÒButthe problem is thatÕs all historical evi-dence,Ó he complains. Showing that anew highway boosted growth in thepast does not mean that building an-other road will have the same impacttomorrow.

Those who argue that more publicconstruction will promote growth in-terpret the increased congestion on theroads and in the sky as a sign that de-mand has outstripped the supply ofthose facilities. Transportation econo-mists, however, counter that such jamsshow that the fees charged for usingthese resources are just too low.

ÒI think infrastructure has what somewould call nonlinear eÝects,Ó suggestsCharles R. Hulten, an economist at theUniversity of Maryland. Sometimes newinfrastructure is critical for growth, suchas before the development of the high-way system; other times, adding anoth-er road might just Þll space.

Aschauer and others applaud Clin-tonÕs emphasis on information tech-nologies in the mix of infrastructurecomponents needed for the 21st centu-ry. ÒDollar for dollar, I think the pro-ductivity beneÞts we can get from tele-communications are signiÞcantÑhigherthan what we would receive from addi-tional highways,Ó he says. Even so, datathat might illustrate the value of suchhigh-tech infrastructureÑfunded by ei-ther the private or public sectorÑare asscarce as blue solid-state lasers.

Not that Aschauer plans to take muchof a hand in building new policies inWashington. Although he helped sparkthe DemocratsÕ infatuation with infra-structure, Aschauer is a greater fan ofthe policies of Alexander Hamilton thanof Bill Clinton. ÒIÕm pro-infrastructure,Óhe says, Òbut IÕm not enamored withmore spending.Ó ÑElizabeth Corcoran

The Return on Infrastructure

THE ANALYTICAL ECONOMIST

All agree infrastructure isimportant. But how much do governmentinvestments really pay oÝ?


ystematic inventories of plots ofwoodlands and Þelds can be ofpractical use in planning how best

to conserve wildlife in a given patch ofland. These surveys show vividly howthe number of species encountered in aplot varies with the amount of land in-spected. They also help to provide aquantitative way to see how human ac-tivity aÝects local biological diversity.With such observations, conservation-ists, ecological planners and policymak-

ers can estimate the smallest amountof land needed to preserve a percent-age of the natural ßora and fauna. Par-ticularly useful in this regard is the re-lation between the diversity of wood-land creatures and plants and the sizeof forest ÒislandsÓ in an urban or sur-burban Òsea.Ó Such relations are techni-cally referred to as species-area curves.

Counting plant species within a lawnis an instructive analogue of such quan-titative methods. (Tabulating things thatcrawl or ßy is diÛcult and tends to liebeyond the amateur level.) I initially de-signed this project as an exercise for a summer course in mathematical geol-ogy and Þeld biology for grade schoolteachers. The teachers have adapted itto be an exercise in exploration andclassiÞcation for children in the lowergrades. But even our preliminary anal-yses have been so informative that Iplan to use the exercise in introducto-ry data analysis and extrapolation forgraduate students. The project can bedone at any level of complexity, fromchildlike exploration to professionalanalysis. Although each level poses its

own important questions about conser-vation, the basic issue that remains ishow much land is needed to sustainspecies diversity.

The teachers and I selected a lawnbehind a parking lot on the Prince-ton University campus. We worked inthree teams of four people. One teamstarted by staking string boundaries onthe lawn in nested blocks. The blocksranged in size from a meter square upto 16 by 16 meters. We set the bound-aries for the largest area Þrst. Becausethe ground bulged slightly (it made the sum of the four angles greater than360 degrees), we fudged the plot into a square by making the diagonals equalin length. We divided this large squareinto four equal areas and then furthersubdivided one corner until the lastblocks were one meter square [see top il-

lustration on opposite page]. A tape mea-sure and 3Ð4Ð5 right triangles came inhandy.

Biodiversity in the Backyard


THE AMATEUR SCIENTIST conducted by Henry S. Horn

TURKEY-TRACKGRASS

SCOTT’STOOTHGRASS

BROAD-LEAVED PLANTAIN

FANCY MARYSPEEDWELL

BERT WEED

DANDELION

ITTY-BITTY TWIN BETTY MYSTERY PLANT

PETITELISA

PRINCETON PARSLEY

MOUSE-EAR CHICKWEED

LINDABERRY

FAT-LEAVEDGRASS

HENRY S. HORN constructs satirical artwork and sings blues, madrigals andchoral music. He is also professor of ecol-ogy and evolutionary biology and directorof the program in environmental studiesat Princeton University. He thanks Bris-tol-Myers Squibb for sponsorship, Eliza-beth Horn for help in designing and di-recting this project, the elementary schoolteachers for their participation and, lastbut not least, Robert May, William Boni-ni, Sheldon Judson, Patricia Matuszew-ski and Amy Wolman for constructivecommentary and moral support.

LEAF SHAPES provided the basis for dis-tinguishing plant species. The Òdiscover-erÓ named the plant, an honor that led toseveral idiosyncratic appellations. Only afew of the 34 species found are shown.


