October 1993 $3.95

OCTOBER 1993 $3.95

Is Los Angeles winning the war on smog? Computers that mimic damaged brains. Vast lava ﬂows that reshaped the earth.

Video proofs, such as this one of a sphere being turned inside out, are transforming mathematics.

32 Clearing the Air in Los Angeles James M. Lents and William J. Kelly

The caustic brown smog that often veils the San Bernardino Mountains attests that the air quality in Los Angeles is still the worst in the U.S. Yet it obscures a remarkable achievement: during the past two decades, pollution has been cut dramatically—even as the city’s population and the number of automobiles clog- ging freeways soared. The cleanup is one that other cities might emulate.

42 Large Igneous Provinces Millard F. Co¤n and Olav Eldholm

Periodically in the earth’s past, massive upwellings of magma have created vast elevated plains, both on land and beneath the sea. Unlike the comparatively steady volcanism at the margins of continental plates, these powerful spasms occurred extremely rapidly in geologic time. Studies of these ancient lava ﬂows indicate they may have profoundly altered the global climate and the course of evolution.

50 Evolutionarily Mobile Modules in Proteins Russell F. Doolittle and Peer Bork

Like necklaces strung from beads, many proteins consist of discrete modules that have distinct structures and functions. Surprisingly, some of these individual domains appear in animal and bacterial cells. Does that imply that they are ancient relics of their common ancestry? Not always, the authors contend; some of them may have jumped across species lines—and done so fairly recently.

58 Electrorheological Fluids Thomas C. Halsey and James E. Martin

They are liquid until an electric current is applied; then they ooze like honey or solidify like gelatin—all in less time than the blink of an eye. The unusual properties of electrorheological ﬂuids, ﬁrst patented in 1947, have suggested applications ranging from automotive clutches to adaptive shock absorbers. Only now are technical impediments to commercialization being overcome.

68 Water-Pollinated Plants Paul Alan Cox

With fickle currents and changing tides, water seems a poor vector to disperse the pollen of flowering plants. But some aquatic species have developed strategies that the mathematics of search theory proves to be quite e¤cient. These adaptations to exploit the physics of fluids enabled terrestrial plants to return to an aquatic environment and are classic examples of convergent evolution. Copyright 1993 Scientific American, Inc. 4 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 rights reserved. 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 stored in 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 mailing 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 Scientific 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.

76 Simulating Brain Damage Geo›rey E. Hinton, David C. Plaut and Tim Shallice

Certain injuries of the brain produce bizarre patterns of errors in reading. The same aberrations can be reproduced in computer models by damaging information pathways. Such simulations add support to current ideas about the nature of dyslexia and the way written language is processed in the brain.

84 Raising the Vasa Lars-Akeº Kvarning

On a sunny day in 1628, the proudest addition to the navy of Sweden’s King Gustavus II Adolphus foundered just minutes into her maiden voyage. Today this magniﬁcent, ﬂawed vessel is the centerpiece of a museum in Stockholm. Here is the story of her 30-year-long salvage and painstaking restoration.

92 TRENDS IN MATHEMATICS The Death of Proof John Horgan, senior writer

Mathematicians have always measured the progress of their search for truth in the precise language of the proof. But computers are putting a new spin on QED. No mere human can verify the accuracy of the enormous calculations in so-called computer proofs. Will mathematicians be forced to accept that their assertions are, at best, only provisionally true, true only until they are proved false?

DEPARTMENTS 14 Science and the Citizen 8 Letters to the Editor The ease of elaboration.... Wealth and health.... Mauled anecdote. The toll of child labor.... A winning strategy.... Gene therapy as a can- 12 50 and 100 Years Ago cer treatment.... Nuclear subs for 1893: A remarkable experiment oceanographers.... Sharks get tu- that only a few could repeat. mors.... Art that evolves.... Jurassic viruses.... PROFILE: Twice Nobelist Frederick Sanger. 112 The Amateur Scientist Fluids that turn solid in a magnetic field. 104 Science and Business 114 Book Reviews Malthusian musings.... The ships Critical of CRADAs.... Is magnetic of Iberia .... Following the flu. resonance really safe?... The battery bottleneck .... Cordless elevators.... Impoverished elderly.... A gas meter 120 Essay : Richard Wassersug that listens to the flow.... THE ANA- The unstoppable human LYTICAL ECONOMIST: Pondering high- pilgrimage to the planets. tech fixes for developing nations. Copyright 1993 Scientific American, Inc. 5 THE COVER shows a scene from a comput- ® er-generated film of a sphere being turned inside out, or everted. The “video proof,” which was produced at the Geometry Cen- Established 1845 ter in Minneapolis, Minn., is based on a topological theorem by William P. Thurston EDITOR: Jonathan Piel of the Mathematical Sciences Research In- BOARD OF EDITORS: Alan Hall , Executive Editor ; stitute (see “The Death of Proof,” by John Michelle Press, Managing Editor ; John Rennie, Horgan, page 92). The rules of topology al- Russell Ruthen, Associate Editors; Timothy M. low the skin of the sphere to be stretched Beardsley; W. Wayt Gibbs; Marguerite Holloway ; and twisted and even to pass through it- John Horgan, Senior Writer ; Philip Morrison, Book Editor ; Corey S. Powell; Philip E . Ross; Ricki self, but the eversion must be completed L. Rusting; Gary Stix; Paul Wallich; Philip M. Yam without the formation of a kink. ART: Joan Starwood, Art Director ; Edward Bell, Art Director, Graphics Systems; Jessie Nathans, Associate Art Director ; Nisa Geller, Photography Editor ; Johnny Johnson, Assistant Art Director, THE ILLUSTRATIONS Graphics Systems Cover illustration by the Geometry Center, University of Minnesota COPY: Maria-Christina Keller, Copy Chief; Nancy L. Freireich; Molly K. Frances; Daniel C. Schleno› Page Source Page Source PRODUCTION: Richard Sasso, Vice President, Production; William Sherman, Production Man- 32–33 Miguel L. Fairbanks (top, 76–77 Courtesy of Tim Shallice ager ; Managers: Carol Albert, Print Production; first six photographs), Ken Janet Cermak, Quality Control; Tanya DeSilva , Biggs/Tony Stone Images 78 Johnny Johnson (top), Prepress; Carol Hansen, Composition; Madelyn (top, last photograph), Boris Starosta (bottom) Keyes, Systems; Eric Marquard, Special Projects; Leo J. Petruzzi , Manufacturing & Makeup; Ad Edwin Maynard/The 79–82 Boris Starosta Tra¤c: Carl Cherebin Environmental Picture CIRCULATION: Lorraine Leib Terlecki, Circulation Library (bottom) 84–85 Courtesy of Vasa Director ; Joanne Guralnick, Circulation Promo- Museum, Stockholm 34–35 Jana Brenning tion Manager ; Rosa Davis, Fulfillment Manager ; Katherine Robold, Newsstand Manager 86–87 Jana Brenning (top), 36 Johnny Johnson Vasa Museum (bottom) ADVERTISING: Robert F. Gregory, Advertising Director. OFFICES: NEW YORK: Meryle Lowen- 38 UPI/Bettmann Newsphotos 88 Hank Iken thal, New York Advertising Manager ; William Buchanan, Manager, Corporate Advertising ; Pe- 39 Alon Reininger/Contact 89–91 Courtesy of Hans ter Fisch, Randy James, Elizabeth Ryan. Michelle Press Images Hammarskiöld, Larsen, Director, New Business Development. Vasa Museum CHICAGO: 333 N. Michigan Avenue, Chicago, IL 42–43 Ian Worpole 60601; Patrick Bachler, Advertising Manager. 92–93 Geometry Center (computer DETROIT: 3000 Town Center, Suite 1435, South- 44 Ocean Drilling Program art), Stephanie Rausser field, MI 48075; Edward A. Bartley, Detroit Man- (left), Patricia J. Wynne (photograph) ager. WEST COAST: 1554 S. Sepulveda Blvd., (right) Suite 212, Los Angeles, CA 90025; Kate Dobson, 94 Robert Prochnow Advertising Manager ; Tonia Wendt. Lisa K. Car- 45 Patricia J. Wynne (left), den, Lianne Bloomer, San Francisco. CANADA: Australian Geological 95 Per Breiehagen/Black Fenn Company, Inc. DALLAS: Gri¤th Group Survey Organization Star (photograph); MARKETING SERVICES: Laura Salant, Marketing (top right), Millard F. Jean E. Taylor, Rutgers Director ; Diane Schube, Promotion Manager ; Co¤n (bottom right) University (computer art) Mary Sadlier, Research Manager ; Ethel D. Little, Advertising Coordinator 46–48 Ian Worpole 98 James T. Ho›man, Edward C. Thayer, INTERNATIONAL: EUROPE: Roy Edwards, Inter- national Advertising Manager, London; Vivienne 49 Roger Ressmeyer/Starlight G.A.N.G., Mathematics Davidson, Linda Kaufman, Intermedia Ltd., Par- Department, University 50–51 Visual Logic is; Karin Oh›, Groupe Expansion, Frankfurt; of Massachusetts at Am- Barth David Schwartz, Director, Special Proj- 52–58 Jared Schneidman Design herst (computer art); ects, Amsterdam. SEOUL: Biscom, Inc. TOKYO: Jessica Boyatt (photograph) Nikkei International Ltd. 59 Thomas C. Halsey 99 David Ben-Zvi, ADMINISTRATION: John J. Moeling, Jr., Publisher ; Marie M. Beaumonte, Business Manager 60–62 Jared Schneidman Design Geometry Center (computer art); Robert SCIENTIFIC AMERICAN, INC. 63 Thomas C. Halsey Prochnow (photograph) 415 Madison Avenue New York, NY 10017 64 © 1993 Universal 100 Johnny Johnson (212) 754-0550 Studios/Amblin Productions; Courtesy 112 Michael Goodman PRESIDENT AND CHIEF EXECUTIVE OFFICER: of Industrial Light & Magic John J. Hanley 113 Jason Kü›er CORPORATE OFFICERS: Executive Vice President 68–71 Patricia J. Wynne 114 Cli› Dwellers (1913), and Chief Financial O¤cer, R. Vincent Bar- by George Wesley Bellows ger ; Vice Presidents: Jonathan Piel, John J. 72–73 Paul Alan Cox Moeling, Jr. (photographs), Patricia (American, 1882–1925); J. Wynne (drawings) Los Angeles County CHAIRMEN OF THE BOARD: Dr. Pierre Gerckens, Museum of Art, Los John J. Hanley 74 Paul Alan Cox Angeles County Funds CHAIRMAN EMERITUS: Gerard Piel PRINTED IN U.S.A.

Simplify, Simplify 1980s, when China started the market pendence. Any government that had to reforms that increased food production face opposition parties, free newspapers Essayist W. Brian Arthur [“Why Do while worsening its distribution. Simi- and regular elections could not a›ord Things Become More Complex?” SCIEN- larly, Cuba has been able to reduce fam- to have famines. During those years of TIFIC AMERICAN, May] observes that fun- ine and malnutrition much faster than famine, the Chinese government was damental designs tend to accrete in- other Latin American countries, includ- under little political pressure to aban- creasingly complex hierarchies of sup- ing Costa Rica and Jamaica, the two Sen don the policies that were directly con- port subsystems as their uses expand, cited as successes. tributing to the problem. A similar gov- until they are supplanted by improved The root of the problem of malnu- ernmental immunity sustained famines designs. Arthur then wonders whether a trition in the world is the skewed con- in Cambodia in the 1970s, in the Soviet general principle underlies this tendency. centration of wealth that is imperme- Union in the early 1930s and in many Yes, a well-known characteristic of able to democratic pressures. Only those nondemocratic regimes—both of the invention: it is much easier to elabo- countries that have redistributed their “right” and the “left”—in sub-Saharan rate than to innovate. wealth have been able to solve the mal- Africa. More democratic regimes in Bots- nutrition problem. wana and Zimbabwe successfully pre- DAVIDSON CORRY vented famine. Seattle, Wash. VICENTE NAVARRO I did not attribute the relatively low Departments of Health Policy, life expectancy in Saudi Arabia to “mis- Arthur writes that “Copernicus’s daz- Sociology and Policy Studies management of health care funds.” Rath- zlingly simple astronomical system, Johns Hopkins University er the main issue concerns the overall based on a heliocentric universe, re- priority that is given to health care and placed the hopelessly complicated Ptol- We wish to take exception to the arti- education (especially female education) emaic system.” Copernicus’s heliocen- cle by Sen. The author draws the inac- compared with other expenditures. To tric system was actually more complex. curate conclusion that life expectancy attribute the achievements of the Indian Because Copernicus used circular orbits in Saudi Arabia is lower than in many state of Kerala to the benign influence to describe the motions of the planets poorer countries because of misman- of the British Empire would be mistaken (as did the Ptolemaic system), he was agement of health care funds. You must on two grounds. First, most of Kerala forced to use epicycles to account for remember that major progress toward remained outside British India. Second, their apparent retrograde motions as modernization in the Kingdom did not the biggest expansions of education and viewed from the earth. Not until Kep- begin until 1970. India, in contrast, has health care in Kerala have come only in ler discovered that the planets move in a long history of modernity, including recent decades, much influenced by left- elliptical orbits around the sun, some British colonization. Thus, Saudi Arabia wing political movements. 50 years later, did the heliocentric sys- has lacked the necessary infrastructure tem emerge in its true simplicity. required to achieve the levels attained by others. What It Was, Wasn’t DERICK W. OVENALL Wilmington, Del. ABDUL-MANNAN TURJOMAN I admire presidential science adviser Scientific Counselor John H. Gibbons as much as the next Royal Embassy of Saudi Arabia person [“Profile,” “Science and the Citi- Vital Priorities Cultural Mission to the U.S.A. zen,” by John Horgan; SCIENTIFIC AMER- Washington, D.C. ICAN, April], but I fervently hope—for In “The Economics of Life and Death” the sake of the nation—that his expertise [SCIENTIFIC AMERICAN, May], Amartya Sen replies: in technology is superior to his knowl- Sen postulates that the problem is not Navarro is right to argue that unequal edge of popular culture. Gibbons quotes only production but distribution of food. distributions of wealth and income have a supposed Tennessee Ernie Ford story Unfortunately, he does not touch on the much to do with undernourishment. about a mountaineer watching a football political conditions necessary for that The need for public services, on which I game. Alas, it wasn’t Tennessee Ernie distribution. The only time he refers to focused, relates to this inequality. Ford, it was Andy Gri¤th in the 1953 this issue is when he writes that, in gen- His argument about India and China comedy routine that made him famous, eral, “democratic” countries such as In- confuses famine with endemic under- What It Was, Was Football. Poor Gibbons dia have been able to avoid famine bet- nourishment and ill health. They are mangled other details, too (although, in ter than “dictatorships” like China. distinct problems. I had, in fact, not- Gri¤th’s original, the tale spinner was The available information, however, ed that “even though postrevolutionary drinking a big orange drink). does not support that statement. Con- China has been much more successful trary to what Sen indicates, China has than India in economic expansion and HOWARD R. COHEN been able to feed its people better than in health care, it has not been able to Los Angeles, Calif. India has. Although China had worse stave o› famine.” Navarro overlooks the health and nutritional levels than India gigantic famine of 1958–1961 in China, Because of the volume of mail, letters in the 1940s and 1950s, China has been which is estimated to have killed from to the editor cannot be acknowledged. able to improve its health and nutrition- 23 million to 30 million people. India Letters selected for publication may be al indicators faster, at least until the has not had a famine since its inde- edited for length and clarity.

OCTOBER 1943 Also in castings it cools into crystalline not the race benefits by this custom, structure, the crystals coarse and full and Dr. A. S. Ashmead, of New York, “For many years it was believed that of holes—possibly from absorbed gas- discusses the question in the Sei-i-Kwai the isolation of a protein in the chemi- es driven o›.’ ” medical journal. In the first place it is cally pure state was nearly impossible. assumed that indirectly the absence of Recently, however, this situation has cow’s milk is most beneficial. The Jap- changed quite completely. The enzymes anese mother feels the compulsion of pepsin, trypsin, and urease and the hor- looking after her own health and diet. mone insulin have all been isolated and Japanese mothers chiefly live on rice, appear to be proteins. One of the bright- OCTOBER 1893 ‘fish, shells, seaweed, and other prod- est chapters in this search for protein in ucts of the sea,’ while wine and beer its chemically pure state is the isolation “It now does not seem improbable are rigidly excluded. The reward of all of the virus of the tobacco mosaic dis- that, when by the power of thought an this meritorious care of motherhood ease. This virus is described as a crystal- image is evoked, a distant reflex action, and childhood is the absolute freedom line protein. The significance of this dis- no matter how weak, is exerted upon from rickets. Again, the author holds covery lies not only in the great advance certain ends of the visual nerves, and, that the transmission of tuberculosis is in the understanding of protein struc- therefore, upon the retina. Helmholtz avoided by the exclusion of cow’s milk tures but in the fact that it connects has shown that the fundi of the eyes from the infant’s dietary.” proteins with measles, yellow fever, the are themselves luminous, and he was common cold, and several other diseas- able to see, in total darkness, the move- “E›orts have been made to teach a es of both plants and animals.” ment of his arm by the light of his own child how to swim by supporting him in eyes. This is one of the most remark- the water and causing him to e›ect the “Motors running on alternating cur- able experiments recorded in the histo- motions of natation. This is the most rent have only a few fixed speeds de- ry of science, and probably only a few practical process. Its inconvenience is pending upon the motor construction men could satisfactorily repeat it, for it that it necessitates the presence of a and the frequency of the current sup- is very likely that the luminosity of the teacher with each pupil, and, in a large plied to them. What industry desires is eyes is associated with uncommon ac- class of children, the teacher cannot the flexibility of the direct-current mo- tivity of the brain and great imaginative occupy himself with each of them for a tor and the e¤ciency of transmission power. It is fluorescence of brain action, very long time. Mr. Devot has been able of the alternating-current system. Elec- as it were.—Nikola Tesla, in a paper to overcome all the di¤culties of the tronics now makes this possible. The read before the Franklin Institute.” preceding method in a very ingenious tubes employed are thyratrons—gas- manner. His apparatus (below) permits eous or mercury-vapor-filled rectifiers “Japanese children are suckled until the pupil to learn in conditions entirely which convert alternating current into their sixth year, and in language un- identical with those that present them- direct current. The portion of the posi- mistakable may be heard asking for selves when he tries to sustain himself tive half cycle during which current the lactatious fountain. In view of the alone in the water. The apparatus is in flows through the tube can be varied almost universal use of cow’s milk in use among the pupils of the Michelet Ly- by merely changing the voltages on the other countries, its exclusion from the ceum, who have been the first to benefit grid with respect to the phase of the diet of the Japanese raises the interest- from the invention of their master, Mr. alternating voltages placed upon grid ing subject of inquiry as to whether or Devot.—La Nature.” and anode.”

“When an optical surface of glass is aluminized, the evaporated molecules, being in a high vacuum, travel, without bumping into other molecules, from the hot metal source to the mirror’s cold surface and are deposited in a non- crystalline metallic ﬁlm having the same degree of polish as that of the glass. As soon as air is admitted, the metallic aluminum begins to oxidize, and this oxide continues to thicken for about 60 days. Why couldn’t all this be as easily accomplished by letting a disk of plain cast aluminum oxidize in the air? Fred B. Ferson, a Biloxi, Mississippi, amateur telescope maker, states it thus: ‘Alumi- num is a metal which absorbs gases readily, and is hard to prevent from taking up impurities when it is cast. Apparatus for teaching swimming

Hard Times Occupational injuries among children are increasing

he sepia photographs and etch- ings lie in the archives of nearly T every industrialized nation: children in factories operating or dodg- ing dangerous machinery, working too many hours for too little pay. Now, for the ﬁrst time since U.S. labor laws protecting children were enacted in 1938, such stark images are being recorded again in record numbers. The number of children working—legally and illegally—has risen dramatically throughout the country in the past decade. Accom- panying this increase are numerous, sometimes grisly, reports of labor law violations as well as occupational injuries—amputations, electrocutions, frac-

tures, burns, lacerations—and deaths. SABA Federal law prohibits anyone under the age of 14 years from working, except in some agricultural jobs, and anyone younger than 18 years from performing certain tasks, such as operating heavy MARK PETERSON machinery or working on construction SWEATSHOPS in New York City’s garment district have been found to employ young sites or with toxins. Nevertheless, near- people illegally. Many such factories have dangerous working conditions. ly every industry employs children of all ages. About ﬁve million teenagers work in the U.S. In 1991 an additional 27,528 Estimating the number of injuries Sinai Medical Center in New York City. children were discovered in illegal jobs, among these children is even more dif- Despite gaps in the data, the Nation- an increase of about 300 percent since ficult. The statistics that are available al Institute for Occupational Safety and 1983, according to the General Account- rely on worker’s compensation data and Health determined that more than 130 ing O¤ce (GAO). The number of youths hospital records that are, at best, in- children died on the job in 1990. Sever- working clandestinely has been estimat- complete. According to child labor ex- al other studies have looked at state cas- ed to be as high as 676,000. perts and pediatricians, children work- es. Landrigan and his colleagues recent- Few studies chronicle the hazards ex- ing for family businesses often do not ly reported in the Journal of the Ameri- perienced by these young people. The report injuries to worker’s compensa- can Medical Association that 9,656 New U.S. Department of Labor keeps com- tion boards. In most states, agricultural York State adolescents received work- prehensive records only of wage and labor is not even covered by worker’s er’s compensation between 1980 and hour infractions: children working for compensation. 1987. Of those, 43.5 percent experienced less than minimum wage or for exces- Other youths, particularly those em- a permanent disability—that is an aver- sive hours. By this measure alone, the ployed illegally, can be unaware of the age of 525 teenagers a year—and 31 country is not doing well. In 1990, the procedure or can be discouraged by children died. year of a major federal e›ort called Op- their employer from making a claim. A In a study to be published in the eration Child Watch, the department un- 1988 survey by the New York State De- American Journal of Industrial Medi- covered 39,000 such violations. In 1991 partment of Labor, for instance, found cine, the Massachusetts Department of and 1992, years when no special opera- that more than half of the teenagers Public Health found that the rate of oc- tion was in place, about 29,000 and 19,- questioned had experienced an injury cupational injury among teenagers was 000 violations were reported, respective- or a wage or hour violation. Only one about twice that of adults. Of the 17- ly. Advocacy groups argue that the ﬁg- third of those injured reported the ac- year-olds whose location of injury was ures are actually much higher. “We have cident. Underreporting may also occur recorded when they were treated in sam- no compunction about saying that there because physicians generally do not re- pled emergency rooms, 26 percent were are over two million child labor viola- cord information related to a young pa- there because of a work-related injury. tions each year,” asserts Je›rey New- tient’s occupation, notes Philip J. Lan- And researchers at the University of man of the National Child Labor Com- drigan of the division of environmen- Washington in Seattle determined that mittee in New York City. tal and occupational medicine at Mount farm work accounted for nearly 50 per-

14 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. cent of the serious injury claims among enforcement as the primary deficiency Sentries and Saboteurs children 13 years old and younger. Ac- in U.S. law and seek to increase such cording to an article in the American e›orts and associated penalties. Mutating patients’ genomes Journal of Public Health, children aged Internationally, conditions look more to suit their medicine 19 years and younger are involved in dire. According to the International La- 25,000 farm accidents around the coun- bor Organization (ILO) in Geneva, some try annually; 300 are killed. 200 million children under the age of octors treating cancer with The reasons for the increases in child 15 years work. The numbers “seem to chemotherapy play the odds. labor are varied. Landrigan and others be increasing in the Third World, where DToo little drug and the diseased cite poverty and immigration as con- you have worsening economic condi- tissue will survive; too much and the tributing factors. Indeed, according to tions and worsening political situations,” damage to healthy cells will be more the Children’s Defense Fund, there are notes Susan E. Gunn of the ILO. In Bra- than the patient can bear. But some re- 14.3 million poor children in the U.S., zil, 18 percent of children aged 10 to 14 searchers believe gene therapy can in- the highest number since 1965. Many years work—a total of seven million. crease the spread. In upcoming clinical of them have di¤culty finding above- “But frankly, it is very hard to generalize trials they intend to insert into cancer board jobs, notes Joseph A. Kinney of from country to country,” Gunn says. patients’ DNA mutations that either the National Safe Workplace Institute in For instance, in Thailand the number make healthy cells more tolerant of ex- Chicago. The GAO has found that minor- of working children rose 34 percent to isting medication or make tumors more ity and low-income children are more an estimated 1.7 million between 1983 vulnerable to it. “This is a totally nov- likely to be employed in hazardous oc- and 1987 while the economy boomed: el approach, but a natural extension of cupations than are their white or high- exports rose 84 percent. gene therapy,” says Albert B. Deisseroth er-income counterparts. If it is di¤cult to find figures on the of the University of Texas M.D. Ander- In New York City, for example, some extent of occupational injuries in the son Cancer Center in Houston. 1,500 sweatshops in the garment indus- U.S., it is virtually impossible worldwide. In June the Recombinant DNA Adviso- try are a large source of illegal work— Isolated reports hint at the extent of ry Committee of the National Institutes and child labor violations. “Most of the the problem. For instance, this spring a of Health approved two protocols for children in these factories are below 18, widely publicized factory fire near Bang- human tests of a strategy that Deis- some as young as eight,” notes Thom- kok left 500 people injured and 188, seroth calls “chemoprotection.” He and as Glubiak, chief of the city’s 20-mem- mostly young women, dead. The work- others plan to genetically modify pa- ber garment district task force. These ers were making dolls for Kader Indus- tients’ bone marrow cells so they can youths are exposed to myriad dangers, trial, a supplier for the manufacturer withstand more courses and higher dos- he notes. “We find unguarded machin- of Cabbage Patch Kids. Whereas in the es of chemotherapy. Human trials are ery, no fire exits, boilers, wiring prob- U.S. more young boys are injured, in also under way to test another gene lems, egress problems, machines too other parts of the world girls are more therapy technique that is essentially a close together.” According to the New likely to be abused. Employers “favor mirror image of chemoprotection. Rath- York State Department of Labor, the hiring girls—they feel girls are more er than inserting “sentries” to guard number of city establishments illegally exploitable,” says Robert Senser of the healthy cells against dangerous drugs, a employing children rose from 19 in Asian-American Free Labor Institute in strategy called molecular surgery sends 1987 to 122 in 1988. Washington, D.C. “Most of the workers in genetic “saboteurs” to make tumor At the same time as more children in the garment sectors are girls. They cells susceptible to a normally harm- have entered the work force, advocates are considered more docile and more less medicine. argue that enforcement of child labor dexterous.” The Texas group headed by Deisse- laws has diminished. Many experts con- The recent attention turned on youth roth plans to enlist 20 to 30 women sider the labor laws to be strong, in- employment is not limited to the U.S.— whose late-stage ovarian cancer has al- deed, most states have legislation that international concern is rising as well. ready bested standard surgery and drug is more stringent in certain aspects than The ILO reports that it is stepping up its treatment. The most e›ective medicine federal law. But they perceive follow-up programs. Although there is no mecha- available for such cases is often Taxol, to be weak. “Active enforcement is ab- nism for enforcement, Gunn notes that a toxin derived from the bark of the Pa- solutely essential,” Newman notes. “This moral pressure can be e›ective. In ad- cific yew tree. Taxol kills cells caught in is one of the areas where you can actu- dition, American consumer groups and the act of dividing; since tumor cells ally stop the crime from happening.” two congressional bills are seeking to multiply more rapidly than do normal Budget cuts and inadequate sta¤ng halt imports of products that have been cells, they are more vulnerable. Unfor- have limited the ability of the federal De- made by children under 15 years of age. tunately, Deisseroth adds, “the major partment of Labor to conduct sweeps, The Indian government has already reside e›ect of Taxol is that it also kills Newman says. The GAO has reported sponded to the threat of sanctions with bone marrow cells,” which produce the that the wage and hour inspectors fol- promises of more crackdowns on child white blood cells of the immune sys- low up on repeat o›enders infrequent- labor violators and a system of labeling tem. Just two or three courses of Taxol ly. And a National Safe Workplace In- to ensure products were made without are enough to deplete most patients’ stitute report stated that only 11 per- exploiting children. According to the marrow so severely that they succumb cent of the inspectors’ time is spent on ILO, about five million Indian children to infection or internal bleeding. child labor; “an establishment that em- work as bonded laborers. Two or three courses are often not ploys adolescents can anticipate a fed- But, as with all aspects of this prob- enough to vanquish the slow-growing eral inspection once every 50 years,” it lem, the impact of trade barriers on the tumors of ovarian cancer. So Deisseroth estimates. The Department of Labor lives of these children has not been wants to remove 10 percent of each pa- counters that its 850 inspectors check studied. “I do not have an opinion yet,” tient’s bone marrow before chemothera- for child labor violations constantly. Gunn says. “It is obviously having an py begins. Using an antibody that binds Regardless of the department’s claim, e›ect, but positive or negative? We just only to younger marrow cells, the re- however, two bills before Congress cite don’t know.” —Marguerite Holloway searchers will try to separate out the

16 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. stem, or progenitor, cells that have yet es a protein that acts as a sentry, es- the more sentries you need to protect to di›erentiate into a particular kind of corting toxic chemicals such as chemo- each cell.” white blood cell. They then plan to in- therapeutic agents out of the cell before By infecting stem cells with MDR- fect these stem cells with a mouse vi- they can do any damage. But for some bearing viruses, the researchers hope rus—crippled so that it cannot repro- unknown reason the gene is turned o› to secure them against Taxol. After each duce and spread—into which has been when a stem cell matures into a white patient completes an ordinary round of inserted the human MDR gene for mul- blood cell, so only stem cells are pro- chemotherapy, she will be injected with tiple-drug resistance. tected and then only weakly. “The guard her own mutated stem cells. Many of All marrow cells naturally have the is easily overrun,” Deisseroth says. “The those cells should carry multiple copies MDR gene. When turned on, it produc- higher the concentration of the drug, of the MDR gene and an extra copy of

Creative Evolution

ike all things cultural, art evolves—a L platitude to most artists, but to a handful, a description of technique. By ap- propriating scientific principles from evolutionary biology and advanced technology from computer graphics, this avant- garde has fashioned a new medium for creative expression: virtual evolution. British artist William Latham designed software to mutate the “genome” of three- dimensional forms. From each generation he selects and breeds the most aes- thetically “fit.” Karl Sims, artist-in-residence at Think- ing Machines, discovers his artworks using a self-mutating program that draws from a genome of equations and transformations. “You quickly evolve equations that you couldn’t design or even understand,” he says. Nevertheless, the process is simple enough that museum-goers can THAM express their taste via interactive exhibits. THAM More dynamically, neural networks add collective behavior to an ecosystem of in- WILLIAM LA telligent paint brushes bred by the genet- WILLIAM LA ic algorithms of Michael Tolson, chief sci- SURVIVAL OF THE MOST INTERESTING is the rule when artists play God and gardener. entist of Xaos Tools. Make way for virtual Beginning with simple objects, William Latham breeds myriad generations of computer- surreality. —W. Wayt Gibbs generated mutations to arrive at his pseudo-organic “White Form” (left). His creations

Thinking Machines KARL SIMS

Thinking Machines SUPERCOMPUTER and image evolution software installed by Karl Sims at the Pompidou Center in Paris let visitors view his art (left) or

KARL SIMS interact with the exhibit to create their own collaborative works (above).

20 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. the regulatory machinery needed to produce it. After a few weeks of recovery time, Deisseroth will begin giving Tax- ol again, in gradually increasing doses. With each course, more of the unmodi- ﬁed marrow cells—greater than 90 percent of the marrow at ﬁrst—will die. But if the progeny of the transgenic stem cells keep making the drug-resistance protein, they should prosper. If all goes

can display characteristic behavior and respond to their surroundings, like the chame- leonic “Mutation in Red Room” (above).

ools

aos T

X OLSON MICHAEL T

“OUR FOUNDER,” part of a series by Michael Tolson awarded the Goldene Nica Award at the 1993 Prix Ars Electronica in Austria.

hall the meek inherit the earth, or is might right? Students of behavior have Sexpended much effort analyzing whether evolution should favor individuals who cooperate or exploiters who go for short-term gains. A computer tournament over a decade ago indicated that a cooperative behavior strategy known as tit-for-tat can beat out exploiters, although it is not stable in more lifelike simulations. Recent computer modeling has uncovered a more resilient formula, dubbed Pavlov, that prevails over tit-for-tat in the evolutionary race and yet is still cooperative—up to a point. Its discoverers believe Pavlov could be a model for many examples of altruism in the natural world. The standard paradigm for analyzing the evolution of cooperation is a co- nundrum called the Prisoner’s Dilemma. This problem considers two sus- pects, imprisoned separately, each contemplating whether to confess and plead for a light sentence (to cooperate) or to blame the other for the crime (to defect). The payoff for each player depends on what the other does. If both cooperate, they each do better than if they both defect or exploit the other. Yet each could do better by defecting than by cooperating [see table]. Although defecting always pays off in a single round, cooperative strategies can be advantageous if the Prisoner’s Dilemma is repeated—as interactions between members of a community often are. When Robert Axelrod, a professor of political science at the University of Michigan, organized a contest of computer programs to play the Prisoner’s Dilemma in the early 1980s, tit-for-tat was the surprise winner. It was a surprise because it was so simple: tit-for-tat starts by cooperating and in subsequent bouts simply repeats what its opponent did in the previous round. Martin Nowak of the University of Oxford and Karl Sigmund of Vienna Uni- versity showed two years ago that tit-for-tat does not endure in simulations in which, as in the real world, occasional mistakes are made. Eventually, they found, it gets bogged down in bouts of backbiting. Recently Nowak and Sigmund have conducted simulations that also model the effect of mutations, or small changes in strategies. The results, published in Nature, were every bit as startling as those from Axelrod’s tournaments. Under a wide range of conditions Nowak and Sigmund found that if they waited long enough, the strategy they call Pavlov usually dominated their modeled populations. A player following Pavlov cooperates after a contest in which both parties cooperate—and also after a contest in which both parties defect. If only one player cooperated on the previous round, however, Pavlov defects. The strategy is so named because, put another way, it repeats, reflexlike, its previous play if it gets one of the two higher payoffs and switches if it comes out behind. A population of Pavlovian individuals does cooperate and reap the atten- dant benefits. But a Pavlovian population has no tendency to start cooperating indiscriminately. The weakness of tit-for-tat, Nowak and Sigmund say, is that mutation allows populations to become more and more cooperative, which eventually leads to an invasion by selfish “always-exploiters.” Nowak and Sigmund note that animals often retaliate against defection. These findings have been PLAYER 2'S MOVE taken as support for the PAYOFF TO idea that animals follow a PLAYER tit-for-tat strategy. But such PAYOFF 2 COOPERATE DEFECT TO PLAYER 1 results also corroborate the idea that animals use a Pav- 3 5 COOPERATE 3 0 lov-like strategy. Pavlov, while cooperative with oth-

MOVE 0 1 DEFECT ers who cooperate, has “no

PLAYER 1'S 5 1 qualms about exploiting a sucker,” according to its dis- COOPERATE AFTER DEFECT AFTER coverers—it keeps on de- THE PRISONER’S DILEMMA is illustrated by fecting against an oppo- the benefit (numbers of points) each player nent foolish enough to go gains depending on whether he or she cooper- on cooperating. And that ates or defects. In sophisticated simulations, the ruthlessness is what allows strategy called Pavlov (shown in green and yel- Pavlov, and cooperation, to low) usually eventually predominates. survive. —Tim Beardsley

22 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. well, the patient’s entire marrow and blood supply will grow to tolerate Tax- ol while the tumors wither away. Deisseroth admits that much of the process he hopes to modify remains unknown. The virus inserts MDR and its replication sequence randomly into the stem cell genome; the researchers rely on luck to avoid causing unwanted mutations. No one knows why MDR is naturally turned o› in mature white blood cells. There may be a good reason, and short-circuiting that “o›” switch may produce side e›ects. Since ovarian cancer does not spread to the bone marrow, there appears to be little danger of making tumors themselves drug resistant. But as a patient’s cumulative dose of Taxol builds, it will take a toll on other tissues. Raising the typical Taxol dosage just 40 percent has been shown to cause neurological problems, says Charles Hesdor›er of Columbia University, whereas other che- motherapeutic drugs can be increased five- to 10-fold without harm. Hesdor›er sponsored the second, recently approved chemoprotection protocol, which will explore other drugs in addition to Taxol and will also include patients with widespread breast cancer or brain tumors. Deisseroth is confident nonetheless. “We’ve tried this out in mice, and it works,” he claims. “As far as we can tell, the cells are not changed by the MDR gene except in their ability to repel toxins.” But Arthur W. Nienhuis, whose group at the NIH has collected much of the animal data cited by Deisseroth and has submitted a third chemoprotection protocol, warns that mice may not be the best model. “In primates— and it seems to be true in humans as well—only 1 to 5 percent of the modified stem cells actually produce MDR, versus 20 to 50 percent in mice,” he says. That may be too little too late. “I think we need to make sure that we’re getting the gene in, that it’s safe and that it produces protein at a reasonable level,” Hesdor›er adds. “Then we can go on to determine whether this system actually works therapeutically.” Cautious optimism also character- izes experiments to test the molecular surgery technique spearheaded by Ken- neth W. Culver of Iowa Methodist Medi- cal Center, Edward H. Oldfield of the NIH and Genetic Therapy in Gaithers- burg, Md. In experiments last year Cul- ver and Oldfield managed to rescue 11 of 14 rats with terminal brain cancer. The first human trial began last De- cember at the NIH. Culver recently won approval for two more trials to begin in November at Iowa Methodist and at Childrens Hospital in Los Angeles. Molecular surgery sidesteps some of

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 23 the special di¤culties brain cancer pre- Sharks Do Get Cancer sents and capitalizes on the unique opportunities it o›ers. Brain tumors often Cartilage cure relies lurk too deep for scalpels to reach, and on wishful thinking the blood-brain barrier can filter out drugs before they reach their target. But once a brain has matured, its cells pseudoscientific myth holds that largely stop dividing, so anything that sharks don’t get cancer. Indeed, multiplies is probably malignant. A that proposition is the title of Culver came up with the idea of a book by I. William Lane and Linda sneaking a gene from the herpes sim- Comac promoting shark cartilage as a plex virus into tumor cells and then “breakthrough in the prevention and killing them with the antiherpes drug treatment of cancer and other degen- ganciclovir. The gene, which codes for erative diseases.” The CBS television the enzyme thymidine kinase and is program “60 Minutes” enthusiastically thus known as HS-tk, can be carried into picked up the idea earlier this year. a cell by a retrovirus only when the cell Unfortunately, John C. Harshbarger, is dividing, so normal brain cells should director of the registry of tumors in be safe. But to get enough of the en- lower animals at the Smithsonian Insti- gineered virus into the brain to do any tution in Washington, D.C., says he has good, Culver had to genetically trans- records of at least 20 cases of cancer in form mouse cells into viral factories. sharks. The registry, which is support- Culver and Oldfield found that when ed by the National Cancer Institute, in- they injected these virus-producing cells cludes shark cancers that originated in into brain tumors in rats, as much as cartilage, as well as cancers of the kid- 60 percent of the tumor cells incorpo- neys, liver and blood cells. (Harshbarg- rated the HS-tk gene, whereas virtually er says data do not exist to determine no normal cells did. When they then whether sharks get cancer more or less gave the rats ganciclovir, they got an- often than do other creatures.) other pleasant surprise. The drug killed In their book Lane and Comac cite not only the infected tumor cells but an article about cartilage by Arnold I. also many malignant cells that did not Caplan, a biologist at Case Western Re- carry HS-tk. In some cases, even when serve University, published in this mag- only 10 percent of the cancerous cells azine in October 1984. Caplan had not- took up the gene, the whole tumor dis- ed that cartilage seemed to contain sub- appeared without causing inflammation stances that inhibit the growth of blood or hemorrhage. vessels and speculated that such sub- This bizarre “bystander e›ect” does stances might someday be used to pre- not seem to extend to normal cells sur- vent tumors from establishing the blood rounding tumors. “We still don’t know supply they need to grow. how this works,” Culver admits. “One Caplan was right: cartilage does in- hypothesis is that it involves intercel- deed contain inhibitors of blood ves- lular communication.” Destruction of sel growth, or antiangiogenesis factors. the circulatory system within the tumor Some have been isolated and are now may also play an important role, e›ec- in clinical trials for treatment of a vari- tively starving to death those tumor ety of cancers as well as stomach ul- cells that do survive the drug. cers. But Caplan says he is “appalled” Human trials have yet to prove that by Lane and Comac’s e›ort to promote molecular surgery is as precise and shark cartilage as a treatment. “This is thorough in people as it seems to be in an extreme interpretation of the data,” rats. And, Culver points out, this tech- Caplan complains. Moreover, he adds, nique is not without its problems and “There’s nothing special about sharks. risks. The virus-producing cells must Why not eat pigs’ knuckles?” be injected close to the cancer for the Caplan is not the only investigator treatment to work, but magnetic reso- troubled by the shark cartilage craze. nance imaging does not always reveal Judah Folkman, an angiogenesis re- every tumor. If the mouse cells (which searcher at Harvard Medical School, also contain HS-tk and so also die when states firmly that there is no evidence exposed to ganciclovir) manage to es- from any controlled study that ingest- cape, they could infect other parts of ing cartilage can treat cancer. What is the body, such as the bone marrow. more, he notes, on the basis of what For patients who have few options is known, “a patient would have to eat and little hope, however, molecular sur- hundreds of pounds of cartilage” to gery and chemoprotection may pro- have any chance of experiencing an long—perhaps preserve—life. If so, this e›ect. Folkman says he has been trying twist on gene therapy could give doctors for years to dissuade Lane from using a new weapon in the frustrating battle his name to promote a shark cartilage against cancer. —W. Wayt Gibbs product.

