199307

JULY 1993
$3.95
Telling reasonable risks from fools chances.

Can sustainable development save the Amazon?
A tick in time: the most precise clocks ever.
Polar dinosaurs, which were adapted to the cold and dark ,

may have outlived their relatives from warmer climates.
Copyright 1993 Scientific American, Inc.
July 1993
32
Volume 269
Number 1
Risk Analysis and Management

M. Granger Morgan
We who live in industrial societies are justly concerned about risk. Hazards as
diverse as AIDS, asbestos in schools and contamination of food and water threaten life and health. As individuals, we daily juggle the chances we take traveling,
eating, smoking, drinking and encountering pathogens. Fortunately, powerful
analytic techniques exist that allow policymakers to assess risk.
42
Viral Quasispecies
Manfred Eigen
The extreme mutability and adaptability of viruses wreaks havoc with the classical notion of species. But where traditional taxonomy has failed, mathematics
may succeed. The author has developed a statistical classication scheme that
provides insights into the evolution of the inuenza virus and the age and origin
of HIV, suggesting new strategies for combating viral diseases, including AIDS.
50
Australias Polar Dinosaurs

Patricia Vickers-Rich and Thomas Hewitt Rich
Australias ability to produce varieties of animals that can be found nowhere else
began at least 100 million years ago, when the continent was one with Antarctica.
At that time, it hosted a population of warm-blooded dinosaurs that had evolved in
the cool climate and dark winters of the polar regions. Their adaptations may have
enabled them to survive longer than others in the frigid late Cretaceous period.
56
Accurate Measurement of Time

Wayne M. Itano and Norman F. Ramsey
A spring-driven watch will get you to the church on time. A cesium clock that loses no more than a second in one million years can time the millisecond ashes of
a pulsar. Yet ever more exacting scientic and navigational requirements demand
even more accurate timepieces. Incredibly, such devices can be madefrom isolated ions suspended in a vacuum and fountains of atoms cooled by lasers.
68
Surgical Treatment of Cardiac Arrhythmias

Alden H. Harken
A number of heart attack survivors later experience an electrical short circuit
that leads to an erratic, rapid heartbeat and sudden death. Because a courageous
banker allowed the author and his colleagues to attempt an untried operation in
1978, the lives of most susceptible patients can now be prolonged. The pathway
of the aberrant impulses is identied and surgically interrupted.
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
Fuzzy Logic
COOL
JUST
RIGHT WARM
Bart Kosko and Satoru Isaka
Too much precision can be a bad thing. An air conditioner that keeps a room at
68 degrees Fahrenheit may make some people uncomfortable. A coeemaker may
produce brew that gives some imbibers the jimjams. But fuzzy programs for camcorders and washing machines enable them to do the job the way you want it done.
84
Edwin Hubble and the Expanding Universe

Donald E. Osterbrock, Joel A. Gwinn and Ronald S. Brashear
The discovery that the universe is expanding did for the 20th century what the idea
of the heliocentric solar system did for the Renaissance. Although others contributed to the concept, the athletic Rhodes scholar from Missouri made the construction
of the universe uniquely his own topic and set the agenda of modern cosmology.
90
TRENDS IN ENVIRONMENTAL SCIENCE
Sustaining the Amazon

Marguerite Holloway, sta writer
The vast rain forest cradles a rich, complex community of plants and animals.
Some humans have lived as part of this web of life for thousands of years. But
others, driven by poverty or by entrepreneurial passion, threaten its existence.
Marguerite Holloway traveled widely with scientists who are seeking to reconcile
the need for economic development with preservation of the irreplaceable ecology.
DEPARTMENTS
20
Science and the Citizen
10
Letters to the Editors

Going around about black holes....
Finding aws before bridges fall.
Are the Rio initiatives losing momentum? . . . Magnifying astronomers resolution . . . . Supercomputer proof . . . .
Kamikaze satellite . . . . A kinder drug
policy . . . . Regenerating hearing . . . .
PROFILE: Howard Hughes Medical
Institute chairman Irving S. Shapiro.
12
50 and 100 Years Ago

1943: How long does it take
to cure vitamin B deciency?
110
Mathematical Recreations
A seamstress grapples with
the topology of garments.
101
Science and Business

114
Humanizing monoclonal antibodies . . . . Acoustic displays for computers . . . . Spin assay . . . . Flagging
British technology . . . . Fractals take
on image compression . . . . THE ANALYTICAL ECONOMIST: Will managed
competition improve health care?
Book Review
A historians shortsighted
vision for the 21st century.
120
Essay : Ralph Gomory

and Hirsh Cohen
How the bottom line can
guide the funding of science.
THE COVER painting shows Allosaurus

hunting by the southern lights in southeastern Australia more than 100 million years
ago, when the region fell within the Antarctic Circle. This specimen is one of the smallest allosaurs, and certainly the latest surviving, yet discovered. It may have owed its
longevity to adaptations to cold and darknessthe very factors thought to have driven the dinosaurs to extinction some 65 million years ago (see Australias Polar Dinosaurs, by Patricia Vickers-Rich and Thomas
Hewitt Rich, page 50).
Page
Source
Page
Source
3233
Nova Press/Sygma
63
Boris Starosta (left),

NIST (right)
3435
Jana Brenning (top),

Armistead Russell and
Gregory McRae (bottom)
64
Boris Starosta
65
Johnny Johnson
6870
Roberto Osti
7273
Laurie Grace (top),

Roberto Osti (bottom)
74
Laurie Grace
77
Michael Crawford
7881
Ian Worpole
85
J. R. Eyerman; LIFE
Magazine, Time Inc.;
courtesy of Henry E.
Huntington Library
4348
Jared Schneidman /JSD
49
Jean-Pierre Prvel/
Sygma
5051
Peter Trusler ; for a stamp

issue entitled Australias
Dinosaur Era by Australia
Post (October 1993)
52
Johnny Johnson ( left),

Frank Coa (right)
53
Patricia J. Wynne (top),

Steve Morton, Monash
University (bottom)
54
55
5657
Peter Trusler;
courtesy of Wildlife in
Gondwana, Reed Books
International (in press)
BOARD OF EDITORS: Alan Hall , Executive Editor ;

Michelle Press , Managing Editor ; John Rennie,
Russell Ruthen, Associate Editors; Timothy M.
Beardsley; W. Wayt Gibbs; Marguerite Holloway ;
John Horgan , Senior Writer ; Philip Morrison ,
Book Editor ; Corey S. Powell; Philip E . Ross; Ricki
L . Rusting; Gary Stix ; Paul Wallich; Philip M. Yam
86
Henry E. Huntington
Library (top), John R.
Hale (bottom right),
Henry E. Huntington
Library (bottom left)
PRODUCTION: Richard Sasso, Vice President, Production; William Sherman, Production Manager ;
Managers: Carol Albert, Print Production ; Tanya
DeSilva , Prepress; Carol Hansen , Composition ;
Madelyn Keyes, Systems; Leo J. Petruzzi , Manufacturing & Makeup; Carl Cherebin (Ad Trac)
CIRCULATION: Lorraine Leib Terlecki, Circulation
Director ; Joanne Guralnick , Circulation Promotion Manager ; Rosa Davis, Fulllment Manager ;
Katherine Robold , Newsstand Manager
ADVERTISING: Robert F. Gregory, Advertising
Director. OFFICES: NEW YORK: Meryle Lowenthal, New York Advertising Manager ; William
Buchanan, Manager, Corporate Advertising ; Peter Fisch, Randy James, Elizabeth Ryan. Michelle
Larsen, Director, New Business Development.
CHICAGO: 333 N. Michigan Avenue, Chicago, IL
60601; Patrick Bachler, Advertising Manager.
DETROIT: 3000 Town Center, Suite 1435, Southeld, MI 48075; Edward A . Bartley, Detroit Manager. WEST COAST: 1554 S. Sepulveda Blvd.,
Suite 212, Los Angeles, CA 90025; Kate Dobson,
Advertising Manager ; Tonia Wendt. Lisa K . Carden, Lianne Bloomer, San Francisco. CANADA:
Fenn Company, Inc. DALLAS: Grith Group
MARKETING SERVICES: Laura Salant, Marketing
Director ; Diane Schube, Promotion Manager ;
Mary Sadlier, Research Manager ; Ethel D. Little,
Advertising Coordinator
INTERNATIONAL: EUROPE: Roy Edwards, International Advertising Manager, London; Vivienne
Davidson, Linda Kaufman, Intermedia Ltd., Paris; Barth David Schwartz, Director, Special Projects, Amsterdam. SEOUL: Biscom, Inc. TOKYO:
Nikkei International Ltd.
8788
Patricia J. Wynne
Peter Menzel; courtesy

of National Geographic
Society (left ), Frank
Coa (right )
89
Mount Wilson and

Palomar Observatories
9091
Ricardo Azoury/ Black Star
Boris Starosta (top),

National Institute of
Standards and Technology
( NIST) ( bottom)
92
Johnny Johnson
93
Marguerite Holloway
PRESIDENT AND CHIEF EXECUTIVE OFFICER:

John J. Hanley
94
Mike Goldwater,
Network/Matrix
95
Marguerite Holloway
CORPORATE OFFICERS: Executive Vice President

and Chief Financial Ocer, R . Vincent Barger ; Vice Presidents: Jonathan Piel , John J.
Moeling, Jr.
9699
Ricardo Azoury/Black Star
CHAIRMEN OF THE BOARD:

Dr. Pierre Gerckens
John J. Hanley
110111
Michael Goodman
CHAIRMAN EMERITUS: Gerard Piel
58
Crown copyright, Science

Museum of London
59
Johnny Johnson
62
EDITOR: Jonathan Piel
COPY: Maria- Christina Keller, Copy Chief ; Nancy

L . Freireich; Molly K. Frances; Daniel C. Schleno
Cover painting by Dimitry Schidlovsky
Johnny Johnson
Established 1845
ART : Joan Starwood, Art Director ; Edward Bell ,

Art Director , Graphics Systems ; Jessie Nathans,
Associate Art Director ; Nisa Geller, Photography
Editor ; Johnny Johnson
THE ILLUSTRATIONS
3841
Boris Starosta
(top and bottom)
ADMINISTRATION: John J. Moeling, Jr., Publisher ;

Marie M . Beaumonte, Business Manager
SCIENTIFIC AMERICAN, INC.

415 Madison Avenue
New York, NY 10017
(212) 754-0550
PRINTED IN U.S. A.
SCIENTIFIC AMERICAN July 1993
LETTERS TO THE EDITORS

More Black Hole Paradoxes
I enjoyed Black Holes and the Centrifugal Force Paradox, by Marek Artur Abramowicz [SCIENTIFIC AMERICAN,
March], very much but was left somewhat puzzled. As we all know, centrifugal force is a ctitious force that appears to exist when a reference frame
is rotated. For example, when a car turns
a corner, unrestrained passengers tend
to continue to move along a straight tangential line, in keeping with Newtons
rst law. Could the explanation of the
paradox be made in terms of real centripetal forces?
STANLEY R. DRAKE
Shawnee, Kan.
The author writes that two astronauts
inside a tube surrounding a black hole
know that the axis of the tube is circular because Bob has measured the curvature of the walls . . . using straight rulers. Is this not impossible, since linearity is dened by the path of light? Would
not the ruler be unable to measure any
curvature because there is no curvature
along the axis of the tube?
RALF PHILIPP
Student, grade 9
Hackley School
Tarrytown, N.Y.
On page 79, the author states that in
any space-time, with or without a gravitational field, light always moves along
geodesics, and therefore it always traces
the geometry of space-time. In a space
warped by a gravitational eld, however, the light rays are curved and in
general do not coincide with geodesics
(emphasis added). Is it left to the reader to choose?
GASTON FISCHER
Observatoire Cantonal
^
Neuchatel, Switzerland
Every so often you publish an article
that reminds me of why I subscribe. The
subject matter of Abramowiczs article
is fascinating, but what is particularly
pleasing is that it is one of the bestwritten scientic articles Ive ever read.
Frankly, it reads like a Borges short story.
DAVID N. SCHWARTZ
London, England
10
Abramowicz replies:
Astronomers study rotating stars by
looking at the rotating reference frame.
They consider both the gravitational
and centrifugal forces acting on the stellar material because the introduction of
those ctitious forces makes the problem much easier. My discussion could
have been in terms of free-falling frames
and centripetal forces, but that would
have obscured the subject.
One can tell whether two identical rulers are straight without referring to the
path of light as the standard. The method is used by professional draftsmen:
they simply lay the rulers beside one
another. If the left and right sides of
each ruler match, they are straight. Of
course, the straight rulers will not appear as straight in a curved space!
Perhaps an analogy will explain why
light trajectories are geodesics in fourdimensional space-time but generally
not in three-dimensional space. Each
great circle on a globe is a geodesic line
on the two-dimensional surface, yet, being a circle, it is not a geodesic line in the
three-dimensional Euclidean space in
which the globe rests.
Inspecting Bridges
In Why Americas Bridges Are Crumbling [SCIENTIFIC AMERICAN, March],
Kenneth F. Dunker and Basile G. Rabbat state that The Silver Bridge disaster [at Point Pleasant, W.Va., in 1967]
happened in part because of poor inspection by local authorities. I am surprised to see that statement in Scientific American because there is not the
slightest factual basis for it.
I was closely associated with the investigation of the collapse, beginning in
January 1968 when I identied the fracture in eyebar 330 as the cause. As a
metallurgical study by the National Bureau of Standards showed, the eyebar
had fractured suddenly because of a
stress corrosion crack less than one
eighth of an inch deep that had started
on the surface of the hole in the eye. The
hole was almost completely lled by the
pin that coupled successive links in the
eyebar chain. The end of the pin and the
hole in the eye were also covered by a
plate that prevented visual inspection.
At the time of the collapse of the Point
Pleasant bridge, an identical bridge was
in service a few miles upstream. Natu-
rally, there was great interest in determining whether its eyebars could be inspected. The best brains in the nondestructive inspection eld concluded
unanimously that it could not be done.
Consequently, the bridge was removed.
JOHN A. BENNETT
Bethesda, Md.
Dunker and Rabbat reply:
We thank Bennett for his clarication.
Ironically, lax inspection noted at the
time of the Silver Bridge collapse helped
to trigger a massive federal bridge inspection program, and yet state-of-theart nondestructive testing would not
have detected the hidden defect.
X-cluded from Credit

Regarding Spot Marks the X, by
John Rennie [Science and the Citizen,
SCIENTIFIC AMERICAN, April], concerning the putative role of the Xist gene in
mammalian X chromosome inactivation,
I wish to make the following clarications. First, the human Xist gene was
discovered during my postdoctoral fellowship at the International Institute of
Genetics and Biophysics in Naples, Italy, and subsequently characterized in a
collaboration with Huntington F. Willards group and mine. Second, the murine Xist gene was discovered independently and reported simultaneously by
Sohaila Rastans group and mine.
ANDREA BALLABIO
Institute for Molecular Genetics
Baylor College of Medicine
ERRATA
The photograph on page 101 of

How Parasitic Wasps Find Their
Hosts [March] shows a potter wasp,
which carries prey to its young, and
not, as suggested, a parasitic wasp.
On page 130 of DNAs New
Twists [March], reference is made
to the linkage of methyl groups to
cysteine. The DNA base in question
is actually cytosine.
Because of the volume of mail, letters

to the editor cannot be acknowledged.
Letters selected for publication may be
edited for length and clarity.
50 AND 100 YEARS AGO

JULY 1943
Experiments during the last eight
years have led to the conclusion that
atoms of gasoxygen, hydrogen, or nitrogenactually dissolve in the crystalline structure of some metals just as
salt dissolves in water. These gas particles then loosen the electrons in this
structure, causing them to be emitted
from the metal more readily when heat
is applied. This explanation, Dr. Harvey C. Rentschler recently told a meeting of the American Physical Society,
should result in longer-lasting tubes
and accomplish important savings in
the size and number of electric batteries, generators, and other apparatus
needed to supply the lament power.
Having acquired a vitamin-B deciency, after several weeks, does it take
a person the same length of time to recover from it? Generalizations are dangerous but, by and large, the neurological and mental recovery is likely to be
rapid; recovery from tissue changes, if
any, probably much slower. A general
idea of the former is imparted by language used in Nutrition Reviews with
regard to deciencies in thiamin (vitamin B1): There is a vast amount of evidence, that journal states, that the administration of thiamin to an animal
acutely decient in thiamin causes a dramatic and prompt remission of the neurologic signs within minutes, and complete recovery within a few hours.
If, as appears to be probable, vegetation exists on Mars, life has developed
on two out of the three planets in our
system where it has any chance to do
so. With this as a guide, it appears now
to be probable that the whole number
of inhabited worlds within the Galaxy
is considerable. To think of thousands,
or even more, now appears far more
reasonable than to suppose that our
planet alone is the abode of life and
reason. What the forms of life might be
on these many worlds is a question before which even the most speculative
mind may quail. Imagination, in the absence of more knowledge of the nature
of life than we now possess, is unequal
to the task. There is no reason, however,
against supposing that, under favorable
conditions, organisms may have evolved
which equal or surpass man in reason
and knowledge of Natureand, let us
12
hope, in harmony among themselves!

Henry Norris Russell.
JULY 1893
A very interesting new mammal has
recently been received at the British Museum in the form of a sh-eating rat
from the mountain streams of Central
Peru. The animal is about the size of a
common house rat, but has a attened
head, strong and numerous whisker
bristles, and very small eyes and ears.
The chief interest of the new form centers in the fact of its being wholly a
sh-eater, and in its having in connection therewith its incisor teeth modied
for catching a slippery, active prey by

the development of their outer corners
into long sharp points, and its intestines
altered by the reduction almost to nil
of its ccum, an organ in vegetarian
Murid always of great size and capacity. There is no other rodent which, as
in the case of Ichthyomys stolzmanni,
as it is proposed to name the new form,
wholly lives on sh, to the exclusion of
a vegetable diet.
It may be of interest to amateur riemen to know the following simple
method for ascertaining the eect of
gravity on a bullet: Sight the rie upon
the target, keeping the sights plumb
above the center line of the bore of the
rie. Mark where the ball strikes. Then
reverse the rie, so as to have the sights
exactly beneath the line of bore. In this
reversed position sight it on the target
as before, and mark where the bullet
strikes. Divide the dierence in elevation of the two bullet marks by 32 and
extract the square root. This will give
the time in seconds that it took the ball
to travel the distance. The distance divided by this time will give the speed
of the bullet per second.J.A.G., Grand
Rapids, Mich.
The Tell-el-Amarna tablets, 320 in
number, were discovered by a fellah woman in 1887 among the ruins of the
palace of Amenophis IV, known as Khuen-Aten, about 180 miles south of Cairo. They have been found to contain a
political correspondence of the very
greatest interest, dating from some
3,370 years back. Many are from Palestine, written by princes of the Amorites, Phenicians, Philistines, etc., the burden of almost all being: Send, I pray
thee, chariots and men to keep the city
of the King, my Lord. Among the enemies against whom help is thus invoked are the Abiri, easily recognized
as the Hebrews. The date xes that of
the Bible (I. Kings vi. 1) as accurate.
The great Ferris wheel
The wonderful merry-go-round designed by Engineer George W. G. Ferris,

of Pittsburgh, Pa., is now completed at
the Columbian Exposition in Chicago.
This curious piece of mechanism forms
a most remarkable and attractive object (left). It carries thirty-six pendulum
cars, each seating forty passengers; thus
one revolution of the wheel carries 1,440
people to a height of 250 feet in the air.
SCIENCE AND THE CITIZEN

The Big Nada?
Inaction may stie
the UNCED agreements
come binding until they have been ratied : 50 nations must approve the climate treaty, 30 the biodiversity treaty.
As of May, only 17 countries had ratied each. And if the pacts take eect
but are not rigorously monitored or enforced, they will become paper tigers,
like the vast majority of international
environmental agreements.
Lack of enforcement could also weaken Agenda 21. Last fall the U.N. set up
a 53-member Commission on Sustainable Development to oversee eorts to
implement the plan. But the commission
has a small sta and no legal power. It
is expected to work much as the U.N.
Commission on Human Rights does, by
using publicity and international opinion to exert moral pressure. There is
no precedent for a group within the U.N.
having the kind of clout that the Commission on Sustainable Development
must have, notes Barbara J. Bramble,
director of international programs at
the National Wildlife Federation. On
the other hand, the U.N. is doing a lot
of unprecedented things.
ALLEN TANNENBAUM Sygma
he United Nations Conference on

Environment and Development
(UNCED) held in Rio de Janeiro
last June was full of sound and fury.
What exactly it signied remains to be
seen. The treaties that were signed have
not yet been implemented. Money that
was pledged has not been forthcoming.
And the group that was established to
enforce Agenda 21, a 40-chapter credo
for sustainable development, has not
cut its teeth.
Nevertheless, many observers and former participants say it is too early to
be disappointed. The Earth Summit did
not revolutionize anything. But it is a
process. We are in a very large river, and
its current does not change direction
easily, comments Susan R. Fletcher,
an expert on international environment
at the Congressional Research Service,
which provides data and analysis to legislators. The major problem is that we
are almost inured to rhetoric. We have
heard so much about doing these things
without actually doing them.
The UNCED conference, which was
attended by delegates and diplomats
from some 178 countries as well as by
thousands of nongovernmental organizations (NGOs), resulted in the creation
of a seemingly strong global political
will and the endorsement of several important policy documents. Along with
Agenda 21, they include the Rio Declaration (a list of environmental and development concerns that ensures national
sovereignty) and a statement about protecting forests.
In addition, two conventionsone to
prevent climatic change and one to conserve biodiversitywere signed by most
countries. You would still be negotiating these conventions today unless
you had the driving force of UNCED,
Fletcher observes. But following signatures with money and muscle is another
matter. The two conventions do not be-
NONGOVERNMENTAL ORGANIZATIONS, groups representing

special interests such as communities or businesses, attended
the 1992 UNCED conference in Rio de Janeiro in force. Their
20
pervasive presence at the meeting and ongoing inuence on

international environmental issues have been regarded as one
of the most positive aspects of the Earth Summit.
At this early stage, the aspect of the

edgling commission that appears to
please environmentalists and development experts the most is the inclusion
of NGOs. So far some 700 organizations
have asked the commission for accreditation, although NGO members such as
Bramble say fewer than 100 will probably be able to maintain a presence at the
U.N. A vote of one third of the members
can serve to exclude an NGOa diculty for groups from developing countries, where some governments have
tried to quell dissenting voices.
Despite potential muzzling, NGO activity is perceived as one of the Earth
Summits successful outcomes. It is
quite a victory that the rules for NGO
participation are modeled on the Rio
conferences rules, explains Hillary F.
French, senior researcher at the Worldwatch Institute. Whether the organizations can maintain this international contact and diplomatic clout depends, in
large part, on funding.
Of course, the entire enterprise of
achieving the goals of the Earth Summit
hinges jointly on national will and whether money can be made available to the
appropriate governments, agencies and
projects. Beyond the problem of getting
countries to contribute 0.7 percent of
their gross national product to the U.N.,
as Agenda 21 stipulates, lies the matter
of how those funds should be used.
Many developing nations do need more
moneyforeign debt is a real burden,
one observer remarks. But these countries do not have a priority for sustainable development. Money going in now
would go into business as usual.
Many environmentalists believe there
is plenty of money around and that it
is just being used incorrectly. If you
were doing a better job with what you
had, you would need a lot less, notes
Ian A. Bowles, legislative director of Conservation International. With regard to
protecting biodiversity, for example, he
argues, we dont need a radically bigger amount of money. We just need to
have it programmed right.
Funding is at the center of debates
about the future of the Global Environment Facility (GEF ). The GEF was established before the Earth Summit to
channel funds for projects in four areaspreventing climatic change as
well as protecting biodiversity, oceans
and the ozone layerthat could serve
as models for sustainable development. Under the joint administration
of the World Bank, the U.N. Environment Program and the U.N. Development Program, the funds pilot program is in the process of distributing
$1.3 billion by December.
At that time, the future of the GEF will
be reviewed. Many organizations argue

that its projects are too big and uncreative and that an alternative fund should
be instituted. In addition, these groups
contend that the GEFs association with
the World Bank ensures environmental
insensitivity. The bank has been sharply
criticized for the environmental damage caused by projects it has supported. We think the GEF should be independent, says Elizabeth Barratt-Brown,
a senior attorney at the Natural Resources Defense Council. There has been a
lot of greening in the language of the
World Bank. But what is really changing in the funding?
Other organizations, such as Conservation International, believe the GEF
should remain aliated with the bank
and serve as a lever to inuence change
there. It is an inherently long-term
proposition, but all these things are
continually advancing. The U.N. and
the World Bank have been around forever, so it is hard to reform suddenly,
Bowles points out.
One notable change since the Earth
Summit involves the U.S.s position on
the environment. During his tenure,
President George Bush refused to sign
the convention on biodiversity in Rio
and was unwilling to meet the goals of
the climate convention to lower carbon
emissions, which contribute to global
warming, to 1990 levels by the year
2000. During his Earth Day address
this past spring, however, President Bill
Clinton emphasized his willingness to
meet the targets for carbon emissions.
Clinton also announced his intention
to sign the convention on biodiversity,
with the addition of what is called an
interpretive statement. Because worries
about U.S. intellectual property rights
had deterred Bush from supporting the
treaty, the new administration sought
to confront the issue by clarifying aspects of the conventiona procedure
that is common to many treaties. The
interpretive statement allows the U.S.
to get a common statement that both
businesses and environmentalists could
live with, Bowles notes. Ratication
depends on getting the wording right
in the statement.
Such changes in national policy seem
to be rare. Some communities and countries, notably the Philippines, have tried
to establish local and national sustainability. But in general, we have seen a
return to business as usual around the
world, says Jacob Scherr, a senior sta
attorney at the Natural Resources Defense Council. These international treatises demand an enormous amount of
attention and energy and should not be
a diversion from needed eorts on the
ground.
Marguerite Holloway
Moonball
Astronomers beat a path
to high resolution
arold A. McAlister of Georgia

State University is an astronomer, but he frequently nds
himself talking baseball when he describes his work. If you built a stadium
on the moon, you couldnt even see it
from the earth through the best optical
telescopes, he begins. But McAlister is
championing a technique called optical
interferometry that would allow earthbound fans to watch the game. With
our array, you could see whos pitching.
By bringing together beams of starlight captured by two or more widely
separated telescopes, McAlister and his
colleagues can achieve the equivalent
resolving power of a single instrument
as large as the distance between the
telescopes. When the beams are combined, the light waves interfere with one
another. Where the peak of one light
wave meets the peak of another, they
reinforce each other; where the peak of
one wave meets the trough of another,
they cancel out.
An electronic detector records the resulting pattern of dark and light areas,
or interference fringes, which can then
be analyzed by computer to extract detailed information about the object being observed. If at least three telescopes
are used, the fringes can be rendered
into images hundreds of times crisper
than even those from the orbiting Hubble Space Telescopeat perhaps one
hundredth the cost.
Many of the most impressive results
reported so far have come from the
Mark III Optical Interferometer on Mount
Wilson in California, which has been operating since 1986. The Mark III consists
of two mobile light collectors that can
be placed as far as 31 meters apart.
The longer the distance between the individual telescopes, the greater the instruments resolving power. At full extension, the Mark III can pick out details
as small as two thousandths of an arc
second, about 100,000 times better than
the human eye can.
The Mark III can measure the outlines
of astronomical objects, but, alas, it cannot make true images. Nevertheless, it
has proved the importance of the concept. Last year Nicholas M. Elias and his
colleagues at the U.S. Naval Observatory made a stunning measurement of a
shell of gas blasting away from Nova
Cygni 1992, a brilliant thermonuclear
detonation that occurred on the surface of a collapsed white dwarf star.
Elias found that 10 days after the exSCIENTIFIC AMERICAN July 1993
21
plosion the shell stretched to 3.8 thousandths of an arc second (the full moon,
in comparison, is 1,800 arc seconds in
diameter). Observations of Nova Cygnis
spectrum revealed the velocity of the
eeing gas. Combining those data with
the Mark III measurement enabled Eliass group to determine that the nova is
about 9,500 light-years from the earth.
Other ndings from the Mark III have
illuminated the shape and structure of
stars. Stars are so distant in comparison
to their diameters that astronomers have
always considered them as unresolvable
point sources of light. But optical interferometers can resolve the disks of many
stars and reveal features on their surfaces. Michael Shao of the Jet Propulsion Laboratory in Pasadena, Calif., estimates that the Mark III has already resolved about 200 stellar disks.
One surprising result of observations
done on the Mark III and elsewhere is the
discovery that stars are not round, says
Richard S. Simon of the Naval Research
Laboratory. Many red giant stars, including the bright star Betelgeuse and the
well-known variable star Mira, exhibit peculiar egglike shapes, presumably because of the huge convection currents
roiling their lmy outer layers. A team

led by Simon has also reported detecting a huge cocoon of hydrogen gas surrounding the hot, highly active blue star
Gamma Cassiopeia. Related work has revealed clouds of titanium oxide billowing o red giants distended surfaces.
Future optical interferometers promise to push the technology and yield
even grander results. A group at the University of Sydney led by John Davis is
busily completing a 640-meter-long optical array that will be able to measure
stellar diameters as small as 50 millionths of an arc second (some 40 times
better than the Mark III and about 1,000
times smaller than the nest details
visible to the Hubble Space Telescope).
One of our key goals will be measuring the pulsations of Cepheids, Davis
relates. Cepheids are a class of pulsating stars whose regular variations in
brightness have been used by cosmologists to establish the distances to remote galaxies. Davis hopes to correlate
direct measurements of Cepheid pulsations with spectroscopic observations
of how fast their surfaces rise and fall.
In much the way that Elias inferred the
distance to Nova Cygni, Davis and his
colleagues will use their information to

derive distances to Cepheidsand thus
help calibrate the size of the universe.
The most exciting results will come
from interferometers that link multiple
telescopes and thereby deliver the longelusive goal of creating true images.
Craig Mackay of the University of Cambridge expects that the universitys 100meter-long interferometer, known as
COAST, will begin producing such images later this year. The instrument initially will target familiar objects such
as red giants and tightly bound binary
stars to make sure were not totally
o base, Mackay says cheerily. Then he
hopes to train COAST on T Tauri stars,
stellar newborns still enshrouded with
disks of gas and dust. Resolving details
around T Tauris will vastly increase astronomers understanding of the process by which stars and planets form.
The Naval Observatory and Naval Research Laboratory are constructing a
similar but larger imaging optical interferometer, the Big Optical Array, which
will begin operation sometime in the
fall. McAlisters planned Center for High
Angular Resolution Astronomy array
will incorporate seven large, 100-centi-
Banzai!
enerally, old satellites dont die; they just fade away.
Yet there are exceptions. This past spring the Japanese Institute of Space and Astronautical Science
( ISAS) decided to send its Hitin satellite into oblivion not
with a whimper but a bang. Rather than flinging the aging
spacecraft into the nether reaches of the galaxy, ISAS piloted it straight into the moon. On April 10, when the 315pound probe crashed at roughly 5,600 miles per hour, it
exploded in a bright flash, throwing up dust and digging
out a crater that astronomers hope will serve as a new
benchmark for planetary science.
Hakan
Svedhem, a physicist with the European Space
Agency, heard rumors of ISASs plans two weeks before
the execution date and scrambled to persuade astronomers to train their telescopes on the moon that night. It
was a great opportunity to observe from the ground a really giant impact as it happens. This has not been done before, Svedhem says.
Three observatories around the world signed on. But as
the kamikaze satellite plunged toward its fiery demise,
the telescope in Irkutsk was jammed up with technical difficulties, and another in Indonesia was rained out. The last
hope was David Allen, an astronomer at the Anglo-Australian Observatory who has a reputation for making difficult observations. If anybody could get this shot, David
could, says Alistair Glasse of the Royal Observatory in Edinburgh. But because of miscommunication about the time
of impact, Allen was unwittingly running behind schedule.
Just seconds before the collision, Allen got the cameras
rolling on the observatorys infrared imaging spectrometer
and recorded half a dozen frames as the flash lit up the lu-
22
nar night. The intensity of the burst and the apparent lack
of a sizable dust cloud make Glasse suspect that Hitin hit
solid rock, converting nearly all its kinetic energy to heat
and light. Svedhem points out, however, that because
ground zero lay about 10 kilometers within the Cimmerian side of the terminator between night and day, a large
dust cloud could easily have been cloaked in darkness.
The deceased was born Muses-A, a small craft designed
to help Japanese astronavigators hone their lunar swingby skills in preparation for a joint mission with the National Aeronautics and Space Administration. Christened Hitin
(a goddess of music) after its launch on January 24, 1990,
the satellite surpassed its one-year life expectancy and after a second year in high earth orbit was sent to wheel
round the moon. While it was en route, Svedhem used the
instrument to collect data on cosmic dust.
Hitins grand finale was not intended to benefit science.
ISAS officials gave only vague explanations for their decisionsomething about leaving fragments for their greatgrandchildren to find, Svedhem reports. But the satellite
may yet attain martyrdom by providing a rare controlled
experiment for planetary scientists. The correlation between the size and velocity of a meteorite and the size of the
crater it creates is based on theoretical calculations and has
never been verified by observations, Svedhem explains. In
this case, we had a very well defined mass and velocity. But
of course we cannot see the crater yet; it is quite small.
Svedhem hopes Hitins successor will pay a visit to the
grave site and send back images of the crater. Meanwhile
he and Glasse will glean all they can from their pictures of
the day the Muses died.
W. Wayt Gibbs
meter telescopes and will be sensitive

to infrared radiationa capability that
reduces atmospheric distortion and improves sensitivity to cool objects such
as dust-cloaked infant stars.
Perhaps the most audacious devices
on the drawing boards are the interferometry arrays proposed to be built late
in this decade around two of the largest
telescopes in the world: the pair of Keck
telescopes on Mauna Kea in Hawaii and
the European Southern Observatorys
Very Large Telescope in Chile. These devices will scrutinize the disks around
young stars, explore the tortured inner
regions of active galaxies and search

for planets orbiting other stars. Shao
estimates that even the extremely ambitious Keck array will have a price tag
of $40 milliona hefty sum, but only
about half the cost of each of the primary Keck telescopes and a tenth the cost
of a single ight of the space shuttle.
Such funds are not immediately forthcoming, however. A cloud of penury
hangs over the eld of optical interferometry. Part of the problem lies in
skepticism within the scientic community. Astronomers tend to be a conservative group. A lot of people consid-
er interferometry to be black magic,

McAlister sighs. Shao hopes the newest
set of devices, including his current
project, a test-bed for the Keck array,
will be able to convince lots of conventional astronomers that interferometry
is a tool that will be useful for them.
Of course, the astronomers are not
the only ones who need convincing, as
McAlister knows only too well. He anxiously awaits approval of the next dollop of funds from the National Science
Foundation. Its contingent on the federal budget, he says. That is always
risky business.
Corey S. Powell
QED for QCD
hadron masses to very good accuracy

to help us look at QCD, says W. Douglas Toussaint, a physicist at the University of Arizona. It would enable us
to compute properties that are useful
for extracting fundamental constants,
such as quark masses.
But the theorys mathematical complexity has made such predictions almost impossible. Perturbation theory,
the main tool of quantum eld physics,
proved inappropriate for a complete description of QCD. It applied only to a limited part of the model. Instead investigators turned to numerical methods on
computer, based on lattice gauge theory, a mathematical framework erected
20 years ago by Kenneth G. Wilson, now
at Ohio State University. The lattice refers to the representation of space as a
scaold, on which quarks rest on connecting sites. The bonds between lattice
points represent the gluons.
To secure answers representative of
the real world, workers must conduct the
calculations as the distance between lat-
tice points shrinks to zero and the number of lattice points increases to innity. In these limits, one should be able to
come up with observable quantities. Indeed, researchers have used lattice QCD
to explain quark connement, which accounts for why no one can see any free
quarks: it would take an innite amount
of energy to isolate a quark.
Coming up with the masses of hadrons has proved even more elusive. The
calculations require that you look at
all possible dierent congurations of
quarks, antiquarks and the chromoelectric eld on the lattice, says Donald H.
Weingarten, who headed the IBM team.
For meaningful results, large lattices are
necessary, and that entails more involved
calculationsmore than 100 million billion arithmetic operations.
Hence the need for a supercomputer.
Weingarten and his IBM colleagues Frank
Butler, Hong Chen, Jim Sexton and Alessandro Vaccarino turned to GF-11, a
massively parallel computer they helped
to develop for QCD calculations. The
A supercomputer backs
the theory of quarks
INTERNATIONAL BUSINESS MACHINES
ts a good thing machines dont get

overtime. Researchers at the IBM
Thomas J. Watson Research Center
have recently completed a set of calculations on a supercomputer that ran
continuously for approximately an entire year. More than an exercise in patience, the task may have provided the
strongest conrmation yet of a current
theory of elementary particles. In particular, the IBM team calculated the masses
of eight hadrons, a family of particles
that includes the proton and neutron,
and showed that the values obtained are
consistent with the masses measured in
the laboratory.
The theory of quantum chromodynamics, or QCD for short, was postulated in the 1970s to describe how the
fundamental particle known as the
quark builds the various hadrons. Two
up quarks and a down quark, for example, create a proton. A so-called chromoelectric eld (based on a property of
quarks called color) holds the quarks
together; the chromoelectric eld is carried by particles called gluons. The QCD
theory was highly successful in enunciating the properties of hadrons in certain kinds of experiments and became
part of the so-called Standard Model,
which unites all the forces of nature except for gravity.
Although experiments can supply data
for hadronsone can simply look up
the mass of the proton in reference
booksa correct theory should be able
to predict such information. Deriving observed values via analytic means would
greatly substantiate the model. Besides
giving physicists condence they have
the right ideas, such derivations suggest
that quantum parameters that cannot
be detected experimentally can be accurately inferred. We would like to get
QUARK CALCULATOR Donald H . Weingarten poses with the supercomputer he

and his IBM colleagues used to compute the masses of subnuclear particles, thereby providing conrmation that the theory of quantum chromodynamics is correct.
23
nomenclature refers to the maximum

speed of the calculations: 11 gigaops,
where a op is a oating decimal point
operation per second. To obtain the hadron masses, the researchers ran the GF11 for more than a year at a sustained
rate of between ve and seven gigaops.
That is roughly equivalent to running
more than 200,000 desktop computers
that use 386 processors.
The result after a yearlong wait? Statistically, the numbers for the hadron
masses are completely consistent with
experiment, Weingarten says. The disagreement for some values ranged from
less than 1 percent to no more than 6
percent. The uncertainties are for the
most part by-products of the statistical
algorithm used in the computer calculations and are not reections of QCD.
In other words, QCD seems to be the
right theory.
Despite their success, Weingartens
calculations rely on a simplication often made to render the mathematics
doable. Called the valence approximation method, it does not fully take into
account an eect that occurs in quantum systemsthe spontaneous creation
and destruction of particles. Quantum
uctuations can cause quark-antiquark
pairs to ash into existence and thereby inuence the system in some way.
Rather than incorporating the process,
the valence approximation assumes that
such virtual pairs act mainly to reduce
the strength of the existing color eld.
The approximation then compensates
for the decrease.
Not everyone thinks the approach is
completely valid. If you leave out a signicant part of the theory, you dont
know what the eect will be, maintains
Norman H. Christ, a physicist at Columbia University. Christ is deriving the values with the full theory, using Columbias supercomputer, the only other machine dedicated solely to QCD reckoning. But Toussaint estimates that calculating from the full theory would require
100 times the computer power of current machines.
The approximation does not undermine Weingartens condence in his results. The agreement between the valence approximation and experiment in a
sense tells you that the full theory must
agree with experiment, he says.
The physicists do concur on one point:
the calculations support the legitimacy
of employing computers in a proof. The
sort of thing we are doing represents
a real qualitative change, Weingarten
muses. What is new is that investigators
accept an experimental uncertainty in
studying and testing a theory in fundamental physics. Its experimental theoretical physics.
Philip Yam
24
A Kinder War
Harm reduction gains ground
as an approach to drug abuse
uring a presidential debate last

October 11, a reporter asked the
candidates whether they would
consider legalizing drugs in order to reduce drug-related violence and crime.
Bill Clinton rejected that option. Declaring that the life of his half brother, Roger, had been saved by his arrest for possession of cocaine (Clinton himself had
authorized the arrest), the president-tobe insisted law enforcement was crucial for combating drug abuse.
Clinton backed up his tough rhetoric
with his rst budget. It called for spending $13 billion in the next scal year on
controlling drugs, almost $1 billion more
than the Bush administration earmarked
for the current year. Clinton allocated
64 percent of the funds for antismuggling programs and law enforcement
(the balance is for education and treatment), only slightly less than Bush had.
Nevertheless, critics of the so-called
war on drugs are hopeful that the new
administration will be willing to try different tactics. Change is in the air, remarks Arnold S. Trebach, a professor of
criminal justice at American University
and president of the Drug Policy Foundation, a nonprot group in Washington, D.C., that espouses an approach to
drugs called harm reduction.
The idea behind harm reduction is
that drug abuse should be viewed as,
at worst, a disease requiring treatment
and not an absolute evil that must be
eradicated at all costs. The essence is
the acceptance of the enduring reality
of drug use, the absurdity of even attempting to create a drug-free society
and the need to treat drug users and
abusers as basically decent human beings, Trebach says.
Support for this viewpoint is growing
at the international level. In the past
three years, representatives of 15 European provinces and cities, including Zurich, Amsterdam and Rome, have signed
the so-called Frankfurt Resolution, which
calls for easing prohibitions on marijuana, free availability of clean needles and
treatment for all who seek it. Trebach
says his foundation, together with the
city of Baltimore, is co-sponsoring a
meeting on the resolution this November. Kurt L. Schmoke, the mayor of Baltimore, who has supported harm-reduction policies, has pledged to sign the resolution at the meeting.
The harm-reduction philosophy also
pervaded a conference on drugs held
in the nations capitol on May 7 and at-
tended by specialists in drugs, among

them law enforcement ocials, judges, physicians and social scientists. Although the audience disagreed on how
to reduce the harm caused by illegal
drugsthe proposals ranged from relatively modest calls for more treatment
to outright legalization of all drugsalmost all concurred that the war waged
by the Reagan and Bush administrations
had been an expensive failure.
Indeed, the annual federal budget for
drug war activities surged from less than
$2 billion in 1981 to more than $12 billion for this scal year. The Bush administration alone spent more than $40
billion to suppress illegal drug use over
four years. More than two thirds of the
funds went toward eorts to decrease
smuggling and to enforce laws.
Federal and state governments also
instituted more severe penalties for drug
violations, including mandatory sentences for those convicted of possession or
sale of drugs exceeding certain amounts.
Consequently, the number of arrests and
convictions for drug violations soared
to record levels. Drug oenders account
for roughly a third of the U.S. prison population, which reached an all-time high
of 883,593 at the end of 1992.
Defenders of strict policies claim their
implementation has reduced the number of people who dabble in illegal drugs,
marijuana in particular. Surveys done
by the National Institute on Drug Abuse
show that casual drug use has fallen
steadily since 1979. Critics contend the
decreases resulted less from law enforcement eorts than from a growing
public awareness of the adverse eects
of all drugs, legal or illegal. They note
that the use of tobacco and alcohol has
also decreased over the past decade.
Moreover, crime and other problems
related to drug abuse and tracking
continue unabated, particularly in poor
urban communities. Overdoses and
medical emergencies caused by cocaine,
heroin and other drugs dropped in 1990
but have risen again since then. The rate
at which intravenous drug users are infected with AIDS continues to grow [see
illustration on page 26 ].
Through his appointmentsif not
his rhetoricClinton has set the stage
for change. At the May conference in
Washington, Attorney General Janet Reno said her experience as state attorney
in Dade County, Florida, a major center
of drug tracking, led her to conclude
that antismuggling programs were having little impact on the cost or availability of drugs. She also complained that
mandatory sentences for nonviolent
drug oenders had decreased the prison
space available for incarcerating more
dangerous criminals.
Reno urged that nonviolent drug offenders be handled with a carrot and
stick approach, in which they can avoid
prison by submitting to a treatment program and staying o drugs; urine tests
would ensure compliance. Such a plan
has been carried out in Dade County during the past four yearswith great success, Reno said. This system has also
been favored by Lee P. Brown, former
commissioner of police in New York
City, whom Clinton named head of the
Oce of National Drug Control Policy.
Some prominent jurists have proposed more radical measures. One is
Whitman Knapp, a senior federal judge
in New York State (famed for having
led a commission that investigated police corruption in New York City two
decades ago). Earlier this year Knapp
announced he would refuse to consider
drug cases subject to mandatory sentencing laws. He subsequently argued
that Congress should repeal all federal
laws banning drug sales or possession
and permit states to devise alternatives
to prohibition.
Opponents of such wholesale decriminalization fear any benets would be
oset by a tremendous upswing in the
abuse of drugs such as cocaine and
heroin. David F. Musto, a historian at
Yale University, suggests in the 1987
book Dealing with Drugs that in 1900,
before opioids were prohibited in the
U.S., the rate of opioid addiction was at
a level never equaled or exceeded.
Trebach challenges this claim. He argues that estimates of the number of
addicts varied wildly at the turn of the
century, as do current estimates; the historical evidence can be used to buttress
any conclusion. The charge that prohi-
bition enforced through the criminal

law has succeeded in reducing the total
number of addicts or the rate of opiate
addiction in the United States cannot
be supported by the evidence at hand,
Trebach states in an upcoming book.
Some opponents of legalization highlight the benets of Prohibition, the period from 1920 to 1933 during which
alcohol was outlawed throughout the
nation. Ethan A. Nadelmann, a public
policy scholar at Princeton University,
acknowledges that consumption of alcohol did indeed fall during Prohibition,
as did public drunkenness and cirrhosis of the liver. Yet he notes that alcoholrelated problems had decreased even
more sharply during World War I as a
result of alcohol rationing and the temperance movement. Moreover, Britain
was more successful than the U.S. at
reducing alcohol consumption and related health problems in the 1920s and
1930s through taxes and restrictions
on hours of sale.
On the other hand, at least one recent experiment in decriminalization
was a spectacular failure. Five years ago
ocials in Zurich designated a park in
which drugs could be used without interference. Zurich recently ended the
experiment after the park became a haven for dealers, prostitutes and addicts
from throughout Europe.
Some experts, while ruling out wholesale decriminalization, have proposed
partial measures. Mark A. R. Kleiman
of the Kennedy School of Government
at Harvard University suggests a policy
that he calls grudging toleration. It
would allow the sale of certain psychoactive drugs through state-regulated
stores but would discourage consump-
tion through such measures as steep

taxes and limits on amounts purchased.
Kleiman thinks alcohol, tobacco and
marijuana might all be included under
this regime, but he would exclude drugs
he considers too harmful, notably cocaine and heroin.
Kleimans proposal aside, the drug reform movement has been more eective at criticizing current approaches
than at suggesting specic alternatives.
To redress that problem, Nadelmann
has helped form the Princeton Working
Group on the Future of Drug Use and
Alternatives to Drug Prohibition, consisting of experts from Princeton and
other institutions. A primary goal of the
group, Nadelmann says, is to devise a
drug regulatory model that eliminates
many of the worst consequences of
drug prohibition without reproducing
the unfortunate consequences of our
alcohol- and tobacco-control policies.
After all, Nadelmann remarks, alcohol
and tobacco remain by far the most
harmful of all drugs. John Horgan
Healing Hearing
Regrowing damaged ear cells
might eventually cure deafness
BILLIONS OF DOLLARS
Surge in Federal Funds for the War on Drugs. . .