Of course, the area may be increased,or the smallest squares subdivided, de-pending on the number of species thatappear during the investigation. A roughcriterion for the right-size area is that amiddle-aged and mildly myopic biolo-gist can walk across it and count about12 obviously diÝerent species. Such anarea will yield about 30 to 40 specieson closer examination.

For familiar plant species, we usedthe common name. A professional ver-sion of the activity would use a tech-nical key to the ßora. For our quanti-tative pattern and just for fun, wedeÞned our own ÒspeciesÓ by diÝer-ences in the leaves. In eÝect, we wereimitating the process by which the truespeciesÕ names came about [see ÒHowMany Species Inhabit the Earth?Ó byRobert M. May; SCIENTIFIC AMERICAN,October 1992].

We set up a ÒmuseumÓ of paper onwhich a ÒcuratorÓ wrote the name ofeach species found and taped a speci-men next to it. While one crew set upthe sampling boundaries, the other twoexplored the region for new species.Any specimen that showed novel fea-tures was taken back to the museum.The investigators compared the speci-men with named species and assessedits novelty in consultation with the cu-rator. If the specimen was truly new, itwas added to the collection. The dis-coverer had the honor of naming it.Without thinking about it, we namedspecies just as professional taxono-mists doÑas often for oneself or for afriend as for deÞning characteristics ofthe specimen, the habitat or relatedplants. Being amateurs, we could aÝordto be whimsicalÑhence, names such asÒHairy HarryÓ and ÒItty-Bitty.Ó

After completing the survey, we add-ed the totals in each block. We also accumulated a running count of thenumbers, starting with those in the smallest square and then adding those in subsequent blocks until we had in-cluded the entire plot. (A sample tallysheet appears at the right.) Even with-out technical analysis, the results pro-voke many interesting observations.Some species are common to nearly ev-ery block; others are rare. Some appearas lone or scattered individuals. Othersare found in clumps of several individ-uals, although the clumps themselvesare unique or scattered. Is there anypattern to which species are commonand widespread, which are clumped,and which are rare and scattered?

To explore for patterns, we plot-ted the number of species against thearea surveyed in several ways. First, wegraphed the cumulative numbers ofspecies for each surveyed square, start-


SAMPLING PLOTS consisted of 13 subdivisions of a 256-square-meter area. Theplots were numbered in a clockwise spiral pattern.

TALLY SHEET kept track of the species found. The red check marks denote thesmallest block number in which the species was encountered. The cumulative totalis the running count of the red checks. The areas are in square meters.

16 METERS

2 METERS

4 METERS

8 METERS

11

21

4 35

67

8

910

1213

1 METER

TURKEYTRACK GRASSSCOTT’S TOOTH GRASSLITTLE BROAD-LEAVED GRASSFAT-LEAVED GRASSWHITE CLOVERALSIKE CLOVERHOP CLOVERYELLOW WOOD SORRELMOUSE-EAR CHICKWEEDJAMES’S 3-LEAFSMOOTH-LEAVED BARBARAHAIRY HARRYBROAD-LEAVED PLANTAINFANCY MARYDANDELIONITTY-BITTYWHITE ASH TREESPEEDWELLBERT WEEDINDIAN STRAWBERRYBROWN-TOP MUSHROOMPETITE LISABOSS MOSSFIELD SPEEDWELLFUZZY CHICKWEEDSCARLET PIMPERNELTWIN BETTYNARROW-LEAVED PLANTAINMYSTERY PLANTNOVA TERRA SHARONPRINCETON PARSLEYPOISON IVYLINDA BERRYERNIE WEEDTOTAL PER BLOCKCUMULATIVE TOTALAREA OF BLOCKCUMULATIVE AREA 25619212864

164832161284321

64 64 641616444111127 32 33 3426262523231917161415 19 13 22121118111913111314

13121110987654321BLOCK NUMBER

SPECIES


ing with the most subdivided corner.This cumulative curve shows that 75percent of our species are found in ar-eas as small as 20 square meters [see il-lustration above].

To test the quantitative pattern wefound against the traditional species-area equation [see box below ], we plot-ted the same data on logarithmic axes.Some of our grade school teachers werewary of logarithms, but the samplingsquares are already scaled multiplica-tively by a factor of two in length, or afactor of four in area. A logarithmicscale is easy to construct for the num-ber of species by marking Þxed inter-vals on linear graph paper with 1, 2, 4,8, 16 and so on. The bilogarithmic plotof our data is a straight line, which con-forms to the theoretical generalizationgiven by the species-area equation.

On the same graph, we plotted thesurveys for each individual block. We

expected the plots to show the samepattern as the cumulative data did, perhaps with a bit of variation and aslightly lower slope and species-inter-cept point. That is because the cumula-tive curve must rise continuously withincreasing area. We discovered to ourdismay that the pattern of the individu-al blocks was somewhat inconsistent.