24 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. Carl A. Luer, an investigator at the says, are already depleted by overﬁshing. thusiasts are persevering. Charles Si- Mote Marine Laboratory in Sarasota, Fla., The National Cancer Institute has ex- mone, a physician in private practice in who appeared on “60 Minutes,” is also amined some cases of supposed im- New Jersey, maintains that a few of his dismayed that his work on antiangio- provements in patients treated with patients have responded favorably. But genesis factors is being used to promote shark cartilage, but the data “did not even he warns against self-medication. shark cartilage. “I feel our factor is a pro- show solid evidence of clinical activity,” Much of the shark cartilage now on tein and can’t be absorbed,” he states. according to Mary S. McCabe, a clinical sale in health food stores, he says, is Some tropical shark populations, he trial organizer at the institute. Still, en- “bogus stu›.” —Tim Beardsley

Run Silent, Run (Not So) Cheap

or decades, polar oceanographers have wanted to rine also plans to surface in holes or cracks in the ice. These F come in from the cold. To do their work, they have had stops would allow researchers to deploy buoys with strings to lug cumbersome sonar and seismic equipment over of sensors that hang under the ice to measure water tem- the surface of polar ice floes. Meanwhile, a few fathoms peratures and salinity. below, the crews of Soviet and U.S. submarines have lived Langseth, however, believes that researchers will be able in relative comfort as they played cat-and-mouse games to take full advantage of a submarine only once it is per- while practicing for the end of the world. manently retrofitted with the necessary instrumentation. A Some of those frostbitten scientists may finally get a high-level study panel of oceanographers recommended chance to do their work in the environment enjoyed by to the navy in 1991 that magnetic and seismic instruments their undersea neighbors. This past summer an SSN-650 for charting the Arctic basin could be towed behind the attack submarine, the Pargo, set off from Groton, Conn., vessel. For the moment, this technology would probably be with five civilian scientists nestled among the more than unacceptable to the navy: submarine commanders assidu- 100 regular crew members. The mission, Submarine Arctic ously avoid any devices that emit substantial amounts of Science Cruise-93, which is to be funded with $3 million radio energy for fear of revealing submarine location and from this year’s defense budget, constitutes the first un- operating characteristics. classified scientific journey on board a nuclear submarine. So the oceanographers could run silent. But can they run On August 11, this Sturgeon-class sub set out on a jour- cheap? A decision about a dedicated research vessel, Lang- ney of more than a month to the Arctic Ocean that would seth says, should come before the planned decommission- take the researchers to the North Pole. Almost half the ing of the entire Sturgeon fleet. Langseth and other scien- mission, 3,300 nautical miles, was to be below the ice. tists have their work cut out: the $8 million to $15 million The “hunter-killer” submarine is most likely still equipped needed to keep a submarine in operation annually could with a cargo of Harpoon antiship missiles and Mk 48 torpe- make a submarine into the oceanographic equivalent of the does in the unlikely event that hostilities break out. But this Superconducting Super Collider. “The big problem is that expedition could mark a step to- the navy doesn’t want to pay for ward perhaps the ultimate de- it, and the scientific communi- fense-conversion project: an at- ty doesn’t want to pay,” remarks tempt to deploy a nuclear sub- George B. Newton, a former marine for the sole purpose of submarine commander who scientific research. helped to organize the mission. Since the first nuclear-powered Are the Russians coming? Our submarine was launched in the former adversaries may ignite 1950s, scientists have recog- a fare war for scientist passen- nized that these vessels might gers. They have twice offered be harnessed to compile a com- to convert one of their nuclear prehensive profile of physical submarines into a research ve- changes in the ice cap, an espe- hicle for the West. cially important task in tracking Hitching a ride on a Russian the possible warming trend in submarine, even as part of a the earth’s climate. “It would be U.S. foreign aid package, may difficult to get this information prove a hard sell at a time any other way,” says Marcus G. when U.S. defense companies Langseth, a scientist at Colum- are laying off workers. Indeed, bia University’s Lamont-Doherty there may be a way to get at Earth Observatory and chairman least some of the information of the science steering commit- with little cost. If Congress ul- tee for this program. timately decides that a multi- The current mission was to million-dollar passage on board Courtesy of U.S. Navy use a type of sonar that can a submarine is an unneed- look upward to judge the vol- WIEN ed joyride, the U.S. research

ume of the ice cap. Meanwhile Y GER community might pressure the a synthetic-aperture radar on a navy to speed up the declas- satellite orbiting over the Arctic HARR sification of Arctic data that would provide topside images STURGEON-CLASS SUB, like this one, was to have they have been gathering for of the same area. The subma- surfaced in the Arctic with ﬁve scientists on board. years. —Gary Stix

26 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. City. Palese and others have been seeking a more elusive quarry: the virus that caused the massive 1918 inﬂuenza pandemic. Such work is particularly challenging because the genes are encoded in RNA, which is less stable than DNA. Virologists want to know more about the 1918 strain because another such pandemic can, in principle, recur. Until recently, workers had to proceed indirectly, inferring the viral structure from antibodies produced by survivors of the infection or extrapolating from the genomes of more recent viral samples. Walter M. Fitch of the University of Cal- ifornia at Irvine has done so by charting the evolutionary trajectories of in- ﬂuenza in humans and pigs. He found that the strains indeed diverged from a

SABA common ancestor around 1918.

ARR Although Fitch believes he has reconstructed the genome of the common ancestor, he says it requires correction STEVE ST by direct sequencing of the 1918 virus. BUGS INSIDE BUGS: Raúl J. Cano and an amber-clad, bacteria-laden insect. While Palese says his attempts to isolate the virus from human remains have been unsuccessful, Robert G. Webster Jurassic Virus? but contend that the entire genome of a of St. Jude Children’s Research Hospi- bacterium could be built in a few years. tal in Memphis says he is sequencing Can’t clone a Tyrannosaur? Yet bringing an ancient bacterium the 1918 virus. He refuses to discuss Then try chicken pox back to life might not be so simple. Mi- the results, however. chael A. Goldman of San Francisco State Another logical RNA target is the HIV University points out that bacterial retrovirus, which causes AIDS. “You can olecular archaeologists immedi- DNA forms loops having quasichromo- go into museum collections and select ately cried foul when they saw somal status—that is, it encodes infor- human tissues preserved in the 19th Mthe visions of dinosaurs cloned mation that lies outside the sequence century, when they used spirit rather from the last meals of amber-entombed of nucleic acid bases. In addition, bac- than formaldehyde,” Morse says. “We mosquitoes in the film version of Mi- teria bequeath proteins as well as nu- know that some diseases killed quickly chael Crichton’s Jurassic Park. Ancient cleic acids to their daughter cells, and in those days. Galloping consumption DNA is too damaged, they argued; it these proteins may influence the ex- was not unlike the presentation of tu- would take centuries to piece it togeth- pression of certain genes. “Already that berculosis in AIDS patients today. May- er. Moreover, the reconstructed genome makes the task di›erent from just mak- be these were individuals with AIDS would be riddled with errors. In short, ing DNA and throwing it into a host who died of secondary infection.” they concluded, extinct organisms must cell’s genome,” he argues. So far, however, the only viral RNA remain so, forever. But Poinar and Cano retort that they to come from old human remains are Not necessarily, say George O. Poinar, will simply shoot at a smaller target. endogenous retroviral genes, so called Jr., and Raúl J. Cano, the entomologist- “We hope to be looking into di›erent because they became integrated in hu- microbiologist team who hold the record kinds of viruses pretty soon,” says Poi- man chromosomes long before Homo for sequencing the most ancient DNA— nar, who is at the University of Califor- sapiens evolved. In a paper that is to that of a Jurassic wood beetle embedded nia at Berkeley. The smallest ones have appear in Biological Anthropology and in amber more than 120 million years genomes no larger than a fair-sized gene the Study of Ancient Egypt, Jaap Goud- ago. They believe it might be possible to and can be made to replicate without smit and his colleagues at the Academic resurrect organisms, albeit ones much helper proteins. One merely inserts the Medical Center in Amsterdam report simpler than Tyrannosaurus rex—a bac- naked viral DNA into a host cell, which that an Egyptian mummy more than terium, say, or a virus. “We’re very close then produces complete viral particles. 5,000 years old has yielded traces of to reconstructing a gene of a bacillus spe- Experts in viral evolution generally agree human endogenous retrovirus type C. cies that lived in the gut of a stingless bee that this procedure can be done. “Juras- More recent studies on mummified 40 million years ago,” says Cano, who is sic virus? Sure,” says Stephen S. Morse monkeys produced similar results. at California Polytechnic State Univer- of the Rockefeller University. Microbiologists do not seem overly sity in San Luis Obispo. “You’d put the The problem, of course, will be isolat- worried about disturbing the sleep of genes into a living species bit by bit, un- ing an ancient virus. No virus fossils are undead germs. Most microbes are innoc- til nothing was left of the modern DNA.” known, although the new techniques of uous, and most laboratories are designed The bacterial gene, known as 16S enzymatic amplification may yet wring to contain pathogens. But the risks are rRNA, helps make ribosomes. So far, the viral genes from ancient remains. “The- not zero. “What was a minor microbe in researchers say, they have strung to- oretically, resurrecting a virus would be the time of the mummies might be lethal gether 1,300 of a possible 1,500 or so possible, but it may be hard to get good today,” Goldman notes. “I would keep in nucleotide bases. They concede that enough DNA,” says Peter M. Palese of mind Michael Crichton’s earlier novel, some genes are harder to reconstruct Mount Sinai Medical Center in New York The Andromeda Strain.”—Philip E. Ross

Revealing the Hidden Sequence his college career. An enthusiastic professor, Ernest Baldwin, enticed him to study the new science of biochemistry. ne might expect a two-time No- His grandfather made a fortune in the Sanger enjoyed the subject so thorough- bel Prize winner to spend his cotton trade and passed it on to his ly that he took advanced courses during Odays raising funds for a world- mother, Cicely. As a boy, Frederick was a fourth, extra year at Cambridge. Two class laboratory, giving lectures to ador- fascinated by nature, collecting every- weeks after the final exams, he says, he ing colleagues and collecting royalties thing from rocks to insects. His early was “very surprised to learn that he had on his best-selling novel. Yet Frederick ambition was to become a doctor, like been awarded a first-class degree.” Sanger seeks neither fame nor fortune. his father, Frederick senior. Yet although In 1940, unlike most of his peers, the Instead the man who built the founda- the younger Sanger enjoyed learning 22-year-old Sanger did not go o› to tions of modern biochemistry lives qui- about the science of medicine, he was fight in World War II. “I was brought up etly in Swa›ham Bulbeck, England, tend- not interested in the art of diagnosing as a Quaker, and I felt pretty strongly ing a garden of da›odils, plum trees and and treating disease. that people should not go around kill- herbs. “I think Sanger hasn’t ing others for any reason,” been recognized as much as he explains. He successfully some, partly because he is defended his position before an undemonstrative person,” a military tribunal and spent says Alan R. Coulson, who the war years pursuing a doc- collaborated with Sanger for torate. Graduate students 16 years at the MRC Labora- were obviously in short sup- tory of Molecular Biology in ply during the war, and the Cambridge. biochemistry department at Forty years ago, Sanger Cambridge was all too hap- was the first to reveal the py to accept a promising complete structure of a pro- student. tein. Then during the 1970s After earning his Ph.D. in he developed one of the first 1943, Sanger joined the lab- techniques for reading the oratory of Cambridge pro- genetic code. “He deserved fessor A. C. Chibnall, who two Nobel Prizes,” says bio- was a pioneer in the field of chemist G. Nigel Godson of protein chemistry. Chibnall New York University Medi- asked Sanger to study insu- cal Center. “He single-hand- lin, the pancreatic hormone edly engineered two revo- that governs the metabolism lutions in biology.” In addi- of sugar. The suggestion led tion, Coulson boasts, Sanger to a 10-year-long project that deserves much of the credit established Sanger as the for laying the groundwork leader of his field. Insulin, for the Human Genome Proj- like other proteins, is made

ect, the multinational e›ort Black Star up of di›erent amino acids. to determine the entire se- During the 1940s research- quence of nucleotides in hu- ers were very aware that the man DNA. chemistry of proteins de- Sanger, now 75, chose to pends on the order in which meet me at one of his favor- LEVENSON DAVID the amino acids are arranged. ite pubs, the Red Lion, near FREDERICK SANGER transformed biochemistry by developing The problem: no one had Cambridge. He has mastered methods for determining the structure of proteins and DNA. found a technique for deduc- the art of understatement ing the sequence. in both appearance and ac- Chibnall and Sanger chose tion—a no-frills kind of guy. Wearing By all accounts, Sanger was not a bril- to work on insulin for several reasons. a red sweater over a plaid shirt, he liant student. In 1936 he was accepted It was available in a pure form, and it orders a chicken sandwich and a half- to the University of Cambridge, but he was a small molecule, at least as pro- pint of lager. He speaks in a shy voice struggled with the basic sciences. Sang- teins go. More important, perhaps, in- that is barely audible above the noise of er received passing grades in chemis- vestigators realized that if they could the restaurant. Sanger is uncomfortable try, but he “bombed” in physics. “I nev- work out the structure of insulin, they talking to journalists, but in the relaxed er won scholarships,” he notes. “I am would understand how it controlled atmosphere of the pub he reveals some not sure I would have been able to at- sugar metabolism—insights that had of what made him one of the great sci- tend Cambridge if my parents had not many implications for medicine. entists of this century. been fairly rich.” To determine the sequence of insu- Sanger was born into a wealthy home. Sanger found his calling early in lin, Sanger began with a simple strate-

30 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. gy that chemists often use to analyze Sanger realized that his techniques for with many similar pieces is more di¤- large compounds: trying to break the cutting the insulin molecule were in- cult to solve than one with many di›er- molecule into fragments and then fig- troducing new bridges between the ent pieces is, the sequence of DNA was ure out how the pieces fit together. Eas- strands. By 1955 he found a way to pre- more di¤cult to decode than the se- ier said than done. To snip insulin into vent the introduction of cross-links and quence of insulin was. pieces of a meaningful size and then succeeded in determining the complete For more than 10 years, Sanger inves- sort them, Sanger tried every trick in structure of insulin. tigated techniques for sequencing DNA, the chemist’s handbook and then came Four years later, Sanger won the No- competing with several laboratories up with some new schemes. bel Prize in Chemistry for the insu- around the world. Then in 1975, Sang- One of Sanger’s important innova- lin work. He was immediately besieged er and Walter Gilbert of Harvard Uni- tions was a method for labeling an end by professors inviting him to teach versity, working independently of one of a protein fragment. These tags made another, developed methods for rapid- it easier to deduce what pieces belonged ly sequencing DNA. Their techniques together. For example, if three amino ac- Sanger’s talent for solving could determine the arrangement of ids—call them A, B and C—are linked nucleotides in segments of DNA 200 or together in some order in a chain, the the right problem at the more units long in a few days. With ear- sequence could be determined in the right time earned him two lier methods, the job would have taken following way. After labeling the chains, Nobel prizes in chemistry. years. The new technique enabled Sang- they could be broken down into individ- er and his collaborators to determine ual amino acids, indicating, say, that B the sequence of 5,375 nucleotides in a was at one end of the chain. Then a new and administrators asking him for ad- virus called ␾X174. sample of the same chains could be vice. But Sanger wanted no part of it. “I In 1980 Sanger won a second No- cut into pieces consisting of two amino have actively tried to avoid both teach- bel Prize, which he shared with Gilbert acids. The final step was isolating all ing and administrative work,” he says. and Paul Berg of Stanford Universi- those pieces that had a B at one end. If “This was partly because I thought I ty. Berg had found a way to insert piec- all the B’s were linked to A’s, then the would be no good at them but also out es of DNA from one organism into the sequence would have to be BAC. of selfishness.” DNA of another. His work launched the While pursuing the insulin sequence, Ironically, in the year Sanger became a technology of recombinant DNA. Sanger tried hundreds of di›erent tech- Nobel laureate, his research began to During his career, Sanger published niques. “Most experiments go wrong,” founder. He had taken the insulin study about one major scientific paper every Sanger sighs. “I didn’t spend too much as far as it could go and was looking for eight years, but his colleagues say each time trying to figure out what went new questions to grapple with. “I think one is a classic in experimental bio- wrong. I just started thinking about the these periods occur in most people’s chemistry. “I don’t publish papers unless next experiment. It prevented me from research careers and can be depressing I have something to write about and getting depressed.” and sometimes lead to disillusion,” he something I am sure about,” he explains. There was one experiment in par- observes. “I have found that the best These days, although Sanger still gives ticular that Sanger would like to for- antidote is to keep looking ahead.” advice to his colleagues at the Laborato- get. In 1947 he spent a year in Uppsa- In 1961 Sanger joined the Laboratory ry of Molecular Biology, he has retired la, Sweden, working in the laboratory of Molecular Biology at the Medical Re- from his research. “The aging process of the eminent biochemist Arne W. K. search Council in Cambridge and hired was not improving my performance in Tiselius. A technique Sanger developed Coulson as his research assistant. “He’s the laboratory,” he complains. “I would seemed to suggest that insulin was not not the kind of guy you pal around with have felt guilty about occupying a space a single chain but rather four cross- right away,” Coulson says. that could have been available to a youn- linked chains. When he presented the Soon after his arrival at the labora- ger person.” results to his mentor, Tiselius suggest- tory, Sanger decided his talents might Reflecting on his research, Sanger ed that they rush to publish a paper to- be useful for analyzing DNA, the mole- finds it hard to recall any moments of gether. “I was rather shocked as he had cule that stores the genetic code. In the great inspiration. He does not subscribe not really contributed anything.” 1950s James Watson and Francis Crick to the “popular idea that scientific prog- But as a junior member of the labo- had figured out that DNA was a long, ress depends on sudden breakthroughs.” ratory, Sanger gave in. “The paper is the double helix made of four di›erent nu- Instead he thinks about many events only one of which I am ashamed,” he cleotides. The arrangement of nucleo- that “were more often associated with comments. Sanger later discovered that tides determines what proteins an or- small and gradual advances.” insulin is actually made of two cross- ganism can make, that is, what genes it Sanger finishes his lunch and invites linked chains, one 30 amino acids long will express. Yet at the time, scientists me to his home. We drive through the and the other 21 long. The larger chain could determine the sequence of nu- countryside to an unpretentious house was by far the most complex structure cleotides for only a very small section surrounded by an elaborate garden. It that a protein chemist of that era had of a DNA strand. seems Sanger is as devoted to garden- ever struggled with. It was not until Sanger set an ambitious goal for him- ing as he once was to research. He and 1952 that Sanger and his co-workers self: sequencing the many thousands his wife of 52 years, Joan, grow a doz- Hans Tuppy and E.O.P. Thompson fig- of nucleotides in the DNA of a virus. Yet en varieties of flowers, cultivate sever- ured out the complete sequence of he and other biochemists soon found al fruit trees and raise chickens. When both chains. DNA more di¤cult to analyze than pro- they are not working in the garden, Sanger then had to decide how the teins for two reasons. First, they had the Sangers entertain their three chil- chains linked together to make an insu- less experience handling DNA than they dren, grandchildren and friends. “I have lin molecule. The problem turned out did proteins. Second, DNA is made from always had a happy and peaceful life,” to be extremely complicated. An initial only four basic building blocks, whereas he remarks. The key, he asserts, “is analysis seemed to show that chains proteins are constructed from some 20 working on the right thing at the right were linked at nearly every point. Then di›erent amino acids. Just as a puzzle time.” —Russell Ruthen

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 31 Clearing the Air in Los Angeles Although Los Angeles has the most polluted skies in the nation, it is one of the few cities where air quality has improved in recent decades

by James M. Lents and William J. Kelly

n some hot, sunny days, the 14 able accomplishment of the citizens of million residents of the Los An- southern California. Los Angeles is one Ogeles area inhale a thick, brown- of the few places in the nation where ish-gray haze, and no one can ignore air quality has improved dramatically its e›ect. The smog obscures the San since the 1970s. From 1955 to 1992 the Bernardino Mountains and the warm peak level of ozone—one of the best California sun; it irritates the eyes and indicators of air pollution—declined nose; it restricts the activities of ath- from 680 parts per billion to 300 parts letes and people who have breathing per billion. The California Air Resources disorders; it injures the lungs of both Board recently documented that popula- young and old. tion exposure to unhealthful ozone lev- Southern California’s air quality is els has been cut in half in just the past the worst in the U.S. Air pollution in decade. Furthermore, the smog levels the region reaches unhealthful levels measured during each of the past three on half the days each year, and it vio- years have been the lowest on record. lates four of the six federal standards All these improvements were achieved for healthful air—those for ozone, ﬁne at a time when human activity in the Los particulates, carbon monoxide and ni- Angeles area was increasing at a rapid trogen dioxide. In 1991 the South Coast rate. Since the 1950s the population Air Basin exceeded one or more federal has almost tripled, from 4.8 million to health standards on 184 days. 14 million; the number of motor vehi- Yet these statistics hide a remark- cles on the road has more than quad- rupled, from 2.3 million to 10.6 million; and the city has grown into one of the most prosperous regions of the world. JAMES M. LENTS and WILLIAM J. KEL- Although the residents of southern LY work together at the South Coast Air Quality Management District (AQMD), the regional air-pollution control agency for the Greater Los Angeles area. Before AIR POLLUTION SOURCES have in- becoming executive o¤cer of AQMD in 1986, Lents headed the air-pollution con- creased in size and number in Los An- trol program for the state of Colorado. geles, yet technical innovation and so- In 1970 he received a Ph.D. in physics cial policy have led to an improvement from the Space Institute at the Universi- in air quality during the past two de- ty of Tennessee. Some 11 years ago Kel- cades. Some typical sources of air pol- ly earned an M.A. in journalism from Co- lution include ( from left to right) indus- lumbia University, and since then he has trial coatings, barbecues, trash inciner- written extensively on the environment. ators, paints, dry cleaners, commercial ovens and motor vehicles.

32 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. California still face and continue to tack- that the smog problem be solved. The lutants in the cool air near the ground le many air pollution problems, they paper also put its money behind its ed- where people live and breathe. The have an advantage in that they have itorial mouth. In 1947 it retained Ray- mountains that surround the region spent 50 years studying the local at- mond R. Tucker, the former smoke reg- compound the problem; they prevent mosphere and experimenting with var- ulation commissioner of St. Louis, to the pollutants from dispersing. ious policies. We hope the cities of all study air pollution in the area. Tucker SRI pointed out that natural materials nations will learn from the experiences identified and investigated several ma- such as dust, pollen, fibers and salt were of Los Angeles [see “The Changing At- jor sources of air pollution, including important components of the haze. But mosphere,” by Thomas E. Graedel and heavy industries, foundries, motor vehi- the institute also recognized that in- Paul J. Crutzen; SCIENTIFIC AMERICAN, cles, backyard incinerators and smudge dustries and motor vehicles contributed September 1989]. pots for protecting crops from frost. to the problem by adding carbon parti- In the same year, the oil industry cles, metallic dust, oil droplets and wa- he movement to clean up the air paid the Stanford Research Institute ter vapor. in southern California began dur- (SRI) to give another perspective on the In the 1950s SRI and Arie J. Haagen- T ing the 1940s, a period of rapid causes and control of pollution. The or- Smit and his colleagues at the Califor- industrialization. At the time, the re- ganization discovered that the hazy nia Institute of Technology began to ex- gion was plagued by sudden “gas at- days were caused in part by a natural amine the chemistry of the atmosphere tacks” that irritated the eyes, dimin- weather phenomenon known as an in- above Los Angeles. Their work and the ished visibility and produced an un- version layer. The warmest part of the research of others have revealed the pleasant odor. Then, as now, the smog atmosphere is, more often than not, complexity of atmospheric chemistry. was so obvious and odious to the pub- that nearest to the ground, but under Automobiles, factories and other sourc- lic that elected leaders were compelled certain conditions a layer of cool air can es release such raw pollutants as hy- to take meaningful action. Yet their ef- slip underneath a stratum of warm air. drocarbons, water vapor, carbon mon- forts provoked strong conflict. Some Such inversions often form o› the coast oxide and heavy metals. When these citizens and industries fiercely resisted of Los Angeles as the Pacific Ocean chemicals are exposed to intense sun- suggestions to clean up sources of pol- cools the atmosphere just above it. Af- shine, they react to yield a vast number lution. But the Los Angeles Times pub- ter ocean breezes blow the air mass in- of secondary pollutants—for instance, lished dozens of editorials demanding land, the inversion layer traps air pol- ozone, nitrogen dioxide, various organ-

Copyright 1993 Scientific American, Inc. ic compounds and acidic particles of ni- does not respect political boundaries. It argued that many industries were trate and sulfate. This concoction then In 1975 the regional governments as clean as they could get, the state interacts with plants and animals, caus- tried voluntarily to consolidate their was not doing enough to clean up cars ing a variety of di›erent e›ects. Many pollution-control programs. Two years and the region might never achieve of these phenomena are still not un- later little progress had been made. So clean air standards. As required by law, derstood, but 50 years ago even less in- the California legislature forged an un- the AQMD adopted air-quality manage- formation was available. easy alliance between the local pro- ment plans in 1979 and 1982, but these In 1953, with the public fearing that grams by creating the South Coast Air were regarded as mostly paper exercis- the Los Angeles haze might become as Quality Management District (AQMD). es. At the time, the federal Clean Air Act bad as London’s “killer” fog, Governor The AQMD was given jurisdiction over required all American cities to achieve Goodwin J. Knight appointed an air- the counties of Los Angeles, Orange and federal standards by 1987, although pollution review committee. Chaired by Riverside and part of San Bernardino— nearly everyone realized that the task Arnold O. Beckman of Beckman Instru- an area of 13,350 square miles. would be impossible in Los Angeles. As ments, the committee proposed ﬁve key Initially, the AQMD was responsible that deadline drew near, however, en- ideas for reducing pollution over the for stationary sources of air pollution, vironmentalists and even some busi- short term. First, they asked that the and the California Air Resources Board ness groups attacked the AQMD for emission of hydrocarbons be reduced was assigned to regulate mobile sourc- its complacency and alleged lax en- by improving procedures for transfer- es such as cars, trucks and buses. In its forcement. In 1987 the legislature re- ring petroleum products. Second, they early years the AQMD adopted a view- structured the AQMD governing board set standards for automobile exhausts. point held by many business leaders. and granted broad powers to the new Third, they encouraged the use of trucks and buses that burned liqueﬁed petroleum gas instead of diesel fuels. Fourth, they considered whether industries that polluted the area heavily should be asked to slow their growth. Fifth, they advocated that the open burning of trash be banned. The committee also hoped that over the long term Los Angeles would devel- SAN GABRIEL MOUNTAINS op a sustained automotive pollution- control program, construct a rapid transit system and start a cooperative program to regulate industrial sources of pollution. Ironically, the report was is- sued while the region’s public train system was being dismantled. Today Los WARM AIR Angeles is trying to get back on track, so to speak, by developing an extensive regional commuter rail network. The Beckman committee’s recommenda- tions eventually grew into a coherent air-quality management plan for the region, but it emerged slowly and was reshaped many times. Soon after the Beckman report, the COOL AIR neighboring counties of Orange, Riv- erside and San Bernardino began their OTHER own pollution-control programs. During POLLUTION the past 40 years, these regions have SOURCES experienced explosive growth in popula- (RESIDENTIAL, COMMERCIAL) tion and in vehicular tra¤c. The control programs in these regions were, for the most part, as energetic and innovative as those in Los Angeles. But residents of these counties soon realized that they needed to coordinate their e›orts; smog

WEATHER PATTERN known as an inversion layer traps air pollution above P A CIFIC OC Los Angeles and neighboring counties. Air cools over the ocean and then is OCEAN BREEZE blown inland, creating a cold layer near the ground and a warmer layer above. The warm layer prevents most of the smog from escaping upward. The sur- rounding mountains keep the polluted air from moving farther inland.

34 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. board. The AQMD now has responsi- than 300,000 families owned such in- the local groundwater. Southern Califor- bility for achieving local, state and fed- cinerators, and many were unwilling nians are now working to clean up the eral standards. to give up the convenience and cost groundwater, but the long-term plans Despite all the political upheaval, savings. But slowly attitudes changed; are, first, to decrease the volume of the state and local governments had today most residents of Los Angeles trash through recycling and, second, to managed to curb many sources of air would consider it shameful to burn reduce further groundwater contami- pollution in the South Coast Air Basin their trash and subject their neighbors nation by improving the management before the AQMD was formed in 1977. to the smoke and smell. of landfills. They gradually developed regulations Waste-disposal managers objected to The ban of backyard incinerators that concentrated on reducing the ma- the ban for a di›erent reason. These an- generated relatively little public resis- jor sources of air pollution: particles alysts realized that the economical al- tance when compared with attempts to from trash incineration, emissions from ternative to burning trash was burying clean up industrial sources of pollu- industry and pollutants from motor it, and therefore they correctly predict- tion. Historically, the control of indus- vehicles. ed that although replacing incinerators trial emissions has been challenging with landfills would reduce air pollu- because of the need to balance the re- n 1958, at the recommendation of tion, the additional landfills would cre- gion’s environmental interests with its the Beckman committee, backyard ate other types of environmental prob- economic needs. The Beckman report of I trash incinerators were banned de- lems. For example, as rain has seeped 1953 quickly led to requirements that spite opposition from the public and through the landfills and carried away industries use vapor-recovery equip- some waste-disposal managers. More soluble materials, it has contaminated ment when they transferred petroleum products, but service stations were not obliged to install such equipment on gasoline pumps until 1978. This equip- SAN BERNARDINO MOUNTAINS ment, while originally somewhat di¤- cult for the motorist to handle, has been streamlined, so it is e›ective and easy to use today. Yet the measure is still resisted by most communities outside of California because of its expense and an undeserved reputation for being cumbersome. Regulations were adopted during the 1960s to eliminate industrial solvents that play a major role in promoting the formation of ozone. The rules a›ected a wide variety of businesses, from SANTA ANA MOUNTAINS construction to auto manufacturing to dry cleaning. To meet the requirements, most industries chose to use nonreac- tive solvents instead of installing control equipment because it was the less ON-ROAD VEHICLES expensive solution. Sadly, the nonre- (CARS, BUSES, TRUCKS) active solvents were later shown to destroy ozone in the stratosphere. (At high altitudes, ozone serves the critical function of shielding the earth from harmful solar radiation.) The solution to this problem is to find truly benign solvents; for example, Hughes Aircraft Company has recently developed a sol- dering flux made from citrus juice. Although Californians succeeded in HEAVY INDUSTRY reducing the quantity of pollution generated by certain industrial sources, their e›orts have been partially o›set in recent decades. Many sources have been introduced as the region’s economy has grown. To compensate for the growth factor, o¤cials instituted, in 1976, “new source” regulations. These CEAN OFF-ROAD VEHICLES required expanding industries to use (PLANES, TRAINS, SHIPS) the cleanest technology available. The rules also specified that if a company planned to start a project that increased emissions, it was required to earn a certain number of credits by reducing emissions from another project. The new source regulations have stim-

PEAK OZONE LEVEL (PARTS PER MILLION)* 1970 0.58 1980 0.49 1990 0.33 POPULATION (MILLIONS)* 1970 9.5 1980 10.9 1990 13.6

VEHICLES (MILLIONS)* 1970 6.4 1980 8.3 1990 10.6 *For Los Angeles, Orange, Riverside and San Bernardino counties

HYDROCARBON EMISSIONS, 1992 NITROGEN OXIDE EMISSIONS, 1992 (TOTAL: 1,375 TONS PER DAY) (TOTAL: 1,208 TONS PER DAY)

ON-ROAD VEHICLES

OFF-ROAD VEHICLES 21% 44% 6%

55% 19% 16% HEAVY INDUSTRY 31% 8% OTHER SOURCES (RESIDENTIAL AND COMMERCIAL)

OZONE BY YEAR OZONE BY REGION 365 350 1980

LOS ANGELES SAN BERNARDINO

300

ORANGE RIVERSIDE 250 PACIFIC OCEAN YS A

200 1991

NUMBER OF D 150

100 < 0.12 PART PER MILLION 0.12–0.19 PPM

0.20–0.35 PPM 50 > 0.35 PPM

0 0 10 20 30 40 50 60 70 80 1977 1982 1987 1992 NUMBER OF DAYS OZONE EXCEEDED 0.20 PART PER MILLION SOURCE: South Coast Air Quality Management District

36 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. ulated considerable innovation. When in southern California, issuing citations the road, the emissions-control equip- researchers compared the best con- to vehicles that smoked excessively. ment was either removed by the own- trol technologies of 1976 with those The crankcase device proved e›ec- er or damaged by the use of leaded fu- available in 1990, they found that the tive on new vehicles, but objections els. The latter problem has abated in amount of pollution generated dur- were raised to equipping old cars with recent years as the price of unleaded ing the manufacture of a product had them. Many auto mechanics were not fuels has moved closer to that of lead- dropped, on average, by 80 percent. Un- trained to install the devices, and ru- ed gasoline. fortunately, many businesses found a mors spread that even properly in- To enforce legislation on automotive way to avoid the regulations. The rules stalled devices caused engine damage. emissions, Los Angeles initiated an in- exempted new projects that would emit That misinformation was the primary spection program in 1976. The regula- daily less than 75 pounds of hydrocar- reason for the state legislature’s deci- tions specified that whenever a vehicle bons or 100 pounds of nitrogen oxides. sion in 1965 to stop requiring the in- was sold, the new owner was required to To exploit this loophole, many indus- stallation of crankcase devices in cars bring the car to a special station where tries expanded by adding several small made before 1963. It was not the first the emissions could be measured with projects instead of a few large ones so time, nor certainly the last, that inaccu- the engine of the car running at normal that all the undertakings would be ex- rate information has become the basis cruising speeds. Los Angeles had many empt. Furthermore, most small busi- of political compromise. di¤culties managing the program, and nesses were exempt, even though they In addition to installation of crank- the state government was reluctant to have accounted for the fastest-growing case devices, Los Angeles tried to re- help. Consequently, the city received in- segment of stationary source emissions duce automotive pollution by encour- numerable complaints from car owners during the past few decades. aging the production and use of clean who were forced to wait in long lines to Growth of these small sources, com- gasolines. In 1960 o¤cials required that get their vehicles tested. bined with increased population and all gasoline sold in the area have a low The California legislature was eventu- motor vehicle use, tended to o›set content of olefins, which are some of ally persuaded to develop a new inspec- cleanup e›orts through the 1980s. Con- the most reactive compounds in petro- tion program when the Environmental sequently, in 1990 the new source reg- leum products. Unfortunately, as the Protection Agency imposed sanctions ulations were revised to include proj- years passed by, the clean gasoline pro- that reduced highway funding. Under ects that introduced small quantities of gram lost its momentum. Then, in the the new “smog check” program, each air pollution. Even then, some exemp- early 1980s, the state government of car in California is tested once every two tions were allowed, and they continue Colorado revitalized the concept, hop- years at a repair shop. Furthermore, the to hurt the e›ort. ing to reduce air pollution problems in state gave local governments the option Denver. Colorado required the blending to conduct tests either while the engine os Angeles’s di¤culties with reduc- of a methanol derivative into gasoline. was idling or while it was running at ing industrial pollutants were sim- In California, meanwhile, the AQMD normal speeds—the latter option being Lilar in one respect to its attempts and the state were pushing for metha- more expensive but more e›ective. The at diminishing automotive emissions. nol and other alternative fuels through program achieved pollution reductions, Although o¤cials identified the prob- regulatory programs. Faced with losing but the results were lower than expect- lem and took action, they were slow to large market shares to methanol and ed. Today the federal government is adjust to the scale of the matter. From other clean fuels, the Atlantic Richfield again pressuring California to create a the start, for example, the Beckman Company in Los Angeles came up with program that lives up to its potential. committee recommended controls on a constructive solution by developing a automotive emissions, but it did not an- truly clean gasoline. These events stimu- ¤cials had much trouble en- ticipate how quickly auto travel would lated California and the federal govern- couraging southern Californi- grow throughout the region. Neverthe- ment to go beyond regulations and in- Oans to share rides, thereby re- less, Los Angeles has succeeded in three troduce new legislation for clean fuels. ducing congestion and emissions. One important areas: promoting the use of Another development in reducing way to promote ride sharing is to re- clean fuels, purifying engine exhaust automotive emissions was the remov- serve lanes on major highways for the and encouraging carpooling and the al of pollutants from the exhausts of use of cars transporting two or more use of public transportation. car engines through the use of catalytic people. When state transportation ad- A significant first step in regulat- converters. California required that be- ministrators proposed restricting one ing automotive pollution was the reduc- ginning with the 1975 model year, all of four lanes of the Santa Monica Free- tion of “blowby” gases. Most car engines new cars have the converters. The conway, however, motorists opposed the that were designed 30 years ago allowed verters and similar devices substan- idea so strenuously that the o¤cials the gaseous by-products of combustion tially reduced emissions, but for sev- quickly withdrew it. Later they found to escape past the piston, through the eral reasons the program was not as that the public was less antagonistic if crankcase and into the atmosphere. In successful as it might have been. First, new lanes were added to existing high- 1960 such blowby gases accounted for researchers discovered that the per- ways for the use of high-occupancy ve- one quarter of the total hydrocarbon formance of a converter decreases as hicles. California now has an extensive emissions in the region. The gases could the car accumulates mileage. Second, network of such lanes. be eliminated by installing a $5 device many motorists tampered with or even In 1987 the AQMD required firms that routed the crankcase emissions to removed the emissions-control equip- with 100 or more employees to o›er the engine intake manifold; in this way, ment. Third, many consumers damaged incentives to institute carpooling. Al- the hydrocarbon fumes were burned their converters by fueling their cars though the regulation generated com- instead of being released. In 1963 the with cheap leaded gasolines instead of plaints, it has been successful. The rule state mandated that the devices be add- the more expensive unleaded varieties. applies to 5,200 work sites, harboring ed to all motor vehicles, new and old. By Indeed, in 1983 the National Enforce- 1.2 million employees. A study of that the middle of the decade, a special po- ment Investigations Center found that group from 1987 to 1992 revealed that lice force was patrolling the roadways in almost one fifth of all the vehicles on the number of employees per vehicle

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 37 rose from 1.13 to 1.24. The program gen oxides and 30 percent of the sulfur and a 74 percent greater risk of asth- has eliminated 90,000 trips a day and oxides and smaller percentages of oth- ma. In addition, women living in those has achieved about half the desired re- er pollutants. The remainder of the pol- areas had a 37 percent greater risk of ductions in emissions. In fact, southern lutants are emitted by households and developing some form of cancer. California is one of the few places in the service-oriented businesses, such as res- Roger Detels of the University of Cal- U.S. where ride sharing has increased taurants, dry cleaners, gas stations and ifornia at Los Angeles studied respira- over the past five years. operations in commercial buildings. tory disease among residents in three Today the motor vehicles and indus- Los Angeles must reduce pollution areas of southern California that di›er tries operating in California are among even further to meet federal health in air quality. During a five-year period, the cleanest in the world. A new car sold standards. Computer projections indi- he periodically questioned and exam- in California emits just one tenth of the cate that southern Californians must ined volunteers, aged seven to 59 years, pollution that a new car did in 1970. cut hydrocarbons by 80 percent, nitro- to determine the health of their respi- Such industries as electric utilities rely gen oxides by 70 percent, sulfur oxides ratory systems. In 1987 Detels discov- almost exclusively on clean-burning by 62 percent and particulates by 20 ered that residents who lived in high- natural gas. Manufacturing plants and percent. The consequences of compla- pollution areas had more symptoms of construction companies use advanced cency could be disastrous. respiratory disease, such as bronchitis paints, solvents and adhesives that have and asthma, than did people who resid- been formulated to minimize pollution. ir pollution in the South Coast Air ed in low-pollution regions. For these reasons and others, southern Basin has already taken a stag- If the citizens of the South Coast Air California has made tremendous prog- A gering toll on residents. A grow- Basin succeeded in meeting federal ress in reducing air pollution. ing body of evidence reveals that the standards for ozone and particulates, a But despite all the successes, pollu- smog is a serious health hazard. 1989 study by the AQMD predicted, tion in the South Coast Air Basin is still In 1991 David Abbey, an epidemiolo- they would gain $9.4 billion in health overwhelming. Average daily emissions gist at Loma Linda University, found a benefits every year. If the residents total 1,375 tons of hydrocarbons, 1,208 correlation between long-term expo- could reduce just particulates to the fed- tons of nitrogen oxides, 4,987 tons of sure to air pollution and the develop- eral standard, they would prevent 1,600 carbon monoxide, 134 tons of sulfur ment of chronic diseases. Abbey stud- premature deaths annually among those oxides and 1,075 tons of particulates. ied 6,340 Seventh-Day Adventists, 62 who su›er from chronic respiratory dis- Transportation, including cars, airplanes, percent of whom lived in the basin (the ease. In addition, they would eliminate trains and ships, adds 47 percent of the remainder lived throughout California). 15 million person-days on which people hydrocarbons, 70 percent of the nitro- In that group of 6,340, those who resid- with respiratory disease are unable to gen oxides, 90 percent of the carbon ed in areas that exceeded government go to work, school and other activities, monoxide, 60 percent of the sulfur ox- standards for suspended particles on and they would reduce the risk of dying ides and 89 percent of the particulates. 42 days or more per year had a higher prematurely from exposure to particu- Industry contributes 26 percent of the risk of respiratory disease, including a lates. (The risk of death is about one in hydrocarbons, 18 percent of the nitro- 33 percent greater risk of bronchitis 10,000—about half the risk a Californi- an faces of dying in a car accident.) If Los Angeles residents were able to achieve the federal standard for ozone, they would eliminate annually 18 million person-days of restricted activity, 65 million person-days of chest discom- fort, 100 million person-days of head- ache, 120 million person-days of cough- ing, 180 million person-days of sore throats and 190 million person-days of eye irritation. These terrible statistics have captured the attention of the federal and state governments. The federal Clean Air Act, which was amended in 1990, gives Los Angeles until 2010 to achieve federal health standards, but the law also requires the region to make incremen- tal progress toward healthful air. The California Clean Air Act demands that the region reduce emissions by 5 percent a year until health standards have been met. Between 1989 and 1991 the AQMD devised an air-quality management plan to respond to the strong mandates of the federal and state law. During the next 17 years, the plan seeks to reduce pollution from virtually all sources and fos- MESSENGERS for the Rapid Blueprint Company in Los Angeles were outfitted ter the development of new and clean- with gas masks in the fall of 1955 so that they would not su›er from the e›ects of er technologies. The AQMD will execute the smog. Shortly thereafter, local o¤cials began air-pollution control programs. the plan with the cooperation of the

38 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. EPA, local governments and state Air Re- sources Board. Each organization has a well-deﬁned role. The AQMD is charged with cleaning up stationary sources and encouraging carpooling. The EPA is to set standards for airplanes, trains and ships that travel through the region. Local governments will work to alle- viate tra¤c through improvements in the transportation infrastructure and expanding mass transit. The Air Re- sources Board is responsible for motor vehicle standards. It has already taken aggressive action that should reduce automotive emissions by 85 percent by the next decade and should increase sales of electric cars.