14
12
10
8
6
4
2
0
1988
1989
1990
YEAR
1991
1992
Has Little Impact on Drug-Related Health Problems

12
110
11
100
10
90
80
I.V. DRUG USERS CONTRACTING AIDS (ANNUAL)
70
1988
1989
1990
1991
1992
SOURCES: Centers for Disease Control (red line), Drug Abuse Warning Network (blue line)
26
8
7
LAURIE GRACE
DRUG-RELATED EMERGENCIES (QUARTERLY)
THOUSANDS
THOUSANDS
120
f your taste for loud music has

forced you to swap your Walkman
for a hearing aid, there is now a
chance that you might someday be able
to switch back. The deafness caused by
loud noise has traditionally been permanent. But researchers have recently
found encouraging signs that humans
may have at least a latent ability to regenerate damaged parts of their inner
ear. Drugs that stimulate that regrowth
could conceivably restore hearing.
To me, its no longer a question of
if but of when we will get regeneration
in humans, predicts Jerey T. Corwin,
a hearing investigator at the University
of Virginia School of Medicine who has
contributed to the new ndings. Thomas R. Van De Water of Albert Einstein
College of Medicine in Bronx, N.Y.,
agrees: Its an exciting time. Ive been
working 25 years in this eld, and all of
a sudden its breaking open.
The focus of their work is the cochlea, a periwinkle-shaped organ of the inner ear. When sound waves strike the
eardrum, the vibrations pass into the
uid lling the cochlea and set in motion the tiny structures called hair cells
that stimulate the auditory nerve. Unfortunately, hair cells are delicate: 120
decibels of Metallica (or Mahler, for that
matter) jolts some hair cells so hard that
they shear o from their foundation.
CHRISTOPHER BURKE Quesada / Burke
Antibiotics, anticancer drugs and some

diseases, such as meningitis, can also
kill hair cells. If too many are lost, the
ear becomes irreversibly deaf.
Up to a point, hearing aids can sometimes compensate for the lost sensitivity by amplifying sounds. In recent years,
cochlear implants have become available; surgeons have inserted about 8,000
of them into deaf patients in the U.S.
These electronic devices allow microphones to stimulate the auditory nerve
directly. They cannot restore normal
hearing, however: at best, and after rehabilitative therapy, the implants permit patients to recognize sounds.
To some audiologists, a better solution would be to repair the injured cochlea. Yet in humans and other mammals,
years of observation indicated that hair
cells did not regenerate naturally. Other animals seemed more resilient. More
than a decade ago Corwin began nding evidence that sharks and amphibians grow new hair cells throughout their
lives. Birds, too, can recover from deafness. Normally, birds dont have turnover of their hair cells, notes Douglas
A. Cotanche of the Boston University
School of Medicine, a pioneer of the
avian work. But if you stress them with
noise or drugs, there is trauma, and new
cells regenerate. For whatever reason,
he says, there was this prejudice that
mammals are too evolved to do that.
New studies are changing that opinion. This past March in Science, Corwin
and Andrew Forge of University College,
London, and their colleagues claimed to
nd regenerating hair cells in the inner
ear of guinea pigs. They had treated the
animals repeatedly with toxic doses of
the antibiotic gentamycin. Within a week,
almost all the hair cells in one region
were gone. Yet a month after the treatments, immature hair cells were visible. In an accompanying paper, they
described what happened when they
repeated the experiment in culture using layers of tissue from the inner ears
of guinea pigs and of humans in their
fties. When antibiotics eliminated the
hair cells, adjacent supporting cells divided and took their place. The supporting cells then dierentiated as hair cells.
Skeptics could still argue that these
results were not pertinent to deafness:
Corwins group had worked with hair
cells from the vestibular system, which
confers the sense of balance, and not
from the cochlea. Van De Water and his
graduate student Hinrich Staecker and
a team from the University of Lige
cleared that hurdle just six weeks later.
Using cochlear tissue from rat pups
only a few days old, they found that the
hair cells did not spontaneously recover. But they also discovered that exten-
EAR CELLS CAN REGENERATE under some conditions, according to Thomas R. Van
De Water (left) and Hinrich Staecker ( right) of Albert Einstein College of Medicine.
sive regrowth could be encouraged in

less than a week if they exposed the tissue to retinoic acida compound related to vitamin A that guides the dierentiation of many cells during embryonic development. Retinoic acid derivatives are now used as wrinkle creams
and treatments for severe acne and have
shown some potential for ghting cancer in recent studies.
Van De Water argues that regeneration of hair cells seems to hinge on two
events: the destruction of the old cells
and the receipt of a chemical stimulus
to spur regrowth by new ones. Mature
mammals may not provide that second
signal eciently.
Although he is excited by the new results, Cotanche points out that no one
has ever really studied to see whether a
regenerated hair cell is functional. The
cells must not only grow back but must
also connect to the nerve. Moreover, as
Corwin explains, the hair cells that are
sensitive to dierent pitches are slightly dierent in their structure.
One of the amazing observations
that have been made, Corwin says, is
that the replacement cells all seem to
develop the appropriate structures and
orientations, apparently in response to
some cues embedded in the tissue environment. Also, all our indications are
that the nerve cells can nd the new
sensory cells, he adds. Both he and Van
De Water suspect the regenerating hair
cells may be releasing chemicals that
attract neuronal connections. Ive never
had much concern that they were eventually going to hook up, Corwin says.
They always get hooked up in sharks
and birds.
The workers caution that therapeutic

regeneration of hair cells in humans will
have to wait a while. Even if retinoic acid
does the trick, getting it into the ear
could be a problem. We dont want people to take large amounts of vitamin A,
because it can be dangerous, Van De
Water warns. Were trying to develop
unique ways of delivering the drug right
to the inner ear tissue using miniature
osmotic pumps. He says that his laboratory is looking into the possibility of
gene therapy, using viruses that could
insert the genes for growth controls into
cochlear cells. Cotanche wonders, too,
whether immature cells might be implanted into the cochleas of deaf adults
and induced to become hair cells. Another issue, Van De Water notes, is that
the neurons that become disconnected
from hair cells often die; he and his coworkers are trying to nd out how to
keep them alive until new hair cells have
emerged.
I expect well have some very good
compounds and protocols worked out
for stimulating regeneration within the
next two or three years, Corwin ventures. He, Van De Water and Cotanche
all agree that it will most likely be a decade before an actual therapy is ready
for testing in patients. Ever the optimist,
though, Corwin says, These discoveries over the past year or so have shaved
10 years o our projections.
Ten years is a long time for some deaf
people to wait. I would love to see us
put the cochlear-implant people out of
business, Van De Water remarks. Im
sure they would love it also. Right now,
though, the cochlear implant is the only
show in town.
John Rennie
27
PROFILE : IRVING S. SHAPIRO

professor at the Rockefeller University
and a former vice president of Merck
Sharp & Dohme. Its almost like a Quahen Irving S. Shapiro turned 65 tors in 1985 for $5.2 billion, thus creat- ker meeting. We have quite rigorous
in 1981, he had completed sev- ing the largest private philanthropy in discussions, but we come to a consenen successful years as chair- the U.S. with total assets of about $7 bil- sus. Id say its a very happy board.
man of E. I. du Pont de Nemours & lion. (Shapiro abstained from the nal
The most visible sign of Shapiros secCompany and earned an enviable repu- selection because of his links with the ond career was the dedication in May
tation as an advocate of ethical stan- biddershe sat on the board of one, of a new $55-million administrative
dards in commerce. He bought a house Boeing, and the other two, General Mo- headquarters and conference center for
in Florida and was ready to retire to a tors and Ford, were major customers Hughes scientists in Chevy Chase, Md.
quiet life of golf. It was not to be. Sha- of Du Pont.) But the complex sale was The institute also spent $281 million
piro, who was trained as a lawyer, says completed at almost exactly the mar- on medical research last year, an eighthe was favored with so many oers kets peak. We had very good fortune, fold increase over the past decade. In
from law rms that I began
the past few years it has into think I was selling myaugurated a series of initiaself short. He took a parttives for supporting science
nership at a major rm and
education. Its fellowship prothen pulled o the most
grams now extend to underchallenging coup of his cagraduates as well as gradureer: turning the troubled
ates, and it operates at liberal
Howard Hughes Medical Inarts and traditionally black
stitute into the richest recolleges as well as at schools
search charity in the world.
that are primarily scientic.
Founded by the reclusive
In addition, the institute has
billionaire and aviator in
awarded grants to museums,
1953 as a tax shelter, the inhealth policy studies and
stitute survived both Howseveral biomedical research
ard Hughes and myriad legal
groups. Expenditures on scichallenges. But in 1987, just
ence education and other
when the institute seemed
grants last year totaled $51.5
to have put its aairs in ormillion; in all, the institute
der, it was rocked by another
expects to spend $318 milscandal. The wife of its then
lion this year.
president, Donald S. FredFor a biomedical investigarickson, had incurred some
tor to receive an appoint$200,000 in decorating exment at one of the 53 Hughes
penses that wound up on
laboratories around the counthe Hughes books; Fredricktry is generally considered
son resigned. Then, in 1990,
the next best thing to a meal
George W. Thorn, chairman
ticket for life. The laboraof the board of trustees, antories, which are associated
nounced his retirement.
with major research hospiThe institute needed a
tals, are the institutes main
leader with a awless repfocus, and the 223 Hughes
utation and an uncommoninvestigators working at
ly steady hand. It turned
them include many of the
to Shapiro, a member of IRVING S. SHAPIRO has run two major science-based concerns, top names in biology.
its board since 1984. Once although he ducked out of every science class he took .
The Hughes approach
again, he was at the helm of
giving funds to exceptional
a multibillion-dollar organization found- Shapiro says with a professionally mod- scientists rather than to particular invesed on science, a eld that by his own est smile. Two years later Shapiro was tigationsmeans that most of the work
admission he knows little about. I instrumental in nally resolving a long- is fundamental in nature. Genetics and
ducked out of every science class I ever running battle with the Internal Rev- immunology, as well as cell biology and
took! he exclaims mirthfully.
enue Service.
structural biology, are the favored areas.
It was Shapiros corporate manageNow, after nearly three years of Sha- We get many letters from people interment acumen, not his technical skills, piros direct leadership, the Hughes ested in a particular disease, asking,
that was at the root of his most signi- Medical Institute may at last be nding Can you do something? It will only take
cant contributions to Hughes Medical. its feet. He leads the board with skill $3 million. I understand their motivaAs a board member, he engineered the and dedication and in ne style, says tions, but thats not our way of looking
sale of Hughes Aircraft to General Mo- trustee Alexander G. Bearn, an adjunct at the world, Shapiro states rmly.
Sciences Unscientic Champion
JOHN MCGRAIL
28
To judge from the number of important advances, the formula works.

Since 1990 Hughes investigators have
found the gene that is defective in neurobromatosis, elucidated the structure of the protein that the AIDS virus uses to enter cells and identied
genes associated with Lou Gehrigs disease and Huntingtons disease, to note
just a few examples. Recently Hughes
has extended its support to researchers who choose to stay at their own institutions, broadening its impact still
further.
The business expertise Shapiro now
applies to biomedical research was rst
exercised when he was helping out in
his familys clothes-pressing business
in Minnesota during the 1930s. For his
Lithuanian-born Jewish parents, it was
tough slogging in terms of money, he
recalls. Shapiro escaped to the University of Minnesota, where he graduated in 1941 with a bachelor of law degree. He spent the next 10 years in
Washington, D.C., in the criminal division of the Justice Department. Shapiro
cites his experience in government as a
source of his sense of corporate responsibility. Franklin D. Roosevelt was
God to the poor people in those days,
he explains.
When Du Pont hired him in 1951,
Shapiro says he felt obliged to point
out his lack of scientic background.
The companys general counsel replied,
as Shapiro remembers it, that he was
wanted not for his scientic knowledge
but for his expertise in the law of conspiracy. During his time at the Justice
Department, Shapiro had made an impression as a whiz in antitrust law. His
experience served Du Pont well. Shapiro distinguished himself by being willing to take calculated risks, but he believes he was also recognized as a fair
player. He climbed the ladder, becoming a vice president in 1970 and chairman and chief executive in 1974.
The appointment caused a sensation,
in part because he was the rst lawyer
in the position but more because there
were then few Jews in top-ranking jobs
in corporate America. Kingman Brewster [a former president of Yale University and ambassador to the Court of
St. James] told me he would not have
been surprised to hear there was a Jewish president of the U.S., but he was
amazed to hear of a Jewish head of a
major U.S. corporation, Shapiro recounts amiably. But Du Pont gave him
a warm reception, and he notes with
satisfaction that many prominent Jews
have since told him he opened their
door into the executive suite.
Shapiro understood from the start the
importance of putting resources into re-
search. To compensate for his personal

unfamiliarity with scientic matters, he
designated a member of Du Ponts executive committee to keep him informed,
and he soon earned a reputation for
decisive action. At the rst shareholders annual general meeting after he
took charge as chairman, he announced
that the company would be borrowing
to nance a $2.5-billion investment in
research and development. It takes a
Cooperation between
business and the
government has
slipped a little bit.
fair amount of intellectual discipline to
recognize that R&D is essential to prepare a company for ones successors,
he reects. Any CEO who is not thinking long-term is tempted to cut.
During the same period, he became
an outspoken champion of corporate
social responsibility as a founder of
the Business Roundtable in 1972. The
roundtable still exists as an association of business executives that, according to its literature, develops positions
which seek to reect sound economic
and social principles. Shapiro tried, he
says, to create the feeling that you have
to be a constructive citizen whatever
your private political beliefs.
He believes the eorts have had an
enduring inuence, especially easing
the almost open warfare in the 1970s
between business and an absolutist
Environmental Protection Agency. But
Shapiro sees troubling signs of deterioration in that entente cordiale. The Business Roundtable, he notes, is less closely identied with government-business
cooperation than it was: It has slipped
a little bit, he says.
Shapiros message secured him a position on a government advisory committee during the Carter years. There
was a price to payduring the Reagan
terms, he says, he was twice approached
about working for the executive branch,
once to consult on defense reforms and
once as an adviser on Middle East politics. But his ties with the Carter administration apparently proved too much
for the White House. Even though years
earlier he had taken his friend George
Schultz on a tour of Israel and Jordan
(which Shapiro says gave the future
secretary of state his education about
the region), the job oers mysteriously
evaporated. They blackballed me, he
states matter-of-factly.
His talent was nonetheless sought by
the legal profession. After retiring from
Du Pont, he promptly took a partnership in the Wilmington oce of Skadden, Arps, Slate, Meagher and Flom, a
powerhouse law rm with oces in 12
countries. Then, in 1984, he became one
of eight prominent citizens appointed
by a Delaware judge as trustees to restore the Howard Hughes Medical Institutes aairs to order. Shapiro threw
himself into rebuilding relations with
ocialdom. The rst thing we tried to
do was end the running sore with the
IRSthat was enough to turn your hair
white, he grimaces.
Shapiro still retains links with Du
Pont and displays his enthusiasm for
industrial research on a wall of his law
oce overlooking the Delaware River.
Behind his desk hangs a copy of a painting of the three Du Pont cousins, greatgrandchildren of the companys founder,
Eleuthre Irne du Pont de Nemours,
who bought the company in 1902 and
turned it into a world leader. These days
Shapiro says he is encouraged by the
Clinton administrations stated intention of implementing a vigorous technology policy. Industry has sometimes
not taken research seriously enough, he
muses. The government might make
mistakes, but I know we also made mistakes when I was at Du Pont.
Shapiro now spends four days a week
at Skadden, Arps and one day a week
on Hughes business. But he makes a
point of going to the institutes scientic briengs, even though he does not
follow many of the reports. I go to put
names and faces together, he explains.
Theres a great value in letting scientists know who I am. When a researcher wrote him recently to take exception
to a Hughes policy on intellectual property, I called her up and said, Lets get
together and talk, he says. You can
do a lot, assuming good faith.
Indeed, the institute has initiated
a grants program to fund investigators in countries such as Mexico, Canada, New Zealand, Australia and Britain.
Shapiro also sees a great opportunity
beckoning in Eastern Europe and the
former Soviet Union, where $10,000
will buy you a lot of science. Plans for
an initiative in those regions are well
advanced.
Impatiently acknowledging a few
of societys more crushing problems,
Shapiro nonetheless predicts that the
U.S. in the next century will be historys richest society in quality of life.
He nds his personal reward when
he reads a popular account of some
biomedical discovery and realizes one
of our people has really moved the
ball forward, and its because we supplied the money and picked the right
person.
Tim Beardsley
29
Risk Analysis
and Management
Inadequate approaches to handling risks
may result in bad policy. Fortunately,
rational techniques for assessment now exist
by M. Granger Morgan
mericans live longer and healthier lives today than at any time in
their history. Yet they seem preoccupied with risks to health, safety and
the environment. Many advocates, such
as industry representatives promoting
unpopular technology or Environmental
Protection Agency staers defending its
regulatory agenda, argue that the public
has a bad sense of perspective. Americans, they say, demand that enormous
eorts be directed at small but scarysounding risks while virtually ignoring
larger, more commonplace ones.
Other evidence, however, suggests that
citizens are eminently sensible about
risks they face. Recent decades have witnessed precipitous drops in the rate and
social acceptability of smoking, widespread shifts toward low-fat, high-ber
diets, dramatic improvements in automobile safety and the passage of mandatory seat belt lawsall steps that re-
M. GRANGER MORGAN has worked for

many years to improve techniques for
analyzing and managing risks to health,
safety and the environment. Morgan
heads the department of engineering and
public policy at Carnegie Mellon University. He also holds appointments in the department of electrical and computer engineering and at the H. John Heinz III
School of Public Policy and Management.
Morgan received a B.A. from Harvard
University, an M.S. from Cornell University and a Ph.D. in applied physics from
the University of California, San Diego.
32
duce the chance of untimely demise at

little cost.
My experience and that of my colleagues indicate that the public can be
very sensible about risk when companies, regulators and other institutions
give it the opportunity. Laypeople have
dierent, broader denitions of risk,
which in important respects can be more
rational than the narrow ones used by
experts. Furthermore, risk management
is, fundamentally, a question of values.
In a democratic society, there is no acceptable way to make these choices
without involving the citizens who will
be aected by them.
The public agenda is already crowded with unresolved issues of certain or
potential hazards such as AIDS, asbestos in schools and contaminants in food
and drinking water. Meanwhile scientic and social developments are bringing new problemsglobal warming, genetic engineering and othersto the
fore. To meet the challenge that these
issues pose, risk analysts and managers will have to change their agenda for
AIR DISASTER in Madrid claimed 183

lives in November 1983. The (small)
chance of dying in an air crash is one of
the prices that society agrees to pay for
rapid, convenient global transportation.
Some risks, including nuclear power generation, have caused fewer deaths but provoked greater calls for regulation, whereas
others, such as automobiles, cause more
deaths but arouse less concern.
evaluating dangers to the general welfare; they will also have to adopt new
communication styles and learn from
the populace rather than simply trying
to force information on it.
hile public trust in risk management has declined, ironically

the discipline of risk analysis
has matured. It is now possible to examine potential hazards in a rigorous,
quantitative fashion and thus to give
people and their representatives facts
on which to base essential personal and
political decisions.
Risk analysts start by dividing hazards into two parts: exposure and eect.
Exposure studies look at the ways in
which a person (or, say, an ecosystem
or a piece of art) might be subjected to
change; eects studies examine what
may happen once that exposure has
manifested itself. Investigating the risks
of lead for inner-city children, for example, might start with exposure studies to learn how old, aking house paint
releases lead into the environment and
how children build up the substance in

their bodies by inhaling dust or ingesting dirt. Eects studies might then attempt to determine the reduction in academic performance attributable to specic amounts of lead in the blood.
Exposure to a pollutant or other hazard may cause a complex chain of events
leading to one of a number of eects,
but analysts have found that the overall result can be modeled by a function
that assigns a single number to any given exposure level. A simple, linear relation, for instance, accurately describes
the average cancer risk incurred by
smokers: 10 cigarettes a day generally
increase the chance of contracting lung
cancer by a factor of 25; 20 cigarettes a
day increase it by a factor of 50. For
other risks, however, a simple dose-response function is not appropriate, and
more complex models must be used.
The study of exposure and eects is
fraught with uncertainty. Indeed, uncertainty is at the heart of the denition
of risk. In many cases, the risk may be
well understood in a statistical sense
but still be uncertain at the level of individual events. Insurance companies

cannot predict whether any single driver will be killed or injured in an accident,
even though they can estimate the annual number of crash-related deaths
and injuries in the U.S. with considerable precision.
For other risks, such as those involving new technologies or those in which
bad outcomes occur only rarely, uncertainty enters the calculations at a higher leveloverall probabilities as well as
individual events are unpredictable. If
good actuarial data are not available,
analysts must nd other methods to
estimate the likelihood of exposure and
subsequent eects. The development of
risk assessment during the past two decades has been in large part the story
of nding ways to determine the extent
of risks that have little precedent.
In one common technique, failure
mode and eect analysis, workers try to
identify all the events that might help
cause a system to break down. Then
they compile as complete a description
EFFECTS PROCESSES
EXPOSURE PROCESSES
RED SUNSETS
SULFUR COMPOUNDS
CHEAP
ELECTRICITY
ACID RAIN
RESPIRATORY
PROBLEMS
INCREASED
PLANT YIELDS
RISK MANAGEMENT PROCESS begins with analysis of the

people and other entities exposed to change, such as in this
illustration, from emissions from a coal-burning power plant
as possible of the routes by which those

events could lead to a failure (for instance, a chemical tank might release its
contents either because a weld cracks
and the tank ruptures or because an
electrical short causes the cooling system to stop, allowing the contents to
overheat and eventually explode). Although enumerating all possible routes
to failure may sound like a simple task,
it is dicult to exhaust all the alternatives. Usually a system must be described several times in dierent ways
before analysts are condent that they
have grasped its intricacies, and even
then it is often impossible to be sure
that all avenues have been identied.
Once the failure modes have been
enumerated, a fault tree can aid in estimating the likelihood of any given mode.
This tree graphically depicts how the
subsystems of an object depend on one
another and how the failure of one part
aects key operations. Once the fault
tree has been constructed, one need
(left ). After the results of exposure have been quantified (second panel ), they must then be ltered through public perceptions, which cause people to respond more strongly to some
only estimate the probability that individual elements will fail to nd the
chance that the entire system will cease
to function under a particular set of
circumstances. Norman C. Rasmussen
of the Massachusetts Institute of Technology was among the rst to use the
method on a large scale when he directed a study of nuclear reactor safety in
1975. Although specic details of his estimates were disputed, fault trees are
now used routinely in the nuclear industry and other elds.
Boeing applies fault-tree analysis to
the design of large aircraft. Company engineers have identied and remedied a
number of potential problems, such as
vulnerabilities caused by routing multiple control lines through the same area.
Alcoa workers recently used fault trees
to examine the safety of their large furnaces. On the basis of their ndings, the
company revised its safety standards to
mandate the use of programmable logic
controllers for safety-critical controls.
They also instituted rigorous testing of

automatic shut-o valves for leaks and
added alarms that warn operators to
close manual isolation valves during
shutdown periods. The company estimates that these changes have reduced
the likelihood of explosions by a factor
of 20. Major chemical companies such
as Du Pont, Monsanto and Union Carbide have also employed the technique
in designing processes for chemical
plants, in deciding where to build plants
and in evaluating the risks of transporting chemicals.
In addition to dealing with uncertainty about the likelihood of an event such
as the breakdown of a crucial piece of
equipment, risk analysts must cope with
other unknowns: if a chemical tank
leaks, one cannot determine beforehand
the exact amount of pollutant released,
the precise shape of the resulting doseresponse curves for people exposed, or
the values of the rate constants governing the chemical reactions that convert
the contents of the tank to more or less
dangerous forms. Such uncertainties are
often represented by means of probability distributions, which describe the odds
that a quantity will take on a specic
value within a range of possible levels.
When risk specialists must estimate
the likelihood that a part will fail or assign a range of uncertainty to an essential value in a model, they can sometimes use data collected from similar
systems elsewherealthough the design of a proposed chemical plant as a
whole may be new, the components in
its high-pressure steam systems will ba-
SUPERCOMPUTER MODEL of ozone concentrations in the Los Angeles basin

( pink , highest ; yellow, lowest) serves as
a starting point for analyses of the risks
of exposure to air pollutants.
34
PERCEPTION PROCESSES
VALUATION PROCESSES
ACID RAIN
RED SUNSETS
CHEAP
ELECTRICITY
INCREASED
PLANT
YIELDS
RESPIRATORY
PROBLEMS
aspects of risk than to others. Ultimately, costs and benets will be weighed. Agreeing on the values used to make decisions and making sure that all relevant eects
are taken into account are crucial , but often neglected, parts of the process.
sically be indistinguishable from those

in other plants.
In other cases, however, historical data
are not available. Sometimes workers
can build predictive models to estimate
probabilities based on what is known
about roughly similar systems, but often they must rely on expert subjective
judgment. Because of the way people
think about uncertainty, this approach
may involve serious biases. Even so,
quantitative risk analysis retains the advantage that judgments can be incorporated in a way that makes assumptions
and biases explicit.
Only a few years ago such detailed
study of risks required months of custom programming and days or weeks
of mainframe computer time. Today a
variety of powerful, general-purpose
tools are available to make calculations
involving uncertainty. These programs,
many of which run on personal computers, are revolutionizing the eld. They
enable accomplished analysts to complete projects that just a decade ago
were considered beyond the reach of
all but the most sophisticated organizations [see box on page 38 ]. Although
using such software requires training,
they could democratize risk assessment
and make rigorous determinations far
more widely available.
fter they have determined the likelihood that a system could expose
people to harm and described
the particulars of the damage that could
result from exposure, some risk analysts
believe their job is almost done. In fact,
they have just completed the preliminaries. Once a risk has been identied
and analyzed, psychological and social
processes of perception and valuation
come into play. How people view and
evaluate particular risks determines
which of the many changes that may
occur in the world they choose to notice and perhaps do something about.
Someone must then establish the rules
for weighing risks, for deciding if the
risk is to be controlled and, if so, how.
Risk management thus tends to force a
society to consider what it cares about
and who should bear the burden of living with or mitigating a problem once
it has been identied.
For many years, most economists and
technologists perceived risk simply in
terms of expected value. Working for
a few hours in a coal mine, eating peanut butter sandwiches every day for a
month, and living next to a nuclear power plant for ve years all involve an increased risk of death of about one in a
million, so analysts viewed them all as
equally risky. When people are asked
to rank various activities and technologies in terms of risk, however, they produce lists whose order does not correspond very closely to the number of expected deaths. As a result, some early
risk analysts decided that people were
confused and that their opinions should
be discounted.
Since then, social scientists have conducted extensive studies of public risk
perception and discovered that the situation is considerably more subtle. When
people are asked to order well-known
hazards in terms of the number of
deaths and injuries they cause every
year, on average they can do it pretty
well. If, however, they are asked to rank
those hazards in terms of risk, they produce quite a dierent order.
People do not dene risk solely as the
expected number of deaths or injuries
per unit time. Experimental psychologists Baruch Fischho of Carnegie Mellon
University and Paul Slovic and Sarah Lichtenstein of Decision Research in Eugene,
Ore., have shown that people also rank
risks based on how well the process in
question is understood, how equitably

the danger is distributed, how well individuals can control their exposure and
whether risk is assumed voluntarily.
Slovic and his colleagues have found
that these factors can be combined into
three major groups. The rst is basically an events degree of dreadfulness (as
determined by such features as the scale
of its eects and the degree to which it
aects innocent bystanders). The second is a measure of how well the risk is
understood, and the third is the number of people exposed. These groups of
characteristics can be used to dene a
risk space. Where a hazard falls within
this space says quite a lot about how
people are likely to respond to it. Risks
carrying a high level of dread, for example, provoke more calls for government intervention than do some more
workaday risks that actually cause more
deaths or injuries.
In making judgments about uncertainty, including ones about risk, experimental psychologists have found that
people unconsciously use a number of
heuristics. Usually these rules of thumb
work well, but under some circumstances they can lead to systematic bias or
other errors. As a result, people tend to
underestimate the frequency of very
common causes of deathstroke, cancer, accidentsby roughly a factor of
10. They also overestimate the frequency of very uncommon causes of death
(botulism poisoning, for example) by as
much as several orders of magnitude.
These mistakes apparently result from
the so-called heuristic of availability.
Daniel Kahneman of the University of
California at Berkeley, Amos N. Tversky
of Stanford University and others have
found that people often judge the likelihood of an event in terms of how easily they can recall (or imagine) examples. In this case, stroke is a very common cause of death, but most people
learn about it only when a close friend
or relative or famous person dies; in
contrast, virtually every time someone
dies of botulism, people are likely to
hear about it on the evening news. This
heuristic and others are not limited to
the general public. Even experts sometimes employ them in making judgments about uncertainty.
nce people have noticed a risk

and decided that they care
enough to do something about
it, just what should they do? How should
they decide the amount to be spent on
reducing the risk, and on whom should
they place the primary burdens? Risk
managers can intervene at many points:
they can work to prevent the process
producing the risk, to reduce exposures,
35
to modify eects, to alter perceptions or

valuations through education and public relations or to compensate for damage after the fact. Which strategy is best
depends in large part on the attributes
of the particular risk.
Even before determining how to intervene, risk managers must choose the
rules that will be used to judge whether to deal with a particular issue and, if
so, how much attention, eort and money to devote. Most rules fall into one of
three broad classes: utility based, rights
based and technology based. The rst
kind of rules attempt to maximize net
benets. Analysts add up the pros and
cons of a particular course of action and
take the dierence between the two. The
course with the best score wins.
Early benet-cost analyses employed
xed estimates of the value of good and

bad outcomes. Many workers now use
probabilistic estimates instead to reect
the inherent uncertainty of their descriptions. Although decisions are ultimately made in terms of expected values, other measures may be employed as well.
For example, if the principal concern is
to avoid disasters, analysts could adopt
a minimax criterion, which seeks to
minimize the harm done by the worst
possible outcome, even if that leads to
worse results on average.
Of course, many tricky points are involved in such calculations. Costs and
benets may not depend linearly on
the amount of pollutant emitted or on
the number of dollars spent for control. Furthermore, not all the pros and
cons of an issue can necessarily be mea-
sured on the same scale. When the absolute magnitude of net benets cannot be estimated, however, rules based
on relative criteria such as cost-eectiveness can still aid decision makers.
Rights-based rules replace the notion
of utility with one of justice. In most
utility-based systems, anything can be
subject to trade-os; in rights-based
ones, however, there are certain things
that one party cannot do to another
without its consent, regardless of costs
or benets. This is the approach that
Congress has taken (at least formally)
in the Clean Air Act of 1970: the law
does not call for maximizing net social
benet; instead it just requires controlling pollutant concentrations so as to
protect the most sensitive populations
exposed to them. The underlying pre-
Risk Analysis in Action
ncertainty is a central element of most problems in- the results are to the dollar values placed on life or health.)
volving risk. Analysts today have a number of softNet social costs, in this model, are simply the sum of
ware tools that incorporate the effects of uncertainty. These control costs and mortality. At $300,000 per death averttools can show the logical consequences of a particular set ed, their most likely value reaches a minimum when TXC
of risk assumptions and rules for making decisions about emissions are reduced by 55 percent. At $3 million, the
it. One such system is Demos, developed by Max Henrion optimum reduction is about 88 percent.
of Lumina Decision Systems
Demos can also calculate a
TXC
in Palo Alto, Calif.
form of correlation between
BASE CONCENTRATION
To see how the process
each of the input variables
works, consider a hypothetiand total costs. Strong correEMISSIONS
HEALTH DAMAGE COEFF
CONCENTRATION
THRESHOLD
cal chemical pollutant, TXC.
lations indicate variables that
REDUCTION
To simplify matters, assume
contribute significantly to the
CONTROL COST COEFF
HEALTH DAMAGE
that the entire population at
uncertainty in the final cost
POPULATION
risk (30 million people) is exestimate. At low levels of polposed to the same dose
lution control, possible variaEXCESS DEATHS
VALUE OF A LIFE
CONTROL COSTS
this makes a model of expotions in the slope of the damsure processes unnecessary.
age function, in the location
The next step is to construct
of the threshold and in the
TOTAL COST
EXPERT EVALUATION
a function that describes the
base concentration of the polrisk associated with any givlutant contribute the most to
UNCERTAINTY
en exposure levelfor examtotal uncertainty. At very high
ANALYSIS
ple, a linear dose-response
levels of control, in contrast,
function, possibly with a BLOCKS in the diagram above can be expanded to call
almost all the uncertainty dethreshold below which there up a window containing graphs and tables for their asrives from unknowns in the
is no danger.
cost of controlling emissions.
sumptions, equations and probability distributions.
Given this information, DeFinally, Demos can commos can estimate the number of excess deaths caused pute the difference in expected cost between the optimal
every year by TXC exposure. According to the resulting cu- decision based on current information and that given permulative probability distribution, there is about a 30 perfect informationthat is, the benefit of removing all uncercent chance that no one dies, about a 50 percent chance tainties from the calculations. This is known in decision
that fewer than 100 people die each year and about a 10 analysis as the expected value of perfect information; it is
percent chance that more than 1,000 die.
an upper bound on the value of research. If averting a sinMeanwhile, for a price, pollution controls can reduce gle death is worth $300,000 to society, this value is $38
the concentration of TXC. (The cost of achieving any giv- million a year; if averting a death is worth $3 million, it is
en reduction, like the danger of exposure, is determined $71 million a year.
by consultation with experts.) To choose a level of polluAlthough tools such as Demos put quantitative risk analtion control that minimizes total social costs, one must ysis within reach of any group with a personal computer,
first decide how much society is willing to invest to pre- using them properly requires substantial education. My colvent mortality. The upper and lower bounds in this exam- leagues and I found that a group of first-year engineering
ple are $300,000 and $3 million per death averted. (Pick- doctoral students first exposed to Demos tended to ignore
ing such numbers is a value judgment; in practice, a cru- possible correlations among variables, thus seriously overcial part of the analysis would be to find out how sensitive estimating the uncertainty of their results.
38
sumption holds that these individuals

have a right to protection from harm.
Technology-based criteria, in contrast
to the rst two types, are not concerned
with costs, benets or rights but rather
with the level of technology available to
control certain risks. Regulations based
on these criteria typically mandate the
best available technology or emissions
that are as low as reasonably achievable. Such rules can be dicult to apply because people seldom agree on the
denitions of available or reasonably
achievable. Furthermore, technological
advances may impose an unintended
moving target on both regulators and
industry.
There is no correct choice among the
various criteria for making decisions
about risks. They depend on the ethical
and value preferences of individuals
and society at large. It is, however, critically important that decision frameworks be carefully and explicitly chosen and that these choices be kept logically consistent, especially in complex
situations. To do otherwise may produce inconsistent approaches to the
same risk. The EPA has slipped into
this error by writing dierent rules to
govern exposure to sources of radioactivity that pose essentially similar risks.
mplicit in the process of risk analysis and management is the crucial

role of communication. If public
bodies are to make good decisions about
regulating potential hazards, citizens
must be well informed. The alternative
of entrusting policy to panels of experts
working behind closed doors has proved
a failure, both because the resulting policy may ignore important social considerations and because it may prove impossible to implement in the face of
grass-roots resistance.
Until the mid-1980s, there was little
research on communicating risks to the
public. Over the past ve years, along
with my colleagues Fischho and Lester B. Lave, I have found that much of
the conventional wisdom in this area
does not hold up. The chemical industry, for example, distilled years of literature about communication into advice
for plant managers on ways to make
public comparisons between different
kinds of risks. We subjected the advice
to empirical evaluation and found that it
is wrong. We have concluded that the
only way to communicate risks reliably
is to start by learning what people already know and what they need to
know, then develop messages, test them
and rene them until surveys demonstrate that the messages have conveyed
the intended information.
In 1989 we looked at the eects of
40
the EPAs general brochure about radon

in homes. The EPA prepared this brochure according to traditional methods:
ask scientic experts what they think
people should be told and then package the result in an attractive form. In
fact, people are rarely completely ignorant about a risk, and so they lter any
message through their existing knowledge. A message that does not take this
ltering process into account can be ignored or misinterpreted.
To study peoples mental models, we
began with a set of open-ended interviews, rst asking, Tell me about radon. Our questions grew more specic
only in the later stages of the interview.
The number of new ideas encountered
in such interviews approached an asymptotic limit after a couple of dozen people. At this point, we devised a
closed-form questionnaire from the results of the interviews and administered
it to a much larger sample.
We uncovered critical misunderstandings in beliefs that could undermine the
eectiveness of the EPAs messages. For
example, a sizable proportion of the
public believes that radon contamination is permanent and does not go away.
This misconception presumably results
from an inappropriate inference based
on knowledge about chemical contaminants or long-lived radioisotopes. The
rst version of the EPAs Citizens Guide
to Radon did not discuss this issue.
Based in part on our ndings, the latest
version addresses it explicitly.
The objective of risk communication
is to provide people with a basis for
making an informed decision; any effective message must contain information that helps them in that task. With
former doctoral students Ann Bostrom,
now at the Georgia Institute of Technology, and Cynthia J. Atman, now at the
University of Pittsburgh, we used our
method to develop two brochures about
radon and compared their eectiveness
with that of the EPAs rst version. When
we asked people to recall simple facts,
they did equally well with all three brochures. But when faced with tasks that
required inferenceadvising a neighbor with a high radon reading on what
to dopeople who received our literature dramatically outperformed those
who received the EPA material.
We have found similar misperceptions
in other areas, say, climatic change. Only
a relatively small proportion of people
associate energy use and carbon dioxide
emissions with global warming. Many
believe the hole in the ozone layer is the
factor most likely to lead to global warming, although in fact the two issues are
only loosely connected. Some also think
launches of spacecraft are the major con-
tributor to holes in the ozone layer. (Willett Kempton of the University of Delaware has found very similar perceptions.)
he essence of good risk communication is very simple: learn what

