On Being a Scientist PNAS-1989-Research Article-9053-74

(1)

Proc. NatL Acad. Sci USA 86 (1989) 9053

On

Being a Scientist

CommitteeontheConduct of Science NationalAcademy ofSciences

Reprintedwithpermission.

(2)

Proc. NatL Acad ScL USA86

(1989)

NATIONAL ACADEMY PRESS 2101Constitution Avenue, NW

Washington, DC 20418

NOTICE: The Council of the NationalAcademyofSciences authorizedthe

formationofthe Committeeonthe Conduct of Science andsubsequently reviewedthecommittee'sreport. The members ofthe committeewerechosen

fortheirspecial competenciesand withregard forappropriatebalance.

The National AcademyofSciences isaprivate, nonprofit, self-perpetuating society ofdistinguished scholars engaged in scientificandengineering re-search,dedicatedtothefurtherance of scienceandtechnologyand totheir use for thegeneral welfare. Upon theauthorityof the chartergrantedtoitbythe

Congress in 1863, the Academy hasamandate thatrequires ittoadvisethe

federalgovernment onscientificandtechnicalmatters. Dr.Frank Press is

presidentof theNationalAcademy of Sciences.

Library of Congress Catalog CardNumber89-62915

Designer FrankPapandrea

Printedinthe United States of America 9054

Report

(3)

Proc. NatL Acad. Sci. USA 86(1989) 9055

Committee

on

the

Conduct of Science

Francisco Ayala Chairman

Department of Ecology andEvolutionaryBiology University ofCalifornia-Irvine

RobertMcCormick Adams

Secretary

SmithsonianInstitution

Mary-Dell Chilton

CIBA-Geigy Biotechnology Gerald Holton

Professor ofPhysics and Professor of the History of Science HarvardUniversity David Hull PhilosophyDepartment NorthwesternUniversity KumarPatel ExecutiveDirector,

Research, MaterialsScience,Engineeringand Academic Affairs Division

AT&TBell Labs

Frank Press President

NationalAcademy of Sciences Michael Ruse

Philosophy Department

University of Guelph,Canada

PhillipSharp

Center for CancerResearchand Department ofBiology

MassachusettsInstitute ofTechnology

Consultant Writer SteveOlson Staff BarbaraCandland LawrenceMcCray ..l

Report

(4)

Proc. Natl. Acad. Sci USA 86 (1989)

Preface

T hisbooklet is written primarily for students who are beginningto doscientificresearch. Itseeks todescribesomeof the basic features of alife in contemporary research and some of the

personalandprofessionalissues that researcherswill encounter in their work.

Traditionally,youngscientistshavelearned aboutthe methods and values of scientific research frompersonal contact with moreexperienced scientists,and suchinteractions remainthebestwayforresearcherstoabsorb what is still a largelytacit codeof professional conduct. Anybeginningresearcher whohas notworked closelywith anexperienced scientist is missingoneofthe most

importantaspectsofascientific education. Similarly, anyexperienced re-searcher who does not pass on to youngerscientistsa senseof the methods and normsofscience is significantly diminishinghis orher contributiontothe field's progress. However, theinformaltransmission of values isnotalways

enough. Changes inscience inrecentyears,includingthegrowingsize of research teams andthequickening paceofresearch, sometimes have had the

effect of reducingcontactbetweensenior andjuniorresearchers. The

increas-ingsocialimportanceandpublicvisibilityofscienceandtechnology also make

itessential that beginning researchersknow how importantthey are to

safe-guardingtheintegrity ofthescientific enterprise.

Some ofthetopics discussed inthisdocument,such as sources of error in

science, scientific fraud, and misappropriationofcredit,have receivedagreat deal of attentionoverthe pastdecade,both withinthe scientificcommunityand

outside it. Inpreparing thisbooklet, the governing council of theNational

AcademyofSciences hopestocontributetothediscussionand tostimulate re-searcherstoidentifyandupholdtheproceduresthatkeep sciencestrong and

healthy.

Oneof the mostappealing features of researchisthe greatdegree ofpersonal freedom accorded scientists-freedomtopursueexcitingopportunities,to

exchange ideas freelywithotherscientists,tochallenge conventional knowl-edge. Excellence insciencerequiressuchfreedoms,and theinstitutionsthat supportscienceintheUnited States have foundwaystosafeguardthem. However,modernscience, whilestrongin many ways, is alsofragilein

importantrespects. Forexample,effortstorestrictthereportingof research resultscanbedevastating.

MostAmericans see astrongscienceasessentialto asuccessful future. Yet

thatgeneroussocial supportis basedonthepremisethatsciencewill be done

honestlyand thatmistakeswill beroutinelyidentified and corrected. The

mechanisms thatoperatewithin sciencetomaintain

honesty

and self-correction mustthereforebe honoredandprotected.Research institutionscansupport thesemechanisms, but it istheindividual researcher whohasboththe capabil-ityand theresponsibilitytomaintainstandards of scientific conduct.

Frank Press President

National

Academy

of Sciences

(5)

Proc. NatLAcad. Sci. USA 86(1989) 9057

Acknowledgments

Theprojectwassupportedby a grantfromthe Richard LounsberyFoundation.

Dissemination costs weresupported bytheBasicScience FundoftheNational

Academyof Sciences, whose contributors include the AT&TFoundation, ARCOFoundation,BPAmerica,DowChemicalCo., E. I. du Pont de Nemours andCo., IBM, MerckSharp& Dohme ResearchLaboratories, Monsanto,and ShellCompaniesFoundation; and theconsortiumfunds ofthe National ResearchCouncil, consistingofcontributions from the following private

foundations:theCarnegie Corporation ofNewYork, theCharlesE.Culpeper Foundation,theWilliam and Flora Hewlett Foundation, the John D. and

CatherineT.MacArthurFoundation,the Andrew W. MellonFoundation,and the RockefellerFoundation.

(6)

Proc. NatL Acad. Sci. USA 86(1989)

On Being

a

Scientist

Paramecium. For years biologists had been trying to induce conjugation between paramecia, a process in which two paramecia exchange genetic material across a cy-toplasmic bridge. Now Sonneborn had isolated two strains of paramecia that he believed wouldconjugate when combined. If successful, his experiment would finally overcome a

major

obstacletostudiesof protozoan

genetics.

Sonneborn mixed the strains together on a slide and put the slide under his microscope. Looking through the eyepiece, he witnessed for the first time what he would later call a "spectacular" reaction: The paramecia had clustered into large clumps and were conjugat-ing. In a state ofdelirious excitement,Sonneborn raced through the halls of the deserted building looking for someone with whom he could share his joy. Finally he dragged a puzzled custodian back to the laboratory to peer through the microscope and witness this marvelous phenomenon.

Moments of scientific discovery can be among the most exhilarating of a scientist's life. Thedesireto observe or understand what no one has ever observed or understood before isoneof the forces that keep researchers rooted to their laboratory benches, climbing through the dense undergrowth of a sweltering jungle, or pursuing the threads ofa difficulttheoreticalproblem. Few discoveries seem to come in a

flash;

most materialize moreslowly over weeks or years. Nevertheless,theprocess can bring great satisfaction. Thepieces fitinto place. The whole makes sense.

Alife in sciencecan entail great frustrations and disappointments as well as satisfactions. Anexperiment can fail because of a technical complication or the sheer intractability of nature. A favorite hypothesis that has consumed months of effort can turn out to be incorrect. Disputes can break out with colleagues over the validity of experimental data, theinterpretationof data, or credit for work done. Setbacks such as these are virtually

im-possibletoavoidin science, and they can strain the composure of both the novice and the mostself-assured senior scientist.

To anobserver of science, the presence of these human elements in research raises an obviousquestion. Science results in knowledge that is as solid and reliable as anything weknow. Science and technology are among humanity's greatest achievements, having

transformednotonly thematerial conditions of our lives but the very way in which we viewthe world. Yetscientific knowledge emerges from a process that is intensely human, aprocess marked by its full share of human virtues and limitations. How is the limited,

falliblework ofindividual scientists converted into the enduring edifice of scientific knowledge?

