The Structure of Scientific Revolutions

ThomasS. Kuhn

TheStructure of Scientific Revolutions Third Edition

The Universityof ChicagoPress Chicagoand Lordon

The Universityof ChicagoPress,Chicago60637 The Universityof ChicagoPress,Ltd., London @ 1962,1970,1996by The Universityof Chicago All rightsreserved. Third edition 1996 Printedin the United Statesof America 050403020100 (cloth) ISBN: 0-226-45807-5 (paPer) ISBN: 0-226-45808-3

345

Data Library of CongressCataloging-in-Publication Kuhn,ThomasS. The structurcof scientificrevolutions/ ThomasS. Kuhn.- 3rd ed' p.cm. andindex. Includesbibliographicalreferences ISBN 0-22645807-5(cloth : alk. paper)' ISBN 0-22645808-3(pbk' : alk'paper) L science--Philosophy.2. Science--History.I. Title. Ql7s.K95 1996 96-13195 501-dc20 CIP of the @ fhe paperusedin this publicationmeetsthe minimum requirements for of Paper Sciences-Permanence for Information Standard National American PrintedLibrary Materials,ANSI 239.48-1992.

Contents Preface vii I. Introduction: A Rolefor History I U. The Routeto NormalScience I0 m. The Natureof Normal Science 23 fV. NormalScienceasPuzzle-solving 35 V. The Priority of Paradigms 43 VI. AnomalyandtheEmergence of ScientificDiscoveries 52 VII. Crisisandthe Emergence of ScientificTheories 66 Vm. TheResponse to Crisis 77 I)(. The NatureandNecessityof ScientificRevolutions 92 X. RevolutionsasChangesof World View I I I XI. The Invisibility of Revolutions 136 )(II. The Resolutionsof Revolutions 144 )(III. ProgressthroughRevolutions 160 Postscript-1969174 Index 2I I

Prefoce The essaythat follows is the first full published report on a project originally conceived almost fffteen years ago. At that time I was a graduate student in theoretical physics already within sight of the end of my dissertation.A fortunate involvement with an experimental college course treating physical sciencefor the non-scientistprovided my ffrst exposureto the history of science.To my complete suqprise,that exposureto out-of-date scientiffc theory and practice radically undermined some of my basic conceptionsabout the nature of scienceand the reasonsfor its specialsuccess. Those conceptionswere ones I had previously drawn partly from scientiffctraining itself and partly from a long-standing avocational interest in the philosophy of science.Somehow, whatever their pedagogicutility and their abstract plausibility, thosenotionsdid not at all fft the enterprisethat historicalstudy displayed. Yet they were and are fundamental to many diicussionsof science,and their failuresof verisimilitudetherefore seemedthoroughly worth pursuing. The result was a drastic shift in my_career plans, a shift from physics to history of science and then, gradually, from relatively straightforward historical problemsback to the more philosophicalconcernsthat had initially led me to history. Except foi a few articles, this essayis the ff-rstof my published works in which these early concernsare dominant.In_somepart it is an attempt to explain to myself and_to friends how I happened to be dt"*tt ito* scienceto its history in the first place. - Yr fi1stopportunity to pursuein depth someof the ideasset forth below was provided by three y"atr as a Junior Fellow of the society of Fellows of Harvard uttiversity. without that period of freedom ihe transition to a new ffeld of study would have beenfar more difficult and might not have beenalhieved. Part of !/ time in thoseyearswas devoted to history of science proper. In particular I continued to study the writings of AlexYll