people already believe, tailor the
communication to this knowledge and
to the decisions people face and then
subject the resulting message to careful empirical evaluation. Yet almost no
one communicates risks to the public
in this fashion. People get their information in fragmentary bits through a
press that often does not understand
technical details and often chooses to
emphasize the sensational. Those trying
to convey information are generally either advocates promoting a particular
agenda or regulators who sometimes fail
either to do their homework or to take
a suciently broad perspective on the
risks they manage. The surprise is not
that opinion on hazards may undergo
wide swings or may sometimes force
silly or inecient outcomes. It is that
the public does as well as it does.
Indeed, when people are given balanced information and enough time to
reect on it, they can do a remarkably
good job of deciding what problems
are important and of systematically addressing decisions about risks. I conducted studies with Gordon Hester (then
a doctoral student, now at the Electric
Power Research Institute) in which we
asked opinion leadersa teacher, a state
highway patrolman, a bank manager
and so onto play the role of a citizens
board advising the governor of Pennsylvania on the siting of high-voltage electric transmission lines. We asked the
groups to focus particularly on the controversial problem of health risks from
electric and magnetic elds emanating
from transmission lines. We gave them
detailed background information and a
list of specic questions. Working mostly on their own, over a period of about
a day and a half (with pay), the groups
structured policy problems and prepared advice in a fashion that would be
a credit to many consulting rms.
If anyone should be faulted for the
poor quality of responses to risk, it is
probably not the public but rather risk
managers in government and industry.
First, regulators have generally adopted a short-term perspective focused on
taking action quickly rather than investing in the research needed to improve
understanding of particular hazards in
the future. This focus is especially evident in regulations that have been formulated to ensure the safety of the environment, workplace and consumer
products.
Second, these ocials have often
NOT OBSERVABLE
UNKNOWN TO THOSE EXPOSED,
EFFECT DELAYED, NEW RISK,
RISKS UNKNOWN TO SCIENCE
DNA TECHNOLOGY
MICROWAVE OVENS
ELECTRIC FIELDS
WATER FLUORIDATION
SACCHARIN
WATER CHLORINATION
ORAL CONTRACEPTIVES
VALIUM
CONTROLLABLE
NOT DREAD, NOT GLOBAL
CATASTROPHIC, CONSEQUENCES
NOT FATAL, EQUITABLE, LOW
RISK TO FUTURE GENERATIONS,
EASILY REDUCED, RISK
DECREASING, VOLUNTARY
NITRITES
POLYVINYL
CHLORIDE
DIAGNOSTIC
X-RAYS
IUDS
ANTIBIOTICS
LEAD (AUTOS)
LEAD PAINT
ASPIRIN
DES
NITROGEN FERTILIZERS
RADIOACTIVE WASTE
NUCLEAR REACTOR ACCIDENTS
PESTICIDES
URANIUM MINING
ASBESTOS
PCBS
NUCLEAR WEAPONS
INSULATION
FALLOUT
SATELLITE CRASHES
MERCURY
COAL-BURNING POLLUTION
VACCINES
SKATEBOARDS
POWER
MOWERS
SMOKING (DISEASE)
SNOWMOBILES
TRAMPOLINES
CARBON MONOXIDE
(AUTOS)
BLACK LUNG
STORAGE
AND TRANSPORT
OF LIQUEFIED
NATURAL GAS
LARGE DAMS
SKYSCRAPER FIRES
TRACTORS
NUCLEAR WEAPONS (WAR)
CHAIN SAWS
HOME SWIMMING
POOLS
UNCONTROLLABLE
DREAD, GLOBAL CATASTROPHIC,
CONSEQUENCES FATAL, NOT
EQUITABLE, HIGH RISK TO
NERVE GAS FUTURE GENERATIONS, NOT
ACCIDENTS EASILY REDUCED, RISK
INCREASING, INVOLUNTARY
UNDERWATER CONSTRUCTION
COAL-MINING ACCIDENTS
SPORT PARACHUTES
ELEVATORS
DOWNHILL SKIING
GENERAL AVIATION
HIGH CONSTRUCTION
RAILROAD COLLISIONS
COMMERCIAL AVIATION
RECREATIONAL BOATING
MOTORCYCLES
BICYCLES
ALCOHOL-RELATED ACCIDENTS
FIREWORKS
AUTO RACING
AUTO ACCIDENTS
HANDGUNS
DYNAMITE
OBSERVABLE
KNOWN TO THOSE EXPOSED,
EFFECT IMMEDIATE, OLD RISK,
RISKS KNOWN TO SCIENCE
RISK SPACE has axes that correspond roughly to a hazards

dreadfulness and to the degree to which it is understood.
adopted too narrow an outlook on the

risks they manage. Sometimes attempts
to reduce one risk (burns from ammable childrens pajamas) have created
others (the increased chance of cancer
from reproong chemicals).
In some instances, regulators have ignored large risks while attacking smaller ones with vigor. Biologist Bruce Ames
of Berkeley has argued persuasively
that government risk managers have
invested enormous resources in controlling selected articial carcinogens
while ignoring natural ones that may
contribute far more to the total risk for
human cancer.
Third, government risk managers do
not generally set up institutions for
learning from experience. Too often adversarial procedures mix attempts to
gure out what has happened in an incident with the assignment of blame. As
a result, valuable safety-related insights
may either be missed or sealed away
from the public eye. Civilian aviation, in
contrast, has beneted extensively from
accident investigations by the National
Transportation Safety Board. The board
does its work in isolation from arguments about liability; its results are
widely published and have contributed
measurably to improving air safety.
Risks in the upper right quadrant of this space are most likely to provoke calls for government regulation.
Many regulators are probably also too

quick to look for single global solutions
to risk problems. Experimenting with
multiple solutions to see which ones
work best is a strategy that deserves
far more attention than it has received.
With 50 states in a federal system, the
U.S. has a natural opportunity to run
such experiments.
Finally, risk managers have not
been suciently inventive in developing arrangements that permit citizens
to become involved in decision making
in a signicant and constructive way,
working with experts and with adequate time and access to information.
Although there are provisions for public hearings in the licensing process for
nuclear reactors or the siting of hazardous waste repositories, the process
rarely allows for reasoned discussion,
and input usually comes too late to
have any eect on the set of alternatives under consideration.
Thomas Jeerson was right: the best
strategy for assuring the general welfare in a democracy is a well-informed
electorate. If the U.S. and other nations
want better, more reasoned social decisions about risk, they need to take steps
to enhance public understanding. They
must also provide institutions whereby
citizens and their representatives can

devote attention to risk management
decisions. This will not preclude the occasional absurd outcome, but neither
does any other way of making decisions. Moreover, appropriate public involvement should go a long way toward
eliminating the confrontational tone
that has become so common in the risk
management process.
FURTHER READING
RATIONAL CHOICE IN AN UNCERTAIN
WORLD. Robyn M. Dawes. Harcourt
Brace Jovanovich, 1988.
READINGS IN RISK. Edited by Theodore S.
Glickman and Michael Gough. Resources for the Future, 1990.
UNCERTAINTY: A GUIDE TO DEALING
WITH UNCERTAINTY IN QUANTITATIVE
RISK AND POLICY ANALYSIS. M. Granger
Morgan and Max Henrion. Cambridge
University Press, 1990.
COMMUNICATING RISK TO THE PUBLIC.
M. Granger Morgan, Baruch Fischho,
Ann Bostrom, Lester Lave and Cynthia
J. Atman in Environmental Science and
Technology, Vol. 26, No. 11, pages 2048
2056; November 1992.
RISK ANALYSIS. Publication of the Society for Risk Analysis, published quarterly by Plenum Publishing.
41
Viral Quasispecies
The standard definition of a biological species does not apply
to viruses. A more expansive and dynamic view of viral
populations holds clues to understanding and defeating them
by Manfred Eigen
ccording to Greek mythology,

when curious Pandora opened a
forbidden box she set loose all
the miseries and evils known to the
world. One of them was undoubtedly
the virusthe very name of which is
Latin for slime, poison and stench. Viruses cause a mind-boggling assortment
of illnesses, ranging from the common
cold to acquired immunodeciency syndrome (AIDS), perhaps the most feared
scourge of modern times.
Viruses have the ability to mystify laypeople and experts alike. Early in their
studies of viruses, investigators became
puzzled by the high mutation rates they
observed: the magnitudes indicated that
viruses must evolve more than a million times faster than cellular microorganisms. If that were true, how could viruses maintain their identities as pathogenic species over any evolutionarily
signicant period? Why didnt they mutate out of existence?
Those questions have generally been
unanswerable within the traditional theoretical framework of biology. Borrowing ideas from both mathematics and
chemistry, however, my colleagues and
I have recently introduced a concept, the
quasispecies, that can illuminate the
problems in new ways. A viral species,
we have shown, is actually a complex,
self-perpetuating population of diverse,
related entities that act as a whole.
The substitution of quasispecies for
MANFRED EIGEN is director of biochemical kinetics research at the Max

Planck Institute for Biophysical Chemistry in Gttingen, where he began his
undergraduate studies in 1951. For his
ground-breaking work in developing
techniques for measuring high-speed
chemical reactions, Eigen was named
as a co-recipient of the 1967 Nobel Prize
for Chemistry. In more recent years the
major focus of his research has been
the signicance of the information concept to molecular evolution and its technological applications.
42
species is not merely semantic. It offers insights into the behavior of viruses.
In the case of AIDS, for example, it helps
in determining when the human immunodeciency virus ( HIV ) rst evolved
and where it may have come from. If
one were to extrapolate only from the
epidemiologic data, AIDS would seem to
have rst appeared in 1979. Our data, in
contrast, suggest that HIV is a very old
virus. Moreover, the quasispecies concept points toward potential treatments
for AIDS and other diseases that have
so far been resistant to vaccines.
o begin to understand viral quasispecies, we must ask ourselves,

What is a virus? In 1959 Nobel
laureate Andr Lwos answer was A
virus is a virus!a truism, perhaps, but
one that cuts to the uniqueness of viruses in the living world. Essentially, a
virus is a genetic program that carries
the simple message Reproduce me!
from one cell to another. Because a virus represents only one or a few of the
messengers vying for the attention of
its host, it must employ certain biochemical tricks to recruit the hosts replication machinery for its selsh purpose. Often those ploys result in the
host cells death.
Viruses fall into many dierent categories, but one way to distinguish
among them is by looking at the molecules that carry their genetic messages.
Perhaps the simplest form of virus is
represented by a single strand of ribonucleic acid (RNA), made up of several thousand individual nucleotide subunits. If this RNA is a so-called plus
strand, it can be read directly by the
hosts translation apparatus, the ribosome, much as the hosts own messenger RNA can. Examples of such plus
strand viruses are the bacteriophage
Q , a parasite of the bacterium Escherichia coli, and the polio-1 virus, which
causes spinomuscular paralysis. Other
viruses encode their messages as minus strands of RNA. Inside a cell, minus
strands must be transcribed into com-
plementary plus strands before viral replication can begin. Inuenza A, one of
the most common epidemic diseases,
is caused by a minus strand virus.
A third class of single-strand RNA
viruses consists of retroviruses. After a
retrovirus infects a host cell, a viral enzyme called reverse transcriptase changes the single strand of viral RNA into a
double strand of deoxyribonucleic acid
(DNA). That DNA can then incorporate
itself into the hosts genome, thereby
making the viral message an inheritable feature of the cell. HIV belongs to
the retroviral family. Its target is the immune system, which ought to provide
protection against the virus.
Because viruses are so dependent on
the replicative systems of their hosts,
scientists generally believe viruses in
their present form must have evolved
after cellular life. It is even possible that
viruses descended from parts of their
hosts genetic programs that turned
their inside knowledge of cells to the
goal of duplicating themselves. Whatever the case, viruses are useful models
for studying how molecules may have
organized themselves into self-perpetuating units at the dawn of life. They
show how information can be generated and processed at the molecular level. The essence of their genetic information is self-preservation, which they
achieve through mutagenesis, reproduction, proliferation and adaptation to a
steadily changing environment.
The genome of a single-strand RNA
virus such as HIV, which comprises only
10,000 nucleotides, is small and simple
compared with that of most cells. Yet
from a molecular standpoint, it is unimaginably complex. Each of those nucleotides contains one of four possible
bases: adenine, uracil, guanine or cytosine. The unique sequence specied by
the genome of HIV therefore represents
just one choice out of 410,000 possibilitiesa number roughly equivalent to a
one followed by 6,000 zeros.
Most such sequences would not qualify as viruses: they could not direct
RETROVIRUS
PLUS STRAND
RNA VIRUSES
LEVIVIRUS
(PATHOGEN OF
BACTERIA)
HUMAN IMMUNODEFICIENCY VIRUS
(CAUSES AIDS)
TOBACCO MOSAIC VIRUS
(PATHOGEN OF PLANTS)
PICORNAVIRUS (CAUSES
POLIO AND OTHER
DISEASES IN ANIMALS)
DOUBLE-STRAND
RNA VIRUS
MINUS STRAND
RNA VIRUSES
REOVIRUS
(PATHOGEN OF PLANTS
AND ANIMALS)
ORTHOMYXOVIRUS
(CAUSES INFLUENZA
AND OTHER DISEASES
IN ANIMALS)
HOST CELL
RHABDOVIRUS
(CAUSES RABIES,
VESICULAR STOMATITIS
AND OTHER DISEASES
IN ANIMALS)
DOUBLE-STRAND
DNA VIRUSES
SINGLE-STRAND
DNA VIRUS
INOVIRUS
(PATHOGEN
OF BACTERIA)
ADENOVIRUS
(CAUSES TUMORS
AND OTHER DISEASES
IN ANIMALS)
MYOVIRUS
(PATHOGEN
OF BACTERIA)
VIRUSES BELONG to many diverse families, which

may be distinguished by the type and activities of
their genetic molecules. In some viruses the genes
are in single or double strands of DNA; in others the
genes are RNA molecules. Some RNA viruses carry
plus strands that can be translated directly by the
host cells protein-making machinery. For minus

strand viruses, the RNA must rst be transcribed
into complementary plus strands. Retroviruses, such
as those that cause AIDS, require that their RNA be
reverse-transcribed into double strands of DNA. Only
a few of the many varieties of viruses are shown.
43
How to Construct a Sequence Space
ne way to study the diverse nucleotide sequences

in the genes of viruses is to map them into a multidimensional matrix called a Hamming sequence space. In
this space, each point represents a unique sequence, and
the degree of separation between points reflects their degree of dissimilarity. The space can be most easily drawn
for short sequences consisting of binary digits. For a sequence with just one position, there are only two possible
sequences, and they can be drawn as the end points of a
line (a). For a sequence with two positions, there are four
00
10
11
01
permutations, which form the corners of a square (b ). The

variations on a three-digit sequence become the corners
of a cube (c), and the variations on a four-digit sequence
are the vertices of a four-dimensional hypercube (d ). Each
higher-dimensional space is built iteratively by drawing
the previous diagram twice and connecting the corresponding points. The sequence spaces for viral genomes are far
more complex than these simple figures because they involve thousands of positions that can each be occupied
by one of four different nucleotides.
100
000
110
010
1000
1100
0100
0000
101
1110
111
0001
001
011
0010
44
living species must show common

traits and must be at least potentially
able to produce ospring by recombining their genetic material. At the genetic level, a biological species is represented by a gigantic variety of diering
DNA molecules.
Biologists generally speak of the wild
type of a species: the form that predominates in a population and that is particularly well suited to the environment
in which it lives. If one found an individual that perfectly embodied that wild
type, its unique sequence of genomic
DNA would specify the wild type at the
genetic level and would occupy a single
point in the sequence space. That view
of the wild type accords with the classical model of natural selection. Although
mutations occur steadily, they presumably disappear because the mutant types
are less t than the wild type. Alternatively, a mutant may have advantages,
in which case it becomes the new wild
type. Either outcome tends to keep all
the members of a species at or very near
one point in a genome sequence space.
hat picture was modied by the

neutral theory advanced in the
1960s by Motoo Kimura of the
National Institute of Genetics in Mishima, Japan. Kimura argued that many
mutations, such as those causing differences in blood types, are neither advantageous nor disadvantageous. Consequently, a small but statistically dened fraction of the neutral mutations
0101
0110
1111
1011
0011
their own duplication. Nevertheless,

even if only a tiny fraction of them are
viruses, the number is still huge. If the
entire universe were completely lled
with hydrogen atomseach about one
trillionth of a trillionth of a cubic centimeter in volumeit could hold only
about 10 108 of them. Hence, an array of
10 6,000 diering RNA sequences is beyond comprehension.
Fortunately, it is not beyond the analytic reach of mathematics. We can construct a theoretical framework that encompasses that vast array and reveals
relations among the elements. To do
so, we must rst develop a geometry
a concept of spacethat would allow
us to represent the informational differences among the sequences as precise spatial distances. In this space, each
nucleotide sequence must occupy a
unique position. The positions must also
be arranged to reect the informational kinship between the sequences. In
other words, each sequence should be
only one unit away from all the other
sequences that dier from it by only
one nucleotide; it should be two units
away from those diering by two nucleotides, and so on.
Sequence space proves to be an invaluable tool for interpreting what a viral species is. The term species is used
in both biology and chemistry. In chemistry, a species is a dened chemical
compound, such as trinitrotoluene or
benzene. In biology, the denition is
not quite as sharp: members of a given
1101
1001
1010
0111
would continuously replace the existing wild type in the population. The genome of a species would therefore drift
steadily but randomly through a certain
volume of sequence space.
Despite those dierences, both the
classical Darwinian and the neutralist
theories favor the idea that wild-type
populations will localize sharply in sequence space after completing an advantageous or neutral shift. Also, both
theories assume that mutations appear
blindly, irrespective of their selective value. No single neutral or advantageous
mutation would occur more frequently
than any disadvantageous one.
That view, however, is not sustained
by the modern kinetic theory of molecular evolution, nor is it backed by experiments with viruses. After all, evolutionary selection is a consequence of
the ability of a genome to replicate itself accurately. Imagine a case in which
the process of replication is so highly
error-prone that no copy resembled its
parental sequence. The resulting population would behave like an ideal gas,
expanding until it lled the sequence
space at a very low density. Selection
acting on such a population could not
dene it or conne it in any way. The
population would lose all its integrity.
If we were to reduce the error rate of
replication progressively, variation in the
population would disperse less and less
as the ospring came to resemble their
parents more and more. At some critical error rate, the eect of selection on
the population would change radically:

the expansive force of mutation would
strike a balance with the compressive
force of selection. The diuse gas of related sequences would suddenly condense into a nite but extended region.
This region in sequence space can be
visualized as a cloud with a center of
gravity at the sequence from which all
the mutations arose. It is a self-sustaining population of sequences that reproduce themselves imperfectly but well
enough to retain a collective identity
over time. Like a real cloud, it need not
be symmetric, and its protrusions can
reach far from the center because some
mutations are more likely than others
or may have higher survival values that
allow them to produce more ospring.
That cloud is a quasispecies.
Biologically, the quasispecies is the
true target of selection. All the members
of a quasispeciesnot just the consensus sequencehelp to perpetuate the
stable population. The tness of the
entire population is what matters, not
the tness of individual members. The
wild type of a quasispecies refers to
an average for all the members, not to
a particularly t individual. Chemically, the quasispecies is a multitude of
distinct but related nucleic acid polymers. Its wild type is the consensus sequence that represents an average for
all the mutants, weighted to reect their
individual frequency. Physically, the
quasispecies is a localized distribution
in sequence space that forms and dissolves cooperatively in very much the
same way that molecules of water pass
through phase transitions as they freeze
or evaporate. Its stability is constrained
by the error threshold, which may be
interpreted as a kind of melting point

for the genome information. The population density at each point of sequence
space depends on the tness value of
that particular sequence. A mathematician would describe the distribution of
sequences in a quasispecies with a vector that refers to the maximum growth
within the set of coupled kinetic equations for all the mutants.
One might wonder why in this model an advantageous or neutral mutant
would have a better chance to occur
than a deleterious one. New mutants
appear at the periphery of the quasispecies distribution, where they are produced by the erroneous copying of mutants already present. Because the population of a mutant in the quasispecies
depends on its degree of tness, welladapted mutants have a better chance
of producing ospring; deleterious mutants produce no ospring at all. Because the chance of nding a well-adapted or advantageous mutant is greatest
in a region of sequence space associated with high tness, there is a large
bias toward producing such well-adapted mutants. Calculations show that this
eect speeds up the evolutionary opportunization of viruses by many orders
of magnitude, as compared with truly
random, unbiased mutations.
ecause the error rate directly determines the size and integrity of
a quasispecies, it is the most telling characteristic of a virus. The error
rate is the probability that an error will
occur when one nucleotide in a sequence
is being copied. It can depend both
on the type of nucleotide substitution
taking place and on its position in the
PERFECT REPLICATION
OF WILD TYPE
HIGHLY IMPERFECT
REPLICATION
POPULATION DYNAMICS of a virus depend on the error rate

of its replication process. These gures are highly simplied
representations of the sequence spaces that might contain a
viral population. If the replication process of a virus were perfectly accurate, all the viral ospring would occupy the same
position in sequence space (a). If replication were highly im-
sequence. The position is important because the ribosome interprets the nucleotides three at a time, in a group
called a codon. In most codons the rst
two positions sufce to specify the amino acid to be incorporated into a protein. Mutations in the rst two positions
may therefore be more stringently maintained by selection. When researchers
speak of the error rate of an entire viral
sequence, they are referring to an average for all the positions.
In general, the error rate of a virus is
roughly proportional to the reciprocal
of its sequence lengththat is, about
one error per replicated sequence. If the
error rate were much larger, almost every replication event would produce an
unt mutation. For an entity that produces as many ospring as a virus, an
error rate reciprocal to the sequence
length is highly signicant. Consider a
typical infection process, which starts
when at least one viable virus enters a
host organism. If that virus is not eradicated, it will replicate. Before an infection is detectable, the viral population
must rise to around 109, which would
take about 30 generations. If the error
rate is more or less equal to the reciprocal of the sequence length, then on
average one error will have been added
in each generation.
Consequently, any two viruses taken
from an obviously infected host are likely to dier from each other at 30 nucleotide positions or more. When researchers rst noticed the sequence diversity of the HIV viruses they found in
individual patients, they thought it was
evidence of multiple infections by different strains. The work of Simon Wain
Hobson of the Pasteur Institute in Par-
QUASISPECIES
perfect, mutant viruses would soon occupy every position in

sequence space (b), and the viral population would lose its
integrity. At some intermediate error rate, however, the viral
population would become a coherent, self-sustaining entity
that resembles a cloud centered on the original consensus sequence (c ). That cloud is a quasispecies.
45
RAND RNA
ND RNA
RA
S ST
ST
OF
PLU
MIN
US
3'
REL
EA
SE
5'
3'
A
RN
ND
RA
ST
US
ND S
BI KE
SE A
ICA ND M
PL
RE A A
RN
IN
S
T
PL O M
US IN
ST US S
RA TR
ND AND
RN
A
RE
LE
E
AS
OF
5'
O
F
CO
AT
5'
RELEASE
OF
5'
5'
PR
OT
EIN
PL
U
OF
RE
LE
AS
E
3'
5'
OF SYN
CO THE
SI
AT
PRO S
TEIN
FO
RM
SUBU
NITS
CO
MB
INE
TO
RN
A
5'
3'
A
RN
ND
RA NA
ST D R
US RAN
PL ST
S
REP
LIC
AND ASE
MA BIN
KE DS
S
M TO
IN
U
A
RN
ND
A
R
ST
EIN
OT NA
PR D R
AT RAN
CO ST
S
U
RE
+ 3
E
AS
IC
L
P
5'
T
NI
IS
ES BU
TH SU
SYN ASE
IC
EPL
OF R
INDS TO REPLICASE INFO

TEIN B
RMA
PRO
TIO
T
A
N IN
CO
RIB
INF OSO
OR M
MA E B
TI IN
ON D
S
IN T
PL O
3'
RIB
OSO
ION
MAT
ME B
INDS TO R LICASE INFOR
EP
SE
3'
5'
A
LE
RE
3'
RIBOSOME
REPLICASE
REPLICASE SUBUNIT
HYPERCYCLES govern the replication of viruses inside host

cells. A hypercycle consists of interlocked feedback loops. In
the replication of the plus-strand bacteriophage virus Q , for
example, the reproduction cycle (tan ) for the genetic information is promoted by a second cycle (blue) involving the
production of a viral replicase enzyme. At the same time, vi-
is has demonstrated, however, that the

diverse HIV sequences in patients are
usually related to one another. His work
clearly conrms that viruses, and immunodeciency viruses in particular, are
quasispecies.
The proliferation of a viral quasispecies is a more complex phenomenon
than the simple replication of a wild
type. Viral replication takes the form of
a hypercycle, a set of interlocking feed46
COAT PROTEIN
PLUS STRAND RNA
MINUS STRAND RNA
ral replication is inhibited by the production cycle (green) of

the viral coat protein, which prevents the synthesis of replicase subunits. The combined inuence of these cycles determines the proportions in which viral components are made
and thereby the rate of viral replication. Because errors can
accumulate in the hypercycle, viruses are prone to mutation.
back loops that describes a regulated coevolution within a cell of the viral genes
and the viral proteins essential to replication that are encoded by those genes.
Michael Gebinoga of the Max Planck Institute for Biophysical Chemistry in Gttingen has quantied the process in vivo
for the Q bacteriophage. He found evidence of two feedback cycles, one based
on the enzyme replicase, which promotes replication, and the other based
on the viral coat protein, which limits

it. The rst molecules of replicase and
other proteins produced by the infectious plus strand are fairly accurate because most copies of the viral genes
in the cell are similar to the originals.
Errors accumulate mostly during later stages in the infection cycle. For that
reason, the synthesis of replicase seems
to occur primarily early after infection.
Yet even viral sequences that make de-
fective proteins are copied because the

replicative machinery acts on all the
strands indiscriminately. When an infected E. coli cell bursts after 40 minutes, it releases around 10,000 phage
particles, of which only 1,000 or less
are infectious.
Analyses of sequence space can reveal
information about the evolution of viral
quasispecies that would otherwise be inaccessible. A straightforward procedure
for studying the evolution would be to
follow the changes in a viral gene over
time. A researcher would need to collect samples of a virus over a period of
many successive years. The diculty is
that even for quickly mutating viruses,
the amount of change that can accumulate in only a few yearssay, the lifetime of a Ph.D. thesisis too small to
measure meaningfully. Hence, the experiment would never be done.
In the mid-1980s Peter Palese of
Mount Sinai School of Medicine found a
better way. He was lucky enough to obtain samples of inuenza A virus that
had been isolated and frozen during
outbreaks of the disease over a span of
about 50 years. Palese and his co-workers analyzed the gene sequence common to those samples. From that information, they plotted the evolutionary
relations among the viruses from each
epidemic. The family tree they created shows the worldwide spread of the
virus from a common source in successive waves during each epidemic. The
tips of the branches are the isolated virus samples; the nodes, or connections
of branches, correspond to the consensus sequences of their shared ancestors.
In collaboration with Walter M. Fitch of
the University of California at Irvine, Palese found for inuenza A an essentially
linear relation between the degree of difference for any two sequences and the
amount of time since their divergence.
Depending on the sequences they examined, two to four mutations appeared
per year. The tip-to-node distances on
the tree, which reected the spread of
individual sequences, corresponded to
roughly ve years of evolution.
nfortunately, the case of inuenza A is as yet unique: no other

collections of viruses that extend
across 50 years currently exist. Nevertheless, other researchers have made
progress by employing a dierent approach. Whereas Palese tracked the evolution of a virus over time, those workers have reconstructed evolutionary
trees by making inferences from the
similarities of dierent viruses and viral strains that abound at approximately the same time. Gerald Myers of Los
Alamos National Laboratory has made
such a tree for the AIDS-causing strain

HIV-1, using samples collected from
1985 to 1987.
The principal dierence between the
tree for HIV-1 and that for inuenza A
virus is the length of their branches.
According to the scheme Myers developed, all the early strains of HIV-1 came
from African sources. Looking at the
tree, we can almost trace the journey
of the virus from that continent to the
rest of the world. Indeed, one can extend the tree even further back into evolution by nding the relations between
HIV-1, HIV-2 and various forms of simian immunodeciency viruses (SIVs).
For determining when these viruses
diverged, it would be helpful if the separation in the sequences could be used
as a measure of evolutionary time. Sadly, the problem is not that simple. If
two long, originally identical sequences
mutate randomly, it is at rst unlikely
that they will undergo the same changes at the same positions. Mutations will
increase their distance from the original consensus sequence, and those
changes will accumulate almost linearly with respect to time.
Eventually, however, when enough mutations have accumulated, some of them
will probably reverse a previous change
or duplicate a change in the other sequence. As a result, the amount of difference between the sequences will decrease or stay constant, and their distance from the original consensus sequence will nally uctuate around a
given value. Past a certain point, then,
the passage of more time does not add
more distance. For a genetic sequence
in which any one of the four nucleotides could occupy any position, that
distance is 75 percent of the total sequence length.
Moreover, the assumption of uniform
substitution probabilities is usually not
correct. Some positions are almost constant because of tness constraints;
some vary at a normal rate, whereas
still others are hypervariable and change
rapidly in response to the selection pressure imposed on them by the immune
response of their host. The constant, variable and hypervariable positions would
each evolve according to a dierent distance-time relation. Applying dierent
relations to an interpretation of the evolutionary distances would give results
for old divergences that diered by orders of magnitude. The lengths of the
branches in the evolutionary trees cannot divulge when new viruses evolved.
Sequence space diagrams can, however. My colleagues Katja Nieselt-Struwe
and Ruthild Winkler-Oswatitsch of Gttingen, Andreas Dress of the mathematics department of Bielefeld University
and I have taken that approach. We developed a mathematical method of analyzing the relations within a quasispecies that we call statistical geometry in
sequence space. That analysis allows
us to determine how often on average
dierent types of changes occur at different positions. It enables us to classify
dierent positions in the viral sequences as constant, variable or hypervariable. From that information, we can deduce roughly how long dierent viral
lineages have existed and the frequency with which dierent types of mutations occur.
hat do the statistical geometries of the inuenza A, polio1 and immunodeciency viruses reveal? For the tree of inuenza A
virus, the probability of mutations that
would parallel or reverse previous changes is small. As Paleses study indicated,
the amount of dierence between strains
of the virus increases almost linearly
over time. An intriguing prediction also
emerges from the data: if all the mutable positions in the virus continue to
change at the indicated rates, the inuenza virus should completely lose its
identity within a few hundred years. Because some positions must be constant,
the inuenza A virus will probably remain a pathogen, because to survive, it
will need to infect humans, but we cannot predict what its pathology will be.
For polio-1 virus, the picture is entirely dierent. In the studied sequence
segment, the nucleotides that occupy
the rst and second positions in each
codon scarcely change at all. Mutations
at those positions must be strongly eliminated from the quasispecies by selection. Conversely, the nucleotides at the
third codon positions are almost completely randomized. As a result, even
though the poliovirus has about the
same error rate as the inuenza virus,
only mutations that do not change the
encoded amino acids appear in the quasispecies. The proteins in the poliovirus are very highly conserved.
The immunodeciency viruses have
a third type of statistical geometry. All
three codon positions are appreciably
randomized for all types of changes. We
have been able to determine the prevalence of constant, variable and hypervariable sites within the gene for an HIV
surface protein that we analyzed. From
that information, we were able to estimate how long it must have taken for
the immunodeciency viruses to have
diverged to the observed degree.
About 20 percent of the positions are
constant, apparently because they are
necessary for HIV to function as a retrovirus. They establish that HIV is the
47
STIRRING
CONTROL
COMPUTER
TURBIDITY
CONTROL
PUMP
NUTRIENT
SOLUTION
PUMP
FERMENTER
FLOW REACTORS
EVOLUTION MACHINES of various types are used in the authors laboratory to study the evasive changes that virus populations can make when subjected to selection pressure. The
machines create systems of cell cultures in which viruses grow
under tightly controlled conditions for many generations. Nutrient solution is pumped into a fermenter in which grow host
cells, such as the bacteria Escherichia coli. These cells are then
descendant of an old viral family. About

70 percent of the positions are variable
and have an average lifetime of about
1,000 years (give or take a few hundred). They seem to give HIV its specic
characteristics. Many of these positions
dier in HIV-1, HIV-2 and the SIV sequences, which indicates that they must
have evolutionarily diverged long ago.
My colleagues and I estimate that it was
600 to 1,200 years ago (or even longer, because more constant positions may
yet be hidden in the data). Contrary to
the evidence of the epidemiologic curves,
therefore, HIV is not a new virus, although its pathogenicity may have varied over the centuries.
About 200 positions in the studied
HIV geneabout 10 percent of the totalare hypervariable and change on
average within 30 years. They provide
the tremendous variability that enables
HIV to thwart the attempts by its hosts
immune system to eliminate it. They
may also be directly responsible for
much of the damage that the virus does
to the immune system. According to a
theory advanced in 1992 by Robert M.
May and Martin A. Novak and their colleagues at the University of Oxford, HIV
uses its capacity for variance to outank the immune response of its host.
48
COLLECTED
SAMPLES
pumped into an array of environmentally controlled vessels

called ow reactors, where the viruses can parasitize their
hosts. Samples of the virus populations can be withdrawn
from the ow reactors for analysis. A computer regulates
components of the system, such as the pumps and the controls for stirring medium turbidity, that determine the growth
conditions and selection pressures on the viruses.
The number of dierent sequences that

result from mutations at hypervariable
sites outruns by far the capacity of the
immune system to generate lymphocytes. If HIV can change at all its hypervariable sites in 30 years, it could exhaust the immune system in only a fraction of that time. The virus can produce mutants that evade the immunologic defenses, particularly because its
infection targets are the T lymphocytes
that control the immune response.
Computer simulations carried out by
the Oxford group verify those predictions. That theory, based on the quasispecies nature of the virus, also satisfactorily explains the decade-long delay
that usually occurs between the initial
viral infection and the fatal state of
the disease, when the immune system
breaks down fairly suddenly. It may take
that many years for HIV to exhaust the
adaptive resources of the immune system. New experiments will test whether
this explanation is correct.
The statistical geometry data also offer insights into ways of ghting HIV
and other viruses. The most common
way to rid an infected individual of a
virus is to stimulate, activate or support the immune system, as a vaccine
does. An awareness of the variational
exibility of viruses suggests that three

additional strategies must also be explored to improve vaccines. One is to
nd stable immunologic features in the
viral quasispecies against which highly specic monoclonal antibodies could
be directed. The second is to create antibodies that can act against a broad
spectrum of the likely mutant viruses
that would otherwise permit a quasispecies to escape attack. The third is to
spot such escape mutants during an
early phase of infection and to outmaneuver them with specic agents before they can produce progeny.
he most fruitful approaches may

vary with dierent viruses. For
example, the immune system can
quickly learn to recognize the almost
constant protein features of the poliovirus. That virus has no chance of surviving if it encounters a vaccinated host.
The real eectiveness of that protection
became apparent only recently when
researchers discovered that the mild
strain of polio-1 virus in the Sabin vaccine diers from the pathogenic wild
type at only two nucleotide positions.
It is entirely possible, therefore, that a
few of the polioviruses from a vaccine
do mutate into a pathogenic state inside
the host. Yet by the time those mutations occur, the immunologic protection
of the host is already practically perfect. The success of the Sabin vaccine
in saving the lives of countless children
is unchallenged.
Inuenza is a quite dierent case, as
are other viruses. The targets for the immune response against inuenza change
steadily. Although the immune system
eventually copes with the virus and
quells the infection, there is no lasting
protection. As a consequence, people
can contract inuenza repeatedly, and
new vaccines must be prepared every
few years. John J. Holland of the University of California at San Diego and Esteban Domingo of the Independent University of Madrid have observed that the
viruses responsible for foot-and-mouth
disease and vesicular stomatitis, an infection of the oral membranes in livestock, behave in a similar way. HIV, with
its many variable and hypervariable positions, mutates even more rapidly and
radically. Vaccines may not have any
lasting value against such infections.
But vaccines are only one way to ght
viruses. The administration of drugs that
block viral replication is an extremely
common therapyand for AIDS it is
currently the sole therapy that is in any
way eective at slowing the progress of
the disease. In theory, articial chains
of RNA could be administered to patients to prevent or eliminate viral infections. Those RNA molecules would
hinder viral replication, either by binding to the viral RNA or by competing
with it for essential enzymes. Specic
factors that interfere with viral replication could also be incorporated into host
cells by genetic technology. Yet all these
approaches may have harmful side effects or would need to clear signicant
technical hurdles.
A further complication is that viruses may be able to mutate around such
obstacles. In my laboratory Bjrn F. Lindemann has used the understanding of
the replicative mechanism of the Q bacteriophage to test one antiviral strategy. He inserted the gene for the viral
coat protein into cells. The cells became resistant to infection because the
coat protein, a natural regulator of the
phages replication, blocked the transcription of viral genes.
Yet this strategy did not work perpetually : given sucient time and generations, the Q bacteriophage adapted by
mutating into a form that ignored the
coat protein signal. Lindemann demonstrated that fact using one of the automated evolution machines developed
recently in my laboratory. In these devices, viruses grow in host cells for extended periods under mild selection
VACCINATION has been extremely eective in controlling polio and some other
diseases. Because the proteins of poliovirus change very little over time, it is relatively easy to nd consistently good immunologic targets. Against more mutable
viruses, such as the AIDS virus, vaccination is much less potent.
pressures. Evolutionary biotechnology,

or applied molecular evolution, as it is
often called, is a rapidly emerging eld
of research that may have many applications in new antiviral strategies [see
Directed Molecular Evolution, by Gerald F. Joyce; SCIENTIFIC AMERICAN, December 1992].
One strategy may be resistant to the
evasive maneuvers of viruses: it would
exploit their nature as quasispecies and
thereby undermine the very basis of
their existence. Even in a successful viral quasispecies, only a small fraction
of the viral sequences in a host cell are
viable. If the error rates of viruses can
be increased moderately, just enough
to cross the critical error threshold that
denes their quasispecies, they would
experience a catastrophic loss of information. The viral quasispecies would fall
apart because it would be producing too
many nonviable mutants.
Using drugs that produce mutations,
Domingo and Holland have demonstrated that this approach works against the
virus that causes foot-and-mouth disease. For such a strategy to work as a
therapy, however, the drugs must change
the error rate of only the viral replicase
and not of enzymes essential to the
hosts well-being. Careful study of replicase mechanisms should bring about
such a possibility of interfering with
virus infection. This strategy would be
precisely the opposite of immunization
therapies that attempt to prevent the
appearance of escape mutants.
As of today, we know little about the
origin of viruses or their role in the evolution of the biosphere. Viruses come
and go: some adapt; others disappear.
The undeniable reality is that an estimated 13 million people worldwide are
infected with HIV. Pandoras box is still
open and releasing new ills. Nevertheless, our growing understanding of viruses suggests that, as in the original
myth, hope has not escaped.
FURTHER READING
MOLECULAR QUASI-SPECIES. Manfred Eigen, John McCaskill and Peter Schuster
in Journal of Physical Chemistry, Vol.
92, No. 24, pages 68816891; December 1, 1988.
ROLE OF GENOME VARIATION IN VIRUS
EVOLUTION. Manfred Eigen and Christof K. Biebricher in RNA Genetics, Vol.
3: Variability of RNA Genomes. Edited
by Esteban Domingo, John J. Holland
and Paul Ahlquist. CRC Press, 1988.
HOW OLD IS THE IMMUNODEFICIENCY VIRUS? Manfred Eigen and Katja NieseltStruwe in AIDS, Vol. 4, Supplement 1,
pages S85S93; 1990.
STATISTICAL GEOMETRY ON SEQUENCE
SPACE. Manfred Eigen and Ruthild Winkler-Oswatitsch in Molecular Evolution:
Computer Analysis of Protein and Nucleic Acid Sequences. Edited by Russell
F. Doolittle. Academic Press, 1990.
THE HYPERCYCLE: COUPLING OF RNA
AND PROTEIN BIOSYNTHESIS IN THE INFECTION CYCLE OF AN RNA BACTERIOPHAGE. M. Eigen, C. K. Biebricher, M.
Gebinoga and W. C. Gardiner, Jr., in Biochemistry, Vol. 30, No. 46, pages 11005
11018; November 19, 1991.
49
Australias Polar Dinosaurs

Their excellent night vision and apparent warm blood
raise a question: Could they have survived icehouse
conditions at the end of the Cretaceous period?
by Patricia Vickers-Rich and Thomas Hewitt Rich
n the Early Cretaceous period, about

100 million years ago, Australia lay
alongside Antarctica, which straddled the South Pole as it does today.
Australias southeastern corner, now
the state of Victoria, lay well inside the
Antarctic Circle. At that time, the region
hosted an assemblage of animals and
plants that lived under climate conditions having no modern analogue. The
average temperature, though low, ap-
LEAELLYNASAURA
pears to have been within the temperate range, yet the sun did not shine
throughout the long winter.
Many dinosaur lineages survived in
this strange environment after they had
died out in other places. At least one
member of the group evolved an adaptation to the cold and to the dark that
is interesting both in itself and for what
it tells of the passing of a biological
epoch. If global cooling indeed killed the
ALLOSAURUS
AUSTRALIAN DINOSAURS ourished in southeastern Victoria during the Early Cretaceous, when the region lay within
50
dinosaurs, as many paleontologists have

suggested, then Australias species were
the ones most likely to have survived
the longest. Did their adaptations to an
already marginal climate help them survive a sharp cooling trend, one that
caught species living on other continents
unprepared?
Although the Cretaceous fossil plants
of southeastern Australia have been
studied for more than a century, the
MUTTABURRASAURUS
the Antarctic Circle. This mural depicts six species that left
fossils there and a sevenththe large iguanodontid Muttabur-
animals remained mostly hidden until

recently. Around 1900 the geologist
William Hamilton Ferguson found two
bones that have had a bearing on later
paleontological workthe tooth of a
lungsh and the claw of a carnivorous
dinosaur, assigned to the theropod genus Megalosaurus. For the next 70 years,
as no further nds joined them, these
bones lay neglected in a cabinet in the
Museum of Victoria. Then, in 1978, two
graduate students at Monash University,
Tim F. Flannery and John A. Long, discovered near Fergusons original site the
rst specimens of a trove of dinosaur
bones embedded in hard sandstones and
mudstones from the Early Cretaceous.
These discoveriesonly an hour and
a halfs drive southeast of Melbourne
encouraged paleontologists to prospect
other coastal sites. In 1980 we struck a
rich lode in the Otway ranges, which
the Victorian government, at our suggestion, has since named Dinosaur Cove.
There, with the help of Earthwatch and
other volunteers and the National Geographic Society, the Australian Research
PTEROSAUR (FLYING)
Council and Atlas Copco, a manufacturer of mining equipment, we have

spent three months out of every year
chiseling, hammering and on occasion
blasting tunnels into the fossil-bearing
strata [see illustration on page 55]. This
project has been the center of our lives
and the lives of our co-workers, our
children and even our parents (two of
whom are paleontologists).
Dinosaur Cove and other sites of similar character were formed when violent,
seasonal streams swept broad oodplains of their accumulated bones and
PATRICIA VICKERS-RICH and THOMAS

HEWITT RICH collaborate in the study of
fossils. Vickers-Rich is a reader in earth sciences and in ecology and evolutionary biology at Monash University in Melbourne,
Australia. She is interested in reconstructing ancient environments, especially those
without modern analogues, and in analyzing rapid biotic change, such as mass extinctions. Rich is curator of vertebrate pa-
ANKYLOSAUR
rasaurusthat has been found only in Queensland, far to the

north. The paucity of large polar dinosaurs may reect a real
plant life, depositing this otsam and

jetsam at the bottom of shallow stream
channels. These deposits appear along
the southern Victorian shore because
only there could gnawing waves expose
the sediments laid down in the rift valley that formed when Australia and Antarctica went their separate ways, as did
the other fragments of Gondwana, the
ancient supercontinent [see illustration
on next page]. Only two fossil sites from
the same period have been found inland,
one in sediments laid down under far
quieter conditions at the bottom of an
leontology at the Museum of Victoria in

Melbourne. He conducts research on the
evolutionary patterns of Mesozoic vertebrates, specializing in primitive mammals
and ornithischian dinosaurs. The Riches received undergraduate degrees in paleontology from the University of California at
Berkeley and doctorates in geology from
Columbia University. They live near Melbourne and have two children.
ATLASCOPCOSAURUS
ORNITHOMIMOSAUR
absence or merely the selective preservation of small bones.