Manypeople think ofscientific research as a routine,cut-and-driedprocess. They associate the nature of scientific knowledge with the process of deriving it and conclude thatresearch is as objective and unambiguous as scientific results. The reality is much different.Researchers continually have to make difficult decisions about how to do their workand how topresent that work to others. Scientists have a large body of knowledge that they can use inmaking these decisions. Yet much of this knowledge is not the product ofscientific investigation, but instead involves value-laden judgments, personal desires, and even aresearcher's personality and style.

This bookletdivides the decisions that scientists make into two overlapping categories. Much of the first half of the booklet looks at several examples of the choices that scien-tists make in their work asindividuals: thetreatment of data, techniques used to minimize bias, the application of values in judging hypotheses. The second half deals largely with questions that arise during the interactions among scientists: the need to report research resultshonestly andaccurately, the proper distribution of credit for scientific work, the difficult problem of reportingmisconduct. A final section touches upon the social context inwhichpersonal andprofessional decisions are made and details a few of the special

ob-ligationsthatscientists have as members of society at large.

(8)

Proc. NatL Acad. Sci. USA86

(1989)

The

nature

of

scientific research

Is

There

a

Scientific

Method?

"Scientists

are

people of

very

dissimilar

temperaments

doing

different

things

in

very

different

ways.

Among scientists

are

collectors,

classifiers and compulsive

tidi-ers-up; many are

detectives

by

tempera-ment

and

many are

explorers;

some are

artists

and

others

artisans.

There

are

poet-scientists and

philosopher-scientists

and

even a

few mystics.

"

Peter B.Medawar,The Artof the Soluble, London:Methuen,

1967,p. 132

T hroughout thehistory of science, some

philosophersandscientists havesoughtto describeasinglesystematic method thatcan be usedtogenerate scientificknowledge. For

instance,oneschool ofthought, datingbackatleastto FrancisBacon in the seventeenth century, points to

obser-vationsasthefundamental source of scientificknowledge.

According to this view, scientists must cleanse their minds ofpreconceptions,sitting down before nature "as a little child," as thenineteenth-century biologist ThomasH. Huxley described it. By gathering facts withoutprejudice,

ascientist willeventuallyarrive atthe correct theory. Somescientists may believe in such apicture of them-selves andtheirwork,butcarryingthisapproachinto practice is impossible. Nature is tooamorphous and diverse forhumanbeingstoobservewithouthavingsome ideas about what they are observing. Scientific under-standing is made possible through the interplay of mental constructsand sensoryimpressions. Scientistsmaybe able tosuspendsomeprior theoreticalorthematic precon-ceptions to view nature from a newperspective,butthey cannotview thephysicalworld without any perspective. Otherformulations of the"scientific method"have been proposed overtheyears, but manyscientistsregard such blanketdescriptions of whatthey do withsuspicion. Perhaps from a distancesciencecanbeorganized into a coherentframework,but inpracticeresearchisasvaried astheapproaches of individual researchers. Some

scientistspostulatemanyhypothesesandsystematically setabouttryingtoweedouttheweaker ones. Others

describe their workasasking questionsof nature: "What

wouldhappenif ... ? Why is itthat ... ?" Some

re-searchersgatheragreatdeal ofdata withonlyvague ideas about theproblem they might be tryingtosolve. Others

developaspecific hypothesisorconjecturethatthey then

trytoverifyorrefute withcarefullystructured observa-tions.

Rather thanfollowingasingle scientific method,

scien-tistsuse abody of methods particulartotheirwork. Some ofthese methodsarepermanentfeaturesofthe scientific community; others evolveovertimeorvaryfrom

disci-plinetodiscipline. Inabroadsense,thesemethods include allofthetechniquesandprinciplesthatscientists applyintheirwork and intheirdealings withother

scientists. Thus, theyencompassnotonly theinformation

scientistspossessabout theempiricalworldbut the

knowledge scientistshaveabout howtoacquiresuch

information.

(9)

Proc. NatL Acad Sci. USA 86(1989) 9061

The Treatment of Data

One goal of methods istocoaxthe facts, untainted by hu-manbias, fromascientific investigation. Inretrospect,

thismay seem astraightforwardprocess, asimple

applica-tion of accepted scientific practicestoaspecific problem.

Butatthe forefronts of research, neithertheproblemnor

the methods usedtosolve itareusually well-defined.

Instead, experimental techniquesarepushedtothe limit,

thesignal is difficultto separatefrom the noise, and unknownsourcesoferrorabound. In suchanuncertain

and fluid situation, pickingoutreliable data points froma massofconfusing and sometimes contradictory

observa-tionscanbeextremely difficult.

Onewell-known example of this difficulty involvesthe

physicist Robert Millikan, whowonthe Nobel Prize in

1923 for his workonthe charge oftheelectron. Inthe

1910s, justasmostphysicistswerecomingto acceptthe

existence oftheelectron, Millikan carriedonaprotracted

and sometimesheateddisputewith theViennesephysicist Felix Ehrenhaftoverthe magnitude ofthesmallest

electri-calcharge foundinnature.Bothmenbased theirfindings on themovementsoftiny charged objects-oil drops, in Millikan's case-in electric fields. Ehrenhaft used allthe

observations he made without much discrimination and eventuallyconcluded that therewas nolower limittothe

sizeofanelectricalcharge that could exist innature.

Millikan usedonly what he regardedashis "best" datasets toestablish themagnitude of the charge andargueagainst

theexistence of Ehrenhaft's "subelectrons." Inother words, Millikan applied methods of data selectiontohis observations that enabledhimtodemonstrate theunitary

charge of the electron.

Millikanhas been criticized fornotdisclosingwhich data heomittedorwhy he omitted those data. Butan

examina-tion ofhis notebooks reveals that Millikanfelt he knew just how far he couldtrusthisrawdata. He often jotted

down in his notebooks what hethoughtweregoodreasons

forexcluding data. However, he glossedoverthese exclusionsinsomeof hispublishedpapers,and bypresent

standardsthis isnotacceptable. Scientistsmustbe willing

toacknowledge the limitationsontheir data iftheyarenot tomisleadothersabout the data'sreliability.

Generalrulesfordistinguishingapriori "good" data

from"bad"cannotbe formulated withmuch clarity. Nevertheless, good scientists have methods that theycan

applyinjudgingthereliabilityofdata, and learning these methods isoneofthegoalsofascientificapprenticeship.

These methodsmaybeuniquetoagiven situation,

dependingonhow andwhyasetof observations isbeing made. Nevertheless, they imposeconstraintsonhowthose

observationscanbeinterpreted. Aresearcher isnotfreeto

selectonlythe data that fit hisorherprior expectations.If certain dataareexcluded,aresearchermusthave justifi-ablereasonsfordoingso.

The Relation Between Hypotheses

and

Observations

Attempts toisolate the facts and nothing butthe facts in scientificresearchcanraisephilosophicalaswellas

meth-odologicalproblems. One prominent difficulty involves

the line ofdemarcation between hypothesesand

observa-tions. For yearsphilosophers have triedto construct

purely observationallanguagesfree oftheoretical con-structs,but they have never been completely successful. Even asimpledescription suchas"Thetemperaturein this roomis 25 degrees

centigrade"

containsahost of theoreti-calunderpinnings. The thermometer usedto measurethe temperatureisacomplexdevice subjecttoitsown

system-aticand random errors. And thequantity beingmeasured is not some fundamentalattribute ofnaturebutdependsin

acomplexway onthemovementsandinteractionsof gas

particles,whicharedescribed intermsof thekinetic theoryofgases, quantummechanics, andso on.

Thetermsused inscience also contributetothe inter-penetration of hypotheses and observations. Forexample,

Anton van

Leeuwenhoek,

the seventeenth-century Dutch microscopist, pridedhimself indescribingwhat hesaw through hislenses withoutanytheoreticalspeculation.