he AQMD and its sibling agencies plan to attack the pollution T problem in three stages. The first stage, dubbed Tier I, includes 135 mea- L.A. MECHANIC at Green Motorworks, a local manufacturer of electric vehicles, sures that can be accomplished using ex- converts a gas-fueled car to battery power. Such initiatives are receiving strong sup- isting technologies and are to be adopt- port from California state agencies as a way to meet ambitious air-quality standards. ed by 1996. The measures limit and reduce pollutants from such sources as electric utilities, motor vehicles, small ization are alternative-fuel vehicles, in- tion because firms can make a profit by businesses and even backyard barbe- dustrial emissions controls for nitrogen cleaning up the air. cues. The electric companies will be in- oxides, and cleaner paints, solvents and During the past 40 years, the citi- stalling low-polluting burners and cata- coatings. zens of Los Angeles have been very ac- lysts on power plants to reduce nitrogen After the AQMD released its ambi- tive in the fight for clean air, but from oxide emissions. Pollution from back- tious plan to clean up the air, the orga- time to time they have neglected the is- yard barbecues will diminish as house- nization quickly realized it could not sue because they feared that air-quality holds use reformulated charcoal light- simply play the role of tough pollution regulations were incompatible with the er fluid and other products for light- cop and expect companies, especially region’s economic interests. Today, de- ing grills. Automotive emissions will de- small ones, to follow the plan. The re- spite occasional doubts about environ- crease because of tailpipe standards as gion’s businesses, which were caught in mental regulation and legitimate con- well as programs encouraging carpools a recession and faced cuts in defense cerns about the post–cold war economy and use of public transportation. spending, were reluctant to invest re- of Los Angeles, residents and business The second stage, or Tier II, will take sources in pollution-control equipment. people seem to recognize the need to the region into the 21st century. These To sustain progress, a di›erent ap- solve the serious air-pollution problems. measures rely on technologies that have proach was required, especially to clean We hope that by 2010 when our chil- just entered the commercial market. up the growing number of small sourc- dren climb the San Bernardino Moun- The list includes a new house paint, es of air pollution. In 1991 the AQMD tains, they will see not a sea of smog developed by the Glidden Company, outlined a series of reforms, which in- but rather the Pacific Ocean and Cata- that does not release hydrocarbons. An- cluded providing free technical assis- lina Island. We would like them to live other initiative is developing automo- tance and starting a guaranteed loan in a region where technological inno- bile engines that run on methanol, nat- program so that small businesses could vation sustains both the economy and ural gas or other alternative fuels. The purchase pollution-control equipment. the environment. But most important, third stage, Tier III, requires technolo- In the long run the task of getting we want them to reap the benefits of gies that have not been fully developed businesses to reduce pollution will hinge healthy, fresh air. but are likely to be available in the next on technological innovation. One of the decade or so. Researchers are fabricat- AQMD’s most powerful incentives for ing paint coatings that dry under ul- developing these clean technologies is traviolet lamps, without emitting any the Regional Clean Air Incentives Market FURTHER READING significant quantity of pollutants. They Program. Under this program, a busi- ENVIRONMENTAL ETHICS: DUTIES TO AND are also working on fuel cells, such as ness is asked to meet emission stan- VALUES IN THE NATURAL WORLD. Holmes Rolston. Temple University Press, 1989. those used in the space shuttle, which dards not only for individual pieces of ONE EARTH, ONE FUTURE: OUR CHANG- will power a kind of zero-emission elec- equipment but also for its entire op- ING GLOBAL ENVIRONMENT. Cheryl S. Sil- tric vehicle. erations and facilities. Those business- ver and Ruth DeFries. National Acade- To help develop these new technolo- es that cut pollution below the standard my Press, 1990. gies, the AQMD is working with govern- will be granted emission-reduction cred- COSTING THE EARTH: THE CHALLENGE ment agencies and private corporations its. Those companies that generate more FOR GOVERNMENTS, THE OPPORTUNITIES through its technology advancement of- pollution than the standard must buy FOR BUSINESS. Frances Cairncross. Har- fice. To date, the o¤ce has contributed enough credits to make up the di›er- vard Business School Press, 1992. EARTH IN THE BALANCE: ECOLOGY AND some $40 million in seed money to fa- ence. The program gives firms flexibility THE HUMAN SPIRIT. Al Gore. Houghton cilitate almost 250 projects. Among the to choose how to reduce emissions, and Mi‹in, 1992. technologies now entering commercial- it also provides an incentive for innova-

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 39 Large Igneous Provinces These vast ﬁelds of lava record powerful but geologically brief pulses of magmatic activity. Their formation may have triggered signiﬁcant changes in the global environment

by Millard F. Co¤n and Olav Eldholm

ometimes getting there is not half Maritime Earth Sciences in Kiel, Ger- trace element content and mix of atom- the fun of geophysical research. many, had as its purpose the investiga- ic isotopes. And they sometimes break SBearing westward into a rising tion of the geologic structures that form through the middle of the normally Southern Ocean gale, the Australian re- when continents tear apart and an ocean placid lithospheric plates. search vessel Rig Seismic was pounded is born. Confounding the expectations In the 1960s the late J. Tuzo Wilson so hard that its anchor broke loose and of most of their colleagues, the research- of the University of Toronto, followed smashed into the forecastle, causing the ers managed to obtain core samples of by W. Jason Morgan of Princeton Univer- waves breaking over the bow to flood igneous rock nearly one kilometer deep, sity, developed a hypothesis that helped parts of the ship. Progress toward the lying under 300 meters of soft sediment. Kerguelen Plateau, an elevated region Although they lie thousands of kilo- of seafloor just north of Antarctica, was meters apart in disparate geologic set- delayed while the ship’s crew secured tings, we recognized fundamental sim- the anchor and welded the ship to keep ilarities in the two plateaus. Our find- it watertight. Several days later the ex- ings, merged with seismic and drilling pedition, including one of us (Co¤n), data gathered by many other workers, arrived at Kerguelen, deployed instru- have helped demonstrate that the Ker- ments for analyzing the structure of guelen Plateau and the volcanic conti- COLUMBIA the ocean bottom and began collecting nental margin o› Norway belong to a RIVER data. The aim of the voyage was to un- class of huge magmatic features gener- FLOOD derstand the origin and evolution of the ically known as large igneous provinc- BASALTS huge underwater plateau. es. These features cover areas of up to In that same year, 1985, but in the oth- millions of square kilometers, and yet er hemisphere, a floating earth science they seem to have formed quite swiftly laboratory and drilling vessel, JOIDES in geologic terms. Resolution, steamed north through stor- The spasms of eruptive activity as- HAWAII my North Atlantic seas to the Vøring Pla- sociated with such rapid outpourings teau o› the coast of Norway. That ex- of lava may have substantially a›ected pedition, led by one of us (Eldholm) and the chemistry and circulation of the Jörn Thiede of the Research Center for oceans and atmosphere. Some of the resulting environmental shifts may have contributed to mass extinctions, including the one in which the dinosaurs van- MILLARD F. COFFIN and OLAV ELD- ished; in contrast, other changes may HOLM have collaborated for the past have promoted biological diversity and three years on a project to develop a fuller understanding of giant igneous the origin of new species. formations. Co¤n is a research scientist The recognition of large igneous prov- at the Institute of Geophysics at the Uni- inces has forced geophysicists to re- versity of Texas at Austin. He received think their notions regarding the struc- his Ph.D. in geology in 1985 from Co- ture of the earth’s interior. The theory of lumbia University; he met Eldholm while plate tectonics tidily accounts for the completing his doctorate work. Eldholm slow, steady volcanic activity that occurs is a professor of marine geophysics at at mid-ocean ridges (where new ocean- the University of Oslo in Norway, a position he has held since 1981. He complet- ic crust is born) and near subduction ed his doctoral work in geophysics at the zones (where old, dense sections of the University of Bergen in 1976. In 1991 he ocean floor sink back into the earth’s took a sabbatical at the University of Tex- hot interior), but it cannot readily ex- as to coordinate his investigations with plain the abrupt outbursts necessary to Coffin’s; a year later Co¤n took a fellow- create large igneous provinces. Although ship in Norway. Both men have spent the rocks that make up these features considerable time on board research vessels collecting data on submarine igne- generally resemble the composition of LARGE IGNEOUS PROVINCES (red re- ous provinces. the lavas that emerge at mid-ocean gions) appear in widely varied geologic ridges, igneous provinces di›er in their settings around the globe. Dots indicate

42 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. to elucidate the phenomenon. The re- of the earth are built on inference. Even 19th century, when they realized that searchers proposed that the earth’s the deepest boreholes extend only about several far-flung volcanic constructions mantle, the vast zone lying below the 10 kilometers. Direct and indirect stud- actually constitute connected flows of crust but above the core, circulates in ies of large igneous provinces are there- basaltic lava. One of the most spectac- two di›erent modes. The dominant one fore vitally important for learning more ular of these is the Deccan Traps in consists of large-scale convection that about the nature of mantle plumes and west-central India. (Trap is Dutch for nudges plates across the surface and how they may ultimately a›ect condi- “staircase,” a reference to the steplike causes continents to drift. But about one tions on the earth’s surface. appearance of the eroded lavas.) Sim- tenth of the heat now escaping from ilar structures include the Columbia the mantle does so in the form of deep- he most fundamental observation River basalts in the northwestern U.S.; rooted, narrow plumes of warmer than about these provinces is that they the North Atlantic volcanic outpourings average material that rise through the T consist of basalt, a common, iron- of the British Isles, the Faeroe Islands mantle. When it reaches the base of the and magnesium-rich rock. The large ig- and Greenland; and the Karoo basalts lithosphere, a plume decompresses and neous features that appear in the middle in southern Africa, to name just a few. partially melts, producing an upwelling of continents, where geologists can eas- The layers of lava in a flood basalt pile of magma and a long-lasting locus of ily collect samples and determine their up several kilometers thick. Individual volcanic activity known as a hot spot. composition, are therefore referred to flows may contain several thousand cu- Some of that magma may erupt as a as continental flood basalts. At the edg- bic kilometers of rock and may extend tremendous flood of lava. Unlike the es of continents, igneous provinces are hundreds of kilometers. slow, steady drifting of continents and called volcanic passive margins; if they Because of the inaccessibility of un- the spreading of mid-ocean ridges, the are located in the middle of the ocean, derwater igneous provinces, earth sci- surfacing of mantle plumes takes place they are dubbed oceanic plateaus. entists only recently became aware of in an erratic and episodic manner. Geologists recognized the existence their resemblance to continental flood All models of the internal structure of continental flood basalts late in the basalts. In 1981 Karl Hinz of the Feder-

SIBERIAN TRAPS NORTH ATLANTIC VOLCANIC MARGIN

DECCAN TRAPS

CHAGOS- LACCADIVE RIDGE

KAROO ONTONG BASALTS MASCARENE JAVA PLATEAU PLATEAU

RÉUNION

BROKEN RIDGE PLATEAU KERGUELEN PLATEAU

locations where seismic reﬂection studies seem to reveal an- square kilometers. Studies of these huge magmatic forma- cient volcanism along continental margins. The largest igneous tions are providing new insight into the internal workings of province—the Ontong Java Plateau—covers nearly two million the earth and how they may affect conditions at the surface.

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 43 al Institute for Geosciences and Natural tion—the total volumes of lava and in- Undersea volcanic margins and oce- Resources in Hannover, Germany, sug- trusive rock in igneous provinces and anic plateaus similarly have been re- gested, based on his analysis of record- the rate at which they formed—is poorly configured by the forces of change. ings of reflected seismic waves, that known. On land, geologists can directly Drill samples and seismic-reflection the submerged margins of many con- measure the area of flood basalts but images of the igneous province along tinents contain extensive, layered lava can only infer their depth. The crust un- the Norwegian continental margin and flows. Since then, investigators have ob- der continents averages 35 kilometers in of the Kerguelen Plateau demonstrate tained improved seismic data that seem thickness. Only the top parts are acces- that when those formations first erupt- to corroborate Hinz’s conclusion. Igne- sible to exploration. Beneath the oceans ed, they stood above sea level and only ous rocks recovered by scientific drilling the situation is even more di¤cult. Sam- gradually subsided to their present po- projects on the continental margins o› pling just the uppermost sections of the sition in deep water. Co¤n has estimat- Ireland and Norway confirm that the five- to 25-kilometer-thick crust in which ed that the Kerguelen and Broken Ridge structures seen in the seismic images oceanic plateaus reside requires the ser- plateaus, along with most of the other do in fact have volcanic origins. vices of an expensive drilling vessel. oceanic plateaus and ridges in the Indi- Analogous seismic studies of the Ker- The passage of time also disguises an Ocean, formed above sea level and guelen Plateau present convincing evi- the true extent of ancient volcanic struc- remained there for as long as 50 mil- dence that oceanic plateaus likewise con- tures. The older igneous provinces are lion years, an ample span of time in sist primarily of volcanic rocks. More often heavily eroded and are therefore which erosion could proceed. recently scientists on board the JOIDES both disfigured and diminished in scale. Despite the paucity of data and the Resolution have recovered direct sam- John J. Mahoney of the University of Ha- physical transformations of large igne- ples of rocks from two immense ocean- waii estimates that the Deccan Traps, ous provinces, we have managed to de- ic plateaus: Kerguelen in 1988 and On- for example, originally covered an area duce the original areas covered by lava tong Java, situated in the Pacific Ocean three times larger than at present. More- for five of the best studied of these for- northeast of Australia, in 1990. The sam- over, in the 65 million years since the mations: the Ontong Java and Kergue- ples consist of basaltic rock, similar to formation of the Deccan Traps, seafloor len-Broken Ridge oceanic plateaus, the that found in continental flood basalts. spreading has apparently dispersed sec- North Atlantic volcanic margins and the Earth scientists have wondered how tions of the original lavas to the Sey- Deccan and Columbia River continental such massive volcanic constructions chelles, the Mascarene Plateau and the flood basalts. The smallest of these—the arise. Unfortunately, the most basic in- Chagos-Laccadive Ridge far to the south Columbia River basalts—encompasses a formation needed to answer that ques- and southwest of India. region larger than the state of New York;

JOIDES RESOLUTION research vessel (left) is the centerpiece of the Ocean Drilling Project, the most extensive international program in earth sciences. The ship incorporates a drill that penetrates the sediments that accumulate on the ocean bottom and collects core samples lying as much as two kilometers deep (below). In this way, earth scientists can directly analyze igneous rock from underwater continental margins and from submerged oceanic plateaus.

44 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. the largest, the Ontong Java, is two rameters change, the seismic waves will igneous provinces. By our calculations, thirds the size of Australia. To evaluate alter course. Seismic analysis reveals a small province such as Columbia Riv- the area and volume of large igneous that the Hawaiian Islands built up as er incorporates approximately 1.3 mil- provinces, geologists realized that they basaltic magma intruded into and col- lion cubic kilometers of rock. The giant needed to examine modern-day ana- lected atop the preexisting oceanic crust. Ontong Java province contains at least logues of the gigantic igneous provinces. Those surface lavas evidently do not 36 million cubic kilometers of igneous contain all of the material associated rock, enough to bury the contiguous o currently active volcanic re- with the islands’ formation, however. U.S. under five meters of basalt. gions even begin to approach Below the islands and seamounts lies an The next stage in understanding the Nthe extinct large igneous prov- anomalous zone of rock distinguished origin and significance of the various ig- inces in magnitude. The most active by the rapid rate at which compres- neous provinces is to learn how quickly modern hot spots (the big island of Ha- sional waves travel through it. Geolo- they formed. Did they accumulate slow- waii and Réunion, an island lying to the gists think the underlying body of rock ly and steadily over tens of millions of east of Madagascar) cover an area of the derived from the same mantle source as years, in much the way that new ocean earth’s crust only about one fifth the did the islands and seamounts. Exist- crust forms at mid-ocean ridges? Or did size of the Columbia River basalts. Nev- ing techniques cannot reveal how much they emerge from a series of volcanic ertheless, studies of the crustal struc- new igneous material was added to the firestorms that swiftly pumped gases ture below Hawaii contribute to the un- previously existing oceanic crust as the and rock fragments into the air and wa- derstanding of the generic forms that islands formed. Any calculated volumes ter and that abruptly transformed the underlie all volcanic hot spots. of mantle plume–derived material con- geology of large areas of the earth? In 1982 Anthony B. Watts and Uri S. tained in the Hawaiian Chain therefore Geologists have expended a great deal ten Brink and their co-workers at the La- must be considered minimal. of e›ort attempting to answer those mont-Doherty Earth Observatory collect- Even so, we were able to establish a questions for the Deccan Traps. This ed seismic data by recording artificially basic relation between seismic-velocity flood basalt has garnered special atten- generated sound waves that are reflect- structure (as well as other geophysical tion because it erupted some 65 million ed and refracted underneath the Hawai- data) and the total volume of materi- years ago, at just about the time of the ian Islands and seamounts. The speed al encompassed within the Hawaiian Is- major extinction at the end of the Cre- of those waves depends on the physical lands. We then used Hawaii as our mod- taceous period, and it may have con- properties (such as density and elastic el for inferring the distribution of vol- tributed to that event [see “A Volcanic constants) of the rock; where those pa- canic material in various ancient, large Eruption,” by Vincent E. Courtillot; SCI-

RIG SEISMIC vessel (top right), operated by the Australian Geological Survey Or- ganization, collects seismic information about the ocean bottom. An air-gun ar- ray creates sound waves that reﬂect o› the oceanic crust (below). Underwater mi- crophones towed behind the ship sense the reﬂected waves, which are then recorded on board the ship. One such recording (bottom right) reveals the igneous structure (area below red line) underlying the Kerguelen Plateau.

MID-OCEAN RIDGE

ONTONG JAVA PLATEAU

LARGE KERGUELEN IGNEOUS PLATEAU PROVINCES

DECCAN TRAPS

NORTH ATLANTIC VOLCANIC MARGIN

COLUMBIA RIVER FLOOD BASALTS

UPPER LOWER Dʹʹ OUTER INNER MANTLE MANTLE CORE CORE RISING PLUMES of hot material migrate through the earth’s of the most voluminous igneous provinces were so huge that mantle; where the head of the plume reaches the surface, a they must have originated at least in part in the lower mantle, large igneous province forms (left). Plumes probably originate most likely at the Dʹʹ. The spheres on the right depict the mini- at the boundary layers between the core and mantle (the Dʹʹ) mum (white) and maximum (dark orange) inferred diameters and between the upper and lower mantle. The parent plumes of the plumes associated with ﬁve major igneous provinces.

ENTIFIC AMERICAN, October 1990]. Rob- versity of Strasbourg, indicates that the meters of ocean crust a year over the ert A. Duncan of Oregon State Universi- Kerguelen Plateau mostly formed with- past 150 million years. In other words, ty, working with Vincent E. Courtillot in a span of 4.5 million years. individual igneous provinces have gen- and Didier Vandamme of the Institute From a geologic perspective, the larg- erated new crust at rates comparable to of Physics of the Earth in Paris and sev- est igneous provinces emerged remark- or greater than that of the global sea- eral other colleagues, has performed ably fast. In comparison, the Rocky floor spreading system. radioactive dating and magnetic analy- Mountains have been rising for more sis on samples from the Deccan Traps. than 40 million years, and the chain of arge igneous provinces build up so The scientists’ results indicate that most Hawaiian Islands and the Emperor Sea- swiftly compared with the churn- of the lavas erupted within a span of mounts have been building for at least L ing of the earth’s deep interior less than one million years. Using simi- 70 million years. Moreover, much of the that they must have derived from single, lar methods, Ajoy K. Baksi of Louisiana volcanic activity associated with the discrete sources. If one knows the vol- State University found that the Colum- provinces may have occurred in short, ume of basaltic rock contained in those bia River flood basalts mostly erupted violent episodes separated by long pe- provinces, one can evaluate the dimen- over one and a half million years. riods of relative quiescence. sions of the hot plumes in the mantle Geologists have few direct samples Once we had collected data on the to- that led to their genesis. Only a fraction on which to judge the ages and rates of tal volume of the major igneous prov- of the plume material actually melts formation of underwater volcanic mar- inces and on how quickly they formed, and reaches the crust, and this fraction gins and oceanic plateaus. The scant we could finally deduce the magnitude presumably is smaller at greater depths, available evidence hints that submarine of the volcanic forces that created them. where increased pressures tend to keep igneous provinces accumulated about The eruptions that built up the Ontong the mantle rocks in a solid state. Hence, as rapidly as those on land. Eldholm es- Java province unleashed between 12 less melting should take place under timates that the bulk of the North Atlan- and 15 cubic kilometers of igneous rock thick continental lithosphere than below tic volcanic province formed in three each year; Deccan volcanism produced thin oceanic lithosphere. Stephen M. Eg- million years or less. John A. Tarduno between two and eight cubic kilometers gins of the Australian National Univer- and his collaborators at the Scripps In- annually. Assuming that the creation of sity and Shen-su Sun of the Australian stitution of Oceanography, in conjunc- igneous provinces, like other eruptive Geological Survey Organization estimate tion with Mahoney, conclude that the processes, occurs in fits and starts, the that the molten, basaltic portion of the Ontong Java province was constructed pace of crustal production may have plume (the part that gives rise to the sur- in less than three million years. Our been far greater in some years. For ref- face volcanism and to subsurface igne- analysis of rock dating, conducted by erence, Roger L. Larson of the Universi- ous intrusions) accounts for from 5 to Hugh Davies, formerly of the Australian ty of Rhode Island estimates that the 30 percent of the plume’s total volume. Geological Survey Organization, and Hu- global network of mid-ocean ridges has We used those numbers to calculate bert Whitechurch, formerly of the Uni- yielded between 16 and 26 cubic kilo- diameters for the thermal anomalies in

46 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. the mantle associated with the five best- linear submarine ridge or a sequence of associated with these changes could studied igneous provinces. For simplic- islands and seamounts. The Hawaiian- strongly influence the evolution of life. ity, we assumed the rising plumes to Emperor chain presumably developed The current fascination with the be spherical because such a shape rep- in this manner, although how a plume greenhouse e›ect and global change has resents the most e¤cient way to trans- could persist for more than 70 million renewed interest in Vogt’s ideas. Ste- port material and so permits the plume years remains a puzzle. phen Self of the University of Hawaii and to move at a plausibly slow velocity Michael R. Rampino of New York Univer- through the earth’s interior. Our analysis s the 1991 eruption of Mount Pin- sity note that the environmental impact suggests that the Ontong Java province atubo in the Philippines docu- of a large igneous province depends in must be derived from a mantle plume A ments, even moderate volcan- part on whether it forms on land or un- at least 600 kilometers, and possibly as ic outbursts can severely damage the derwater. Ocean plateaus and volcanic much as 1,400 kilometers, in diameter. local environment. Yet such geologic passive margins modify the geometry of That size carries special significance events are incidental twitches compared the ocean basins and alter the global sea for geologists. It indicates that large with the convulsions of magmatic ac- level. We estimate that accumulation of plumes must contain at least some ma- tivity during the formation of large ig- the Ontong Java plateau material elevat- terial from the lower mantle, more neous provinces. One can thus begin to ed sea level by about 10 meters. Ocean- than 670 kilometers below the surface. imagine that those ancient eruptions ic plateaus and volcanic margins may At that depth, the velocity of seismic had profound consequences. In 1972 block or constrict circulation in ways waves changes abruptly, probably be- Peter R. Vogt of the Naval Research that influence large-scale circulation, cause of a change in the mineral struc- Laboratory first proposed that the sur- erosion and sedimentation, especially ture of the mantle rocks. The issue facing of a mantle plume would lead to if the volcanic activity occurs in a sen- of whether the whole mantle mixes or physical and chemical changes around sitive location between ocean basins. whether the upper and lower parts of the globe; the environmental e›ects Because they are much denser and the mantle behave as independent systems that circulate separately strongly divides geophysicists. SEA LEVEL INFERRED GLOBAL (METERS ABOVE TEMPERATURE Our work favors models that allow BLACK PRESENT LEVEL) (DEGREES CELSIUS) SHALE MASS for at least some interaction between 100 150 200 250 0 5 10 15 20 25 FORMATION EXTINCTIONS the upper and lower mantles. In our 0 view, the largest plumes originate in the lower mantle, most likely at the Dʹʹ, 10 a region having unusual seismic proper- COLUMBIA RIVER FLOOD BASALTS ties that lies just outside the core. Small- 20 er plumes may arise at the 670-kilometer-deep transition zone between the 30 upper and lower mantle. Regardless of where mantle plumes originate, their attributes and e›ects 40 at the surface depend strongly on the temperature, composition and physical 50 state of the material they encounter just NORTH ATLANTIC VOLCANIC MARGIN beneath the lithosphere. These factors, 60 combined with the local strength of the DECCAN TRAPS lithosphere, determine the volume, tim- 70 ing and position of the eruption at the surface. When plumes rise under conti- 80 nental masses, they may help force the continents apart at a weak spot and 90 then induce the formation of extensive volcanic features along the margins of TIME (MILLIONS OF YEARS AGO) 100 the rifted landmass [see “Volcanism at Rifts,” by Robert S. White and Dan P. 110 McKenzie; SCIENTIFIC AMERICAN, July KERGUELEN PLATEAU 1989]. Under certain circumstances, the 120 plume may penetrate the thick central ONTONG JAVA PLATEAU regions of continental blocks and give birth to a continental flood basalt. 130 If the upwelling mantle plume surfaces underneath the seafloor, it may give 140 rise to an oceanic plateau. Laboratory experiments suggest that a drawn-out 150 tail of hot material should lag behind ENVIRONMENTAL EFFECTS from the eruption of large igneous provinces include fluc- the spherical head of the plume, yielding tuations in the global sea level. Powerful volcanic activity could alter the chemistry a long-lived, focused source of magma. and circulation of the atmosphere and oceans, influencing the evolution of life. Volca- Over millions of years, plate motions noes release carbon dioxide, which may contribute to greenhouse warming; warmer cause the ocean floor to migrate over the temperatures promote the formation of black shale, in part by enhancing global bio- site of the hot spot. Where lava erupts mass production. In contrast, several mass extinctions of the past 150 million years at the surface, it gradually constructs a coincide with the appearance of igneous provinces, hinting at a causal relationship.

GAS, PARTICLES CHANGE IN ACID RAIN AND CHEMICALS SEA LEVEL

HEAT, GAS, PARTICLES AND CHEMICALS UPWELLING

HYDRO- THERMAL VENT

HOT SPOT

PHYSICAL AND CHEMICAL EFFECTS accompany the appear- ted during eruptions would alter the composition of the land, ance of large igneous provinces. Volcanoes emit carbon diox- air and water. Topographic changes associated with exten- ide, which can elevate the global temperature. Sulfurous parti- sive volcanism would modify oceanic circulation and change cles and dust in the stratosphere could induce acid rain and sea level; heat and hydrothermal activity cause upwelling, block out sunlight. Heavy metals and other chemicals emit- further transforming underwater conditions. more massive than the atmosphere, the bility in the Dʹʹ layer initiated so-called often carry sulfurous particles into the oceans are far more able to absorb and superplumes that ultimately supplied stratosphere, where they are converted dilute gases and heat, a factor that the Ontong Java and Kerguelen oceanic to tiny droplets of sulfuric acid, accord- tends to ameliorate the consequences of plateaus, along with several other small- ing to Rampino and Self. Basaltic lavas volcanic eruptions. On the other hand, er plateaus in the Pacific Ocean. Ken- emit about 10 times as much sulfur submarine volcanism and associated neth G. Caldeira of Penn State and Ram- per unit volume as do silica-rich lavas; hydrothermal activity may leach trace pino have run computer models of the Charles B. O¤cer of Dartmouth College metals such as arsenic, which are poi- geochemical cycle based on their as- and his co-workers deduce that if the sonous to marine life. Heat produced sumptions about the quantity of car- gases and particles produced during by submarine eruptions induces ocean- bon dioxide in the atmosphere during basaltic eruptions were shot into the bottom waters to rise to the top, altering that era. From the models, the workers stratosphere, they could cause short- the circulation of the surface waters and infer that worldwide temperatures av- term plagues of acid rain, worldwide disrupting the organisms living there. eraged from 7.6 to 12.5 degrees Celsius darkening and global cooling. Richard B. Carbon dioxide emitted at an under- higher than today’s mean, if one takes Stothers of the National Aeronautics and sea volcanic site spreads through the into account the di›erent geography Space Administration Goddard Space ocean. In so doing, the compound could and higher sea level at the time. Flight Center, along with several others, increase the alkalinity of the seawater, Where volcanic eruptions occur on hypothesizes that massive lava foun- thereby a›ecting both marine life and dry land, they directly alter the physics tains and vigorous atmospheric con- climate. If elevated levels of carbon di- and chemistry of the atmosphere. Alan vection taking place over eruptive vents oxide lead to global warming, ocean cir- R. Hu›man of Exxon Exploration Com- in continental flood basalts could sup- culation should turn sluggish, in which pany in Houston calculates that a single ply the su¤cient upward momentum case it would carry less dissolved oxy- flood basalt event that generates 1,000 to inject material into the stratosphere. gen. Oxygen-deprived waters may con- cubic kilometers of lava (the volume of tribute to the formation of carbon-rich a typical flow in the Columbia River he powerful Lakagigar eruption black shale, as documented by the late province) emits 16 trillion kilograms of in Iceland in 1783–84 illustrates Seymour O. Schlanger of Northwestern carbon dioxide, three trillion kilograms T the potential disruptive e›ects of University, Michael A. Arthur of Penn- of sulfur and 30 billion kilograms of flood basalt volcanism. Although the La- sylvania State University and Hugh C. halogens. The thousands of such epi- kagigar eruption poured out only about Jenkyns of the University of Oxford. sodes that must occur in the accumula- 15 cubic kilometers of lava, local tem- The atmosphere must have absorbed tion of an individual large igneous prov- peratures declined noticeably in the fol- an extensive outpouring of carbon diox- ince would modify the atmosphere in lowing years. About three quarters of all ide around 120 million years ago, when ways that would dwarf the e›ects of livestock in Iceland died, probably the the tempo of volcanism peaked. Larson modern, human-generated pollutants. result both of the deterioration of the speculates that a massive thermal insta- Explosive eruptions of silica-rich rock climate and of the emissions of acid gas;

48 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. the resulting famine killed about one Another significant but less celebrat- fourth of Iceland’s human population. ed change in the global environment Dust veils, fog and haze appeared over took place 10 million years after the most of Europe and adjacent parts of demise of the dinosaurs, during the em- Asia and Africa for many months after placement of the North Atlantic volcanic the eruption. margins. At that time, many deep-ocean Scientists are uncovering clear evi- foraminifera and land mammals be- dence that the environmental conse- came extinct, and hydrothermal activity quences of flood basalt volcanism have was high. Moreover, David K. Rea and in fact contributed to mass extinctions. his co-workers at the University of Mich- The most severe extinction in terrestri- igan, along with Ellen Thomas of Yale al history occurred 248 million years University, find evidence of major trans- ago, when the Siberian Traps formed. formations in deep water and in atmo- At that time, roughly 95 percent of all spheric circulation. marine species perished; the ensuing Analysis of the oxygen isotopes taken evolutionary free-for-all marked the first in by foraminifera indicates that ocean appearance of the dinosaurs. temperatures 55 million years ago, dur- The biological repercussions of the ing the early Eocene epoch, were warm- eruption of large igneous provinces er than at any other time during the may depend in part on the state of the past 70 million years. Eldholm, working global environment at the time. When with Thomas, recently postulated that the environment is already stressed by the balmy ocean surface of the early Eo- other factors, such volcanism may trig- cene could have resulted from the out- ger rapid climatic, oceanographic and gassing of carbon dioxide during the biotic changes. When the environment eruptions along the North Atlantic vol- is robust, few e›ects may show up in canic margins. Ash layers dating from the geologic record. 55 million years ago cover large regions Oddly enough, the most voluminous of northwestern Europe, a¤rming the igneous province, the Ontong Java, pro- violence of the volcanism. Higher atmo- duced virtually no detectable extinc- spheric temperatures could have led to tions. On the contrary, its formation the formation of a layer of warm water coincides with the deposition of black atop the oceans at high latitudes. Such shale, which is suggestive of an interval a warm surface layer would tend to re- of enhanced biological activity. We pro- sist mixing with underlying, cooler wa- pose that the deleterious e›ects associ- ters. The consequent changes in deep- ated with the Ontong Java were mini- water circulation could have been fatal LAVA FLOW from Kilauea Volcano in mal in part because it erupted under- for many ocean-bottom species. Hawaii only hints at the magnitude and water, as indicated by Ocean Drilling Even the relatively small Columbia Riv- potential environmental impact of the Program studies led by Loren Kroenke er flood basalts coincided with a mass flows incorporated into large igneous of the University of Hawaii and Wolf- extinction 16 million years ago. At about provinces. Individual sheets of lava in gang H. Berger of the Scripps Institu- that time, the earth began to experience the Columbia River flood basalts, for ex- tion of Oceanography. We also suspect an ongoing cycle of ice ages, as Maureen ample, extend for hundreds of kilome- that the global environment was in a E. Raymo of the Massachusetts Institute ters and stand several meters thick. resilient state 120 million years ago, of Technology points out. Perhaps the although such an assertion is di¤cult global environment was already so frail to quantify and prove. In contrast, the that even a moderate eruption could natural causes of global change. Earth emplacement of the smaller Kerguelen have had a substantial impact. scientists are now intensifying their ef- and Broken Ridge provinces, which took Obviously, geologists have—literal- forts to produce better seismic images, place about 110 million years ago, co- ly and figuratively—only just scratched to undertake more field and laboratory incided with a mass extinction as well the surface of large igneous provinces. studies, to perform additional model- as with major deposits of black shale. Current knowledge of these formations ing and to conduct more scientific drill- The eruptions that built up the Dec- amply demonstrates that they contain ing. Those labors promise even better can Traps, and simultaneous volcanism crucial information about the internal insight into the intimate links between along the margins of western India and workings of the earth and about the the earth’s inner and outer realms. along the Seychelles and the Mascarene Plateau, took place approximately 65 million years ago—just about the time that nearly half of all species of life, in- FURTHER READING cluding the dinosaurs, went extinct. Sci- CONTINENTAL FLOOD BASALTS. Edited by 19, No. 10, pages 963–966; October 1991. entists continue to debate whether to J. D. Macdougall. Kluwer Academic Pub- MAGMATISM AND THE CAUSES OF CONTI- attribute those changes to the impact lishers, 1988. NENTAL BREAK-UP. Edited by B. C. Storey, of a sizable asteroid or to more earthly HOTSPOTS, MANTLE PLUMES, FLOOD BA- T. Alabaster and R. J. Pankhurst. Special explanations. Even if an asteroid was SALTS, AND TRUE POLAR WANDER. R. A. Publication, No. 68. Geological Society of the primary agent of the extinction, the Duncan and M. A. Richards in Reviews of London, 1992. Geophysics, Vol. 29, No. 1, pages 31–50; LARGE IGNEOUS PROVINCES: CRUSTAL Deccan eruptions may have contribut- February 1991. STRUCTURE, DIMENSIONS, AND EXTERNAL ed to an environmental deterioration GEOLOGICAL CONSEQUENCES OF SUPER- CONSEQUENCES. M. F. Co¤n and O. Eld- that could have magnified the reper- PLUMES. Roger L. Larson in Geology, Vol. holm in Reviews of Geophysics (in press). cussions of the impact.

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 49 Evolutionarily Mobile Modules in Proteins Many proteins consist of a fairly small set of modular elements. How these units spread and multiplied during evolution is not altogether clear, but a pattern may be emerging

by Russell F. Doolittle and Peer Bork

olecular biologists and bio- to the argument that the subdivision of into a characteristic shape under a de- chemists have learned in re- genes into separate coding parts is a far fined set of circumstances is called a do- Mcent decades that many pro- more recent development by evolution- main, but that formal definition is sel- teins consist of domains, or discrete ary standards. dom applied with rigor. More often than blocks of amino acids. Many of these Proteins are long chains of small mol- not, the term is applied to any part of a domains have well-defined functions ecules called amino acids. Twenty dif- protein that can be defined as structur- that contribute to the overall activity of ferent kinds of amino acids, each with ally distinct from the rest. Some small a protein. Furthermore, some of these its own shape and chemical character, proteins are completely embodied in a modular units frequently move about make up all the proteins found in na- single domain; many others are com- within and between proteins during ture. All the properties of a protein de- posed of two or more domains; and evolution. The evolutionary mobility of pend on which of the 20 amino acids some are made up of many domains, these modules is not restricted to hops are used in its construction and, partic- the shapes of which may be very simi- within the genetic material of a single ularly, on the order in which they are lar or very di›erent. species: in some cases, the modules are strung together. Most notably, the ami- apparently able to travel laterally across no acid sequence determines how the species lines—even moving from ani- protein will fold up into an active, three- mal cells to bacteria, for example. dimensional body. Because a compartmental delineation The length of a sequence plays an has also been found in the regions of important part in that determination. genes that encode proteins, many biol- Chains of amino acids—usually called ogists have been convinced that these polypeptides—may be up to many thou- structural features are reflections of the sands of amino acid units long. (The same underlying phenomenon. They be- record to date is for titin, a muscle pro- lieve that each genetic coding region cor- tein that contains upward of 30,000 responds to a specific structural feature amino acid residues.) Short chains, how- in a protein. We and our colleagues take ever, are not large enough to have suf- a somewhat di›erent point of view. The ficient intramolecular attachments to weight of evidence, in our opinion, falls lock themselves into a single conformation; they tend to flop from one form to another. Ordinarily, when a polypeptide RUSSELL F. DOOLITTLE and PEER chain contains 30 or 40 residues, it be- BORK analyze the structures of proteins gins to have enough internally cohesive to gain insights into evolution. Doolit- forces to give it a predominant shape, tle is professor of biology at the Center although it may still need additional sta- for Molecular Genetics of the Univer- bilization from bound metal ions or di- sity of California, San Diego. He began sulfide bonds between pairs of the ami- his studies of protein evolution while he no acid cysteine. was a graduate student in biochemistry In a constant environment, any pro- at Harvard University in the late 1950s. Doolittle is a member of the National tein containing more than some mini- Academy of Sciences and other scientif- mum number of amino acid units will ic associations. This article is his third always fold itself in the same way. That for Scientific American.Bork is a project environment may be the dilute salt so- leader at the Max Delbrück Center for lution that constitutes many biological Molecular Medicine in Berlin and a visit- fluids or the greasy confines of a biolog- ing scientist at the European Molecular ical membrane; it can also include near- Biology Laboratory in Heidelberg. He also holds a teaching appointment at Hum- by proteins or even other parts of the boldt University in Berlin. same long polypeptide chain. A sequence that folds spontaneously CD2 DOMAIN 2

50 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. The most straightforward way to iden- ties of proteins in this way, one can closely resembled features of some oth- tify a domain is to determine its struc- learn much about their evolution. er proteins he had seen. Specifically, a ture through x-ray di›raction of protein part of the enzyme that bound to a co- crystals or nuclear magnetic resonance ntil the early 1970s, the conven- factor had obvious counterparts in oth- (NMR) studies. Once the amino acid se- tional wisdom about how pro- er dehydrogenases. quence of a domain has been identi- Uteins evolved centered mostly What made those structural similar- fied, researchers can find other related on “duplication and modification.” The ities noteworthy was that they did not domains without recourse to structural gene for a given protein is occasional- always occur in the same relative parts studies: they can simply look for amino ly duplicated by various recombination of the molecules. It seemed as though acid sequences that are similar to those processes during which genetic infor- during evolution the unit had moved of familiar structures. That shortcut is mation is exchanged between strands around within the linear amino acid se- extremely helpful because many more of DNA. Sometimes the duplication re- quence without losing its function of amino acid sequences are known than sults in a second gene that can then binding to a cofactor. Rossmann sug- are structures from x-ray di›raction or undergo further modification or muta- gested that proteins were constructed NMR. Even in the absence of experimen- tion to produce a new protein with a of modules—what we would now con- tal determinations, it is often possible novel function. Alternatively, the dupli- sider domains—that had appeared ear- to infer the existence of a domain on cated DNA can be in tandem, in which ly in life’s history and had been assem- the basis of sequence alone. By looking case the original protein is elongated bled in di›erent combinations. at the structural and sequence similari- and may assume novel properties as a His observation presented a possibili- result. As comparisons of amino acid ty for protein evolution that greatly aug- sequences make obvious, these kinds of internal duplications have clearly given rise to many extended proteins, rang- GENETICALLY MOBILE MODULES have ing from small ones, such as the bacte- been found in many proteins. Two types rial ferredoxins that have only 56 ami- of these modules, or domains, are shown no acid residues, to large ones, such as here. The Fn3 and the GHR domains are bacterial beta-galactosidase, which has examples of fibronectin type III modules. more than 1,000. The PapD and CD2 domains are immunoglobulin domains. These modules are A hidden aspect of protein evolution linear sequences of amino acids that can came to light about 20 years ago, how- fold themselves into consistent, recog- ever, when Michael G. Rossmann of Pur- nizable structures with specific biochem- due University determined the three-di- ical properties. During evolution, these mensional structure of the enzyme lac- domains can move as discrete units tate dehydrogenase by x-ray di›raction. from one protein to another, which helps One part of that molecule, he noticed, new types of proteins to appear.