Discussion suggested possible caus-es. One group admitted to being lessthan thorough in their surveys. Theywere more interested in the morpholo-gy of what they found than in the num-bers. Several admitted to accumulatingfatigue during the second hour of crawl-ing around the larger plots. It is possiblethat lapses by one group or by a few in-dividuals were compensated by othersin the cumulative data, hence explain-ing the consistency of those data. It is also possible, however, that we un-derestimated the slope of the species-

area curve for our lawn. In any case, theteachers were so impressed with theregularity of the cumulative data thatthey started an animated conversationabout how to conduct more careful sur-veys the next time.

The discussion led to further ques-tions. Can our results be safely extrap-olated to areas larger than those sam-pled? How much area would be neededto preserve 50 percent, or even 90 per-cent, of the regional lawn species? Howwould the diversity of plants in real Òis-landsÓ of lawn in a paved parking lotdiÝer from marked-oÝ samples of thesame size in a continuous lawn? Whatinsights does this analysis give into theplanning of urban parks?

This exercise is just a conceptualmetaphor for some far more practicaluses of species-area curves. It is, howev-er, a large empirical step toward makingyour own surveys of trees, shrubs, vines,wildßowers, ferns, mushrooms or veg-etables in patches of various sizes. Thenplot the number of species against area,think about the results and take yourdata to the next meeting of your localplanning board.


FURTHER READING

THE FRAGMENTED FOREST: ISLAND BIO-GEOGRAPHY THEORY AND THE PRESERVA-TION OF BIOTIC DIVERSITY. Larry D. Har-ris. University of Chicago Press, 1984.

WEEDS. Alexander C. Martin. Western Pub-lishing Company, 1987.

ECOLOGICAL DIVERSITY AND ITS MEASURE-MENT. Anne E. Magurran. Princeton Uni-versity Press, 1988.

NATURE RESERVES: ISLAND THEORY ANDCONSERVATION PRACTICE. Craig L. Sha-fer. Smithsonian Institution Press, 1990.

THE DIVERSITY OF LIFE. Edward O. Wilson.Belknap Press/Harvard University Press,1992.

AREA (SQUARE METERS)

NU

MB

ER

OF

SP

EC

IES

40

30

20

10

00 100 200 300

100

1 10 100 1,00010

AREA (SQUARE METERS)

NU

MB

ER

OF

SP

EC

IES

CUMULATIVE NUMBER OF SPECIESNUMBER OF SPECIES PER BLOCK

a b

SPECIES-AREA CURVE shows that the cumulative number ofplant species encountered increases with the area surveyed(a). A logarithmic graph of the data reveals a straight line (b),

which provides the constants c and z [see box below]. The plotsof the number of species per block, however, were inconsistent.Experimenter fatigue is a possible reason for the inaccuracy.

Deriving the Species-Area Curve

or many groups of organisms, thenumber of species encountered

increases as the area increases. Asuitable relation can be expressed as

S = c Az

where S is the total number of spe-cies observed in a surveyed area, Ais the area surveyed and c and z areconstants fitted to the data.

Taking the logarithm of both sidesgives

log S = log c + z log A.

This equation is an empirical gen-eralization. Many researchers are cur-

rently trying to pose theories that“predict” it. The reality of this equa-tion can be tested, for a given regionand group of organisms, by plot-ting surveys on logarithmic scales of both species and area to see if they conform to the generalization of astraight line. If they do, then the re-lation can be characterized by onlytwo fitted parameters, c and z.

As appropriate as this equationmay be, the species-area curve is of-ten more rhetorically convincing asan argument for conservation if thenumber of species and area are plot-ted linearly. Then it is clear that ef-forts to find as many species as pos-sible have diminishing returns.

F


Heavenly Objects

SKY CALENDAR, by the Abrams Plan-etarium (Michigan State University,East Lansing, MI 48824) ($6 per year,12 monthly black-and-white sheets).EPCOT SKY CALENDAR 1993, by theAbrams Planetarium and Stephan VanDam. Hyperion, 1992 (114 Fifth Ave.,New York, NY 10011) ($12.95, wall cal-endar in color). WANDERERS IN SPACE:EXPLORATION AND DISCOVERY IN THE

SOLAR SYSTEM, by Kenneth R. Lang andCharles A. Whitney. Cambridge Univer-sity Press, 1991 (paperbound, $24.95).IS ANYONE OUT THERE? THE SCIENTIF-IC SEARCH FOR EXTRATERRESTRIAL IN-TELLIGENCE, by Frank Drake and DavaSobel. Delacorte Press, 1992 ($22).

The celestial action we all perceivewith unaided eye is at its liveliestaround sunset and sunrise, and

it plays close to those decisive direc-tions as they march seasonally alongyour horizon. Watch there every dayyou can, and follow the moon and real-

ly bright points up into the vault of thenight sky, too.