However,hisdescriptionswereanything but theory-neutral. When heexaminedthe waterstandingin the gutteroutside hiswindow,someof themicroscopic creatureshesaw wereprobably Euglena. Todayweknow

that thesesingle-celled organisms containchlorophyll and

are moreclosely relatedtoplantsthananimals. But because thecreaturesmoved,vanLeeuwenhoekcalled

them"animalcules,"not"planticules."

Termssuchas"energy,""grossnationalproduct," "pion,""blackhole,""intelligence quotient," and "gene"

areclearly derived from particulartheories and obtain

muchoftheirmeaning from their rolesinthesetheories.

Butsuch theoreticaltermscantakeonalife of theirown and begradually transformedinto moreobservational terms.Similarly,as termsbecome unmooredfrom their original theories, the potentialtomisuseormisunderstand them increases.

The Risk

of

Self-Deception

Awarenessoftheinroads thattheorycanmakeinto observationsserves asavaluablereminderof theconstant danger of self-deceptioninscience. Psychologistshave

shown thatpeople haveatendencyto seewhat theyexpect to seeandfailtonotice whatthey believe shouldnotbe

there. For instance, duringtheearlypartof thetwentieth

centuryoneof themostardent debates inastronomy con-cerned thenatureof whatwerethen knownas

spiral

nebulae-diffusepinwheels of lightthat

powerful

tele-scopesrevealedtobequitecommonin the

night sky.

Someastronomersthought that these nebulaewere

spiral

galaxies like the

Milky

Wayatsuchgreatdistances that individualstarscouldnotbe

distinguished.

Others

(10)

Proc. NatL Acad ScL USA 86

(1989)

lieved that theywere clouds of gas within our own galaxy. One astronomer inthelattergroup,Adriaanvan Maanen ofthe MountWilsonObservatory, sought to resolve the issue bycomparingphotographs of the nebulae taken several years apart. Aftermaking a series of painstaking measurements, vanMaanenannouncedthat he had found

roughlyconsistent unwinding motions in the nebulae. The detection of suchmotions indicatedthatthespiralshad to be withintheMilkyWay,since motionswould be impos-sibletodetectin distant objects.

VanMaanen'sreputation caused many astronomers to accept agalacticlocation forthe nebulae. Afewyears

later, however,van Maanen'scolleague Edwin Hubble,

usingthe new 100-inchtelescopeatMount Wilson,

con-clusively demonstratedthatthenebulae were in fact

distantgalaxies;vanMaanen'sobservations hadtobe wrong.Studies ofhis procedures have not revealed any in-tentionalmisrepresentationor sourcesofsystematic error.

Rather, hewasworkingatthe limits ofobservational

accuracy,andhesawwhatheexpectedto see.

Self-deceptioncantakemoresubtleforms. Forexample,

aresearcher may stop adatarun tooearly because the ob-servations conformtoexpectations, whereasalongerrun

mightturnup

unexpected

discrepancies. Insufficient

repetitions of anexperimentare a common causeof

invalidconclusions,as arepoorlycontrolled experiments.

Methods and

Their Limitations

Overtheyears, scientistshavedevelopeda vast arrayof methods thataredesignedtominimize

the

kinds of

prob-lemsdiscussedabove. Atthe mostfamiliarlevel, these methodsincludetechniquessuch asdouble-blind trials,

randomization ofexperimental subjects, and the proper use

ofcontrols,whichareallaimedatreducing individual sub-jectivity. Methodsalsoinclude theuseof tools in

scien-tificwork,both themechanical tools usedtomake

obser-vations and theintellectualtools used to manipulate

abstractconcepts.

The term"methods"canbeinterpretedmorebroadly. Methodsinclude thejudgments scientistsmakeabout the interpretationorreliability ofdata. They also includethe

decisions scientists make about which problemsto pursue orwhen toconcludeaninvestigation. Methods involve the

waysscientistsworkwitheachotherandexchange information. Takentogether, these methods constitute the craft ofscience, andaperson's individual application of

thesemethodshelps determinethatperson's scientific style.

Somemethods, suchasthose governing the design of experimentsorthe statisticaltreatmentof data,canbe

writtendownandstudied. (The bibliography includes

several booksonexperimental design.) Butmanymethods arelearnedonlythrough personal experience and

interac-tions withother scientists. Some are even harder to

describeorteach. Many of theintangible influenceson

scientific

discovery-curiosity,

intuition,

creativity-largely defy rational analysis, yet they are among the tools that scientists bring to their work.

Althoughmethods are an integral part of science, most of

themare nottheproductofscientificinvestigation. They have been developed and their use is required in science becausethey have beenshown to advance scientific knowledge. However, even if perfectly applied, methods cannot guaranteethe accuracy of scientific results. Experi-mentaldesignis often as much an artasascience;tools canintroduceerrors;andjudgments aboutdatainevitably

rest onincompleteinformation.

Thefallibility of methodsmeansthatthere isno

cook-book approachtodoing science,noformula that can be

appliedormachine thatcanbe builttogeneratescientific

knowledge. Butscience wouldnotbesomuch fun if there were. The skillfulapplicationof methodsto achallenging problem isoneof thegreatpleasuresof science. The laws of nature are not apparent in oureverydaysurroundings, waitingtobe pluckedlike fruit froma tree. Theyare

hidden andunyielding, and the difficulties ofgrasping

them addgreatlytothesatisfaction ofsuccess.

Values in

Science

When methods aredefinedasall ofthetechniques and principles that scientists applyintheirwork, it is easierto seehowtheycanbeinfluencedby humanvalues. Aswith

hypotheses, human valuescannotbeeliminated from science,andtheycansubtly influencescientific investiga-tions.

Theinfluence of values isespeciallyapparent

during

the formulationorjudgment of hypotheses. At anygiven time, several competing hypothesesmayexplain the

avail-able factsequally well, and eachmay suggestanalternate routefor further research. Howshouldoneselectamong

them?

Scientists andphilosophers haveproposed severalcriteria

by whichpromising scientific hypothesescanbe distin-guished from less fruitfulones. Hypothesesshould be

internally consistent,sothat

they

donotgenerate contra-dictory conclusions. Their abilityto

provide

accurate predictions, sometimesinareasfar removed fromthe

original domainofthe

hypothesis,

is viewedwith great

favor. Withdisciplines in which

prediction

is less

straight-forward,suchasgeologyorastronomy,goodhypotheses should be abletounify disparate observations. Also highly prizedaresimplicity anditsmorerefinedcousin, elegance.

The above values relatetothe

epistemological,

or knowl-edge-based, criteria

applied

to

hypotheses.

But valuesofa

different kindcanalsocomeintoplayinscience. Histori-ans,sociologists, and other students of sciencehave shown thatsocial andpersonal values unrelatedto

epistemologi-cal

criteria-including

philosophical, religious, cultural,

political,and economic values-canshape scientific judgmentinfundamentalways. For

instance,

inthe

nine-teenthcenturythegeologist Charles

Lyell championed

the 90762

Report

(11)

Proc. NatL Acad Sci. USA 86

(1989)

9063 conceptofuniformitarianismin geology,arguingthat

incremental changes operating over longperiodsof time haveproduced the Earth's geological features,not

large-scalecatastrophes.However,Lyell's

preference

for this

still importantideamayhave dependedasmuchonhis

religiousconvictionsas onhis geological observations.He favored thenotion ofaGod whoisanunmovedmoverand

does notintervene in His creation. SuchaGod, thought

Lyell, would produce a world where the same causes and

effects keepcycling eternally, producingauniform geological history.

Theobvious questioniswhetherholdingsuchvaluescan harm a person'sscience. In many cases the answer has to be yes. The historyofscience offersmanyepisodesin whichsocialorpersonal values ledtothepromulgationof

wrong-headed ideas. Forinstance,pastinvestigators produced "scientific" evidence forovertly racistviews, evidencethat wenowknowtobewhollyerroneous. Yet atthe time the evidencewaswidely accepted and contrib-utedto

repressive

socialpolicies.