PapD DOMAIN 1 GHR DOMAIN 1 Fn3 (10TH DOMAIN)

COLLAGEN XII

10 GIANT MUSCLE PROTEIN TITIN

n n 9

RECEPTOR TYROSINE KINASE EPH

PROTEIN KINASE SRC

COLLAGENOUS TRANSMEMBRANE NO KNOWN HOMOLOGIES TO REPEATED BLOCKS TRIPLE HELIX REGION OTHER PROTEINS OF DOMAINS n

Fn1 Fn2 Fn3 VWA TSPO Ig TYR-KIN CRR S3 SH2

LIKE BEADS IN A NECKLACE, domains appear as distinct sub- tin, collagen XII and the muscle protein titin, contain many units of modular proteins. Some proteins, such as fibronec- repeats of the same few domains. mented what could be accomplished human version) that is tightly folded biologists happened on an unexpected purely through duplication and modifi- and pinned together by three disulfide feature of genes that seemed to o›er an cation. If new proteins could be created bonds. Similar domains have been iden- explanation. As James Watson and Fran- by recombining the components of oth- tified by their sequences in more than cis Crick had learned in 1953, genetic er ones, then protein diversity could 100 proteins, in which they occur from information is inscribed in the double grow explosively. one to more than 30 copies. helices of deoxyribonucleic acid (DNA) Rossmann’s suggestion proved pre- molecules. Just as proteins are strings of scient. The amino acid sequences for he functions of many of these amino acids, DNA molecules are strings numerous large proteins have been de- modules are not altogether clear, of nucleotide bases. The DNA nucleotide termined, and many have the kind of T but many of them do bind to or sequences are copied, or transcribed, as grossly repetitive structure that one recognize particular substances. There complementary strands of messenger might expect of a string of mobile mod- is a family of lectins that bind to various ribonucleic acid (RNA). Inside cells, min- ules. For instance, the protein fibronec- carbohydrates. Similarly, the immuno- iature factories called ribosomes as- tin is made up of two long chains, each globulin domain, a feature of antibod- semble proteins from the instructions of which contains more than 2,000 ies and other molecules in the immune in the RNA: each codon, or three-base amino acid residues. Casual inspection system, is well known for its binding ca- sequence in the RNA, corresponds to revealed that the chains of this large pabilities. Some domains may serve as an amino acid. extracellular protein consist of several recognition tags that identify a protein The surprising observation made in runs of three di›erent types of repeat- as “belonging” to a certain tissue. Many the mid-1970s was that the DNA coded sequences. The repeats, which are appear to be mere linkers or spacers, in- ing for a polypeptide can be interrupt- referred to as Fn1, Fn2 and Fn3, have nocuous connecting units. Finally, some ed by noncoding sequences—arrays of lengths on the order of 45, 60 and 100 may have no function at all. It appears, bases that do not correspond to the amino acids, respectively. (The repeats therefore, that many domains can and sequence of amino acids found in the are “imperfect,” which means that all do move within and between proteins ultimate protein product. The noncod- the repeats of a given type are not ex- during evolution. As long as no damage ing sequences are excised by a splicing actly identical.) Presumably, each type or loss of function results from a shuf- mechanism before the messenger RNA of repeat can fold up independently as fle, the evolutionary cost of maintaining strand is translated into a polypeptide. a true domain, and the overall protein a domain in a new setting is insignifi- The discovery of those interruptions must be like a long necklace made from cant. That result is a corollary of the the- in genes prompted Walter Gilbert of three types of beads. ory of neutral evolution, but it might be Harvard University to suggest that the Surprisingly, sequences similar to phrased as a rule borrowed from profes- noncoding sequences, which he named Fn1, Fn2 and Fn3 were subsequently sional basketball: “No harm, no foul.” introns (for intervening sequences), fa- observed in a huge assortment of other When Rossmann first suggested that cilitated the exchange of the coding animal proteins. Much the same is also modular units might move within and parts of the genes, which he termed true for a number of other identified between proteins, no one pondered very exons (for expressed sequences). The domains. The small protein called epi- seriously the genetic mechanisms that thought was that the additional distance dermal growth factor consists of a sin- might account for those rearrangements. between coding segments would lead gle domain (53 amino acids long in the Not long afterward, however, molecular to proportionately more opportunities

52 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. for recombinations, which depend on all exons are evolutionarily mobile and bile is that only true domains—those random breaks in the DNA. It was sug- correspond to potential modular build- that can fold completely and indepen- gested that similarities between the se- ing blocks in proteins. dently—will be able to survive in new quences of introns might promote mis- In our view, this notion is mistak- protein settings. Smaller, less self-su¤- alignments and unequal crossovers of en on two counts. One is that, as Lázló cient sequences would be unable to fold DNA during recombination, which would Patthy of the Institute of Enzymology and would lose their identity. Moreover, make the rearrangement of genes easi- in Budapest first pointed out, all exons if a shu‹ed unit were to land between er. Although there was no basis at that can be shu‹ed, but only a fraction of two exons that were not themselves true time for supposing that introns would such shu‹ed units will be genetically domains, then the product of the gene have similar sequences, subsequent re- compatible—that is, many of them can- receiving the addition might not be able search has shown that introns are a ha- not be sensibly translated in some new to fold itself properly either. ven for large numbers of mobile genet- positions. When an intron falls in a cod- These two factors, one genetic and ic elements. The similar sequences in ing sequence, it will occupy one of three the other structural, contribute great- those mobile elements can contribute types of positions: squarely between two ly to why mobile domains are so often to genetic delinquencies during meio- codons (type 0), between the first and found in one another’s company. Not sis, the cell division process that gives second positions of a codon (type 1) or infrequently, proteins that contain one rise to eggs and sperm. between the second and third positions mobile domain contain others; some (type 2). If that intron and the coding proteins are mosaics that contain as f course, many creatures do not sequence adjacent to it are shuffled into many as five di›erent commonly shuf- engage in meiosis; have those a new location, the intron must adopt fled modules. These types of proteins Oorganisms missed out on a the same type of position—otherwise are both genetically and structurally tol- splendid way to assemble new proteins? the shifted codons will be translated im- erant of the shu‹ing process. Introns that interrupt protein coding are properly and a nonsensical amino acid The observation that many modules found only in the DNA of eukaryotes, or- sequence will result. If introns were dis- are encoded by exons has been inter- ganisms with discrete nuclei. The genes tributed randomly, we would expect to preted as support for the idea that the for bacterial proteins do not contain in- find that after a shu‹e, only one third primeval organisms assembled all their trons: every set of three bases corre- of the new exon combinations would be proteins from an inventory of exon-en- sponds to an amino acid in the protein. in phase. Curiously, the overwhelming coded primitive structural components. (A few types of introns that do not in- majority of the genes for the most fre- Several points argue against such an interrupt proteins have been found in quently shu‹ed modules are flanked terpretation, however. For one, simple bacteria, but they are not properly part by introns that are of type 1. arithmetic proves that the hypothetical of this discussion.) The other fundamental reason why early exons would have been too small The absence of introns from bacterial only some exons are evolutionarily mo- to produce protein components that protein genes led Ford Doolittle of Dal- housie University and James Darnell of the Rockefeller University to suggest independently that bacteria may have Sizes of Some Mobile Domains possessed introns in the past but had lost them. Presumably, their genomes Evolutionarily mobile domains vary in size. Some hold their shape in part had been streamlined during evolution because of disulfide bonds between pairs of their cysteine residues. to make their replication more e¤cient. Other domains are stable without such bonds. In short, the introns had been around since the beginning of life; it was the DOMAINS CONTAINING APPROXIMATE NUMBER NUMBER OF short coding sequences that were en- DISULFIDE BONDS OF AMINO ACID UNITS CYSTEINES gendered separately. Somatomedin B 40 8 Doolittle and Darnell’s proposal led to Complement C9 40 6 a still unsettled debate about whether EGF 45 6 introns appeared “early” and are fun- Fn1 45 4 damental to the origin of all proteins Fn2 60 4 or whether they came “late.” The latter “Apple” (Sushi) (ßGP26) 61 4 proposition was advanced by Thomas Ovomucoid 68 6 Cavalier-Smith, now at the University of VWF-C 72 8 British Columbia, who theorized that in- Kringle 80 6 trons might be invasive bits of nucleic Kunitz 80 6 acids, referred to as transposable ele- Link ~100 4 ments, that originated in the symbiotic Scavenger receptor 110 6 organisms that eventually became the Fibrinogen-related domain 250 4 or 6 mitochondria and other organelles of eukaryotic cells. His idea has been ex- NO DISULFIDE BONDS tended by Donal Hickey of Ottawa Uni- Collagen 18 versity, among others. Leucine-rich 25 As it turns out, the stretches of DNA “Gla” 35 that encode the evolutionarily mobile Collagen-binding 50 modules in proteins are frequently, but Lectinlike 100 not always, flanked by introns. In other Fn3 ~100 words, the structural units in many pro- SH2 100 teins are encoded by exons. That obser- SH3 100 vation fostered a widespread belief that

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 53 CODING SEQUENCE WITH NO INTRONS TYPE 1 INTRON INTERRUPTS A CODON AFTER FIRST POSITION C C C G G G A A A C C C C C C G G G A A A C C C C C C G G A A A C C C CODON INTRON TRANSPOSITION FRAMESHIFTED TYPE 0 INTRON FALLS BETWEEN CODONS OF EXONS CODONS EXON EXON TYPE 2 INTRON INTERRUPTS A CODON AFTER SECOND POSITION

C C C G G G A A A C C C C C C G G G A A A C C C C C C G G G G A A A C C C

INTRON INTRON

INTERRUPTED STRUCTURE of some genes that encode pro- sequences in three di›erent positions. If exons and introns teins divides them into exons and introns (expressed and in- moved randomly, transpositions between di›erent types of intervening sequences). Triplets of bases in exons are translat- trons would cause frameshift mutations. That problem counts ed as amino acids in proteins. Introns can interrupt coding against the idea that all exons encode movable modules. could fold on their own. The average out all the others. None of the exons, tween these phenomena raises a thorny size of the known exons in vertebrate it seems, has ever shu‹ed itself out of problem. Perhaps the evolution of these genomes today is 135 nucleotide bases, the domain. Thus, the mere presence introns did facilitate exon shu‹ing. On which corresponds to a polypeptide of a of introns in a gene is apparently not the other hand, it is not impossible that mere 45 amino acids. A sequence that enough to cause exons to be mobile. these introns often delineate the cod- short usually needs auxiliary stabiliza- That the vast majority of identified ex- ing regions for domains because that tion to fold into a stable conformation. ons has never been found in more than placement is advantageous for intron Keep in mind, too, that the propo- one setting argues against simple and propagation. If an intron interrupted a nents of the “intron early” theory con- indiscriminate mobility. sequence coding for a domain, it might tend that introns are constantly being survive in that situation (if it did not vi- lost over time. They have been forced here are other examples of mov- olate the phase rule noted earlier). It to that conclusion by the sporadic occur- able units that contain introns would not spread further, however, be- rence of introns in di›erent species. This T within their coding sequences. cause the exons on its boundaries could inconstant pattern could be the result One of the first movable modules to be not stand alone and thus could not be of either gain or loss of introns, but if identified is referred to as a kringle (be- moved independently. Conversely, if the one is wedded to the notion that introns cause it resembles a type of Danish pas- intron lands between regions coding were there from the beginning, then try by that name). It is made of about for independently folding units, it can the only explanation is loss. According- 80 amino acid units and contains three spread to other locations along with the ly, the earliest exons would have been characteristic disulfide bonds. It is quite exons. So exon shu‹ing may be only even smaller, encoding polypeptides similar to the Fn2 domain, di›ering only incidental to the survival of introns. that could not reasonably be expected in the number of residues between its One way to learn more about the evo- to fold into domains on their own. cysteine residues, and some workers lution of movable modules is to look Another argument against the use of do not distinguish between the two. In at the structure and dispersal of one in modern mosaic proteins as examples of some of its settings, the gene for the particular. Our favorite is Fn3, the fibro- early building blocks concerns the dis- kringle is split by an intron, but so far nectin type III domain. Like kringles, tribution of domains among proteins. no one has found half a kringle in any Fn3 units are sometimes split by sin- Far and away, the majority of mobile protein. gle introns, but they have never been modules known to date are found exclu- A further point in favor of the “intron observed with fewer than their full sively in animal proteins. At this point, late” theory is that the introns that in- complement of 90 to 100 amino acids. we do not really know when or where terrupt coding regions are much more The two of us independently followed most of them arose initially. Perhaps the common in plants and animals than the discovery of Fn3 in various proteins trail of their evolution has been partly among the earliest diverging eukary- for several years. In early reports the obscured by extensive sequence chang- otes. No introns at all have been found sequence was turning up only in ani- es in the related domains of plants, fun- in primitive eukaryotes, such as Giardia mal proteins, so we were both surprised gi and protozoa. As we shall discuss, lamblia. Furthermore, modular proteins when, in 1990, workers at Niigata Uni- the fact that three-dimensional struc- have been identified in plants for which versity in Japan reported the existence tures are more persistent than sequenc- no recognizable counterpart is known of an Fn3 domain in a bacterial protein. es in an evolutionary sense may allow in animals, and vice versa. Finally, there Our common interest came to light at a this puzzle to be solved. is indirect evidence that the modular meeting in Italy in 1991, whereupon we Besides all the evidence that most ex- assembly of some bacterial proteins oc- decided to join forces in a comprehen- ons are not evolutionarily mobile, some curred so recently that they must have sive inventory of Fn3 occurrences. mobile domains are also clearly not the evolved without the aid of introns. All To that end, we screened a data base products of single exons. One large do- this evidence suggests that protein-in- of protein sequences by various means, main, first observed in the fibrinogen terrupting introns appeared after the including a pattern-searching algorithm molecule of vertebrate animals, is com- evolution of eukaryotes. that one of us (Bork) had devised with posed of 250 amino acids. In some pro- So some exons encode domains, but Christian Grunwald when both were at teins, the gene for that domain contains most exons do not; those that do can of- the Central Institute for Molecular Biolo- multiple introns. Yet none of these indi- ten be genetically duplicated and shuf- gy in East Berlin. We uncovered well over vidual exons has ever been found with- fled. The issue of cause and e›ect be- 300 unique occurrences of the Fn3 se-

54 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. quence motif, which established that it domain had been present in a common may assimilate itself into the new host’s was indeed a true domain, should there ancestor of prokaryotes and eukaryotes, DNA. Bacteria can be transformed by have been any doubt. The 300 occur- we would have expected its radiation to picking up DNA from their surround- rences actually represented 67 di›erent have continued along the major lines ings, which may include decaying ani- proteins, not counting the same proteins of descent and to be represented in all mal cells. Many bacteria also carry genes from di›erent species. Sixty of these these groups. on small rings of DNA, or plasmids, were from animals and seven from bac- that they can exchange with other bacteria. None of the sequences identified he idea that the gene for a domain teria. In theory, all these mechanisms were from plants, fungi or unicellular can migrate between distantly re- represent opportunities for the hori- eukaryotes. T lated organisms may seem out- zontal transmission of genes. The obvious questions to be asked landish at first. Common experience Allowing that some bacterium did ob- were: Did bacteria and animals both in- shows that genes are transmitted verti- tain the gene for an Fn3 domain from herit the domain from some mutual an- cally, from one generation of organisms an animal source, when did it happen? cestor, or did one of the groups some- to the next. It is nonetheless sometimes All that the phylogenetic tree can indi- how acquire it from the other? If the possible for genes to be transmitted hor- cate is that it happened within the past domain was present in the common an- izontally as well, not only between spe- billion years, after the divergence of cestor of prokaryotes and eukaryotes, cies but across distant lineages. Some vi- animals from plants and fungi. To pin then why has it not been found in fun- ruses can pick up small genes from one down the date, we need to know the gi and plants? With the aid of a com- host and carry them to another host; average rates of sequence change along puter, we aligned all the Fn3 sequences in rare instances, the transposed gene both the bacterial and animal lines of that we could find and constructed a crude phylogenetic tree based on their similarities. Because trying to accom- TIR3 plish that with all 300 sequences was MKC1 cumbersome, we began by using repre- TWE4 MUSCLE sentative sets of sequences from all the TWD1 bacterial proteins and from only the TWD2 most di›erent of the animal ones. PCC2 It was soon clear that something was NCHU amiss. The bacterial sequences were IBH2 simply too similar to the animal ones to DPA1 have descended from a common ances- CLCF tor two billion years ago. Instead the evi- CHB2 dence—including the computer-generat- CHB1 ed phylogenetic trees—favored the view AME3 BACTERIA that somehow the bacteria had acquired GLC2 the Fn3 domain from an animal source. GLC3 There were several reasons to think GLC1 so. First, it was often the case that an AML2 enzyme from one bacterial species car- AML1 ried the domain but that the same en- AME2 zyme from other species did not, which implied that the domain was structural- AME1 ly and functionally expendable. It must PXE1 therefore have been a late addition to LRH1 TNH1

speciﬁc bacteria. Moreover, the Fn3 se- EXTRACELLULAR quences appeared sporadically and in FNH10 clusters but always in a characteristic UNH1 set of extracellular enzymes; if bacteria UNH2 were losing copies of Fn3 over time, one FNH9 might expect to ﬁnd surviving copies in FNH1 a more varied set of proteins. TNH7 Lastly, although the bacteria that had TNH2 Fn3 domains were of diverse types, they 45TW

did have certain features in common. OTHER PROTEINS EPH1 CYTOKINES AND All are soil bacteria that obtain their AXH1 food from commonly found polymers, such as cellulose and chitin, released CDH1 during decomposition. Vast numbers of CMM5 other types of bacteria have been ex- I2M1 amined, but none of them have the Fn3 I7H1 domain in any of their proteins. More GRBF than half the genome sequence of the KLH2 intestinal bacterium Escherichia coli is EVOLUTIONARY RELATIONS of 39 Fn3 domains from various sources appear in known, and not one hint of an Fn3 se- this phylogenetic tree. Using a computer, the authors created the tree by compar- quence has surfaced. The same can be ing the sequences of the domains from animals and bacteria. Surprisingly, the bac- said for the large number of studied terial domains are closely related to some from animals, which suggests that bac- fungal and plant sequences. If the Fn3 teria obtained the genes for those domains from animals.

TRANSFORMED BACTERIUM DEAD BACTERIUM CELL

HORIZONTAL TRANSMISSION

VERTICAL TRANSMISSION

DAUGHTER DAUGHTER DAUGHTER BACTERIA CELLS BACTERIA WITH ANIMAL GENE

HORIZONTAL TRANSMISSION OF GENES could explain how by the absorption of DNA from their environment. If a bac- bacteria obtained domains from animals. Genes pass vertically terium picked up the DNA for a domain from a dead cell, it between generations of cells. Bacteria can also be transformed might have transmitted the gene to its progeny. divergence. For the animal proteins, we eventually to catch the act of transmis- about one out of every 50 animal pro- can estimate the rate by comparing se- sion in progress by finding a phage that teins (50 out of 2,500 known animal quences from various creatures whose carries a gene for an Fn3 domain. Now sequences, independent of species re- divergence times are shown in the fos- that a number of bacterial Fn3 sequenc- dundancies). We estimate that about 25 sil record. Unfortunately, we do not have es have been found, it may be possible modules are very commonly strewn comparable information about the bac- for us to synthesize short tags of DNA throughout animal proteins, much as terial sequences. (Whereas some micro- that will bind to the Fn3 genetic units. Fn3 is. More than 100 others are found fossils corresponding to bacteria have Used in conjunction with the DNA am- in more than one setting but less fre- been reported, there is certainly not yet plification technique called the poly- quently than the first group. Tracing the an interpretable evolutionary hierarchy merase chain reaction, those tags could pedigree and diaspora of these modular for them as there is for animals.) We help identify the genes in bacteriophag- units is a major challenge that should can see, however, in both the animal es or other vectors. reveal much about every aspect of the and bacterial proteins a tendency for Of course, where the Fn3 domain evolution of all living things. tandem duplication of Fn3 sequences— originated is still a mystery. Did it first that is, within any one protein contain- appear in animals? Or are we just un- ing more than one Fn3 domain, the able to identify its ancestral forms by se- sequences of those domains are often quence comparison? Three-dimensional FURTHER READING adjacent and usually very similar. That determinations show that the Fn3 struc- THE GENEALOGY OF SOME RECENTLY observation implies that the duplication ture is suspiciously similar to immuno- EVOLVED VERTEBRATE PROTEINS. Rus- sell F. Doolittle in Trends in Biochemical of the DNA for the Fn3 domain must globulin domains. Three-dimensional Sciences, Vol. 10, No. 6, pages 233–237; be relatively recent. analyses by x-ray crystallography and June 1985. The timing of the horizontal trans- NMR have made it possible to trace the PROTEIN MODULES. Martin Baron, David mission and of the genetic duplications immunoglobulin domain back to pro- G. Norman and Iain D. Campbell in is critical to our understanding of how teins in prokaryotes, including PapD, Trends in Biochemical Sciences, Vol. 16, these genetic units spread. So far as is a “chaperone” that helps other proteins No. 1, pages 13–17; January 1991. known, contemporary bacteria do not fold, and also to a bacterial enzyme that MODULAR EXCHANGE PRINCIPLE IN PRO- TEINS. Lázló Patthy in Current Opinion have introns in their protein-coding digests cellulose. It is interesting that in Structural Biology, Vol. 1, No. 3, pag- gene sequences. If they ever did have the immunoglobulin domain, as origi- es 351–361; June 1991. introns in their coding sequences, how nally defined, contains a disulfide bond MOBILE MODULES AND MOTIFS. Peer Bork long has it been since those introns dis- that pins its two sides together, but the in Current Opinion in Structural Biology, appeared? Unless it was fairly recent- more primitive forms, some of which Vol. 2, No. 3, pages 413–421; June 1992. ly, the genes for Fn3 must have been persist in vertebrate animals today, lack PROPOSED ACQUISITION OF AN ANIMAL spreading without the aid of introns. that bond. It is these primitive forms PROTEIN DOMAIN BY BACTERIA. Peer One likely possibility is that the Fn3 that are most similar to the Fn3 domain. Bork and Russell F. Doolittle in Proceed- ings of the National Academy of Scienc- domain is being spread among soil bac- We can anticipate that other instances, Vol. 89, No. 19, pages 8990–8994; teria by a promiscuous bacteriophage es of pirated modules will be uncovered. October 1, 1992. (bacterial virus) or plasmid. We hope By our census, the Fn3 domain occurs in

56 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. Electrorheological Fluids Some liquids solidify instantly when exposed to an electric ﬁeld. Such protean materials may give engineers quicker, more adaptive machines

by Thomas C. Halsey and James E. Martin

he T-1000—a nearly indestruc- ents and more stable chemistry. Struc- the fluid is an insulator, no appreciable tible robot that was brought to tural weakness, however, remains a seri- electric current flows through it. Con- T life in the movie Terminator 2: ous impediment. If practical electrorhe- sequently, the charge on the plates rap- Judgment Day—could deftly transform ological fluids are to be designed, it is idly builds up to a high voltage, just as from liquid to solid. When pierced by a important to have a firm understand- thousands of volts of static electricity bullet, the robot’s liquid-metal skin ing of the electrical interactions among can easily accumulate on one’s skin. flowed into the hole; when broken into suspended particles that give rise to Winslow’s patented fluid was a very small pieces, the T-1000 melted and strong solids. exciting discovery, and the future of then congealed into its original form. Electrorheological fluids have fasci- electrorheological fluids looked bright. Electrorheological fluids promise some nated scientists not only because of Automotive engineers could imagine, of the adaptability that the film’s pro- their useful physical properties but also for instance, constructing an electro- ducers imagined. because of their complexity. The parti- rheological clutch. The fluid inside this An electrorheological fluid is a sub- cles in an electrorheological fluid orga- clutch would couple a car’s engine to stance whose form changes in the pres- nize themselves into a variety of intri- its drive shaft, which delivers power to ence of electric fields. Depending on the cate structures. At one extreme, when the wheels. When solidified, the fluid strength of the field to which it is sub- the fluid flows freely, the particles move jected, an electrorheological fluid can independently of one another. At the run freely like water, ooze like honey other extreme, when the fluid solidifies, or solidify like gelatin. Indeed, the sub- the particles link together to form thin ATTRACTION stance can switch from one state to an- chains and thick columns that are visi- other within a few milliseconds. Electro- ble to the naked eye. rheological fluids are easy to make; they In studying electrorheological fluids, consist of microscopic particles sus- we are thus probing the behavior of 55° pended in an insulating liquid. Yet they matter somewhere between the typical are not ready for most commercial ap- liquid and solid, quite distant from or- plications. They tend to su›er from a dinary liquid-state or solid-state phys- number of problems, including struc- ics. Recent research has provided many REPULSION REPULSION tural weakness as solids, abrasiveness new insights into the interaction be- as liquids and chemical breakdown, es- tween particles in these unusual states. pecially at high temperatures. It remains to be seen whether such find- Researchers have recently learned to ings will contribute to the design of make fluids with less abrasive ingredi- stronger fluids. ATTRACTION

nvestigators began taking an inter- INTERACTION between charged parti- est in the electrorheological e›ect cles is the key to the electrorheological THOMAS C. HALSEY and JAMES E. I around 1947, when Willis M. Wins- e›ect. A particle that has a positively MARTIN have made major contributions low obtained the first patent on an elec- charged pole and a negatively charged to the field of electrorheology during the trorheological fluid after some eight pole can either repel or attract another past few years. Since 1987 Halsey has years of observations. It was quickly dis- particle polarized in the same manner been an assistant professor of physics in covered that something as simple as (above). When particles are side by side, the James Franck Institute at the Univer- cornstarch dispersed in corn oil could the repulsion between like poles forces sity of Chicago. After earning his Ph.D. them apart. When particles align vertical- in physics from Harvard University, he make a good electrorheological fluid. This edible home recipe exhibits the ly, the attraction between unlike poles became a Sloan Research Fellow and a brings them together. In the illustration Presidential Young Investigator. In 1986 electrorheological e›ect when it is sub- at the right, a polarized particle in the he served as a visiting scientist in theo- jected to a strong electric field in the repulsive zone (orange) of another par- retical physics at the Center for Nuclear range of 1,000 volts per millimeter. The Studies in Saclay, France. Martin is a se- ticle circles around into the attractive e›ect can be produced, for example, by zone (purple). If many polarized parti- nior member of the technical sta› at placing the fluid between two metal Sandia National Laboratories. He joined cles suspended in a fluid interact in this plates one millimeter apart and charg- the sta› at Sandia in 1981 after earning way, the particles will form chains, and a Ph.D. in physical chemistry from the ing the electrodes to 1,000 volts. That the fluid will exhibit the electrorheolog- University of Washington. voltage may seem high, but the power ical e›ect; that is, the fluid will thicken requirement is actually small. Because or even solidify.

58 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. would force the drive shaft to rotate. As cause the flow rate of an electrorheo- into the shock absorber during the re- it liquefied, the fluid would allow the en- logical fluid can be controlled by the bound. Unfortunately, the viscosity of gine to disengage from the drive shaft application of an electric field, novel oil is very sensitive to temperature. On and spin freely, as if in neutral gear. concepts such as the adaptive shock a cold day the oil becomes thicker, so Such a clutch would have few parts to absorber have drawn interest from the the shock absorber is a bit sti›. Fur- wear or fail. And with a response time automotive industry as well as from thermore, because the oil heats and be- of a few milliseconds, it would make proponents of magnetically levitated comes thinner with repeated compres- purely mechanical clutches look slow. trains. (Although such trains do not sion, the shock’s damping ability nearly It was also hoped that electrorheologi- ride on rails, adaptive shock absorbers vanishes on a long, bumpy road, where cal fluids would lead to valveless hy- could damp the e›ects of quick stops it is needed most. One way to compen- draulic systems, in which solidifying and starts.) sate for these di¤culties is to introduce fluid would shut o› flow through a thin An ordinary shock absorber consists a mechanical system that adjusts the section of pipe. of a cylinder with a piston inside that size of the shock absorber’s orifice as Those ideas are still attractive, but when compressed pumps a thick, vis- the need arises. the truly exciting possibilities for elec- cous oil through a small orifice. The flu- Electrorheological fluids o›er the pos- trorheological fluids involve adaptable id absorbs the force of an impact by its sibility of a shock absorber that pro- devices that take advantage of the flu- slow, sticky motion, and a larger orifice vides response times of milliseconds id’s continuously variable viscosity. Be- then enables it to return more quickly and does not require mechanical ad-

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 59 justments. A shock absorber can be Investigators have known for some ond, and, since opposites attract, the made by filling a piston with electro- time that the origin of the electrorheo- two particles feel a net attractive force. rheological fluid and adding electrodes logical e›ect lies in a phenomenon If the first particle somehow man- and a microprocessor. As the piston termed electrical polarizability. In an ages to come alongside the second, moves, the microprocessor rapidly mod- electric field—between, say, a pair of aligning like the letters in the Roman ulates the voltage across the electrodes charged metal plates—the positively numeral II, then the negatively charged to change the fluid’s viscosity. Such charged protons inside a particle will be bottom of the first is closest to the neg- fast control would make it possible to attracted toward one electrode while the atively charged bottom of the second, increase the fluid’s thickness in mid- negatively charged electrons will be at- and likewise for the positively charged stroke, so as to damp the e›ect of a tracted toward the opposite electrode. tops. Because likes repel, the two parti- pothole. The fluid could then be made The result is that the positive and neg- cles feel a net repulsive force. Of course, thinner again for a quick recovery. A ative charges shift somewhat inside the if the particles approach each other at single shock absorber could thus adapt particle, each in an opposite direction. some other angle, the attractive or re- to a variety of vehicles and operating A particle in an electric field thus be- pulsive force is less than it is in the ex- conditions. comes an electric dipole, whose positive treme cases. In general, if a line con- Yet existing electrorheological flu- and negative poles are analogous to a necting the two particles is within 55 ids fail when subjected to heavy pun- magnet’s north and south poles. The degrees of the vertical, the particles at- ishment. The compounds are often too strength of this e›ect varies according tract; otherwise, they repel. soft as solids, and the high voltages re- to the electrical polarizability of the sub- Even if initially repelled, the mobile quired to solidify them, if frequently stance. If the polarizability is large, then particles tend to circle around one an- applied, can heat volatile liquid ingredi- large amounts of electric charge shift other. Their trajectories curve, like those ents and eventually boil them away. To over large distances toward the parti- of comets, in response to the forces at- address these problems, we are investi- cles’ opposite poles. If the polarizabili- tracting or repelling them. At the same gating how the electric field causes soli- ty is small, then the particles hardly re- time, no matter in which direction the dification. Ideally, we would like to have spond at all to an applied electric field. particles are headed, their poles point fluids that become strong solids at low- either up or down toward the elec- er voltages than are now needed. hen the fluid contains many trodes. Together these forces bring vir- polarized particles, the electri- tually all particles into a configuration W cal interactions become com- in which oppositely charged poles are plicated. The simplest situation in- close together [see illustration on pre- volves two particles in a strong electric ceding page]. Then the attractive forces field, each particle having its positive take over, and the particles stick to- pole on top, for argument’s sake, and gether. Successive particles line up end its negative pole on the bottom. The to end in much the same way to create force of electrical attraction or repul- long chains, like pearls on a string. This sion between these two particles de- behavior is analogous to the familiar pends on the way they approach each alignment of iron filings along the field other. If the first moves up directly be- lines of a bar magnet. The chains of neath the second, then the positively particles in an electrorheological fluid charged top of the first is closest to the quickly grow from one end of the fluid negatively charged bottom of the sec- container to the other. Formation of such chains is responsible for the solidification of the fluid. Some particles readily form chains LASER when suspended in one kind of fluid BEAM but not when suspended in another, and researchers are still working to sort out the messy details. One potential complication is that the particles are immersed in liquid whose molecules can also be polarized. If the polarizabil- CHAIN ity of the liquid is exactly the same as COLUMN that of the particles, then the polarization of the liquid masks that of the particles, and chains will not form. In general, electrorheological fluids work best when the particles polarize more easily than do the molecules of the fluid. Still, chains are only an intermediate LASER LIGHT shining through a solidi- step in the evolution of structure in fied electrorheological fluid produces a electrorheological fluids. Single chains pattern that o›ers clues about how the slowly migrate toward nearby parallel particles of the fluid are organized. In chains to combine as thicker columns, this experiment each particle acts like a lens because each is made of glass which combine to form still thicker col- and is much larger than the wavelength umns, and so on. What is the ultimate of light. A pattern of circles implies a solid structure? Recent theoretical work three-dimensional structure known as the and experimental observations have re- body-centered tetragonal lattice (above). vealed the answer.

60 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. First, an analysis of the complex elec- ern Illinois group was able to infer that like a figure eight. At first, the figure- tric fields around large numbers of par- the structure of their solidified electro- eight pattern is large and di›use. As ticles o›ers clues about the structure rheological fluid was a common, order- the columns thicken and the distance we might expect. As we have already ly arrangement of spheres (a three-di- between them increases, the light trav- seen, dipolar particles repel one anoth- mensional lattice) known as the body- eling through the fluid encounters fewer when side by side, so the appearance centered tetragonal structure, in which er obstacles, so scattering decreases and of columns made of parallel chains at the particles in adjacent chains are in- the lobes in the figure eight become first seems mysterious. Why don’t such deed staggered. smaller and brighter. columns disintegrate into their con- Although the attractive forces be- This rough rule lets us interpret the stituent chains? This does not happen tween chains are strong enough to cre- patterns of light seen in our experi- because a particle is attracted to or ate such a lattice, they are too weak to ments. When the electric field is o›, and repelled from a neighboring chain not account for the relatively rapid aggre- there are no chains or columns in the only by the force from the few nearby gation (over a few minutes) observed sample, the light is scattered uniformly. chained particles but also by the force by the Southern Illinois group. Reality, But when the field is turned on, the light from many distant particles in the chain. it seems, is more complicated than any exiting the sample begins to concen- Columns stay together because the di- neat analysis of electrical forces. trate into two lobes. As time progresses, polar force is long-range, and particles In experiments, columns form faster the light becomes brighter near the cen- respond to the sum of forces from all than expected because the chains have ter of the figure eight. This brightening other dipoles in the fluid. imperfections. For at least two reasons, indicates less scattering as the columns To understand why the columns particle chains are not perfectly straight, thicken and more space opens up be- hold together, consider a “test” particle ordered structures. First, chains some- tween them. Column growth is initial- placed near a chain. The test particle will times fail to span the entire distance ly very fast, but after a few minutes its be repelled by the nearest particle in the between the electrodes, and, second, progress virtually stops, because the chain, just as side-by-side particles re- they are shaped, in part, by the Brown- attractive force between the columns is pel one another. But the test particle ian motion of the particles. Browni- too weak for continued growth. Because will be attracted by the chained parti- an motion, first observed in 1827 by calculations show that the solidifying cles well above and below it because the Scottish botanist Robert Brown, is a fluid may get stronger as the columns the test particle will be in the 55-degree ceaseless random motion of particles attractive zone of most of the particles suspended in a fluid; it is caused by from the top and bottom of the chain. collisions between a particle and hordes To be even more precise, the test parti- of fluid molecules. In an electrorheologi- cle is drawn to a position slightly above cal fluid the suspended particles dance one particle in the chain and slightly chaotically about their average position below another. Consequently, if many in the chain as they are struck by fluid test particles are placed near the chain, molecules hitting them from every di- they will be attracted to it and each rection. Thus, although the chain may other and can form a second chain and be straight on average, at any given time a third, and so on. In this way, the di- it is kinked by Brownian motion. This polar forces between particles encour- slight deformation increases the force age the chains to form columns in of attraction between chains and cat- which the particles in adjacent chains alyzes the aggregation of chains into are staggered. columns. To obtain a better understanding of ian-jie Chen, Robert N. Zitter and column growth, the authors, along with LASER Rongjia Tao of Southern Illinois Judy Odinek of Sandia National Labo- BEAM T University recently confirmed the ratories, employed an optical method existence of such a structure. In their quite di›erent from the technique de- experiment, they used an electrorheo- veloped by Chen, Zitter and Tao. Instead logical fluid made of glass spheres with of directing light through chains made a relatively large diameter of 40 mi- from relatively large glass spheres, we crons, suspended in a silicone oil. Im- aimed a laser beam in a direction per- mediately after the electric field was pendicular to the chains and used much COLUMN applied, chains formed. Within a few smaller spheres—less than a nanome- minutes the chains aggregated into col- ter in diameter. Light scattered from umns with a roughly circular cross sec- columns made of such particles carries CHAINS tion approximately six millimeters in with it information concerning the spac- diameter. ing and width of the columns. In gen- To determine the structure within eral, the exiting light makes a pattern these columns, the investigators utilized the focusing properties of the glass spheres, treating the columns as stacks FIGURE-EIGHT PATTERN appears when of lenses that could transmit light. Light light penetrates an electrorheological flu- propagated from sphere to sphere un- id consisting of particles approximately til it exited from the column. The last as small as the wavelength of light. This layer of spheres acted as a set of light pattern indicates that the particles have sources and created a pattern of over- organized themselves into columns. As lapping spots [see illustration on page the distance between the columns in- 60]. By analyzing this pattern, the South- creases, the figure eight becomes smaller.

SLIP ZONE

SHEAR SHEAR

NO FORCE MODERATE FORCE EXTREME FORCE

SHEAR SHEAR

NO FORCE MODERATE FORCE EXTREME FORCE TWO THEORIES describe how chains of particles might inhibit center of the fluid, and this “slip zone” grows as the shearing the flow of electrorheological fluids. In the Klingenberg and Zu- force increases. In the theory of Halsey, Martin and Adolf koski theory (top row), chains bend and stretch as their ends (bottom row), chains float throughout the fluid. As the shear- adhere to shearing electrodes. The chains finally break in the ing force increases, the chains tilt and fragment. thicken, these measurements demon- Bingham plastic does not flow at all. For applications such as the elec- strate that keeping the electric field on Beyond this threshold, the flow rate in- trorheological clutch, maximizing the for periods up to several minutes may creases in proportion to the applied yield stress is very important. A fluid improve the mechanical properties of stress minus the threshold stress. Most with a high yield stress will withstand a electrorheological fluids. of us encounter a Bingham plastic ev- high torque, or rotational force, before These fluids, after all, need to with- ery morning, in the form of toothpaste. flow begins and the clutch slips. One stand mechanical forces. Knowing their This type of rheology creates the “plug” way to increase the fluid’s resistance to static structure is not enough; we must flow that enables us to squeeze a neat stress is to apply a higher voltage. The also understand how the structure be- cylindrical plug of nonrunning paste amount of separation achieved between haves when squeezed or stretched. Such onto a toothbrush. Like toothpaste, elec- positive and negative charges on the issues fall into the domain of rheology, trorheological fluids resist a certain particle dipoles increases in proportion which is the study of the flow proper- amount of stress but begin to flow when to the voltage applied, and the attrac- ties of materials. (Almost all materials, more force is applied. tive force between particles goes up as even solids, flow under su¤ciently high the square of this charge separation. pressure, although sometimes they do tress often builds up in an elec- The yield stress is therefore propor- so by imperceptible amounts.) trorheological fluid when the elec- tional to the square of the applied volt- In general, the sti›est materials flow Strodes start to move—one plate age. Increasing the voltage, however, very little, whereas the weakest ones toward the right, say, and the other to- also increases power consumption and flow at the slightest provocation. One ward the left. This is called a shearing heats the electrorheological fluid. If the variable a›ecting the flow rate of a flu- force. If we apply a small force, the fluid gets too hot, it might decompose. id is the stress, or force, applied to it. chains of particles in the fluid will tilt Hence, in designing strong electrorheo- For simple fluids such as water, the flow and stretch but not break. Yet when the logical fluids, one goal is to minimize rate is proportional to the stress: water chains are tilted, the attractive force be- the voltage needed to achieve a given flows faster when you pump harder. On tween particles is diminished, because yield stress. the other hand, the flow rate is inverse- the chains are no longer aligned straight The voltage needed depends largely ly proportional to the viscosity, or thick- up and down in the electric field. In- on particle polarizability. A few years ness, of the fluid. Viscous fluids such as stead the chained particles move to- ago it was thought that the strength of thick oil flow slowly at a given stress, ward a slanted configuration that gets a solidified fluid would not increase dra- compared with water, for instance. ever closer to forcing the particles side matically with polarizability. We know But electrorheological fluids in an by side. If the chains tilt too far, the better now, thanks to studies of high- electric field are not so simple. Their particles slip out of the zone in which ly polarizable ceramic particles con- behavior resembles that of a family of they mutually attract, the chains break ducted by Douglas B. Adolf, Robert A. materials known as Bingham plastics. and the material flows. The amount of Anderson, Terry J. Garino and Bradley Below a certain threshold of stress, a stress at failure is called the yield stress. G. Hance of Sandia National Laborato-

62 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. ries. The investigators have shown that are examining how these structures in- strength grows dramatically as polar- hibit the flow of the liquid in which they ization increases. There is much opti- are suspended and how this flow then mism that highly polarizable particles distorts and weakens the solid struc- will lead to very strong electrorheologi- tures so that the liquid keeps flowing. cal fluids. So far there is no comprehensive the- Electrical conductivity is also a con- ory. One group has proposed a model cern. Di›erences in the conductivity of of electrorheological flow based on ob- the particle and the liquid solvent are servations of a cornstarch fluid, where- especially significant if the applied elec- as our group, experimenting with other tric field is on for more than a hun- fluids, has a di›erent explanation for dredth of a second or so. Consider non- their thick, resistant flow. conducting particles suspended in a Daniel J. Klingenberg, now at the Uni- weakly conducting liquid. Immediately versity of Wisconsin, and Charles F. Zu- after the field is turned on, the particles koski of the University of Illinois stud- will polarize. But charges will also begin ied the flow of a cornstarch electrorhe- to flow through the liquid and will accu- ological fluid at stresses larger than the mulate on the surfaces of the polarized yield stress. They found that shearing particles. Shortly thereafter, the inter- the electrodes gave rise to two kinds actions between particles will be deter- of structure. Near the plates, tilted col- mined not so much by their intrinsic umns of particles formed. In the central polarization as by clouds of charge en- region between plates, a “slip zone” de- circling each particle in the liquid. The veloped in which the chains were bro- basic force between particles is still di- ken and the fluid moved freely. As the polar, but the microscopic origin of this rate of shearing increased, the slip zone force is di›erent, and the yield stress increased in width and lubricated the may be di›erent. relative motion of the tilted columns Other problems arise with conducting adhering to the plates. particles in an insulating liquid. When Working with Adolf, we have pro- conducting particles polarize, negative- posed an alternative picture of shear- ly charged electrons do not just shift ing in electrorheological fluids in which inside molecules but actually migrate the adhesion of columns to the elec- from molecule to molecule across the trodes is not dominant. If the columns particle, leading to a strong polariza- are free to drift, the attractive forces tion and strong attractive forces. But if between particles tend to align the col- the particles contact each other, charge umns perpendicular to the electrodes. flows between them. In that case, one Yet the flowing fluid tries to rotate them of the particles will have a net positive so that they are parallel to the elec- charge, and the other will be negatively trodes and the direction of flow. These charged. Instead of linking together, forces arising from the fluid flow are these particles will migrate toward op- called viscous forces. At some angle of posite electrodes, since the electrodes column tilting, the viscous forces that exert a more powerful pull than that of are trying to turn chains parallel to the any single small particle. This phenom- flow will balance the electrical forces enon has been seen in suspensions of that are attempting to keep them per- conducting particles and is colloquial- pendicular. We expect to see such tilt- ly referred to as the “bouncing ball” ef- ed columns throughout the fluid as it fect. The exchange of charges between yields to shearing stress and begins conducting particles can be prevented to flow. by coating the particles with a thin in- We also expect the columns or chains sulating layer. Such coatings allow the to shorten as the rate of shearing in- particles to become highly polarized, creases. Because a long chain has more and the result is that the fluid displays surface area than a short one, the vis- a strong electrorheological e›ect. cous forces act more strongly on a long chain. In fact, the strong viscous forc- high yield stress is not the only es will rotate very long chains beyond useful property of an electrorhe- the angle at which the particles attract A ological fluid. In some applica- one another. Therefore, the chains will tions, the fluid should yield at a low break up as the particles begin to re- stress, like toothpaste, yet continue to pel. This phenomenon sets an upper resist motion as the stress increases. limit on the length of the chains that SOLIDIFICATION occurs in stages when To control such a fluid’s varying thick- can appear at a given shearing rate in an electric field is applied to a disor- ness, we need to learn more about the a given electric field. These relatively dered fluid (top). Particles take only mil- behavior of electrorheological fluids af- short chains will nevertheless dramati- liseconds to form long “pearl” chains ter they yield and begin to flow. We and cally increase the viscosity of the fluid. (middle). A few seconds later the chains several other research teams are there- The situation is analogous to pumping can be seen aggregating into columns fore investigating what kinds of struc- water through a pipe filled with pine (bottom). These photographs show sili- tures evolve as fluids are stressed. We needles. The fact that the water must ca particles one nanometer in diameter.