Prefoce andre Koyr6 and ffrst encounteredthose of Emile Meyerson, H6ldne Metzger, and AnnelieseMaier.r More clearly than most other recent scholars,this group has shown what it was like to think scientiffcallyin a period when the canons of scientiffc thought were very different from those current today. Though I increasinglyquestiona few of their particular historicalinterpretations, their works, together with A. O. Loveioy's Great Chain of Being, have been secondonly to primary sourcematerialsin shapingmy conceptionof what the history of scientiffc ideascan be. Much of my time in thoseyears,however,was spent exploring fields without apparentrelation to history of sciencebut in which researchnow disclosesproblemslike the oneshistory was bringing to my attention. A footnote encounteredby chance led me to the experimentsby which JeanPiagethas illuminated both the various worlds of the growing child and the process of transitionfrom one to the next.2One of my colleaguesset me to reading papersin the psychologyof perception,particularly the Gestalt psychologists;another introduced me to B. L. Whorf's speculationsabout the effect of language on world view; and W. V. O. Quine openedfor me the philosophical puzzlesof the analytic-syntheticdistinction.sThat is the sort of random exploration that the Society of Fellows permits, and only through it could I haveencounteredLudwik Fleck'salmost unknown monograph,Entstehung und. Entu;icklung einer usis1 Particularly infuential were Alexandre Koyr6, Etud.es Galll4ennes (8 raols.; Paris, 1939); Emile Meyerson, Identity ard Reality, trans. Kate Loewenberg ( New York, 1980 ); H6l0ne Metzger, Lei dnarines chlmiqtns en Frarrce du illbit du XVlle d la fin du )(Vllle stdcle (Pans, 1923), and Nerotoa, Stalil, Boeilwaoe a Ia doailrc chimiquc (Paris, 1930); and Anneliese Maier, Db Vorhufet GaIh ("Studien zur Nahrrphilosophie der Spltscholastik"; Leis im 74. Iahrhutderf Rome, f949). 2 Because they &splayed concepts and processesthat also emerge directly from the history of science, two sets of Piaget's investigations proved particularly important: ihe Childs Cotrceptbn of C"ausotity, Ea"ns. Mar'j,orie Cibain (Loidoru (Paris, f946). 1930), and Les rwtions de moutsement et de oltesse clvzfenlont 8 Whorfs papers have since been collected by B. Langtnge, Carroll, John (New-Yoik, Thouglt, atd, Realitg-Seleaed Lee Wlwt Wfitings of Beniaiin f 956). Quine has presented his views in "Two Dogmas of Empiricisrn," reprinted in his From a Logical Pokrt ol Viruu: (Cambridge, Mass., l95g), pp. 20-,1O.

Yiii

Preloce settsclwftlichen Tatsache (Basel, 1935), an essay that antici-

ences to either these works or conversationsbelow, I am debted to them in more ways than I can now reconstruct or evaluate. During my last year as a Junior Fellow, an invitation to lecture for the Lowell Institute in Boston provided a first chance to try out my still developing notion of science.The result was a seriesof eight publie lectures,deliveredduring March, 1951, on "The Quest for PhysicalTheory." h the next year I began to teach history of scienceproper, and for almost a decadethe problems of instructing in a field I had never systematically studied left little time for explicit articulation of the ideas that had first brought me to it. Fortunately, however, those ideas proved a sourceof implicit orientation and of some problemstructure for much of my more advancedteaching.I therefore have my studentsto thank for invaluable lessonsboth about the viability of my views and about the techniquesappropriate to their effectivecommunieation.The sameproblemsand orientation give unity to most of the dominantly historical, and apparently diverse,studiesI have published since the end of my fellowship. Severalof them deal with the integral part played by one or another metaphysic in creative scientific research. Others examinethe way in which the experimentalbasesof a new theoryarc accumulatedand assimilated by men committed to an incompatibleolder theory. In the processthey describe the type of developmentthat I have below called the "emergence"of a new theory or discovery.There are other such ties besides. The ffnal stage in the development of this essay began with an invitation to spend the yeir 1958-59 at the cinter for Advancedstudiesin the Behavioralsciences.once again I was able to give undivided attention to the problems discussed below.Even more important,spendingthe year in a community tx

Prefoce composed predominantly of social scientists confronted me with unanticipated problems about the differences between such communities and those of the natural scientists among whom I had been trained. Particularly, I was struck by the number and extent of the overt disagreementsbetween social scientistsabout the nature of legitimatescientificproblemsand methods.Both history and acquaintancemade me doubt that practitioners of the natural sciencespossessfirmer or more perm_anentanswersto such questions than their colleaguesin socialscience.Yet, somehow,the practiceof astronomy,physics, chemistry, or biology normally fails to evoke the controversies

time provide model problems and solutions to a community of practitioners. Onee that piece of my puzzle fell into place, a draft of this essayemergedrapidly. The subsequenthistory of that draft need not be recounted here, but a few words must be said about the form that it has preservedthrough revisions.Until a ffrst versionhad been com-

much indebted to them, particularly to Charles Morris, for wielding the essentialgoad and for advising me about the

an essayrather than the full-scale book my subiect will ultimately demand. Sincemy most fundamentalobiective is to urge a changein x