Attitudesregarding thesexesalsocanleadtoflawsin

scientificjudgments. Forinstance,someinvestigators who

havesoughttodocument theexistenceorabsenceofa

relationship betweengender andscientificabilities have allowedpersonalbiases todistortthe design of their studiesor theinterpretation of their findings.Such biases cancontributetoinstitutionalpoliciesthathave caused females andminoritiestobeunderrepresentedinscience,

with a consequentlossofscientific talent anddiversity. Conflicts of interestcausedby financial considerations areyetanothersourceof values thatcanharm science.

Withtherapiddecrease in time betweenfundamental discoveryand commercialapplication, privateindustry is

subsidizing a considerable amount of cutting-edge

research.This commercial involvementmaybring researchers intoconflict with industrial managers-for instance,overthe publication of discoveries-oritmay

biasinvestigations in the direction of personal gain.

The above examples are valuablereminders of the danger ofletting values intrude intoresearch. Butit does not

follow that socialandpersonalvaluesnecessarilyharm

science. The desiretodoaccurateworkis a social value. Sois thebelief that knowledge willultimately benefit ratherthan harmhumankind. Onesimplymust acknowl-edge that values do contributetothe motivationsand

conceptual outlook of scientists. Thedangercomeswhen scientistsallow values tointroduce biases into their work that distort theresults ofscientificinvestigations.

Thesocial mechanismsof sciencediscussed lateract to minimizethedistorting influences of social and personal values. Butindividual scientistscanavoidpitfalls by tryingtoidentifytheir own valuesand theeffectsthose

values haveontheir science. One of the bestways todo

this isby studyingthehistory,philosophy, andsociology ofscience.Humanvalueschangeveryslowly,andthe lessons of thepastremainof greatrelevancetoday.

Judging

Hypotheses

Values emerge intoparticularly sharp relief whena long-establishedtheorycomesinto conflict withnew

observa-tions. Individualresponses tosuchsituationsrange be-tween twoextremes. At oneendof thespectrumis the

notionthatatheorymustberejectedorextensively modifiedas soonas oneofitspredictionsisnotborneout byanexperiment. However,history is full ofexamples in

whichthiswouldhave been premature becausenotenough

wasknown to makeanaccurateprediction. Aclassic

ex-N

RAYS

Self-delusion isnot adanger only for individual scientists. Sometimesanumberof scientists can get caught upin scientificpursuitsthat later prove to be unfounded.

Oneofthemostfamousexamples of such "pathological science" is the history ofN rays. Inthe first fewyearsofthe twentiethcentury,shortlyafterthe discoveryof X raysby theGermanphysicistWilhelm Roentgen,thedistinguished

FrenchphysicistReneBlondlot announced that he had discovered a new type of radiation. Blondlot named the new

radiationNraysafterthe University of Nancy, where he was professor ofphysics. The rays weresupposedly produced

byavarietyof sources,includingelectricaldischarges withingasesand heatedpieces of metal; they couldberefracted through aluminumprisms; and they could be detected by observing faint visual effectswhere the rayshit phosphorescent orphotographic surfaces. Within afewyears,dozensofpapersdescribing the properties ofN rayshadbeenpublished in journals byeminentscientists.

Otherscientists, however,found itimpossibletoduplicate the experiments. One such scientistwastheAmerican physicistRobert W.Wood, who traveledtoBlondlot'slaboratory in 1904towitness theexperiments for himself. After viewingseveral inconclusiveexperiments,Wood was shown anexperimentby Blondlotin whichNraysgenerated by a

lampwerebentthroughanaluminumprismand fell on aphosphorescentdetector. At onepointintheexperiment,

Wood tookadvantageof the room's darknesstosurreptitiouslyremovethealuminumprismfromtheapparatus.

Never-theless,Blondlot continued to detect thevisualsignalsthat hebelievedwerecausedbyNrays.

Inanarticle in Naturepublished shortly after his visit,Wood wrotethat hewas"unableto report asingle observation

whichappearedtoindicatetheexistence of therays." Scientific workonN rays sooncollapsed, and previous results wereshowntobeexperimental artifactsortheresultofobservereffects. YetBlondlot continued tobelieve in the existenceofN raysuntilhisdeathin 1930.

(12)

Proc. NatL Acad Sci USA 86(1989) ampleinvolves Charles Darwin'sdefenseof the theory of

evolution.After Darwin presentedhis theory,physicists argued that theageof theEarth-then calculatedtobe between24 million and 100millionyearsbasedonthe lossofthe heat generated by the Earth'sformation-could

notpossibly be long enough forDarwinian evolutionto haveoccurred. Doggedly, although admittedly rather

miserably,Darwinhung on. Only afterhis death was he

vindicated. When physicistsdiscoveredradioactivityand

realized that naturalradioactive heatingmustbeincluded inthe Earth's heatbudget,there proved to beplentyof timefornaturalselectiontohaveproduced today's

spe-cies.

On theotherhand, history also containsmanyexamples of scientistswhoheldontoanoutdatedtheoryafter it had beendiscredited. Humanbeingshave a strongtendencyto clingtolong-established ideaseveninthe face of

consid-erableopposing evidence. Atrend in the datacanalways

beresistedbyciting uncertaintiesinthe observationsorby

supposingthatunknown factorsareatwork.

Hanging on for awhileto afavorite butembattled ideais

often anecessity duringthe initial stages ofresearch. But scientists must also learn to give way inlight ofnewand moreinsistentevidence. Knowingwhy anidea isso

appealing,orwhy countervailing evidence issostrongly resisted,canhelp a persondevelopthis fine sense of

dis-crimination.

Peer

Recognition

and

Priority of

Discovery

"A

large number

of

incorrect

conclusions

are

drawn

because the

possibility of

chance

occurrences

is

not

fully

consid-ered.

This

usually

arises

through lack of

proper

controls and

insufficient

repeti-tions.

There

is

the

story

of the research

worker in nutrition who

had

published

a

rather surprising conclusion concerning

rats.

A

visitor

asked him ifhe could

of the evidence. The researcher

re-plied, Sure, there's the

rat.

"

-E.Bright Wilson, Jr.,AnIntroductiontoScientific Research,

NewYork:McGraw-Hill, 1952,p.34

Humanvalues are also anintegral part of the forces that

motivatescientists. Theseforcesarenumerousand

psy-chologically complex. Theyincludecuriosityabout the natural orsocial world,the desire to better the human

condition, andafeelingof awe, whetherreligiousor secular,atdiscerningtheworkings ofnature.

Anotherimportantmotivatingforce inscienceisadesire forrecognition byone's peers. Oneof thegreatest rewardsscientistscanexperience istohavetheirwork ac-knowledgedandpraised by other scientistsand

incorpo-ratedinto theircolleagues' research. Sometimesthe quest forpersonal creditcanbecome

counterproductive,

aswhen

time,energy, or evenfriendshipsarelosttopriority disputesoradhominempolemics. Butastrongpersonal

attachmenttoanideaisnotnecessarilyaliability. Itcan evenbeessentialindealingwith the greateffortand

frequentdisappointments associated with scientific research.

Inscience,thefirstperson or grouptopublisharesult

generallygetsthe lion's share of creditforit,evenif

anothergroupthathasbeenworkingontheproblemmuch

longerpublishes thesameresultjustalittle later. (Actu-ally,priorityis dated from whenascientificjournal re-ceivesamanuscript.) Oncepublished, scientificresults become the

public

propertyoftheresearch

community,

but theirusebyotherscientistsrequiresthattheoriginal discovererbe

recognized.

Onlywhen results have become 9064

Report

(13)

Proc. NatL Acad Sci. USA 86 (1989) 9065 commonknowledgearescientistsfree to use them without

attribution.

Indeciding when to make a result public, a scientist weighsseveral competing factors. If a result is kept private,researcherscancontinue to check its accuracy and useit tofurther their research. Butresearcherswhorefrain from publishingrisk losing credit to someone else who

publishesfirst. Whenconsiderationssuch aspublic

acclaim or patentrightsareadded tothe mix,decisions

about whentopublishcanbedifficult.