The shepherd tending his ßocks need-ed no schedule to draw his attention,but we who dwell indoors do. For someyears now, the astronomer-artists atthe Abrams Planetarium have made an extremely useful aide memoire, sim-ple enough for tyros yet helpful to theexperienced, including those who use binoculars or small telescopes. It hasbeen noticed here on more than oneJanuary past. Their calendar oÝers abox for every day of the year, a monthof boxes placed on a single page. Eachbox holds a small scaled drawing ofjust what sights to look for and when,clear weather assumed at the horizon(meant for viewers in middle northernlatitudes). On the reverse of each pagea neatly simplifed star map for themonth urges your gaze up from thehorizon and into the night hours. Textin the margins oÝers more data andmuch context. All you need to know isthe time of day and the directions ofeast and west.

Now the measured and beautifuldance unfolds with the year, linkinghappy watchers in a human traditiontoo old to estimate. A few 1993 eventsmay draw you in: near dawn on 15 Jan-uary, three luminaries will stand in ashort straight line that aims right downat the still hidden sun; they are last-quarter moon, bright star Spica andbrilliant Jupiter. If you miss that Þnearray, a quite similar line will formagain before dawn on 21 November,with Mercury and Venus replacing themoon. Venus and bright star Reguluspose nearby at dawn during a coupleof weeks in September, becoming onthe 21st a mask of two unequal catÕseyes gleaming only a ÞngerÕs breadthapart. A total eclipse of the moon willtake place near midnight between 28and 29 November, mid-totality comingeverywhere about 1:26 A.M. Easterntime. The Pleiades and the Hyades clus-ters will provide a jeweled backdrop tothe darkened and reddened moon.

This year a very attractive revision ofthe Abrams Sky Calendar has appearedunder the aegis of DisneyÕs EPCOT Cen-ter at Orlando. The original frugal oÝ-set pages, each evoking a single sheetstraight out of the copier, are still avail-able. But the little day boxes have meta-morphosed into striking color throughthe work of a well-known New York car-tographic designer. The material hasbeen turned into a stapled calendar,each day box doubled in size and drawnagainst a colorful sky whose blues codefor dawn, dusk or deep night. The restof each monthÕs double spread showsthe monthÕs star maps in color or largeimages from NASA, among them mapsof the near and far sides of the moon,JupiterÕs Great Red Spot orÑmost won-derfulÑa big showy Voyager close-upof Saturn, the rings curving past theirown black shadow across the brightgolden globe. That big calendar is justright on a wall that many passersbymay consult; the loose pages provide asbefore an easy portability worthwhilefor travelers and oÝer somewhat moreÞne-print detail, much simpler starmaps for beginners and lower price.

Wanderers is a scrupulous yet imagi-native introduction to what we knowabout our solar system and how wegained the knowledge. The two well-known astronomer-authors have avoid-ed equations entirelyÑnot one equalsign is to be foundÑbut they have

BOOK REVIEWS by Philip Morrison


WINDS ON SATURN form counterßowing eastward and westward patterns in the po-lar regions, as shown in this color-enhanced illustration from Wanderers in Space.


packed their book with data and graphsthat summarize well what the probeshave found. A number of works of art,mostly from our century (plus the Bot-ticelli portrait of foam-born Aphrodite),help to extend rich factual images intoapt metaphor. Particularly memorableis the circular poem-image by AnnieDillard mapping the wind regimes ofthe earth. The story of the moonÕs ori-gin is given the latest and most plausi-ble explanation: a glancing impact froma Mars-sized collider knocked a ring ofmatter out of both intruder and form-ing earth; that ring soon condensedinto our outsized low-density moon.The book is at once an introductorytext of stature and a thorough, seriousand readable report for general read-ers, with much compact reference data.Chemical issues are less fully treatedthan one might wish, a limitation notuncommon among astronomers.

A few examples of argument and ex-hibit: an Ansel Adams photograph al-lows comparison between the dark mir-ror of the moon and the brighter gran-ite of Half Dome, each lit by the samesun at the same distance; a photographof auroral glow from space clariÞeswhy dwellers in a band along the Cana-dian border west of the Great Lakeshave the best auroral view anywhereoutside of the polar regions; a chapteron asteroids, meteors and meteoritestraces their lineage, even to a picture oficy Antarctic hills that hold a rich lodeof the little fallen asteroid debris; Voy-

ager 2 discloses strange moons andstranger multiple rings around the plan-ets beyond Saturn. So the chaptersmarch in good order outward past everyplanet, to close with the trillion cometsof the remote solar fringe. The lastchapter orders us more in time than inspace, examining the dimly lit birth ofour solar system.