Prefoce the perception and evaluationof familiar data, the schematic characterof this first presentationneed be no drawback. On the contrary,readerswhoseown researchhaspreparedthem for the sort of reorientationhere advocatedmay find the essayform both more suggestiveand easierto assimilate.But it has disadvantagesas well, and thesemay iustify ^y illustrating at the very start the sorts of extensionin both scopeand depth that I hope ultimately to include in a longer version.Far more historical evidenceis available than I have had spaceto exploit below. Furthermore, that evidencecomesfrom the history of biological as well as of physical science.My decisionto deal here exclusively with the latter was made partly to increasethis essay's coherenceand partly on grounds of present competence.In addition, the view of scienceto be developedhere suggeststhe potential fruitfulness of a number of new sortsof research,both historical and sociological.For example,the manner in which anomalies,or violations of expectation, attract the increasing attention of a scientiffc community needs detailed study, as does the emergenceof the crises that may be induced by repeatedfailure to make an anomalyconform. Or again, if I am right that eachscientific revolution alters the historical perspective of the community that experiencesit, then that change of perspective should afrect the structure of postrevolutionary textbooksand researchpublications. One such effect-a shift in the distribution of the technical literature cited in the footnotes to researchreports-ought to be studied as a possibleindex to the occurrenceof revolutions. The needfor drasticcondensationhas alsoforced me to forego discussionof a number of maior problems. My distinction betweenthe pre- and the post-paradigmperiodsin the development of a scienceis, for example,much too schematic.Each of the schoolswhose competition characterizesthe earlier period is guided by somethingmuch like a paradigm;there are circumstances,though I think them rare, under which two paradigms can coexistpeacefully in the later period. Mere possessionof a paradigm is not quite a sufficient criterion for the developmental transition discussedin SectionII. More important, exxl

Prefoce cept in occasionalbrief asides,I have said nothing about the role of technological advance or of external social, economic, and intellectual conditions in the development of the sciences. One need, however, look no further than Copernicus and the calendarto discoverthat external conditions may help to transform a mere anomaly into a source of acute crisis. The same examplewould illustrate the way in which conditions outside the sciencesmay influencethe range of alternativesavailable to the man who seeksto end a crisis by proposing one or another revolutionary reform.r Explicit consideration of effects like these would not, I think, modify the main thesesdevelopedin this essay,but it would zurely add an analytic dimension of ffrst-rateimportancefor the understandingof scientific advance. Finally, and perhaps most important of all, limitations of spacehave drastically affected my treatment of the philosophical implications of this essay'shistorically oriented view of science.Clearly, there are such implications, and I have tried both to point out and to document the main ones.But in doing so I have usually refrained from detailed discussion of the various positions taken by contemporary philosophers on the correspondingissues.Where I have indicated skepticism,it has more often been directed to a philosophical attitude than to any one of its fully articulated expressions.As a result, someof thosewho know and work within one of trhosearticulated positions may feel that I have missedtheir point. I think they will be wrong, but this essayis not calculated to convince them. To attempt that would have required a far longer and very different sort of book. The autobiographical fragments with which this preface r Thesefactorsare discussedin T. S. Kuhn, The CopemlcanReoohnbn:Phtpy AfiotwmV tury Astronomgin the Deoelopment Deoelopment_of of Western Westen firougl* flwugl* (Cambridge, Mass., 1957), pp. 12?-32, 27|.l-^71. Other effectsof external intellectual and-economic cundifio-ni upon condiuons condiuorxr upon substaDtive substantivescieDtrtrc development evelopment are are ruusrated illustrated in illusEated scientiffc development in my mv Dalrers. mv DaDers. Daners.