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 63 ULTIMATE IN ADAPTIVE MATERIALS was portrayed in Ter- changing readily from solid to liquid. The technology of elec- minator 2. The Terminator camouflages itself as a tile floor, trorheological fluids is, of course, far less sophisticated. push its way through a fine mesh in- the electrorheological e›ect increases chanical controls will be crucial. A ro- creases the force necessary to achieve with particle size, large particles—with bot agile enough to perform feats that a certain flow rate, so water combined diameters much greater than one nano- are fairly routine for human beings— with pine needles has a high viscosity, meter—are preferred. But large parti- say, catching a baseball—may need a just as electrorheological fluids do be- cles tend to sink to the bottom of the hydraulic system that takes advantage yond the yield stress. fluid. (Also, fluids containing large par- of the rapid response of electrorheo- The behavior of many electrorheolog- ticles are more abrasive than are sus- logical fluids. ical fluids can probably be explained by pensions of smaller particles.) The sec- We do not need robots to replace hu- a combination of the model of Kling- ond problem is that an activator such man athletes, but imagine, for instance, enberg and our theory. Near the elec- as water tends to evaporate or undergo a computer-controlled device that could trodes, a relatively quiescent region sim- chemical reactions with the other in- extrude the fine metal wire used to con- ilar to the rest state of the solidified gredients, leading to a short operating nect integrated circuits. When winding electrorheological lattice will form. In life, especially at high temperatures. this delicate wire on a spool, the robot the intermediate region, this structure would have to control tension precise- will break up into shorter, tilted chains he first generation of electrorheo- ly: the wire might droop or break unless of particles. logical fluids, developed by Wins- the machine quickly sensed the slight- Electrorheological fluids today are no T low, was kept moist by exposure est change in tension and responded. longer the fairly simple mixtures they to humid air and contained gritty silica Such a device, with a fast electrorheo- were in the 1940s. In addition to the particles. About 10 years ago James E. logical clutch, has already been test- suspending liquid and the particles, Stangroom and a group at the Cran- ed in prototype. Others may soon fol- they often contain a dispersant and field Institute of Technology in England low—machines, in e›ect, with super- an activator. The dispersant keeps the developed a second generation of less human “joints” that instantly lock or particles from sticking to one another abrasive fluids with particles made of loosen with extremely high precision. in the absence of an electric field. With- polymers but still containing water. They would be no match, of course, for out the dispersant, particles sometimes The next step, which is now under way, the T-1000 Terminator but would be clump together or fail to disperse even is to create a third generation of water- tough competitors in the technological when the strength of the electric field free fluids with small, highly polariz- marketplace. is zero. Such temperamental fluids are able particles. of little use. The activator, which is of- Meanwhile the machines that might ten water or sometimes alcohol, con- use electrorheological devices have also FURTHER READING tains impurities—often dissolved salts. changed since the 1940s. The auto- The mechanism by which the activator motive industry is still interested in INDUCED FIBRATION OF SUSPENSIONS. W. M. Winslow in Journal of Applied works is not entirely clear, but it is gen- adaptive materials with millisecond re- Physics, Vol. 20, No. 12, pages 1137– erally believed that water, repelled by sponse times, but so is another industry 1140; December 1949. an oily suspending liquid, collects on that did not exist half a century ago. To- ELECTRORHEOLOGICAL FLUIDS AS COL- the particle surfaces. The dissolved salts day simple robots do much of our in- LOIDAL SUSPENSIONS. Alice P. Gast and then polarize in an electric field, add- dustrial work. If the recent revolution C. F. Zukoski in Advances in Colloid and ing their charge to the particles’ intrin- in semiconductor technology is to lead Interface Science, Vol. 30, Nos. 3–4, sic polarization. to the robotics revolution envisioned pages 153–202; December 1989. These standard formulations have two by several generations of science-fiction ELECTRORHEOLOGICAL FLUIDS. T. C. Hal- sey in Science, Vol. 258, pages 761–766; general problems, aside from structural writers, adaptive materials that couple October 30, 1992. weakness. First, because the strength of fast microprocessors to equally fast me-

64 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. Water-Pollinated Plants Once thought to be mere aberrations of nature, these ﬂowering aquatic species provide evidence for the evolutionary convergence toward e¤cient pollination strategies

by Paul Alan Cox

n a sunny October day in 1787, tion for flowering plants on land. These that reinvaded the water. Hydrophilous the botanist Filippo Cavolini plants use wind and animals to dissem- plants thus retain many features of the Orowed a boat in the Bay of Na- inate their gametes over dry terrain. reproductive systems of their terres- ples, continuing a pursuit that had so But Cavolini’s finding of water as a trial counterparts, much as whales and far proved futile. Four years previous- pollination vector has recently been re- porpoises kept their mammalian traits ly, divers had brought him a species considered in light of search theory and after their return to the ocean. of eelgrass, Zostera marina, that bore di›usion physics. The results indicate Water pollination should not be con- fruit. Cavolini had immediately realized that hydrophilous, or water-based, pol- fused with reproduction in such non- that the plant was special: it must flow- lination systems cannot be considered flowering plants as algae, ferns, mosses er, unlike the seaweeds and other ma- mere quirks of natural history, as pre- and liverworts. Although these species rine algae he had collected. But he had viously thought. Rather water-pollinat- require water—say, dewdrops or water- yet to discover its blossoms. Cavoli- ed plants should now be viewed as com- fall spray—they do not produce pollen; ni’s quest finally ended that day as he pelling cases of convergent evolution, they simply disperse their motile ga- reached a Zostera population that grew because unrelated hydrophilous species metes into the water. Nor is hydrophi- in the bay. He became tremendously have evolved similar strategies that pro- lous pollination similar to the aerial excited to find “a spike [flower cluster] mote e¤cient pollination. Although they pollination of other aquatic flowering not unlike that of a grass floating on are uniquely adapted to their watery plants, such as water lilies, that raise the water.” environment, their ancestors can be their flowers above the surface to where Later, examining the pollen under ultimately traced to flowering plants insects or wind can pollinate them. his laboratory microscope, Cavolini was stunned to see that the grains were not small and round, like those of terrestri- MALE FLOWER of the surfgrass Phyllospadix scouleri (left) releases filamentous al plants: “I found it [the Zostera pol- pollen that forms snowflakelike rafts. Currents carry the pollen mass to female len] di›erent from that of other plants, flowers (right), where the pollen grains contact the stigma, resulting in fertilization. being oblong like little eels, which, with a sudden and brisk motion, exploded and scattered their sperm in a twin- kling of an eye.” Based on this observation, he drew a startling conclusion: Zostera must reproduce by pollination in the open water. Cavolini’s discovery was remarkable. Because water causes the pollen of terrestrial plants to burst, botanists had regarded the aquatic environment as inimical to pollination. Pollen, which transports and protects a plant’s genetic information, is a necessary adapta-

PAUL ALAN COX is dean of general and honors education and professor of botany at Brigham Young University, where he received his B.S. After earning his master’s at the University of Wales and a Ph.D. from Harvard University, he was awarded a Miller Research Fellowship at the University of California, Berkeley, and later a Melbourne Research Fellow- ship at the University of Melbourne in Australia. He enjoys backpacking, skiing and rafting with his family.

Copyright 1993 Scientific American, Inc. Not uncommon taxonomically or ge- the existence of God in his influential ographically, water-pollinated species book Natural Theology.) Darwin found occur in 31 genera in 11 di›erent fami- that the submerged male plants release lies. They range in latitude from north- flowers that float to the surface. Once ern Sweden to southern Argentina and there, the petals recurve, elevating the in altitude from 40 meters below sea flower on a meniscus of water that en- level to 4,800 meters high in the An- ables it to glide with the slightest breeze. des, where Elodea potamogeton impedes Two sterile anthers function as tiny navigation on Lake Titicaca on the Peru- sails, while two fertile anthers hold the Bolivia border. They are ecologically di- spherical pollen of Vallisneria aloft. Each verse as well, extending from tropical floating female flower, anchored by a rain forests to seasonal desert pools. long stalk to the bottom of the pond, Freshwater species feed fish and wild- creates a slight depression in the water’s fowl, whereas marine species support surface. As the male flower falls into the sea turtles, manatees and shellfish. depression created by the female flower, it transfers pollen to the stigma, thus ow do hydrophilous plants suc- e›ecting pollination. cessfully use water as a pollina- In 1792 Cavolini expanded on his ear- Htion vector? Water would ap- lier studies of Zostera, showing that pear to be a poor medium in which to another mechanism—directly releasing disperse pollen. In addition to its pro- pollen into the water—existed for flow- pensity to cause most pollen grains to ering aquatic plants. Cavolini also re- burst, water can be unpredictable. Cur- ported a second sea-grass species, Cy- rents can be quite erratic, and tides can modocea nodosa, that releases its noo- abruptly submerge plants or leave them dlelike pollen directly onto the surface far above the surface. of the water. He wrote that “white cap- Historical accounts provide some in- illary pollen...covered for a great space formation about how the plants exploit- the surface of the water around.” Such ed the water. Erasmus Darwin, grand- pollen was protected from bursting be- father of Charles Darwin, described the cause it could rapidly equilibrate with remarkable pollination system of the the seawater. freshwater plant Vallisneria more than two centuries ago. (Darwin’s description was so compelling that the Reverend William Paley cited it as evidence for

Copyright 1993 Scientific American, Inc. path and A is the search area.) Koop- STAMEN man’s equation had immediate military significance: even a modest increase in FLOWER the range of a ship-borne radar dramatically increases the probability of detecting a fixed target during a random search. Koopman’s equation applies not only FEMALE MALE to ships but also to pollen grains in the hunt for stigmas. The equation makes it possible to set a lower bound on the search e›iciency of pollen grains if the grains do not trace completely random paths. If the probability of encountering a target is low (a reasonable assumption in aquatic pollination), then a small increase in the size of the search vehicle will substantially heighten the VALLISNERIA probability of hitting a stigma lying on the surface. ABOVE-SURFACE POLLINATION is one reproductive strategy that hydrophilous, or The evolutionary significance of Koop- water-pollinated, species rely on. The pollen grains are kept above the water, man’s equation is clear for Vallisneria, where they are transported via floating male flowers that plow into female flowers the plant studied by Erasmus Darwin. to effect pollination. Examples are Enhalus, Lagarosiphon and Vallisneria. Here the floating male flowers sweep a path several thousand times wider than an individual pollen grain can, thus in- Darwin’s and Cavolini’s observations of search theory, first developed dur- creasing search e¤ciency. A similar pol- received only passing interest from oth- ing World War II by the mathematician lination system occurs in a di›erent er investigators, who continued to con- Bernard O. Koopman and his colleagues species: the African pondweed Lagaro- centrate on animal and wind pollina- at the Antisubmarine Warfare Opera- siphon, studied by C.D.K. Cook of the tion. Only in the past two decades have tions Group of the U.S. Navy. In a sim- University of Zurich. As in Vallisneria, botanists asked broader evolutionary ple but elegant derivation, Koopman male flowers are released underwater questions about hydrophilous pollina- showed that if a search vehicle traces a and float to the surface. There three tion. Are there general principles that somewhat random path in two dimen- sterile stamens become erect, function- apply to all hydrophilous pollination sions, its probability of encountering a ing like miniature sails. The long, fer- systems? What features should charac- fixed target rapidly increases with the tile stamens, which unfold parallel to terize e¤cient systems for pollination width of the path it traces. (Mathemati- the water’s surface, bear sticky pollen on a two-dimensional plane such as the cally, the relation is p = 1 – e–wL/A,where at the end of the ramrodlike filaments. surface of water? What shape and size p is the probability of hitting a target, e The male flowers link together in large should the pollen be? Should pollen be is the base of natural logarithms, w is rafts, colliding with the female flowers dispersed singly or in groups? the width of the path swept by the floating on the water. Such questions fall within the realm search vehicle, L is the length of the But what about such plants as the one Cavolini studied, Z. marina, which releases its pollen directly onto the water’s surface? Are the predictions of wide search paths likewise verified? To STIGMA answer these questions, R. Bruce Knox of the University of Melbourne in Aus- tralia and I examined the pollination POLLEN biology of several genera of hydrophilous plants in the South Pacific. One MALE FEMALE sea grass studied was Halodule pinifolia, which grows along sandy coasts in the ANTHER Fiji Islands. While conducting fieldwork with P. B. Tomlinson of Harvard Univer- sity and John A. West of the University of California at Berkeley, I found entire Halodule populations to be almost completely exposed during the low spring HALODULE tides. Then, the white stamens of Halo- dule slowly release the floating long, noodlelike pollen onto the surface of the sea. The pollen noodles have coatings of POLLEN ON WATER’S SURFACE is a second method. To form pollen rafts, some proteins and carbohydrates that make species release filamentous or oval pollen; others have spherical pollen embedded them sticky. They adhere to one anoth- in long tubes. Stigmas tend to be threadlike in marine species such as Amphibolis er, forming rafts that resemble snow- and Halodule and cuplike in freshwater ones such as Ruppia and Lepilaena. flakes. Millions of these floral search ve-

70 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. hicles are carried along as the tide returns to the shallow pools where the filamentous stigmas of the female plants float. The collision of the search vehicles with the stigmas on the water’s surface MALE FEMALE results in pollination. Many other sea-grass genera also assemble search vehicles that widen the search path, although the production varies slightly. In Amphibolis in Austra- lia and Thalassodendron in Kenya, the male flowers float to the surface and there release filamentous pollen that assembles into latticelike rafts. Some unrelated sea-grass genera lack noodlelike pollen but compensate in other ways. In Fiji the small oval pollen grains of Halophila ovalis do not form rafts POLLEN but are embedded inside sticky muci- THALASSIA laginous tubes that assemble into rafts, appearing as floating feathers. In Aus- UNDERWATER POLLINATION is a third method of reproduction. In some species, tralia the freshwater species Lepilaena filamentous pollen or pollen grains in mucilaginous strands are carried underwater cylindrocarpa releases spherical pollen by currents. In others, pollen denser than water is released. Examples include Tha- in floating mucilaginous mats that re- lassia, Syringodium and a species of the freshwater genus Lepilaena. semble tiny omelettes. These mats fall into depressions caused by the female flowers. operate in three dimensions—that is, dom hunt is far more e¤cient in two Thus, it is clear that many hydroph- below the surface [see “Submarine Pol- dimensions because the search path is ilous plants—Vallisneria, Halodule, Hal- lination,” by John Pettitt, Sophie Duck- recurrent: any stigma on the plane will ophila and Lepilaena, among others— er and Bruce Knox; SCIENTIFIC AMER- eventually be hit given enough time. together represent an example of con- ICAN, March 1981]. The same cannot be said for a three-di- vergent evolution: unrelated species Pollination strategies in three dimen- mensional search, even given an infi- have evolved similar means (floating sions seem to be less common than the nite amount of time [see “Brownian Mo- search vehicles) that serve to increase two-dimensional varieties, probably be- tion and Potential Theory,” by R. Hersch pollination e¤ciency. Of course, some cause they are intrinsically less e¤- and R. J. Griego; SCIENTIFIC AMERICAN, hydrophilous species have unique ad- cient. In three dimensions, pollen and March 1969]. aptations, mechanisms that perhaps are stigmas are dispersed throughout a vol- Three-dimensional pollination has indicative of evolutionary whimsy. What ume rather than concentrated on a sin- tended to evolve in hydrophilous plants military strategist could fail to be in- gle plane (such as the water’s surface), that ordinarily would have little oppor- trigued by the horned pondweed Zan- and so more pollen is needed. The con- tunity to engage in two-dimensional pol- nichellia palustris? Cook and I and our sequences of random motion also dif- lination—that is, in plants that are al- respective students You Hao Guo and fer greatly in three dimensions. A ran- ways submerged. An example is the tur- Rebecca Sperry found Zannichellia to form ominous floating clouds that slowly drop round pollen grains onto the stigmas below. Fortunately, in this case, being hit by such a botanical depth charge results not in flower destruction POLLEN but plant reproduction. And what golf- er could fail to be delighted by the pollination system of another freshwater MALE FEMALE plant, Hydrilla verticillata, also studied by Cook? Here, male flowers float to the water surface and explosively catapult their pollen through the air to the floating female flowers. Because Hydrilla pollen is destroyed by contact with water, only the holes-in-one count. Given the vast amount of pollen released, the probability that a stigma is struck is actually fairly high.

o far I have discussed the pollina- PHYLLOSPADIX tion systems of plants whose pol- Slen is transported above or on the MIXED MODE of pollen transport, used by Phyllospadix and Zostera, combines sur- surface of the water. In such cases, the face with submarine pollination. Surface pollen tends to be slightly more filamen- pollen moves in two dimensions. Some tous than submarine pollen. Underwater delivery occurs in subtidal zones—that is, species have pollination systems that in populations that never reach the surface, even during low tide.

250

200

150

100 NUMBER OF HITS PER 10,000 TRIALS 50

0 DISKLIKE SPHERE RODLIKE SHAPES SHAPES OPTIMAL POLLEN SHAPE for underwater pollination was or as disks. The rods proved best at striking a target and determined by analysis on a supercomputer. Spheres were compare well with actual submarine pollen grains, such as deformed into ellipses of equal volume, either as long rods those of Phyllospadix (photograph). tle grass Thalassia testudinum, a Carib- now at Princeton, David A. Olson, then below the water’s surface also revealed bean plant that Tomlinson, Thomas Elm- at New York University’s Courant Insti- pollen, indicating that submarine polli- qvist of Umea˚ University in Sweden and tute of Mathematical Sciences, and I nation can take place. I studied on the island of St. Croix. Tha- confirmed these numerical experiments Yet another example of mixed sur- lassia releases its round pollen under- analytically. The equations also revealed face and submarine pollination occurs water, embedded in elongated strands something the supercomputer experi- in the plant originally studied by Ca- of mucilage. Having been driven under- ments had missed: shapes that are only volini: Z. marina. A.W.A.M. de Cock of water by waves, the strands collide there slightly disklike are the least e›icient Catholic University in the Netherlands with the sti›, bristly stigmas of the fe- at three-dimensional search. Still, the found that although Zostera pollen re- male plants. A similar underwater pol- general conclusion holds. The search leased underwater in aquariums slowly lination mechanism of embedding round vehicles of T. testudinum in the Carib- sinks, pollen dispersed at low tide in pollen on long, mucilaginous strands bean Ocean and L. bilocularis in New intertidal populations floats on the sur- is a feature of an unrelated freshwater Zealand—each consisting of pollen at- face. Roger H. Laushman of Oberlin Col- species: Lepilaena bilocularis, found at tached to a mucilaginous strand—rep- lege, Mary H. Ruckelshaus of the Uni- Lake Ellesmere on South Island of New resent highly e¤cient solutions to the versity of Washington and I, working Zealand. problems of locating objects scattered at Friday Harbor Marine Laboratories I wanted to see if this pearls-on-a- in three dimensions. in the San Juan Islands, o› Washington string approach by these geographical- Some hydrophilous species mix both State, confirmed de Cock’s report. We ly remote and taxonomically distant two- and three-dimensional strategies. found that pollen rafts resembling snow- species represented an e¤cient three- For example, Enhalus acoroides, a sea flakes float on the surface; however, we dimensional strategy. Together with my grass I studied in the Indonesian island also observed linear bundles of noodle- mathematics students Tyler Jarvis, now of Banda, flowers only during low spring shaped pollen dispersed beneath the at Princeton University, and Scott Cro- tides. Like Vallisneria, the male flower surface of the sea. mar, now at Rutgers University, I ran buds float to the water’s surface, open a series of numerical experiments on and are captured within the rumpled ¤cient search methods are not a supercomputer at the IBM Scientific petals of the female flowers. As the tide the only way to increase the Computing Center in Palo Alto, Calif. We rises, the petals of the female flowers Echances of pollination. Evolution deformed spheres elliptically to varying close, trapping the male flowers within. can also select for stigmas that can degrees while keeping their volume con- But Enhalus disperses some pollen di- readily be found. Unfortunately, the de- stant, thus producing a variety of long rectly on the water’s surface. sign of targets that could easily be hit strings or flattened Frisbee-like disks. The surfgrass Phyllospadix scouleri was not considered a high-priority task We then determined how frequently also disperses noodlelike pollen both by the military strategists who devel- each shape encountered randomly po- on and below the surface. Studying a oped search theory. So, much of the sitioned targets. In our simulations we population along the northern Califor- analysis of target design has fallen to found that all the elliptical deformations nia coast, Tomlinson, my student Kevin botanists. were more e¤cient in three-dimensional Nieznanski and I observed floating pol- From the stigma’s point of view, searches than were the spheres [see il- len search vehicles of Phyllospadix col- there are at least two possible ways to lustration above]. lide with female stigmas. But electron heighten the chances of being struck Recently my student David Smith, micrographs of stigmas collected far by a pollen grain or pollen raft. First,

500

375

250 NUMBER OF POLLINATIONS

125

0 1 2 3 4 5 6 7 8 9 10 STIGMA LENGTH (ARBITARY UNITS)

OPTIMAL STIGMA SHAPE in surface pollination was deter- frequently than they did with the round stigmas. The results mined by deforming spheres into increasingly longer ellipses. compare favorably with stigmas of real plants, such as that Pollen grains collided with the long, elliptical stigmas more of Zostera marina (photograph, with pollen strands clinging). the e›ective area of the target can be percomputer simulations, James A. Se- anther to the stigma of the same flow- broadened. Second, the stigma’s shape thian of Berkeley and I found that in two er. Genetic homogeneity could also re- can be altered to increase the probabil- dimensions, filamentous stigmas are sult from infrequent flowering and large ity of encounter. far more likely to be hit than are circu- clone sizes in a population. The floral structures of many water- lar stigmas. The extremely long, thread- pollinated plants embody the first so- like stigmas of the sea-grass genera lthough the mechanics of water lution. Some blossoms, such as those Halodule, Halophila and Thalassoden- pollination has not been able to of Vallisneria and Enhalus, create de- dron seem to confirm the prediction A detail the genetics of hydroph- pressions in the water’s surface, form- [see illustration above]. ilous populations, it does o›er clues to ing a target area much larger than a Field observations also contradict the explain why such plants followed an stigma or even an entire female flower. assumptions made by some previous evolutionary path di›erent from their Another example is provided by Lepi- workers, who thought that hydrophi- terrestrial counterparts. The noodlelike laena cylindrocarpa in Australia. Usher lous plants rarely flower or pollinate. I shape of the pollen is one distinctive Posluszny of the University of Guelph have found that flowering and cross- feature of hydrophilous plants. To study pointed out to me a small bract that pollination (the transfer of pollen from the emergence of such pollen, C. J. holds three separate stigmas together one individual plant to another) to be Humphries of the London Natural His- before fertilization. Thus united, the common, though sometimes ephemer- tory Museum and I have attempted to three stigmas of the female flower, each al, events. The frequency of flowering— reconstruct the family tree of the sea- resembling a baseball mitt, cause a dip and by implication cross-fertilization— grass family Cymodoceaceae. We found in the surface of the water. can directly affect genetic diversity of that this family, which includes Thalas- The oscillatory motion of some plant the populations. By using a technique sodendron, Amphibolis, Syringodium, parts can also enlarge the e›ective tar- called starch-gel electrophoresis, Laush- Halodule and Cymodocea, shares a com- get area. In a slow-moving stream, the man has found cross-pollination in Zos- mon ancestor with Posidonia (a sea elongated stalk of Ruppia marina func- tera populations to be relatively fre- grass found in the Mediterranean and tions much like a windshield wiper: quent, particularly in the intertidal area Western Australia) and Zosteraceae, a it causes the flowers to sweep gently where surface pollination happens. family that encompasses Zostera, Phyllo- back and forth, collecting the snowflake- Yet other investigators find little di- spadix and the Australian genus Hetero- shaped pollen rafts. Floral structures versity. Based on his studies of the zostera. Our analysis suggests that this may also have localized e›ects. Joseph freshwater species Ceratophyllum de- common ancestor had noodle-shaped D. Ackerman, then at Cornell Univer- mersum, Donald H. Les of the Universi- pollen. Thus, noodlelike pollen prob- sity, observed how submerged flower ty of Wisconsin at Milwaukee believes ably evolved only once, explaining why clusters of Zostera a›ect underwater hydrophilous plants may in large part other unrelated sea-grass genera such flow patterns. He found that under lab- be genetically uniform. Such uniformi- as Halophila, Thalassia and Enhalus still oratory conditions the altered flow can ty could arise from self-pollination. For retain round pollen similar to that of concentrate pollen near the stigmas. instance, C. Thomas Philbrick of Ran- terrestrial plants. Investigation into the second possible cho Santa Ana Botanic Garden in Clare- Yet another striking feature of hy- means of increasing target e¤ciency— mont, Calif., has found that in a sub- drophilous plants has to do with dioe- changing target shapes—has begun only merged species of Potamogeton, minute cism—the separation of male and fe- in the past few years. In a series of su- air bubbles can carry pollen from the male flowers into distinct plants. More

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 73 BREACHING THE WATER’S SURFACE during the spring low terey, Calif. (left), have to release pollen or flowers on the tides is the only time many marine hydrophilous plants, surface. The pollen of many such hydrophilous species as- such as a population of Phyllospadix along the coast of Mon- semble into rafts, such as those of Amphibolis (right). than half of all hydrophilous species cytoplasm within the pollen grain. I plants do not produce large, showy are dioecious. In contrast, only about 3 found that Zostera pollen, for example, flowers, as do many terrestrial plants, or 4 percent of terrestrial species are has a strikingly short lifetime: almost I find them to be exquisitely beautiful. dioecious. Dioecism in hydrophilous all the pollen dies within eight hours The plants provide opportunities for plants might have evolved because of after release. In contrast, determining both amateur and professional bot- the physical incompatibility between the life span of terrestrial pollen is anists to make original observations the mechanisms of waterborne pollen much more di¤cult: terrestrial pollen in natural history and to study more dispersal and pollen capture within a generally must be stained to determine closely convergent evolution in aquatic single flower. A combined floral mor- viability. environments. phology would likely result in a flower Furthermore, the fate of individual capturing its own pollen before release, pollen grains of surface-pollinated spe- defeating the ability of the plants to cies is easily determined. One can sim- cross-fertilize. ply examine, frame by frame, videotapes FURTHER READING Another possible reason for the ob- of the water surface and trace the paths SEARCH AND SCREENING. B. O. Koopman. served high rate of dioecism is that of the grains. This approach is clearly Pergamon Press, 1980. it helps to prevent inbreeding. Because impractical with plants whose pollen is HYDROPHILOUS POLLINATION. P. A. Cox many hydrophilous populations con- carried by wind or animals. in Annual Review of Ecology and Sys- sist of large clones, inbreeding can re- Hydrophilous species may yield in- tematics, Vol. 19, pages 261–280; 1988. sult if male and female parts occur sights into patterns of geographic co- TWO-DIMENSIONAL POLLINATION IN HY- on the same plant. Alternative mecha- lonization and the genetics of plant DROPHILOUS PLANTS: CONVERGENT EVO- LUTION IN THE GENERA HALODULE (CY- nisms of outbreeding typical of terres- populations because the populations MODOCEACEAE), HALOPHILA (HYDRO- trial plants probably would not func- of freshwater hydrophilous plants are CHARITACEAE), RUPPIA (RUPPIACEAE), tion as well in hydrophilous plants. For in essence reproductively isolated. Wa- AND LEPILAENA (ZANNICHELLIACEAE). P. example, the maturation of male and fe- terborne pollen cannot move between A. Cox and R. B. Knox in American male flowers at di›erent times, a phe- ponds. Gene flow between populations Journal of Botany, Vol. 76, No. 2, pages nomenon typical in land plants, would must therefore transpire primarily by 164–175; February 1989. probably be di¤cult to synchronize in transfer of fruits or small vegetative SURFACE AND SUBMARINE POLLINATION IN THE SEAGRASS ZOSTERA MARINA L. a large sea-grass clone. pieces that break away. Vigorous clon- Paul A. Cox, Roger H. Laushman and The unusual features of hydrophilous al growth in most water-pollinated Mary H. Ruckelshaus in Botanical Jour- plants render them not only fascinat- plants also makes them ideal for eco- nal of the Linnaean Society, Vol. 109, ing but also extremely useful as study logical experiments: one can expose No. 2, pages 281–291; June 1991. organisms for many problems in the the same genotype to many di›erent UNDERWATER POLLINATION, THREE-DI- evolutionary ecology and demography environments. MENSIONAL SEARCH, AND POLLEN MOR- of plants. The life span of hydrophilous Compared with the widespread in- PHOLOGY: PREDICTIONS FROM A SUPER- pollen is easy to determine because the terest in wind and animal pollination, COMPUTER ANALYSIS. P. A. Cox, S. Cro- mar and T. Jarvis in Pollen and Spores: pollen grains of many hydrophilous little work has been done on hydroph- Patterns of Diversification. Edited by S. plants display cytoplasmic streaming. ilous pollination since Cavolini first Blackmore and S. H. Barnes. Oxford This feature, readily observed under a discovered Zostera flowers in the Bay University Press, 1992. microscope, is simply the movement of of Naples. Although water-pollinated

74 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. Simulating Brain Damage Adults with brain damage make some bizarre errors when reading words. If a network of simulated neurons is trained to read and then is damaged, it produces strikingly similar behavior

by Geo›rey E. Hinton, David C. Plaut and Tim Shallice

n 1944 a young soldier su›ered a During the past few years, computer ter shapes into the meaning of a word. bullet wound to the head. He sur- simulations of brain function have ad- When John C. Marshall and Freda I vived the war with a strange disabil- vanced to the point where they can be Newcombe of the University of Oxford ity: although he could read and compre- used to model information-processing analyzed G.R.’s residual problems in hend some words with ease, many oth- pathways. We have found that deliberate 1966, they found a highly idiosyncratic ers gave him trouble. He read the word damage to artificial systems can mimic pattern of reading deficits. In addition antique as “vase” and uncle as “nephew.” the symptoms displayed by people who to his many semantic errors, G.R. made The injury was devastating to the pa- have sustained brain injury. Indeed, visual ones, reading stock as “shock” tient, G.R., but it provided invaluable building a model that makes the same and crowd as “crown.” Many of his misinformation to researchers investigat- errors as brain-injured people do gives readings resembled the correct word in ing the mechanisms by which the brain us confidence that we are on the right both form and meaning; for example, comprehends written language. A prop- track in trying to understand how the he saw wise and said “wisdom.” erly functioning system for convert- brain works. Detailed testing showed that G.R. ing letters on a page to spoken sounds We have yet to make computer mod- could read concrete words, such as ta- reveals little of its inner structure, but els that exhibit even a tiny fraction of ble, much more easily than abstract when that system is disrupted, the pecu- the capabilities of the human brain. Nev- words, such as truth. He was fair at read- liar pattern of the resulting dysfunction ertheless, our results so far have pro- ing nouns (46 percent correct), worse may o›er essential clues to the original, duced unexpected insights into the way at adjectives (16 percent), still worse undamaged architecture. the brain transforms a string of let- at verbs (6 percent) and worst of all at

BRAIN IMAGES show damage to the language-processing areas tomography scans, made by Cathy J. Price and her colleagues of patients with acquired dyslexia, which can now be mod- at the MRC Cyclotron Unit in London, measure activity of the eled by artiﬁcial neural networks. (These positron-emission brain in successive horizontal slices, starting at the top. Low

76 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. function words, such as of (2 percent). reflect how the semantic route oper- Finally, he found it impossible to read ates in isolation. Later empirical find- GEOFFREY HINTON, DAVID PLAUT and wordlike nonsense letter strings, such ings suggest that this account is over- TIM SHALLICE use artificial neural net- as mave or nust. simplified, but the notion of a semantic works to investigate the behavior of the brain. Hinton is the Noranda Fellow of Since then, clinicians have studied route is still generally accepted. It now the Canadian Institute for Advanced Re- more than 50 other patients who make seems likely that deep dyslexics not only search and professor of computer sci- semantic errors in reading aloud, and lose their phonological route but have ence and psychology at the University of virtually all of them show the same damage somewhere along the semantic Toronto. He has been studying represen- strange combination of symptoms. In one as well. tation and learning in neural networks 1973 Marshall and Newcombe described for more than 20 years. Plaut is a post- two contrasting types of acquired dys- he hypothesis that reading de- doctoral research associate in the psy- lexia. So-called surface dyslexics mis- pends on multiple routes that can chology department at Carnegie Mellon University, where he earned his doctor- read words that are pronounced in an Tbe separately damaged has proved ate in computer science in 1991. Shal- unusual way, often giving the more reg- fruitful in classifying patients but less lice is professor of psychology at Univer- ular pronunciation; a surface dyslexic useful in understanding the precise na- sity College, London, where he received might read yacht as “yatched.” In con- ture of their injuries. Max Coltheart of his doctorate in 1965. His research has trast, a “deep” dyslexic patient like G.R. Macquarie University in Australia and mainly focused on what can be under- might read yacht as “boat.” Eleanor M. Sa›ran of Temple University stood about the normal cognitive sys- To explain the existence of these have both proposed, for example, that tem by studying impairments resulting from neurological disease, with oc- two types of dyslexia, Marshall and the reading of deep dyslexics may bear casional forays into the organization of Newcombe proposed that the informa- a strong resemblance to that of patients short-term memory and the nature of tion processed in normal reading trav- who have only the right hemisphere of will and consciousness. els along two distinct, complementary their brain functioning. routes. Surface dyslexics retain the pho- This explanation, however, provides nological route, which relies on com- little insight into the highly characteris- mon spelling-to-sound correspondenc- tic pattern of errors that typically oc- dles information. These descriptions ob- es. Deep dyslexics, meanwhile, retain curs in acquired dyslexia. Any detailed viously cannot be subjected to the kinds the semantic route, which allows the explanation of how errors arise and of injuries that brain cells may incur. meaning of a word to be derived direct- why they form consistent patterns re- As a result, we have turned to neural ly from its visual form (when it can be quires a model of how that information networks—idealized computer simula- derived at all). A person reading words is processed in each route—and of how tions of ensembles of neurons. We have aloud via the semantic route derives this processing goes wrong when the developed networks that perform the pronunciation entirely from meaning. neural circuitry is damaged. Psycholo- role of the semantic route, and then we According to Marshall and Newcombe, gists often use abstract, algorithmic de- have selectively removed connections the errors produced by deep dyslexics scriptions of the way that the brain han- between neurons to see how their be-

levels of activity appear in blue and high levels in white.) rior regions. The other has sustained damage to the parietal One patient (top row) has lost almost all function in the left and temporal lobes of the left hemisphere, regions generally hemisphere of the cerebral cortex, except for the most poste- believed to be crucial for processing language.

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 77 the correct pattern of semantic features w/o/l/f for each word, we had to set the weight on each connection to the appropri- PHONOLOGICAL ate value. These weights are set not SYSTEM by hand but rather through a learning procedure—an algorithm for program- WOLF WOLF “WOLF” ming neural networks. To teach a network a task, one starts with random VISUAL VISUAL SEMANTIC SPEECH weights and then presents the network PROCESSES WORD-FORM SYSTEM repeatedly with a “training set” of input SYSTEM patterns (in this case, letters in specified positions). The algorithm adjusts the weights after each training run to reduce the di›erence between the network’s output and the “correct” response. TWO PATHWAYS in the brain are responsible for the mental processing and pronunciation of written words. One (the phonological route) derives pronuncia- eural-net workers have known tion from spelling, the other (the semantic route) from meaning. Deep dyslexics since the 1950s how to adjust have lost the phonological route completely and have su›ered damage to the se- Nweights in simple, two-layer net- mantic route as well. works, but training networks with a greater number of layers is more di¤- cult. In particular, it is not immediately havior changes. A few years ago we de- with the pattern of connections among obvious how to set the weights on the signed a simple network to mimic the neurons, determine the computation connections from the input units to the semantic route and found that damag- that the network performs. intermediate units because there is no ing any part of it could reproduce sev- The first version of our network conway to determine, a priori, which inter- eral of the symptoms of deep dyslexia. sisted of three sets of units: “graph- mediate units should be active for any We have since made more detailed mod- eme” units, each of which represented given input and output. els to learn which aspects of neural- a particular letter in a specific position During the 1980s, however, neural- network architectures were responsible within the word; “sememe” units that net researchers developed a number of for this behavior. We have also extend- represented the meanings of words; di›erent methods for training multi- ed the approach to account for addi- and a layer of intermediate units that layer networks. These methods appor- tional symptoms of deep dyslexia. make it possible to learn complex asso- tion changes to the connection weights Our models of the semantic route con- ciations. A completely general network of each layer according to their consist of interconnected units representing would require 26 grapheme units for tribution to the error. Over the course neurons. Each neuron unit has an activi- each position within a word, but we of many training cycles, the resulting ty level (between 0 and 1) that depends used a simplified vocabulary that per- weights converge to yield a network that on the inputs it receives from other neu- mitted a smaller number. The graph- produces the correct results. Depending rons. Connections between units have eme units in the first position were all on the initial random weights, learning an adjustable weight that specifies the consonants, for instance, and those in may result in any of a number of sets extent to which the output of one unit the second, all vowels. of weights, each of which leads the net- will be reflected in the activity of the The sememe units do not correspond work to produce correct answers for unit it is feeding. These weights, along directly to individual word meanings its training inputs. (For further details but rather to semantic features that de- of the learning procedure, see “How scribe the thing in question. The word Neural Networks Learn from Experi- cat activates such units as “mammal,” ence,” by Geo›rey E. Hinton; SCIENTIF- “has legs,” “soft” and “fierce.” Units IC AMERICAN, September 1992.) WWWrepresenting such semantic features as In theory, these learning procedures “transparent,” “tastes strong,” “part of can get stuck in so-called local min- limb” or “made of wood” remain qui- ima—configurations of weights that escent. Our network has 68 sememe are incorrect but for which any small units representing both physical and change would only make the network’s functional attributes of a word’s defini- errors worse. In practice, however, a tion. Each word that we chose was rep- network almost always learns nearly resented by a di›erent combination of optimal solutions. In addition, some of NEURON active and inactive sememe units. the learning procedures are more bio- To make our neural network produce logically plausible than others, but our results do not seem to depend on which method we use. We suspect that even if the brain uses a quite di›erent learning IDEALIZED NEURON is the basis for procedure, the resulting neural circuit- artificial neural networks. It sums the ry will still resemble the structure that WW W W weighted inputs that it receives from oth- our network develops. Thus, our expla- er neurons (bottom) and generates an activation level between 0 and 1. It then nation of what happens when the net- passes this activation (through weight- work is damaged may be correct even ed connections) to other neurons. The if its learning procedures are not. set of weights and connections in a neu- Although our initial network, with ral network determines its behavior. one intermediate layer, could learn to

78 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. map word-forms to their semantic fea- intermediate layer can direct the learn- neural network. The original network tures, it was not really satisfactory. It ing of nets containing multiple inter- was static—any given input would cause had a strong tendency to map very sim- mediate layers or even networks whose the network to produce a corresponding ilar inputs (such as cat and cot) to sim- units are connected in cyclical fashion. output pattern, and that pattern did not ilar outputs unless subjected to exces- The most natural way to implement change as long as the input stayed con- sively long training. We addressed this this cleanup mechanism is with a feed- stant. The output of the new network, problem by adding another layer of back loop. The output of the sememe however, is dynamic; it settles gradual- “cleanup” neurons. If the original set of units goes to the cleanup units, and ly into a stable pattern. connections produces a sloppy answer, their output goes to the inputs of the Consequently, we have found it useful the new units will change it to produce sememe units. Each time activity flows to think of the network’s output not just exactly the correct semantics. The num- around the loop, the influence of the as a list of active semantic features but ber of word meanings is limited, so the cleanup units on the sememe units rather as motion through a multidimen- pathway from the input need only get (and vice versa) will yield a pattern of sional “semantic space,” whose coordi- the activities of the sememe units clos- semantic features that is closer to the nates are defined by all the semantic er to the correct meaning than to any correct one. features that the network can represent. other. The same learning techniques The feedback loop introduces a new Every point in the space corresponds to that succeed on networks with a single characteristic into the behavior of our a specific pattern of activity among the

CLEANUP UNITS

SEMEME UNITS

INTERMEDIATE UNITS

GRAPHEME UNITS

NEURAL NETWORK FOR READING contains four layers. The tation in terms of semantic features, such as size, edibility or ﬁrst responds to the letters in each word. Connections be- aliveness. “Cleanup” units are connected to sememe units in tween input and intermediate units and between intermedi- a feedback loop that adjusts the sememe output to match the ate and “sememe” units convert the word-form to a represen- meanings of words precisely.