Is our system unique? It remains pos-sible that in all the cosmos it is we alonewho inhabit some wanderer of circum-stellar space, although Òmany astrono-mers Þnd it hard to believe.Ó Radio as-tronomer and senior co-author of thefourth title above, Frank Drake hasfound it impossible to hold that partic-ular belief ever since he was a talentedÞrst-year graduate student, when hisdiscipline was growing up in America.He was the Þrst person to muse eÝec-tively on how radio might open the wayto purposeful signals between the stars.One night, all alone at the Harvarddish, he heard a Òstrikingly regular sig-nalÓ above the stellar noise source thatwas his thesis topic. Its channel fre-quency was even oÝset to match theexpected motion of the star cluster hewas recording. The detection came to


the eager student as a stunning mira-cle, one manifested only to him in allthe world; it did not surprise him whensoon his hair started to turn white! (Heconcedes that there is a family tenden-cy to early graying.) But he investigatedhis miracle; when he pointed the dishaway from the cluster, the signal contin-ued loud and clear. Some quite earthlytransmitter, probably a military sourcenot bound by usual rules, had chosen a radio channel supposed to be heldopen for astronomy.

Within a few years the new Ph.D. wasresponsible for observations at the bignew dish of the National Radio Astron-omy Observatory at Green Bank, W.Va.Winning the consent of his famous andstarchy Director, he quietly made readythe receiving equipment at hand to lis-ten to a few nearby sunlike stars. Be-fore his Project Ozma could begin in1959, its rationale and possibilities be-came public knowledge. His own headyidea was in the air, but published Þrstby a pair of Cornell physicists, outsid-ers to radio astronomy who had neverheard of Drake or of Ozma (one ofthem was the present reviewer).

The entire book is as warm, candidand intimate as that anecdote. Drake isnow fully established, a prematurelywhite-haired, fatherly, innovative lead-er of research, sometimes an observato-ry director himself. DrakeÕs Ozma wasfollowed by three decades of debateand pioneering receiver searches by in-dividual astronomers in many lands. To-gether they act rather like one personwho searches for a needle in a vast hay-stack by sorting a handful of hay eachday as she passes by. But now an expertNASA team, starting with a share of timeon two top American radio dishes, is lis-tening systematically, greatly empow-ered by a modest investment in equip-ment that has applied the astonishingpower of modern computing to listento a multitude of channels all at once.

In this book, both names and ideasmake news. The reader will encounterfascinating little scenes of turbulenceand of concord, among such people asCarl Sagan, Timothy Leary (then in pris-on), Senator William Proxmire, FatherTheodore Hesburgh and the late Dr. Io-sif Shklovsky. An international cast ofother men and women, from Karl F.Gauss and Nikola Tesla to contempo-rary radio astronomers and electronicswizards, has been entangled somehowor other in the same inquiry: Is anyoneout there?

The great dish at Arecibo has justspent a couple of months in the initialsearch period with NASAÕs new gear.The grand Þnd they dream of is Ònot asimple one to make. If there were once


cockeyed optimists. . . there arenÕt anynow.. . . We should be willing to sweatand crawl and wait.Ó Yet we wait inhope; the only judgment that counts isthe verdict after meticulous search.

Earthly Creatures

WALKERÕS MAMMALS OF THE WORLD,by Ronald M. Nowak. Fifth edition intwo volumes. Johns Hopkins UniversityPress, 1991 ($89.95).

Any mammalian browser will enjoythe diversity of one class of ver-tebrates laid across the 1,600

pages of this succinct, authoritativeand most readable natural history. Itlists genus after genus until all 1,100recent genera have been described atleast brießy. The particulars emphasizethe nature of the living animal in itsrange and any relation it may have toour own abundant, capable, destructiveyet hopeful species of the genus Homo.The length of discussion varies with in-terest and knowledge; the small catsearn an indulgent 20 pages with many a Þne slinky photograph; the cheetah, asingle famous genus and species, hasone page and one photograph. Ninehundred genera are illustrated by clearblack-and-white photographs of a liv-ing specimen of some typical species.

The originator of this much admiredreference, Ernest P. Walker of the Na-tional Zoo, took 30 years to preparethe Þrst edition of his book, written tointerest and instruct the general read-er and serve at the same time as a Þrst resort for knowing professionals. Hebrought it out in 1964; this edition, likethe fourth edition in 1983, was updat-ed and improved by his able continua-tor, a Virginia mammalogist, very muchin WalkerÕs spirit. The apparatus of in-dexes and tables is easy to use, basedon both the scientiÞc and vernacularnames. There are nearly 6,000 citationsof the literature.

What can you Þnd in this treasury?Look at random; soon a portrait stopsthe eye. This little creature is a narrow-footed marsupial ÒmouseÓ of north-western Australia, with a tail that arcsupward like a rainbow, twice the lengthof its mousy body. Look more pur-posefully; you will Þnd well treated allthe seven species of rats commensalwith humans, especially the black ratfrom Malaysia so abundant in the trop-ics and the temperate-zone ÒNorwayÓrat from north China. Try the largest of land carnivores, the brown bear wecall the grizzly. It has a wide naturalrange, from the Alaskan islands all theway around the hemisphere to Eastern


Siberia, even including the mountainsof North Africa.