Social mechanisms

in

science

The

Conmunal Review

of

Scientific

Results

G iventhemorassof

preconceptions,

fallible

methods,andhuman valuesdescribed in the previouspages, a personmightwonderhow

sciencegetsdoneatall. Yet the

large

and rapidly expandingbody ofscientific knowledge,resistant tochangeandeminently successful in its practical applica-tion,atteststothe tremendoussuccessof theenterprise. The linkbetween thetwodomains, between the volatile microcosm ofindividual scientistsandthe solid macro-cosmof scientificknowledge, lies largelyin thesocial

structureof the scientificcommunity.

Ifscientistswerepreventedfromcommunicating with

eachother, scientificprogresswouldgrindto ahalt.

Science isnotdone inisolation;noris it done from first principles. Scientific research takes place withinabroad socialandhistoricalcontext,whichgivessubstance, direction, and,ultimately, meaningtothework of

individ-ualscientists.

Researchers submittheirobservations and hypothesesto thescrutiny of othersthroughmanyinformalandformal mechanisms. They talktotheircolleagues and supervisors

inhallways andoverthetelephone, airing their ideas and modifyingthem in thelight ofthe responsesthey receive. They give presentationsatseminarsandconferences, exposingtheirviewsto abroader but stilllimited circle of colleagues. Theywrite uptheir resultsand sendthemto

scientificjournals, which inturnsend thepaperstobe

scrutinizedby reviewers. Finally, whenapaperhasbeen

published,

it isacceptedorrejected by the communityto

theextentthat it isusedorignored by other scientists.

Ateach stage,researchers submittheir work to be examinedby others with thehope that itwillbeaccepted.

This process ofpublic,systematic skepticism is criticalin

science. Itminimizestheinfluenceofindividual subjec-tivity by requiring that research results be accepted by

otherscientists. Italso isapowerfulinducement for

re-searcherstobecritical of theirown

conclusions,

because theyknowthat theirobjectivemustbetoconvince their

ablestcolleagues,including those with contrastingviews.

Bypassingthe standardroutesofvalidationcan short-circuit theself-correctingmechanisms of science. Scien-tists who releasetheir results

directly

tothe

public-for

example, throughapress conference called to announcea discovery-riskadverse reactions later iftheir resultsare showntobemistakenor aremisinterpretedbythemedia orthepublic. Publication inascientificjournalincludes

important

aspects ofquality

control-particularly,

critical review

by

peers whocandetectmistakes, omissions,and alternativeexplanations. Ifinformation transmitted

throughthemassmediacannotbesubstantiatedlater,the

publicmaynotbelieveother,morecareful researchers. For thisreason, many journals do not accept papers whose results havebeenpreviouslypublicizedbytheir authors. Whenapressrelease iswarranted,itshould be scheduled

onlywhenpeer review iscomplete(normally, in

conjunc-tion withpublicationinascientific

joumal).

Whilepublicationinapeer-reviewedjournalremains the standard means ofdisseminatingscientific results, other methods of communication aresubtlyalteringhow scientistsdivulgeand receiveinformation. The increased use ofpreprints,abstracts,andproceedingsvolumes and

technologies suchascomputer networks and facsimile machines aresimultaneously increasing the speed of com-munication andloosening the network of social controls imposed on formal publication. These new methods of communication are often simply elaborations of the informalexchanges that pervade science. But reliance on such means ofinformation exchange should not be allowed to weaken themechanisms ofquality control that operatesoeffectivelyin science.

Replication and the Openness

of

Communication

Therequirement that results be validatedby one's peers explains why scientific papers must be written in sucha way that theobservations in them can bereplicated. How-ever, actual replication in science is selective: it tends to be reserved forexperiments with unusualimportance or for

experimentsthat conflict with anacceptedbodyof work. Mostoften,scientists who hear or read aboutaresult that affects their own research buildonthat result. If

some-thinggoeswrong with thesubsequent work,researchers may then return to theoriginalresults and attempt to duplicate them.

Scientists build onprevious results because it is not

practical(or necessary)toreconstructalltheobservations and theoreticalconstructsthat go into an

investigation.

They

make the

operating

assumption

that

previousinvestigators performedworkasreportedand adheredtothemethodsprescribed by thecommunity. If that trust ismisplacedand thepreviousresultsare inaccu-rate,the truth willlikelyemergeasproblems arisein the

ongoinginvestigation. But monthsoryearsofeffort may be wasted in the process. Thus, thesocialstructureof sci-enceminimizeserrorsin thelongrun

through

peer

(14)

Proc. NatLAcad Sci USA 86 (1989)

"As the world of

science has grown

in size and

in

power,

its deepest problems

have

changedfrom the epistemological to

the

social.... The increase and

improve-ment

ofscientific knowledge

is a

very

spe-cialized and delicate

social process,

whose continued health and vitality under

new conditions is

by no means takenfor

granted.

"

Jerome Ravetz, ScientificKnowledgeand Its SocialProblems, Oxford, England:ClarendonPress, 1971, p.10

verification. But in the short term science operates on a

basis oftrustandhonestyamongitspractitioners.

The needfor skeptical reviewofscientific results isone reasonwhyfireeand opencommunicationis soimportant

inscience. Different scientistscanreview thesamedata

and, drawingontheirowntheories andvalues,differ in

their interpretations of those data.Thebenefits ofopenness

donotnecessarily imply, however,thatall scientific data

should beavailabletoallpersonsin allcircumstances. In

theinitial, sometimes bewilderingstagesofresearch,a

scientist is entitledto aperiod of privacyinwhichdataare notsubjecttopublic disclosure. This privacy allowsthe

creativeprocess tocontinuewithoutfear ofprofessional embarrassment and allowsindividualstoadvance their worktothepoint at which they can haveconfidencein its accuracy. Manyscientistsarevery generousindiscussing their preliminary theoriesorresultswithcolleagues,and some evenprovide copies of raw data to others prior to

public disclosuretofacilitate related work. The standards ofscienceencouragethesharingof dataand other research toolsatthisstage,butthey donotdemand it.

After publication, scientistsexpectthat data andother researchmaterials willbesharedupon request.Sometimes

thesematerialsare toovoluminous, unwieldy,orcostly to

sharefreely and quickly. Butinthose fieldsinwhich sharing is possible,ascientist who isunwillingtodivulge research datatoqualified colleaguesruns a greatrisk of notbeingtrusted orrespected. Becauseofthe continued needforaccess todata, researchers should keepprimary data foraslongasthere isanyreasonableneed toreferto

them. Ofcourse,researchers who sharetheir datawith others should receive full credit fortheuseofthose data.

Thesharing of dataand otherresearchtoolsissubjectto certain constraints. Individualsrequestingsuch

informa-tion needtohave demonstratedanabilitytodevelop conclusions relevanttothe field ofinquiry fromrawdata.

Scientistsalsoare notobligedtoshareresearch materials withpeople who theysuspect areacting solelyonthebasis ofcommercialorother private interests. Forinstance,a

university biologist wouldnotbeobligatedto turn over a potentially valuablereagenttoscientists in industry.

However,scientists shouldnotdenyrequestsfor accessto primary data because of professional jealousy.

Inresearchthathasthepotential of being financially profitable,openness canbemaintained bythe grantingof

patents. Patentsofferprotectionforthe commercial promise ofascientific discovery inreturnfor

making

the

resultspublic. However,patenting isnotalwaysan option. Therefore,manyscientists,particularlyinindustry

but also inacademia,mustmaintainsomelevel of secrecy intheir work. Scientistsworkingonweaponsor defense-related research alsogenerallyacceptthe necessityfor secrecy insome areas. Butscientistsworkingunder such

conditions shouldrecognize the

potential

dangers of

secrecyinfostering unproductiveresearch andshielding

results from professional scrutiny. 9066 Report

(15)

Proc. NatL Acad Sci USA 86 (1989) 9067

Scientific

Progress

If thereis one thing on which almost all scientists would agree,it is thatscienceis aprogressive enterprise. New observations andtheories survive thescrutiny of scientists

and earn aplacein the edifice ofscientific knowledge

becausetheydescribe thephysicalorsocial worldmore completely or more accurately. Relativistic mechanicsisa morethoroughdescription of whatweobservethan New-tonianmechanics. TheDNAmoleculeisadouble helix.