Peter Trusler painted the mural for Australia Post.
51
EQUATOR
AUSTRALIA
TODAY
T
RC
IRCLE
IC C
DINOSAUR
COVE
AN
TA
108 MILLION
YEARS AGO
SOUTHERN SUPERCONTINENT began to break up more than

100 million years ago, when a rift valley formed between Australia and Antarctica (left ). Stream channels in the valley re-
ancient lake. This inland site has therefore yielded some uncommonly well preserved specimens.
It must be noted that southeastern
Australias dinosaurs are known from a
mere 5,000 individual bones and two
partial skeletons. Just a few hundred of
the bones can be assigned to a given
species or genus. What they lack in number, however, they make up in scientific interest.
All eorts at interpretation revolve
around the estimation of temperature,
for which two methods have been tried.
Robert T. Gregory of Southern Methodist University and his associates infer
Australian paleoclimate from the ratio
of oxygen 18 to oxygen 16 trapped in
ancient rocks. They nd that mean annual temperatures probably approached
zero degrees Celsius but might have
reached as high as eight degrees C above
zero. Such values occur today in Hudson Bay, Saskatchewan (zero degrees C),
and in Minneapolis and Toronto (eight
degrees C above zero).
Robert A. Spicer of the University of
Oxford and Judith Totman Parrish of
the University of Arizona instead deduce temperature from the structure
of ancient plants, arriving at the somewhat higher mean annual temperature
of 10 degrees C. Their research has
demonstrated that polar Australia supported conifers, ginkgoes, ferns, cycads,
bryophytes and horsetails but only a
few angiosperms, or owering plants,
identiable by a sprinkling of pollen.
The angiosperms were then just beginning to spread into new niches. Perhaps they got their start by exploiting
weedy ecological systems in the rift val52
ceived bones gathered by oodwaters that periodically swept

these broad plains. The bones, together with clay and silt, created the fossil-bearing formations of Dinosaur Cove (right ).
leys that formed as the supercontinent

split apart.
Evergreens, which provided forage in
all seasons, had thick cuticles and other
structural features that indicate adaptation to cold or dryness (perhaps brought
on by winter freezing). Deciduous plants
oer another climatic clue: they seem to
have lost all their leaves at once. These
mass falls may have been triggered
by darkness or cold. Drought, however,
probably did not serve as a constant
cuethe sedimentary record and the
abundance of ferns and bryophytes argue for conditions that were moist in
all seasons except perhaps winter.
f the higher estimate of mean temperature is correct, Australia was

both temperate and subject to a
period of continuous darkness every yeara combination with absolutely no modern counterpart. The winter
night lasted between six weeks and four
and a half months, depending on the
true paleolatitude. Because the lower extreme of temperature would then have
fallen well below the mean, most of the
vertebrates preserved as fossils must
have lived quite close to their thermal
limits. Some, such as lungsh, cannot
now breed in waters colder than 10 degrees C.
If, on the other hand, the lower mean
temperature is correct, it becomes more
than a normal scientic challenge to
understand how this paleocommunity
functioned at all. Before seriously attacking this problem, scientists will rst
have to demonstrate that it exists. To
rene the estimate of the average annual temperature, a multidisciplinary
team is comparing oral, geochemical

and other forms of evidence.
Nothing in this fauna is quite so peculiar to the region as the koala is today,
for although the species and genera were
local, they belonged to cosmopolitan
families. Yet their adaptations are striking, as is the fact that some survived beyond the time of demise for their families elsewhere.
Among such anachronismsor relictsare the labyrinthodont amphibians, ancestors of modern amphibians
and reptiles. Most paleontologists had
thought this group went extinct by the
Jurassic, some 160 million years ago.
In the past 15 years, however, Michael
Cleeland and Lesley Kool of Monash University found three jaws from this group
in Victorian sediments dating from the
Early Cretaceous. Two of the jaws were
unmistakable, because their teeth had
the labyrinthine infolding of the enamel that gives this group its name. At least
one large species of labyrinthodonts
lived in polar Australia 115 million years
ago, several million years after the group
had died out elsewhere.
How did they survive? We suspect that
the cool weather preserved the animals
from competition with crocodiles, which
were probably poorly adapted to the
conditions prevailing in southeastern
Australia until the onset of climatic
warming during the last ve million
years of the Early Cretaceous. The hypothesis rests on the fact that contemporary crocodilians now live in waters
no colder than 10 degrees C, whereas
some modern frogs and salamanders
can be active in meltwater from snow.
Another late survivor was the famil-
iar Allosaurus, a carnivorous theropod.

Elsewhere in the world this animal
ranged up to ve meters in height, but
the southeastern Australian specimen
stood no more than two meters high
hardly taller than a human. This pygmy is the latest-surviving allosaur that
has yet been found. It remains unclear
whether this species also owed its longevity to some niche that cold climate
may have carved out for it. The discovery of juvenile forms (but no eggshells,
as yet) does suggest that these dinosaurs were not just casual visitors but
lived near the pole for much of the year,
using the area as a nursery during the
period of maximum sunlight.
Unlike the allosaurs, many dinosaurs
of Australia were not the last in their
lineage; some may have been the rst.
At least two and perhaps as many as
four families of dinosaurs have been recognized that include forms which are
either the oldest or among the oldest
of their kind. For instance, the ornithomimosaurs, carnivores of ostrichlike
size and appearance, are manifestly
primitive and among the oldest within
this group; only a Late Jurassic species
from East Africa predates the Australian form. The elongated, slender hind
limbs of the Australian species made
them the gazelles of the dinosaur world,
able to escape from predators and to
run down prey. The ornithomimosaurs
probably originated in Gondwana and
spread northward to join the later Cretaceous faunas of North America and Eur-
asia, where they enjoyed wide success.

Two very small theropods remain unidentied, but one seems to resemble
an egg-eating oviraptosaur, known until now exclusively from the younger Cretaceous rocks of North America and
Asia. These groups may also have an origin in Gondwana.
Yet another dinosaur group that has
recently been identied belongs to the
neoceratopsians, or horned dinosaurs.
Identication is tentative, as it is based
on just two ulnae (part of the lower
arm), but the similarity to Leptoceratops,
a browser the size of a sheep, is uncanny. Previously, all neoceratopsian records dated from the Late Cretaceous
and, with the exception of a few bones
from Argentina, came from the Northern Hemisphere. This dinosaur family
may also have arisen in the southern
supercontinent.
The Early Australian Cretaceous also
reshaped forms that continued to ourish in other regions. By far the most
successful such group consisted of the
hypsilophodontid dinosaurs. These animals, most of them hardly larger than
a chicken, were bipeds built for speed,
with large hind legs, small but well-developed hands, substantial tails and
for the most partherbivorous habits.
They thus resembled wallabies in both
shape and ecological role.
The family Hypsilophodontidae was
common throughout the world from
the Middle Jurassic to Late Cretaceous
times, but its prominence reaches an
ACUTE NIGHT VISION is suggested by the eyes and brain of

Leaellynasaura amicagraphica , a hypsilophodontid shown
here at life size (above, right ). The large eyes were common
to all hypsilophodontids and may have helped the group dom-
absolute and relative peak in the Victorian sediments. Not only do hypsilophodontids constitute most of the dinosaur remains, they are also represented
by four to ve genera, depending on the
taxonomic criteria one uses, and ve to
six species. Other areas, some much
more richly endowed with dinosaur species, never harbored more than three
kinds of hypsilophodontids at a time.
Something clearly favored the diversication of this group in polar Australia.
particularly intriguing adaptation

of at least one species of polar
hypsilophodontid is suggested
by the magnicently preserved brain
cast of Leaellynasaura amicagraphica
(named after our daughter, friends of
the Museum of Victoria and the Nation-
inate an environment marked by seasonal darkness. This hypothesis may also explain the huge optic lobes, of which the
left one can be seen at the rear of this natural brain cast (below; bump at far right), formed when silt solidied in the skull.
53
al Geographic Society). The brain, unusually large for a dinosaur of this size,
bears the marks of optic lobes whose relative size is easily the greatest ever documented in a hypsilophodontid.
How is one to interpret these enlarged
lobes? We hypothesize that they enhanced the animals ability to see in the
dark, enabling them to forage eectively during the long winter months. There
would have been no lack of food then,
for those capable of seeing it: the herbivores could have lived o evergreens
and deciduous leaf mats, and the carnivores could have hunted the herbivores.
This hypothesis also explains why this
group came to dominate the polar environment in the rst place. Hypsilophodontids everywhere in the world had
large eyes and, presumably, acute vision.
That trait could have given them their
foothold in polar Australia. Once estab-
lished in this protected environment,

the hypsilophodontids could have competed with one another to produce the
observed diversity of genera and species, perhaps all sharing hypertrophied
optic lobes.
If the animals foraged at night, they
must have been active at freezing or subfreezing temperatures. This feat goes far
beyond the cold tolerance of any modern reptile, even the New Zealand tuatara, Sphenodon punctatus, which can
remain active at ve degrees C provided it can sun itself. Leaellynasaura could
have survived solely by maintaining a
constant body temperature, eating frequently, as birds do in wintertime.
Pterosaurs, ying reptiles, and ankylosaurs, heavily armored dinosaurs,
also appear, but in such fragmentary
remains as to tell the student little of
the animals lives. Much can be gleaned
BONE TO STONE: Leaellynasaura as it might have appeared in

the process of becoming a fossil. A bone assemblage from an
54
from one handful of teeth, however, for

they come from plesiosaurs. These longnecked reptiles, not themselves dinosaurs, generally paddled the seas, but
here they inhabited fresh water in the
ancient valley between Australia and
Antarctica. They thus recall the Ganges
River dolphin, one of the few cetaceans
that live in fresh water.
The sauropods alone are absent. These
giants, familiar from the example of
Apatosaurus (or Brontosaurus, as it is
more popularly known), lived at that
time in Australias lower latitudes. None,
however, has been found further south
nor, indeed, in any of the nine Cretaceous polar dinosaur sites so far identied in both hemispheres. The only polar sauropod yet discovered is the much
older (early Jurassic) Rhoetosaurus from
northeastern Australia.
The apparent restriction of these large
individual could have fossilized in this way only if the stream

channel was choked o, forming an oxbow or billabong.
HARD ROCK makes hard work for these volunteer paleontologists. Full-scale mining techniques (left ) and explosives (right)
dinosaurs to lower latitudes in the Cretaceous of Australia may be real or merely an artifact of sampling. We worry
about this question because the oodwaters that broke out of rain-swollen
rivers would have collected small and
medium-size bones but left large ones.
The body of a sauropod would have
stayed put rather than oating to a place
where many specimens were concentrated in the small ood channels, which
were no more than ve to 10 meters in
width and 20 to 30 centimeters in depth.
Yet we suspect there was an underlying tendency toward small body size in
these polar environs. None of the hypsilophodontids, it must be remembered,
stood taller than a human, and most
were barely knee-high. The dwarf Allosaurus matches the smallest we have examined in the North American collections. The ornithomimosaur is equally
unprepossessing, and the protoceratopsid and the ankylosaur are no bigger than a sheep. A single fragment of
a claw constitutes our sole record of a
forma carnivore, apparently similar
to Baryonyx of Englandwhich may
have measured up to eight meters in
length.
This pattern contradicts the scaling
laws that Bergmann and Allen formulated in the 19th century. According to
these laws, animals in a given lineage
tend to become larger and more compact as the average temperature of their
environment falls. This trend is exemplied by the comparison of mountain
lions in Canada with pumas of Central
America and of human populations in
the subarctic and tropical zones.
Other factors also determine body
dimensions, especially the size of the
territory in which a population lives.
Individuals found on islands are often smaller than their mainland counterparts. For example, there were dwarf
elephants on the ancient Mediterranean
extract fossil-bearing slabs, which tend to fracture along the

planes containing the largest treasures.
islands, and pygmy mammoths were

recently found in 4,000-year-old sediments on islands o the north coast of
Siberia. Dwarsm may be a response to
selective pressure to increase the number of individuals so as to ensure a gene
pool diverse enough for the species to
survive in a restricted area. This eect
has also been noted on peninsulas
and ancient southeast Australia was a
peninsula of the Gondwana landmass.
The dinosaurs on that peninsula were
trapped virtually at the end of the earth.
Their direct path north was blocked
by a vast inland sea, which they could
have passed only by going hundreds of
kilometers to the west before wheeling
about to the north. At the end of such
labors, they would have been able to
catch, at most, an hour of sun a day in
winter. Migration would have made little sense for such small animals.
Less formidable barriers sealed in the
dinosaurs of the one other polar site that
has yielded large quantities of fossils:
the north slope of Alaska. The dinosaurs
there had a clear north-south corridor
along which they could migrate with
ease. It is signicant that those dinosaurs were bigat least equal in size to
caribou, wildebeest and other modern
animals that migrate.
ne must question whether animals so superbly adapted to the

cold and the dark could have
been driven to extinction by an articial winter, such as is supposed to have
followed a cataclysmic event at the
boundary between the Cretaceous and
Tertiary formations. It is proposed that
the cataclysm, perhaps a collision with a
comet or asteroid or a series of volcanic eruptions, suused the atmosphere
with a blanket of dust, excluding sunlight and freezing or starving most animals to death.
We suspect, however, that no such
articial winter could have killed the dinosaurs unless it lasted for a long time,
certainly more than a few months. Otherwise at least a few of the polar dinosaurs would have survived the cataclysm. Of course, it is possible that some
other development had already ended
the reign of southern Australias dinosaurs by the end of the Cretaceous.
Arthur Conan Doyle once dreamed
of a plateau in South America that time
forgot, where dinosaurs continued to
reign. Reports earlier this year that
dwarf mammoths survived to early historical times, in islands o the coast
of Siberia, give force to such speculation. If dinosaurs found a similar haven
in which they outlived the rest of their
kind, then we think polar Gondwana,
including southeastern Australia, is a
likely place to look for it.
FURTHER READING
EVIDENCE FOR LOW TEMPERATURES AND
BIOLOGIC DIVERSITY IN CRETACEOUS
HIGH LATITUDES OF AUSTRALIA. P. V.
Rich, T. H. Rich, B. E. Wagstaff et al. in
Science, Vol. 242, pages 14031406; December 9, 1988.
OXYGEN ISOTOPIC COMPOSITION OF CARBONATE CONCRETIONS FROM THE LOWER CRETACEOUS OF VICTORIA, AUSTRALIA. R. T. Gregory, C. B. Douthitt, I. R.
Duddy et al. in Earth and Planetary Science Letters, Vol. 92, No. 1, pages 27
42; February 1989.
POLAR DINOSAURS AND BIOTAS OF THE
EARLY CRETACEOUS OF SOUTHEASTERN
AUSTRALIA. T. H. Rich and P. V. Rich in
National Geographic Research, Vol. 5,
No. 1, pages 1553; Winter 1989.
CONTINENTAL CLIMATE NEAR THE ALBIAN SOUTH POLE AND COMPARISON
WITH CLIMATE NEAR THE NORTH POLE.
J. T. Parrish, R. A. Spicer, J. G. Douglas
et al. in Geological Society of America,
Abstracts with Programs, Vol. 23, No. 5,
page A302; Annual Meeting, 1991.
55
Accurate Measurement of Time

Increasingly accurate clocksnow losing no more than
a second over millions of yearsare leading to such advances
as refined tests of relativity and improved navigation systems
by Wayne M. Itano and Norman F. Ramsey
ew people complain about the accuracy of modern clocks, even if

they appear to run more quickly
than the harried among us would like.
The common and inexpensive quartzcrystal watches lose or gain about a second a weekmaking them more than
sucient for everyday living. Even a
spring-wound watch can get us to the
church on time. More rigorous applications, such as communications with interplanetary spacecraft or the tracking of
ships and airplanes from satellites, rely
on atomic clocks, which lose no more
than a second over one million years.
WAYNE M. ITANO and NORMAN F.

RAMSEY have collaborated many times
before writing this article: Itano earned
his Ph.D. at Harvard University under the
direction of Ramsey. Itano, a physicist at
the Time and Frequency Division of the
National Institute of Standards and Technology in Boulder, Colo., concentrates
on the laser trapping and cooling of ions
and conducts novel experiments in quantum mechanics. He is also an amateur
paleontologist and fossil collector. Ramsey, a professor of physics at Harvard,
earned his Ph.D. from Columbia University. He has also received degrees from
the University of Oxford and the University of Cambridge, as well as several honorary degrees. A recipient of numerous
awards and prizes, Ramsey achieved the
highest honor in 1989, when he shared
the Nobel Prize in Physics for his work
on the separated oscillatory eld method and on the atomic hydrogen maser.
56
There might not seem to be much

room for the improvement of clocks or
even a need for more accurate ones.
Yet many applications in science and
technology demand all the precision
that the best clocks can muster, and
sometimes more. For instance, some
pulsars (stars that emit electromagnetic radiation in periodic bursts) may in
certain respects be more stable than
current clocks. Such objects may not
be accurately timed. Meticulous tests
of relativity and other fundamental
concepts may need even more accurate
clocks. Such clocks will probably become available. New technologies, relying on the trapping and cooling of atoms and ions, oer every reason to believe that clocks can be 1,000 times
more precise than existing ones. If history is any guide, these future clocks
may show that what is thought to be
constant and immutable may on ner
scales be dynamic and changing. The
sundials, water clocks and pendulum
clocks of the past, for example, were
suciently accurate to divide the day
into hours, minutes and seconds, but
they could not detect the variations in
the earths rotation and revolution.
clocks accuracy depends on the

regularity of some kind of periodic motion. A grandfather
clock relies on the sweeping oscillation
of its pendulum. The arm is coupled to
a device called an escapement, which
strikes the teeth of a gear in such a way
that the gear moves in only one direc-
tion. This gear, usually through a series

of additional gears, transfers the motion
to the hands of the clock. Eorts to improve clocks are directed for the most
part toward nding systems in which
the oscillations are highly stable.
The three most important gauges of
frequency standards are stability, reproducibility and accuracy. Stability is
a measure of how well the frequency
remains constant. It depends on the
length of an observed interval. The
change in frequency of a given standard might be a mere one part per 100
billion from one second to the next,
but it may be largersay, one part per
10 billionfrom one year to the next.
Reproducibility refers to the ability of
independent devices of the same design to produce the same value. Accuracy is a measure of the degree to
which the clock replicates a dened interval of time, such as one second.
Until the early 20th century, the most
accurate clocks were based on the regularity of pendulum motions. Galileo
had noted this property of the pendulum after he observed how the period of oscillation was approximately independent of the amplitude. In other
words, a pendulum completes one cycle in about the same amount of time,
no matter how big each sweep is. Pendulum clocks became possible only
after the mid-1600s, when the Dutch
scientist Christiaan Huygens invented
an escapement to keep the pendulum
swinging. Later chronometers used the
oscillations of balance wheels attached
TRAPPED MERCURY IONS, separated by about 10 microns, uoresce under illumination by ultraviolet light (photograph).
The ions are held by oscillating electric elds generated by
electrodes (cutaway diagram). Static electric potentials (not
shown) prevent the ions from escaping through the ends of
the trap. Strings of trapped ions may lead to new timing devices more stable than conventional atomic clocks.
to springs. These devices had the advantage of being portable.

Considerable ingenuity went into
improving the precision of pendulum
and balance-wheel clocks. Clockmakers would compensate for temperature
changes by combining materials with
dierent rates of thermal expansion.
A more radical approach came in the
1920s, when William H. Shortt, a British engineer, devised a clock in which a
slave pendulum was synchronized to
a free pendulum. The free pendulum
oscillates in a low-pressure environment
and does not have to operate any clock
mechanism. Instead it actuates an electrical switch that helps to keep the
slave pendulum synchronized. As a result, the period of the Shortt clock is
extremely stable. These clocks had an
error of a few seconds in a year (about
one part per 10 million) and became
the reference used in laboratories.
The next major advance in timekeeping was based on the development
of quartz-crystal electronic oscillators.
The frequency of such devices depends
on the period of the elastic vibration of
a carefully cut quartz crystal. The vibrations are electronically maintained
through a property of such crystals
called piezoelectricity. A mechanical
strain on the crystal produces a low
electric voltage; inversely, a voltage induces a small strain.
The quartz vibrates at a frequency
that depends on the shape and dimensions of the crystal. In some wristwatches, it is cut into the shape of a
tuning fork a few millimeters long. In

other timepieces, it is a at wafer. The
quartz is connected to an electric circuit that produces an alternating current. The electrical feedback from the
quartz causes the frequency of the circuit to match the frequency at which
the crystal naturally vibrates (usually
32,768 hertz). The alternating current
from the circuit goes to a frequency divider, a digital electronic device that
generates one output pulse for a xed
number of input pulses. The divider
also actuates either a mechanical or
digital electronic display.
In the late 1920s Joseph W. Horton
and Warren A. Marrison, then at Bell
Laboratories, made the rst clock based
on a quartz-crystal oscillator. In the
1940s quartz-crystal clocks replaced
Shortt pendulum clocks as primary laboratory standards. These clocks were
stable to about 0.1 millisecond per day
(about one part per billion). Relatively
inexpensive, quartz clocks continue to
be extensively used. The timekeeping elements of common quartz watches and
clocks are simplied and miniaturized
versions of quartz frequency standards.
Quartz wristwatches became common
once the ability emerged to cut the
quartz into thin, tuning-fork shapes reliably and to manufacture miniature, lowpower digital electronic components.
Yet quartz-crystal clocks prove inadequate for many scientic applications,
such as tests of relativity. According
to Albert Einsteins calculations, gravity
distorts both space and time. The dier-
ence in gravitational potential causes

time to pass more quickly high in the atmosphere than it does on the surface.
The dierence is slight. Time runs about
30 millionths of a second per year faster
at the top of Mount Everest than it does
at sea level. Only atomic frequency standards achieve the requisite precision.
he quantized energy levels in atoms and molecules provide the

physical basis for atomic frequency standards. The laws of quantum mechanics dictate that the energies of a
bound system, such as an atom, have
certain discrete values. An electromagnetic eld can boost an atom from one
energy level to a higher one. The process can also work in reverse. If the
atom is in a high energy level, it can
drop to a lower level by emitting electromagnetic energy.
The maximum amount of energy is
absorbed or emitted at a denite frequencythe resonance frequency, or
the dierence between the two energy
levels divided by Plancks constant.
This value is sometimes called the Bohr
frequency. Such frequencies make ideal
time standards because they are extremely stable. Time can be kept by observing the frequencies at which electromagnetic energy is emitted or absorbed by the atoms. In essence, the
atom serves as the master pendulum
whose oscillations are counted to mark
the passage of time.
Although we have described general
quantum properties, the eects exploitSCIENTIFIC AMERICAN July 1993
57
ed in atomic clocks are slightly more

complicated. In most atomic clocks the
energy that atoms absorb or release actually results from transitions between
so-called hyperne energy levels. These
levels exist because of an intrinsic property of particles known as the magnetic
moment. Electrons and the nuclei of
most atoms spin about their axes as if
they were tops. In addition, they are
magnetized, like compass needles oriented along their axes of rotation. These
axes can have dierent orientations
with respect to one another, and the
energies of the orientations may dier.
These positions correspond to the hyperne levels. The nomenclature comes

about because the levels were rst observed in spectroscopy as small splittings of spectral lines.
On paper, standards based on atomic processes are ideal. In practice, perfection is elusive. Atoms do not absorb
or emit energy precisely at the resonance frequency. Some energy is spread
over a small interval surrounding the
frequencya smearing of frequencies,
so to speak. All else being equal, the
precision to which the resonance frequency can be measured is inversely proportional to this smearing. The
greater the spread, the less precise the
measurement. The spread is often expressed in terms of the quality factor,
or Q , which is equal to the resonance
frequency divided by the frequency
spread. In many cases, the higher the
resonance frequency, the higher the Q .
Furthermore, smearing is often inversely proportional to the time the atom is
in the apparatus. In those situations,
the Q of the resonance, and hence the
precision of the measurement, increases as the measuring time increases.
The motions of the atoms also introduce uncertainty by causing apparent
shifts in the resonance frequencies.
Such changes appear because of the
Doppler eect. The phenomenon can
be divided into rst- and second-order
shifts if the atoms are moving much
slower than the speed of light. The
rst-order Doppler shift is an apparent
change in the frequency of the applied
electromagnetic wave as seen by a moving atom. The amount of the shift is
proportional to the velocity of the atom.
If the atom moves in the same direction as the wave does, the shift is to a
lower frequency. If the atoms motion
is opposed to that of the wave, the
shift is to a higher frequency. If the directions are perpendicular, the rst-order shift is zero.
The second-order Doppler shift
comes about as a consequence of time
dilation. According to relativity, time
slows down for objects in motion; a
moving atom sees a slightly dierent
frequency than does a stationary counterpart. The eect on the resonance frequency is usually much smaller than
the rst-order shift. The second-order
shift is proportional to the square of the
atomic velocity and does not depend
on the relative directions of the atom-
MASTER PENDULUM of this 1920s Shortt

clock oscillates in an evacuated enclosure. It actuates an electrical switch to
synchronize a slave pendulum, which
drives the clock mechanism.
58
ic motion and the electromagnetic wave.

Several other factors aect the quality of the information. Atoms in the system may collide with one another; the
impacts add noise to the signal. The
surrounding environment can perturb
the resonance frequencies. Defects in the
electronic equipment, stray electromagnetic elds and the ever present thermal
radiation all introduce errors. Therefore,
a good atomic frequency standard not
only must establish a steady, periodic
signal but also must minimize these
potential errors.
ne of the earliest and now widely used methods to sidestep

many of these diculties is
called atomic beam resonance, pioneered by I. I. Rabi and his colleagues
at Columbia University in the 1930s.
The atoms emerge from a small chamber, exit through a narrow aperture and
then travel as a beam. The entire instrument can be shielded from stray
magnetic and electric elds and insulated from external sources of heat.
Perhaps more important, collisions of
atoms are virtually eliminated, because
the entire device is housed in a long,
evacuated chamber. The pressure in
the chamber is so low that the atoms
are unlikely to strike anything before
reaching the other end.
In simplied form, atomic beam resonance involves three steps. The rst is
to select only those atoms in the appropriate energy level. This selection is accomplished by using a specially shaped
magnetic eld, which acts as a kind of
lter. It allows atoms in one energy level to pass and blocks all others by bending the beam. Only atoms in the correct
energy level are bent the correct amount
to reach and pass through the aperture
that serves as the entrance to the cavity.
The second and crucial step is to
send the selected atoms into another
energy level. The task is accomplished
by passing the atoms through an oscillating microwave eld inside a cavity.
The atoms will go to another energy
level only if the frequency of the applied oscillating microwaves matches
their Bohr frequency.
The third step is to detect those atoms that have changed energy levels.
At this point, the beam of atoms passes through another magnetic eld lter, which allows only atoms in the correct energy level to strike a detector that
records the atoms as current ow. An
abundance of such atoms will exist if
the frequency of the applied oscillating
microwaves precisely matches their
natural frequency. If the frequency of
the applied microwave eld is o the
mark, fewer atoms change their energy
Resonance Frequency
tomic frequency standards depend on the quantization of

the internal energies of atoms or
molecules. A pair of such energy
levels, shown here as levels E1 and
E 2, is associated with an atomic resonance. The resonance frequency f, at which it absorbs or
emits electromagnetic radiation, is
f = (E 2 E 1)/h, where h is Plancks
constant. The radiation, however,
is not precisely f but instead is
spread over a range near f, called
Df. The precision to which f can
be measured is proportional to
the quality factor, Q , defined by
Q = f/Df. The higher the Q , the
more stable the clock.
E2
ENERGY
levels, and so fewer will strike the detector. One knows, therefore, that the
applied microwaves match the natural
frequency of the atoms if the number
of atoms striking the detector is maximal. An electronic feedback mechanism, called a servo loop, keeps this
value constant. If it nds that the current from the detector is falling o, it
changes the frequency of the applied
eld until the current reaches a maximum again.
By keeping the current from the detector at a maximum, the servo loop
maintains the frequency of the applied
microwave eld at the natural frequency of the atoms. To measure time, one
couples the applied eld to a frequency divider, which generates timing pulses. By analogy, the atoms represent the
quartz crystal in a watch or the master
pendulum in a Shortt clock. The applied microwave eld is the oscillating
circuit or the slave pendulum, which
actually drives the clock mechanism.
Minor variations of the atomic beam
standard exist. For example, in some devices the atoms that undergo a change
in energy level are made to miss, rather than strike, the detector. Not much
dierence in accuracy exists, however.
Rather all the versions to some extent
represent trade-os in terms of size,
cost and complexity.
A more important modication of
the atomic beam came in 1949, when
one of us (Ramsey) invented the socalled separated oscillatory eld method. Instead of irradiating the atoms
with a single applied eld, this technique relies on two elds, separated by
some distance along the beam path. Applying the oscillating eld in two steps
has many benets, including a narrowing of the resonance and the elimination
of the rst-order Doppler shift. Jerrold
R. Zacharias of the Massachusetts Institute of Technology and Louis Essen and
John V. L. Parry of the National Physical Laboratory in Teddington, England,
adapted this method to working frequency standards in the mid-1950s.
Currently the separated oscillatory
eld method provides the most reproducible clocks. The best ones are located at a few national laboratories, although smaller and less accurate versions are commercially available. The
clocks rely on cesium, which has several
advantages over other elements. It has a
relatively high resonance frequency
about 9,192 megahertzand low resonance width, which lead to an excellent
Q . Cesium can also be detected readily
and eciently; all that is needed is a hot
metal lament. When a cesium atom
strikes the lament, it ionizes and becomes observable as electric current.
f
PHOTON
E1
The Q s of these standards are about

100 million, exceeding the Q of quartz
wristwatches by a factor of several thousand. The greatest reproducibilities are
about a part per 10 14. The best cesium
frequency standards are so much more
reproducible than the rate of rotation
and revolution of the earth that in
1967 the second was dened as 9,192,631,770 periods of the resonance frequency of the cesium 133 atom.
ne of the most promising improvements in cesium atomicbeam standards is the use of

optical pumping to select the atomic
states. Beginning in the 1950s opticalpumping techniques were developed by
Francis Bitter of M.I.T., Alfred Kastler
and Jean Brossel of the cole Normale
Suprieure and others. In this method,
light, rather than a magnetic eld, selects atoms in the desired states. Before
the atoms are subjected to the microwave eld, radiation from a laser is used
to drive (or pump) the atoms from one
energy level into another. In fact, one
can control the number of atoms in energy levels by tuning the frequency of
the light.
After the atoms have been irradiated by the microwave eld, they pass
through a second light beam. Only atoms occupying the correct energy level
absorb this light, which they quickly re-
emit. A light-sensitive detector records

the reemissions and converts them
into a measurable current. As in atomic
beam resonance that relies on magnetic selection, one knows that the applied
microwave eld matches the natural
frequency of the atoms if the current
from the detector is at a maximum.
Using light instead of magnets has
many advantages. Perhaps the most crucial is that, with the right optical-pumping techniques, all the atoms in the
beam can be put into the desired energy
level. Magnetic selection merely lters
out those that are in the other energy
levels. Hence, the signal strength from
optical pumping is much higher than it
is from magnetic selection. Researchers
at various laboratories are developing
optically pumped cesium atomic-beam
clocks. One such clock, at the National
Institute of Standards and Technology
(NIST) in Boulder, Colo., has recently become the primary frequency standard
for the U.S. Designated NIST-7, it has an
expected error of one second in about
one million years, making it many times
more stable than its predecessor.
There is an optically pumped atomic clock that is available commercially. Such a clock is based on the 6,835megahertz, hyperne resonance of rubidium 87. Rather than moving through
the apparatus as a beam, the rubidium
atoms are contained in a glass cell. The
cell also houses a mixture of gases that
prevents the rubidium atoms from colliding with the cell walls. A discharge
lamp containing rubidium vapor, rather than a laser, irradiates the atoms. A
photovoltaic sensor on the opposite
side of the cell detects changes in the
amount of light absorbed by the atoms. The atoms are prepared, the microwaves applied and the light detected
in one cell. As a result, rubidium clocks
can be made to t in a cube about 10
centimeters on a side. In contrast, cesium beam clocks can extend from about
50 centimeters to more than ve meters. Rubidium clocks are also much
less expensive than are cesium ones.
The drawback is that the rubidium
devices are generally less accurate and
less reproducible. The Q of rubidium
standards is about 10 million, a factor
of 10 less than the cesium beams quality factor; their reproducibility is only
about a part per 10 10. Shifts in the resonance frequency mostly account for
the poor reproducibility. The frequent
collisions of the rubidium atoms with
other gas molecules cause the shifts.
But the rubidium standards short-term
stabilities are goodin fact, better than
those of some cesium atomic beams.
The atomic clocks described thus far
work in a rather roundabout wayby
59
detecting a change in some signal, such

as the number of atoms striking a detector, as the frequency of the applied oscillatory eld shifts. One way to make
use of the radiation emitted by the atoms more directly relies on the principle of the maser (an acronym for microwave amplication by stimulated
emission of radiation). In 1953 Charles
H. Townes and his associates at Columbia invented the rst maser, which
was based on ammonia. Beginning in
1960, Ramsey, Daniel Kleppner, now at
M.I.T., H. Mark Goldenberg, then at
Harvard University, and Robert F. C.
Vessot, now at the Harvard-Smithsonian Center for Astrophysics, developed
the atomic hydrogen maser, the only
type that has been used extensively as
an atomic clock.
In this instrument, a radio frequency discharge rst splits hydrogen molecules held in a high-pressure bottle
into their constituent atoms. The atoms emerge from a small opening in
the bottle, forming a beam. Those in the
higher energy level are focused by magnetic elds and enter a specially coated
storage bulb surrounded by a tuned,
resonant cavity.
In the bulb, some of these atoms will
drop to a lower energy level, releasing
photons of microwave frequency. The
photons will stimulate other atoms to
fall to a lower energy level, which in
turn releases additional microwave photons. In this manner, a self-sustaining
microwave eld builds up in the bulb
thus the name maser. The tuned cavity around the bulb helps to redirect
photons back into the system to maintain the stimulated emission process.
The maser oscillation persists as long
as the hydrogen is fed into the system.
A loop of wire in the cavity can detect
the oscillation. The microwave eld in-
DETECTOR
MAGNET B
MAGNET A
MICROWAVE
CAVITY
SOURCE
SERVO
MECHANISM
MICROWAVES
DIGITAL COUNTER
AND DISPLAY
FREQUENCY
DIVIDER
ATOMIC-BEAM frequency standards provide the most accurate, long-term timekeeping. Conventional atomic clocks rely on magnets (a). Atoms in the correct energy level are deected by magnet A through the microwave cavity. Microwave elds oscillating at the resonance frequency of the atoms drive some of them into a second energy level. These atoms are deected by magnet B so as to strike a detector. The
servo mechanism monitors the detector and maintains the frequency of the applied
microwaves at the resonance frequency. To keep time, some of the microwaves are
duces a current in the wire, which leads

out of the cavity to a series of circuits.
The circuits convert the induced current to a lower frequency signal suitable for generating timing pulses.
The resonance frequency in the hydrogen maser is about 1,420 megahertz, which is much lower than the
resonance frequency of cesium. But because the hydrogen atoms reside in the
bulb much longer than cesium atoms do
in a beam, the masers resonance width
is much narrower. Consequently, the Q
of a hydrogen maser standard is about
10 9, exceeding the Q of the cesium
atomic clock by an order of magnitude.
In addition, a hydrogen maser has the
TUNED CAVITY
MICROWAVE
PHOTONS
DIGITAL COUNTER
AND DISPLAY
MAGNET
SOURCE
highest stability of any frequency standard, better than one part per 10 15.
Unfortunately, the masers superior
attributes last just for a few days. Beyond that, its performance falls below
that of cesium beams. The stability decreases because of changes in the cavitys resonant frequency. Collisions between the atoms and the bulb shift the
frequency by about one part per 10 11.
One way to overcome the problem is
to operate the hydrogen maser at low
temperatures. This condition allows
more atoms to be stored (thus resulting
in a stronger signal) and reduces electronic noise. Coating the inside of the
bulb with superuid liquid helium also
enhances performance. This substance
acts as a good surface against which the
hydrogen atoms can bounce. More eective magnets, better coating substances
and servo loop techniques that keep the
cavity resonance centered on the atomic resonance are other approaches now
being taken to improve maser stability.
ATOMIC HYDROGEN MASER relies on a self-sustaining microwave eld to serve as a

frequency standard. Hydrogen atoms in the correct energy level are deected by a
magnet into a storage bulb. Some atoms will drop to a lower level, releasing a microwave photon. The photon stimulates other atoms to drop to a lower level, which
produces more photons. The process quickly builds up a microwave eld in the bulb.
The eld induces an alternating current in a wire placed in the cavity. The tuned cavity helps to redirect the photons back into the bulb to maintain the process.
lthough the cesium atomic-beam

frequency standard is the most
accurate, long-term standard
we have, several breakthroughs have
indicated that it is possible to fabricate
even more precise clocks. One of the
most promising depends on the resonance frequency of trapped, electrically
charged ions. Trapped ions can be suspended in a vacuum so that they are
almost perfectly isolated from disturbing inuences. The ions themselves
stay well separated from one another
62
FREQUENCY
DIVIDER
STORAGE BULB
b
DETECTOR
MICROWAVE
CAVITY
SOURCE
LASER B
SERVO
MECHANISM
LASER A
MICROWAVES
DIGITAL COUNTER
AND DISPLAY
FREQUENCY
DIVIDER
directed to a device that divides the frequency into usable timing pulses. Optically
pumped standards (b) use light rather than magnets to select atoms. Laser A pumps
the atoms into the right energy level, preparing them to be excited by the microwaves. Only atoms placed in the correct energy level by the microwaves absorb light
from laser B. They quickly reemit that energy, which is sensed by a photodetector.
An optically pumped clock using cesium atoms at the National Institute of Standards
and Technology, called NIST-7, now keeps time for the U.S. (photograph).
because they have the same electric

charge. Hence, they do not suer collisions with other particles or with
the walls of the chamber. Ions can be
trapped for long periods, sometimes
for days.
Two dierent types of traps are
used. In a Penning trap, a combination
of static, nonuniform electric elds and
a static, uniform magnetic eld holds
the ions. In a radio frequency trap (often called a Paul trap), an oscillating,
nonuniform electric eld does the job.
Each type of trap has its own characteristic shortcoming. The strong magnetic elds of Penning traps can alter the
resonance frequency. The electric eld
in Paul traps can create heating eects
that cause Doppler shifts. The kind of
trap chosen depends on its suitability
for a particular experimental setup.
Workers at Hewlett-Packard, the Jet
Propulsion Laboratory in Pasadena,
Calif., and elsewhere have fabricated experimental standard devices using Paul
traps. The particles trapped were mercury 199 ions. This ion was selected
because it has the highest hyperne
frequency40.5 gigahertzof all the
atoms that are appropriate for the
trapping technique. A few million such
ions are caught between the electric
elds generated by electrodes. Then
the ions are optically pumped by ultraviolet radiation from a lamp. Subsequent operation resembles that of the
optically pumped standards, but the
maximum Q s of trapped-ion standards
exceed 10 12. This value is 10,000 times
greater than that for current cesium

beam clocks. Their short-term stabilities are also extremely good, although
they do not yet reach those of hydrogen masers. The second-order Doppler
shift limits the reproducibility to about
one part per 10 13.
The Doppler shifts can be greatly reduced by laser cooling. In 1975 David J.
Wineland, now at NIST, Hans G. Dehmelt of the University of Washington,
Theodor W. Hnsch, now at the University of Munich, and Arthur L. Schawlow
of Stanford University rst proposed
such a technique. In essence, a beam
of laser light is used to reduce the velocities of the ions. Particles directed
against the laser beam absorb some of
the laser photons momentum. As a result, the particles slow down. To compensate for the Doppler shifting as the
particle moves against the laser, one
tunes the beam to a frequency slightly
lower than that produced by a strongly
allowed resonance transition.
Many laboratories are developing frequency standards based on laser-cooled
ions in traps. A standard based on beryllium 9 ions, laser-cooled in a Penning
trap, has been constructed. Its reproducibility is about one part per 10 13,
limited as it is by collisions of the ions
with neutral molecules. Improvements
in the quality of the vacuum should
signicantly increase the reproducibility because the uncertainty of the second-order Doppler shift is only about
ve parts per 10 15.
During the past few years, there have
been spectacular developments in trapping and cooling neutral atoms, which

had been more dicult to achieve than
trapping ions. Particularly eective laser cooling results from the use of three
pairs of oppositely directed laser-cooling beams along three mutually perpendicular paths. A moving atom is then
slowed down in whatever direction it
moves. This eect gives rise to the designation optical molasses. Several investigators have contributed to this breakthrough, including William D. Phillips
of NIST in Gaithersburg, Md., Claude
Cohen-Tannoudji and Jean Dalibard of
the cole Normale Suprieure and Steven Chu of Stanford [see Laser Trapping of Neutral Particles, by Steven Chu;
SCIENTIFIC AMERICAN, February 1992].
Neutral-atom traps can store higher
densities of atoms than can ion traps,
because ions, being electrically charged,
are kept apart by their mutual repulsion. Other things being equal, a larger
number of atoms results in a higher
signal-to-noise ratio.
The main hurdle in using neutral
atoms as frequency standards is that
the resonances of atoms in a trap are
strongly aected by the laser elds. A
device called the atomic fountain surmounts the diculty. The traps capture
and cool a sample of atoms that are
then given a lift upward so that they
move into a region free of laser light.
The atoms then fall back down under
the inuence of gravity. On the way up
63
LASER
and again on the way down, the atoms

pass through an oscillatory eld. In this
way, resonance transitions are induced,
just as they are in the separated oscillatory eld beam apparatus.
The Q of such a device can be higher
than that of an atomic beam because
the time between the two passages can
be longer. Experiments on atomic fountains have been done by Chu and his
co-workers at Stanford and by Andr
Clairon of the Primary Laboratory of
Time and Frequency in Paris and Christophe Salomon of the cole Normale
Suprieure and their co-workers.
Much current research is directed
toward laser-cooled ions in traps that
resonate in the optical realm, where frequencies are many thousands of gigahertz. Such standards provide a promising basis for accurate clocks because
of their high Q . Investigators at NIST
have observed a Q of 10 13 in the ultraviolet resonance of a single laser-cooled,
trapped ion. This value is the highest Q
that has ever been seen in an optical or
microwave atomic resonance. Because
he variety of high-performance
frequency standards that exist
today might seem to obviate the
need for future devices of even greater
performance. After all, current atomic
clocks are so accurate that they have
redened some of our basic units. As
mentioned earlier, the second is now
based on the resonance frequency of
the cesium atom. Also by international
agreement, the meter is dened as the
distance light travels in 1/299,792,458
of a second. The voltage unit is maintained by the characteristic frequency
associated with a voltage that appears
in a so-called Josephson junction in a
superconducting circuit.
There are, however, applications that
tax the capacity of modern clocks. Radio astronomy is a good example. Astronomers often use several telescopes
spaced thousands of kilometers apart
LASER
of technical diculties, none of the optical frequency clocks constructed so

far, however, has operated over extended periods.
MICROWAVE
CAVITY
MICROWAVES
LASER
LASER
4
LASER
DETECTOR
LASER
ATOMIC FOUNTAIN uses atoms that have been cooled and trapped by six
laser beams (1). The vertical beams then briey impart an upward velocity
to the atoms. The atoms rise, passing through the microwave cavity on the
way up (2) and again on the way down (3). The rest of the process resembles optically pumped atomic-beam standards: the atoms pass through another laser beam (4 ), and their uorescence is recorded by a photodetector
(5 ). Servo loops and frequency dividers (not shown) generate timing pulses.
64
PROBE
LASER
INTERNATIONAL
DATELINE
MERIDIAN
OF GREENWICH
to study a stellar object, a technique