No form is stranger than the pert little dwarf mongoose of East Africa.They dwell in harmonious groups of afew dozen, living among crevices andtree roots, seeking insects and smallfruits by day. The groups are matriar-chal families, ruled by a dominant fe-male, the queen, with her consort. Onlythat pair produce oÝspring, and theysuppress sexual activity among theworking class, who form a hierarchy.The youngest members rank highest,mainly free to forage, though some-times posted as group guards. Othersnurse, tend and baby-sit the helplessyoung during a very long period of immaturity. When the queen dies, thegroup may split up. The analogy to so-cial insects is plain; this mongoose ge-nus closely resembles in behavior astrange social rodent of the region, theburrowing naked mole rats.

This is a book many libraries andmany a reading family should keep onopen shelves to help the puzzled andto tempt the curious.

Step by Step

THE STAIRCASE, by John Templer, Vol.I: HISTORY AND THEORIES, Vol. II: STUD-IES OF HAZARDS, FALLS, AND SAFER DE-SIGN. MIT Press, 1992 ($55 for the set;Vol. I, $27.50; Vol. II, $32.50).

The staircase is as treacherous asit is beautiful, at once Òone of themost dangerous manufactured

objectsÓ and an art form with a glori-ous history. These two slim, beautifullymade and illuminating volumes dis-close these two sides of the topic. TheGeorgia Tech author, architect and en-gineer, who sees himself within the clas-sical tradition, is persuasive that realunderstanding can come only by prob-ing an issue from many directions.

Among us bipeds, steps are Neolithicin age or even older. They began asnotched tree trunks leaning against awall, a forked portion at the top to pre-vent rolling, and as ladders. Ladderscan be designed for easy removal, akind of moat against unwanted en-trants: see Taos Pueblo. Typical laddersare steep, like the shipÕs companion-way, so steep that you must face theladder to go down it. At last, the oldestproper stair arrives, a simple, straightßight of steps, wide enough tread andlow enough gradient to allow the userto face forward when going either upor down.

Up the designers climb, too, theirstaircases increasing in scale as they

change in intent and surroundings,from old Delphi past Michelangelo, onto Le Corbusier and the Pompidou Cen-ter. Straight stairs may be walled in,open on one side, or open on both;some stairs patently invite and oth-ers daunt. Helical stairs, known earlier,became common only after the fall ofRome. Such structures save costly ßoorspace and are defensible against forcibleentry, but they require stonemasons oftrue skill.

The next evolutionary step is combi-nation of the modules. Straight ßightsjoin ßat landings. Dogleg stairs turn180 degrees at a landing; two mirroredarcs of steps ßow together at a balco-ny, often behind columns. Monk, Doge,Pope and King are given even granderstairs. These presage the baroque Òstair-cases of honor,Ó major integrating fea-tures of the whole ediÞce, built bothout-of-doors and within. A century andmore ago these left the aristocratic hallsto engage pleasure-seeking crowds ofburghers, until that theatrical staircasein the Paris Opera would give access to all public parts of the building, cast-ing throngs of well-dressed spectatorsthemselves as part of the spectacle. Itsproud architect declared, ÒThe Opera isthe staircase.Ó Romantic Victor Hugothought the place Òcomparable to NotreDame,Ó but musical Claude Debussyheld it rather like a Turkish bath.

In our age, elevators and escalators,and even ramps, begin to overshadowstaircases, although public stairs remainvisible, serving their old functions withnewer designs; often even bare-bonedÞre stairs are boldly exposed.

The second volume looks squarelyinto the cruel face of stairs, to analyzethe injurious accidents that take placethere by the millions every year. ÒMuchof the suÝering is unnecessary.Ó Likecars, soap and medicines, stairs are arti-facts whose use entails some risk. Oneremedy is too little used: adequate re-search, both for accident prevention andfor the reduction of injuries. It is asthough only Òa dozen small studiesÓ ofautomobile safety had been made since1900. ÒThe automobile would still bethe perilous machine it was at the turnof this century. The stair still is.Ó

Epidemiology opens the study. Theepidemic of falls is worldwide. Expect a misstep once in 2,000 stair uses, onethat can be corrected before a fall, oneminor accident among 30 such mis-steps, one hospitalization for each 50minor injuries, and one fatality amongsomewhat fewer than 200 hospitaliza-tions. The elderly are at particular riskfrom this plague. (This reviewer, elder-ly and somewhat inÞrm, will testify tolong fear and avoidance of risky stairs.)


CORRESPONDENCE

Reprints are available; to order, writeReprint Department, SCIENTIFIC AMERI-CAN, 415 Madison Avenue, New York,N.Y. 10017-1111 or fax inquiries to(212) 355-0408.

Back issues: $6 each ($7 outside U.S.)prepaid. Most numbers available. Cred-it card (Mastercard/Visa) orders fortwo or more issues accepted. To order,fax (212) 355-0408.

Index of articles since 1948 availablein electronic format. Write SciDex™, SCI-ENTIFIC AMERICAN, 415 Madison Ave-nue, New York, N.Y. 10017-1111, or fax(212) 355-0408.