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 79 ogists and others had wondered how damage near the input—the visual part of the reading system—could cause semantic errors. According to our models, these errors arise naturally as the cleanup neurons use semantic information to try to make sense of the output of the damaged earlier stages. The notion of attractors helps to explain another anomaly in the data as well. Almost all patients who make semantic errors also make some visual errors—they confuse a word like cat with a visually similar word like cot. They do not, however, make the sounding-out errors of surface dyslexics (“loave” for love or “deef ” for deaf ). This invari- able connection between semantic errors and visual errors is odd. Some patients must have damage solely to the later stages of their processing systems, and one would intuitively expect them to make only semantic errors. After implementing our neural-network model, we discovered to our great surprise that damage to the semantic cleanup circuit sometimes caused visual errors. Retrospectively, we can understand why: the earlier layers of an undamaged network can a›ord to produce somewhat similar semantic outputs for the words cat and cot because the cleanup circuit will steer each to its proper meaning. But when the cleanup circuit is damaged and the shapes of each attractor change, the output of the sememe units may fall into the attractor for a visually similar but semantically unrelated word. GRAPHEME UNITS INTERMEDIATE UNITS SEMEME UNITS CLEANUP UNITS This explanation did not initially oc- ACTIVATION LEVELS of neurons in the network change with time as the net pro- cur to us because it relies on the idea cesses the word bed. At first, many of the sememe units are activated to varying that the boundary of the attractor for degrees, but interaction with the cleanup units strengthens the activation of some cat can come very close to the one for and weakens that of others until the output converges. cot even though the two words are semantically dissimilar. One would expect the attractors for many other meanings sememe units, but only a few of those able for understanding how our net- to come between those for cat and cot. patterns correspond to valid meanings. work operates and how it can make the In a two-dimensional space this intui- The correct meanings of words are same semantic errors that dyslexics do. tion is correct: if we choose 40 points points in semantic space. If we damage the network by randomly at random to represent word meanings changing the weights in the cleanup and construct fairly compact attractors he first three layers of the net- mechanism, for example, the boundar- around each point, the attractors for work, seen according to this per- ies of the attractor for each word will dissimilar meanings will not come any- T spective, take a word-form and change. As a result, if the network is in where near one another. convert it to a position somewhere in a region in semantic space where it was It is very dangerous, however, to as- semantic space. Activity in the cleanup previously drawn to one word, it may sume that the same is true in spaces layer then draws the output of the net- now be drawn to a semantically relat- that have many dimensions. Our network to the point corresponding to the ed one instead. Alternatively, if we dis- work represents 68 semantic features closest meaning. The region around rupt the pathway coming from the in- in its sememe units, and so the attrac- each word is what physicists and math- put, the network’s initial output may be tors for each of its 40 words reside in a ematicians know as a point attractor— closer to the meaning of a semantical- 68-dimensional space. It turns out that whenever the network’s initial output ly related word than to the meaning of in 68 dimensions, the midpoint between appears within a certain region, the net- the word originally presented. any two randomly chosen points is al- work’s state will inexorably be drawn This result clears up one of the first most certainly closer to each of those to one position within the region. puzzles presented by deep dyslexia: points than it is to any of 38 other ran- This notion of a semantic space dot- why damage to any part of the brain’s dom points. Consequently, the attracted with attractors representing the semantic route produces an essentially tors for cat and cot can have a common meanings of words has proved valu- similar pattern of misreadings. Neurol- border without any other attractors get-

80 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. ting in the way. Avoiding obstacles is errors characteristic of deep dyslexia. quently, the word’s category can still be easy in 68-dimensional space. Interestingly enough, our network determined. not only reproduces the obvious visual lthough our network was able to and semantic errors of deep dyslexia, it ne symptom of deep dyslexia mimic both the correct and dys- also mimics some of the subtler char- that our models did not initial- A functional mapping of word- acteristics of the disorder. For instance, Oly address is the way in which forms to meanings, that does not mean patients occasionally make “visual then patients have more trouble reading ab- its architecture is the only possible one semantic” errors, in which a semantic stract words than concrete ones. This for the brain’s semantic processing route. confusion seems to follow a visual one. phenomenon appears to be an integral To determine the range of possible al- G.R. would read sympathy as “orches- part of the syndrome because abstract- ternatives, we investigated the e›ects tra” (presumably via symphony). Our ness—a semantic property—increases of damage on several di›erent archi- networks also produce these errors— the probability of visual errors. Further- tectures, each designed to evaluate one sometimes reading cat as “bed,” via cot. more, when patients make such mis- aspect of the original network design. When severely damaged, our network readings, the responses they come up We programmed versions of the neu- also exhibits a strange e›ect that occurs with tend to be more concrete than the ral network that contained connections when patients have a lesion so large original word presented. among the sememe units and ones that that their semantic representations are We based our approach to simulat- lacked such connections; we also pro- distorted beyond recognition and they ing this e›ect on the proposal, made grammed some networks so that each cannot ﬁnd a word at all. Such patients by Gregory V. Jones of the University neuron in one layer was connected to are unable to identify the word they are of Warwick in England and others, that every neuron in the succeeding layer trying to read, but they can often still concrete words are easier for deep dys- and others whose connections were decide which category it falls into, say, lexic patients because they evoke a sparse. In addition, we moved the clean- “animal” versus “food.” Under similar more consistent and detailed meaning. up units so that they performed their circumstances, our network no longer In terms of our network, a concrete work ahead of the sememe units, and stabilizes at the attractor corresponding word has more semantic features than we combined the cleanup units with the to a particular word—indeed, the attrac- does an abstract one. For example, post intermediate layer. We even changed tors for several words may have merged. has 16 features ranging from “size be- the arrangement of neurons in the in- Nevertheless, the network’s output does tween one foot and two yards” to “used put layer to alter the way that words stabilize within a larger volume of se- for games or recreation.” In contrast, were represented and added an out- mantic space wherein the correct word past has only two features: “has dura- put network that converted meanings and its relatives once resided. Conse- tion” and “refers to a previous time.” We to strings of phonemes, so that the system actually spoke. Most of the architectural details are “BED” “COT” “CAT” irrelevant. The speciﬁc way the visual input is represented is not important as long as words that resemble each other visually produce similar patterns of activity in the input layer. The only crucial ingredient is the existence of attractors—if there are no cleanup units “downstream” of the damage, the network does not exhibit the pattern of

BED

SEMANTIC SPACE has many dimensions, corresponding to the semantic features (only a three-dimensional approximation is drawn here). The meanings of particular words are points in semantic space. When the authors’ neural network reads b a word, interaction between sememe a and cleanup units causes any word-form that is mapped into a region of semantic space near the meaning of a word (col- COT c ored regions) to converge on that mean- COT (DAMAGED) ing (dots). If the network is damaged so that the boundaries of these so-called attractors shift, a word can be misread as a semantically similar one—“cot” for bed, CAT for example (a). Semantic errors may also occur if damage causes a word-form to be mapped to a slightly di›erent point in semantic space (b). Such a network SEMANTIC SPACE can make visual errors because visually similar words will initially be mapped to nearby points in semantic space, even if the stable points of the attractors they fall into are quite distant (c).

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 81 surprising reversal: the damaged network reads concrete words less well and produces more visual errors than with abstract words. This type of lesion and pattern of performance are consistent SPEECH with what is known about the single, SYNTHESIZER CLEANUP enigmatic patient with “concrete-word UNITS dyslexia,” studied by Elizabeth K. War- rington at National Hospital in London. Not only did he have much more trouble PHONEME reading concrete words than abstract UNITS ones, he also did better matching spoken abstract words with pictures. This consistency suggests that his problem k/a/t lay at the level of the semantic system. Our account of the error pattern of deep dyslexia relies on the properties of a neural network that transforms INTERMEDIATE one representation (a visual word-form) UNITS into another, arbitrarily related representation (a set of semantic features). CLEANUP One would expect similar error patterns UNITS to result from damage to other cognitive processes that involve an arbitrary transformation to or from a semantic space. Moreover, neuropsychologists SEMEME have already described somewhat simi- UNITS lar error patterns in deep dysgraphia, a disorder of writing, and deep dyspha- sia, a disorder of word repetition. This additional evidence suggests that our model may have a wider validity than we originally supposed. More important, however, it marks the success- INTERMEDIATE ful use of a new technique for under- UNITS standing how the brain works. Our work di›ers from other explanations for deep dyslexia (and, with few exceptions, other explanations for neuropsychological phenomena in general) in the kinds of hypotheses that we frame. Instead of verbally characterizing each component in a complex neural mechanism and re- GRAPHEME lying on intuition to tell us how damage UNITS "CAT" will a›ect its behavior, we simulate that mechanism, damage it and watch to see what happens. We have found that many of our hunches were wrong. This SPEAKING NETWORK adds another set of three layers to the original reading neural net. It converts sequences of letter shapes to semantic representations and discovery suggests that detailed com- maps those in turn to sequences of phonemes that can be fed to a speech synthe- puter simulations will play a crucial role sizer. This network is particularly useful because it does not require researchers to in furthering understanding of how the make potentially biased judgments about what word (if any) corresponds to a per- brain normally processes information turbed pattern of semantic features, as may be generated when the network is about language and of how that func- damaged to simulate dyslexia. tion is disrupted by injury or disease. designed a new vocabulary containing is more redundancy in their semantic 20 pairs of four-letter words di›ering activity patterns. Hence, there is more FURTHER READING by a single letter, one concrete and the structure that the cleanup units can use LESIONING AN ATTRACTOR NETWORK: IN- other abstract. On average, the concrete to make the network converge on the VESTIGATIONS OF ACQUIRED DYSLEXIA. words had about four times as many se- proper meaning. The abstract words, G. E. Hinton and T. Shallice in Psycho- mantic features as did the abstract ones. which have less redundancy in their se- logical Review, Vol. 98, No. 1, pages 74– After the network had been trained mantic patterns, must rely more heavi- 95; January 1991. to pronounce the words, we found that ly on the feed-forward pathway, where DEEP DYSLEXIA: A CASE STUDY OF CON- NECTIONIST NEUROPSYCHOLOGY. D. C. lesions to any part of the network “up- visual inﬂuences are the strongest. Plaut and T. Shallice in Cognitive Neu- stream” of the cleanup units reproduced Because correct recognition of con- ropsychology, Vol. 10, No. 5, October the e›ects of abstractness. The concrete crete words relies more on the cleanup 1993. words cause fewer errors because there circuit, severe damage there leads to a

82 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. Raising the Vasa This Swedish man-of-war foundered on her maiden voyage and slept for three centuries at the bottom of Stockholm Harbor. Here is the story of her resurrection

by Lars-Ake˚ Kvarning

n the 1620s Gustavus II Adolphus, and all,” as the Council of the Realm the lines, however. The skipper, Göran king of Sweden, was on his way to described the disaster in a letter to the Mattson, related that the vice admiral I becoming one of the leading war- king. The Danish ambassador, Erik and the captain of the Vasa had tested lords in Europe. He controlled Finland, Krabbe, reported that the drowned “are the stability of the ship while she was Estonia and Livonia, and he had just said to be more than half a hundred per- being fitted, by having 30 men run to won the small part of Russia that touch- sons, among them some women and and fro across the upper deck. After the es on the Gulf of Finland. By thus ex- children who wanted to follow their hus- third cycle, they had had to call a halt cluding the czar from the Baltic, he had bands into the archipelago.” to prevent the Vasa from capsizing on nearly made that sea into a Swedish lake. Three centuries later the wreck of the the spot. Something was clearly wrong To secure his power, he needed to main- Vasa was found, raised and restored, with the vessel’s design, but still no ac- tain a strong fleet in the Baltic. There- making the ship a monument to her tion was taken. fore, in January 1625, the king commis- builders and rescuers alike. The war- It can be gathered from the records sioned the construction of four ships. ship’s resurrection was worthwhile both that most observers believed the bal- The most magnificent was the Vasa. for the intrinsic interest of the opera- last compartment had not been made Named after the Swedish royal family, tion itself and because the prize has so large enough to balance the extraordi- she was one of the largest warships of much to tell about the vessel’s makers. nary double row of gundecks that the the time. A ship is not only a society in miniature; king had specified. Indeed, the ship- On August 10, 1628, a bright, sunny it is also the product of a larger society. wrights maintained that they had built day, crowds gathered at Stockholm Har- bor to see this new addition to the fleet he loss of one of the biggest war- leave for the island naval base not far ships in Europe, on her maiden from the city. Members of the crew went T voyage and in sheltered waters, aloft to set four of the Vasa’s 10 sails— constituted a catastrophe—one that foresail, foretop, maintop and mizzen. called for a scapegoat. Interrogations of Cannons fired salutes, and the cheer- the captain of the ship, Söfring Hans- ing throng on the quays and shores son, and other o¤cers began soon af- watched as the imposing vessel began ter they were rescued. The shipbuilder, to move slowly in the light breeze, her Henrik Hybertson, had died the year be- colors and gilding shining. fore, and the responsibility of complet- Suddenly the cheers changed to ing the project had been taken over by shrieks of horror as a gust of wind his widow and brother in cooperation caught the ship in the harbor. She heeled with the assistant shipbuilder, Hein Ja- over to port and raised a little only to cobson. These people also had to sub- list again, the water now rushing in mit to questioning. through the open gunports. Then the Had the guns been properly secured? Vasa foundered, “under sails, pennants The principals could not prove it, but in the summer of 1961 marine archaeologists indeed found the cannon carriag- es standing in their proper places, the LARS-AKE˚ KVARNING is president of remains of their securing ropes still Sweden’s National Maritime Museums coiled around the axles. Had the stone and director of the Vasa Museum in Stockholm, where he has worked since ballast been taken on properly? A wit- 1964. Before that he was associated ness replied in the a¤rmative that it with the folk department of the Nordic had been and that the space for it down Museum and with the Swedish Central in the hold was filled. Office of National Antiquities. He holds Signs of trouble may be read between a licentiate in ethnology from Uppsala University and has directed a documen- tary project to photograph national mon- uments from the air. He serves as a re- RESTORED MAGNIFICENCE of the Vasa serve captain in the life guard regiment is apparent in the stern, which sports of the Grenadiers. such royal symbols as a crown, a coat of arms and griffins.

84 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. the vessel just as contracted and in ac- cordance with orders received directly from the king. Gustavus Adolphus had wanted a strong navy to protect the supply lines leading to the theater of war on the other side of the Baltic. He also intend- ed to blockade the rich Polish ports so that his agents could collect customs duties from cargo ships calling there. It seems that his ambitions grew after he placed his original order and that the Vasa was begun as a smaller ship and completed as a larger one. This royal intervention was perhaps the real reason for the disaster. The

6 3 4 5 2 STEP 1 0 METERS 100 200 shipwrights of the day did not make leather clothing to protect against the and so they abandoned it to the sea, construction drawings and had no cold and climb on a platform hanging ending a salvage operation that was means of calculating the stability of under the bell, his head and chest pro- perhaps the most impressive underwa- their creations. Trial and error was their jecting into the bubble of air. Then the ter enterprise of the premodern era. only guide. When the builders scaled bell would be lowered down to the The divers were lucky as well as cou- up the plans of the Vasa by adding a sunken wreck. rageous. The limited supply of air cut second gundeck, they left only enough the dives short enough to prevent dan- room in the hold for about 121 tons of rancesco Negri, an Italian priest gerous quantities of nitrogen from dis- stone ballast, less than half of what and explorer, has left posterity this solving, under high pressure, in the was needed. Yet had they attempted to Feyewitness account of von Trei- men’s blood. A too rapid return to the increase the amount, the lower gun- leben’s operations: surface would then have caused the gas decks would have come down danger- to precipitate in bubbles, an often fatal ously close to the waterline. When he had signaled and been malady commonly known as the bends. Many others made mention of the hauled up after having been under water The Vasa was more or less forgotten king’s personal interest in the build- at least a quarter of an hour, he brought until 1956, when Anders Franzén, a pri- ing of the vessel. The shipwright, when up with him a heavy plank of oak with vate researcher, located her on the bot- asked about whom or what to blame iron mountings that he had caught with tom of Stockholm Harbor. Franzén for the debacle, sighed: “Only God his hook. I asked him through an inter- owed his success not to chance but to knows....” With God and the king in- preter if he could have stayed down any insight. He realized that the Baltic was volved in this fashion, there was only longer. He said yes, up to half an hour, not salty enough to support the ship- one way out. The case was dropped. but no longer. It was possible—as I be- worm Teredo navalis—an elongated, A ship of this size, equipped with lieve, although I forgot to ask him—that wormlike clam—and that its waters bronze cannons worth a fortune, could the air enclosed by the bell would have therefore ought to have preserved the not but become an immediate object of become too hot from his breathing, and wrecks of wooden ships extraordinarily salvage. Only three days after the calam- probably the water would have exces- well. The Baltic’s spoils are indeed with- ity, the Council of the Realm authorized sively tormented his legs with cold. The out parallel, for no other sea of such Ian Bulmer, an Englishman, to make the man trembled, although he was born in low salinity has borne so many fleets. attempt. He achieved little. Although the the country and was strong and used to Franzén concentrated on men-of-war principles of such operations were well physical strain. It was then towards the from the 16th and 17th centuries. He known, the salvage crews lacked the end of October of last year, 1663. studied the archives and got indications force to lift a prize as heavy as the Vasa. of the positions of 12 sunken ships A Swede named Albrect von Treileben When below, the diver could hardly that he found particularly interesting came on the scene in 1663, five years move and could not see at all. Still, the and then set his mind on finding the after introducing a diving bell that made salvagers managed to bring up most Vasa. One day, while dragging his grap- submarine sites far more accessible. of the Vasa’s 64 cannons, including 48 pling iron in Stockholm Harbor, it Thus equipped, he and his divers set 24-pounders weighing more than a ton hooked something big and immovable. out to salvage the cannons and other each. The remainder of the wreck had Divers went down and confirmed that loose items of value. A diver would don little to interest the people of the day, an enormous hull of black oak was

ivers armed with water jets blasted six tunnels through ters—about eight feet—so it could be towed shoreward. Dthe clay beneath the sunken wreck (inset at left). They When the ship set down again, the workers refilled the pon- then drew steel cables through the tunnels, fixed either end toons, tightened the cables and repeated the process. After of each cable to parallel banks of half-submerged pontoons 18 steps, they reached shallower water. There, after two and tightened them, forming a taut cat’s cradle. By bailing years of preparation, jacks mounted on the pontoons final- out the pontoons, the workers could lift the wreck 2.5 me- ly lifted the ship to the surface.

KASTELLHOLMEN

16 METERS

18 16 17 13 14 15 11 12 7 8 9 10

400 500 600

standing on its keel on the bottom. It Franzén’s dream of recovering the vaging Company, undertook the task was the Vasa. In and around her were ship invited ridicule from several quar- free of charge. strewn the skeletons of 25 people. A few ters, but he was able to attract two There was no lack of imaginative sug- of them were found inside the ship weighty collaborators. The Swedish Navy gestions. One wag proposed using Ping- during the excavation, but most were agreed to organize the training of its Pong balls to ﬂoat the Vasa to the sur- turned up by workers who scoured the divers around the needs of the project, face. Another recommended freezing underlying seaﬂoor after the vessel had and the Broströms Shipping Company, the ship in a block of ice, which would been raised. through its subsidiary the Neptun Sal- then bob to the surface, where it could

RAISED AT LAST, the Vasa’s hull was still held together by SHOWER OF POLYETHYLENE GLYCOL impregnated the Vasa the original wooden pegs, although all the iron bolts and nails over a period of 18 years. Smaller objects were treated more had rusted away many years before. quickly by immersion in baths.

fficers bunked in back, with the highest Oranking in the main cabin and the juniors just above. Provisions included salt fish and meat, sour bread and beer. The helmsman steered by swinging the whip staff vertically, then pressing down with all of his weight through the bearing. The combined action shifted the tiller horizontally, thereby moving the 10 rudder slightly—a surprisingly cumbrous design.

5 13 14 4 UPPER GUNDECK

6 LOWER GUNDECK 11

ORLOP DECK

9 12 7

1. POOP-ROYAL CABIN 8. BALLAST 2. OFFICERS’ CABIN 9. PROVISIONS STORE 3. MAIN CABIN 10. MIZZEN MAST 4. STEERAGE CABIN 11. BARBER-SURGEON STATION 5. HELMSMAN’S STATION 12. RUDDER 6. BILGE PUMP 13. TILLER 7. POWDER MAGAZINE 14. WHIP STAFF

88 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. be towed to a suitable place for melting. ologists who were to excavate the ves- and boric acid as a safeguard against The salvage company, however, fa- sel from within were vaccinated against fungus and applied the mixture to the vored a traditional technique. It ran ca- tetanus, typhus and other communica- wood. The PEG e›ectively embalms the bles under the Vasa, sparing her tim- ble diseases. Thus protected, they found wood by sti›ening its cells, thus pre- bers by snaking the cables through tun- and registered about 25,000 objects, venting the wood from changing its nels bored in the seabed by two-faced about half of them deriving from the shape or cracking. Additional protec- water guns, known as Zetterström noz- ship’s structure. tion was provided by climate-control zles. The forward jet cuts through the All the finds were to return to their systems that assured a constant tem- clay while weaker backward jets offset proper places during reconstruction, the perature and an ideal humidity. the recoil and clear away loosened ma- aim of which was to restore the Vasa Wooden objects of manageable size terial. Divers had to work within the as far as possible to her original shape. went into big baths, where the concen- growing tunnels, with four meters of To that end, in the fall of 1961 an alu- tration and temperature of the solution mud and sludge, a 333-year-old ship minum shelter was constructed around could be increased slowly. This very ef- and 32 meters of water directly over- the Vasa, which still rested on the pon- ficient form of the treatment took 18 head. Their worst nightmare, as they toon. The pontoon was then towed to a months for the oak and about a year toiled in the complete darkness, was temporary museum, where work could for the softwoods. The ship, however, that the ship’s bottom would give way proceed under the eyes of the public. could not be dipped in this fashion. The under the weight of the ballast stones The challenge was to replace the water workers had to sprinkle the solution on inside, trapping them in their holes. in the wood with something more last- her, inside and outside, a tedious pro- ing without allowing the wood to crack, cess of saturation that started in 1961, t was a breathless moment when shrink, warp or misbehave in any way. three years before I joined the restora- the Vasa at last sat within her steel Not only the 1,100 tons of structural tion team, and ended only in 1979. At I cat’s cradle, ready for the tug that components had to be thus preserved that point, the hull was slowly dried. would loosen her from the clay that had but also 700 sculptures and carved or- imprisoned her for so long. Pumps emp- naments and many thousands of ob- n intriguing preservation prob- tied pontoons attached to the cables, jects made of textiles, leather and metal. lem was o›ered by the six sails which floated up, tightening the cables No one had ever attempted to pre- A that survived from the Vasa’s until they were as tense as violin strings. serve such a mass of wood. But two original 10. They were discovered in a The clay at last let go its grip, and the Swedes—Bertil Centervall, a conserva- sail locker on the orlop, the lowest con- Vasa hung free in the cables. This sus- tor, and Rolf Morén, an engineer—had tinuous deck. The cloth was in such bad pension put a tremendous strain on invented a process to preserve fresh condition that mere handling might the hull, and it would have collapsed if wood by impregnating it with polyethy- have caused it to crumble into dust. It not for the strength remaining in its lene glycol (PEG). Because the polymer was possible, however, to work the ma- ancient oaken timbers. Now everything worked by replacing water in fresh terial as long as it was kept in water. went very smoothly. wood, there was reason to hope that it Large baths were therefore prepared in Because the pontoons could not lift might do the same in ancient water- which the remains of the sails could be the Vasa more than 2.5 meters at a time, logged wood—and it did. Workers dis- unfolded, cleaned and dried in alcohol the process had to ratchet through 18 solved the PEG in water, added borax and xylene. Because of their fragility, steps. The Vasa would rise a bit, then the workers would tow her to shallower water and flood her pontoons until she set down on the bottom again. Remark- ably, the impressions the Vasa left on the seafloor at each stage are clearly visible today, appearing like a giant’s footsteps on the sonar scan. At last the ship reached a site from which the final lift could be made with ease. Huge jacks mounted on the pontoons hauled the prize to the surface, at which point pumps that had been installed inside the hull began to bail it out. Thus assisted, the Vasa floated on her own keel into dry dock, where a concrete pontoon was waiting. There, in 1961, for the first time in 333 years, the Vasa revealed her imposing hull to an astonished public. To protect the delicate, waterlogged wood, the caretakers of the man-of-war constantly sprayed her with water until permanent preservation could be arranged inside an appropriate shelter. It was also necessary to get rid of the vast amount of mud that had collected in the ship—a black sludge loaded with MODEL OF THE VASA, constructed at one tenth the original size, took four crafts- archaeological finds and, perhaps, bac- men four years to complete. It now sits alongside the reconstructed hull in the teria. To be on the safe side, the archae- Vasa Museum, which faces Stockholm Harbor.

Copyright 1993 Scientific American, Inc. SCIENTIFICCopyrightAMERICAN 1993 ScientificOctober American, 1993 Inc. 89 SMALL OBJECTS afford perhaps the thereby restoring the iron while purg- side. Of course, we always mark such clearest view of life on board the Vasa. ing injurious compounds. reconstructions. The vessel’s brilliant colors are evi- The next problem consisted in put- The restored Vasa has revealed many denced in the fully restored heraldic de- ting the ship back together. Fortunate- unexpected aspects of the shipwright’s vice, whose sheaf represents the word ly, the hull was built with a double sys- craft. The steering system, for instance, Vasa, the name of the Swedish ruling tem of iron bolts and wooden pegs, so is strangely awkward. The tiller stretched family (far left, top). Other items in- that the disappearance of the bolts did 11 meters from the top of the rudder clude (clockwise from left) clothing, not cause the wooden planks to collapse to the eye of the whip sta›—a vertical mess kits, officers’ tableware, apothe- in a heap. The main part of the hull beam that connected at a right angle to cary materials, sailmaking equipment, ammunition, games, coins and toiletries. could therefore come up in one piece. the tiller and projected up to the deck The stern castle, however, had broken through a roller bearing. To steer, the down, and the beak head—a protrud- helmsman first had to tilt the whip ing construction at the prow—and oth- sta› and then bear down, pressing with they were a¤xed to a backing of glass fi- er parts had fallen o›. Consequently, all his weight through the bearing. The bers by means of a plastic solution made workers had to identify and locate many combined action would move the tiller especially for this purpose. The solution thousands of structural components, to the side. To center the rudder, he had has the same refractive index as the ranging from heavy beams to tiny bits to haul the whip sta› back up through glass fibers, making them invisible. of wood—a gigantic jigsaw puzzle to be the hole in the bearing. The wrought iron had long since rust- assembled without benefit of blueprints. It appears that the shipwright was fol- ed away, with the exception of the low- lowing a tradition that had made sense est rudder pintle and gudgeon (the pin he restorers hesitated over the only when vessels were smaller. In the and the pivot on which the rudder had propriety of filling in gaps in the Vasa, the great length of the tiller limit- turned), which had been pushed down T archaeological record. At first ed the turning radius of the rudder to into the bottom clay and preserved in they were able to dodge the question be- 14 degrees—seven to the left and seven relatively good condition. In 1992 we cause such a wealth of original parts re- to the right. That paltry distance, com- also found a hidden bolt in very good mained. But soon they reached a stage bined with the relatively small surface condition when we were preparing for at which connecting structures had to of the rudder, meant that most of the the masts to be restepped. be made to exploit some of the surviv- steering had to be done by manipulat- The cast-iron cannonballs were pro- ing pieces. By then, however, we real- ing the sails. tected by their low carbon content and ized that we wanted to rebuild the ship The most striking feature of the Vasa thus resisted oxidation more e›ective- as completely as possible and that very is her gorgeous ornamentation, so dif- ly. Many of them now weighed hardly few missing parts would have to be ferent from the modern subordination more than a tennis ball, but their rust- supplied by our own people. Symmetry of form to function. The Vasa may well ed hulks retained their original shape helped enormously: a ship is the same illustrate the change in taste more effec- and size. Because these rust balls would on either side, so when anything was tively than can architectural remains, have fallen apart in air, we dried them missing we were usually able to model for one expects to see decoration in in hydrogen at 1,060 degrees Celsius, it on its counterpart from the other buildings but not in military hardware.

90 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. Yet the Vasa is festooned throughout 1961, notably the combination of an longings were recovered. The simple with sculptures prominent enough to electron microscope and an energy-dis- sailors had things of wood or clay for attract the eye during an engagement. persive x-ray analyzer. The Vasa, it is the most part. The o¤cers’ quarters in They filled several purposes: to encour- clear, was as bright with colors and as the stern castle contained delicate Swed- age friends, intimidate enemies, assert resplendent with gold as the altarpiece ish glassware, porcelain, pewter and claims and impress the world with this of a baroque church. So far we have brass, as well as goods imported from picture of power and glory. used this information to restore only a other parts of Europe. One pewter flask The decoration of the stern, for in- few sculptures, among them the Vasa still contained clear alcohol, 66 proof, stance, spun a story of the Vasa, the family crest. that seems to be arrack from the East Swedish royal family for whom the ship Life on board the warship would have Indies. I can testify, from personal ex- was named. The ornamentation begins been hard, even in the 1620s, when few perience, that the liquor was good. at the top with a woodcut of the young sea battles were waged. But peace also In June 1960, King Carl XVI Gustaf in- King Gustavus Adolphus, a royal crown had its horrors: more than 400 men augurated the Vasa Museum in Stock- held over his head by two gri¤ns. This would have been crammed in a con- holm. There the royal warship rests, sur- is a political statement, for it contra- fined space. Hygienic conditions on the rounded by exhibits of the world from dicts King Sigismund III of Poland— ships of the day were shocking, and the which she comes—a graphic testimoni- Gustavus Adolphus’s first cousin—who food was bad even in port, let alone af- al to another time and place. also laid claim to the Swedish crown. ter stewing in the hold for a few weeks. The statement is reinforced by the frieze The men and the o¤cers, too, were thus just below, which reads “G A R S,” for exposed to a range of diseases. Gustavus Adolphus Rex Sueciae, desig- FURTHER READING nating Gustavus Adolphus king of Swe- sunken ship is a kind of time cap- THE SWEDISH WARSHIP WASA. Lars-Ake˚ den. Further down, the Swedish nation- sule. Indeed, it is to be preferred Kvarning and Bengt Ohrelius. Macmil- al coat of arms appears and, below it, to the contrivances of the cor- lan, 1974. A WHY WASA CAPSIZED. Curt Borgenstam the family crest of the Vasa dynasty. nerstone cache and burial chamber be- and Anders Sandström. Wasa Studies Vasa is an old Swedish word for “sheaf,” cause it provides an utterly unself-con- No. 13. National Maritime Museum, pictured on the coat-of-arms. scious snapshot of daily life. Ships, 1984. Hardly any traces of color remained their equipment and their cargoes con- THE POWER AND THE GLORY: THE SCULP- on the waterlogged sculptures, the tain a wealth of information about tech- TURES OF THE WARSHIP WASA. Hans wood of which had turned to a rather nology and commerce, wars and diplo- Soop. Almqvist and Wiksell/Coronet even dark brown. On closer inspection, macy, common people and the spirit of Books, 1986. bits of gold leaf still glittered here and the age. THE ROYAL WARSHIP VASA. Björn Land- there, leading us to believe that many The Vasa was not fully manned and ström. Stenström Interpublishing, 1988. VASA. Erling Matz. Vasa Museum, 1990. of the sculptures had been gilded. We equipped when she went down. The THE VASA MUSEUM: AN OLD SHIP IN A were doubtful as to other colors, but a ship lacked, for instance, 300 marines NEW HOUSE. K. A. Adrup. National number of applicable techniques to dis- from the normal military complement. Board of Public Building, 1990. cern them have been developed since Nevertheless, thousands of personal be-

egend has it that when Pythagoras and his followers discovered the theorem that bears his name in the sixth L century B.C., they slaughtered an ox and feasted in celebration. And well they might. The relation they found between the sides of a right triangle held true not sometimes or most of the time but always—regardless of whether the triangle was a piece of silk or a plot of land or marks on papyrus. It seemed like magic, a gift from the gods. No wonder so many thinkers, from Plato to Kant, came to believe that mathematics o›ers the purest truths humans are permitted to know. That faith seemed rea¤rmed this past June when Andrew J. Wiles of Princeton University revealed during a meeting at the University of Cambridge that he had solved Fermat’s last theorem. This problem, one of the most famous in mathematics, was posed more than 350 years ago, and its roots extend back to Pythagoras himself. Since no oxen were available, Wiles’s listeners showed their appreciation by clapping their hands. But was the proof of Fermat’s last theorem the last gasp of a dying culture? Mathematics, that most tradition-bound of intellectual enterprises, is undergoing profound changes. For millennia, mathematicians have measured progress in terms of what they can demonstrate through proofs—that is, a series of logical steps leading from a set of axioms to an irre- futable conclusion. Now the doubts riddling modern human thought have ﬁnally infected mathematics. Mathematicians may at last be forced to accept what many scientists and phi- losophers already have admitted: their assertions are, at best, only provisionally true, true until proved false. This uncertainty stems, in part, from the growing complexity of mathematics. Proofs are often so long and complicated that they are di¤cult to evaluate. Wiles’s demonstration runs to 200 pages—and experts estimate it could be ﬁve times longer if he spelled out all its elements. One observer as-

“VIDEO PROOF” dramatizes a theorem, proved by William P. Thurston of the Mathematical Sciences Re- search Institute (left), that establish- es a profound connection between topology and geometry. The theorem shows how the space surround- ing a complex knot (represented by the lattice in this scene) yields a “hyperbolic” geometry, in which parallel lines diverge and the sides of pentagons form right angles. The computer-generated video, called Not Knot, was produced at the Ge- ometry Center in Minnesota.

Copyright 1993 Scientific American, Inc. serted that only one tenth of 1 percent vocates doing away with proofs alto- time when funds and jobs are scarce, of the mathematics community was gether, some practitioners think the va- young mathematicians cannot a›ord to qualified to evaluate the proof. Wiles’s lidity of certain propositions may be ignore these exhortations. claim was accepted largely on the ba- better established by comparing them There are pockets of resistance, of sis of his reputation and the reputa- with experiments run on computers or course. Some workers are complaining tions of those whose work he built on. with real-world phenomena. “Within the bitterly about the computerization of Mathematicians who had not yet exam- next 50 years I think the importance of their field and the growing emphasis ined the argument in detail nonethe- proof in mathematics will diminish,” on (oh, dirty word) “applications.” One less commented that it “looks beauti- says Keith Devlin of Colby College, who of the most vocal champions of tra- ful” and “has the ring of truth.” writes a column on computers for No- dition is Steven G. Krantz of Washing- Another catalyst of change is the tices of the American Mathematical So- ton University. In speeches and articles, computer, which is compelling mathe- ciety. “You will see many more people Krantz has urged students to choose maticians to reconsider the very nature doing mathematics without necessarily mathematics over computer science, of proof and, hence, of truth. In recent doing proofs.” which he warns could be a passing fad. years, some proofs have required enor- Powerful institutional forces are pro- Last year, he recalls, a National Science mous calculations by computers. No mulgating these heresies. For several Foundation representative came to his mere human can verify these so-called years, the National Science Foundation university and announced that the agen- computer proofs, just other comput- has been urging mathematicians to be- cy could no longer a›ord to support ers. Recently investigators have pro- come more involved in computer sci- mathematics that was not “goal-orient- posed a computational proof that of- ence and other fields with potential ap- ed.” “We could stand up and say this is fers only the probability—not the cer- plications. Some leading lights, notably wrong,” Krantz grumbles, “but mathe- tainty—of truth, a statement that some Phillip A. Gri¤ths, director of the Insti- maticians are spineless slobs, and they mathematicians consider an oxymoron. tute for Advanced Study in Princeton, don’t have a tradition of doing that.” Still others are generating “video proofs” N.J., and Michael Atiyah, who won a David Mumford of Harvard Universi- in the hopes that they will be more per- Fields Medal (often called the Nobel ty, who won a Fields Medal in 1974 for suasive than page on page of formal Prize of mathematics) in 1966 and now research in pure mathematics and is terminology. heads Cambridge’s Isaac Newton Insti- now studying artificial vision, wrote re- At the same time, some mathemati- tute for Mathematical Sciences, have cently that “despite all the hype, the cians are challenging the notion that likewise encouraged mathematicians to press, the pressure from funding agen- formal proofs should be the supreme venture forth from their ivory towers cies, et cetera, the pure mathematical standard of truth. Although no one ad- and mingle with the real world. At a community by and large still regards

A Splendid Anachronism?

hose who consider experimental mathematics and with scientists, Wiles worked in virtual solitude for seven T computer proofs to be abominations rather than inno- years. He shared his ideas with only a few colleagues to- vations have a special reason to delight in the conquest of ward the end of his quest. Fermat’s last theorem by Andrew J. Wiles of Princeton Uni- Wiles’s proof has essentially the same classical, deductive versity. Wiles’s achievement was a triumph of tradition, run- form that Euclid’s geometric theorems did. It does not in- ning against every current in modern mathematics. volve any computation, and it claims to be absolutely—not Wiles is a staunch believer in mathematics for its own probably—true. Nor did Wiles employ computers to repre- sake. “I certainly wouldn’t want to see mathematics just sent ideas graphically, to perform calculations or even to being a servant to applications, because it’s not even in compose his paper; a secretary typed his hand-written notes. the interests of the applications themselves,” he says. He concedes that testing conjectures with computers The problem he solved, first posed may be helpful. In the 1970s comput- more than 350 years ago by the er tests suggested that a far-fetched French polymath Pierre de Fermat, is proposal called the Taniyama conjec- a glorious example of a purely math- ture might be true. The tests spurred ematical puzzle. Fermat claimed to work that laid the foundation for have found a proof of the following Wiles’s own proof. proposition: for the equation X N + Y N Nevertheless, Wiles doubts he will = Z N, there are no integral solutions take the trouble to learn how to per- for any value of N greater than 2. form computer investigations. “It’s The efforts of mathematicians to find a separate skill,” he explains, “and if the proof (which Fermat never did you’re investing that much time on a disclose) helped to lay the founda- separate skill, it’s quite likely it’s tak- tion of modern number theory, the ing you away from your real work study of whole numbers, which has on the problem.” recently become useful in cryptogra- He rejects the possibility that there phy. Yet Fermat’s last theorem itself may be a finite number of truths ac- “is very unlikely to have any applica- cessible to traditional forms of in- tions,” Wiles says. quiry. “I disagree vehemently with Although funding agencies have the idea that good theorems are run- been encouraging mathematicians to ning out,” he says. “I think we’ve collaborate, both with each other and barely scratched the surface.”

94 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. computers as invaders, despoilers of the sacred ground.” Last year Mumford proposed a course in which instructors would show students how to program a computer to find solutions in advanced calculus. “I was vetoed,” he recalled, “and not on the grounds—which I expected—that the students would complain, but because half of my fellow EXPERIMENTAL MATHEMATICIAN Jean E. Taylor of Rutgers University seeks the teachers couldn’t program!” rules governing minimal surfaces by studying real phenomena, such as soap bub- That situation is changing fast, if bles, and computer-generated ones, such as idealized crystals (left). the University of Minnesota’s Geometry Center is any indication. Founded two years ago, the Geometry Center occu- ly and doing calculations on comput- bubbles with a sophisticated computer pies the fifth floor of a gleaming, steel ers. Last year some of the center’s facul- graphics program. She has also graduat- and glass polyhedron in Minneapolis. ty helped to found a journal, Experimen- ed from soap bubbles to crystals, which It receives $2 million a year from the tal Mathematics, that showcases such conform to somewhat more complicat- National Science Foundation, the De- work. “Experimental methods are not ed rules about minimal surfaces. To- partment of Energy and the university. a new thing in mathematics,” observes gether with Frederick J. Almgren of The center’s permanent faculty mem- the journal’s editor, David B. A. Epstein Princeton and Robert F. Almgren of the bers, most of whom hold positions else- of the University of Warwick in England, University of Chicago (her husband and where, include some of the most promi- noting that Carl Friedrich Gauss and stepson, respectively) and Andrew R. nent mathematicians in the world. other giants often performed experi- Roosen of the National Institute of Stan- On a recent day there, several young mental calculations before constructing dards and Technology, Taylor is trying sta› members are editing a video dem- formal proofs. “What’s new is that it’s to mimic the growth of snowflakes and onstrating how a sphere can be mashed, respectable.” Epstein acknowledges that other crystals on a computer. Increas- twisted, yanked and finally turned in- not all his co-workers are so accepting. ingly, she is collaborating with materi- side out. In a conference room, three “One of my colleagues said, ‘Your jour- als scientists and physicists, swapping computer scientists from major univer- nal should be called the Journal of Un- mathematical ideas and programming sities are telling a score of high school proved Theorems.’ ” techniques in exchange for clues about teachers how to create computer graph- how real crystals grow. ics programs to teach mathematics. Bubbles and Tortellini Another mathematician who has Other researchers sit at charcoal-col- prowled cyberspace in search of novel ored NeXT terminals, pondering luridly A mathematician who epitomizes the minimal surfaces is David A. Ho›man hued pictures of four-dimensional “hy- new style of mathematics is Jean E. of the University of Massachusetts at percubes,” whirlpooling fractals and lat- Taylor of Rutgers University. “The idea Amherst. Among his favorite quarry are tices that plunge toward infinity. No that you don’t use computers is going catenoids and helicoids, which resemble paper or pencils are in sight. to be increasingly foreign to the next the pasta known as tortellini and were At one terminal is David Ben-Zvi, a generation,” she says. For two decades, first discovered in the 18th century. “We Harpo Marx–haired junior at Princeton Taylor has investigated minimal surfac- gain a tremendous amount of intuition who is spending six months here explor- es, which represent the smallest possi- by looking at images of these surfaces ing nonlinear dynamics. He dismisses ble area or volume bounded by a curve on computers,” he says. the fears of some mathematicians that or surface. Perhaps the most elegant In 1992 Ho›man, Fusheng Wei of Am- computers will lure them away from and simple minimal surfaces found in herst and Hermann Karcher of the Uni- the methods that have served them so nature are soap bubbles and films. Tay- versity of Bonn speculated on the exis- well for so long. “They’re just afraid of lor has always had an experimental tence of a new class of helicoids, ones change,” he says mildly. bent. Early in her career she tested her with handles. They succeeded in repre- The Geometry Center is a hotbed of handwritten models of minimal sur- senting these helicoids—the first discov- what is known as experimental mathe- faces by dunking loops of wire into a ered since the 18th century—on a com- matics, in which investigators test their sink of soapy water. puter and went on to produce a formal ideas by representing them graphical- Now she is more likely to model proof of their existence. “Had we not

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 95 been able to see a picture that roughly ciples of nonlinearity. In general, the spits out solutions, which are then iter- corresponded to what we believed, we color, or “state,” of each cell is deter- ated, or fed back, into the equation. would never have been able to do it,” mined by the state of its neighbors. A The mathematics underlying the set Ho›man says. change in the state of a single cell trig- had been invented more than 70 years The area of experimental mathemat- gers a cascade of changes throughout ago by two Frenchmen, Gaston Julia ics that has received the lion’s share the system. and Pierre Fatou, but computers laid of attention over the past decade is One of the most celebrated of cel- bare their baroque beauty for all to known as nonlinear dynamics or, more lular automata was invented by John see. When plotted on a computer, the popularly, chaos. In general, nonlinear H. Conway of Princeton in the early Mandelbrot set coalesces into an image systems are governed by a set of sim- 1970s. Conway has proved that his au- that has been likened to a tumorous ple rules that, through feedback and tomaton, which he calls “Life,” is “unde- heart, a badly burned chicken and a related e›ects, give rise to complicated cidable”: one cannot determine wheth- warty snowman. The image is a fractal: phenomena. Nonlinear systems were in- er its patterns are endlessly variegated its fuzzy borders are infinitely long, vestigated in the precomputer era, but or eventually repeat themselves. Scien- and it displays patterns that recur at computers allow mathematicians to ex- tists have seized on cellular automata as di›erent scales. plore these systems and watch them tools for studying the origin and evolu- Researchers are now studying sets evolve in ways that Henri Poincaré and tion of life. The computer scientist and that are similar to the Mandelbrot set other pioneers of this branch of mathe- physicist Edward Fredkin of Boston Uni- but inhabit four dimensions. “The kinds matics could not. versity has even argued that the entire of complications you get here are the Cellular automata, which divide a universe is a cellular automaton. kinds you get in many di›erent scienc- computer screen into a set of cells More famous still is the Mandelbrot es,” says John Milnor of the State Uni- (equivalent to pixels), provide a partic- set, whose image has become an icon versity of New York at Stony Brook. ularly dramatic illustration of the prin- for the entire field of chaos since it was Milnor is trying to fathom the proper- popularized in the early 1980s by Be- ties of the four-dimensional set by ex- noit B. Mandelbrot of the IBM Thomas J. amining two-dimensional slices of it Watson Research Center. The set stems generated by a computer. His prelimi- from a simple equation containing a nary findings led o› the inaugural issue complex term (based on the square root of Experimental Mathematics last year. of a negative number). The equation Milnor, a 1962 Fields Medalist, says he

HELICOID WITH A HOLE (bottom left) was discovered last year by David A. Ho›- man of the University of Massachusetts at Amherst and his colleagues, with the help of computer graphics. Edward C. Thayer, one of Ho›man’s graduate students, recently found a structure (below) that mimics the pattern of certain polymers.