that dramatically increases the resolution [see Radio Astronomy by VeryLong-Baseline Interferometry, by Anthony C. S. Redhead; SCIENTIFIC AMERICAN, June 1982]. Two radio telescopes
spaced 10,000 kilometers apart have an
eective angular resolution more than
one million times better than either telescope alone. But to combine the data
from each telescope appropriately, investigators need to know precisely
when each telescope received the signal.
Present-day hydrogen masers have the
stability required for such observations. More stable clocks may be needed for space-borne radio telescopes.
Highly stable clocks are essential for
the best tests of relativity. Timing measurements of millisecond pulsars, some
of which are as stable as the best atomic
clocks, oer evidence for gravity waves.
In 1978 Joseph H. Taylor, Jr., and his associates at Princeton University found
that the period of a binary-pulsar system has been slowly varying by just the
amount that would be expected for the
loss of energy by gravitational radiation, as predicted by general relativity. Greater precision can be achieved
if measurements are taken over many
years, so clocks with better long-term
stability would be useful.
In other tests of relativity, Vessot
and his colleagues conrmed the predicted increase in clock rates at high altitudes. They sent on board a rocket a
hydrogen maser and measured the
small, relativistic clock shift to within
an accuracy of 0.007 percent at an altitude of 10,000 kilometers. Highly stable clocks have also been used by Irwin
I. Shapiro, now at the Harvard-Smithsonian Center for Astrophysics, to observe the relativistic delay of a light signal passing by the sun.
Ultraprecise timekeeping has more
practical applications as wellmost
notably, for navigation. The location of
Voyager 2 as it sped by Neptune was
determined by its distance from each
of three widely separated radar telescopes. Each of these distances in turn
was obtained from accurate measurements of the eight hours it took for
light to travel from each telescope to
the spacecraft and return.
Navigation is, of course, also important on the earth. One of the latest applications of precise clocks is the satellite-based assemblage called the Global
Positioning System, or GPS. This system
relies on atomic clocks on board orbiting satellites. The GPS enables anyone
with a suitable radio receiver and computer to determine his or her position to
approximately 10 meters and the correct time to better than 10 7 second.
10
6 5
4 3
+1
+2
+3
+4
+5
+6
+7
+8
+9 +10 +11
+12
12
Coordinating Time Scales
n the article, we discuss the measurement of an interval of time, such as

a second or a minute. This process requires only a good clock. But to be
able to state that an event happened at a particular time, say, 22 seconds after 12:31 P.M. on July 5, 1993, requires synchronization with a clock that is,
by mutual agreement, the standard. The worlds standard clock exists on
paper as an average of the best clocks in the world. The International Bureau
of Weights and Measures in Svres, France, is responsible for coordinating
international time. This coordinated time scale is called International Atomic
Time, or TAI.
Many users require a time scale that keeps pace with the rotation of the
earth. That is, averaged over a year, the sun should be at its zenith in Greenwich, England, at noon. The day as determined by the apparent position of
the sun is irregular but on the average longer than the 24 hours as defined
by TAI. To compensate, another time scale, called Coordinated Universal
Time, or UTC, is specified by occasionally adding or subtracting a whole
number of seconds from TAI. These seconds, or leap seconds, are inserted
or deleted, usually on December 31 or June 30, to keep UTC within 0.9 second of the time as defined by the rotation of the earth. The record of leap
seconds must be consulted to determine the exact interval between two
stated times.
Two observers monitoring the same

satellite can synchronize their clocks to
within a few nanoseconds.
It is expected that the GPS will have
widespread practical applications, such
as pinpointing the positions of ships,
airplanes and even private automobiles.
The GPS was used during the 1991 Persian Gulf War to enable troops to determine their positions on the desert.
Commercial receivers can be purchased
for less than $1,000, although these
civilian versions are limited to an accu-
racy of about 100 meters because of deliberate scrambling of the signals transmitted from the satellites. A full complement of 24 satellites would give 24hour, worldwide coverage. The system
is nearly complete.
These and other applications show
the importance of time and frequency standards. The anticipated improvements in standards will increase the effectiveness of the current uses and open
the way for new functions. Only time
will tell what these uses will be.
FURTHER READING
pages 301320; September/October 1983.
FROM SUNDIALS TO ATOMIC CLOCKS: UNPRECISE MEASUREMENT OF TIME. N. F. RamDERSTANDING TIME AND FREQUENCY. J.
sey in American Scientist, Vol. 76, No. 1,
Jespersen and J. Fitz-Randolph. Dover,
pages 4249; January/February 1988.
1982.
TIME AND FREQUENCY. Special issue of ProHISTORY OF ATOMIC CLOCKS. N. F. Ramceedings of the IEEE, Vol. 79, No. 7; July
sey in Journal of Research of the Nation1991.
al Bureau of Standards, Vol. 88, No. 5,
65
Surgical Treatment
of Cardiac Arrhythmias
To save the life of a doomed patient, the author and his colleagues
developed a now standard surgical procedure for correcting
lethally fast heartbeats in many people susceptible to them
by Alden H. Harken
n 1978 a vice president of a bank

in Philadelphia collapsed at work
when his heart began to beat dangerously fast. Fortunately, his co-workers were able to administer cardiopulmonary resuscitation immediately, keeping him alive until emergency medical
workers arrived. He was soon brought to
the Hospital of the University of Pennsylvania, where I was a junior member
of the surgical faculty.
Little did either of us know that within weeks of this episode we would participate together in making a small piece
of surgical history. Desperate to prevent the bankers imminent death, my
colleagues and I devised a new surgical
treatment to correct the underlying disturbance that caused his heart to malfunction. Since then, hundreds of other
patients have been aided by this therapy. At the same time, further research
has expanded insight into why our treatment strategy, born of necessity, proved
so useful.
I well remember our initial evaluation
of the bankers medical condition because we were in for a surprise. When he
rst appeared at the hospital, we suspected he had suered a heart attack
(myocardial infarction): the death of cardiac muscle after blockage of an artery
feeding that tissue. But tests told a different story. Indeed, the muscle was in
ALDEN H. HARKEN, a practicing cardiac surgeon, is professor and chairman

of the department of surgery at the University of Colorado Health Sciences Center in Denver. He earned his M.D. from
Case Western Reserve School of Medicine
in 1967. After completing his surgical
residency at Peter Bent Brigham Hospital
and Childrens Hospital, both in Boston,
he joined the Hospital of the University of Pennsylvania in 1976. Harken has
held his current posts since 1984.
68
good shape, except for a small area that

had been damaged during a heart attack several years before.
His heart had malfunctioned now because it became suddenly and lethally
unstable electrically. The electrical wiring system that regulates the heartbeat
induces the cardiac muscle to contract
and thus push blood into the arterial circulation some 72 times a minute.
The mans muscle had begun to receive
much more frequent signals, leading to
abnormally fast pumping. If the heart
beats too rapidly, its interior chambers
do not have time to ll with blood. Because the organ cannot eject something
it does not receive, delivery of blood to
the bodys tissues, including to the cardiac muscle itself, can drop precipitously, causing the heart to stop.
Although we had originally expected
to nd evidence of a new heart attack,
we were also aware that the bankers
electrical derangement was not unique.
Six years earlier Hein J. J. Wellens, then
at the University of Limburg in the Netherlands, observed that excessively fast
pumping occurred in certain patients
months or years after a heart attack.
We understood as well that medications designed to prevent arrhythmias,
or abnormal heartbeats, could restore
proper functioning in some people, and
so we tried every type available. Each
failed. In a span of three weeks at the
hospital, the banker seriously extended
his metaphysical line of credit, suering three additional cardiac arrests. To
let him leave under those conditions
would most assuredly have been fatal
and he knew it.
At the time, I was privileged to be
working with Mark E. Josephson and
Leonard N. Horowitz, who specialized
in diagnosing cardiac electrical abnormalities. They concluded that the bankers trouble stemmed from a disturbance
known as a reentrant electrical circuit
in the heart. That being the case, we

thought we might be able to interrupt
the circuit surgically.
o follow our logic, it helps to know

a bit about how the hearts electrical system controls cardiac activity. The heart, which is divided into
four chambers, is essentially a ball of
muscle (myocardium) lined by conduction tissue: unique bers that form a
kind of internal nervous system. These
special bers convey electrical impuls-
LIFESAVING OPERATION involves excising ap of diseased muscle (lined area

in image at right), about three square
centimeters in area and several millimeters thick, from the inner surface of a patients heart. When successful, the surgery halts propagation of impulses through
a pathway known as a reentrant circuit,
which may arise months or years after
a heart attack and can fatally disturb
normal cardiac rhythms. The surgeon
has entered the left ventricle through
an incision (broken line in inset) in dead
scar tissue (shaded area in inset) left by
the heart attack . Clamps hold back the
edges of the incision.
es swiftly to the entire cardiac muscle.

In response to the impulses, the muscle contractsrst at the top of the
heart and slightly thereafter at the bottom. As contraction begins, oxygen-depleted, venous blood is squeezed out of
the right atrium (one of two small upper chambers) and into the larger right
ventricle below. Then the ventricle ejects
the blood into the pulmonary circulation, which resupplies oxygen and delivers the blood to the left side of the heart.
In parallel with the events on the right,
the muscle pumps newly oxygenated
blood from the left atrium into the left
ventricle and, from there, out to the aorta, which distributes it to every part of
the body.
The signal giving rise to these machinations emanates from a cluster of conduction tissue cells collectively known
as the sinoatrial node. This node, located at the top of the right atrium, establishes the tempo of the heartbeat; hence,
it is often referred to as the cardiac pacemaker. It sets the tempo simply because
it issues impulses more frequently than
do other cardiac regions, once about every 830 milliseconds. If something provoked another part of the heart to re
at a faster rate, as occurred in the banker, it would become the new pacemaker. Although the sinoatrial node can respond to signals from outside the heart,
it usually becomes active spontaneously. In other words, it is on automatic pilot, a capability known as automaticity.
Such automaticity stems from the
unique leakiness of the membrane encasing nodal cells. As is true of the membrane surrounding muscle cells and neurons, the nodal cell membrane is studded with pumps that transport ions into
and out of the cell. The net result of this
exchange is the creation of an electrical
potential, or unequal charge distribution, across the membrane. Yet unlike
muscle and nerve cells, which maintain
their resting potential until they are
jogged by an outside stimulus, nodal
cells allow certain ions to leak back out
of the cells. This outow reduces the
membrane potential to a critical value.
At that point, the membrane permits a ood of other ions to rush back
into the cells. This onslaught momentarily depolarizes the cells (eliminates
the membrane potential) and actually
reverses the membrane polarity. Such
depolarization constitutes an impulse.
After the impulse is generated, cells
SCAR TISSUE
INCISION
CLAMP
AREA
BEING
EXCISED
CLAMP
SURGICAL
SCISSORS
repolarize and prepare for ring anew.

Impulses born at a cell in the sinoatrial node typically speed instantly through
the rest of the node; from there, they
course through the entire heart in the
span of 160 to 200 milliseconds. Traveling along conduction bers, they rst
race across both atria and then regroup
at the atrioventricular node, a cellular
cluster centrally located atop the ven-
tricles. After a pause, they course down

the ventricles along a conduction cable
that divides into two branches known
as conduction bundles; these further
ramify to form arbors of thinner projections called Purkinje bers. One arborized bundle serves each ventricle,
sending signals rst along the surface
of the septum (a wall dividing the two
ventricles) to the tip of the heart (the
BLOOD FLOW IN HEART

PULMONARY
SUPERIOR
ARTERY
VENA
AORTA
CAVA
apex) and, from there, up along the inner surface of the external (lateral) walls
to the top of the ventricle.
As impulses from the conduction bers reach muscle, they activate the overlying cells. Muscle cells, too, are capable
of relaying impulses, albeit more slowly than do conduction bers. The cells
of the endocardium (the inner surface
of the wall ) depolarize rst and relay
ELECTRICAL CONDUCTION IN HEART
AORTA
LEFT
PULMONARY
VEINS
SINOATRIAL
NODE
LUNG
INFERIOR
VENA
CAVA
FT
LE RIUM
AT
HT
LVE
VA
M
RIU
AT
SE
ATRIOVENTRICULAR
NODE
E
CL
FT
LE NTRI
E
V
LV
VA
specialized electrical conduction system (green in large heart) normally regulates the steady beating of the
heart. The impulses (black arrows in
image at right) that induce pumping are
issued at set intervals from the sinoatrial node (large green oval at top left), or
the cardiac pacemaker. From there,
they race to the atrioventricular node
(above the ventricles) and, after a brief
pause, speed down along the septum
to the bottom of the heart and up its
sides. Meanwhile the impulses also migrate from the conduction fibers across
the overlying muscle, from the endocardium to the epicardium, thereby triggering the contractions that force blood (arrows in small diagram above) through
the heart and into the arterial circulation. The spread of electricity through a
healthy heart gives rise to the familiar
electrocardiogram at the bottom right.
The P wave (purple) and QRS wave (red )
form as impulses pass through the atria
and ventricles, respectively; the T wave
(black) arises as cardiac cells, which cannot be stimulated for a while after they
fire, recover their excitability.
70
UM
PT
RIG
The Making of a Heartbeat
LATERAL
WALL
HT CLE
RIG NTRI
VE
CONDUCTION
BUNDLE
ENDOCARDIUM
EPICARDIUM
PURKINJE
FIBER
APEX
Q
S
the impulses through the thickness of

the muscle to the outer surface (the epicardium). Depolarization, in turn, triggers contraction.
osephson and Horowitz suggested

that diseased cells had distorted
this normal ow of electricity in
the bankers heart. After a heart attack, many cells surrounding the resulting scar (the group of cells killed by lack
of blood delivery) continue to live but
are abnormal electrically; they may conduct impulses unusually slowly or re
when they would typically be silent.
These diseased areas, my co-workers
indicated, might perturb smooth signaling by forming a reentrant circuit in
the muscle: a pathway of electrical conduction through which impulses can cycle repeatedly without dying out. In our
patients case, the circuit was thought to
be in the left ventricle, where his heart
attack, in common with most others, occurred. (Activation of reentrant circuits
some time after a heart attack is now
believed to take place in a sizable number, perhaps 10 percent, of the roughly
1.2 million Americans who suer heart
attacks every year.)
Passage of impulses through a reentrant circuit can be envisioned by imagining a wave of impulses encountering, say, the bottom of an oval scar in
the left ventricle. On reaching the scar,
the wave would split in two, to detour
around both sides of the dead area. If
diseased cells somehow interrupted impulses propagating along one of those
branches, impulses might still ow up
along the opposite branch and over the
top of the oval. Then they might traverse
the previously blocked path and return
to the beginning of the circuita region
we call the origin.
If this circuit were negotiated slowly
enough, the origin would have repolarized and become responsive once again
to stimulation. (Between the time cells
depolarize and repolarize, they are generally refractory, or incapable of responding to new impulses.) In that case,
the impulses could reexcite the origin,
sending impulses back into the diseased circuit and also out to the rest of
the ventricular muscle. Despite the slow
conduction, the impulses could complete the circuit in a shorter time than
the interval between normal heartbeats.
Hence, persistent cycling could enable
the origin of the circuit to become the
new pacemaker and to provoke sustained ventricular tachycardia: excessively rapid pumping by the ventricles.
We knew that continuous passage
through reentrant circuits could occur
in humans because Wellens had established that fact in the 1970s. Fortunately
for us, he also introduced a procedure

for determining whether a quiescent circuit lurks in a patient who survives a lifethreatening episode of tachycardia and
whether any existing drugs can prevent
renewed activation of the pathway. A
physician threads an electrode known
as a pacing catheter into the heart and
issues a series of specically timed impulses. Initiation of sustained premature
heartbeats conrms that a patient harbors a reentrant pathway. ( In contrast,
impulses delivered to a healthy heart
would yield only single contractions that
would not be repeated.) Next, the individual is given an antiarrhythmic drug.
If paced stimuli now fail to trigger sustained tachycardia, the nding implies
the drug should be helpful.
When Josephson and Horowitz performed the procedure on the banker,
they found they could indeed induce
persistent tachycardia and that, sadly,
no antiarrhythmic medications could
aid him. I met with the two of them
soon afterward in their tiny, windowless
catheterization laboratory. Knowing our
patient carried a life-threatening electrical pathway inside his heart, we began wondering if we might prevent its
activation by surgically removing all or
part of the culprit circuit, especially the
origin. We realized the plan could fail,
or that by removing the tissue, we might
actually create other problems. But we
were out of options.
efore proceeding, we had to develop a way to locate the renegade pacemaker. We hoped we
might nd it by analyzing signals reaching an electrode placed directly on the
inner or outer surface of the heart [see
bottom illustration on next page]. More
specically, we planned to induce sustained tachycardia with a pacing electrode. During each heartbeat, we would
measure electric currents produced at
a single site (consisting of a small cluster of cells) along the diseased border
of the heart attack scar. We would start
at a position arbitrarily designated as
12 oclock and proceed around the
clock face back to the beginning.
We would delineate the circuit by
comparing the time of electrical activation in each region against that seen in
healthy tissue. Regions that generated
currents before the healthy tissue did
would be revealed as belonging to the
circuit; the area that became excited earliest would be the pacemaker. We could
not rely on standard electrocardiography for this purpose because it lacked
the specicity we needed. Familiar electrocardiogram tracings, made by attaching electrodes to the skin, reect the
summed activity of many thousands of
cells in the heart; they cannot identify

the precise swatch of muscle that is depolarized at any given moment.
Our general approach made sense, but
no one had ever attempted to map the
ow of signals in the living, pumping
chambers of the human heart by recording directly from the organs surface.
We had no idea whether we could obtain decipherable results. The next day
I was scheduled to remove a cancerous
lung from a dierent patient. He kindly
agreed to let us try to detect signals
directly from the outside of his heart.
To our delight, we could clearly discern
when a wave of impulses crossed any
point on the muscle.
I was now ready to discuss our proposed strategy with the banker. Not
knowing whether the origin of the circuitthe zone of earliest activation
was closer to the inside or outside of
the cardiac muscle, we intended to map
both the inner and outer surfaces. We
planned to reach the interior by opening
the heart through the existing scar. (Cutting into healthy tissue would, after all,
destroy new tissue unnecessarily.) If we
found the troublesome region, we proposed to remove it surgically. To keep
blood moving through the patients body
during the operation, we should have
to attach him to a heart-lung machine.
This device diverts unoxygenated blood
into an articial lung. Blood containing
oxygen is then pumped back into the
arterial circulation via the aorta.
People often call physicians courageous, but it was our patient who was
brave. After I described our therapeutic
strategy in great detail, he posed the
dreaded question: How many times
have you done this before? I told him,
Never. Then he asked how many times
anyone had performed the operation
previously. I informed him it was untried. Despite these unsettling answers,
he gave me a condent smile and said,
Go ahead.
The next morning we were able to
pinpoint and excise the region of earliest activity, which turned out to reside
on the inside surface. (Today we know
that virtually all reentrant pathways
weave through cells in or close to the
endocardium.) Our patient not only resumed banking but also went on to become the county tax assessor. I lost
track of him a few years ago, but as of
a decade after our treatment, he had suffered no further arrhythmias.
Not everyone who has the surgery is
as lucky as the banker was, however. Of
all the patients who undergo the procedure after surviving an episode of persistent tachycardia, approximately 9 percent succumb either during the operation or within a month after it. On the
71
DISEASED
CELLS
DEAD
CELLS
2
IM
S
UL
ES
PASSING AROUN
DP
AT
C
n addition to assembling survival

statistics, we have discovered since
1978 that reentrant pathways need
not be as large as we originally thought.
Those occurring at a microscopic level
can be equally pernicious. In fact, microanatomic reentrant circuits seem to
be the most common form of all.
The notion that microcircuits could
exist was rst suggested in the early
1970s by another surgeon: James L. Cox,
then at Duke University. He argued that
a small bit of mottled tissue, consisting
of diseased cells interspersed with islands of dead cells, could set up the
conditions needed to establish reentrant
tachycardia. In such a microscopic circuit, impulses that encounter a divided
pathway at an entryway to a mottled
patch would split and travel along both
routes [see top illustration on this page].
As is true of larger, macro reentrant
circuits, impulses propagating along
one branch would encounter a one-way
blockade. At the same time, impulses
owing along the other branch would
meander through a maze of diseased
cells and return along the previously
blocked lane.
If conduction through the diseased
tissue were suciently slow, the impulses would come back to the entryway, or
origin of the circuit, after that site was
no longer refractory. Excitation of the
site would then stimulate the ventricular muscle to contract and, at the same
time, would send the impulses back into
the microcircuit again and again. Instead of traveling along the circumference of a scar, then, a reentrant circuit
could trace a recursive path through a
more localized maze of cells in the diseased boundary between a heart attack
scar and fully healthy tissue.
Two of my colleagues, Glenn J. R.
Whitman and Michael A. Grosso, decided to test this idea in the early 1980s.
They were able to create small heterogeneous zones consisting of mixed dead
and living but diseased cells in the ventricles of test animals. These animals,
not previously susceptible to the electrical induction of self-sustaining tachycardia, became highly prone to it.
Whitman and Grosso assumed that if
the mottled tissue were at fault, killing
all the cells in the patch should restore
appropriate electrical activity in the
heart. Instead of wandering through a
dangerous maze, impulses encounter72
BLOCKADE
ENTRYWAY
WAVE
OF IMPULSES TO REENTRANT CIRCUIT
REENTRANT CIRCUIT, a closed loop through which impulses can cycle repeatedly,
has formed in a patch of cardiac muscle populated by diseased tissue (beige) and
islands of dead cells (brown). Continuous electrical activity in such a circuit can
usurp the normal pacemaker function of the sinoatrial node, leading to ventricular
tachycardia (a dangerously fast heartbeat). Persistent cycling begins when a wave
of impulses (blue arrows in 1 ) enters the patch and divides (2 ) at the entryway. If
ing the homogeneous patch of killed

tissue would either be extinguished or
zoom around it through adjacent healthy
cells. Sure enough, when the mottled
patches were destroyed, the predisposition to arrhythmia vanished.
These ndings revealed that mottling could set the stage for reentrant
tachycardia. They also provided the
hindsight needed to explain why a dif-
ferent surgical treatment tested by us

and others in various patients had not
worked well. Believing that the scar itself was somehow responsible for the
electrical disturbances, we had previously removed only the dead tissue. Whitman and Grossos work indicated that
this approach was doomed to failure because it left the true culpritthe zone
of mixed living and dead cellsin place.
other hand, 80 percent of surgically

treated patients live for at least a year
without recurrence of tachycardia, and
60 percent survive for ve years or more.
The candidates most likely to do well
are those whose heart muscle is damaged least.
RECEDING
IMPULSES
IMPULSES
REENTERING CIRCUIT
signals owing along one pathway (two-headed black arrow) encounter a temporary blockade (vertical lines), those propagating along a second pathway (thin blue
arrow) may return to the entryway (3 ), or origin, through the previously blocked
alley. If the entryway has regained excitability (4 ), the returning impulses will reexcite that area. They may thus fan out through healthy muscle (large green arrows)
and proceed back through the worrisome circuit again and again (thin green arrow).
Yet we still faced two signicant puzzles, one scientic and one clinical. Why
is it that reentrant circuits do not become active every time the heart beats
in susceptible patients? In other words,
why can people often survive for months
or years before deadly disturbances of
rhythm arise? We also wondered how
we might noninvasively identify patients at risk for reentrant tachycardia
RECORDINGS OF ELECTRICAL ACTIVITY

1
2
3
4
5
6
7
8
before they experienced a potentially

life-threatening episode.
The simplistic explanation for why a
reentrant circuit does not jump into
action with each heartbeat seemed to
be that impulses red by the sinoatrial
node cannot cycle repeatedly through
the troublesome pathway. At the end of
the rst cycle, they return to a still refractory starting site. Blocked from reentering the circuit, they go no further.
Unfortunately, this explanation did not
clarify how persistent cycling does arise.
We now think it is triggered when, in a
case of exquisite bad luck, an electrically irritable cell lying adjacent to a reentrant pathway res spontaneously in
a narrow window of time between one
activation of the sinoatrial and atrioventricular nodes and the next.
We came to this conclusion after reviewing research reported in the late
1970s by our colleagues E. Neil Moore
and Joseph F. Spear of the Hospital of
the University of Pennsylvania. By impaling cells on tiny, needlelike electrodes,
Moore and Spear were able to track
changes in the membrane potentials of
single, diseased cardiac cells taken from
the area surrounding heart attack scars.
After healthy cells depolarize, they repolarize smoothly. In the diseased cells,
by contrast, the membrane potential
uctuated markedly during the repolarization period.

We presumed that these uctuations
would sometimes progress to premature depolarization, or ring of an impulse. If an irritable cell happened to lie
next to a reentrant pathway, it might
well insert an impulse into the worrisome channel during the interval between normal heartbeats.
This insertion might activate a reentrant circuit, whereas an impulse originating at the sinoatrial node would not,
because recent passage of an impulse
through a pathway can alter the electrochemical characteristics of that pathway
and slow conduction of a subsequent
signal. Thus, the impulse delivered by
the irritable cell could pass through the
circuit more slowly than would a prior
signal originating at the sinoatrial node.
If delivery of the wayward impulse were
timed properly, the impulse propagating through the circuit would return to
the entryway at a most devastating moment: after the site regained excitability
(and so could relay the impulse onward)
but before the sinoatrial node red for
a second time (thereby gaining control
of the heartbeat). Hitting a receptive
target, the impulse might proceed to
run many unimpeded laps around the
lethal circuit.
ur second problemreadily
identifying patients at risk for
reentrant tachycardiawas resolved masterfully by our co-worker
Michael B. Simson, a person of many
talents. Aside from being a superb cardiologist, he is, as I sometimes say, an
enthusiastic sports-car hack and computer driver. Steering his beat-up sports
car home one night after sitting in on
one of our surgical research meetings,
he began to ponder the electrical noise,
or seemingly random signals, emanating from the hood of his car. If he simply monitored the currents reaching the
hood, he reasoned, the resulting data
would be indecipherably chaotic. But if
he wanted to track the electrical impulses coming specically from his distributor, he might well discern them by
signal averaging.
In this procedure, he would record
the voltage and direction (the electrical
vector) of currents owing toward and
9
10
11
12
TIME (MILLISECONDS)
CARDIAC MAPPINGperformed by moving an electrode (blue) over diseased tissue surrounding a heart attack scarenables a surgeon to identify a reentrant circuit. In one patient, recordings of electrical activity (tracings at right) made at 12
sites (inset at left) around the inner surface during induced ventricular tachycardia
delineated a large, recurring electrical circuit that began at the one oclock position
(tracing 1) and progressed clockwise around the circumference of the scar. A surgeon can block impulse propagation in such a circuit by excising the region of earliest excitation, which can be assumed to be the entryway to the circuit.
73
160
BEFORE SURGERY
AFTER SURGERY
MICROVOLTS
120
VENTRICULAR
EXCITATION
80
ELECTRICAL ACTIVITY
DURING NORMALLY
QUIET PERIOD
40
0
0
100
200
TIME (MILLISECONDS)
300
100
200
TIME (MILLISECONDS)
300
SPECIAL ELECTROCARDIOGRAMS made by a technique called signal averaging

were produced before (left) and after (right) a patient underwent surgery to prevent the continuous passage of impulses through a reentrant pathway. The jagged
red line at the left reects slow, self-limited electrical conduction through the worrisome pathway during the normally quiet period immediately following ventricular excitation. Flattening of the line after surgery means slow conduction through
the potentially deadly circuit has been eliminated. Signal-averaged electrocardiograms can now identify patients at risk for reentrant arrhythmias.
away from the hood during particular

phases of rotation by his distributor rotor. If he summed the signals obtained
by repeated measurements in a given
phase, random currents would tend to
cancel one another out, leaving a record
of only those produced by the rotor. Dividing the result by the number of readings made in a selected phase would give
him a measure of the current generated
by the distributor in that phase.
It then occurred to Simson that he
might apply much the same approach
to screen heart attack victims for susceptibility to reentrant tachycardia. Perhaps signal averaging would enable him
to detect very slow electrical activity persisting after the normal ow of signals
passed through the ventricles. Most of
the extra activity he found would reect
impulses propagating belatedly through
a potentially dangerous reentrant channel. Put another way, Simson thought
he could place electrodes on the skin,
as for a standard electrocardiogram,
but then record only those currents
produced in the 40 milliseconds immediately after formation of the familiar
QRS wave seen on electrocardiograms.
( The QRS wave reects the spread of
impulses through the ventricles.) Heart
cells are generally quiet at that point,
giving rise to a at line on the electrocardiogram tracing. Signal-averaged deviations from this normal pattern would
signify slow conduction in a reentrant
pathway.
Simson spent that night in his basement building a signal-averaging device.
The next day Josephson, Horowitz and
I were scheduled to remove tissue that
74
had earlier caused reentrant arrhythmia in one of our patients. Before surgery, Simson attached his new recorder
to the patient and noted, as expected,
that there was a urry of electrical activity in the usually quiescent span following ventricular excitation. But was
the signal, in fact, an indication of late
impulse conduction in a reentrant circuit? The answer would be yes if the
uctuations disappeared after the operation. The surgical procedure went
well. Josephson and Horowitz identied the circuit, and I excised the entryway. After surgery, Simson reattached
his device to the patient. The post-QRS
uctuations were gone.
We had come a long way since 1978.
We had learned why our surgical approach, initially designed by guesswork,
is useful. It interrupts the diseased anatomic pathway that, in response to aberrant ring by a nearby cell, gives rise to
the repeated ow of impulses through
a recursive circuit. Moreover, we had
gained the ability to identify noninvasively patients at risk.
t the University of Colorado,

where I moved in 1984, we use
Simsons screening test routinely. We usually wait two or three months
after a heart attack to be sure we are not
detecting a predisposition to automatic tachycardias. For a week or so after a
person has a heart attack, dying cells often re when they should be silent. This
behavior can cause the heart to beat prematurely. If the cell depolarizes repeatedly, the activity could lead to fast beating, and sometimes failure, of the heart.
A tendency to automatic tachycardia

generally resolves within a few weeks,
as the sputtering cells expire.
If a propensity for reentrant tachycardia is discovered after a suitable waiting
period, and if medications do not suffice, patients can consider other treatment options. I speak of more than one
choice because surgery is no longer the
only therapeutic alternative to drugs. A
device known as an implantable debrillator has been available since 1980.
When the heart begins to beat quickly,
the machine issues a shock that depolarizes the entire heart instantly, giving
the sinoatrial node a chance to resume
its pacemaker function.
About half as many patients die from
complications of the implantation procedure for the device as from consequences of undergoing our surgery. But,
in contrast to the surgery, the device offers only palliation, not a cure. Recipients continue to face episodes of tachycardia and may lose consciousness each
time they are shocked back into normal rhythm. Consequently, they cannot
drive or engage in other activities where
sudden blackouts could be dangerous.
If surgery to eliminate a reentrant circuit is deemed the better therapy for a
given patient, it can now be obtained at
many medical centers.
Overall, it is fair to say that the majority of patients who survive a heart
attack are not vulnerable to reentrant
arrhythmias. Perhaps half of the small
group who are susceptible can be treated with medication. Of those who do
not respond to drugs, however, as many
as 80 percent are likely to die from their
electrical abnormality within a year after their rst bout of reentrant tachycardia unless they receive some other
therapy. It is reassuring to know that
for many of those individuals the courage of a Philadelphia banker has permitted a cure.
FURTHER READING
OBSERVATIONS ON MECHANISMS OF VENTRICULAR TACHYCARDIA IN MAN. H.J.J.
Wellens, D. R. Duren and K. I. Lie in Circulation, Vol. 54, No. 2, pages 237244;
August 1976.
SURGICAL ENDOCARDIAL RESECTION FOR
THE TREATMENT OF MALIGNANT VENTRICULAR TACHYCARDIA. A. H. Harken,
M. E. Josephson and L. N. Horowitz in
Annals of Surgery, Vol. 190, No. 4, pages 456460; October 1979.
CARDIAC ARRHYTHMIAS. A. H. Harken
in Care of the Surgical Patient, Vol. 1:
Critical Care. Edited by D. W. Wilmore,
M. F. Brennan, A. H. Harken, J. W. Holcroft and J. L. Meakins. Scientic American Medicine, 1992.
Fuzzy Logic
The binary logic of modern computers often falls short
when describing the vagueness of the real world.
Fuzzy logic oers more graceful alternatives
by Bart Kosko and Satoru Isaka
omputers do not reason as brains

do. Computers reason when
they manipulate precise facts
that have been reduced to strings of zeros and ones and statements that are
either true or false. The human brain
can reason with vague assertions or
claims that involve uncertainties or value judgments: The air is cool, or That
speed is fast or She is young. Unlike
computers, humans have common sense
that enables them to reason in a world
where things are only partially true.
Fuzzy logic is a branch of machine
intelligence that helps computers paint
gray, commonsense pictures of an uncertain world. Logicians in the 1920s
rst broached its key concept: everything is a matter of degree.
Fuzzy logic manipulates such vague
concepts as warm or still dirty and
so helps engineers to build air conditioners, washing machines and other devices that judge how fast they should
operate or shift from one setting to another even when the criteria for making those changes are hard to dene.
When mathematicians lack specic al-
BART KOSKO and SATORU ISAKA are

pioneers in the development of fuzzy
logic systems. Kosko holds degrees in
philosophy and economics from the University of Southern California, a masters
in applied mathematics from the University of California, San Diego, and a Ph.D.
in electrical engineering from the University of California, Irvine. He is an assistant professor of electrical engineering
at U.S.C., a governor of the International Neural Network Society and the program co-chair of the 1993 World Congress on Neural Networks. Isaka specializes in fuzzy information processing in
the research and development division
at Omron Advanced Systems in Santa
Clara, Calif. He is interested in applying
machine learning and adaptive control
systems to biomedical systems and factory automation. He received his M.S. and
Ph.D. degrees in systems science from
U.C.S.D., in 1986 and 1989, respectively.
76
gorithms that dictate how a system

should respond to inputs, fuzzy logic
can control or describe the system by
using commonsense rules that refer
to indenite quantities. No known mathematical model can back up a truckand-trailer rig from a parking lot to a
loading dock when the vehicle starts
from a random spot. Both humans and
fuzzy systems can perform this nonlinear guidance task by using practical but
imprecise rules such as If the trailer
turns a little to the left, then turn it a
little to the right. Fuzzy systems often
glean their rules from experts. When no
expert gives the rules, adaptive fuzzy
systems learn the rules by observing
how people regulate real systems.
A recent wave of commercial fuzzy
products, most of them from Japan, has
popularized fuzzy logic. In 1980 the
contracting rm of F. L. Smidth & Company in Copenhagen rst used a fuzzy
system to oversee the operation of a
cement kiln. In 1988 Hitachi turned over
control of a subway in Sendai, Japan,
to a fuzzy system. Since then, Japanese
companies have used fuzzy logic to direct hundreds of household appliances
and electronics products. The Ministry
of International Trade and Industry estimates that in 1992 Japan produced
about $2 billion worth of fuzzy products. U.S. and European companies still
lag far behind.
Applications for fuzzy logic extend
beyond control systems. Recent theorems show that in principle fuzzy logic
can be used to model any continuous
system, be it based in engineering or
physics or biology or economics. Investigators in many elds may nd that
fuzzy, commonsense models are more
useful or accurate than are standard
mathematical ones.
t the heart of the dierence between classical and fuzzy logic is

something Aristotle called the
law of the excluded middle. In standard
set theory, an object either does or does
not belong to a set. There is no middle
ground: the number ve belongs fully

to the set of odd numbers and not at
all to the set of even numbers. In such
bivalent sets, an object cannot belong
to both a set and its complement set or
to neither of the sets. This principle preserves the structure of logic and avoids
the contradiction of an object that both
is and is not a thing at the same time.
Sets that are fuzzy, or multivalent,
break the law of the excluded middle
to some degree. Items belong only partially to a fuzzy set. They may also belong to more than one set. Even to just
one individual, the air may feel cool,
just right and warm to varying degrees.
Whereas the boundaries of standard
sets are exact, those of fuzzy sets are
curved or taper o, and this curvature
creates partial contradictions. The air
can be 20 percent cooland at the
same time, 80 percent not cool.
Fuzzy degrees are not the same as
probability percentages, a point that has
eluded some critics of the eld. Probabilities measure whether something will
occur or not. Fuzziness measures the degree to which something occurs or some
condition exists. The statement There
is a 30 percent chance the weather will
be cool conveys the probability of cool
weather. But The morning feels 30 percent cool means that the air feels cool
to some extentand at the same time,
just right and warm to varying extents.
The only constraint on fuzzy logic is
that an objects degrees of membership
in complementary groups must sum to
unity. If the air seems 20 percent cool,
it must also be 80 percent not cool. In
this way, fuzzy logic just skirts the bivalent contradictionthat something
is 100 percent cool and 100 percent not
coolthat would destroy formal logic.
The law of the excluded middle holds
merely as a special case in fuzzy logic,
namely when an object belongs 100
percent to one group.
The modern study of fuzzy logic and
partial contradictions had its origins early in this century, when Bertrand Russell found the ancient Greek paradox at
the core of modern set theory and logic. According to the old riddle, a Cretan
asserts that all Cretans lie. So, is he lying? If he lies, then he tells the truth
and does not lie. If he does not lie, then
he tells the truth and so lies. Both cases lead to a contradiction because the
statement is both true and false. Russell found the same paradox in set theory. The set of all sets is a set, and so it
is a member of itself. Yet the set of all
apples is not a member of itself because its members are apples and not
sets. Perceiving the underlying contradiction, Russell then asked, Is the set
of all sets that are not members of themselves a member of itself? If it is, it isnt;
if it isnt, it is.
Faced with such a conundrum, classical logic surrenders. But fuzzy logic says
that the answer is half true and half
false, a 5050 divide. Fifty percent of
the Cretans statements are true, and
50 percent are false. The Cretan lies 50
percent of the time and does not lie the
other half. When membership is less
than total, a bivalent system might sim-
plify the problem by rounding it down

to zero or up to 100 percent. Yet 50 percent does not round up or down.
In the 1920s, independent of Russell, the Polish logician Jan /Lukasiewicz
worked out the principles of multivalued logic, in which statements can take
on fractional truth values between the
ones and zeros of binary logic. In a 1937
article in Philosophy of Science, quantum
philosopher Max Black applied multivalued logic to lists, or sets of objects,
and in so doing drew the rst fuzzy set
curves. Following Russells lead, Black
called the sets vague.
Almost 30 years later Lot A. Zadeh,
then chair of the electrical engineering
department at the University of California at Berkeley, published Fuzzy Sets,
a landmark paper that gave the eld its
name. Zadeh applied /Lukasiewiczs logic to every object in a set and worked
out a complete algebra for fuzzy sets.
Even so, fuzzy sets were not put to use
until the mid-1970s, when Ebrahim H.
Mamdani of Queen Mary College in London designed a fuzzy controller for a
BUILDING MACHINES THAT UNDERSTAND the vagueness of

the real world is the goal of those working on fuzzy logic.
steam engine. Since then, the term fuzzy logic has come to mean any mathematical or computer system that reasons with fuzzy sets.
uzzy logic is based on rules of

the form if . . . then that convert
inputs to outputsone fuzzy set
into another. The controller of a cars air
conditioner might include rules such as
If the temperature is cool, then set the
motor speed on slow and If the temperature is just right, then set the motor speed on medium. The temperatures (cool, just right) and the motor
speeds (slow, medium) name fuzzy sets
rather than specic values.
To build a fuzzy system, an engineer
might begin with a set of fuzzy rules
from an expert. An engineer might dene the degrees of membership in various fuzzy input and output sets with
sets of curves. The relation between the
input and output sets could then be
plotted. Given the rule If the air feels
cool, then set the motor to slow, the inputs (temperature) would be listed along
Fuzzy logic can model and control nuances overlooked by

the binary logic of conventional computers.
77
one axis of a graph and the outputs

(motor speed ) along a second axis. The
product of these fuzzy sets forms a
fuzzy patch, an area that represents the
set of all associations that the rule forms
between those inputs and outputs.
The size of the patch reects the rules
vagueness or uncertainty. The more
precise the fuzzy set, the smaller it becomes. If cool is precisely 68 degrees
Fahrenheit, the fuzzy set collapses to a
spike. If both the cool and slow fuzzy
sets are spikes, the rule patch is a point.
The rules of a fuzzy system dene a
set of overlapping patches that relate a
full range of inputs to a full range of
outputs. In that sense, the fuzzy system approximates some mathematical
function or equation of cause and effect. These functions might be laws that
tell a microprocessor how to adjust the
power of an air conditioner or the speed
of a washing machine in response to
some fresh measurement.
Fuzzy systems can approximate any
continuous math function. One of us
(Kosko) proved this uniform convergence theorem by showing that enough
small fuzzy patches can suciently cover the graph of any function or input/
output relation. The theorem also shows
that we can pick in advance the maximum error of the approximation and be
sure there is a nite number of fuzzy
rules that achieve it. A fuzzy system rea-
sons, or infers, based on its rule patches.