Ourapelike ancestorswalkederectbefore brain sizes

greatly increased.

Given theprogressive natureofscience,alogical

question is whether scientists can everestablishthat a

particular theory describes theempirical world with completeaccuracy. The notion isatemptingone,anda numberof scientists haveproclaimed thenearcompletion

of research inaparticular discipline (occasionallywith comical resultswhen thefoundations ofthatdiscipline shortly thereafterunderwentaprofoundtransformation).

Butthe nature of scientific knowledge arguesagainstour everknowingthat agiven theoryis the final word. The reasonlies in the inherentlimitationsonverification. Scientists canverifyahypothesis, saybytestingthe valid-ity ofaconsequencederived fromthat hypothesis. But

verificationcanonly increase confidencein atheory,never provethetheory completely,because aconflictingcase canalwaysturnupsometimeinthe future.

Becauseof the limitsonverification, philosophershave

suggestedthat amuchstrongerlogicalconstrainton scientific theories isthattheybefalsifiable. Inother

words, theories must have thepossibilityofbeing proved

wrong,because thenthey can bemeaningfully tested againstobservation. This criterionoffalsifiabilityisone way todistinguish scientific from nonscientific claims. In thislight,theclaims ofastrologersorcreationistscannot bescientific because these groups willnotadmitthat their ideas can befalsified.

Falsifiability isastrongerlogical constraintthan

verifiability, butthebasic problem remains. General state-mentsaboutthe worldcan neverbe absolutelyconfirmed

onthe basis of finite evidence,and all evidence is finite. Thus, science is progressive, but it is an open-ended progression. Scientifictheories are always capable of

being reexaminedandifnecessaryreplaced. Inthissense, anyoftoday's mostcherishedtheories may prove to be

onlylimiteddescriptionsof theempirical worldand at leastpartially "erroneous."

Human Error

in

Science

Error caused by the inherent limits on scientifictheories canbe discoveredonly through the gradual advancement

ofscience, buterrorofa morehuman kind alsooccursin

science. Scientistsare notinfallible;nordotheyhave

limitlessworking timeor access tounlimitedresources. Eventhe mostresponsiblescientistcanmake an honest

mistake. When sucherrors arediscovered, theyshouldbe acknowledged,

preferably

inthesamejournal inwhichthe mistakeninformationwaspublished. Scientists whomake

suchacknowledgments promptly and graciouslyare not

usually

condemned

by colleagues. Others can imagine makingsimilar mistakes.

Mistakesmade whiletryingtodoone'sbestare

toler-ated inscience;mistakes madethrough negligentworkare not. Haste,carelessness, inattention-anyof anumber of faultscanleadtoworkthat doesnot meetthe standards demanded inscience. Inviolating the methodological standardsrequired byadiscipline,ascientistdamagesnot only hisorherownwork butthework

of

othersaswell.

Furthermore, because thesourceof the error may behard toidentify, sloppinesscancostyearsofeffort, bothforthe scientistwho makestheerrorandforothers who tryto buildonthatwork.

Somescientistsmayfeelthat thepressuresonthemare aninducementtospeedratherthancare.

They

may

believe, forinstance, that they haveto cutcornersto compilealong list of publications.Butsacrificing quality

THE

HISTORICAL ORIGINS OF PRIORITY

The systemofassociating scientific prioritywithpublication took shape during the seventeenthcenturyin theearlyyears ofmodernscience. Eventhen,atension existed between theneed of scientiststohaveaccess tootherfindingsanda desiretokeepworksecret sothat others wouldnotclaimit astheirown. Scientists ofthetime, includingIsaacNewton,

wereloathetoconvey newsoftheir discoveriestoscientific societies for fear thatsomeoneelsewouldclaimpriority,a fear thatwasfrequentlyrealized.

To ensurepriority,manyscientists, includingGalileo,Huygens, and Newton,resortedtoconstructinganagrams

describingtheirdiscoveriesthatthey would then make knowntoothers. Forinstance, thelaw"mass timesacceleration equalsforce" could bedisguisedas"a remote,facilequestionscaresclams"

(though

Newtonwould have constructed his anagramsinLatin). Later, ifsomeoneelsecameupwith thesamediscovery, the original discoverercouldunscramble

the anagramtoestablishpriority.

The solutiontotheproblemof

making

newdiscoveriespublicwhileassuringtheir authors creditwasworkedoutby

HenryOldenburg,the secretaryoftheRoyalSociety ofLondon. Hewon overscientistsbyguaranteeing rapid

publica-tion in thePhilosophicalTransactions ofthesocietyaswellas

the

officialsupportof thesociety incasethe author's prioritywasbroughtintoquestion. Thus, itwasoriginally

the

needto ensureopencommunication insciencethatgave risetothe conventionthatthe firsttopublishaviewor afinding,notthe firsttodiscoverit,getscreditfor thediscovery.

(16)

Proc. NatL. Acad. Sci. USA 86 (1989) tosuch pressures islikelytohaveadetrimentaleffecton a

person'scareer. Thenumber ofpublicationstoone's name,though a factorinhiring or promotion decisions, is notnearly asimportantasthequalityof one's overall work. Tominimize pressure to publish substandard work, anincreasing numberof institutionsareadopting policies

thatlimit the number of papers considered when evaluat-ing an individual.

Fraud in

Science

Thereis asignificantdifference betweenpreventableerror inresearch, whether causedbyhonest mistakes orby sloppy work,andoutright fraud. Inthecaseof error, sci-entists do not intend topublish inaccurate results.But whenscientists commitfraud, theyknowwhattheyare doing.

Of all theviolations of the ethos ofscience,fraud is the

gravest. Aswith error, fraud breaks the vital link between humanunderstandingand theempirical world,alinkthat is science's greateststrength. Butfraud goesbeyonderror toerode thefoundation oftrust onwhichscience is built.

Theeffects of fraudonotherscientists,intermsof time

lost,recognition forfeitedtoothers,andfeelingsof

personal betrayal,canbedevastating. Moreover,fraudcan directlyharmthosewhorelyonthefindingsofscience,as when fraudulent results become thebasis ofamedical

treatment. Moregenerally, fraud underminesthe confi-dence andtrustofsocietyinscience,withindirectbut

potentiallyserious effectsonscientificinquiry.

Fraud has beendefinedtoencompassawidespectrum ofbehaviors. Itcanrangefromselecting only thosedata that supportahypothesisand

concealing

therest ("6cook-ing"data) to changing thereadingsto meetexpectations

("trimming" data)tooutright fabrication ofresults.

Thoughit mayseemthatmakingupresults is somehow

FRAUD AND

TIHE

ROLE OF

INTENTIONS

The acid test ofscientific fraud is the intention to deceive, butjudging the intentions ofothers is rarely easy. The case ofWilliam Summerlin illustrates both situations: an instance of blatantfraud and a previous history in which the origins ofserious discrepancies are harderto determine.

In 1973 Summerlin came to the Sloan-Kettering Institute for Cancer Research in New York, where he subsequently

became chief ofa

laboratory working

on

transplantation immunology.

For the

period

of some days or weeksbefore

transplantation,

the immune reaction thatusually causes the

transplant

to be

rejected

could be avoided. The work had become well-known to scientists andto thepublic.