98 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. occasionally performed computer experiments in the days of punch cards, but “it was a miserable process. It has become much easier.” The popularity of graphics-oriented mathematics has provoked a back- lash. Krantz of Washington University charged four years ago in the Mathe- matical Intelligencer that “in some circles, it is easier to obtain funding to buy hardware to generate pictures of fractals than to obtain funding to study algebraic geometry.” A broader warning about “specula- tive” mathematics was voiced this past July in the Bulletin of the American Mathematical Society by Arthur Ja›e of Harvard and Frank S. Quinn of the Vir- ginia Polytechnic Institute. They suggested that computer experiments and correspondence with natural phenomena are no substitute for proofs in es- UNEARTHLY LANDSCAPES emerge when a computer generates “slices” of a four- tablishing truth. “Groups and individu- dimensional map similar to the well-known Mandelbrot set. John Milnor of the State als within the mathematics community University of New York at Stony Brook studies similar two-dimensional images in have from time to time tried being less order to understand the properties of the complex mathematical object. compulsive about details of arguments,” Ja›e and Quinn wrote. “The results have been mixed, and they have occa- Indeed, veteran computer enthusiasts at least the ﬁrst 10 billion integers. In sionally been disastrous.” know better than most that computa- 1984, however, more extensive com- Most mathematicians exploiting com- tional experiments—whether involving putations revealed that eventually—at puter graphics and other experimental graphics or numerical calculations— numbers as high as 10(10 70 )—the pat- techniques agree that seeing should can be deceiving. One cautionary tale tern predicted by Mertens vanishes. not be believing and that proofs are involves the Riemann hypothesis, a fa- One potential drawback of comput- still needed to verify the conjectures mous prediction about the patterns dis- ers is that all their calculations are they arrive at through computation. “I played by prime numbers as they march based on the manipulation of discrete, think mathematicians were contem- toward inﬁnity. First posed more than whole numbers—in fact, ones and ze- plating their navels for too long, but 100 years ago by Bernhard Riemann, ros. Computers can only approximate that doesn’t mean I think proofs are ir- the hypothesis is considered to be one real numbers, such as pi or the square relevant,” Taylor says. Ho›man o›ers of the most important unsolved prob- root of two. Someone knowledgeable an even stronger defense of traditional lems in mathematics. about the rounding-o› functions of a proofs. “Proofs are the only laboratory A contemporary of Riemann’s, Franz simple pocket calculator can easily in- instrument mathematicians have,” he Mertens, proposed a related conjecture duce it to generate incorrect answers remarks, “and they are in danger of be- involving positive whole numbers; if to calculations. More sophisticated pro- ing thrown out.” Although computer true, the conjecture would have provid- grams can make more complicated and graphics are “unbelievably wonderful,” ed strong evidence that the Riemann elusive errors. In 1991 David R. Stoute- he adds, “in the 1960s drugs were un- hypothesis was also true. By the early myer, a software specialist at the Uni- believably wonderful, and some people 1980s computers had shown that Mer- versity of Hawaii, presented 18 experi- didn’t survive.” tens’s proposal did indeed hold for ments in algebra that gave wrong an-

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 99 swers when performed with standard perimental mathematics. As in all other Purists may have a harder time ignor- mathematics software. kinds of experiments, you can do them ing William P. Thurston, who is also an Stephen Smale of the University of wrong.” But he emphasizes that compu- enthusiastic booster of experimental California at Berkeley, a 1966 Fields tational experiments, intelligently per- mathematics and of computers in math- Medalist, has sought to place mathe- formed and analyzed, can yield more re- ematics. Thurston, who heads the Math- matical computation on a more secure sults than the old-fashioned conjecture- ematical Sciences Research Institute at foundation—or at least to point out the proof method. “In every other field of Berkeley and is a co-director of the Ge- size and location of the cracks running science there are a lot more experimen- ometry Center (with Albert Marden of through the foundation. Together with talists than theorists,” Wolfram says. “I the University of Minnesota), has im- Lenore Blum of the Mathematical Sci- suspect that will increasingly be the peccable credentials. In the mid-1970s ences Research Institute at Berkeley and case with mathematics.” he pointed out a deep potential con- Michael Shub of IBM, he has created “The obsession with proof,” Wolfram nection between two separate branches a theoretical model of a computer that declares, has kept mathematicians from of mathematics—topology and geome- can process real numbers rather than discovering the vast new realms of phe- try. Thurston won a Fields Medal for just integers. nomena accessible to computers. Even this work in 1982. Blum and Smale recently concluded the most intrepid mathematical exper- Thurston emphasizes that he be- that the Mandelbrot set is, in a technical imentalists are for the most part “not lieves mathematical truths are discov- sense, uncomputable. That is, one can- going far enough,” he says. “They’re ered and not invented. But on the sub- not determine with certainty whether taking existing questions in mathemat- ject of proofs, he sounds less like a dis- any given point on the complex plane ics and investigating those. They are ciple of Plato than of Thomas S. Kuhn, resides within or outside the set’s hir- adding a few little curlicues to the top the philosopher who argued in his 1962 sute border. These results suggest that of a gigantic structure.” book, The Structure of Scientific Revo- “you have to be careful” in extrapolat- Mathematicians may take this view lutions, that scientific theories are ac- ing from the results of computer ex- with a grain of salt. Although he shares cepted for social reasons rather than periments, Smale says. Wolfram’s fascination with cellular auto- because they are in any objective sense These concerns are dismissed by Ste- mata, Conway contends that Wolfram’s “true.” “That mathematics reduces in phen Wolfram, a mathematical phys- career—as well as his contempt for principle to formal proofs is a shaky icist at the University of Illinois. Wol- proofs—shows he is not a real mathe- idea” peculiar to this century, Thurston fram is the creator of Mathematica, matician. “Pure mathematicians usually asserts. “In practice, mathematicians which has become the leading mathe- don’t found companies and deal with prove theorems in a social context,” he matics software since first being mar- the world in an aggressive way,” Life’s says. “It is a socially conditioned body keted five years ago. He acknowledg- creator says. “We sit in our ivory tow- of knowledge and techniques.” es that “there are indeed pitfalls in ex- ers and think about things.” The logician Kurt Gödel demonstrated more than 60 years ago through his incompleteness theorem that “it is impossible to codify mathematics,” Thur- ston notes. Any set of axioms yields statements that are self-evidently true but cannot be demonstrated with those axioms. Bertrand Russell pointed out even earlier that set theory, which is the basis of much of mathematics, is rife with logical contradictions related to the problem of self-reference. (The self- contradicting statement “This sentence is false” illustrates the problem.) “Set theory is based on polite lies, things we agree on even though we know they’re not true,” Thurston says. “In some ways, the foundation of mathematics has an air of unreality.” Thurston thinks highly formal proofs are more likely to be flawed than those appealing to a more intuitive level of understanding. He is particularly enam- ored of the ability of computer graphics

PARTY PROBLEM was solved after a vast computation by Stanislaw P. Radziszow- ski and Brendan D. McKay. They calculated that at least 25 people are required to ensure either that four people are all mutual acquaintances or that ﬁve are mutual strangers. This diagram, in which red lines connect friends and yellow lines link strangers, shows that a party of 24 violates the dictum.

he continuing penetration of computers into mathe- and logic “that have stumped people for years.” Another T matics has revived an old debate: Can mathematics is Siemeon Fajtlowicz of the University of Houston, inven- be entirely automated? Will the great mathematicians of tor of a program, called Graffiti, that has proposed “thou- the next century be made of silicon? sands” of conjectures in graph theory. In fact, computer scientists have been working for de- None of these achievements comes close to satisfying cades on programs that generate mathematical conjec- the “profound effect” criterion, according to David Mum- tures and proofs. In the late 1950s the artificial-intelli- ford of Harvard University, a judge for the prize. “Not now, gence guru Marvin Minsky showed how a computer could not 100 years from now,” Mumford replies when asked to “rediscover” some of Euclid’s basic theorems in geometry. predict when the prize might be claimed. In the 1970’s Douglas Lenat, a former student of Minsky’s, Some observers think computers will eventually surpass presented a program that devised even more advanced our mathematical abilities. After all, notes Ronald L. Gra- geometry theorems. Skeptics contended that the results ham of AT&T Bell Laboratories, “we’re not very well adapt- were, in effect, embedded in the original program. ed for thinking about the space-time continuum or the Rie- A decade ago the computer scientist and entrepreneur mann hypothesis. We’re designed for picking berries or Edward Fredkin sought to revive the sagging interest in avoiding being eaten.” machine mathematics by creating what came to be known Others side with the mathematical physicist Roger as the Leibniz Prize. The prize, administered by Carnegie Penrose of the University of Oxford, who in his 1989 Mellon University, offers $100,000 for the first computer book, The Emperor’s New Mind, asserted that computers program to devise a theorem that has a “profound effect” can never replace mathematicians. Penrose’s argument on mathematics. drew on quantum theory and Gödel’s incompleteness the- Some practitioners of what is known as automated rea- orem, but he may have been most convincing when dis- soning think they may be ready to claim the prize. One is cussing his personal experience. At its best, he suggest- Larry Wos of Argonne National Laboratory, editor of the ed, mathematics is an art, a creative act, that cannot be Journal of Automated Reasoning. He claims to have devel- reduced to logic any more than King Lear or Beethoven’s oped a program that has solved problems in mathematics Fifth can.

to communicate abstract mathematical conjecture, which stated that four hues people are all mutual acquaintances or concepts to others both within and out- are su¤cient to construct even an infi- at least Y are mutual strangers? This side the professional community. Two nitely broad map so that no identical- number is known as a Ramsey number. years ago, at his urging, the Geometry ly colored countries share a border. In Previous proofs had established that Center produced a computer-generated some respects, the proof of Appel and 18 guests are required to ensure that “video proof,” called Not Knot, that dra- Haken was conventional—that is, it con- there are either four mutual acquain- matizes a ground-breaking conjecture sisted of a series of logical, traceable tances or four strangers. In their proof, he proved a decade ago [see illustration steps proceeding to a conclusion. The Radziszowski and McKay showed that on pages 92 and 93]. Thurston men- conclusion was that the conjecture could the Ramsey number for four friends or tions proudly that the rock band the be reduced to a prediction about the five strangers is 25. Socialites might Grateful Dead has shown the Not Knot behavior of some 2,000 di›erent maps. think twice about trying to calculate the video at its concerts. Since checking this prediction by Ramsey number for greater X’s and Y’s. Whether Deadheads grok the sub- hand would be prohibitively time-con- Radziszowski and McKay estimate that stance of the video—which concerns suming, Appel and Haken programmed their proof consumed the equivalent of how mathematical objects called three- a computer to do the job for them. 11 years of computation by a standard manifolds behave in a non-Euclidean Some 1,000 hours of computing time desktop machine. That may be a rec- “hyperbolic” space—is another matter. later, the machine concluded that the ord, Radziszowski says, for a problem Thurston concedes that the video is dif- 2,000 maps behave as expected: the in pure mathematics. ficult for nonmathematicians, and even four-color conjecture was true. The value of this work has been de- some professionals, to fathom, but he bated in an unlikely forum—the news- is undaunted. The Geometry Center is The Party Problem paper column of advice-dispenser Ann now producing a video of yet another Landers. In June a correspondent com- of his theorems, which demonstrates Other computer-assisted proofs have plained to Landers that resources spent how a sphere can be turned inside out followed. Just this year, a proof of the on the party problem should have been [see cover illustration]. Last fall, more- so-called party problem was announced used to help “starving children in war- over, Thurston organized a workshop by Stanislaw P. Radziszowski of the torn countries around the world.” Some at which participants discussed how Rochester Institute of Technology and mathematicians raise another objection virtual reality and other advanced tech- Brendan D. McKay of the Australian Na- to computer-assisted proofs. “I don’t be- nologies could be adapted for mathe- tional University in Canberra. The prob- lieve in a proof done by a computer, matical visualization. lem, which derives from work in set the- says Pierre Deligne of the Institute for Paradoxically, computers have cat- ory by the British mathematician Frank Advanced Study, an algebraic geometer alyzed a countertrend in which truth is P. Ramsey in the 1920s, can be phrased and 1978 Fields Medalist. “In a way, I obtained at the expense of comprehensi- as a question about relationships beam very egocentric. I believe in a proof bility. In 1976 Kenneth Appel and Wolf- tween people at a party. What is the if I understand it, if it’s clear.” While rec- gang Haken of the University of Illinois minimum number of guests that must ognizing that humans can make mis- claimed they had proved the four-color be invited to guarantee that at least X takes, he adds: “A computer will also

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 101 make mistakes, but they are much more intensified recently with the advent of di¤cult to find.” a technique that o›ers not certainty but Others take a more functional point only a statistical probability of truth. of view, arguing that establishing truth Such proofs exploit methods similar to is more important than giving mathe- those underlying error-correction codes, maticians an aesthetic glow, particularly which ensure that transmitted messag- if a result is ever to find an application. es are not lost to noise and other ef- Defenders of this approach, who tend fects by making them highly redundant. to be computer scientists, point out that The proof must first be spelled out pre- conventional proofs are far from im- cisely in a rigorous form of mathemat- mune to error. At the turn of the cen- ical logic. The logic then undergoes a tury, most theorems were short enough further transformation called arithmet- to read in one sitting and were pro- ization, in which “and,” “or” and oth- duced by a single author. Now proofs er functions are translated into arith- often extend to hundreds of pages or metic operations, such as addition and more and are so complicated that years multiplication. may pass before they are confirmed Like a message transformed by an by others. error-correction code, the “answer” of a The current record holder of all con- probabilistic demonstration is distribut- ventional proofs was completed in the ed throughout its length—as are any early 1980s and is called the classifica- errors. One checks the proof by query- tion of finite, simple groups. (A group ing it at di›erent points and determin- is a set of elements, such as integers, ing whether the answers are consistent; together with an operation, such as ad- as the number of checks increases, so dition, that combines two elements to does the certainty that the argument is get a third one.) The demonstration con- correct. Laszlo Babai of the University sists of some 500 articles totaling near- of Chicago, who developed the proofs ly 15,000 pages and written by more two years ago (along with Lance Fort- than 100 workers. It has been said that now, Carsten Lund and Mario Szegedy the only person who grasped the entire of Chicago and Leonid A. Levin of Bos- proof was its general contractor, Daniel ton University), calls them “transpar- Gorenstein of Rutgers. Gorenstein died ent.” Manuel Blum of Berkeley, whose last year. work helped to pave the way for Babai’s Much shorter proofs can also raise group, suggests the term “holographic.” doubts. Three years ago Wu-Yi Hsiang of Berkeley announced he had proved The Uncertain Future an old conjecture that one can pack the most spheres in a given volume by Whatever they are named, such proofs stacking them like cannonballs. Today have practical drawbacks. Szegedy ac- some skeptics are convinced the 100- knowledges that transforming a con- page proof is flawed; others are equal- ventional demonstration into the prob- ly certain it is basically correct. abilistic form is di¤cult, and the result Indeed, the key to greater reliability, can be a “much bigger and uglier ani- according to some computer scientists, mal.” A 1,000-line proof, for example, is not less computerization but more. could easily balloon to 1,0003 (1,000,- Robert S. Boyer of the University of 000,000) lines. Yet Szegedy contends Texas at Austin has led an e›ort to that if he and his colleagues can simplify squeeze the entire sprawling corpus of the transformation process, probabilis- modern mathematics into a single data tic proofs might become a useful meth- base whose consistency can be verified od for verifying mathematical proposi- through automated “proof checkers.” tions and large computations—such as The manifesto of the so-called QED those leading to the four-color theorem. Project states that such a data base will “The philosophical cost of this e¤cient enable users to “scan the entirety of method is that we lose the absolute cer- mathematical knowledge for relevant tainty of a Euclidean proof,” Babai not- results and, using tools of the QED sys- ed in a recent essay. “But if you do have tem, build upon such results with re- doubts, will you bet with me?” liability and confidence but without Such a bet would be ill advised, Levin the need for minute comprehension of believes, since a relatively few checks the details or even the ultimate founda- can make the chance of error vanish- tions.” The QED system, the manifesto ingly small: one divided by the num- proclaims rather grandly, can even “pro- ber of particles in the universe. Even vide some antidote to the degenerative the most straightforward conventional e›ects of cultural relativism and nihil- proofs, Levin points out, are suscepti- ism” and, presumably, protect mathe- ble to doubts of this scale. “At the mo- matics from the all-too-human willing- ment you find an error, your brain may ness to succumb to fashion. disappear because of the Heisenberg The debate over computer proofs has uncertainty principle and be replaced

102 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. by a new brain that thinks the proof is correct,” he says. Ronald L. Graham of AT&T Bell Lab- oratories suggests that the trend away from short, clear, conventional proofs that are beyond reasonable doubt may be inevitable. “The things you can prove may be just tiny islands, exceptions, compared to the vast sea of results that cannot be proved by human thought alone,” he explains. Mathema- ticians seeking to navigate uncharted waters may become increasingly dependent on experiments, probabilistic proofs and other guides. “You may not be able to provide proofs in a classical sense,” Graham says. Of course, mathematics may yield fewer aesthetic satisfactions as investigators become more dependent on computers. “It would be very discour- aging,” Graham remarks, “if somewhere down the line you could ask a computer if the Riemann hypothesis is correct and it said, ‘Yes, it is true, but you won’t be able to understand the proof.’ ” Traditionalists no doubt shudder at the thought. For now, at least, they can rally behind heros like Wiles, the con- queror of Fermat’s last theorem, who eschews computers, applications and other abominations. But there may be fewer Wileses in the future if reports from the front of precollege education are any guide. The Mathematical Scienc- es Research Institute at Berkeley, which is overseen by Thurston, has been holding an ongoing series of seminars with high school teachers to ﬁnd new ways to entice students into mathematics. This past January Lenore Blum, the institute’s deputy director, organized a seminar devoted to the question “Are Proofs in High School Geometry Obsolete?” The mathematicians insisted that proofs are crucial to ensure that a result is true. The high school teachers de- murred, pointing out that students no longer considered traditional, axiomatic proofs to be as convincing as, say, visual arguments. “The high school teachers overwhelmingly declared that most students now (Nintendo/joystick/MTV generation) do not relate to or see the importance of ‘proofs,’ ” the minutes of the meeting stated. Note the quotation marks around the word “proofs.”

FURTHER READING ISLANDS OF TRUTH: A MATHEMATICAL MYSTERY CRUISE. Ivars Peterson. W. H. Freeman and Company, 1990. THE PROBLEMS OF MATHEMATICS. Ian Stewart. Oxford University Press, 1992. PI IN THE SKY: COUNTING, THINKING, AND BEING. John D. Barrow. Oxford Universi- ty Press, 1992.

CRADA Mania Will joint research make U.S. industry competitive?

hen economic competitiveness became a front-burner issue W during the 1980s, Congress passed laws to entice companies to adopt technologies developed in government laboratories and encourage government researchers to team up with industry. The Clinton administration, keen to secure beneﬁts from research, is going even further down that road. It is telling the Department of Energy’s huge national laboratories, as well as the Na- tional Institutes of Health and other agencies, that they must make their expertise available to industry. But change is bringing problems in its wake. To show that they are solidly on board, o¤cials and laboratory directors point proudly to the explosion in cooperative research and development

agreements, or CRADAs. These agree- Sandia National Laboratories ments between government laboratories and industry or universities allow participants to share the costs of a collaboration. In return, the nongovern- ment partner is granted some rights RANDY MONTOYA over any o›spring, in the form of intel- ENERGY SECRETARY Hazel R . O’Leary tours Sandia National Laboratories in April af- lectual property. The national laborato- ter signing a CRADA with Sematech. With her is Senator Je› Bingaman of New Mexico. ries have been allowed to have CRADAs only since 1989, but they now boast close to 500. Their directors talk expan- rizing them was not accompanied by prompted Congressman Ronald L. Wy- sively of devoting 20 percent of their guidelines. The result was a profusion den of Oregon, a longtime scourge of budgets to CRADAs and like arrange- of CRADAs with widely di›erent terms. the industry, to demand to know why ments within a few years. The NIH, for one, has been changing its the NIH had not limited the price that Opinions on the value of CRADAs policies. Dinah Singer, an immunologist Bristol-Myers Squibb charges for Tax- vary widely in industry and even with- who chaired a committee that draft- ol, a drug whose anticancer properties in the government, despite their mush- ed CRADA guidelines for the NIH,com- were discovered by the National Cancer rooming growth. Some critics in Con- plains that some of the health institutes’ Institute. gress say CRADAs can easily turn into a early CRADAs failed to define clearly The inspector general of the Depart- fire sale of taxpayer-supported research. enough the area they covered. “It’s clear ment of Health and Human Services, in And companies complain that even with research scope should be limited to a report now under review at the de- the new laws, red tape means it is still where an invention has been made in a partment, also finds cause for concern too hard to collaborate with researchers federal laboratory and development is that the NIH’s CRADAs do not protect in government laboratories. “In many to be transferred to the private sector. the public interest. Fears that research- ways the jury is still out on these ar- It’s not appropriate for a CRADA to ap- ers could be left with divided loyalties rangements,” says Thomas H. Moss, di- ply to concepts or ideas,” she says. have prompted an interagency commit- rector of graduate studies and research CRADAs with drug companies are tee under the Department of Commerce at Case Western Reserve University in particularly controversial. A major un- to draw up conflict-of-interest guide- Cleveland. “It’s easy to imagine that peo- resolved question for the NIH is how— lines for federal employees in CRADAs. ple in government are moving to achieve or whether—to require companies that The most common complaint from the maximum number of CRADAs in- develop drugs based on an NIH discov- academic researchers and industry is dependently of substance.” ery to sell them at a “fair” price. Phar- the delay engendered by legal review of Indeed, the rush into CRADAs has maceutical companies refuse to divulge agreements. The story of Stephen M. been haphazard. The 1986 law autho- how they set their prices. Their reticence Boyle, a researcher at the Virginia-Mary-

104 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. land Regional College of Veterinary Med- from providing real working capital for Sciences in Ann Arbor, Mich. The DOE icine who has a CRADA with the Walter small businesses.” has recently indicated a willingness to Reed Army Institute of Research, is typi- Large companies have their own relax the U.S. preference provision. cal. Boyle says that although it took him gripes. The Department of Energy has The DOE has also refused to o›er in- and his collaborators at Walter Reed clashed with General Motors and the demnification against product liability only an hour to agree on a one-page Ford Motor Company over the DOE’s arising from joint work beyond a nar- statement covering their work on a hu- policies on product liability and a “U.S. row limit. But the limits the agency man brucellosis vaccine, it took law- preference” provision. Until recently the is proposing are unacceptable to indus- yers six months to approve it. The for- DOE required that products developed try, Sheridan asserts. Moreover, he says, mal document, Boyle recounts ruefully, from CRADA research be substantial- industry views as “unreasonable” at- is eight pages long and has the work ly manufactured in the U.S. or be other- tempts by the DOE to specify in advance statement attached at the end. wise beneficial to the domestic econ- the permissible field of application of The General Accounting O¤ce re- omy. But global companies—the princi- a technology developed in a CRADA. ported recently that the Department of pal users of DOE-developed technol- “Many of our industry members have Energy takes between three and five ogy—find these and other restrictions concluded that CRADAs do not work times longer to sign o› on a CRADA, on overly burdensome, says John J. Sheri- and have turned their attention else- average, than either the army or the Na- dan, director of technology sourcing at where,” Sheridan told Congress. tional Institute of Standards and Tech- the National Center for Manufacturing He supports instead alternative routes nology. At the DOE, eight months has been the average, and a year is not unusual. One reason, according to James E. Wells, Jr., an associate director for energy and science issues at the ac- A Wand for the Meter Reader counting o¤ce, is “the desire of some o¤cials to create a highly visible, sep- hen gaslights were introduced into homes and businesses in the late arately funded CRADA program in or- W 19th century, suppliers billed according to the number of lamps at the der to justify continued support for the establishment. After heating and cooking made gas consumption more vari- department’s weapons laboratories.” In able, the bellows meter was developed. A complicated mechanism, this in- addition, the DOE has a separate fund strument originally used a pair of inflatable sheep bladders linked by crank- for CRADAs, so it has to have a for- shaft and connecting rods to ratchet- and gear-driven dials. mal competitive selection and approv- Some 100 years later the bellows have been changed to rubber, but the al process. The National Institute of same ungainly design is still in use all over the world. The mundane meter Standards and Technology, in contrast, is the most common home appliance that is not directly purchased by con- which has less formal procedures, has sumers. More than 10 million are sold annually to gas utilities, supporting a been able to approve CRADAs in as lit- $700-million-a-year market. But not for much longer. These unsightly cast- tle as two weeks. iron lawn ornaments may be supplanted by a new electronic version devel- Another reason for the Department oped after five years of work by government boffins Down Under. of Energy’s slowness is that its national Researchers at Australia’s Commonwealth Scientific and Industrial Research laboratories are managed by indepen- Organization (CSIRO), in a joint project with the utility AGL Australia, have dent contractors. That means agree- come up with a smaller, less expensive and highly accurate alternative. The ments have to be monitored by DOE developers believe the new meters will last twice as long and reduce the po- headquarters in Washington, D.C. Sec- tential margin of error from as high as 10 to 0.2 percent. The devices may retary of Energy Hazel R. O’Leary told help make the monthly visit of the meter reader a piece of nostalgia, like the Congress in late July that the depart- hurdy-gurdy man coming down the block. ment will shortly produce a simplified, The CSIRO unit, a solid-state device about the size of a videotape cartridge, approved model CRADA for small busi- measures the velocity of an ultrasonic sound burst to determine gas flow. nesses. Moreover, by November labora- The 500-gram, battery-powered meter ricochets bursts of acoustical energy tory directors will be empowered to au- across a pair of transducers along a pipe 14 millimeters in diameter. The thorize CRADAs that involve expendi- sound pulse travels faster with the gas flow than against it. The fluid velocity tures of less than $500,000. O’Leary can be solved knowing just the upstream and downstream times along with said she will also cut the time for ap- the fixed distance between the detectors. A small, low-power custom micro- proving a CRADA at DOE headquarters processor handles all the control, calculation and display functions. from 32 weeks to fewer than 16. The meter is projected to have a useful operating life of more than 25 Delays are not industry’s only com- years and requires a battery change once every eight years. The key to this plaint. Because the DOE attracts about long duty cycle was the development of a highly efficient piezoelectric 10 times more CRADA proposals than transducer that converts the vibration of the sound into an electric signal. it can approve, many companies ex- The CSIRO meter can be read with an infrared wand, similar to a television pend substantial e›orts drafting and remote control, at a distance of three meters. negotiating a proposal only to have it CSIRO is not alone. For the past 10 years, gas companies throughout the turned down months later, notes John world have sponsored efforts to develop a more compact electronic system. L. Sullivan, software counsel of Cray In 1987 British Gas held a competition for a new design, for which it re- Research, Inc. Roger G. Little, chairman ceived 32 submissions. It chose to fund development projects from Sie- of the Solar Energy Industries Associa- mens, the German electronics giant, and Gill, a small British start-up compa- tion, told a House subcommittee in ny. Their systems are still shrouded in confidentiality. Meanwhile meter June that the requirement that indus- readers in California will be waving wands this autumn in a test of the trial partners match the value of gov- CSIRO meter. —Joshua Shapiro ernment contributions “keeps CRADAs

SCIENTIFIC AMERICAN October 1993 105 Copyright 1993 Scientific American, Inc. for technology transfer, such as direct funding of industrial research consortia. Yet while the NIH is seeking to re- strict CRADAs to narrow fields of work, the DOE is moving in the opposite direction, seeking larger CRADAs that involve industrial consortia. The conclusion that CRADAs have been less than a resounding success is supported by a study conducted by J. David Roessner and Anne F. Wise of the Georgia Institute of Technology. The KAMIL UGURBIL University of Minnesota study, which focused on industrial and HIGH-STRENGTH MRI SCANS from the University of Minnesota Medical School show federal laboratories, concluded that the areas of cortex both before (left) and during (right) stimulation by flashing lights. proportion of research-intensive companies that have contacts with federal laboratories was unchanged over the Magnetic Apprehensions But many of the safety studies used past five years. In other words, some machines of even lower field strengths companies are interacting more now Radiologists call for more or were conducted on cell cultures or than they were—but the pool of indus- testing of MRI’s effects animals, not people. Although function- trial collaborators is not expanding. al imaging has been demonstrated on Surprisingly, the companies in Roess- 1.5-tesla machines retrofitted with hard- ner and Wise’s report that did work with new generation of fast and very ware to allow faster scanning, scientists federal laboratories said that expecta- powerful magnetic resonance im- are looking to higher strength machines tions of licensing technology were not A aging (MRI) machines is provid- to provide the image resolution they the main reason for their forging links. ing scientists with an unparalleled view seek. There are only half a dozen or so Rather, it was access to the unique re- not just of tissues but also of the meta- experimental machines in the U.S. oper- sources of the laboratories. According bolic changes associated with function- ating in the neighborhood of four tes- to Roessner, technology-based compa- ing brains and other organs. Recent las, but more are being built. nies still see federal laboratories as rel- investigations have shown the activity Lawrence Berkeley Laboratory wants atively unimportant sources of new in specific areas of the brain during to build a 10-tesla machine. That field ideas, in comparison with universities thought, conversation or dreaming. But strength exceeds that of the six-tesla and other companies. This was not the even as the exciting findings are re- magnets in the proposed Superconduct- result that the legislation encouraging ported, a few radiologists are voicing ing Super Collider. CRADAs sought. concern that another round of testing Assessing MRI safety has proved dif- Congress, bu›eted by the conflict- is needed to evaluate biological e›ects ficult because the devices produce an ing opinions, is now considering ad- and safety issues for these advanced exotic combination of three disparate ditional measures to improve technol- scanners. “A lot of the work should be electromagnetic fields. In addition to ogy transfer in general and CRADAs repeated,” says Frank Shellock, a radi- the unchanging or static magnetic field, in particular. Legislation is pending in ologist and co-chairman of the safety another magnetic field, called a gradi- both the House and the Senate; most committee of the Society of Magnetic ent field, increases in intensity across of the e›orts focus on making it eas- Resonance Imaging. the body to provide the scanner with ier for industry to find a federal part- Shellock is concerned because to take information about the location of the ner. But easier access could give rise to rapid snapshots of physiological pro- tissues being imaged. Furthermore, coils more problems unless care is taken to cesses, the rate that scanner fields are emit rapid pulses of electromagnetic en- avoid the sort of situation that the not- pulsed through the body has been ac- ergy at radio frequencies to induce the for-profit Scripps Research Institute in celerated in the new machines, and the hydrogen nuclei in the body to move La Jolla, Calif., found itself in earlier magnetic field, already 30,000 times that out of alignment with the static field. this year. of the earth, has in some cases been in- The measurement of this misalignment Scripps was about to sign a large creased by a factor of three or four. “I yields an image. research agreement with Sandoz, the don’t feel you can look at the data ac- The initial focus of safety testing was Swiss-based pharmaceutical company, cumulated at the lower field strengths radio-frequency energy, because scien- when word leaked out. For $300 million and at di›erent radio frequencies and tists have long known that it can heat over 10 years, the agreement would assume that the same safety issues have body tissues, although the small tem- have given the drug company rights to been addressed at high fields,” Shel- perature rise caused by present MRI the commercial fruits of federally sup- lock notes. machines has been judged of no real ported research at Scripps. The deal Most scientists and radiologists are concern. Shellock, however, worries that was denounced by the then director of comfortable with the safety of the lower- di›erent radio-frequency pulse sequenc- the NIH, Bernadine Healy, as giving San- speed machines, with fields of about one es designed to improve the image con- doz “excessive control,” and it is now to 1.5 teslas, now in clinical use—in fact, trast and to shorten the time between being renegotiated. they are widely viewed as a benign alter- images for “functional” imaging could The day when Los Alamos and Sandia native to x-rays. “There have been thou- induce significant heating. For a healthy National Laboratories might be working sands of people scanned, and the iden- person, the temperature rise would be principally for an industrial conglomer- tifiable problems are mostly failure of similar to sitting at length in the sun ate is probably still far o›. But their di- the operator or the machine in one form on a hot day. But Shellock says ill or se- rectors had better be ready to face tough or another,” says Robert Balaban, chief dated patients with poor temperature questions as industrial influence at the of the laboratory of cardiac energetics regulation “may not be able to with- laboratories grows. —Tim Beardsley at the National Institutes of Health. stand those temperatures.”

106 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. The new machines speed imaging scientist. But the static magnetic field Mag Lift primarily by switching gradient fields in the high-field, four-tesla machines faster. But the rapid fluctuation of field has caused a small percentage of volun- Japan’s engineers push strengths has the potential to stimulate teers to report feeling nauseated, seeing the envelope for elevators nerves. At the threshold of stimulation, flashing lights or tasting metal in their people begin to experience sensations mouths—sensations that scanner manu- such as a light tapping on the nose, the facturers say can be reduced by having ong before Chicago’s Sears Tower small of the back or the inside of the a patient avoid moving his or her head or New York’s World Trade Cen- thigh. A subject in functional imaging ex- during an examination that may last Lter was built, Frank Lloyd Wright periments at Lawrence Berkeley Laborato- about half an hour. conceived of a city contained in a single ry reported that at levels not much high- To be certain these e›ects are benign, mile-high skyscraper. Wright’s dream er than the threshold, the feeling resem- Shellock and others are calling for more never went further from his drawing bled an electric shock to the abdomen. scrutiny. The only conference to date board than the pages of co›ee-table Despite these prospects, many radiol- devoted solely to the safety of MRI was books on architecture. Had the struc- ogists believe that rapidly switched gra- held in 1991 by the New York Academy ture been built, it would have been the dients are not harmful, a position with of Sciences. It drew a rare congregation world’s tallest walk-up. The hike would which the Food and Drug Administra- of radiologists, biologists and electrical have been necessary because conven- tion concurs. The FDA has recently al- engineers. “I was struck by how little tional elevator technology reaches its lowed an exception to its guidelines, so knowledge or widespread agreement limits at about 800 meters. Above that some machines now in clinical use oper- there is about the biological e›ects,” height, elevator cables become too thick ate just below the threshold of stimula- says radiologist Je›rey Weinreb, direc- and too heavy for practical use. tion. “Peripheral nerve stimulation can tor of MRI at the New York University Decades after Wright’s death in 1959, serve as an early warning to prevent Medical Center, who was a session chair- engineers in Japan with a similar pen- cardiac stimulation and is not in itself man. “The models and the numbers are chant for futuristic indulgences are plan- harmful,” says Whit Athey, a senior FDA not in.” —Karen Fitzgerald ning structures with similar spans—but

Chronologically Privileged

-SPAN viewers who watched Congress this spring tip- the poverty level 25 percent above that used for official C toe around the subject of Social Security cuts may statistics—“it’s widely believed that the thresholds are kept have smiled to hear politically correct legislators refer def- down so that poverty doesn’t look as bad as it is,” Rendall erentially to their elderly constituents as “the chronologi- explains—the researchers found that the poverty rate for cally privileged.” One point of their debate is whether the those age 65 and older dropped from 15.1 to 8.9 percent. senior citizenry is financially privileged as well. Yeasayers It dropped further for some groups of seniors than oth- point to U.S. Census Bureau statistics that show a poverty ers. “Blacks and other minorities tend to have lower as- rate since 1985 running about 10 percent lower for elders sets than whites, and the gap in assets is much bigger than for Americans as a whole. Naysayers question the bu- than the gap in incomes,” Speare observes. Elderly white reau’s method, which sets a lower poverty level for retirees households owned median total assets worth $84,760, in order to compensate for reduced work-related expenses. compared with a median of $12,121 for black elderly. To help clarify the situation, the American Association Rendall and Speare found a poverty rate of 28.3 percent of Retired Persons’ Andrus Foundation joined with the for black pensioners, 4.2 times that of elderly whites. University of Wisconsin to commission Michael S. Rendall The sociologists are quick to point out that their figures of Cornell University and Alden Speare, Jr., of Brown Uni- depend on a critical—and questionable—assumption: that versity to study poverty among the elderly. In addition to retirees will sell off their assets evenly and die broke, or measuring income by traditional methods, the sociolo- nearly so. They defend this postulate with three argu- gists counted elders’ assets and government subsidies. ments. First, Rendall maintains, whereas “rich elderly may The results may have come as a surprise. Despite raising accumulate substantial bequests, poorer elderly do not.” Second, Rendall contends, poverty by definition is invol- untary. “If an elderly person chooses to live in poverty in 35 INCOME ONLY order to leave a large bequest to his children, does the pub- 30 lic call this person poor? We argue that you don’t.” Finally, INCOME Speare cites current welfare policies. “Most government pro- AND ASSETS 25 grams disqualify people if their total assets exceed a very POVERTY RATE small amount. The assumption is that people should spend 20 FOR ALL AGES all their assets before they ask for assistance,” even though AND RACES this may leave them destitute for the rest of their lives. 15 OFFICIAL POVERTY RATE The researchers advocate policies based instead on new 10 FOR THE ELDERLY types of retirement financing, such as reverse equity mort- gages. “You make an agreement with your bank to give up 5

POVERTY RATE (PERCENT) all or part of your home when you die in exchange for current income,” Speare elaborates. Such options are not wide- 0 WHITE BLACK HISPANIC ly available, however, and the sociologists emphasize that SOURCE: Michael S. Rendall, Cornell University; Alden Speare, Jr., their model is a better measure for the prevention, rather

Brown University; U.S. Bureau of the Census (1984 data). LISA BURNETT than diagnosis, of poverty. —W. Wayt Gibbs

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 107 this time they are going both up and sei Corporation of Tokyo, one of Japan’s down. And they think they have a way most prominent construction compa- to break the elevator height barrier by nies, has built a small model in its EX- eliminating cable altogether. CEED 2000 project. Extreme crowding in Japanese cities The issue that needs to be resolved has drawn attention to the use of under- is safety. Cut the power to a linear mo- ground space. At the Geospace Engi- tor elevator and it will drop, though not neering Center, a subsidiary of the En- precipitously. This will require new con- gineering Advancement Association of cepts in braking systems. When brake Japan (ENAA), companies that include failure could result in a fall of a kilome- NEC, Fujitsu, Hitachi, Mitsubishi Elec- ter or more, anything less than perfec- tric and Fuji Electric have been carrying tion is unthinkable—and perfection is out preliminary work on these concepts an engineering parameter that is still in for seven years. ENAA is backed by Ja- short supply. —Robert Patton, Tokyo pan’s powerful Ministry of Internation- al Trade and Industry. The key to the cordless elevator is Power Pack the same linear motor technology now being applied to the development of Batteries are the bottleneck magnetically levitated rail transporta- for portable electronics tion systems. Under consideration is a linear synchronous motor that propels a passenger car along the tracklike guide- omputer and telecommunica- way running the length of the elevator tions manufacturers consider a shaft. The motor works by the interac- Csix-month product cycle an unre- tion of an electromagnetic field from markable phenomenon. For battery mak- electric coils on the guideway with mag- ers, eight to 10 years is a better-than-av- nets on the car. erage time to market. This disparity is In theory, the linear motor elevator, turning into a barrier for electronic de- free of cables, can travel up or down any signers trying to condense a panoply of distance at all as long as the power is o¤ce equipment into devices that can on. Another advantage might be that the rest within a palm or, at worst, fit the elevator can travel horizontally as easily dimensions of an airplane’s food tray. as it can vertically. In the subterranean “The battery is holding back addi- city, the same cars that transport urban tional system improvements,” says Mi- troglodytes to and from the surface can chael P. Dooley, director of strategy and move them horizontally from home to business development for Motorola En- o¤ce or theater or school. Moreover, ex- ergy Products. Even Motorola’s suppli- pensive, space-hogging elevator shafts ers do not disagree. William Bowden, a would no longer be limited to a single Duracell manager, admits that “the pace car. At any moment, several cars could of technological change in my industry be descending in the “down” shaft, and is glacial.” as each reaches the bottom of the shaft, Packing more power into a mixture of it could simply move sideways and chemicals stu›ed into tiny cans labeled then ascend the “up” shaft. AA is a laborious process not conducive ENAA is also looking skyward in what to the exponential gains in performance it calls Tokyo Eco-polis City 1000. This traditional for chip makers. Even so, proposed project involves the erection companies such as Motorola are trying of a building 1,000 meters tall that will to prod suppliers such as Duracell to try house 75,000 people, while providing harder. Toward this end, battery makers 700 hectares of o¤ce and commercial are attempting to improve their reaction space. ENAA has proposed the creation times by focusing on the great light of “large-scale man-made structures, hope: lithium. The lightest metal, it gives rich in greenery and free from the trau- up electrons more easily than does any mas of commuting”—a not-so-oblique other material used in batteries. reference to the travails of getting home Electrochemists have targeted lithi- in Tokyo. um as the best candidate for providing Fuji Electric, under contract with high-capacity, rechargeable batteries. Al- ENAA, has constructed a linear motor though tiny lithium disposable camera elevator to test the concept and has batteries are on the shelf in pharmacies, recently launched a two-year e›ort to manufacturers of rechargeable batteries test a larger model capable of moving have struggled with the tendency of this on both a vertical and horizontal path. highly energetic material to smoke or Working prototypes big enough for rid- to catch fire when exposed to repeat- ers are a long way o›. Current plans call ed charges and discharges. This occurs for a new model that measures some six because lithium tends to build up, or meters in height. But that does not stop “plate,” unevenly on the anode (the nega- Japanese engineers from dreaming. Tai- tive electrode) during the charging cycle.