Two or more rules convert any incoming number into some result because
the patches overlap. When data trigger
the rules, overlapping patches re in
parallelbut only to some degree.
Imagine a fuzzy air conditioner that
relies on ve rules and thus ve patches
to match temperatures to motor speeds.
The temperature sets (cold, cool, just
right, warm and hot) cover all the possible fuzzy inputs. The motor speed sets
(very slow, slow, medium, fast and very
fast) describe all the fuzzy outputs. A
temperature of, say, 68 degrees F might
be 20 percent cool (80 percent not cool)
and 70 percent just right (30 percent
not just right). At the same time, the air
is also 0 percent cold, warm and hot.
The if cool and if just right rules
would re and invoke both the slow and
medium motor speeds.
The two rules contribute proportionally to the nal motor speed. Because
the temperature was 20 percent cool,
the curve describing the slow engine
speed must shrink to 20 percent of its
height. The medium curve must shrink
to 70 percent. Summing those two reduced curves produces the nal curve
for the fuzzy output set.
In its fuzzy form, such an output
curve does not assist controllers that
act on binary instructions. So the nal
step is a process of defuzzication, in
which the fuzzy output curve is turned

into a single numerical value. The most
common technique is to compute the
center of mass, or centroid, of the area
under the curve. In this instance, the
centroid of the fuzzy output curve
might correspond to a motor speed of
47 revolutions per minute. Thus, beginning with a quantitative temperature
input, the electronic controller can reason from fuzzy temperature and motor speed sets and arrive at an appropriate and precise speed output.
All fuzzy systems reason with this
re-and-sum techniqueor something
close to it. As systems become more
complex, the antecedents of the rules
may include any number of terms conjoined by and or disjoined by or. An
advanced fuzzy air conditioner might
use a rule that says, If the air is cool
and the humidity is high, then set the
motor to medium.
SET THEORY underlies the dierence between standard and fuzzy logic. In standard logic, objects belong to a set fully or not at all (top left ). Objects belong to a
fuzzy set only to some extent (top right ) and to the sets complement to some extent. Those partial memberships must sum to unity (bottom). If 55 degrees is 50
percent cool, it is also 50 percent not cool.
uzzy products use both microprocessors that run fuzzy inference algorithms and sensors that
measure changing input conditions.
Fuzzy chips are microprocessors designed to store and process fuzzy rules.
In 1985 Masaki Togai and Hiroyuki Watanabe, then working at AT&T Bell Laboratories, built the rst digital fuzzy
chip. It processed 16 simple rules in
12.5 microseconds, a rate of 0.08 million fuzzy logical inferences per second.
Togai InfraLogic, Inc., now oers chips
based on Fuzzy Computational Acceleration hardware that processes up to two
million rules per second. Most microprocessor rms currently have fuzzy
chip research projects. Fuzzy products
largely rely on standard microprocessors that engineers have programmed
with a few lines of fuzzy inference code.
Although the market for dedicated fuzzy chips is still tiny, the value of microprocessors that include fuzzy logic already exceeds $1 billion.
The most famous fuzzy application
is the subway car controller used in
Sendai, which has outperformed both
human operators and conventional automated controllers. Conventional controllers start or stop a train by reacting
to position markers that show how far
the vehicle is from a station. Because
the controllers are rigidly programmed,
the ride may be jerky : the automated
controller will apply the same brake
pressure when a train is, say, 100 meters from a station, even if the train is
going uphill or downhill.
In the mid-1980s engineers from Hitachi used fuzzy rules to accelerate,
slow and brake the subway trains more
smoothly than could a deft human operator. The rules encompassed a broad
78
MEMBERSHIP
(PERCENT)
NONFUZZY SET
100
COOL
FUZZY SET
COOL
100
0
50
60
70
AIR TEMPERATURE
50
60
70
AIR TEMPERATURE
FUZZY SET AND ITS COMPLEMENT

NOT COOL
COOL
NOT COOL
MEMBERSHIP
(PERCENT)
100
50
0
50
60
70
AIR TEMPERATURE (DEGREES FAHRENHEIT)
BLAST
OUTPUT (Y): MOTOR SPEED
FUNCTION:
X
Y
IF WARM,
THEN FAST
FAST
MEDIUM
IF JUST RIGHT,
THEN MEDIUM
IF COOL,
THEN SLOW
SLOW
STOP
COLD
JUST
RIGHT
COOL
WARM
HOT
INPUT (X): AIR TEMPERATURE

CENTROID
100
PERCENT
PERCENT
100
70
COOL
JUST
RIGHT
20
70
SUMMED SETS
MEDIUM
20
SLOW
0
50
60
68 70
AIR TEMPERATURE (DEGREES FAHRENHEIT)
80
APPLICATION OF FUZZY LOGIC to the control of an air conditioner shows how manipulating vague sets can yield precise instructions. The air conditioner measures air temperature and then calculates the appropriate motor speed. The
system uses rules that associate fuzzy sets of temperatures,
such as cool, to fuzzy sets of motor outputs, such as slow.
Each rule forms a fuzzy patch. A chain of patches can approx-
range of variables about the ongoing

performance of the train, such as how
frequently and by how much its speed
changed and how close the actual speed
was to the maximum speed. In simulated tests the fuzzy controller beat an automated version on measures of riders
comfort, shortened riding times and
even achieved a 10 percent reduction in
the trains energy consumption. Today
the fuzzy system runs the Sendai subway during peak hours and runs some
Tokyo trains as well. Humans operate
the subway during nonpeak hours to
keep up their skills.
Companies in Japan and Korea are
building an array of fuzzy consumer
goods that oer more precise control
than do conventional ones. Fuzzy wash-
10
47
MOTOR SPEED (REVOLUTIONS PER MINUTE)
imate a performance curve or other function (top). If a temperature of 68 degrees Fahrenheit is 20 percent cool and 70
percent just right (bottom left), two rules re, and the system tries to run its motor at a speed that is 20 percent slow
and 70 percent medium (bottom right ). The system arrives
at an exact motor speed by nding the center of mass, or centroid, for the sum of the motor output curves.
ing machines adjust the wash cycle to

every set of clothes, changing strategies
as the clothes become clean. A fuzzy
washing machine gives a ner wash
than a dumb machine with xed commands. In the simplest of these machines, an optical sensor measures the
murk or clarity of the wash water, and
the controller estimates how long it
would take a stain to dissolve or saturate in the wash water. Some machines
use a load sensor to trigger changes in
the agitation rate or water temperature.
Others shoot bubbles into the wash
to help dissolve dirt and detergent. A
washing machine may use as few as 10
fuzzy rules to determine a wide variety
of washing strategies.
In cameras and camcorders, fuzzy
logic links image data to various lens

settings. One of the rst fuzzy camcorders, the Canon hand-held H800, which
was introduced in 1990, adjusts the autofocus based on 13 fuzzy rules. Sensors measure the clarity of images in six
areas. The rules take up about a kilobyte
of memory and convert the sensor data
to new lens settings.
Matsushita relies on more rules to
cancel the image jitter that a shaking
hand causes in its small Panasonic camcorders. The fuzzy rules infer where the
image will shift. The rules heed local
and global changes in the image and
then compensate for them. In contrast,
camcorder controllers based on mathematical models can compensate for no
more than a few types of image jitter.
79
in a health management system for

large rms. The fuzzy systems use 500
rules to diagnose the health of some
10,000 patients and to draw up personalized plans to help them prevent disease, stay t and reduce stress. Other
companies, including Hitachi and Yamaichi Securities, have built trading programs for bonds or stock funds that
use fuzzy rules to react to changes in
economic data.
he Achilles heel of a fuzzy system is its rules. Almost all the

fuzzy consumer products now
on the market rely on rules supplied by
an expert. Engineers then engage in a
lengthy process of tuning those rules
and the fuzzy sets. To automate this
process, some engineers are building
adaptive fuzzy systems that use neural
networks or other statistical tools to
rene or even form those initial rules.
Neural networks are collections of
neurons and synapses that change
their values in response to inputs from
surrounding neurons and synapses.
The neural net acts like a computer because it maps inputs to outputs. The
neurons and synapses may be silicon
components or equations in software
that simulate their behavior. A neuron
adds up all the incoming signals from
other neurons and then emits its own
response in the form of a number. Signals travel across the synapses, which
have numerical values that weight the
ow of neuronal signals. When new input data re a networks neurons, the
synaptic values can change slightly. A
neural net learns when it changes the
value of its synapses.
Fuzzy Sets and Paradoxes
elations between sets show the paradox at the heart

of fuzzy logic. In standard sets, an object either does
or does not belong to a set: the glass of water is either
empty or full. Fuzzy sets cover a continuum of partial sets,
such as glasses only half full (left).
Two objects or two sets define a two-dimensional space
(right). Standard set theory holds only at the corners of
the square, where the values are either 1 or 0. The midpoint, the fuzziest point in the square, cannot be rounded
off to a corner.
1/2
ONE OBJECT
80
Depending on the available data, networks can learn patterns with or without supervision. A supervised net learns
by trial and error, guided by a teacher.
A human may point out when the network has erredwhen it has emitted a
response that diers from the desired
output. The teacher will correct the responses to sample data until the network responds correctly to every input.
Supervised networks tune the rules
of a fuzzy system as if they were synapses. The user provides the rst set
of rules, which the neural net renes
by running through hundreds of thousands of inputs, slightly varying the fuzzy sets each time to see how well the
system performs. The net tends to keep
the changes that improve performance
and to ignore the others.
A handful of products in Japan now
use supervised neural learning to tune
the fuzzy rules that control their operation. Among them are Sanyos microwave oven and several companies washing machines. Sharp employs this technique to modify the rules of its fuzzy
refrigerator so that the device learns
how often its hungry patron is likely to
open the door and adjusts the cooling
cycle accordingly. So far the neural net
must learn o-line in the laboratory,
from small samples of behavior by average customers. In time, researchers
at such groups as Japans Laboratory for
International Fuzzy Engineering and the
Fuzzy Logic Systems Institute hope to
build fuzzy systems that will adapt to
the needs of each consumer.
Supervised networks do have drawbacks. Tuning such systems can take
hours or days of computer time be-
Fuzzy sets of two objects are points of the square, such

as the point A. The union and intersection of a fuzzy set A
and its complement A* lie at the corners of an interior
square. In contrast, the union or intersection of a standard set and its complement results in the total set or the
empty set at the squares corners, respectively.
(01)
SECOND OBJECT
Systems with fuzzy controllers are often more energy ecient because they
calculate more precisely how much power is needed to get a job done. Mitsubishi and Koreas Samsung report that
their fuzzy vacuum cleaners achieve
more than 40 percent energy savings
over nonfuzzy designs. The fuzzy systems use infrared light-emitting diodes
to measure changes in dust ow and so
to judge if a oor is bare. A four-bit microprocessor measures the dust ow to
calculate the appropriate suction power and other vacuum settings.
Automobiles also benet from fuzzy
logic. General Motors uses a fuzzy transmission in its Saturn. Nissan has patented a fuzzy antiskid braking system,
fuzzy transmission system and fuzzy
fuel injector. One set of fuzzy rules in
an on-board microprocessor adjusts the
fuel ow. Sensors measure the throttle
setting, manifold pressure, radiator water temperature and the engines revolutions per minute. A second set of fuzzy
rules times the engine ignition based on
the revolutions per minute, water temperature and oxygen concentration.
One of the most complex fuzzy systems is a model helicopter, designed
by Michio Sugeno of the Tokyo Institute of Technology. Four elements of
the craftthe elevator, aileron, throttle
and rudderrespond to 13 fuzzy voice
commands, such as up, land and
hover. The fuzzy controller can make
the craft hover in place, a dicult task
even for human pilots.
A few fuzzy systems manage information rather than devices. With fuzzy
logic rules, the Japanese conglomerate
Omron oversees ve medical data bases
2/3
(11) TOTAL
SET
A
CENTRAL
POINT
1/3
A*
(00)
1/4
EMPTY SET
3/4
(10)
FIRST OBJECT
How Fuzzy Systems Learn Rules
daptive systems called neural networks can help fuzzy systems learn rules. A neural network accepts
pairs of input and output data, such as temperatures and
motor speeds for air conditioners, and groups them into
a small number of prototypes, or classes. Within the network, each prototype acts as a quantization vectora
list of numbersthat stands for the synapses feeding into
a neuron. When a new data point enters the network,

it stimulates the neuron associated with the prototype
that matches the data most closely. The values of the
winning synapses adjust to reflect the data they are receiving. As the data cluster, so do the quantization vectors, which define rule patches. More data lead to more
numerous and precise patches.
FIELD OF COMPETING
NEURONS
RULE PATCH
OUTPUT
SYNAPTIC VECTOR
QUANTIZATION
VECTOR
DATA POINT
INPUT
cause networks may converge on an inappropriate solution or rule or may fail

to converge at all. Neural researchers
have proposed hundreds of schemes to
alleviate this problem, but none has
removed it. Even after a lengthy tuning
session, the nal rules may not be much
better than the rst set.
Rather than relying on an expert to
supply a training set of data and correct a network in the process of learning, unsupervised neural networks learn
simply by observing an experts decisions. In this way, an adaptive fuzzy
system can learn to spot rule patterns
in the incoming data. Broad rule patches form quickly, based on a few inputs.
Those patches are rened over time.
nsupervised neural networks

blindly cluster data into groups,
the members of which resemble
one another. There may be no given
right or wrong response or way to organize the data. The algorithms are simpler, and, at least in theory, the network
need run through the data just once. (In
some cases, when data are sparse, the
neural net must cycle through them repeatedly.) Unsupervised learning is thus
much faster than supervised learning.
With numerical inputs and outputs supplied by an expert or a physical pro-
cess or even an algorithm, an unsupervised neural network can nd the rst

set of rules for a fuzzy system. The
quality of the rules depends on the
quality of the data and therefore on
the skills of the expert who generates
the data. At this point, there are fewer
unsupervised than supervised adaptive
fuzzy systems. Because unsupervised
networks are best used to create rules
and supervised networks are better at
rening them, hybrid adaptive fuzzy
systems include both.
Most fuzzy systems have been control systems with few variables. That
trend happened because most of the
rst fuzzy logic engineers were control theorists and because a control loop
regulates most consumer products. The
challenge for the next generation of fuzzy research will be tackling large-scale,
nonlinear systems with many variables.
These problems can arise when people
try to supervise manufacturing plants
or schedule airline ights or model the
economy. No experts may be able to
describe such systems. Common sense
may falter or not apply. The neural nets
that must learn the rules for modeling
these hard problems may have little or
no data to go on.
A further problem is that, like any
other mathematical or computer mod-
el, fuzzy logic falls prey to the curse

of dimensionality: the number of fuzzy rules tends to grow exponentially as
the number of system variables increases. Fuzzy systems must contend with a
trade-o. Large rule patches mean the
system is more manageable but also
less precise. Even with that trade-o,
fuzzy logic can often better model the
vagueness of the world than can the
black-and-white concepts of set theory.
For that reason, fuzzy logic systems may
well nd their way into an ever growing number of computers, home appliances and theoretical models. The next
century may be fuzzier than we think.
FURTHER READING
VAGUENESS: AN EXERCISE IN LOGICAL
ANALYSIS. Max Black in Philosophy of
Science, Vol. 4, No. 4, pages 427455;
October 1937.
FUZZY SETS AND APPLICATIONS: SELECTED PAPERS BY L. A. ZADEH. Edited by R.
R. Yager et al. Wiley-Interscience, 1987.
NEURAL NETWORKS AND FUZZY SYSTEMS.
Bart Kosko. Prentice-Hall, 1991.
PROGRESS IN SUPERVISED NEURAL NETWORKS. Don R. Hush and Bill G. Horne
in IEEE Signal Processing Magazine,
Vol. 10, No. 1, pages 839; January
1993.
81
Edwin Hubble
and the Expanding Universe
More than any other individual, he shaped
astronomers present understanding of an expanding
universe populated by a multitude of galaxies
by Donald E. Osterbrock, Joel A. Gwinn and Ronald S. Brashear
uring the 1920s and 1930s, Edwin Powell Hubble changed the
scientic understanding of the
universe more profoundly than had any
astronomer since Galileo. Much as Galileo banished the earth from the center
of the solar system, Hubble proved that,
rather than being unique, the Milky Way
is but one of untold millions of galaxies, or island universes. Hubbles work
also helped to replace the notion of a
static cosmos with the startling view
that the entire universe is expanding,
the ultimate extension of Galileos deant (if apocryphal) assertion, Yet still
it moves. Although many researchers
contributed to those revolutionary discoveries, Hubbles energetic drive, keen
intellect and supple communication
skills enabled him to seize the problem of the construction of the universe
and make it peculiarly his own.
Hubbles early years have become enmeshed in myth, in part because of his
desire to play the hero and in part because of the romanticized image of him
recorded by his wife, Grace, in her journals. Many accounts of Hubble conse-
DONALD E. OSTERBROCK, JOEL A.

GWINN and RONALD S. BRASHEAR share
a keen interest in Hubbles place in the
history of astronomy. Osterbrock is a former director of Lick Observatory at Mount
Hamilton, Calif., where he now researches
and teaches astronomy. He has written
many books and is currently working on
a biography of George Willis Ritchey, a
noted American telescope maker and designer. Gwinn has spent more than 30
years at the University of Louisville, where
he is a professor of physics; he has concentrated on optics, atomic spectroscopy
and the history of astronomy, including
Hubbles early life. Brashear earned his
masters degree in 1984 from Louisville,
where he met Gwinn. Brashear is curator
of the history of science at the Henry E.
Huntington Library in San Marino, Calif.
84
quently bear little relation to the real

man. For example, his biographers often
state that he was a professional-caliber
boxer, yet there is no evidence to support that claim. Likewise, the oft-repeated story that he was wounded during
service in World War I seems to have
no basis in fact.
Even without such embellishments,
Hubbles life has the ring of an all-American success story. The Hubble family
established itself in Missouri in 1833,
when Edwins great-grandfather rode
horseback from Virginia to settle in what
is now Boone County. Edwins grandfather and his father, John P. Hubble, were
hard-working, moderately prosperous
insurance agents. His mother, Virginia
Lee James Hubble, was ninth in direct
descent from Miles Standish. With characteristic overexuberance, Grace Hubble described her husbands ancestors
thus: Tall, well-made, strong, their bodily inheritance had come down to him,
even to the clear, smooth skin that
tanned in the sun, and the brown hair
with a glint of reddish gold. They had
handed down their traditions as well,
integrity, loyalty as citizens, loyalty to
their families . . . and a sturdy reliance
on their own eorts.
dwin was born in Marsheld, Mo.,

on November 20, 1889, the third
of seven children who survived.
In 1898 the Hubbles moved to Evanston, Ill., and, two years later, to Wheaton, where Edwin showed a robust aptitude for both academics and athletics.
He became a star athlete at Wheaton
High School, especially on the track
team. During those years, Edwin also
consistently placed in the top quarter
of his class. At the age of 16, he entered the University of Chicago, where
he continued to display his twin talents.
He earned high grades in mathematics,
chemistry, physics, astronomy and languages. Despite being two years young-
er than most of his classmates, he won

letters in track and basketball.
After graduation, Hubble received a
Rhodes scholarship. Even then, it was
far from clear what direction his wideranging abilities would nally take.
When he went up to Queens College
at the University of Oxford in October
1910, Hubble read (studied) jurisprudence; both his father and grandfather
hoped he would become a lawyer. After completing the course in two years
with second-class honors, he switched
to Spanish in his third year. He also continued to participate actively in sports.
Edwin remained a Rhodes scholar for
three years and then rejoined his family, who had since moved to Kentucky.
He claimed to have passed the bar and
practiced law in Louisville. According
to Hubbles friend Nicholas U. Mayall, he
chucked the law for astronomy, and I
knew that even if I were second-rate or
third-rate, it was astronomy that mattered. No record of Hubbles bar exam
or of his law practice exists, however.
Rather, during this time he taught and
coached basketball at the high school in
New Albany, Ind., just across the Ohio
River from Louisville.
Uninterested in the law and unsatised by his life as a teacher, Hubble decided to return to his true passion, astronomy. In the spring of 1914 Hubble
wrote Forest Ray Moulton, his astronomy professor at Chicago, to ask about
the possibility of returning as a graduate student. Moulton enthusiastically recommended him to Edwin B. Frost, the
director of Yerkes Observatory in Williams Bay, Wis., where the graduate program was given.
Frost gladly accepted Hubble as a student and gave him a scholarship covering tuition and living expenses. The
American Astronomical Society was to
meet at Northwestern University in nearby Evanston in August 1914; Frost suggested that Hubble come north in time
EDWIN HUBBLE drew on his observing experience, personal

drive and access to top facilities to make a series of ground-
for the event. Thus it happened that

Hubble was present at the meeting when
Vesto M. Slipher, a quiet, modest astronomer at Lowell Observatory, fueled a
controversy as a result of his latest studies of nebulae.
Nebulae was a blanket term used by
astronomers for centuries to designate
faint, cloudy objects that, unlike comets, do not change in position or appearance. The nature of these objects deed
easy explanation. In 1755 Immanuel
breaking cosmological discoveries. He is seen here in the observing cage of the 200-inch Hale telescope, circa 1950.
Kant had postulated that some nebulae

might be so-called island universes
self-contained systems of stars like our
own Milky Way. That notion excited
many scientists imagination but could
not be easily proved.
During the 19th century, improved telescopic observations showed that many
nebulae denitely consist of clouds of
luminous gas. One noteworthy class of
spiral-shaped nebulae looks distinctly
unlike the others, however. By the be-
ginning of the 20th century, many astronomers had come to believe that spiral nebulae were in fact distant galaxies
composed of a multitude of stars; skeptics continued to argue that these objects were nearby structures, possibly infant stars in the process of formation.
At the 1914 Astronomical Society
meeting, Slipher personally presented
the rst well-exposed, well-calibrated
photographs of the spectra of spiral nebulae. Those photographs displayed light
85
HUBBLES FAMILY helped to instill in him an appreciation of

self-motivation and hard work. Here he is seen with his cousins; Hubble appears third from the left (top left). At the time
of his graduation from the University of Chicago (top middle), Hubble considered a career in law. This pose of Hubble
from the nebulae after it had been split

into its constituent rainbow of colors.
The spectra of stars contain dark lines,
called absorption lines, caused by radiation captured by atoms in the stellar
atmospheres. Slipher showed that the
spectra of spiral nebulae exhibit the
kinds of absorption-line spectra characteristic of collections of stars.
Moreover, Slipher found that the wavelengths of the nebular absorption lines
are signicantly oset from where they
appear in the laboratory. Such changes
in wavelength, or Doppler shifts, are
caused by the motion of the object emitting the light toward or away from the
86
with his sister Lucy, a Red Cross nurse, captures his sense of
pride in uniform (top right ). Hubble often went on weekend
outings, such as this one with his high school students and
their family members (bottom left ). Late in life, he came to enjoy the solitary pleasures of y-shing (bottom right ).
earth. The Doppler shifts indicated that

the spiral nebulae observed by Slipher
are moving far faster than is typical for
stars within the Milky Way and hence
strengthened the argument that the spiral nebulae are not part of our galaxy.
When Hubble arrived at Yerkes Observatory, he found a near-moribund institution engaging in routine research.
Undaunted, Hubble set out in his own
direction. Possibly inspired by Sliphers
presentation, he began an intriguing program of photographing nebulae using
the observatorys underused 24-inch reecting telescope. That work grew into
his Ph.D. thesis, Photographic Investiga-
tions of Faint Nebulae, which contains

foreshadowings of his later research on
galaxies and on cosmology.
Hubble described and classied the
numerous small, faint nebulae that did
not resemble diuse clouds of gas and
noted that most of those nebulae are
not spiral but elliptical. He emphasized
that the distribution of faint nebulae
avoids the parts of the sky near the Milky Way and that many of them appear
clustered together. Hubble left no doubt
as to the direction of his thinking. Suppose them to be extra-sidereal and perhaps we see clusters of galaxies; suppose them within our system, their na-
EARTHBOUND
OBSERVER
Ph.D. research had imparted a rm direction to his investigations.

Milton L. Humason, who later collaborated with Hubble on his studies of the
distant universe, vividly recalled Hubbles rst observing session at Mount
Wilson: Seeing that night was rated as
extremely poor on our Mount Wilson
scale, but when Hubble came back from
developing his plate he was jubilant. If
this is a sample of poor seeing conditions, he said, I shall always be able to
get usable photographs with the Mount
Wilson instruments. The condence and
enthusiasm which he showed on that
night were typical of the way he approached all his problems. He was sure
of himselfof what he wanted to do,
and of how to do it.
ubble promptly returned to the

question of the nature of the
nongalactic nebulae, those
whose members tend to avoid the galactic plane and concentrate in high galactic latitudes. At the time, he had not
yet fully accepted the notion that these
objects are galaxies outside our own.
He initiated his detailed study of nongalactic nebulae by concentrating on
the irregular object NGC 6822 (NGC
denotes J.L.E. Dreyers New General Catalog of nebulae). By 1923, using the
100-inch telescope, he had found several small nebulae and 12 variable stars
within NGC 6822.
This endeavor suffered a happy interruption in 1924, when Hubble married
Grace Burke Leib, a widow whose rst
husband had died in an accident in a
coal mine in 1921. Grace had met Edwin
during a trip to Mount Wilson in 1920
and idolized him for the rest of her life.
Years later she recalled her electrifying
rst impression of Edwin as an Olympian, tall, strong and beautiful, with the
shoulders of the Hermes of Praxiteles,
and the benign serenity.
GALAXY 1
1,000
INDIVIDUAL
GALAXIES
500
0
GROUPS
OF GALAXIES
0
106
2 X 106
DISTANCE (PARSECS)
GALAXY 2
CHANGE IN DISTANCE
CHANGE
IN DISTANCE
TIME 1
The newlyweds honeymooned in

Carmel and in Europe. Hubble then returned to Pasadena, where he turned his
attention to M31, the famous Andromeda galaxy. Drawing on his well-honed
observing skills and the unequaled capabilities of the giant 100-inch telescope,
Hubble managed to resolve distinctly
six variable stars in M31. To Hubbles
mind, the achievement unequivocally
conrmed that M31 is a remote system
composed of stars and, by implication,
that other, fainter spiral nebulae are
even more distant galaxies.
Although the brightest star that Hubble detected was only about 18th magnitude at its maximum60,000 times
fainter than the dimmest stars visible
to the naked eyehe managed to make
83 measurements of its brightness, an
impressive feat at the time. From these
data, he determined that the stars
brightness rose and fell in the manner
characteristic of a class of stars known
as Cepheid variables. These stars were
of tremendous interest because the period of variation of a Cepheid directly
relates to the stars absolute luminosity. By comparing that luminosity with
his observations of the stars apparent
brightness, Hubble was able to deduce
VELOCITY (KILOMETERS PER SECOND)
ture becomes a mystery, he wrote. Hubbles thesis is technically sloppy and

confused in its theoretical ideas, but it
clearly shows the hand of a great scientist groping toward the solution of a
profound problem.
In October 1916 Hubble corresponded with George Ellery Hale, the director
of Mount Wilson Observatory in Pasadena, Calif. Hale was scouting for sta
members to work with the 100-inch reecting telescope nearing completion
on Mount Wilson. Hale oered Hubble
a job, conditional on the completion of
his doctorate. Hubble hoped to nish
the following June and then take the job
at Mount Wilson.
Those plans were scuttled by the entry of the U.S. into World War I on April
6, 1917. Hubble, a former Rhodes scholar who had deep emotional ties to Great
Britain, decided that patriotic duty took
precedence over scientic pursuits. On
May 15, just three days after receiving
his degree, he reported for duty and began an ocer training course.
Hubble thrived in the military setting.
He became a battalion commander and
next a major in the 86th Black Hawk
Division. In September 1918 the division embarked for Europe, but, much
to Hubbles regret, the Armistice arrived
before he went into combat. He summed
up his military career to Frost in these
words: I am still a major. I barely got
under re and altogather [sic] I am disappointed in the matter of the War.
However, that Chapter is nished and
another is opening.
Hubble was discharged on August 20,
1919, whereupon he immediately went
to Mount Wilson. He had arrived at an
opportune time. The observatory now
had two huge reecting telescopes, a
60-inch and the new 100-inch, then the
largest in the world. Hubbles studies
at Yerkes had taught him how to make
ecient use of such telescopes, and his
TIME 2
TIME 1
TIME 2
RELATIVE DISTANCE
HUBBLES LAW states that galaxies appear to be receding at

speeds in direct proportion to their distance from the observer. A graph of distance versus velocity (top ), which Hubble
published in 1929, plainly shows the linear relation. That relation is caused by the fact that the universe is expanding
uniformly in all directions. The schematic illustration (bottom)

shows why that is so. Galaxy 2 is three times as faraway as galaxy 1, so it has moved three times as far during the duration
from time 1 to time 2. Earthbound observers therefore see galaxy 2 receding at three times the rate of galaxy 1.
87
GALAXY CLASSIFICATION scheme developed by Hubble in the 1920s assumed

that elliptical galaxies develop arms and turn into normal or barred spiral galaxies.
Astronomers have rejected Hubbles evolutionary theory but still use his diagram
and nomenclature to categorize the wide variety of galaxy shapes.
the distance to the star and its surrounding galaxy.

A demon of energy, Hubble made discoveries in rapid succession. By the end
of 1924 he had identied several more
Cepheid variables in M31 and in M33, a
spiral galaxy in the constellation Triangulum. He sent a paper describing his
results to the joint American Astronomical SocietyAmerican Association for
the Advancement of Science meeting in
Washington, D.C., where the eminent
astrophysicist Henry Norris Russell read
the paper for him on January 1, 1925.
Hubble had deduced that both M31
and M33 are about 930,000 light-years
distant, far beyond the known bounds
of the Milky Way. In the words of Joel
Stebbins, the astronomical societys secretary, Hubble had succeeded at last in
bringing conrmation to the so-called
island universe theory.
In contrast to the rapid progress of
his research, Hubble published his results in a leisurely fashion, trying to be
as certain as possible before committing them to print. The rst paper containing Hubbles results from his studies of variable stars in spiral nebulae
appeared later in 1925. Based on observations of Cepheids in NGC 6822,
he declared it to be the rst object definitely assigned to a region outside the
galactic system. Hubbles epochal paper, A Spiral Nebula as a Stellar System, presenting his evidence that M33
also is a galaxy lying far outside our
own, appeared in 1926. Not until 1929
did Hubble wrap up this phase of his research in a long paper that nally laid
88
out in full detail his studies of the distance to and stellar composition of M31.
In addition to determining the distance to the extra-galactic nebulae,
Hubble sought to classify them and to
comprehend their diversity of structure.
Starting in 1923, he ordered the nebulae according to a perceived pattern of
evolution. As an elliptical nebula ages,
Hubble argued, it forms arms that enlarge and open up. To explain the existence of barred spirals, in which the spiral arms trail from a barlike formation,
Hubble suggested that the elliptical
could evolve into either a normal spiral
or the barred form. The resulting double-pronged classication diagram is still
widely used, though stripped of its evolutionary implications.
y far Hubbles most remarkable

bequest to modern science was
his research showing that we live
in an expanding universe. Hints of that
phenomenon had begun to surface well
before Hubble tackled the problem. During the 1910s, Slipher had noticed that
most of the galaxies he observed seemed
to be moving away from the sun. By
1924 a trio of astronomersCarl W.
Wirtz in Germany, Knut E. Lundmark in
Sweden and Ludwik W. Silberstein in
the U.S.had assembled evidence that
particularly small, and presumably distant, spiral nebulae are receding from
the Milky Way faster than nearby ones.
Their data were marginal, however, and
Silbersteins polemical style and wild
claims actually hindered the acceptance
of the concept of an expanding universe.
Clearly, further progress required a

bigger telescope; the 100-inch telescope
on Mount Wilson was ideal for the task.
In the mid-1920s Hubble was busy measuring the brightnesses of stars in spiral galaxies, so he enlisted his colleague
Humason to determine the galaxies radial velocitiesthat is, their motion toward or away from the observer. Humason calculated those motions by measuring the displacement of the emission
and absorption lines in the galaxies
spectra. When an object is receding, its
light is shifted toward the red end of
the spectrum, resulting in a displacement called a redshift.
To search for signs of an expanding
cosmos, Hubble needed to know not
only the velocities at which galaxies are
moving but also how faraway those galaxies are. Hubbles familiarity with measuring cosmic distances greatly assisted this eort. In each of the few galaxies whose distances he had determined
by observing Cepheids, Hubble derived
the mean absolute magnitude of the
brightest stars and of the galaxy as a
whole. He then extrapolated from these
results to determine the distances to galaxies far beyond M31 and M33.
Hubbles rst paper on a velocity-distance relation, published in 1929, sent
shock waves through the astronomical
community. The ndings he presented
became part of the basis for the big
bang theory underpinning modern cosmology. In his paper Hubble emphasized
that, in contrast to some previous researchers, he used only those nebular
[galactic] distances which are believed
to be reliable. The paper combined his
distance estimates with precision measurements of radial velocities collected
by Slipher, Humason and others.
From these data, Hubble derived a linear relation between velocity and distance, written as v = Hd in modern notation. The statement that the velocities
of recession of distant galaxies are proportional to their distances is now
known as the Hubble law, and the constant H is referred to as the Hubble constant. The Hubble law implies that the
universe is expanding: velocities seem
to increase as one looks progressively
farther outward from any point within
the universe.
Between 1931 and 1936 Hubble and
Humason extended the law of redshifts
to increasingly great distances. Hubble
had immediately grasped that his ndings supported the notion of an expanding universe, but he never completely accepted that the redshifts resulted only
from the radial motions of galaxies. He
consistently maintained that the motions
inferred from the redshifts should be
described as apparent velocities.
Hubble estimated Hnow understood

as the expansion rate of the universe
to be about 500 kilometers per second
per megaparsec (one megaparsec equals
about three million light-years). In retrospect, Hubbles value of the constant
was much too large. Astronomers today
nd values between 50 and 100 kilometers per second per megaparsec. Hubbles errors arose chiey because he severely underestimated the absolute magnitudes of the brightest stars in the galaxies he observed. Nevertheless, it was
a splendid rst eort.
fter his investigation of the law of

redshifts, Hubble concentrated
his eorts in observational cosmology. In particular, he sought to measure how the apparent density of extremely remote galaxies changes with
distance, in order to determine the overall geometry of the universe. He collaborated with physicists Richard C. Tolman and Howard P. Robertson of the
California Institute of Technology, who
were at the forefront of research into
theoretical models of the curvature of
space. In this work, Hubble proved less
than successful, mostly because of the
diculty in accurately measuring the
brightness of dim, distant galaxies.
Although his later undertakings never equaled his remarkable ndings made
during the 1920s and early 1930s, Hubble continued to sway strongly the direction of astronomical research and to
make that research better known to the
public. His books The Realm of the Nebulae and The Observational Approach
to Cosmology inspired a generation of
young astronomers and physicists. Nobody who read Hubbles books or heard
his lectures could ignore him; his compelling personality seemed less like
those of other astronomers than like
those of the movie stars and writers who
became his friends in the later years of
his life. Hubble often affected English
clothes, expressions and mannerisms,
habits that colleagues who knew of his
Missouri background found unconvincing but that contributed to a dramatic
public image.
In an eort to push the boundaries of
observational cosmology, Hubble helped
to organize the construction of a 200inch Hale telescope on Mount Palomar
in California. The advent of World War
II delayed the construction of the telescope; soon after the Japanese attack
on Pearl Harbor, Hubble joined the sta
of the U.S. Armys Ballistics Research
Laboratory at Aberdeen, Md., where he
applied his early astronomical training
to directing calculations of artillery-shell
trajectories. The 200-inch telescope nally began operation in 1948. Hubble
SPIRAL GALAXY M33 was an early target of Hubbles eorts to determine the distance scale of the universe. Based on his observation of Cepheid variable stars in
M33, Hubble proved that it lies far beyond the known limits of the Milky Way and
hence must be an independent galaxy comparable to our own.
continued his research at Palomar until

he died at the age of 63, felled by a
stroke on September 28, 1953.
After Hubbles death, Humason and
then Rudolph L. Minkowski used the
200-inch telescope to detect ever more
remote, swiftly receding galaxies. Walter Baade recalibrated the period-luminosity relation for Cepheid variable stars
and discovered that all Hubbles galactic distances, huge though they seemed,
needed to be multiplied by a factor of
two. Hubbles former student Allan
Sandage, along with innumerable other
researchers, has continued to rene the
value of the Hubble constant. The triumphant work of Hubbles successors
stands as eloquent tribute to the innovative thinking and boundless energy that enabled Hubble to redirect the
course of modern cosmology.
FURTHER READING
THE EXPANDING UNIVERSE: ASTRONOMYS
GREAT DEBATE, 19001931. Robert Smith.
Cambridge University Press, 1982.
MAN DISCOVERS THE GALAXIES. Richard
Berendzen, Richard Hart and Daniel Seeley. Columbia University Press, 1984.
THE OBSERVATIONAL APPROACH TO COSMOLOGY: U.S. OBSERVATORIES PRE
WORLD WAR II. Donald E. Osterbrock in
Modern Cosmology in Retrospect. Edited by R. Bertotti, R. Balbinot, S. Bergia
and A. Messina. Cambridge University
Press, 1990.
SELF-MADE COSMOLOGIST: THE EDUCATION OF EDWIN HUBBLE. Donald E. Osterbrock, Ronald S. Brashear and Joel
A. Gwinn in Evolution of the Universe of
Galaxies: Edwin Hubble Centennial Symposium. Edited by Richard G. Kron. Astronomical Society of the Pacic, 1990.
89
TRENDS IN ENVIRONMENTAL SCIENCE
SUSTAINING THE AMAZON

by Marguerite Holloway, sta writer
Can scientists reconcile the

inevitability of economic
development with the
preservation of rain forests?
ig, brown and sedate, two toads blink

in the beam of the ashlight, while
nearby pairs of green eyes, the phosphorescent markings of click beetles, move
lazily along the leaf-laden oor of the forest.
On this night, Barbara L. Zimmerman is trying to nd what she calls the ultimate frog
for a group of visitors she is accompanying
through the dark tangle of the Brazilian rain
forest. The tropical biologist can hear her
quarrys barklike call, but the vivid emerald
creature, Phyllomedusa bicolor, is nowhere
to be seen.
Instead the nocturnal search reveals the
red irises of P. tarsius, the penultimate green
tree frog; Plica plica , a mint-chocolate-chipcolored lizard; the two enormous, poisonous
toads, Bufo marinus ; as well as a tireless, resonant frog, Osteocephalus taurinus, that resides in the camps drainpipe. The Amazon is
like this: look for one thing, nd another.
Zimmerman, who formerly stalked frogs in
this forest 70 kilometers north of the city of
Manaus to study their mating patterns, now
pursues them primarily for pleasure. These
days, her research centers on people, a shift
in emphasis that is increasingly common in
the Amazon. She has come here this time
as part of a team of tropical scientists who
are reviewing a project as well as possibilities for collaborative studies. Inevitably, their
discussions turn to ways of reconciling the
preservation of the rain forest with economic development.
Many of the scientists seem resigned to the
prospect that the rain forest will, in large part,
be cut down even as they struggle to understand the biological wealth of this tropical ecosystem. The ve million square kilometers of
HARVESTING BRAZIL NUTS and other products may allow economic growth without destroying the Amazon.
Researchers are debating the potential of such practices.
MARGUERITE HOLLOWAY spent a month in the Brazilian rain

forest to report this article. Traveling from one end of the Amazon basin to the other, she interviewed scientists working at remote research stations, native preserves, extractive reserves,
ranches and frontier logging camps.
91
AMAZON RAIN FOREST extends into nine countries and covers 60 percent of Brazilsome 510 million hectares (about 1.3 billion acres).
the linchpin of the 1992 United Nations

Conference on Environment and Development, but the fashionable phrase,
which now appears in nearly every grant
or loan application, is still being dened.
Thus, researchers are seeking pieces of
a mosaic of development, the overall
pattern of which, like the web of biodiversity, can only be guessed at.
Slowly, however, some scientic underpinnings for sustainable development appear to be emerging. Sometimes
the answers seem to contravene assumptions about what might be ecologically sustainable. Scientists are investigating how to use the millions of hectares of formerly productive pasture
that lie abandoned throughout the Amazon. Once cut down, the rain forest does
not necessarily become a desert but
rather a eld or a secondary forest that
can again be harnessed for ranching,
logging or agriculture. If this land can
be returned to productivity, it would reduce pressure to cut virgin forest. Indeed, a group of researchers recently
challenged dogma by suggesting that
cattle ranchinglong thought to be an
ecological and economic anathema
may be a viable land use.
Whatever the project or nding, controversy is never far o in Brazil. The
country is so vast and the economic,
social and political problems so complicated that every issue is labyrinthine: follow the thread, and there are often several Minotaurs. One of the most
contentious topics involves the future
of indigenous people and their land. It
is on one of these Indian reserves, with
a people called the Kayap, that Zimmerman has chosen to work.
Zimmerman and her sponsorsthe
environmental group Conservation International and the David Suzuki Foundation, a Canadian nonprot organizationare setting up a scientic research
station on remote Kayap territory in
the southern reaches of the Amazon.
The villagers of Aukre, which means
place where the river bottom is noisy,
have set aside 5,000 hectares for the
eld station within the 500,000 hectares of rain forest that they control.
The project will characterize the biodiversity of a region that has not been
extensively studied, Zimmerman explains. The ora and fauna of the Amazon vary so greatly that adjacent sites
can shelter entirely dierent species.
One hectare in the eastern Amazon may
have 120 tree species, whereas a nearby hectare may have 170. The woods
around Aukre may have creatures or
plants that are not found, or are infrequently seen, elsewhere in Brazil.
Many scientists have shied away from
working with Indians because the issues
92
Amazon tree cover make up the largest

continuous expanse of tropical rain forest remaining in the world. Although
such forests cover only 7 percent of the
planets land surface, they are inhabited
by some 50 percent of the plants and
animals found on the globe (estimates
of which range from a total of two million to 30 million species). Most disturbing of all is the fact that nobody really knows how many. But as many as
27,000 species may be consigned to extinction every year, calculates Harvard
University biologist Edward O. Wilson.
This destruction is the result of forests
being felled at a rate of 1,800 hectares
(about 4,500 acres) every hour.
The impact of these extinctions is
multifold: the rain forests have profound philosophical, spiritual, cultural,
scientic and economic signicance. Because Brazil is considered to have the
highest species diversity on the earth,
it is the epicenter of eorts to stem deforestation. So far the country has lost
only an estimated 12 percent of its forest, a sign of hope to many researchers

and development experts. Yet the country has a debt burden of $121 billion,
staggering poverty and one of the most
unequal distributions of wealth in the
world. Any successful attempt to preserve the forest must necessarily meet
human economic objectives.
By working with the people who live
in and use the forestincluding Indians,
rubber tappers, ranchers, loggers, farmers and industrialistsscientists such
as Zimmerman hope to strike a balance
between extremes of conservation and
exploitation that will protect species and
threatened environments while fostering economic development and reducing poverty. Do what you can dojust
do something, urges Don E. Wilson,
head of the biodiversity program at the
Smithsonian Institution.
These eorts are among the myriad
attempts to implement what is called
sustainable development. The idea was
EQUATOR
MACAPA
MANAUS
BELM
Paragominas
ACRE
RIO AMAZONAS
BRANCO
PAR
Xapuri
RONDNIA
Aukre
B R A Z I L
MATO
GROSSO
BRASLIA
RIO DE
JANEIRO
SO PAULO
PACIFIC
OCEAN
TROPIC
OF CAP
RICOR
N
PRTO
ALEGRE
ATLANTIC
OCEAN
facing them are so volatile. It is a

an hour by canoe from Aukre.
land mine, comments Anthony B.
Bright-blue morpho butteries it
Anderson, program ocer at the
down trails pockmarked with taFord Foundation in Brazil. There
pir tracks, and the wolf whistle
are politics within the tribes and
of screaming pia birds rings out
without, as well as all of these
from the canopy.
misconceptions about what the
Zimmerman intends to ll this
Indians are supposed to be.
place with graduate students and
The debate about what indigeresearchers who will study the fornous people should do with their
est and work with the Kayap. She
land is bitter. As many as 79 milnotes that natives have a dierlion hectares have been or are in
ent way of knowing the forest
the process of being designated
for instance, men called mateiros
as reserves, according to Stephan
can identify hundreds of trees by
Schwartzman of the Environmensmell alone. She expects that such
tal Defense Fund. To many sciencollaboration will yield medicinal
tists and environmentalists, this
plants or new crops and perhaps
expanse represents a perfect sea strategy for the sustainable harries of places to preserve and
vesting of mahogany.
study biodiversity as well as naIn turn, the project is expected
tive cultures. The Kayap, for into bring income to Aukre. Scienstance, control 10 million hectists will buy local products, ecotares, a region approximately the
logical tours may visit the site
size of Ontario.
under Zimmermans stewardship
Yet these lands are also rich in
and y-shermen may explore the
minerals and timber and so are
rivers varied population of sh.
ripe for economic development.
Collaborative studies with the KaMany are illegally inltrated by
yap could also lead researchers
loggers and gold miners, such as KAYAP TRIBESMAN, Bastian, waits in a canoe that to medicinal plants or agriculturthe territory of the Yanomami in transports him between a forest reserve and his vil- al techniques that bring some northern Brazil, where violence, lage, Aukre. He is working on a eld station that may nancial return. Although the town
disease and mercury poisoning, bring researchers and revenue to Aukre.
makes money harvesting Brazil
the last a result of the gold extracnut oil for Body Shop Internationtion process, are pervasive. But
al, a British company that develthe erce reputation of Kayap warri- tunistic. In the past, they used resourc- ops and sells natural beauty products, it
ors has successfully deterred encroach- es sustainably, but not always.
is not enough.
ments. Some Kayap chiefs have opted
If Zimmermans project is successful,
The fervor about the future of Indito sell mahogany and mining rights, a it could help settle the debate by sup- an lands is echoed in arguments about
decision that has aggravated environ- plying a source of income for Indians another kind of reserve: extractive rementalists. Greenpeace, the internation- while simultaneously maintaining bio- serves. These areaswhich are also set
al environment group, recently sought diversity. Aukre is an interesting choice aside by the government and whose
to enforce a ban on mahogany logging for such a site because it has experi- boundaries are also rarely enforced
on the tribal landsan action that has, enced each facet of the controversy. The are places where local people harvest
in turn, aggravated certain Kayap. It is villages chief, Paiakan, is famous as a natural products, including latex, palm
estimated that about 40 percent of the spokesperson for protecting rain forests hearts and Brazil nuts, without damagAmazons mahogany treeswhich, in and tradition, as well as for being a skill- ing the rain forest. Such reserves were
general, only grow in a strip across the ful politician who could unify previous- made famous by Francisco Alves Menlower portion of the rain forestthrive ly warring tribes in the face of develop- des Filho, known as Chico Mendes, a
on reserves.
ers. Yet other villagers have disagreed rubber tapper from the town of Xapuri
with his view. In addition, Paiakan can in the western state of Acre. Because of
Noble Savage
no longer easily be an advocate for in- his eorts to organize the rubber tapdigenous people in the Amazon: he has pers to ght for their land, Mendes was
Development experts and conserva- reportedly been charged with attempt- murdered by cattle ranchers in 1988.
tionists, many of whom struggle to help ed rape and is awaiting trial.
Extractive reserves, of which there are
establish reserves, expect that Indians
The region around the reserve is as now 14 covering a total of three million
should maintain the rain forest intact. scarred as the tribe. The ight north- hectares, confer land rights to squatThis view has been shaped, in part, by west from Redenoa frontier town ters and people whom large landholdanthropological studies documenting whose economy is based on minerals ers have dispossessed. ( In Brazil, 4.5
tribes sustainable use of the forest and and mahogany from Kayap landto percent of the landowners control 81
their sophisticated agricultural systems. Aukre oers a view of the entire spec- percent of the farmland.) The presence
Others experts argue, however, that sus- trum of development. Ranches, logged of these powerful ranchers is felt for
tainability has been achieved by default: plots and rivers disgured by gold min- the entire four-hour drive from the city
the forest is large enough to recover ers are slowly replaced by uninterrupt- of Rio Branco to Xapuri. The road is
from alteration, and populations are ed forest. On the ground, the new re- anked by pastures, and the landscape
small enough to have limited impact. search site appears intact, although it is verdant but eerie: ghostlike Brazil nut
The idea of the noble savage has real- is bordered on one side by an old log- trees, lone relics of the forest, stand in
ly confused the issue, Anderson notes. ging road. The station stands in the the green elds, their bleached branchIndians, like everyone else, are oppor- middle of a clearing just o a river, es luminous against the backdrop of
93
dark, stormy sky. A law prohibits cutting these majestic trees, but once pasture is burnt around them, the Amazon
sun inevitably desiccates them.
Disagreements about extractive reserves pivot on questions of economic viability. Scientists on one side argue
that insucient markets exist for many
of the products. By studying past market cycles, Alfredo Kingo Oyama Homma of the Brazilian Enterprise for Agricultural Research ( EMBRAPA ) found
that when a forest product such as latex becomes commercially important
it is inevitably introduced into higheryielding plantations; in 1991 about 60
percent of Brazils natural rubber came
from plantations. Or the material is
made synthetically. As a result, the price
plummets, and small-scale extraction
ceases to be protable.
Other researchers counter that the
matter has not been settled. One study
conducted in Peru several years ago
found that 72 products could be harvested from a one-hectare plot for an
annual yield of $422. The investigators
concluded that this approach was more
lucrative than the one-time logging profit of $1,000. On the other hand, Anderson notes, if you took a random sample of forests here in the Amazon, you
would not nd many that would have
such a high value of nontimber forest
products. In addition, the Peruvian
products could be marketed regionally, a situation that does not yet exist
for many of the reserves in the Amazon.