However, other scientists werehaving trouble replicating Summerlin's work. Another immunologist at Sloan-Kettering was

assigned

to repeat some ofSummerlin's experiments, but he, too, could not make the experiments work. As doubts were growing, Summerlin began a series ofexperiments in which he grafted patches of skin from black mice onto white mice. One morning as Summerlin was carrying some ofthe white miceto the director of the institute to demonstrate his progress, he took a

felt-tipped

pen from his pocket anddarkened some oftheblack skingrafts on two white mice. After the meeting, a laboratory assistant noticed that the dark color could be washed away with alcohol,

and within a few hours the director knew ofthe incident. Summerlin

subsequently

admitted his deception to the direc-tor andto others.

Summerlin was

suspended

from his duties and a six-member committee conducted a review oftheveracity ofhis scientific work and his

alleged misrepresentations concerning

that work. In particular, inaddition to reviewing the "mouse incident," the committee examined a series ofexperiments in which Summerlin and several collaborators had

transplanted

parts ofcorneas into the eyes ofrabbits. The committee found that Summerlin had incorrectly and

repeatedly exhibited or reported oncertain rabbits as each having had two human

corneal

transplants, one unsuccessful

from a fresh cornea and the other successful from a cultured cornea. In fact, only one cornea had beentransplanted to

each

rabbit,

and all were unsuccessful.

When asked to

explain

this serious discrepancy, Summerlin stated that he believed that the protocol called foreach

rabbit to receive a fresh cornea in one eye and a cultured cornea in theother eye. Summerlin subsequently admitted

that he did notknow and was not in a position to know which rabbits had undergone this protocol, and that he only

assumed what

procedures

had been carried out on the rabbits he exhibited. After reviewing the circumstances of what the

investigating

committee characterized as "this

grossly misleading

assumption,"

thereport ofthe investigating com-mittee stated: "The

only possible

conclusion is that Dr. Summerlin was responsible for initiating and perpetuating a

profound

and serious

misrepresentation

about the results oftransplanting cultured human corneas to rabbits."

The

investigating

committee concluded that "some actions ofDr. Summerlin over a considerable period of time were not those ofa

responsible

scientist." There were indications thatSummerlin may have been suffering from emotional

illness,

and the committee's report recommended "thatDr. Summerlinbe offered a medical leave of absence, to alleviate his

situation,

which may have been exacerbated bypressure of the many obligations which he voluntarily

undertook." The reportalso stated that, "for whateverreason," Dr. Summerlin's behaviorrepresented "irresponsible

conduct thatwas

incompatible

with

discharge

of his

responsibilities

in the scientific community." 9068 Report

(17)

Proc. NatL Acad Sci USA 86(1989) 9069

"We thus begin to see that

the

institu-tionalized practice ofcitations and

refer-ences

in the

sphere oflearning

is not a

trivial matter While

many

a

general

reader-that is, the lay reader located

outside the domain ofscience and

schol-arship-may regard the lowlyfootnote

or

the remote

endnote or the

bibliographic

parenthesis as a

dispensable nuisance,

it

can be

argued

that these are in truth

central to the incentive

system

and an

underlying sense

ofdistributive justice

that do much to

energize

the

advancement

ofknowledge.

"

Robert K.Merton,"The MatthewEffectinScience, II: Cumula-tiveAdvantage and theSymbolism of IntellectualProperty,"Isis

79(1988):621

moredeplorablethancookingortrimmingdata, all three areintentionally misleading and deceptive.

Instancesof scientific fraud have received a great deal of

publicattention in recent years, which may have exagger-atedperceptions of its apparent frequency. Over the past fewdecades,several dozen cases of fraud have come to light in science. These cases represent a tiny fraction of the total output ofthelarge and expanding research

community. Ofcourse,instancesofscientificfraud may goundetected,ordetectedcasesof fraudmaybe handled privately within research institutions. Butthereisagood reasonforbelieving the incidenceof fraud insciencetobe

quite low. Becausescienceis a cumulative enterprise, in

which investigatorstestandbuild on the workoftheir

predecessors, fraudulent observationsandhypothesestend

eventuallytobe uncovered. Science could not be the

successful institution it is if fraudwere common. The

social mechanisms of science, and in particular the

skepticalreviewandverification of published work,act to

minimizetheoccurrenceof fraud.

The

Allocation

of Credit

Fraud may be the gravestsininscience,buttransgressions that involvethe allocation of credit and responsibilityalso

distorttheinternalworkings ofthe profession. Inthe

standard scientific paper,creditisexplicitly acknowledged

in twoplaces: at the beginning in the listofauthors, and at the end inthe list of referencesorcitations(sometimes accompanied by acknowledgments). Conflictsoverproper attributioncanariseinbothplaces.

Citationsserve anumber ofpurposesinascientific

paper.Theyacknowledgethe work ofotherscientists, directthe reader towardadditionalsourcesofinformation, acknowledge conflicts with other results,andprovide

supportfortheviewsexpressed in thepaper. More

broadly,

citations

place

apaper within itsscientificcontext, relating ittothepresent stateof scientificknowledge.

PATENT

PROCEDURES

Insome areasofresearch,ascientistmaymakeadiscoverythathas commercialpotential. Patenting isa meansof protectingthatpotential while continuingtodisseminatetheresults of the research.

Patentapplications involve such issuesasownership, inventorship, and licensing policies. Inmanysituations, ownershipofapatentisassignedto aninstitution, whetherauniversity,acompany,or agovernmental organization. Some institutions shareroyalty incomewith theinventors. Universities andgovernmentlaboratories

usually

havea policy of licensing inventionsinamannerconsistent with thepublic interest,atleastincasesinwhich federal fundshave

supportedthe research.

Scientists whomaybedoing patentable work haveanobligationtothemselves andtotheiremployersto

safeguard

intellectualpropertyrights. Particularly in industryorinanationallaboratory, thismayinvolve promptdisclosure ofa valuablediscoverytothepatentofficial of theorganization in which the scientist works. Italso entailskeeping accu-rately dated notebook recordswritten ininkinaboundnotebook, ideally witnessed andsigned byacolleague who isnot acoinventor. Datascribbledinpencilonscrapsofpaperinterleavedinloose-leafnotebooks,besidesbeing profession-allyundesirable,areofno useinapatentdispute.

Under U.S.patentlaw,apersonwho inventssomething firstcanbegrantedapatentevenifsomeoneelsefilesaclaim firstsolongaswitnessedlaboratoryrecords demonstratetheearlier invention. Anypublicdisclosureofthediscovery priortofiling foraU.S. patentcan

jeopardize

worldwidepatentrights.

(18)

Proc. NatL Acad Sci USA 86(1989) Citationsarealsoimportant because they leave a paper

trail for later workerstofollow in case things start going wrong. If errors crop up in a line of scientific research,

citationshelp in tracking down thesourceof the discrepan-cies. Thus,in additiontocredit, citations assign responsi-bility. Theimportance of this function iswhy authors should dotheirbest toavoidcitationerrors, a common

problem in scientificpapers.

Science is bothcompetitive and cooperative. These opposing forces tendtobeplayedoutwithin "invisible colleges," networks of scientistsinthesamespecialtywho

readanduseeachother's work.Patternsofcitationswithin these networksareconvoluted andsubtle.Ifscientists cite work by other scientiststhat they have usedinbuilding theirowncontributions,theygainsupportfrom theirpeers

butmaydiminish theirclaimsoforiginality. On theother hand, scientists who failtoacknowledgethe ideas of others tendtofind themselvesexcluded from the fellow-ship of theirpeers. Suchexclusioncandamageaperson's

scienceby limitingtheinformalexchange of ideas with

otherscientists.

Itisimpossibletoprovideasetof rules that would

guaranteetheproperallocation of creditincitations. But scientists haveanumberofreasons tobe generous in their

attribution. Mostimportant, scientists haveanethicaland

professional obligationtogive othersthecreditthey deserve. Thegolden ruleofenlightened self-interestis

alsoaconsideration:Scientists whoexpect tobe treated fairly by othersmust treatothersfairly. Finally, giving

propercredit isgood for science. Sciencewillfunction mosteffectively if those who participate in it feel that they aregettingthecreditthey deserve.Onereasonwhy scienceworksaswellasitdoesisthatitisorganizedso that natural humanmotivations, suchasthedesiretobe

acknowledgedfor one'sachievements,contribute to the overallgoalsof theprofession.