108 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. That is apparently what happened when Moli Energy in Vancouver produced more than a million rechargeable cells for Japan’s Nippon Telephone and Telegraph. These lithium metal recharge- ables, as they are called, were quickly pulled o› the market in 1989 after 10 of them “vented”—smoke or flame was released from a battery vent because of a chemical interaction between the plat- ed anode and the liquid electrolyte. Battery companies have since focused their attention on ways to produce more benign lithium products. One lithium rechargeable that is beginning to reach the market, for example, is called a lithium-ion, or “rocking chair,” battery. During the charging cycle, a lithium ion “rocks” between a cathode (positive electrode) made of a material such as an oxide of cobalt, nickel or manganese and an anode of carbon or graphite. The important thing is that lithium-ion cells do not experience the dangerous build- up of metallic lithium on the anode. Rather, the lithium atoms insert themselves between layers of graphite or carbon in the anode, a process called intercalation. These lithium-ion cells store two to three times the energy per weight of conventional nickel-cadmium cells. The voltage—3.6 volts—enables a portable- electronics designer to gain a much- sought weight saving for a computer or camcorder because one cell can be used instead of two or three. But circuitry used to recharge nickel-cadmium or nickel metal-hydride cells needs to be redesigned for the lithium cells. Development of the lithium-ion battery has become the preoccupation chiefly of Japanese electronics manufacturers. Their domination of this technology, like the lead they have snared in flat-panel displays, ensures close com- patibility between component engineers, who draw up plans for individual batteries, and systems engineers, who design the consumer products powered by these cells. “We’re not only behind in the manufacturing but in the basic research for these batteries,” says Je›ery R. Dahn, a professor of physics at Si- mon Fraser University, who was formerly research director for Moli Energy. Evidence of the U.S. lag in development could be seen at the meeting of the Electrochemical Society in Hono- lulu in mid-May. Sanyo and Matsushi- ta reported increases as high as 50 percent in energy-storage capacity for lithium-ion prototypes over a commercial version of the battery that Sony Ener- gytec, a subsidiary of the electronics giant, has begun incorporating into cam- corders and cellular phones. Besides Sony, A&T Battery (a joint venture of

Copyright 1993 Scientific American, Inc. SCIENTIFIC AMERICAN October 1993 109 electrolyte is still too reactive, and it tends to crystallize. Despite the uncertain prospects, a Long Island company, Moltech, has already agreed to take out a license on the Arizona State electrolyte. 3M and others have also shown an interest in negotiating a license. The risks in new battery development have led to a bumper crop of consortia. The consumer market could, in fact, re- ceive a much-needed boost from the U.S. Advanced Battery Consortium (US- ABC), a joint e›ort of automakers and battery manufacturers to develop lithium-polymer and other batteries for an electric vehicle. In addition, the Micro- electronics and Computer Technology Corporation in Austin, Tex., has been trying to gather battery and electronics manufacturers into a collaboration that would adapt lithium-polymer technology to portable devices. And the Ad- vanced Research Projects Agency plans to distribute $5.7 million largely for research on lithium-polymer cells.

T PROCHNOW Getting battery companies, consumer electronics manufacturers and national ROBER laboratories to work together has often GREAT LIGHT HOPE is researcher Jean-Marie Tarascon’s ambition for this lithium- turned into a legal morass. It has taken ion cell that he helped to develop at Bell Communications Research. Sandia National Laboratories more than two years to set up a three-year consortium, with $300,000 in annual funding, Asahi and Toshiba) was slated to scale ers can be overcome, engineers craft- to adapt a carbon material created for up to production of hundreds of thou- ing the new lithium cell dream of plac- the moribund “Star Wars” program into sands of cells a month. ing the thin-layered cells in shapes that an electrode for lithium-ion batteries. The only U.S. institution to present a allow them to conform to the design of Duracell, a prospective participant, was lithium-ion paper at the conference was a phone or computer case. Amalgamat- unable to reach agreement with Sandia Bell Communications Research (Bell- ing these films into a producible battery and so was left out of the collaboration. core), the research arm of the regional still remains an unmet challenge. But the “The unfortunate thing is that when we telephone companies. Bellcore is inter- industry hopes that machinery for mak- started, we had a material that was two ested in a battery that could serve as ing cellophane and other plastic films times better than what Japan had, but a backup power source for the opti- can be adapted for the high-speed man- during the period of negotiating with cal-to-electronic conversion devices for ufacturing of polymer batteries. the lawyers, the Japanese have made a fiber-optic communications. Jean-Marie A lithium-polymer battery could ex- lot of advances,” says Samuel C. Levy, a Tarascon, leader of a Bellcore team, fash- ceed the energy capacity of its lithium- member of the technical sta› at Sandia. ioned a cathode from lithium, manga- ion cousin because it need not carry the An upstart in the lithium-polymer nese and oxygen. Manganese is less ex- burden of a carbon or graphite anode business is Valence Technology of San pensive than the cobalt in the batteries that acts merely as a storage material Jose, Calif. A securities analysis report developed by Japanese manufacturers. for the lithium. Instead lithium would by a Valence stock underwriter raised Bellcore, which like its Baby Bell par- constitute the entire anode. But such a eyebrows by suggesting that the bat- ents is prohibited from manufacturing, solid-state battery will become a real- teries could be produced for less than is licensing the technology to an undis- ity only if engineers can find polymers a dime for each watt-hour of energy. closed U.S. company. The big three U.S. that are more conductive and that do “That’s less than what they would pay manufacturers—Duracell, Eveready and not degrade after 150 or so cycles. to buy the lithium,” Dahn comments. Rayovac—also have in-house research Some lithium-polymer developers Even so, Motorola signed an exclusive teams working on lithium-ion batteries. have tried to make the batteries con- agreement with Valence to buy $100- The Japanese lead in lithium-ion bat- duct more electricity by adding small million worth of the start-up compa- teries may cause U.S. companies to con- amounts of a conductive liquid solvent ny’s batteries if Valence can meet the centrate on another lithium technolo- to the polymer. But the cell becomes big electronics manufacturer’s technical gy still secreted in the laboratory. This a candidate for the same safety prob- and cost specifications. The announce- material is a lithium-polymer combina- lems experienced by Moli Energy. ment sent the company’s stock soaring tion. Anodes of lithium-polymer batter- Basic research on the polymer electro- for a time. “We have nothing commit- ies consist of a layer of thin lithium foil lytes is inching forward. A team at Ari- ted up front,” says Dooley of Motoro- alternating with a solid polymer elec- zona State University found that adding la. “The message we’re trying to con- trolyte, such as polyethylene oxide. A a small amount of polymer to lithium vey is that it’s pretty clear that Motoro- material like vanadium oxide is used as salts produced conductivity comparable la doesn’t want to sit around and wait the cathode. to that found in many liquid electro- until the battery industry does some- If substantial developmental barri- lytes. But the rubbery “polymer-in-salt” thing new.” —Gary Stix

ADigital Fix for the Third World? India, still record transactions by hand. “In these places, what you may see are enclaves of high-expertise engineers and he notion that technology can fix Telecommunications linkages in Gha- high-skilled technicians with telecom- the economic problems of the de- na, for example, might tie cocoa produc- munications links to wherever they’re T veloping world dates back to the ers more closely to customers in world getting their orders—and a barbed-wire early years of the post-colonial era. The markets, allowing them to wheel and fence between them and the rest of the “green revolution” of the 1960s held deal more rapidly and to increase prof- economy,” says W. Brian Arthur, a pro- out the hope that the labors of Western its that could eventually flow back to fessor of economics at Stanford Univer- plant geneticists could eliminate food farmers. Service jobs that require man- sity. “The proceeds of what they do may shortages by raising crop yields. ual processing of data are being moved go to some Swiss bank, so they may as The information age has begun to from back o¤ces in high-priced coun- well be on the moon.” raise similar expectations. African heads tries to provide jobs for city dwellers Africa also has measured some gains of state, such as Flight Lieutenant Jerry in the developing world. from information technology and tele- J. Rawlings, who staged the overthrow communications, but a huge gap re- of a civilian government in Ghana more mains in the ability to meet most coun- than a decade ago and last year was Information technology tries’ needs. The cocoa board in Ghana elected as president, sound a little like uses Morse code to communicate with executives from Silicon Valley. Rawlings may bring limited benefits o¤ces all over the country. now talks about an “enabling environ- to the work force The government in Ghana now al- ment” for e¤cient telecommunications of developing nations. lows private companies to compete with in a country where many employees its own Posts and Telecommunications spend about half their time going from Corporation, which provides 48,000 tele- one place to another because of the un- phone lines, according to an article in a reliability of the telephones. Moreover, networks could keep aca- journal called African Technology For- Rawlings’s neighbor, Felix Houphouet- demics in developing nations in touch um. Ghana could accommodate at least Boigny, longtime leader of Ivory Coast, with colleagues in the North. The Na- 200,000 telephone lines, observers es- voices a similar hopeful refrain: “Thanks tional Science Foundation is contemplat- timate. But at $1,000 to install a hand- to informatics, technological shortcuts ing the funding of U.S. university pro- set—more than the nation’s average per to development exist today and are grams to set up data banks that would capita income—demand for privately within the reach of everyone. We are let both U.S. and Latin American engi- supplied telephones will remain limited. not therefore doomed to remain unde- neers and scientists share information Instead, where information technolo- veloped forever.” over the Internet data network. gy has become a boon is in the newly in- For all its promise, the allure of skip- Dogging all these new-age ideas is the dustrialized nations, those with the or- ping over several generations of tech- fact that the green revolution and other ganizational and planning skills to make nology has fallen into disfavor in some technological salvations were not exact- full use of the technology. Singapore circles. Economists often adhere to the ly unmitigated successes. Agricultural saw a dramatic increase in revenues “product cycle” approach to di›usion aid definitely improved crop yields for from software and services during the of technology. This trickle-down theo- larger farmers with the know-how and 1980s. The small nation’s advances re- ry holds that developing nations ab- capital to buy and use hybrid seed, fer- sulted from an aggressive government sorb new technology only after it has tilizer and the other accoutrements of a policy of building telecommunications become widely established in industri- midwestern U.S. farm. But in India and networks and promoting a domestic alized countries. Costa Rica assembles Pakistan, the poorest subsistence farm- software industry. Other nations—Ven- baseballs for the major leagues; Du Pont ers and unlanded laborers were still ezuela and Mexico are two—have wit- turns out aerospace composites. Oth- left on the margins. nessed a rise in supply of telecommu- er economists also point to the billions Similarly, the benefits of PCs and nications services after privatization of of dollars wasted on aid that recipients pocket phones have been confined to a government telephone monopolies. have been unable to integrate into their narrow segment of society in developing Indeed, some economists worry that economies, from tractors without gas countries. In India, a more than $200- information technology may actually stations to steel mills without transport million software industry has emerged, draw jobs away from the developing for ore or finished products. centered around Bombay and the south- world, further exacerbating the gap be- Many preach an enlightened cousin ern city of Bangalore. Software houses tween rich and poor. Just-in-time in- of the trickle-down approach known as there have crafted, among other applica- ventory control methods and comput- appropriate technology. And the infor- tions, programs that let U.S. consumers erized manufacturing have become a mation advocates argue that high-tech buy gasoline automatically with a cred- form of intellectual and organization- electronic tools are appropriate—$1,000 it card, according to Pyramid Research, a al capital that outstrips the importance microcomputers can do the work that consulting firm in Cambridge, Mass. Yet of raw labor costs. In short, Ashoka required $1-million mainframes in the less than than 1 percent of Indians have Mody, a World Bank economist, warns, industrialized world only a decade ago, phones, compared with 96 percent of “the low wages in the developing coun- and cellular telephones can fend o› the Americans. Most branches of the countries can get lost in the noise of cost ac- lengthy process of laying land lines. try’s largest bank, the State Bank of counting.” —Gary Stix and Paul Wallich

Making Fluids into Solids with Magnets could scrape a piece of iron with a good metal file; or you could gather the iron filings produced by a metal lathe. nyone can turn a liquid into a sol- fields. These electrorheological fluids The size of the filings will determine id using an ice tray and a freez- have been incorporated in prototype the magnitude of the magnetorheologi- A er. But can you solidify a fluid brake systems, clutches, shock absorb- cal e›ect observed. The filings should and then liquefy it within a few sec- ers, engine mounts, actuators, control be large enough to identify single parti- onds? Actually, all you need is a strong valves and even artificial joints [see cles with a magnifying glass but prefer- magnetic field and a simple recipe for a “Electrorheological Fluids,” by Thomas ably less than half a millimeter long. magnetorheological fluid. C. Halsey and James E. Martin, page 58]. Filings that are too large can easily be Such mixtures consist of particles Yet amateur scientists would be wise ground to a smaller size with a spoon suspended in a fluid. When the sub- to experiment with magnetorheological and a hard surface. Be aware that iron stance is exposed to a magnetic field, fluids before they attempt anything with will scratch the surface. the particles align in such a way that the electrical counterparts. Most electro- The type of oil needed to suspend the the mixture becomes solid. When the rheological fluids require high voltages. filings does not matter much. It just has field is removed, the particles return to For example, several thousand volts to be more viscous than water, to slow their random state, and the substance would be needed to solidify a mixture down the rate at which the filings set- liquefies. The degree of solidification de- of corn oil and cornstarch. (The actual tle. I have used corn oil, linseed oil, sili- pends on the inherent properties of the amount of electric current that would cone oil and motor oil. Corn oil is per- fluid and the strength of the applied flow through the liquid would be quite haps best: it is cheap and cleans up rel- magnetic field. low, however.) atively easily with soap and water. Magnetorheological fluids are not as Although present magnetorheologi- The filings and oil should be mixed in common as a similar class of materials cal fluids are not as versatile as their a small, clear container, such as a clean that solidify in the presence of electric electrical cousins, they do demonstrate aspirin bottle, a plastic cup or, as I used, the e›ect that so excites many research- a 100-milliliter beaker. Just make sure ers. The experiment described here can that you can see down into the bottle. be performed on the kitchen table for The fluid should consist of approximate- DANIEL J. KLINGENBERG is an assis- less than $20. ly 25 parts by weight oil to 100 parts tant professor of chemical engineering A simple magnetorheological fluid is by weight iron filings. The exact propor- at the University of Wisconsin–Madison. made of iron filings and corn oil. You tions of the constituents, however, are He received his Ph.D. from the Universi- can find iron filings at a toy store, at not critical to observing some kind of ty of Illinois at Urbana-Champaign and has written several papers on electrorhe- your workbench or at a machine shop. magnetorheological e›ect. ological fluids. Specifically, you could empty the filings As for the magnetic field, almost any out of a toy magnetic sketch pad; you type of strong, permanent magnet will

a b

IRON FILINGS in corn oil produce a simple magnetorheolog- ing sti›, stringy masses that can be pushed to one side of the ical fluid. The mixture remains liquid (a) until a magnet is container (b). The material reverts back to its original liquid brought close to it. The filings line up between the poles, form- state once the magnetic field is removed.

112 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. work: horseshoes, bars, even the ones INSULATED WIRE from old loudspeakers. (Refrigerator magnets are typically far too weak.) CLEAR PLASTIC Magnets are often rated by the number CONTAINER of pounds they can pull. You should buy the most powerful ones available. I purchased two “100 pound” bar magnets from a local hardware store for about $12 apiece. Electromagnets are also an option. Such magnets are made by winding thin, insulated wire around a metal bar and connecting the ends of the wire to a bat- ELECTROMAGNET FREE ROD tery. The magnetic field will be strongest near the ends of the coiled wire. It does not matter whether you wind the wire from one end to the other or back and forth several times. Just be sure that the wire is always wound in the same direction (that is, clockwise or counter- clockwise) and that the winding is tight. MAGNETORHEOLOGICAL Two layers of windings over a couple FLUID (IRON FILINGS IN OIL) of inches of the core should su¤ce. + – To amplify the strength of the magnetic field, you can increase the power supplied to the electromagnet by con- POWER SUPPLY necting several batteries in series. Note, MAGNETORHEOLOGICAL BRAKE relies on a disk mounted on a rod. Turning on however, that the electromagnet draws the magnetic field prevents the rod from rotating. Batteries rather than a plug-in a lot of current and will quickly drain power supply can be used, although they will be drained quickly. the batteries. A power supply that can deliver several volts may be a better choice. Although it is quite fun trying tach end to end to one another, extend- use electromagnets, I suggest wiring to create a good electromagnet, it is ing from the poles of the magnet. Now several perpendicular to the disk in the hard to beat a permanent magnet in move the container away from the mag- fluid. Be sure to wind the wire around terms of both time and convenience if net. The filings-and-oil mixture will rap- each core in the same direction. your goal is to play with a magnetorhe- idly change back into a fluid: the clumps Before the magnetic field is applied, ological fluid. fall apart, and the surface becomes the rod rotates with little or no resis- Once you have your fluid and magnet fairly smooth and horizontal. If you do tance. (To make the rotation more vis- ready, you can observe magnetorheol- not observe the e›ects, you probably ible, you may wish to bend the end of ogy. First, stir the fluid with a piece of either need to get a stronger magnet the rod or attach some kind of han- wood or plastic (the filings will adhere or need to change the consistency of dle.) Applying the magnetic field solid- to a magnetizable object), keeping the the fluid. ifies the fluid. The rod is now di¤cult magnet at least six inches from the con- As a variation, repeat this experi- to turn. tainer. You should notice that the com- ment with di›erent proportions of the As you may notice with this combi- bination of filings and oil behaves as a filings and oil. The optimum combina- nation of iron filings and corn oil, such fluid: it continues to fill the container tion will depend on the size of the fil- fluids often do not solidify robustly and tends to form a smooth, flat surface ings and the type of oil. Also, attempt and tend to break down over time. Re- after you stop stirring. It has the con- to vary the distance between the poles search in industry and academia has sistency of maple syrup. of the magnet and the container or else made significant progress in improv- Bring the magnet near the contain- use a stronger magnet. ing these fluids for applications. For ex- er. If you are using a horseshoe-shaped A much more challenging project is ample, the Lord Corporation, based in magnet, put the container just inside to create a simple magnetorheological Cary, N.C., plans to market magnetorhe- the two legs. If you have two magnets, brake [see illustration above]. Machin- ological fluids that promise to use much place the opposite poles on either side ing skills may be needed here. Insert less magnetic energy to function than of the container, facing each other. The the end of a rod into a plastic container the formulation described here. But sev- closer the poles are, the stronger the filled with the magnetorheological flu- eral years will pass before these fluids magnetic field in the fluid will be. If id. The rod should be made of some- will drive and stop our cars. you have just one magnet, place one thing that cannot be magnetized, such of the poles directly on the side of the as plastic or aluminum. You might try container. attaching a plastic disk to the end of the Stir the fluid again and notice how rod in the fluid, because a larger surface FURTHER READING it begins to behave like wet sand. Dis- area in the fluid will yield a more pro- FERROHYDRODYNAMICS. R. E. Rosensweig. tinct, stringy clumps should form at the nounced e›ect. To keep the fluid from Cambridge University Press, 1985. bottom of the container. These clumps leaking out, you may need to place a ELECTRORHEOLOGICAL FLUIDS: PROCEED- INGS OF THE SECOND INTERNATIONAL should retain their shape and remain rubber O-ring around the rod where it CONFERENCE. Edited by J. D. Carlson, suspended in the fluid after you re- enters the container. A. F. Sprecher and H. Conrad. Technom- move the stirrer. The clumps appear Now all you need to do is immerse ic Publishing, Lancaster, Pa., 1990. stringy because the filings tend to at- the container in a magnetic field. If you

The Demographic Transition he spreads before us in 60 diagrams and a clearly declining growth rate, save figures and a couple of dozen tables. only Africa, where the growth rate is A CONCISE HISTORY OF WORLD POPU- Jump first to news exciting enough currently about at plateau. The author LATION, by Massimo Livi-Bacci. Trans- to warrant public celebration in the says firmly: “The fact remains that de- lated by Carl Ipsen. Blackwell Publish- streets. Across one graph field, two mographic change in the poor world in ers, 1992 ($49.95; paperbound, $17.95). curves compare two groups of coun- recent decades has on average pro- tries, the rich with the poor. The curves ceeded rapidly as compared to the path his readable account by a Floren- plot the smoothed annual percentage previously followed by the rich.” Of tine demographer and statistician rate of population increase over the course, no extrapolation can itself give T is plainly in the highest civic spir- past three centuries. We all know that a guarantee, but all signs are strong and it. Massimo Livi-Bacci displays the main the industrialized countries have man- clear. Absolute numbers will grow as strands of the thick skein of human aged control over both birth and death. the abounding young become parents, head counts unusually well. Their net growth curve peaked broadly but relative rates have begun a steady The book opens with the simpler around 1900 at a rate a little more than decline. mathematics of the discipline, expound- 1 percent a year and has by now fallen For demographers, our common fu- ed by example. Then it puts side by about fourfold. The poor countries too ture is played out across an abstract side the conflicting explanations that have entered the same transition. Their “strategic space,” a simple if fateful plot have governed attitudes toward popu- peak percentage rate of growth was whose vertical axis shows the average lation growth ever since the days of Rev- reached as recently as 1970 and has number of children born per woman, erend Thomas Malthus. Fully half the fallen steadily ever since, much more as its horizontal one marks the average text treats a striking contrast that has steeply than the comparison peak of expectancy of life at birth. How these opened between countries rich and poor 1900 among the well o›. To be sure, it are estimated is summarized. The two and its decisive correlates, both social had risen to almost 2.5 percent a year, parameters are far from independent, and economic. Throughout, the author so it has a long way to fall. for mortality among young women of grounds his views on the numbers, well Each major region of the develop- course reduces the number of children chosen and meaningful numbers that ing world taken separately also shows expected. (Note that abundant sperm matters here much less than rarer eggs; in less biological discourse the societal status of women is a major index to demographic realities.) On such plots, concave curves that dip from upper left to lower right mark the locus of all points that correspond to a particular annual percentage growth. Historical populations from the Paleolithic (largely conjectural), firmer numbers from 18th-century Europe, on to India and Russia about 1900, and a dozen more dot a crescent-shaped band with annual net growth rates all between 0 and 1 percent. That is the old enduring regime, once universal. In contrast, another plot shows the strategic shifts that have occurred since 1800 among 17 countries in Europe. The old ellipse that then enclosed all their dots has slid steadily downward and rightward. By 1980 they all occu- pied a still tighter ellipse, centered on a small negative growth rate. Once the number of children per woman clus- tered between four and seven, and the expectancy of life held around 30 years. The 1980 values show an average of fewer than two children per woman, and an expectancy of life beyond 75 years. City life dominates now; the demand for children’s labor drops; the CLIFF DWELLERS, a 1913 painting by George Wesley Bellows, is used as the cover cost and length of their education rise; illustration of A Concise History of World Population. women have more independence and

114 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. marry later in life; the control of births unresolvable only if we insist on find- rivals the postponement of death. Sur- ing hard and fast rules to explain com- vivorship has “increased immensely.” plex phenomena.” Both outcomes are A good century ago only one 15-year- only contingent scenarios, not fateful CORRESPONDENCE old in 10 could expect to reach the age compulsions. of 80; now about half of us do. But most Real human adaptation, studied here of the aging of the population has come over a variety of economic and social Reprints are available; to order, write not from better survival of the old but variables, is much more subtle. Our au- Reprint Department, SCIENTIFIC AMERI- from reduction in the young popula- thor sees the chronicle as a process of CAN, 415 Madison Avenue, New York, tion because of so few recent births. choices and constraints, both extended N.Y. 10017-1111 or fax inquiries to From now on, improvements in mortal- over time, neither of them ruled by any (212) 355-0408. ity will need to benefit the old; there is single simple factor. Hard constraints “no room for improvement” among the are surely real; more than a billion of Back issues: $6 each ($7 outside U.S.) young, for 98 percent of those born the poorest of us humans still live hoe prepaid. Most numbers available. Cred- reach reproductive age. (These are fig- in hand under the old regime. And it card (Mastercard/Visa) orders for ures for Italy; the U.S. with its cruel leg- choices are clearer; in China by mandate two or more issues accepted. To order, acy of slavery still has ample room to and in easier India as well, limitation of fax (212) 355-0408. better the vital statistics of our children.) fertility is genuine. All save a few of the The tribute to war, famine and pesti- national dots drift predictably along lence declines, too. A graph shows the the decisive path to population stabili- Index of articles since 1948 available crude death rate for Sweden between ty. The bottom line a century ahead is in electronic format. Write SciDex™, SCI- 1740 and 1920. Before 1820 we see not easy to calculate, but the wide trend ENTIFIC AMERICAN, 415 Madison Ave- three deadly peaks, each a few years to workable control is plain, a trend nue, New York, N.Y. 10017-1111, or fax wide, during which death rates more not only in our growing abilities but in (212) 355-0408. than doubled. As the decades passed, the increasingly common purpose of all the peaks became pinched between the improvement of human life. Photocopying rights are hereby grant- two trends—a slowly falling baseline ed by Scientific American, Inc., to li- and a faster falling maximum of deaths. braries and others registered with the There is by 1920 a 10-fold decline in the 16th-Century Shipwrights Copyright Clearance Center (CCC) to variation of the death rate. The new re- photocopy articles in this issue of SCI- gime has begun, substantially relieved VANGUARD OF EMPIRE: SHIPS OF EXPLO- ENTIFIC AMERICAN for the fee of $3.00 from the fear of sudden death. Grim war RATION IN THE AGE OF COLUMBUS, by per copy of each article plus $0.50 per remains as do emerging diseases, but Roger C. Smith. Oxford University Press, page. Such clearance does not extend as long as we fight o› novel disease 1993 ($35). to the photocopying of articles for pro- and keep the nuclear peace, their epi- motion or other commercial purposes. sodic toll will be small. ortuguese ports have a resource Correspondence and payment should Between 1950 and 1985 the dots that unique in Europe: by departing on be addressed to Copyright Clearance describe two dozen of the most popu- the summer trades, ships could Center, Inc., 27 Congress Street, Sa- P lous of the poorer lands show a general sail downwind toward the ocean islands lem, Mass. 01970. Specify CCC Ref- procession along the well-marked Euro- and America and yet count on fair erence Number ISSN 0036-8733/93. pean route, not yet nearly as far along winds out of the west to speed them $3.00 + 0.50. as are Europe and the neo-Europes over- home again as the seasons passed. seas (plus Japan). A worldwide demo- Young Prince Henry the Navigator Editorial correspondence should be graphic transition is patently in full soon determined to press his country’s addressed to The Editors, SCIENTIFIC progress. Conceding at least one dou- future at sea steadily southward along AMERICAN, 415 Madison Avenue, New bling of the head count in the next cen- the African coast. He died in 1460, York, N.Y. 10017-1111. Manuscripts tury, we still don’t face runaway growth. about when his skilled captains had are submitted at the authors’ risk and What about theories? In these com- first felt their way to the equator, but will not be returned unless accompa- plex human a›airs, so rich in feedback his pioneering quest survived him. Por- nied by postage. loops and in feed-forward anticipations, tugal and her neighbor, Spain, alone it is no easy matter to tease out true would send their flags around the Advertising correspondence should causes. Old Malthus could argue logical- world, until by about 1600 their empires be addressed to Advertising Manager, ly that population growth with dimin- experienced overt competition from the SCIENTIFIC AMERICAN, 415 Madison Av- ishing returns to farmers’ e›ort would other states of Atlantic Europe. No small enue, New York, N.Y. 10017-1111, or reduce per capita production, increase part of Portuguese seapower came from fax (212) 754-1138. poverty and bring both hard times and that early understanding of Atlantic a population decline. So it can; indeed, winds and currents. Subscription correspondence should the scheme fits the cycles of his time in These Iberians, major early users of be addressed to Subscription Manager, France and England. But about a genera- inanimate energy, moved by windpow- SCIENTIFIC AMERICAN, P.O. Box 3187, tion ago, another scholar of agricultural er and fought by gunpowder. In this Harlan, IA. 51537. The date of the last economics, Ester Boserup, argued quite crisply focused narrative we are shown issue of your subscription appears on di›erently and equally convincingly. in some detail how it was all done, from each month’s mailing label. For change He held that the demand for more food the loans made by Italian bankers to the of address notify us at least four weeks calls out more intensive labor and new master lodestones kept in Seville to pre- in advance. Please send your old ad- production methods, and hence, even- pare strong magnetic needles for the pi- dress (mailing label, if possible) and tually, more food, and then still more lots. We read chapter by chapter just your new address. people at the table. “This dilemma is how the ships of Lisbon and Seville were

116 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. built, then rigged and outfitted, manned, provisioned and, of course, armed. These self-assured and ruthless explor- ers were paramilitary; their superior weapons gave them victory, to the east against all marine powers from the Swa- hili ports to Japan, to the west against no real opposition at sea. The details are rich. But the freshest topic any landlubber will find here is how we come to know all this and how well we know it now. We have shipyard contracts and legal inventories, copious sealed state records of all the voyages and what they found, artist-embellished maps and texts that bear plenty of images of the outsides of ships. But we have little counterpart to blueprints from those wonderful craftsmen whose skills, tools, sketches and rules of thumb fashioned the complex hardware of planetary enterprise. After 1600, scholars began to describe all they could learn. Modern Iberian scholarship in particular has sifted the records, recalcu- lated ship performances, compiled technical glossaries and sought the sources and nature of wrought-iron cannon, hempen canvas and saltpeter, even to Columbus’s own experienced recommendation for “the third part of the bread- stu› to be good biscuit...and not old.” What is most wonderful is that within the past decades we have found even more intimate and reliable data not on paper but in this world of three dimensions. A meticulous ship’s model was given in votive o›ering long ago to a church near Barcelona. The four-foot carved model represents an actual medium-sized ship, probably a little earlier than those of Columbus. The model, now in Rotterdam, has been studied inside and out by medical optical instruments, videotaped and itself replicated. It solves several knotty puzzles of ship construction. A still better time machine is the old shipwrecks found on both sides of the Atlantic [see “Raising the Vasa,” by Lars- ˚Ake Kvarning, page 84]. A few vessels of about the period have been un- earthed along the Po and in England. One proud battleship of Henry VIII, a royal carrack, capsized and sank in 1545 while readying for battle with the French o› Portsmouth. It was resurrect- ed a decade ago. The coasts of the New World have turned up remains that are even closer to the issue. On the Yuca- tán coast, in southern Texas, then north to the Caicos reefs and the Great Baha- ma Bank, the divers have found and the scholars identified half a dozen mer- chantmen of the Spanish Main, all from the first half of the 16th century. On the sea bottom, they look like oval mounds

SCIENTIFIC AMERICAN October 1993 117 Copyright 1993 Scientific American, Inc. of boulders, for a pile of waterworn of weeks or months, and still it does. stones ballasted every ship of the day. But now we understand its RNA core Around, above and beneath those and its surface proteins, tiny spikes that stones lie the hardware, weapons, stores induce the human immune system to and whatever is left of the wooden ship form antibodies. The antibody record in itself. One Basque venture into whaling the blood of elderly patients demon- sank in the cold waters of Red Bay, Lab- strates that since 1890 all influenza epi- rador, after slipping its moorings late in demics have come from only a few of 1565: a rubber casting of the center of the possible subtypes of the virus par- the ship now preserves full-sized a fine ticles (we do not know why). example of “sixteenth-century ship- The virus genome can mutate, chang- wrightery.” An exciting summation ends ing the molecular code that forms the this engaging book. spike-tip proteins, so that the virus is But the handy fore-and-aft sail of the able to evade immune attack until new caravels, the sternpost rudder, the mag- custom antibody is formed. But the prin- netic compass, the cannon, gunpowder cipal changes we see year after year in and crossbows that armed these ships epidemic human flu are far less subtle; for war are all of them inventions of Chi- they result not from merely defensive nese origin, practical for centuries be- point mutations at the spike ends but fore they found the way west. The au- from the analogue of chromosome as- thor, perhaps too intent, has not recog- sortment among the eight gene seg- nized that fact by a single word, even ments of the viral core of RNA. A new as an aside. Not invention but develop- mix of genes can bring not simply a ment is here on center stage. changed infectivity but new virulence. With a modified protein the particles may no longer cause a self-limited ill- A Swift Viral Tra¤c ness but can instead turn more deadly by one or even two powers of 10. EMERGING VIRUSES, edited by Stephen We share influenza A with our do- S. Morse. Oxford University Press, 1993 mesticates—pigs, turkeys, chickens and ($39.95). horses. The infection can move both ways; all U.S. hog farms have sick pigs thousand centuries ago our sapi- each fall. Sometimes their virus shares ent forebears first quit the an- the spike type of the human strain-of- A cestral home, leaving a life pat- the-year, although swine virus usually tern wherein protohumans had long lacks the ability to spread among hu- equilibrated with local infections. They mans. Whence, then, the new flu almost multiplied greatly in the temperate every year? world—Eden until Cain, prolific farmer “The ultimate reservoir of influenza and stockman, would largely dispossess is in the aquatic birds of the world.” All his foraging sister and hunter brother. 13 or 14 subtypes of influenza are found The old disorders of our scattered kind in wild ducks, where they cause no dis- turned into a fitful tale of universal ease at all but replicate temperately, childhood infections in the more popu- mostly in the duck intestinal tract, to lated settlements. Those pathogens sur- spread and mix within hospitable wa- vive on numerous new human births, ters. “Every August you can go to Cana- even though their attacks confer lifetime da and...isolate your own kind of in- immunity on each survivor: witness fluenza virus.” Wanderers like gulls and measles and smallpox. In the past centu- turnstones maintain the flu virus in the ry another grand flourish of human spring; migrant ducks harbor it in the numbers appeared, diseases kept at bay fall. An epidemic of flu among New by a thin modernity, whether in medi- England harbor seals in 1980 clearly de- cine, in food or in well water. Today we rived from that avian reservoir of virus. restlessly transcend all geography at jet (The severe seal plague of 1988 in Eu- speed, while farmers still sow and reap rope was not the flu but a virus related many new fields at the feather edges of to canine distemper and to the terrible living forests. Of course, a swift viral cattle rinderpest.) tra›ic challenges what we once imag- Airborne transoceanic spread of flu vi- ined as a balanced and tolerant regime. rus did not wait for Boeing; avian virus It is influenza A that causes the teary, types suggest two ancient flight paths. achy fever a hundred million of us un- The path is north-south from Siberia in dergo every year worldwide. Its agent the Old World and from the Canadian is an old but steadily renascent virus, Arctic southward in the New; only a little surely the respiratory epidemic clearly mixing occurs across the Alaskan Straits. described in Europe since medieval The inference from long field studies is times. Out of the east it came again and now strong: the gene reassortment that again, to rage and to go away in a matter year after year generates new types of

118 SCIENTIFIC AMERICAN October 1993 Copyright 1993 Scientific American, Inc. human flu proceeds in those myriad ponds where the farmers of Asia grow paddy rice. Within those unfreezing waters, ducks, pigs and humans together shed and acquire virus particles. Passage through avian host cells usually selects for virus genes that do not spread well between humans. But the potential for a murderous pandemic is always present in those chancy, infectious and indispensable ponds. The influenza A pandemic of 1918–19, called in error “Spanish” flu, turned out to be both highly infectious and uniquely virulent. That edition of the flu took as many human lives as any single war has ever done. This volume presents 28 brief chapters in dazzling variety, only a handful of them in too narrow a jargon. Among the authors you find a historian, a mathematician, clinicians, virologists, evolu- tionists, a student of mosquitoes, epi- demiologists with strange tales from the Pampas or the West African bush, molecular biologists and a few very worried senior public health o¤cials. Troubling lists of a dozen newly emer- gent viruses are here. As an instance, dengue fever is making a reappearance in the Americas under new urban conditions. Used tires arrive in Texas from Asia by sea; some are piled up outdoors, soon to house tiny pools of rainwater, each perfect to awaken the stowaway eggs of a proved vector of dengue fever, the Asian tiger mosquito. A science- fiction episode is presented here, an exciting, serious extrapolation of what a virulent new tropical disease might mean to the U.S. Our country is slowly becoming unready; our tropical expertise and our labs for virulent infections are for the most part the veterans of old wars. Nor does the world have an ade- quate virus early-warning system, a clear international need that would double the scope of the World Health Organi- zation, an agency whose total budget is now about the size of that of a single big urban hospital. Viral RNA lacks proofreading and editing adaptations common in the longer, stabler DNA genomes. That ensures many replicative errors, in analo- gy to the earliest days of still unstable life. Viruses recklessly gamble in a risky game we more elaborated creatures cannot a›ord to enter. We still note fre- quent interaction between cell DNA and virus RNA. Such transfers mark our entire evolutionary history; we dare not now underestimate what viral change can yet do. Equilibrium is no longer seen as unchanging refuge; the peaks and troughs of life’s endless novelty demand vigilance and insight.

Tadpoles from Heaven may be dumb, but it is a uniquely human activity played explicitly on the surface of the planet that, good, bad or ike the cat that came back, the collaboration is whether gravity is essen- otherwise, defines our place in the uni- space station Freedom once again tial for the development of vertebrates. verse. When the astronauts who trav- Lgrips the attention of the U.S. Con- I consider this to be a serious topic wor- el into the heavens also do such human gress. Year after year Congress tries to thy of investigation, whether or not Con- things, they reinforce the humanity that snu› out the space station, and year af- gress ever approves a moon/Mars ini- links us to them and thus to the heav- ter year it squeaks through. What could tiative. As an earthbound biologist, who ens above. be more surreal than a space station that began his career as a naturalist, how- It is not totally clear what kind of sci- doesn’t exist yet won’t go away? ever, I have typically viewed NASA’s ac- entific information we will gain from a Why won’t it? Given the more press- tivities with cynicism. With so much visit to Mars, but I suspect that the trip ing problems on the home planet, why to learn about the universe, I could not will take place in my lifetime. It will is such an adventure contemplated for understand why NASA undertook what probably take a quarter of a century and even a nanosecond? The human explo- so often seemed to resemble pure pub- more money than anything that human- ration of space will be (excuse the pun) licity stunts. kind has attempted before, but it will astronomically expensive. We have all I remember, for example, learning, happen. It will happen not because it is heard that robots could do the job more back in 1969, that an astronaut had essential for our survival and certainly cheaply and more safely. Why, then, are just swung a golf club at a golf ball on not because it will answer any pressing we, the taxpayers, repetitively entertain- the moon. My first thought was, “What scientific questions. It will happen being multibillion-dollar initiatives to get a dumb use of taxpayers’ money. The cause it is a step toward the gods. In a few of our kind into orbit, o› to the guy’s wasting millions of dollars when sum, space exploration elevates our ex- moon and eventually to Mars? Why have he could be doing serious science. Fire istence, and NASA sanctifies our lives in we not declared these latter proposi- him! Let him find his own transporta- a way that no other government activi- tions positive lunacy (in the metaphori- tion home.” My involvement with NASA ty could ever hope to do. cal sense) and left them at that? and space biology has not much altered Because of its high cost, however, any The above are not just rhetorical ques- my skepticism about the agency’s commission to Mars will probably have to tions—they deserve answers. I feel some mitment to science. But it has forced me be an international endeavor. It could responsibility for finding those answers to ponder NASA’s relationship to the even be argued that to start planning since I happen to be involved with re- populace at large. this hajj to the heavens now could be a search ultimately related to human positive step toward global coopera- interplanetary exploration. Last year I hat the space agency has pro- tion. Dare I say that one small step for collaborated with scientists at the Na- vided us with is a primal link NASA could become, so to speak, one tional Aeronautics and Space Adminis- W to the gods. It isn’t science, but giant leap toward peace on earth and tration on a project to raise tadpoles it may still be worth millions. Think goodwill toward humankind? in zero gravity. The mission was a suc- about what humankind has invariably Despite all that, the fact is that I am cess. With the help of the astronauts asked of religion and what we now get not very religious. If I am correct in pre- on board the space shuttle Endeavor, from NASA: we speak of heaven “above,” suming that human space exploration my colleagues at NASA produced the with the earth below. Through reli- is motivated more by spirituality than first vertebrates conceived in space and gion, we have historically “transcended” by science, I suppose I should give up then brought to the earth as free-liv- our earthly burden. The gods were in the pretense that there is much value ing organisms. My part of the project the skies, and the names of the planets in doing “space biology.” Certainly that was to determine whether those little still remind us of that. Before we con- is the position held these days by most aliens, “our tadpoles from heaven” as I quered gravity with rockets, religion biologists, who vehemently oppose the took to calling them, were normal. Ba- was our only path of ascent above the space station. I take a di›erent stand. I sically, they were. mundane. think it is sheer hubris to believe that I Thus, we now know that it is possi- Today we treat our shuttle flight crew or any mere mortal could stop the pil- ble for a lower vertebrate to be born with the deference that was reserved grimage to other planets. There is noth- in space and, if brought to earth while for high priests in previous centuries. ing, however, in the religious quest that still young, to develop normally. It re- Astronauts are now the intermediaries forbids atheists and agnostics from go- mains to be determined whether mam- between ourselves and something still ing along solely for the science. Let us mals can produce normal o›spring in intangible, but unquestionably higher. try, then, to see that some real science, zero gravity. But the results to date Our astronauts, like the priests of old, some high-quality science, gets done with amphibians are promising. At least, will visit Mars. It is Mars the planet, not along the way. they do not yet preclude peoplekind the god, but the emotional appeal is or other kinds of organisms from resid- still the same. RICHARD WASSERSUG is a dual ing beyond the earth’s gravitation and We have to be careful not to deify American and Canadian citizen and is leading normal reproductive lives. our astronauts completely as they as- based at Dalhousie University in Hali- The core question behind my NASA cend to the heavens. Golf on the moon fax, Nova Scotia.