Nevertheless, Anderson and others argue that there may be a compromise. I
am one of the proponents, so I am suspect, Anderson cautions. But I like to
think that I am somewhere in the middle ground. Maybe the term extractive
reserve is a misnomer. These are areas
zoned to protect the rights of people
who live in the forest. Anderson explains that once land claims are guaranteed, extractivists begin to look for supplemental livelihoodssuch as raising
livestock or planting additional crops.
Extraction is undergoing a transformation, and therefore it is potentially a basis for sustainable development. The
implications of this economic diversication for the preservation of biodiversity are not yet established.
Many scientists and nongovernmental organizationsCultural Survival, a
Boston-based group of anthropologists,
and the New York Botanical Gardens
Institute for Economic Botany, among
themare working with rubber tappers
and Brazil nut gatherers to nd other
products, such as medicinal plants and
fruits, and to open new markets. Some
are addressing the overwhelming problems of communication and transportation. Many individuals who harvest forest products travel for days to sell them.
Getting out of Xapuri, even at the outset of the rainy season, illustrates the
challenge of reaching the market. Gomercindo Clovia Garcia Rodrigues, a
member of the extractivist cooperative

of Xapuri, hitches a ride with a group of
travelers back to Rio Branco so he can
visit his dentisthe says he was shot
in the jaw by ranchers a while ago, and
the wound continues to hurt. As he describes the woes of driving at this time
of year, the white Fiat sinks in the red
muck of the road, for the rst time. After the fourth escape, we are covered
with a thick, rust-colored patina of mud.
The windshield wiper breaks, the rain
blinds and the battery fails. Rodrigues
laughs: the Amazon has made his point.
Amazon Microcosm
RUBBER TAPPER extracts latex . As the pattern of diagonal lines shows, tappers repeatedly cut each of the rubber trees on their land to collect the white liquid.
At the opposite end of the country,

in the eastern state of Par, researchers
are working with Rodriguess traditional foes: ranchers and loggers. In a valley capped by smoke, res from dozens of sawmills smolder. The black and
gray plumes hang in the humid air of a
late afternoon. The only motion is that
of children playing alongside the barren road, tin cans tied to the bottoms
of their feet like shiny platform shoes.
Although the image of a deforested
Amazon is by now a familiar one, a
close look at this landscape reveals an
unusual collaboration. Christopher Uhl
of Pennsylvania State University, Daniel
C. Nepstad of the Woods Hole Research
Center and Adalberto Verissimo, Paulo
Barreto and their colleagues at the Amazon Institute of Man and the Environment (IMAZON) are working with owners and employees at local sawmills
and farms. They are documenting ways
of reclaiming land that has been abandoned and of managing forests for timber. Uhl and Verissimo are fond of describing the region near the town of
Paragominas, where they have focused
their eorts, as a microcosm for the
Amazon. They believe that if the forces
driving deforestation are understood
and altered here, the lessons can be ap^
plied to states such as Rondonia
and
Mato Grosso, which, by 1988, had both
lost some 24 percent of their forest.
Par encompasses one third of the
Brazilian Amazon and had lost 13 percent of its forest cover by 1990. The
area opened for development in the
mid-1960s, when completion of a highway connected the city of Braslia to Belm, the hub of Brazils rubber boom.
At that time, the government promoted
cattle herding by oering subsidies
often as great as 75 percent of the costs
of starting the ranchto settlers. The
government also oered property to
anyone who could demonstrate that
they had cleared land, a policy that led
to spasms of deforestation.
Several problems quickly emerged.
94
BOLUSES OF LATEX are stacked in a

courtyard at the marketing cooperative
in the town of Xapuri after they are
brought in from the forest.
Because the forest covets its nutrients,

its soils are relatively poor. Although
ranchers slashed and burned the forest,
releasing phosphorus, nitrogen, potassium and other riches into the ground,
the infusion was short-lived. After a few
seasons of overgrazing, grasses no longer grew. A riot of unruly and, to cattle,
unappetizing vegetation soon covered
the elds. By the 1980s ranching was
not economically worthwhile. The government proved as unsupportive as the
soil. A scal crisis dried up subsidies,
making clear-cutting and cattle raising
even less attractive.
A new industry emerged in ranchings
stead. The native forests of southern
Brazil had by then been for the most
part leveled, and ranchers realized they
could aord to rejuvenate degraded
pastures by selling logging rights to
their property. More than 100 species
of wood, which had been viewed as impediments to pasture, were harvested.
By 1990 there were 238 sawmills near
Paragominas.
Today logging is the areas principal
industry, and it is just as unsustainable
as ranching was. IMAZON researchers
have determined that for each tree that
is removed in a process called selective
logging, 27 other trees that are 10 centimeters or more in diameter are severely injured, 40 meters of road are created and 600 square meters of canopy are
opened. Recovery of the forest is slow
even under these conditions: it can take
70 years for selectively logged forests to
resemble their original state.
Nevertheless, Verissimo, Uhl and others think sustainable approaches to
logging and ranching are possible. I
am much more optimistic than I was
ve years ago, Verissimo comments
as he bounces in a car on the way to
a ranch, Fazenda Vitria. At the study
station there, he climbs a tower that
stands at the intersection of the four
distinct kinds of land that characterize the region: primary, or pristine, forest; partially regrown, or secondary,
forest; pasture that is in use; and abandoned pasture. Verissimo points to the
old pasture, explaining that the forest
is slow to invade such sites, in part, because some seed-dispersing birds do
not venture into open spaces.
Although there are about 10 million
hectares of abandoned pasture in the
Brazilian Amazon, scientists are just beginning to monitor what happens to
land once it has been burned, grazed
and then left. The investigations support one of the most exciting possibilities for sustainable development. If already deforested areas could be eciently used for agriculture, ranching or
logging, pressure could be taken o the
primary rain forest. There is no need
to deforest when two thirds of the already deforested land is not being used
in a reasonable way, Uhl notes.
Alternatively, if such lands could be
restored or coaxed back into forest,
biodiversity could be better protected.
Understanding which trees, plants and
creatures reestablish themselves rst
and under what conditions could allow researchers to speed up the recovery. Uhl cautions, however, that making
the process economically attractive is
critical. Restoration makes us all feel

good, but it doesnt necessarily address
the key issue, he warns. Putting money into it without economic development is not the best thing to do.
Studies of pastures and ranching have
already led Uhl and his colleague Marli
M. Mattos of Woods Hole to the somewhat heretical conclusion that, under
certain conditions, cattle ranching can
be protable. They observe that if ranchers manage abandoned land intensivelythat is, by introducing fertilizer or
by planting better adapted species
cattle herds for beef and for dairy products can be lucrative. In fact, ranchers
could increase their prots as much as
vefold at the same time that employment could rise. (Given estimates of
95
MEASURING PRIZE HARDWOODS is one

of the rst things a logger does to select
the tree he will cut, demonstrates Paulo
Barreto (right) of the Amazon Institute
of Man and the Environment ( IMAZON ).
population in Brazil246 million by

2025considerations of employment
are fundamental.)
The two scientists note that the drawbacks to intensive ranching are signicant. Fire is becoming more and more
common, Uhl states. It will increase
as landscapes are altered. The high forest is by nature quite resilient to re,
but as soon as you begin to alter it, it is
much more prone to burning. According to Nepstad of Woods Hole, more
than one half of the land in an 80kilometer radius around Paragominas
burned in 1992.
There are other ecological costs as
well. Nepstad has shown in studies of
soil and root systems that pastures hold
only 3 percent of the aboveground car96
bon and 20 percent of the root carbon

that forests do. Moreover, some 220
tons of carbon are released when the
slash from each hectare of forest is
burned. This emission has implications
for global warming: tropical forest loss
is estimated to have contributed one
quarter of the carbon released into the
atmosphere during the past decade.
Evapotranspirationthe means by
which roots absorb water from the soil
and release itis also reduced in pasture because the root systems are relatively shallow and cannot reach deep
deposits of moisture. In the Amazon,
where 50 percent of precipitation may
be generated by the ora, changes in
evapotranspiration could cause alterations in local climate and rainfall pat-
terns that adversely aect agriculture.

Reductions in rainfall, in turn, cause
the local temperature to increase, hampering vegetation and crops.
The ranching conclusions reveal some
of the problems of dening the word
sustainable. It is a really tricky term.
There are all sorts of ways to measure
sustainability, from an environmental,
an ecological and a social perspective.
It is a rare land use indeed that could
fulll all three criteria, observes Anderson of the Ford Foundation. Dont get
me wrong, I am not saying ranching is
good, but it is not necessarily as bad as
it seemed in the earlier literature. The
question is not whether or not ranching
is suitable, because it is there to stay.
Research on sustainable logging in
Paragominas is not as controversial. In
a forested site adjacent to the study
tower, IMAZONs Barreto describes one
simple strategy he has designed to help
combine forest management and profit. Loggers routinely send workers into
the forest to mark and cut the prize
trees. Equipment then follows and creates roads. Barreto has found that by
plotting the chosen trees in a simple
software program, a shorterand therefore less destructive and less expensivenetwork of roads can be designed.
In addition, injury to the surrounding trees can be reduced by as much as
30 percent. Pruning vines in a selectively logged forest can also permit light to
reach seedlings, accelerating the growth
of commercially important species. If
loggers implemented these two techniques, Barreto says, forests would take
35 instead of 70 years to reach their
prelogged stature. It would reduce
their prots only slightly to get sustainability, Verissimo asserts.
Because internal and external demand
for Amazonian timber is expected to
grow, more research on forest management strategies is needed. In 1976,
14 percent of the countrys sawn wood
came from the rain forest; by 1986 the
share had grown to 44 percent. The
number of wood processors grew from
about 100 in 1952 to more than 3,000
in 1992. Unlike domestic demand, however, international demand may catalyze
changes in logging practices. Currently
less than 5 percent of the worlds supply of tropical hardwood comes from
Brazil, but supplies in Asia are plummeting. Yet outside demand for timber could be quite positive if nongovernmental organizations are involved
in promoting green labeling, Uhl says.

Retailers want to know where this
wood is coming from. This little step is
tremendously important.
Regardless of ndings such as those
in Paragominas, research amounts to
nothing if it is not implemented. Because Brazil has no coherent Amazonian policy, the IMAZON workers and
their colleagues have suggested that
countywide plans would be eective.
The Brazilian Constitution of 1988 conferred more power to counties; therefore, these local governments could enact legislation requiring that specied
land be used intensively for pasture and
logging. Such laws would generate jobs
for the region and preserve primary forest. Uhl says ranchers and mill owners
are beginning to implement aspects of
this management strategy as they realize that resources are nite and that
well-placed investments in intensication will help guarantee sustainability.
Survival, rather than prot, is the
consideration behind another principal
cause of deforestation. For impoverished farmers, shifting cultivation often provides the only source of livelihood, as the ubiquitous small, smol-
dering plots of trees alongside roads

testify. Natural resources are the bleeding valve for the poor; that is where
they turn to survive, observes Ralph M.
Coolman, a tropical ecologist at North
Carolina State University, who is currently working with EMBRAPA .
On the elds of an old water-bualo
ranch and former rubber and palm plantation near Manaus, Coolman and his
colleagues are investigating agroforestry
systems. Agroforestry combines animal
husbandry with farming and forestry.
The practice has a long history, but scientic agroforestry is relatively new,
says Erick C. M. Fernandes, also of North
Carolina State and EMBRAPA. By determining which plants are best suited to
particular soils, the researchers hope to
be able to increase the subsistence and
prot yield of shifting cultivars, while
letting them stay indenitely on the deforested land. They are identifying species that are best suited to xing nitrogen and replacing other nutrients.
The eort to balance preservation and
economics that characterizes the work
of Fernandes and Coolman, the Paragominas group and Zimmermans team
can also be seen at one of the longest-
SAWMILL NEAR PARAGOMINAS, a town in the eastern Amazon, is one of 238 in the vicinity. Although Brazil supplies less
than 5 percent of the worlds hardwood, this share is expect-
98
running programs in the rain forest. This

study, the brainchild of biologist Thomas E. Lovejoy, was created in 1979 as a
collaboration between the National Institute for Amazon Research (INPA) and
the World Wildlife Fund. Originally called
the Minimum Critical Size of Ecosystems
Project, it is now administered jointly
by INPA and the Smithsonian Institution, where Lovejoy serves as assistant
secretary for external aairs.
Business and Biodiversity
The program was established to answer a fundamental question of ecology : What is the smallest size an island
(or, in this case, a tract of forest) can be
and still maintain its biodiversity? The
ndings have direct implications for
protecting species. Because landowners
in the Amazon are required by law to retain 50 percent of their forests, scattered woods of all dierent sizes and
shapes can be found everywhere, often
surrounded by pasture.
By studying these various-sized plots
and monitoring the changes in species
composition, Lovejoy and his colleagues
substantiated the prediction that larger
ed to increase as stocks in Asia dwindle. Researchers are

working with loggers and sawmill owners here to determine
how to practice sustainable logging.
forest fragments harbor more species

than do smaller ones. But they also
found that size alone was not always the
determining factor. Instead it was crucial to understand the habitat requirements of each species. The project provided evidence that corridorswooded
paths that connect one fragment to anotherfacilitated the movement of certain species between fragments. Because
it is hard to nd large, uninterrupted
parcels of land in many areas throughout the world, corridors are now incorporated in plans for wildlife preserves.
The project recently included an additional focus: the study of restoration,
including the regrowth of pasture. Its
new name, the Biological Dynamics of
Forest Fragments Project, reects this
emphasis. Graduate students and researchers are expected to supply data
for forest management, soil reclamation
and commercial harvesting. The shift
also ensured continued funding. According to tropical scientists, money for sustainable development is easier to obtain than are grants for basic research.
Lovejoy believes that another key
to saving the rain forest lies in working with business. In the mid-1980s he
developed the idea of debt-for-nature
swaps. Recognizing that nations burdened with foreign debt could devote
few resources to protecting ecology,
Lovejoy proposed that conservation organizations repay a portion of the debt
in exchange for a guarantee that land
would be protected. Alleviating debt
would both maintain biodiversity and
strengthen the economy.
Building on this idea, Lovejoy and
others are drawing attention to the nancial importance of biodiversity. For
example, pharmaceutical products, genetic stocks for agricultural products
and timber species can be found in
these forests. By publicizing this economic potential and the importance of
conserving the resource base for future
prot, Lovejoy says companies can be
brought into the environmental loop
an approach that, he admits, is perceived
with wariness on the part of many
green organizations. Lovejoy, who in
the course of one day easily moves
from sitting on the dirt oor of a hut
in a Kayap village swatting blackies
to dining with the states governor at a
table dressed with crystal and silver,
argues that sustainable development
BIOLOGIST THOMAS E. LOVEJOY stands

in the rain forest that he rst visited in
1965. His Amazon research and, more recently, his advisory work have increasingly sought to nd economically successful ways of preserving biodiversity.
and the protection of biodiversity must

be practiced at all levels of society.
To this end, he is advising the Brazilian Foundation for Sustainable Development, a consortium of 24 of the largest Brazilian companies, formed after
the 1992 Earth Summit in Rio de Janeiro. The foundation plans to support research that will dene sustainable development and to help companies act
in environmentally sound ways. It will
also nance the Forest Fragments Project. Although the rst step for sustainable development is in local communities, it is not only for them, says Angelo
dos Santos, a foundation ecologist who
has worked with INPA and Lovejoy. Because of global climatic change, we need
to work with big enterprises. We need a
whole new approach.
At a December meeting at Camp 41
the Tiany of sites at the Forest Fragment Project because of its ample hammocks and good cooksLovejoy, Zimmerman, Wilson and biologists from
Brazil, Panama, Costa Rica, Peru and
Mexico met to evaluate the project and
to consider future studies. David B. Clark
of La Selva Biological Station in Costa
Rica pointed out that the only large
tracts of forest left in his country are in
parks. I am trying to be philosophical
about it, Clark says. Most reserves in
South America are going to be in this
situation. Development is going to be
right up against the protected areas.
As the scientists walked through the
rain forest, noting dierences and similarities between this region and those
they study, it was apparent that no one
was able any longer to devote attention
exclusively to the study of biodiversity.
Many of us were not trained as administrators but as research biologists, the
Smithsonians Wilson comments. Yet
we nd ourselves, out of necessity, doing conservation work. If they do it well
enough, human beings and the rest of
Amazonian nature may yet nd a way
to survive together. After all, in the Amazon you look for one thing and keep
nding others.
FURTHER READING
THE FATE OF THE FOREST: DEVELOPERS,
DESTROYERS AND DEFENDERS OF THE
AMAZON. Susanna Hecht and Alexander
Cockburn. HarperPerennial, 1990.
BIODIVERSITY. Edited by E. O. Wilson.
National Academy Press, 1992.
THE LAST NEW WORLD: THE CONQUEST
OF THE AMAZON FRONTIER. Mac Margolis. W. W. Norton and Company, 1992.
SUSTAINABLE AGRICULTURE AND THE ENVIRONMENT IN THE HUMID TROPICS. National Research Council. National Academy Press, 1993.
99
SCIENCE AND BUSINESS

Try, Try Again
of this new generation of antibodies

against a variety of diseases.
The major obstacle to medicinal use
of murine antibodies proved to be the
very immune systems they were intended to bolster. Many patients produce
human antimouse antibodies ( HAMA )
when they are given shots of murine
monoclonals. The ensuing battle between the mouse and human proteins
reduces the drugs eect, quickly clearing it from the bloodstream and often
causing allergic reactions. Another problem, explains Robert A. Fildes, chief executive ocer of Scotgen Biopharmaceuticals, is that when you put a mouse antibody into a human being, it does not
turn on the human immune system as
eectively as a human antibody would.
Human antibodies have proved dicult to make outside the body, however.
Mouse monoclonals, the Nobel Prize
winning discovery of Csar Milstein and
Georges Khler in 1975, are made by
fusing a mouses B lymphocytea white
blood cell in the spleen that produces
antibodies but is dicult to culture
with a cancerous melanoma cell to produce an immortal hybrid cell, or hybridoma. Such cells create antibodies and
replicate furiously in a fermentation
Making antibodies more useful

by making them more human
LONED ANTIBODIES PROMISE MEDICAL REVOLUTION, screamed the
front-page headline of the New
York Times science section on August
5, 1980. Throughout the next decade,
dozens of biotechnology rms struggled to turn monoclonal antibodies
(MAbs) harvested from mouse tumors
into new drugs to treat everything from
cancer to bacterial infections. They have
yet to deliver. Despite years of intense
research, only one MAb-based therapeutic drug, OKT3 from Ortho Pharmaceuticals, has gained Food and Drug
Administration approval for the U.S.
market.
But enthusiasm for monoclonals
seems to be enjoying something of a renaissance. A handful of companies are
now perfecting techniques for manufacturing fully human and humanized
antibodies that promise to surmount
some of the problems that have plagued
murine monoclonals. They are buoyed
by results from early clinical trials
tank. Murine antibodies can thus be

mass-produced for nearly any antigen
injected into a mouse. Unfortunately,
however, human B cells are loath to fuse
with cancerous cells.
The ability of murine MAbs to home in
on a particular cellular target won them
a niche in the lucrative market for diagnostic tests and cancer imaging. Monoclonals that bind to proteins on the surface of cancer cells can be attached to
radioisotopes to help doctors see clusters of malignant cells invisible to xrays, CT scans and magnetic resonance
imaging. And because imaging agents
need be given just once or twice, HAMA
reactions are less of a problem. Even so,
only one imaging agent, a murine monoclonal for diagnosing colon cancer, has
been approved to date. Look back over
10 years: there has been huge success
with mouse MAbs as in vitro diagnostic
testsits a billion-dollar industry worldwide, Fildes says. But in that same period, virtually every company that has
tried to develop mouse antibodies for
therapy has failed miserably.
Witness Xoma, which was within
shouting distance of winning FDA approval for its MAb-based sepsis drug
last year before nal clinical trials failed
Antibody Anatomy
ntibodies are variations on a theme. Natural antibodies

(left) are proteins composed of amino acid chains
twisted into complex three-dimensional shapes. They lock
onto antigenstarget moleculesthat match the shape
and chemistry of the binding sites (shown in red) on their

arms. Geneticists can convert mouse antibodies to various humanized forms (right) by combining the genes for
murine binding sites with those for a human framework.
NATURAL ANTIBODIES
GENETICALLY ENGINEERED ANTIBODIES
RADIOISOTOPE
JARED SCHNEIDMAN/JSD
PROTEIN DESIGN LABS, INC.
MOUSE (MURINE)
ANTIBODY
CHIMERIC
ANTIBODY
ANTIBODY
FRAGMENT (Fab)
RADIOLABELED
ANTIBODY
GRAFTED
MURINE
BINDING
SITES
HUMAN
ANTIBODY
HUMANIZED
ANTIBODY
SINGLE-CHAIN
ANTIBODY
BI-SPECIFIC
ANTIBODY
101
Monoclonal Antibody Therapeutics in Development

Murine Antibodies
IDEC Pharmaceuticals: Phase III trials* for B cell lymphoma.
ImmunoGen: Phase III for various cancers; phase I and II for small-cell
lung cancer.
Immunomedics: Phase I and II for colorectal cancer and B cell lymphoma.

NeoRx: Phase I for lung and ovarian cancer.
Humanized Antibodies
Cytogen: Preclinical evaluation for breast, lung and ovarian cancer.
Genentech with NeoRx: Phase II for breast and ovarian cancer.
IDEC Pharmaceuticals: Preclinical testing of primatized macaque/
human antibody for rheumatoid arthritis.
Protein Design Labs: Phase II for graft versus host disease; phase I and
II for acute myeloid leukemia; preclinical tests for lung, colon and breast
cancer, inflammation, lupus and multiple sclerosis.
Scotgen Biopharmaceuticals: Preclinical testing of 12 antibodies against

cancer, cytomegalovirus, HIV and atherosclerotic plaque.
Human Antibodies
Protein Design Labs: Completed phase I and II for retinitis and hepatitis
B; preclinical tests for herpes and chicken pox.
*Marketing approval of drugs by the Food and Drug Administration generally requires three
phases of human clinical tests: the first for safety and toxicity, the second for efficacy on a
small scale and the final for large-scale, statistically significant effectiveness.
to show sucient eectiveness. Xoma

has since lost its CEO, shut down clinical
trials of four other products and slashed
its sta by 25 percent. Nevertheless, the
size of the pot of gold at the end of this
rainbow has induced many companies to
continue working on therapeutic MAbs;
industry analysts peg the potential market at $3 billion for cancer alone.
Most biotech rms now seem to agree
that therapeutic antibodies will have to
beat HAMA to succeed. Here is where it
helps most to be human. Actually, an antibody need not be entirely human to
fool a patients immune system. Simply
replacing the constant region, or stem,
of the Y-shaped murine antibody with
its human equivalent produces a chimeric antibody that may sneak through undetected. Greg Winter of the British Medical Research Council improved on that
idea in the mid-1980s by splicing the
genes for just the binding sites of a
mouse antibody alongside the genes for
the rest of a human antibody to make a
humanized antibody.
Getting the humanized version to
lock onto the same antigens as the original takes some work, however. You
sometimes have to tinker with the amino acid sequences in the framework of
the humanized antibody in order to get
the binding sites in the right threedimensional form, says Fildes, whose
company has used Winters technique
to humanize more than 20 antibodies
taken from mice. The trick is to know
which ones to changeits really an art.
102
A start-up company in California is

making a science of that art. Protein Design Labs (PDL), which went public early in 1992, has used protein-modeling
software to compare the amino acid sequence of the binding sites with a library of other antibodies whose spatial
structure has been determined by x-ray
crystallography. Once the model picks
out which mouse amino acids to keep
and which to replace, PDL synthesizes
the new genes, inserts them into mamalian cells and cultures the cells to
produce the humanized antibodies.
So far PDL has humanized more than
a dozen antibodies against viruses, cancers and even adhesion molecules that
play a role in inammation. In the three
clinical trials PDL has conducted to date
with its humanized MAbs, we have not
seen any HAMA response, claims Peter
Dworkin, PDLs spokesperson.
The fastest arrows in PDLs quiver are
not its 90 percent human monoclonals,
however, but entirely human MAbs produced by cell lines that the company
purchased from Sandoz in April. From
the spleen of a car crash victim, Sandoz
managed to isolate human B cells that
produce antibodies against cytomegalovirus, the cause of an eye infection in 20
percent of AIDS patients that often leads
to blindness. By fusing this B cell with a
human/mouse hybridoma, Sandoz created triomas, which thrive in culture
and produce many dierent antibodies
against the virus. For $4.5 million, PDL
got four such triomas, two of which
anti-cytomegalovirus and antihepatitis

Bhave already completed initial clinical trials for toxicity and ecacy.
Triomas still depend too much on
luck and labor to be widely useful. GenPharm International thinks it has a better way to produce human MAbs. Using
its experience producing transgenic animal models for cancer and cystic brosis research, the company engineered a
mouse that produces only human antibodies. Researchers turned o genes for
each of the mouses two native antibody
chains and then inserted and activated
genes for human antibody chains. The
result is a mouse strain that is able to
make any human antibody, including antibodies to human antigens, says Jonathan MacQuitty, the companys CEO.
GenPharm also recently announced
that after four years and $10 million
it had created a second line of human
MAb-producing mice by using articial
chromosomes genetically engineered in
yeast cells to transfer huge blocks of
DNA from humans to the rodents. Unlike the rest of GenPharms products,
we wont sell these mice, says Howard
B. Rosen, director of corporate development. We will work with researchers
on a collaborative basis. Its sort of the
mouse that lays a golden eggwed hate
for people to just run o with it.
Researchers who have isolated a human antibody may soon use a technique rened at the Scripps Research
Institute to mass-produce it in bacteria.
Scripps molecular immunologist Dennis R. Burton amplies the antibodys
DNA and clones it into display viruses. Each virus is a little unit, with one
antibody gene inside and the protein
coded by that gene on the outside, Burton explains. When poured onto a plate
coated with the target antigen, viruses
containing antibodies that match that
antigen stick. The rest are washed away.
Burton then infects Escherichia coli with
the selected viruses, converting them
into bacterial antibody factories. The
process does have a serious limitation,
Burton concedes. Because bacteria lack
the sugars needed to make the stem of
the antibody, they reproduce only the
armsa fragment known as Fab. Although this fragment alone will neutralize a lot of viruses, we dont know
how it works, he admits.
Even so, several biotech rms are looking at antibody fragments with newfound respect. A whole antibody is too
large to penetrate the surface of many
tumors; fragments may be able to go
places whole MAbs cannot. But fragments are removed from the bloodstream in hours rather than days. Its
a two-edged sword, says Thomas J.
McKearn, president of Cytogen. An in-
tact MAb will pass through the circulatory bed of a tumor many more times
than will a fragmented version.
Researchers at Scotgen and elsewhere
are also joining two dierent antibodies
to make X-shaped bi-specic antibodies that attach to a diseased cell on one
end and an immune cell on the other,
dragging immunologic soldiers to the enemy. Medarex recently began clinical trials on such a drug for breast and ovarian cancer. The problem is making these
things, Fildes complains. Its a long,
tortuous and very expensive process.
Indeed, all these new technologies are
considerably more expensive than churning out old-fashioned murine MAbs. And
there are no guarantees. Centocors human antisepsis monoclonal fared little
better than Xomas murine version in
nal ecacy trials. After investing $250
million in development, Centocor has
seen just $31 million in sales. McKearn,
for one, takes this as a sign that it is too
early to dismiss mouse monoclonals.
The literature is replete with examples
where people have ignored common
sense and simply jumped on the bandwagon of humanization, he says. But of
course sometimes the bandwagon goes
straight to the party. W. Wayt Gibbs
Earcons
Audification may add
a new dimension to computers
n the early days of computers, an

enterprising programmer would often place a portable radio on top of
one of these primitive machines. The
internal processing circuitry would act
as a transmitter whose signals could be
picked up a block away if a radio were
tuned to the proper frequency. Each program would send out a characteristic
stream of whistles and clicks that indicated that the software was running as
it should.
These days, history is repeating itself. Software engineers struggling to
program parallel processors have enlisted the help of electronic sound synthesizers to decipher the eectiveness
of their improvisations. A bell tone or a
violin note may reveal that processors
13 and 15 are not running, letting the
programmer then rework the software
to keep them busy.
Computer scientists, in league with
electronic musicians, have begun to add
a sound track not only to the often tan-
gled workings of parallel programs but

also to huge stores of data that comprise medical images, partial dierential equations and even census gures.
Last fall, for example, 35 followers of
what has come to be called sonication,
audication or, even less sonorously,
data-driven sound descended on the
Santa Fe Institute for the rst International Conference on Auditory Display.
This tiny community of researchers
is trying to bring another dimension to
the communication between machine
and user. Just as the ear can distinguish the oboe from the other instruments in a symphony orchestra, cybernetic audiophiles believe sound may
complement or, in some cases, even surpass visual graphics.
Acoustic displays may reinforce visualization or add new information. A
change in pitch, duration or loudness,
for instance, corresponds to a variation
in a value from the collected data. Some
workers assert that existing technology
severely overloads the visual. Yet, they
point out, computers ignore the sense
of hearing except for an occasional wild
eep or bong from a neglected piece of
software. Communicating information
by means of an aural cue is a legacy of
Putting a Spin on Parasites
drian Parton used to spend his days pondering molecular biology. Then, in 1991, he heard a talk given
by biophysicist Ronald Pethig of the University College of
Wales, in which he described how electric fields can be
used to manipulate particles. That set Parton, who works
for Scientific Generics, a technology consulting firm in
Cambridge, England, thinking about rotating fields.
A rotating electric fieldeasily produced from four electrodescreates a torque on small particles that can make
them spin if the frequency is right. Parton has shown that
the effect can function as the basis for an extremely sensitive assay to detect microscopic parasites and even single biological molecules. Co-inventors Parton and Pethig
call it the electrorotation assay.
Scientific Generics has now embarked on a joint venture with Severn Trent, a water utility in England, to develop and license equipment based on the principle. One of
the first applications that will be pursued by the new company, Genera Technologies, is an apparatus to assess water-supply contamination by Cryptosporidium, a common
waterborne parasite that caused an outbreak of sickness
in Milwaukee, Wis., this past spring. Parton maintains that
the assay could reliably detect other small beasties as
well, such as Giardia, an intestinal parasite.
What is more, Parton says, the assay can outperform
ELISA (which stands for enzyme-linked immunosorbent assay), a widely used laboratory technique for detecting small
quantities of specific proteins. Genera plans to pursue several applications in the health care and food industries.
To detect Cryptosporidium, Parton first concentrates the
particulates from a 1,000-liter sample of water into two
microliters. He then adds microscopic beads coated with

an antibody that binds to the organisms oocysts (the form
found in contaminated water). The rotating field makes
the oocysts bound to the beads spin. The number of spinning oocysts can be recorded electronically.
At low frequenciesaround 1,000 hertzlive specimens
rotate in the opposite direction to the applied field. Dead
specimens, in contrast, rotate at higher frequencies but
spin in the same direction as the field. The assay can thus
assess the viability of Cryptosporidium as well as its numbers. At frequencies of about a megahertz, you are looking at the ionic properties of the membrane, Pethig explains. At still higher frequencies, you are looking inside
the cell. A test that normally takes eight hours can be done
in 30 minutes by electrorotation assay, Parton declares.
To detect targets smaller than oocysts, such as bacteria
or protein molecules, Parton adds larger antibody-coated
beads. The beads themselves are observed through a microscope. Any targets present bind to the beads, which
makes them spin. Parton has discovered that in samples
with very few targets, only a few beads rotatebut those
that do so rotate very fast. That makes the assay exquisitely sensitive.
Parton says he has detected viral proteins in amounts
as small as a billionth of a billionth of a gram. Commercial
ELISAs now available are about a million times less sensitive than the levels that Parton claims. If beads are coated
with nucleic acids instead of antibodies, Parton speculates, the technique could rival the polymerase chain reaction, currently the most sensitive test for particular sequences of genetic material.
Tim Beardsley
103
the Geiger counter, sonar systems and

the alarms that ring in the cockpits of
jet ghter aircraft. Engineers in the eld
already see far beyond that simpleminded prompting. Sound may be the best
means of depicting the ordering of several overlapping events, say Joan M.
Francioni and Jay Alan Jackson of the
University of Southwestern Louisiana.
The husband-and-wife team have produced scalelike tones that will let a parallel programmer tell when a processor
becomes active or when it receives a
message from another processor. Their
choice of synthesized sounds even won
plaudits from music acionados at various conferences.
Other workers explore complementary
strategies, in which sound and graphics reinforce each other. At the University of Massachusetts at Lowell, Stuart
Smith, a professor of computer science,
and his colleagues have combined visual and sound displays to exhibit variables ranging from infrared radiation
from satellite imagery to heat ows in
dierential equations. Tiny zigzagging
lines represent up to 20 variables graphically. Each connected segment, the zig
or the zag, is a distinct variable, its value
represented by the angle. Together they
create an image whose texture varies
across the screen like a eld of rippling
wheat.
The impression of an undulating topography is reinforced by tones that
are reminiscent of Karl-Heinz Stockhausens greatest hits. Like an auditor
of one of the avant-garde German composers works, the observer does not inspect each value separately but looks
for clusters of similar sounds and visual cues. Staccato clicks and pops, translated from the brightness values for two
merged magnetic resonance images, are
supplemented by a crackling that sounds
like an egg frying as a cursor is passed
over an area showing a brain tumor.
Sound can pick out information that
is dicult to discern in the image. Each
person in a sampling of 1980 census
data for scientists, engineers and technicians becomes a jagged line, whose segments represent marital status, level of
education, sex and birthplace. These
same variables can be represented by
changes in sound. The category for sex,
for example, is delineated by an electronic violin. A long, bowed note reveals a
male; a high-pitched pizzicato note is
the female. (The political correctness of
the sundry tones has yet to be debated
by technical standards groups.) As the
mouse moves the cursor up the screen
the big wage earners reside at the top
the continuous tone from the bow gets
well exercised. A short pluck in the upper corner, a lone woman, demonstrates
106
how sound can pick out dierences in

the data that would be dicult to distinguish within the mass of visual cues.
Smith and other researchers in this
auditory enclave are still asking themselves how much information can be
clearly understood by listening to a data
base. Training a listener to identify the
single highly paid woman or heat-ow
value may be a pointless pursuit. What is
sought, instead, is the distinctive phrasing in the ow of synthesizer notesa
data melodythat will enable the listener to, say, locate a tumor.
The eld is too young to have developed the conventions that would allow
a set of characteristic notes to be recognized as cancerous tissues or heat
ows. Smiths group recently received
a $147,000 grant from the National Science Foundation to work with an experimental psychologist on determining how well the ear can hear the subtle changes in pitch, loudness and other sound traits within an aural display.
You can arrange a clever scheme for
visualization and audication, but the
question is how will people perceive it,
Smith says.
Sound-display artists hope the ear will
in some situations surpass the eye
because of its superb ability to extract
meaning from a noisy environment.
Gregory Kramer of the Santa Fe Institute and W. Tecumseh Fitch of Brown
University found that a group of university students who were assigned to
role-play as medical doctors were able
Keeping the Sun Shining on British Technology

ichael Faraday, on being asked by a prime minister of the day what
good were his researches into electricity, is said to have answered, Sir,
one day you will tax it. The successors to Faradays interrogator may
need to be reminded of the 19th-century physicists vision. Margaret Thatcher,
despite her professed admiration for Faraday and her chemistry degree, gave
Adam Smiths invisible hand such a free rein during the 1980s that spending on
research and development in Britain is now less, in relation to gross domestic
product, than it is in the U.S., France, Switzerland, Germany or Japan.
A policy that prohibited government support for near-market research and
development further battered Albions technical prowess. The hope, of course,
was that industry would pick up anything worthwhile that the government let
drop. But the plan didnt work. The Confederation of British Industry (CBI) notes
that U.K. patenting activity decreased by 6.4 percent between 1985 and 1990,
for example, while it increased in Germany, the U.S. and Japan.
Britains industrial decline now seems to have persuaded the prime minister,
John Major, that it might be worth giving the Newtons, Watts and Darwins of
the next century some encouragement. The government has for the past several months been conducting a full-scale policy review aimed at goading the scientific goose into laying more golden eggs. As in the U.S., where the Clinton administration has embraced technology policy, there is increasing talk of a government role in fostering innovation.
The man in charge of the exercise is William Waldegrave, a soft-spoken humanities graduate who enjoys the formal title of Chancellor of the Duchy of
Lancaster. More meaningfully, he heads a new independent office of science
and technology, the first cabinet minister with that portfolio since the 1960s. In
June, Waldegrave was expected to publish his proposals for a complete shakeup of the British scientific enterprise.
His work is cut out for him: Waldegraves office has received more than 800
pieces of advice. The most common complaint is one familiar in the U.S.: Britons, academics and officials like to say, are as good as they ever were at doing
science, but industry fails to develop the commercial spin-offs. Industry
charges that the government should be doing more to help. We see no sign of
strategic direction in the way resources are allocated, said Howard Davies, director general of the CBI, last November.
Interviewed in his Whitehall office, Waldegrave points out that counts of citations of published papers indicate that Britain is second only to the U.S. in scientific standing. Britain has a powerful science base, he says. But, he admits,
we have continuing uncertainties about whether were good at relating it to societal goals. Waldegrave, who has previous experience in industrial research as
well as in science policy-making, contends that leaving decisions to industry
has given British technical expertise a certain random, spotty character.
Much in favor with industry is the suggestion that the U.K. adopt a R&D tax
credit similar to that of the U.S. Yet Waldegrave is pretty skeptical about using
to follow multiple aural variables on a

computerused to represent body temperature, heart rate, blood pressure and
ve other variablesbetter than they
could on a graphic display.
Indeed, the next-generation computer operating system may enhance visual
icons with sounds. When someone else
accesses your computer, these earcons,
as they are whimsically called, might reproduce the tap-tap of approaching footsteps. If you saw things through the
same cone of vision as your narrow computer screen, you couldnt get a drivers
license in the United States, says William Buxton, a professor of computer
science at the University of Toronto and
a researcher at Xeroxs Palo Alto Research Center. Were already overload-
ing the visual channel. One key solution

is o-loading appropriate things to the
auditory channel.
Some things are just waiting to sound
o. Chris Hayward, a seismologist at
Southern Methodist University, wants to
delve into what can be learned by speeding up and listening to the various subsonic shock waves of an earthquake. R.
Kevin McCabe, a computer scientist at
the National Aeronautics and Space Administration Ames Research Center in
Mountain View, Calif., has written a programa software microphonethat
does digital sampling and playback of
a computer simulation of acoustic data
from a jet turbine blade. The days may
be gone when a computer was meant
to be seen and not heard. Gary Stix
OFFICE OF PUBLIC SERVICE AND SCIENCE
tax policy to stimulate economic activity. The government has also dismissed as too expensive proposals
that it fund Faraday Centrestechnology-transfer institutes modeled on
the German Frauenhofer Institutes
to foster mixing among academic and
industrial scientists.
Rather Waldegrave sees a case for
improved direct support for strategic research in key areas, that is, work
somewhere between completely unfocused blue sky investigation and
product development. And he would
like to see a little more coordination
and a little more structured interchange between government, academia and industry.
There are misgivings about the applied focus of the governments re- WILLIAM WALDEGRAVE is presiding over
view in academic quarters. And the rst complete shake-up of British sciWaldegrave insists he can promise ence in nearly two decades.
no increases in support for civilian
research. The $1.7 billion planned for the 199394 fiscal year was intended to
be a modest boost, but scientific organizations argue that the devaluation of
the pound turned it into a decrease.
The budget strictures mean there is likely to be no quick relief for impoverished scientists. Pay is almost unbelievably low by U.S. standards. A trainee researcher relying on the Science and Engineering Research Council for support,
for example, has to survive the first three years after graduation on a stipend of
less than $7,000 a year, a third of the national average wage. The government
seems to have accepted the need for a remedy, and William D. P. Stewart, who is
the governments chief scientific adviser, is sympathetic to proposals to make
blue sky grants for young researchers easier to obtain. But, he warns, any increases in support for research will have to be coupled with demonstrated increases in efficiency.
Britain will be looking for ways to improve efficiency in international big science projects as well, says Stewart, who is a botanist. He declares that the
country cannot be party to the construction of expensive research facilities that
stem from national pride (a dig at the U.S.s Superconducting Super Collider).
Rather he favors planned international collaborations.
Increasingly, British scientists say they are willing to recognize the financial
realities of research, and some are even anxious to see their work commercialized. But dispelling the British economic malaise will be a tall order, even for a
scientific magic wand.
Tim Beardsley
Practical Fractal
Mandelbrots equations
compress digital images
hen IBM researcher Benoit B.