Credit and

Responsibility

in

Collaborative

Research

"Whether or not you agree that

trim-ming

and

cooking

are

likely

to

lead on to

downrightforgery,

there

is

little

to

sup-port

the argument that trimming and

cooking

are

less

reprehensible

and

more

forgivable.

Whatever the rationalization

is,

in

the last analysis

one can no

more

be

a little

bit dishonest than

one can

be

a

little

bit

pregnant. Commit any

of

these

three

sins

and your

scientific

research

career is in

jeopardy

and

deserves

to

be."

C. IanJackson,Honor inScience, New Haven, Conn.: Sigma Xi, The Scientific Research Society, 1984,p. 14

Successful collaborationwith others isoneof themost rewarding experiencesinthe lives ofmostscientists. It canimmenselybroadenaperson's scientificperspective

and advancework farbeyond whatcanbeaccomplished alone. Butcollaboration alsocangeneratetensions betweenindividuals andgroups. Collaborative situations arefarmorecomplexnowthantheywere ageneration

ago. Manypapers appearwithlargenumbers of coau-thors, andanumberofdifferent laboratoriesmaybe

involved, sometimesindifferent countries. Expertsinone field maynotunderstand incompletedetail thebasis ofthe workgoingoninanother. Collaborationthereforerequires agreatdealof mutual trust andconsiderationbetween the

individuals andgroupsinvolved.

Onepotential problemareaincollaborative research involves thelisting ofapaper'sauthors. In manyfields the earliera nameappearsinthelist of authorsthegreater 9070

Report

(19)

Proc. NatL Acad

Scd

USA 86 (1989) 9071 the impliedcontribution,butconventions differ greatly

amongdisciplinesand amongresearch groups. Sometimes thescientist with the greatest name recognition is listed first, whereasin other fields the research leader's name is always last. In somedisciplines,supervisors' namesrarely

appear onpapers,whileinothers the professor'sname appears onalmost every paper that comes out of the lab.

Well-established scientistsmaydecide to list their names after thoseofmorejunior colleagues,reasoning that the youngerscientiststherebyreceive a greater boost in reputation than theywould if theorder were reversed. Someresearchgroupsandjournals avoid these decisions

by simplylisting authors alphabetically.

Frank and opendiscussion ofthedivision of credit within research groups, as early in the process leading to a

publishedpaperaspossible,canavoid laterdifficulties. Collaboratorsmustalso haveathorough understandingof

the conventions in aparticularfield to know ifthey are

being treated fairly.

Occasionallya nameisincludedinalist of authorseven thoughthatpersonhadlittleornothingtodo with the

genesisorcompletion ofthe paper. Such"honorary authors"dilutethecreditduethepeoplewhoactuallydid the workand makethe properattributionof creditmore difficult. Somescientific journalsnow statethat a person should belistedastheauthor ofapaperonlyif that person madeadirect and substantialcontributiontothepaper. Of

course,suchterms as"direct" and "substantial"are

them-selves opentointerpretation. Butsuch statementsof principle help changecustomarypractices, which is the onlylastingway todiscouragethepracticeofhonorary authorships.

Aswithcitations,authorlistings establishresponsibility aswellascredit. When a paperis showntocontain error, whethercausedbymistakes orfraud,authorsmight wish todisavowresponsibility,sayingthattheywerenot involvedin the part of the papercontainingthe errors or thattheyhad verylittletodowith thepaperingeneral.

However,anauthorwhoiswillingtotake credit fora papermustalso bearresponsibilityforitscontents. Thus,

unlessresponsibility is apportioned explicitly inafootnote orinthebody ofthe paper, the authors whose names appearon apaper must bewillingtoshareresponsibility

for all of it.

Apportioning

Credit Between

Junior

and

Senior

Researchers

Thedivisionofcreditcanbeparticularly sensitive when it

involvespostdoctoral, graduate,orundergraduatestudents ontheonehand andtheirfacultysponsorsonthe other. In

thissituation, differentroles andstatuscompoundthe

difficultiesofaccording recognition.

Anumber ofconsiderationshavetobeweighedin

determining theproperdivisionof creditbetweenastudent

or research assistant and asenior scientist, and a range of

practicesareacceptable. Ifasenior researcher hasdefined

and put aproject into motion and a junior researcher is invited tojoin in, major credit may go to the senior researcher, evenifatthemomentof discovery thesenior

researcherisnot

present.

Just asproductioninindustry

entailsmorethan workersstanding at machines, science

entailsmorethan the single researcher manipulating

equipment or solvingequations. Newideas must be generated,lines ofexperimentation established,research

fundingobtained, administrators dealt with,courses

taught, the laboratory kept stocked,informedconsent obtained fromresearchsubjects,apparatusdesignedand

built, andpaperswritten anddefended. Decisionsabout

howcredit istobeallottedfor these and many other

contributionsarefarfrom easy andrequire seriousthought

andcollegialdiscussion.If indoubt aboutthedistribution

ofcredit,aresearchermusttalkfrankly with others, including the seniorscientist.

Similarly, when a student or research assistant is making anintellectualcontributionto aresearchproject,that

contribution deservestoberecognized. Senior scientists arewell awareoftheimportanceofcreditin the reward systemofscience, andjunior researcherscannotbe expectedtoprovide unacknowledged labor if theyare actingasscientificpartners. Insuch cases,junior

re-searchersmaybe listed ascoauthorsorevensenior authors,dependingonthework, traditions withinthefield,

and arrangementswithin theteam.

Plagiarism

Plagiarismis the mostblatant formofmisappropriationof credit. A broad spectrum of misconductfallsinto this category,rangingfrom obvious thefttouncredited

para-phrasingthatsomemightnotconsider dishonestatall. In alifetime ofreading, theorizing,andexperimenting, a person'swork willinevitablyincorporate and overlapwith thatof others. However, occasionaloverlapisonething; systematic, unacknowledgeduseofthetechniques, data,

words orideas of others isanother.

Erring

ontheside of excessgenerosity in attributionis best.

Theintentionaluseof another's intellectualproperty withoutgiving creditmay seem moreblameworthythan theactions ofapersonwho claimstohaveplagiarized

becauseof inattentionorsloppiness. But,asin thecaseof

fraud,the harmtothevictim is thesameregardlessof

intention.Furthermore,giventhedifficulty ofjudging intentions, thecensureimposed by the scientific commu-nityislikelytobeequallygreat.

Specialcare mustbe takenwhendealingwith unpub-lished materialsbelongingtoothers, especiallywith grant

applications and papersseen orheardpriortopublication orpublic disclosure. Suchprivileged materialmust notbe

exploitedordisclosedtootherswhomightexploitit. Sci-entists alsomustbeextremely carefulnottodelay

publica-tionordenysupporttowork thattheyfindtobe competi-tivewith theirowninprivileged communication.

Scrupu-loushonesty is essentialinsuchmatters.

On Being a Scientist PNAS-1989-Research Article-9053-74

On

Being a Scientist

(1989)

Report

Committee

on

the

Conduct of Science

Report

Preface

honesty

Academy

Acknowledgments

On Being

a

Scientist

Contents

major

genetics.

flash;

(1989)

The

nature

of

scientific research

Is

There

Scientific

Method?

"Scientists

people of

dissimilar

temperaments

doing

different

things

in

different

Among scientists

collectors,

classifiers and compulsive

detectives

by

tempera-ment

and

explorers;

artists

and

others

artisans.

There

poet-scientists and

philosopher-scientists

and

few mystics.

"

The Treatment of Data

The Relation Between Hypotheses

and

Observations

centigrade"

Leeuwenhoek,

The Risk

of

Self-Deception

spiral

powerful

night sky.

spiral

Milky

distinguished.

(1989)

unexpected

Methods and

Their Limitations

the

discovery-curiosity,

Values in

Science