"Consenratioln of Energy as an Example rle of Simultaneous Simultaneous Discovery," Discovery," er*;/lcol er*;bol koblemt ln koblems lnthe the Hfrtory HMor{'olof Science, Science, ed. ed.-trlarshall Marshall Clagett Clagett ((Madison,liris., Madison, lg59), "E-ngineering Precedent pp. 821-56; 821-5-6; "Engineering kecedent for the Work o[ o{ Sadi Carnot," Carnot,' Archloes l* 'Sadi tenatUnules thi*oire d,asccbtwes, XIII ( 1960), 247-5li and Carnot and

the CagnardEngine," Isis, LII ( 196l ), 567:l4.It is, therefore,only with lespect to the problens iliscussedin tl'is essaythat I take the role of externil factors t6 be rninor.

xii

Prefoce opens will serve to acknowledgewhat I can recognize of my main debt both to the works of scholarship and to the instihrtions that have helped give form to my thought. the remainder of that debt I shall try to dischargeby citation in the pagesthat follow. Nothing said above or below, however, will more than hint at the number and nature of my personalobligationsto the many individuals whose suggestionsand criticisms have at one time or another sustainedand directed my intellectual development. Too much time has elapsedsince the ideas in this essay began to take shape; a list of all those who may properly ffnd some signs of their infuence in its pageswould be almost coextensivewith a list of my friends and acquaintances.Under the circumstances,I must restrict myseUto the few most significant infuences that even a faulty memory will never entirely suPPress. It was |ames B. Conant, trhenpresident of Harvard University, who ffrst introduced me to the history of scienceand thus initiated the transformation in my conception of the nature of scientiffc advance.Ever since that processbegan, he has been generousof his ideas, criticisms, and time-including the time required to read and suggestimportant changesin the draft of my manuscript. Leonard K. Nash, with whom for ffve years I taught the historically oriented c€urse that Dr. Conant had started, was an even more active collaborator during the years when my ideas ffrst began to take shape,and he has been much missedduring the later stagesof their development.Fortunately, however, after my departure from Cambridge, his place as creative soundingboard and more was assumedby my Berkeley colleague, Stanley Cavell. That Cavell, a philosopher mainly concernedwith ethics and aesthetics,should have reachedconclusions quite so congruent to my own has been a constant sourceof stimulation and enoouragementto me. He is, furthermore, the only person with whom I have ever been able to explore my ideas in incomplete sentences.That mode of communication attests an understanding that has enabled him to point me the way through or around severalmaior barriers encourtered while preparing my first manuscript. xill

Prefoce Since that version was drafted, many other friends have helped with its reformulation. They will, I think, forgive me if I name only the four whose contributions proved most farreachingand decisive:Paul K. Feyerabendof Berkeley,Ernest Nagel of Columbia,H. PierreNoyesof the LawrenceRadiation Laborator/, and my student,John L. Heilbron, who has often worked closelywith me in preparing a ffnal versionfor the press. I have found all their reservationsand suggestionsextremely helpful, but I have no reasonto believe (and somereasonto doubt) that either they or the othersmentioned above approve in its entirety the manuscriptthat results. My ffnal acknowledgments, to my parents,wife, and children, must be of a rather different sort. In ways which I shall probably be the last to recognize,eachof them, too, has contributed intellectualingredientsto my work. But they havealso,in varying degrees,done somethingmore important. They have, that is, let it go on and evenencouragedmy devotionto it. Anyone who haswrestledwith a project like mine will recognizewhat it has occasionallycost them. I do not know how to give them thanks. T. S. K. lBxlrrr.gv, Cer.rronxr.l February 1962

xtY

A Rolefor History l. Introduclion;

from which eachnew scientiffcgenerationleams to practice its trade. Inevitably, however, the aim of such books is persuasive and pedagogc; a concept of science drawn fiom them is no more likely to fft the enteqprisethat produced them than an image of a national culture drawn from a tourist brochure or a language text. This essayattempts to show that we have been misled by them in fundamental ways. Its aim is a sketchof the quite different concept of science that can emerge from the historical record of the researchactivity itseU. Even from history, however, trhat new concept will not be forthcoming if historical data continue to be sought and scrutinized mainly to answer questions posed by the unhistorical stereotype drawn from science texts. Those texts have, for example,often seemedto imply that the content of scienceis uniquely exemplified by the observations,laws, and theories described in their pages.Almost as regularly, the same books have been read as saying that scientific methods are simply the onesillustrated by the manipulative techniquesused in gathering textbook data, together with the logical operations employed when relating those data to the textbook's theoretical generalizations.The result has been a concept of sciencewith profound implications about its nature and development. If scienceis the constellationof facts, theories,and methods collected in current texts, then scientistsare the men who, successfully or not, have striven to contribute one or another element to that particular cunstellation.Scientiffcdevelopmentbecomes the piecemealprocessby which these items have been