Mandelbrot published The Fractal Geometry of Nature a decade ago, few imagined that the beautiful, innitely detailed designs he called
fractals might one day also improve television reception and help make pictures
a part of everyday computing. Two who
did were Michael F. Barnsley and Alan
D. Sloan, mathematicians at the Georgia
Institute of Technology. They reasoned
that because images of the real world
tend to consist of many complex patterns that recur at various sizesin other words, fractalsthere ought to be a
way to translate pictures into fractal
equations. Images so coded would require fewer data and thus less disk space
to store and less time to transmit.
By 1987 the researchers had worked
out enough details to patent their idea,
garner $500,000 in start-up capital from
Norwegian investors and form Iterated
Systems in Norcross, Ga. But turning potential to prot proved dicult. Iterated found a formidable competitor in
JPEG, an image compression technique
developed by the International Standards Organizations Joint Photographic Experts Group. JPEG uses a well-understood mathematical procedure, the
discrete cosine transform, to compress
les by 90 to 95 percentequivalent to
compression ratios of 10:1 to 20:1. JPEG
enjoys another advantage: because it is
an industry standard, it is free.
The mathematics behind fractal compression, on the other hand, has only recently left the crib. Relatively speaking,
very little research has been done to
date, Barnsley admits. Two or three
years ago, when people looked at my pictures, they said, This stu looks terrible; it isnt going to workyoure dreaming. But we bet on continual dramatic
improvement, and weve seen it.
Indeed, fractal compression has at
last begun to challenge JPEG in the three
areas that matter: speed, quality and
strength. To a certain extent, the three
are mutually exclusivestronger compression yields a smaller image le but
takes longer and creates more artifacts.
JPEG takes as long to decompress an image as to compress it, and it produces
blocky pixelation as compression ratios rise above about 20:1.
Iterated takes a dierent tack. Its algorithm works by treating an image as
a jigsaw puzzle composed of many overlapping pieces, some of which are similar. The software takes each piece and,
107
ITERATED SYSTEMS Courtesy of The Postcard Factory
using a fractal formula, transof entrepreneur, he explains.

forms its shape, size and color
Hell come to meetings and give
until it matches another part of
a sales pitch for 45 minutes inthe picture. To do this for each
stead of a mathematics talk. This
piece of the puzzle takes some
drives mathematicians up the
time; fractal compression is quite
wall.
slow. But it produces a much
Academic ire does not seem
smaller image le, which conto be hindering Iterateds busitains just the numbers needed
ness, however. Five years and
to specify the mathematical re$18 million after its inception,
lations between the pieces, and
Iterated nally started making
not those needed to actually
money. At the end of last year,
draw each piece.
we passed breakeven, and cash
This approach has some
revenues have increased by over
unique advantages. Reconstruct10 percent a month each of the
ing the image from the numbers
past six months, boasts Louisa
is relatively quick work: IteratAnson, a vice president. Our
eds video software can decomgoal is to go public in 1995. The
press tens of images each seccompany has grown to more
ond without any special hardthan 56 employees (32 of whom
ware. Quality depends on the acwork in research and developcuracy of the numbers; giving
ment), oers half a dozen or so
the compression algorithm more
image compression products and
time or a dedicated computer
is working to establish a footprocessor thus yields better pichold in the nascent computer
tures. And the artifacts generatvideo market.
ed at higher compression ratios
Video will pose a greater chaladd an impressionistic blur, ralenge than still-image compresther than blockiness, to the imsion. The current video compresage. What is more, because fracsion standard recently nalized
tal images are encoded by equaby the Motion Picture Experts
tions, they have no inherent
Group (MPEG) is well entrenched
sizean attractive quality to
and near a second version. AT&T,
publishers who want their prodMicrosoft, Philips and numerous
ucts to work as well on a laptop FRACTAL COMPRESSION technique patented by Iter- other giants have announced
screen as on a 21-inch monitor. ated Systems reduces data by translating a picture into plans to sell MPEG-based video
Barnsley expects that in the long fractal equations. The higher the compression, the more decoders within the year.
term, such resolution indepen- the image is distorted. Clockwise from top left: King of
Barnsley cares little. If nothspades at original, 2.4-megabyte size, then compressed ing else, he maintains, the resodence will prove irresistible.
So far, however, most of Iter- to 129, 63 and 31 kilobytes.
lution independence of fractals
ateds customers have chosen
will triumph in the end. Iteratfractal compression for its brute pow- eectively across public telephone lines, eds long-term goal is to produce MPEGer. Microsoft licensed Iterateds software says company president Steve Swift. Tay- compatible chips that use fractals to
to compress the nearly 10,000 pictures son Information president Peter A. Rich- improve television reception and from
it put on Encarta, its CD-ROM encyclo- ardson says his company routinely ob- there push into broadcasting. To that
pedia, because fractal compression pro- tained 100:1 compression when it added end, the company sought and received
duced much more aesthetically pleas- fractals to its education software, used a $2-million grant from the National Ining results than JPEG at high compres- by the University of Guelph in Ontario stitute of Standards and Technology to
sion ratios, says Jay Gibson, who man- to create more than 500 undergraduate help it develop a fractal compression
ages the project. Racal Radio in Read- course modules. We chose fractals be- chip for high-denition television. Once
ing, England, put fractal compression cause JPEG is pushing the limit of its ca- we get the reception circuitry in place,
into its PICTOR system for similar rea- pabilities, Richardson says. Fractals were in a position to oer better transsons. We wanted to send data les over can go and go and go.
mission as well, Barnsley says. Its kind
high-frequency radio, which is currentHow far they can go is a matter of of a Trojan-horse strategy.
ly limited to some hundreds of bits per some dispute. Barnsley has claimed efIn the meantime, Barnsley sees a niche
second, explains Peter W. Goddard, Ra- fective compression ratios of up to in CD-ROM video for computers. MPEG
cals marketing manager. Using fractals, 2,500:1. But sometimes people expand requires special decompression chips
we were able to get a 768-kilobyte pic- the image and then claim compression to play video. Iterated has teamed up
ture le down to 10 kilobytes or low- relative to the expanded image, notes with TMM in Thousand Oaks, Calif., to
ersmall enough to transmit in two to Michael F. Shlesinger, who as director of develop software that plays broadcastthree minutes.
physics at the Oce of Naval Research quality video straight from a CD-ROM.
Other companies are attempting even funded Barnsleys early research. Its Were going out full blast at Comdex
higher compression ratios. Kalman Tech- just a trick. That trick has some math- [the computer trade show] with video,
nologies used fractal compression to ematicians crying fraud and heaping Barnsley beams. By this fall well be a
add video mail to its HydraWorks com- scorn on Barnsley. Shlesinger attributes major player in the desktop computer
munications software. We found com- the controversy to a conict of person- video market. But rst Barnsley and
pression ratios of about 600:1, which is alities. Barnsley stopped playing the role his mathematicians must nish their
what we need to send video mail cost- of professor and started playing the role homework.
W. Wayt Gibbs
108
THE ANALYTICAL ECONOMIST

Health Care without Perverse Incentives
o traditional economists, the U.S.

health care system looks as if it
had been designed to produce
market failure. Customers have relatively little information about the product
they are buying and must rely largely on
sellers for advice. Sellers are paid according to how many procedures they perform, regardless of their eectiveness.
The price consumers pay for services
bears only a distant connection to the
amount they use. The middle and upper
classes eectively buy their health care
in prepaid annual chunks, either directly or as part of their compensation at
work; the elderly and the poor have
theirs paid for out of tax revenues. In
all cases, the sum at stake when people
decide whether to see a doctor is only a
small fraction of the actual cost.
These perverse incentives have caused
health care expenditures to soar and
have led providers to exclude many
people from the market. The U.S. now
spends $800 billion a year to leave close
to 40 million citizens without adequate
care. This spending diverts resources
from other issues on the national agenda and hurts the countrys competitive
position. For example, manufacturers
complain that health insurance premiums add more than $2,000 to the price
of an average automobile.
To combat these escalating costs, the
Clinton administration has gathered
more than 500 experts to create a new
health care system. The central feature
of the plan is supposed to be managed competition, in which groups of
consumers (either companies or individuals) bargain with medical consortia
to purchase care at a reasonable price.
Many economists, however, are skeptical. They argue that the task force is
ignoring fundamental economic principles, which put ecient, competitive
health care markets in the same class
as powdered unicorn horn. There is
no way of arranging health care nancing that does not lead to perverse incentives, contends Uwe E. Reinhardt of
Princeton University: We proved that
years ago.
As long as doctors are paid on a feefor-service basis, critics argue, they will
overtreatand fear of malpractice litigation tends to encourage unnecessary
tests and procedures. As long as pa-
tients rely on insurance, they will demand more treatment. Economic theory
suggests that forcing patients to cover a
percentage of their medical bills would
curb spending. But, as Henry J. Aaron
of the Brookings Institution observes, if
patients paid enough to make them really cost-conscious, a single bout of serious illness would spell nancial ruin
(as it does today for the uninsured).
Furthermore, Aaron points out, most
health care dollars are spent during episodes of acute illness, when patients
have little control over charges. Indeed,
in another example of perversity, conventional limits on out-of-pocket costs
may mean that patients have more nancial incentive to postpone a checkup than to avoid a major operation.
Health maintenance organizations
(HMOs), which oer to meet all of a
consumers medical needs for a xed
fee, have been cited as a major ingredi-
Ecient, competitive
health care markets may
be in the same class as
powdered unicorn horn.
ent in the managed competition plan.
But, as Reinhardt and others note, these
groups create a dierent and opposing
set of perverse incentives for physicians
and patients. Doctors receive a xed salary and so do best nancially by oering as little care as possible; patients
have paid up front and so prot (if that
is the right term) by insisting on maximum treatment. The HMO itself has an
incentive to sign up a maximum of subscribers and to minimize the time each
one spends with a physician.
Given this set of imperfect alternatives, critical economists say, the objective of health care reform must be to
downplay the inherent aws in this particular market rather than making futile
attempts to eliminate them. But is this
insight useful? Most of the suggestions
they oer are modications to the current status, not radical departures in the
vein of, say, Great Britain or Canada.
Reinhardt proposes reducing perverse
incentives by setting physicians incomes according to the number of pa-
tients they treata little like setting up

myriad microscopic HMOsbut allowing patients to choose freely among
doctors, as HMOs do not. The reward
for undertreatment will be balanced by
the threat that patients will take their
money elsewhere.
Although the ability to switch physicians is important, it is not enough: the
cost of nding out whether one has the
wrong doctor is quite high. So the next
step is to monitor the quality of care
and make that information available.
Publicly disseminated physician proles would help patients choose wisely,
encouraging both competition and cost
containment.
Indeed, Alan M. Garber of Stanford
University believes medical quality assurance will eventually reshape the market: We are faced with the problem of
trimming fat from health care, and we
are going to be cutting out treatments
that do some goodbut not very much
and at high cost. Much of the recent
debate over booming expenditures has
focused on the question of whether
high-technology medical interventions
actually improve care.
But studies to determine when expensive medical procedures are vital and
when they are placebos will carry their
own costs. Administrative expenses already eat up nearly one fth of every
health care dollar in the U.S. In addition, as Victor R. Fuchs of Stanford
comments, such numbers will do nothing to answer such crucial questions as
Are dying patients treated with compassion? or How long does a bedridden hospital patient lie in urine before
someone responds to a call? New measures of eectiveness may have to be
developed if humane medicine is not to
fall victim to cost-cutting.
Perhaps the most paradoxical problem in health care reform is that if managed competition or negotiated caps on
medical spending do succeed in controlling costs, policymakers will be stiing one of the past decades most potent sources of new jobs. Health care has
been almost a textbook case of the transition from the smokestack industry to
the service economy. Every dollar spent
on administration is a dollar of income
to someone, Fuchs says. With almost
one seventh of its gross national product dependent on health industries, the
U.S. may be in a classic double bind.
Paul Wallich and Marguerite Holloway
109
MATHEMATICAL RECREATIONS
The Topological Dressmaker
he oor was littered with bits

of paper pattern, pins and snippets of cloth. Jemima Duddlepuck
whistled as she worked on her new summer dress, while in a corner her dog,
Pucciwho was rather lazy and very
fatsnored happily. It was the rst time
Jemima had ever tried to make a dress
with a lining. All that remained was to
sew the two together, so that the seams
ended up inside, between lining and
dress, invisible from the outside.
She looked at the instructions. Turn
the lining inside out. Slip it over the
dress. Machine around the neck and
armholes. Turn the completed dress the
right side out. Ever so gingerly, she
machined around the armholes and the
neck. The hems, of course, were left unattached. Done! Oh, it was going to be
really nice. Jemima began to turn it inside out.
Funny. Was something stuck? No, if
she turned it back to the start, then everything seemed okay. But that nal
twist eluded her. Half an hour later
she hurled the balled-up wad of material into the corner of the room and
burst into tears. All that work for nothing. Then, having vented her frustration, she sat down to think the matter
through. Jemima believed in learning
from mistakes.
Had she made some parts the wrong
shape? No, that wasnt likely. If you
couldnt turn a slightly distorted dress
inside out, then you couldnt turn a perfectly made one inside out either. Whether or not you could turn something inside out was a topological questionthe
answer didnt change if the object was
continuously deformed. That meant two
things. First, there was something wrong
with the instructions in the pattern
bookbecause whatever mistakes shed
made, they had to be very small ones.
Second, there had to be a topological reason why the instructions were wrong.
What could she remember about topology? Well, there was the idea of topological equivalence. Two objects are
the same, or topologically equivalent,
if one can be deformed continuously
into the other. You can bend them and
stretch them and shrink them and twist
them, but you cannot cut them or tear
them or glue bits together that werent
glued together to begin with. That was
110
the basic ideabut what kind of things

did topologists study? There was a lot
of stu about Klein bottles and Mbius
strips, surfaces with only one side. But
that wasnt likely to be relevant to problems about things with two sides. Then
there was some other stu about surfaces with holes. That was more likely
to be useful.
She recalled that hole is a rather ambiguous word in topology. For instance,
people usually say that a donutor torushas a hole in it. But the hole isnt
really in the donut at all. Its just a place
where the donut isnt. Whereas if you
take a car inner tubealso a torusand
cut out a patch of rubber, then you
really do get a hole.
All right, she thought, Ill call the kind
of hole you cut out with scissors a hole,
HANDLE
HOLE
TUNNEL
Some topological terminology
and the kind where the surface just isnt

Ill call a tunnel. One way to create a tunnel is to add a handle to a simpler surface [see illustration above]. By adding
handles one at a time to a sphere, you
rst get a torus, with one tunnel through
the middle. And then theres the twohandled torus, the three-handled torus
and so on. She knew these were often
called the two-holed torus and so on,
but she wasnt keen to use the hole
terminology without being very careful
indeed what she meant.
Now, the instructions were to turn
the lining inside out, so that its outside
was on the inside; machine the neck and
armholes; and then turn both lining
by Ian Stewart
and dress inside out so that the inside

of the lining was next to the inside of
the dressno, the outside of the inside
of the . . . no, the inside-out lin. Oh,
heck. Terminology for sides was going
to be at least as much a problem as terminology for holes.
Start again. The dress has two sides:
the outside, which is what everyone sees
when you wear it, and the inside, which
is what they dont see. The lining also
has an inside and outside, but those are
confusing words. So call the side of the
lining that ts next to the skin the skinside and the other side of it the skirtside because it goes next to the dress.
Well, as mnemonics go, those werent
too bad. It reminded her of George A.
Strongs parody of Hiawatha:
When he killed the Mudjokivis,
Of the skin he made him mittens,
Made them with the fur side inside,
Made them with the skin side outside,
He, to get the warm side inside,
Put the inside skin side outside.
So the sides of the material, reading outward, have to end up in this order: skinside, skirtside, inside, outside.
And they had to be machined so that
the stitches around the armholes and
the neck were hidden between skirtside and inside.
Topologically, both the dress and lining were equivalent to disks with three
holes. The rim of each disk was the hem,
and the holes were the two armholes
and the neckhole. What the pattern book
said was to place the disks so that the
skinside was next to the outside; machine all three holes so that the stitches were on the skirtside/inside surfaces [see top left illustration on opposite
page] and then turn the whole ensemble inside out.
So the question was, what happened?
Well, she thought, you can deform the
whole ensemble topologically until it
becomes a two-handled torus with two
holes cut in it. (To see why, imagine sliding both ends of the two armhole tubes
onto the neckhole tube and then bending the dress hem up and the lining hem
down and shrinking them [see top right
illustration on opposite page].) So the
problem becomes: Can you turn a twohandled torus inside out if there are
two holes in it?
It was kind of hard to think about, so
Jemima started with something simpler.
OUTSIDE
DRESS
INSIDE
SKIRT
SIDE
LINING
SKIN
SIDE
DRESS PLUS LINING is topologically equivalent to two disks,

each with three holes, sewn together along three edges (left).
Can you turn an ordinary torus inside

out when one hole is cut in it? And the
answer, of course, was yes. The idea was
to ip the entire thing through the hole,
but then the handle was like a tunnel
bored into a sphere rather than a handle stuck on the outside. But you could
This conguration in turn is topologically equivalent to a twohandled torus with two holes, hems of dress and lining (right).
imagine sticking a nger into the tunnel and straightening it. Then the part
of the surface outside the tunnel is a
rather distorted handle, which can be tidied up to look like a neater one. Then
you got something that looked just like
a torus again [see illustration below].
So topologically, there was no diculty.

She realized she could play the same
game on the two-handled torus, too.
She sketched a series of pictures of the
process. First, turn the whole thing inside out through its holes, rather like
reversing a sock (or in this case an ankle warmer because of the two holes).
Then the two handles on the outside
became two tunnels on the inside [see
illustration on next page]. But then you
can stick your nger into each in turn
and pull it out to create a handle on the
outside. And then all you have to do
is twist each handle around, and you
end up with the original ensemble, but
turned inside out.
Hmm. You could do it topologically,
then. Strange, because when she tried it
with an actual dress, it didnt work.
6
7
8
How to turn a torus inside out through one hole
111
TO TURN the ensemble inside out, reverse the whole thing. Pull the tunnels out to form
handles. Twist the handles into correct position. Note, however, where seams end up.
neck when she had another thought.

Topologically, each hole was just as
good as the others. You had to start by
machining some seam. The neck was as
good a place to start as any.
Lets experiment, she mused. She
pulled the lining inside the dress, skinside inside, then skirtside next to inside
of dress, then the outside of the dress
well, outside.
Great! She even found that she could
poke the arms of the lining down the
arms of the dress, just where they ought
to end up. The only trouble was, she
couldnt machine around the seam; the
stitches would be on the outside.
So turn just the arm inside out. She
pulled the arm of the lining inside the
dress, turning it inside out like a sock.
Then she did the same with the arm of
the dress. She burrowed into the ensemble from the hem end, ashlight in
hand. It certainly looked as if the arm
could be restored to its correct position. All she had to do was push the sewing machine up inside the ensemble
and then
Stupid. Sheepishly, she backed out of
the heap of crumpled cloth. She reached
in through the neck and pulled the
sleeve out into the open. She ran the
seam through the machine, then turned
it inside out and back to its proper position. Now the dress really was nished. She tried it on and pirouetted in
front of the mirror. Brilliant!
I ought to write a book about this, she
reected. Topology for Tailors. I could
throw in all the party tricks, like taking o your vest without removing your
jacket.
By now her mind was in overdrive.
Suppose you were a Martian with 17
arms. How many hems would be needed to make a Martian ensemble, dress
plus lining? Seventeen? No, just two. Use
the same trick of machining the neck
only and then dealing with each arm in
turn by pulling it through the neckhole,
inside out. Unfortunately, there werent
any Martians. So she wouldnt get to try
it outPucci, stop that, your nose is
all wet! She pushed the little dog gently away.
Say, Pucci had four arms, and its
tail could poke out through the hems.
Jemima took the dogs measurements,
tut-tutting at the gures for the waistline. She cut shapes from remnants of
the cloth, whistled, sewed, twisted and
turned. Yes, even with four arm holes,
the method worked ne. Now to see if
the dress tted the dog.
It tted, in the sense that the measurements were okay. But fat dog into
small hole wont go.
In the real world, there are nontopological constraints.
112
Why not? Dresses, unlike topological

spaces, cant be stretched or shrunk.
They can be crumpled up and twisted
around, though. It was possible that the
material nature of the dress changed
the answer, but she had a feeling that
something both simpler and more fundamental was involved. She picked up
the ruined mess from where she had
ung it and smoothed it out. She could
wriggle her arm between the various
surfaces, and it really did seem as if
there were three handles. But near the
armholes and neckhole everything was
confused, as if the lining and the dress
were getting mixed up. There were three
handles, all right, but they just werent
in the right places.
Aha! Shed forgotten to think about
the seams. They had to end up in the
right places, neatly separating dress
from lining. But did they?
She drew the seams on her pictures.

No, they didnt. When the surface was
stretched, to create the handles from the
tunnels, and then twisted to put the handles in the right place, all three seams
ended up in completely the wrong places. (E. C. Zeeman of the University of
Oxford has proved that there is no way
to make them end up in the right places.)
So that was the topological obstruction
to turning the ensemble inside out. You
couldnt do it if the seams had to end
up where they started.
And that gave her another idea. Maybe the ensemble wasnt ruined after all.
She could unpick the stitches and try
to nd some other method of making
the seams. She would be on her own,
thoughobviously, the instructions on
the pattern were crazy.
Jemima unpicked one armhole, then
the other. She was about to start on the
BOOK REVIEW
by Gerard Piel
Can History Stop

Repeating Itself ?
PREPARING FOR THE TWENTY-FIRST CENTURY, by Paul Kennedy. Random House,
1993 ($25).
114
CLIFFORD HAUSNER Tony Stone Images
n 1988 Paul Kennedy welcomed the

Bush administration into oce with
the advice that the U.S. had begun
to follow the British Empire into the inevitable second phase of The Rise and
Fall of the Great Powers, as his popular book was entitled. Now the J. Richardson Dilworth Professor of History at
Yale University welcomes the Clinton administration with the word that there is
more to history than history.
In Preparing for the Twenty-First Century, he asserts that political leaders
must reckon with three crises approaching from outside of history as it is usually told. Kennedy sees them, inextricably interconnected, as the current doubling of the world population, the peril
in which the crowding human presence
places the global ecosystem and the disruption of social institutions as well as
the environment by technologies that
are taking over life functions, including
those of the human nervous system.
It augurs well that a historian should
thus bring into his reckoning the longterm continuities of the human experience. With better comprehension of
those realities, historians and the political leadership they seek to instruct
might begin to nd ways to bring history to stop repeating itself.
In part one of his book, Kennedy gives
the current exigencies a compelling presentation. He states the predicament of
population growth in a single, trenchant
paragraph. Humankind numbered one
billion at the end of the 18th century,
when Thomas Malthus rst estimated
its size and made his familiar prediction of the cruel consequences to follow from what he perceived to be its
inexorable and accelerating growth. In
fulllment of Malthusian misgivings, the
number of the earths inhabitants doubled in the next century and then doubled again in the next half century. By
1990 it had increased to 5.3 billion.
Having given Malthus his due, Kennedy points to a new phenomenon in
human population dynamics: increases
have all but halted in the industrial countries. Even among todays fast-expand-
EXTENSIVE SLUMS in Rio de Janeiro kept the 1992 United Nations Conference on
Environment and Development conscious of the worlds most serious problems.
( The photograph does not appear in the book under review.)
ing populations of the developing

world, Kennedy says, demographers
expect average family sizes to decline
in the future, as urbanization and other factors cause a demographic transition and numbers begin to stabilize.
The demographic transition has a
longer existence in human experience
than this passing mention suggests.
Demographers see global population
moving from the near-zero growth that
persisted up to 1600 at high death and
birth rates, with life expectancy as short
as 25 years, back again to near-zero
growth at low death and birth rates,
with life expectancy exceeding 75 years.
The population explosion began in the
17th century in the now industrial countries. Representing one quarter of the
worlds inhabitants today, these nations
have nearly completed the demographic transition. The leading edge of industrial revolutionsanitation and vaccination, mass literacy, the green revolution in agriculturehas crossed into
the developing countries. In this trend,
demographers nd grounds to forecast
stabilization of the population at a number technology and the earths nite resources can sustain in civilized comfort.
But that is decades awayeven if
those forecasts are correct, Kennedy
concludes. Before we reach what is
termed global replacement fertility levels, which United Nations authorities
believe may occur around 2045, the population supertanker will have moved a
long way.
It must indeed make a dierence, as
Kennedy argues, whether human beings
stabilize their numbers at 10 billion or
20 billion. This, not the nightmare increase to terminal misery, is the population crisis.
The human presence already perturbs
the life cycles of the planetary ecosystem. The industrial revolution, by making all the citizens in some nations rich
and by lengthening the life expectancy
of people even in the poorest nations,
has amplied that inuence. In their different ways, the rich, by their appetites,
and the poor, in their increasing number, devastate the local, regional and global environment.
Kennedy gives rst priority to the
threat of global warming. (The evident
eects of the thinning of the ozone layer on photosynthesis may soon give that
threat higher priority.) Observation has
established a measurable increase in the
current concentration of carbon dioxide in the atmosphere; that is the result
principally of the fourfold increase since
1950 in the combustion of fossil fuels.
The geologic record shows the average
ambient global temperature rising and
falling in correlation with change in that
concentration. Whatever scientic uncertainty qualies the present threat,
the rise in sea level and rearrangement

of the climate map consequent to even
a small increase in global temperature
in the past should be compelling action now.
What is more, the economic development of the poor countries, crucial to
the timing of their demographic transition and so to the size of the ultimate
population, portends another fourfold
multiplication of the output of energy.
Without alternatives to fossil fuels, the
environment appears in Kennedys prognosis to be damned if they do and
damned if they do not make the demographic transition.
Kennedy nds technology posing
equal dilemma to the social arrangements of humankind. Dematerialized
manufacturing technologies, no longer
bound to the source of their raw materials, allow transnational corporations
to seize the comparative advantage of
poverty and build their factories in lowwage preindustrial countriesor sta
them with robots at home. The extended, collective human nervous system arrayed in the high-velocity, high-capacity
global communications network conducts 24-hour-a-day trading and speculation in national currencies, 90 percent of [it] . . . unrelated to trade or capital investment. In the borderless world
economy the waning of the national
state erodes a peoples capacity to
control its own aairs.
Along with factory ghost towns, bankrupted by the transnationalization of
manufacturing, the U.S. has prairie
ghost towns depopulated in the trivialization of agriculture by new technology from the life sciences. The same
biotechnology that ensures increase of
food supply in pace with the growing
numbers of people may deprive preindustrial villagers of their livelihoods by
signicant relocation of agricultural
production (or its substitutes) out of the
developing world or by in vitro production of food in situ.
On territory more familiar to historians in part two, Kennedy conducts a
comprehensive survey of nations and
regions, extrapolating from their present situations and trends possible courses for each into the 21st century. He assesses their prospective ups and downs
and their chances to arrive at mid-century as winners or losers. His assessments hold no surprises.
Japan is thus a likely winner. That is
qualied in large part by the degree to
which dematerialized technology may
incorporate the Japanese world-bestriding industrial enterprises into the planetary family of transnationals. At present, Japanese corporations remain devoutly Japanese. For the celebrated tigers of East Asia, Kennedy makes an
116
almost equally hopeful but separate appraisal. Here he fails to reckon with the
reality of the Greater East Asia Co-Prosperity Sphere. The tigers and still bigger new partners, Indonesia, Malaysia
and Thailand, run huge decits in their
trade with Japan; they oset them by a
combined surplus in their trade with
the U.S. that is equal to Japans. Surely,
the future of this empire, now in such
good working order, strongly conditions
the future of each of its members, starting with Japan.
India and China, with one third of the
worlds people, command consideration
together. Population growth reduces
their impressive achievements in economic development to small per capita
gains, Indias smaller than Chinas. Equity in distribution of gains, more so in
China than in India, has advanced both
nations into the demographic transition,
again China ahead of India. Together
they are bound to make huge additions
to the ultimate world population and
still huger contributions to the concentration of carbon dioxide in the atmosphere. Economic change will put their
unity as nations under stress. For individual Indians and Chinese, life will be
hard but less brutish and short.
In assessment close to prophesy, Kennedy sees the calamity of Africa below
the Sahara extending without remission
into the 21st century. Second only to
that of Africa, the poverty of Latin America promises to resist whatever development its countries achieve. In this connection, Kennedy does not give enough
weight to the inequality, amounting to
inequity, in income distribution prevailing in the region: the top 20 percent income group in Brazil keeps 62.6 percent of the gross national product and
allots 2.4 percent of it to the bottom
20 percent.
Hardest to decipher, Kennedy decides,
are the futures of the erstwhile U.S.S.R.
and its former buer states in Eastern
Europe. Russia is bearing the brunt of
the Soviet triple crisisthe economic
incompetence of a lopsided industrial
system, the uncertain legitimacy of the
successor political powers and the rush
of ethnic minorities to self-determination. However it weathers the near term,
Russia faces the long-term ethnic revenge of population explosion in the
newly sovereign Siberian republics.
( They, as well as Kennedy, may call it
revenge, but they owe their population explosions to Russian promotion
of economic development in the Soviet
internal Third World.) Poland, the former Czechoslovakia and Hungary are
prospective fast adjusters, with economies recovering to 1989 levels as early
as next year. Even the slow adjusters
Romania and the Balkansget there
ahead of Russia and the other fragments

of the U.S.S.R.
The alternative futures of the Western European countries hang on their
progress toward economic and political
community. Concentration of economic
power in their transnationals clouds the
political scene: it hardens national borders even as it transcends them. Europe
must continue under pressure of migration from poverty in West Asia and
North Africa, unrelieved by fundamentalist Muslim rebellion. Despite all uncertainties, science, technology and civilization seem to hold out the good life
to the next generations of Europeans.
Five years have given Kennedy no reason to adjust the trajectory on which
he pointed the U.S. in his earlier book.
He stoutly maintains his declinism
against the protests that have come
meanwhile from die-hard supply-side
publicists. The U.S. has a combined public, business and household debt approaching 200 percent of the worlds
biggest GNP. It borrows $100 billion a
year from the world economy. It loses
one fth of its children from school in
their high school years. The number of
candidates for advanced degrees in its
universities is declining. Homicide rates
per capita are ve times the worst found
in other industrial countries. The U.S.
crowds more than a million convicts into
its prisons, among them 40 times as
many black prisoners, proportionately,
as South Africa at the peak of apartheid. It holds family incomes at mid1970s levels, even as the average compensation of the managers of failing industrial enterprises ascends from 40 to
90 times their payroll averages. It celebrates gains in productivity calculated
from increases in applications for unemployment insurance. The U.S. can
have no plan for the 21st century without becoming a dierent kind of country. For that, short of a sucient shock
to complacency, the prospect is small.
That, in turn, would require leadership
very different from the sort demonstrated by recent incumbents of the White
House, whether it concerned domestic
decits or global population and environmental issues.
On those issues that intrude from the
world outside, Kennedy has little help
to oer the leaders of the U.S. and the
other industrial nations. He is misled
on his venture into unfamiliar territory. The 21st century is not irrevocably
committed to fossil fuels and climatic
change. Investment now in promising
alternative energy conversion cycles can
meet the fourfold increase in energy demand without further opaquing of the
carbon dioxide window. The capture of
sunlight by photovoltaic cells and storage of the energy in hydrogen are al-
117
CORRESPONDENCE
Reprints are available; to order, write
Reprint Department, SCIENTIFIC AMERICAN, 415 Madison Avenue, New York,
N.Y. 10017-1111 or fax inquiries to
(212) 355-0408.
Back issues: $6 each ($7 outside U.S.)
prepaid. Most numbers available. Credit card (Mastercard/Visa) orders for
two or more issues accepted. To order,
fax (212) 355-0408.
Index of articles since 1948 available
in electronic format. Write SciDex, SCIENTIFIC AMERICAN, 415 Madison Avenue, New York, N.Y. 10017-1111, or fax
(212) 355-0408.
Photocopying rights are hereby granted by Scientific American, Inc., to libraries and others registered with the
Copyright Clearance Center (CCC) to
photocopy articles in this issue of SCIENTIFIC AMERICAN for the fee of $3.00
per copy of each article plus $0.50 per
page. Such clearance does not extend
to the photocopying of articles for promotion or other commercial purposes.
Correspondence and payment should
be addressed to Copyright Clearance
Center, Inc., 27 Congress Street, Salem, Mass. 01970. Specify CCC Reference Number ISSN 0036-8733/93.
$3.00 + 0.50.
Editorial correspondence should be
addressed to The Editors, SCIENTIFIC
AMERICAN, 415 Madison Avenue, New
York, N.Y. 10017-1111. Manuscripts
are submitted at the authors risk and
will not be returned unless accompanied by postage.
Advertising correspondence should
be addressed to Advertising Manager,
SCIENTIFIC AMERICAN, 415 Madison Avenue, New York, N.Y. 10017-1111, or
fax (212) 754-1138.
Subscription correspondence should
be addressed to Subscription Manager,
SCIENTIFIC AMERICAN, P.O. Box 3187,
Harlan, IA. 51537. The date of the last
issue of your subscription appears on
each months mailing label. For change
of address notify us at least four weeks
in advance. Please send your old address (mailing label, if possible) and
your new address.
118
ready in pilot-plant development with

government funding in Japan and Germany. Biotech reneries of synthetic nourishmentfeedstocks not speciedare unlikely to obviate farmlands
and villagers to cultivate them.
With respect to the demographic transition in the preindustrial countries and
the economic assistance needed to encourage it, Kennedy has even less help
to oer. He settles for distribution of
contraceptives to the natives and, in an
improvement on that condescension,
the encouragement of female literacy.
It is true that people in the developing
countries are motivated to control fertility at much lower incomes than those
at which contraception became widespread in the industrial countries. Thus,
they are making the demographic transition in a much shorter time. An extensive literature on economic development
supports the necessity, however, of appropriate physical and cultural setting
for promotion of female literacy and
motivation of fertility control: potable
if not running water, a modicum of sanitation and schoolhouses, to start with.
Kennedy refers to a UN Conference
on Trade and Development and $125
billion a year to pay for new environmental programs . . . swiftly trimmed to
a request for an additional $5 to $10
billion a year, in recognition of todays
political and economic realities. Whatever happened at that UNCTAD meeting in Paris in 1990, that $125 billion
stays there on the green-baize table at
which the rich and poor countries are
negotiating the future of the planet and
its inhabitants. At the U.N. Conference
on Environment and Development in Rio
de Janeiro in June 1992, the then 172
rich and poor member states adopted
Agenda 21. In preparation for the 21st
century, this calls for annual investment
in the economic development of the
poor countries at the rate of $600 billion. The poor nations are to supply
nearly $500 billion in the form of labor
and natural resources. The $125 billion
is the share of the rich countries. It is
specied in the form of the catalytic
technology transfer necessary to put
those people and resources to work.
Agenda 21 stands as the rst international agreement premised on the demographic transition. Its great objective
is to hasten the arrival of the human
population at a smaller ultimate size.
While it is a nonbinding agreement, it
does set out a global consensus and
political commitment at the highest levels on what needs to be done about
population, environment and development, comprehended in a program of
sustainable development. By action of
the U.N. General Assembly in December,
Agenda 21 is now the agenda of the
U.N., at the very least an agenda for ongoing debate in the General Assembly.
Implementation of Agenda 21 relies,
of course, on the forthcoming of that
$125 billion a year. As Kennedy notes,
the Nordic countries (and Canada) are
laying out 0.7 percent of their GNP in
economic assistance to the developing
countries. They are only making good
on a promise in which all the industrial
countries joined 20 years ago. In 1961,
on a motion made by John F. Kennedy,
they had promised 1.0 percent of their
GNP to make the 1960s the Decade of
Development. When the 1960s went
into history as the decade of disappointment for the preindustrial countries, the
industrial countries committed to the
more realistic figure of 0.7 percent.
By no coincidence, that is precisely the
percentage of the GNP of the industrial
world that yields $125 billion. At Rio,
the rich countries, with the exception
of the U.S., promised to bring their outlays for economic assistance up to the
0.7 percent as soon as possible. The
European Community and Japan were
ready to make a down payment on their
promise. They refrained out of diplomatic courtesy to the Bush reelection
campaign and are waiting for the Clinton administration to settle into oce.
If Agenda 21 has been so little celebrated in the U.S. mass media as to
make this mention of it the rst that
has called it to the readers attention,
more will be heard of it during this last
decade of the 20th century. Agenda 21
now has its own undersecretary general for sustainable development installed
at U.N. headquarters in New York City
and a determined caucus of developing
countries in the General Assembly. It is
gathering the support, moreover, of the
citizen movements around the world
that put environmental protection agencies into nearly every national government during the past two decades. At
Rio, in their own Global Forum, 20,000
environmental zealots learned that poverty connects the integrity of the environment to development.
Anxious to create jobs, especially in
the smokestack industries, President Bill
Clinton may soon nd it desirable to
seek to increase the U.S. annual outlay to
foreign aid above the present paltry
and misdirected 0.2 percent of the countrys GNP. His vice president is on record advocating a global Marshall plan.
More sympathetic acquaintance with
the relevant political and economic realities, as well as with continuities of
human experience less familiar to him,
might have encouraged Paul Kennedy to
urge bolder action.
GERARD PIEL is chairman emeritus of
Scientic American.
ESSAY
by Ralph Gomory and Hirsh Cohen
Science: How Much Is Enough?
he military defense of the U.S. has

been the best articulated motivation for the support of basic science. Also important has been a widespread belief that scientic leadership
would translate into both economic power and a higher standard of living. The
disappearance of the threat from the Soviet Union has undermined one of these
rationales. The slow growth of the American economy, contrasted with the rapid
progress of Japan, which is not a world
leader in academic science, has eroded
the other.
Does basic research have something
to oer when economic progress and
improving the environment have displaced the military threat on the nations
agenda? If so, how much basic research
is needed?
The history of the computer answers
the rst question. The transistor could
not have been invented without years
of basic research that built a quantum
mechanical understanding of crystalline
solids. Without that work neither the device itself nor the subsequent gains in
computing power accompanied by declining cost per circuit element would
have been possible. We would not have
the computer industry as we know it.
The research in solid-state physics
that preceded the invention of the transistor exemplies one kind of relation
between science and economic progress.
In this case, it is the emergence of useful technology from research pursued
without a visible industrial goal. At the
next stage are areas of science such as
high-temperature superconductivity that
have not generated new industries but
do have discernible economic potential.
Then there are new science-based industries closely dependent on fundamental research that already yield products. Biotechnology is one. It is not a
large industry today, but it has enormous potential. This kind of infant industry, moving at a fast technical and
competitive pace, requires continuous
input of new science to maintain its
expansion.
At the more developed stage reached
by the semiconductor and pharmaceutical industries, technical directions have
stabilized. In such industries, corporate
R&D laboratories meet much of the scientic needs. Then there are industries
120
in which technological change itself is

slow. For these the tie to current science
is correspondingly weak. Automobile
manufacturing is an example of an industry that requires only the science that
is available to everyone.
For industries of either of these last
two types, ones that are like semiconductors or like automobiles, productivity gains and market dominance come,
for the most part, from improving design and production processes, perceptive marketing strategies and managing
people to benet from their most creative capabilities. In some of these industries, we observe that Japan, in spite
of its lack of prominence in academic
science, has become a global power.
For edgling science-based industries,
then, the linkage to research must be
close and continuous. Industries that are
already well on their way can be sustained by the work of their own laboratories. For mature industries, the main
contribution of science may be the
graduates who can use the knowledge
available to all.
learly, science does have an economic role. How do we decide

what science to support and at
what level to support it?
Our proposition is a simple one: all
major elds of U.S science should be
funded at a level that will keep them
among the worlds leaders. Being one of
the leaders will enable us to be among
the rst to participate in the birth of
new industries. This basic economic rationale must apply to all disciplines, because we do not know where the next
signicant event will take place. Moving
quickly when something exciting does
happen in a eld may even call for increasing the support of that scientic
area. Action may be necessary downstream as well.
We are suggesting that the economic
criterion for deciding how much to support American science is a comparative
one. Can comparisons by disciplines
actually be carried out? The many studies conducted by the National Research
Council, the opinions of researchers in
the eld and the number of rst-rate
scientists in a particular area all provide
useful indicators.
We believe the goal of having a world-
class scientic enterprise can be obtained without increasing the federal

budget. The reason is that in the U.S.
most elds of science probably already
satisfy our test.
For those sciences that directly support the infant industries we have described, we propose that the U.S. should
maintain a denite advantage. We might
expect, for example, to see strong support for those parts of molecular and
cellular biology and immunology that
assure a primary position in therapeutic and diagnostic technology and in agricultural biotechnology. We would expect to see such support for condensedmatter physics, for chemistry and for
the material sciences that contribute to
the development of products based on
high-temperature superconductivity.
None of these considerations should
exclude funding based on other objectives and values. They might include
contributions to health, to the environment or to other societal goals or a decision to emphasize a eld because of
its intellectual excitement. The economic rationale based on the comparative
world-class criterion, however, should
put a rm oor under the support of
basic science.
In some cases, for example, space science, we may be spending billions more
than anyone else. Funds from such elds
should be redirected to areas in which
our advantage is less overwhelming. In
areas of inquiryparticle physics, for
onethat depend on a small number
of expensive, experimental installations,
our standing could probably be maintained more easily by dealing with the
complexities of international cooperation than by absorbing the enormous expense of an unneeded rivalry for world
leadership.
We believe this approach would lead
to a more predictable, stable scientic
environment in which the funding could
be determined on a long-term basis. It
would help science to be a more attractive career for both the established and
the young scientist. It would assure that
our leading scientists would be well
funded and that our country would be
in a position to benet economically
from their work.
RALPH GOMORY is president and
HIRSH COHEN is program ocer for
industrial studies at the Alfred P. Sloan
Foundation in New York City.

199307

Uploaded by

Document Informationclick to expand document informationi

Document Informationclick to expand document information

Copyright:

Available Formats

199307

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

199307

Uploaded by

Copyright:

Available Formats

JULY 1993

Telling reasonable risks from fools chances.

Polar dinosaurs, which were adapted to the cold and dark ,

Risk Analysis and Management

Australias Polar Dinosaurs

Accurate Measurement of Time

Surgical Treatment of Cardiac Arrhythmias

Copyright 1993 Scientific American, Inc.

Bart Kosko and Satoru Isaka

Edwin Hubble and the Expanding Universe

TRENDS IN ENVIRONMENTAL SCIENCE

Sustaining the Amazon

Science and the Citizen

Letters to the Editors

50 and 100 Years Ago

Science and Business

Copyright 1993 Scientific American, Inc.

Essay : Ralph Gomory

THE COVER painting shows Allosaurus

Boris Starosta (left),

Jana Brenning (top),

Laurie Grace (top),

Jared Schneidman /JSD

Peter Trusler ; for a stamp

Johnny Johnson ( left),

Patricia J. Wynne (top),

BOARD OF EDITORS: Alan Hall , Executive Editor ;

Peter Menzel; courtesy

Mount Wilson and

Ricardo Azoury/ Black Star

Boris Starosta (top),

PRESIDENT AND CHIEF EXECUTIVE OFFICER:

CORPORATE OFFICERS: Executive Vice President

Ricardo Azoury/Black Star

CHAIRMEN OF THE BOARD:

CHAIRMAN EMERITUS: Gerard Piel

Crown copyright, Science

EDITOR: Jonathan Piel

COPY: Maria- Christina Keller, Copy Chief ; Nancy

Cover painting by Dimitry Schidlovsky

ART : Joan Starwood, Art Director ; Edward Bell ,

ADMINISTRATION: John J. Moeling, Jr., Publisher ;

SCIENTIFIC AMERICAN, INC.

SCIENTIFIC AMERICAN July 1993

Copyright 1993 Scientific American, Inc.

LETTERS TO THE EDITORS

SCIENTIFIC AMERICAN July 1993

X-cluded from Credit

The photograph on page 101 of

Because of the volume of mail, letters

Copyright 1993 Scientific American, Inc.

50 AND 100 YEARS AGO

SCIENTIFIC AMERICAN July 1993

hope, in harmony among themselves!

for catching a slippery, active prey by

The great Ferris wheel

The wonderful merry-go-round designed by Engineer George W. G. Ferris,

Copyright 1993 Scientific American, Inc.