Photocopying rights are hereby grant-ed by Scientific American, Inc., to li-braries and others registered with theCopyright Clearance Center (CCC) tophotocopy articles in this issue of SCI-ENTIFIC AMERICAN for the fee of $3.00per copy of each article plus $0.50 perpage. Such clearance does not extendto the photocopying of articles for pro-motion or other commercial purposes.Correspondence and payment shouldbe addressed to Copyright ClearanceCenter, Inc., 27 Congress Street, Sa-lem, Mass. 01970. Specify CCC Ref-erence Number ISSN 0036-8733/93.$3.00 + 0.50.

Editorial correspondence should beaddressed to The Editors, SCIENTIFIC

AMERICAN, 415 Madison Avenue, NewYork, N.Y. 10017-1111. Manuscriptsare submitted at the authorsÕ risk andwill not be returned unless accompa-nied by postage.

Advertising correspondence shouldbe addressed to Advertising Manager,SCIENTIFIC AMERICAN, 415 Madison Av-enue, New York, N.Y. 10017-1111, orfax (212) 754-1138.

Subscription correspondence shouldbe addressed to Subscription Manager,SCIENTIFIC AMERICAN, P.O. Box 3187,Harlan, IA. 51537. The date of the lastissue of your subscription appears oneach monthÕs mailing label. For changeof address notify us at least four weeksin advance. Please send your old ad-dress (mailing label, if possible) andyour new address.


It is time that controls injury. Thetime scale of a fall is set by gravity andstep size. The recovery from a misstepdemands swift response by the nervoussystem. The reßex action to extend thefoot takes enough time to fall two inch-es. Falls of less than that height are un-retarded by muscular action and al-ways lead to a jolt. Willed responsecomes too late to control the eÝect ofencountering an unforeseen seven-inchstep. The functional stretching reßexthat might soften a fall also comes toolate. Thus, Òshort, unexpected fallsÓ of-ten cause remarkable injuries; oncecontrol is lost the fall may continue,unless indeed handrails oÝer a secondchance desperately reached for. (This re-viewer regularly blesses well-designedand well-built handrails.)

Step geometry is a perpetual concernof architects. Alvar Aalto recalls a pagein the Inferno where Dante says theworst thing there is that all the stairshave the wrong proportions. Modernbuilding codes still adopt empirical pro-portions Þxed in a rather idealized wayby Fran�ois Blondel in Paris in 1675.They were so clearly put that they havetaken the force of law today in manyplaces, although peopleÕs foot and legdimensions have changed since, andBlondelÕs units are often even miscon-verted. Design tables are given here forbetter safety and comfort, the dimen-sions based on 20th-century work, notyet deÞnitive.

Chapters on surfaces resistant to slip,on designing stairs safe for crowds, afull treatment of railings, and subtlestudies of how people use stairs follow.Since walking gait is repetitive, preci-sion is required; a ÒnoiseÓ in step dimen-sions of more than 3/16 inch is a hazard.Lighting and the contrast and visibilityof step edges, even distraction by theviews a stair opens, are all sharply rele-vant to safe design. Transitions mattergreatly; 70 percent of all stair accidentsoccur on the Þrst three or on the lastthree steps of a ßight. ÒSoftÓ stair de-sign (but not too soft), using forgivingshapes and surfaces like up-to-date au-tomobile dashboards, lies just ahead. AÒquick design checklistÓ with muchquantitative material ends the book.

The most interesting images in thebook record human behavior on stairsover long periods, favored paths wornin stone as plainly as through grassylawns. Erosion diÝers between ascentand descent. Coming down, the foot-step rounds the step nosing; going up,it wears the treads concave. Study theold stone steps at Wells Cathedral orobserve the paths of users in the NewYork subways: ÒHuman behavior onstairs is largely consistent over time.Ó


am a black professor at a majorwhite university. As an immunolo-gist, I am well known for my work,

both in this country and abroad. I amodd in that there arenÕt many black sci-entists doing what I do. Achieving suc-cess in science is diÛcult, but my experi-ences have convinced me that it is hard-er for blacks, who must harbor greaterambition and develop a very thick skin.

How did I Þnd my way here? A privi-leged background is certainly not thereason. I was born in 1936 in a ruralcommunity just outside Annapolis, Md.The state was then unabashedly segre-gated. The elementary school I attendedwas a two-room frame building that saton blocks and lacked running water andcentral heating. I had only two teachersthrough those elementary years, andthey both helped me appreciate thepower of knowledge at a young age.

One incident stands out. I was in thesecond grade. My teacher, Mrs. Minor,sent me on an errand to the local store. Igave the clerk a dollar bill and received a handful of change, which I did notcount. I returned with the item and thechange for Mrs. Minor. Before I couldturn around, she informed me that theclerk had short-changed me by a nickel.There was no punishment, but she didemphasize that it was the principle in-volved, and not the nickel, that was im-portant. I was more than capable of per-forming the task given me, but I did notuse that knowledge to ensure a fair trans-action. The message to this seven-year-old was that he had better start using hisbrain. I have never forgotten that lesson.

As an elementary student I was oblivi-ous to the unfairness of having poorerfacilities than white students had. Per-haps this is not so surprising: our schoolreßected our standard of living. I wasin my teens before our house had run-ning water and central heating. My fa-ther worked as a laborer at the NavalAcademy, my mother as a domestic. Iam unusually lucky to have had parentswith such vision and commitment: de-spite their modest income, they man-aged to set aside enough to pay a sub-stantial part of my undergraduate edu-cation and that of my two brothers.

When I entered secondary school, I be-came quite aware of the inequities, butmy own passion for learning and thededication of my teachers enabled me tothrive. In that period I developed a pas-

sion for science that was nurtured by myteachers and admired by my classmates.My high school experiences left me feel-ing well educated and conÞdent of suc-ceeding at college. In those days gradu-ates of my black high school continuedtheir education at black institutions. Forreasons I cannot clearly explain, I decid-ed I would break tradition and attend awhite university. This turned out to bethe most signiÞcant educational deci-sion I ever made, because it propelledme into the white academic and intel-lectual world, where I have remained.

After mulling through cataloguesfrom a number of schools, I ap-plied only to Ohio State Univer-

sity, where I was accepted without prob-lem. During freshman orientation I dis-covered that about 5,900 of my 6,000freshman classmates were white. At onesession we were informed that only onein three of us would be around for grad-uation. I knew I would be among them.Indeed, I thrived in the competitive andsomewhat impersonal environment atOhio State and graduated in 1958 witha B.S. in microbiology. I had six or sevenclose black classmates. We stayed in thesame dorm, had the same disciplinedstudy habits and did well. We viewedour achievements not so much as blackstudents but simply as students.

I chose to stay at Ohio State for grad-uate school to pursue research in mi-crobiology. After many futile attemptsto Þnd a summer job in a laboratory, Iturned to Matthew Dodd, from whom Ihad taken two courses in immunology.Although I received AÕs in both cours-es, I had initially avoided Dodd becausehis gruÝ manner intimidated me. HeoÝered me a job, and the immunologicexperiments in which I participated thatsummer spurred me to work with himtoward a Ph.D. in immunology. It turnedout I was not DoddÕs Þrst black student.Two others had preceded me.

Following graduation in 1962 and ayear of postdoctoral work at Ohio State,I tried to Þnd an academic position. Iwas singularly unsuccessful, and Þnal-ly I accepted a research post with theFood and Drug Administration in Cin-cinnati. For the most part, I used theposition for my own research interestsand was among the Þrst to show thatstaphylococcal enterotoxins have pro-found eÝects on the immune system. I

was also one of the Þrst researchers tostudy the role of interferons in immuneregulation. The interferon studies wereinitiated in collaboration with Sam Bar-on at the National Institutes of Health.

Baron left the NIH for chairmanshipof the department of microbiology atthe University of Texas Medical Branchat Galveston and in 1975 invited me tojoin the department as an associateprofessor. Three years later I was pro-moted to full professor with tenure. Astime passed, my program grew but sodid personal strain between Baron andme. So in 1983 I accepted a visiting pro-fessorship at the University of Floridaat Gainesville. In 1989 I was promotedto graduate research professor.

In addition to being successful in sci-ence, blacks who aspire to high academ-ic position must cope with unique barri-ers. Here at the University of Florida, forexample, I had broad faculty support formy appointment but little administra-tive backing. President Marshall Criserhad to step in and insist. It seems thatpowerful forces must sometimes get in-volved in order to bring qualiÞed blackscholars into the academic community.

What role do the professional soci-eties play in the life of a black scien-tist? Not a large one, from my experi-ence. As an example, I have been anelected member of the American Asso-ciation of Immunologists (AAI) since1969. Until recently, the AAI has shownlittle or no interest in encouraging theparticipation of qualiÞed blacks. Notplaying a role in professional organiza-tions is more signiÞcant than it sounds:it means fewer contacts in the variouscirclesÑthe journals, the academic com-mitteesÑthat shape oneÕs career.

I feel that my success comes from acombination of things. I discovered ear-ly that I could Òcut itÓ cognitively in anykind of environment. I have an insa-tiable appetite for discovery that feedsand drives my love of research. I amambitious and want to achieve. I am in-ternally tough, which helps me func-tion in an environment of people whodo not much care for me. It is sad, but Ifeel that this is what it takes to succeedas a black scientist in a white intellectu-al environment. Thank God, it hasnÕtdriven me crazy.


ESSAY by Howard M. Johnson

The Life of a Black Scientist

HOWARD M. JOHNSON is a graduate

research professor in the department ofmicrobiology and cell science at the Uni-

versity of Florida at Gainesville.


the-eye.eu american 1993-2013... · t he pitohui sounds like what it is: something to spit out. the...

Documents