fhe Sfrucfureof ScientificRevofutions added,singly and in combination,to the ever growing stockpile that constitutesscientifictechnique and knowledge.And history of science becomes the discipline that chronicles both these successiveincrements and the obstaclesthat have inhibited their accumulation.Concernedwith scientiffcdevelopment,the historian then appearsto have two main tasks.On the one hand, he must determineby what man and at what point in time each contemporaryscientiffcfact, law, and theory was discoveredor invented. On the other, he must deseribeand explain the congeries of error, myth, and superstition that have inhibited the more rapid accumulation of the constituents of the modern sciencetext. Much researchhasbeendirectedto theseends,and somestill is. In recent years, however, a few historians of science have been finding it more and more difficult to fulffl the functions assignsto that the concept of development-by-accumulation them. As chroniclers of an incremental proctss, they discover that additional researchmakesit harder, not easier,to answer questionslike: When was oxygen discovered?Who first conceived of energy conservation?Increasingly,a few of them suspect that these are simply the wrong sorts of questionsto ask. Perhapssciencedoesnot developby the accumulationof individual discoveriesand inventions.Simultaneously,these same historians confront growing difffculties in distinguishing the "scientific"componentof past observationand belief from what their predecessorshad readily labeled "elTor" and "superstition." The more carefullythey study, say,Aristoteliandynamics, phlogistic chemistry, or caloric thermodynamics,the more certain they feel that thoseonce current views of nature were, as a whole, neither less scientific nor more the product of human idiosyncrasythan those current today. If theseout-of-datebeliefs are to be called myths, then myths can be produced by the samesorts of methodsand held for the samesorts of reasons that now lead to scientific knowledge.If, on the other hand, they trre to be called science,then sciencehas included bodies of belief quite incompatiblewith the oneswe hold today. Given thesealternatives,the historian must choosethe latter. Out-of-

2

A Rolefor HistorY lnlroduction: date theoriesare not in principleunscientificbecaurytley have been discarded.That c[oice, however,makesit difficult to see scientificdevelopmentas a Processof accretion.The samehistorical researchthat displays the difficulties in isolating individual inventions and discoveriesgives ground for profound doubts about the cumulativeprocessthrough which theseindividual contributionsto sciencewere thoughtto havebeencompounded. The result of all these doubts and difficulties is a historiothough one that js graphic revolution in the study of scie_nce, ititl i.t its early stages.Gradually, and often without entirely realizing they are doing so,historiansof sciencehav-eb-egunto asknew-sortsof questionsand to tracedifferent,and often less than cumulative, developmentallines for the sciences.Rather than seekingthe Permanentcontributionsof an older scienceto dur presentvantage,they attempt to display the historical integrily of that sciencein its own time. They ask, for -example, no"t a'bout the relation of Galileo'sviews to those of modern science,but rather about the relationshipbetweenhis viewsand and immethoseof his group,i.e.,his teachers,contemporaries, diate srrccesiotsin the sciences.Furthermore,they insist uPon other similar onesfrom studying the opinionsof that grouP ""qfrom that of modern scivery difierent ttte "ieripointlusually ence-th at givesthose opinions th e m aximum internal-cnherp-Bglland the clJsestpossiblefit to nature. Seenthrough the works that result, worfs perhapsbest exemplifiedin the writings of the same Ale&pdre_K'g6, icience does not seem altogether older historioin the writers by discussed one tie as enteryrise historical these least, at implication, By tradition. g.uplii" Jt,tii"r suggestthe possibility of a new image of science.This essayaims"fodelineatethat image by making explicit someof implications. the new historiography's What aspectsof science will emerge !o prgminence in the courseof this eflort? First, at least in order of presentation,is theinsufficiencyof qtrSgdgJ-o-g,ry-e$g9S!ry9t-bythemselves,to .+---L

a*-*'

--''"Y_'

conclusion to many sorts of sciendic@stantive tific questionJ. Instructed to examine electrical or chemical Ph"-

TheSfruclureol ScienliffcRevolutions nomena,the man who is ignorant of theseffeldsbut who knows what it is to be scientiftc may legitimately reach any one of a number of incompatible conclusions.Among those legitimate , possibilities, the particular conclusionshe does arrive at are J probably determined by his prior experiencein other ffelds, by / I the accidents of his investigation, and by his own individual makeup. What beliefs about the stars, for example, does he bring to the study of chemistry or electricity? Which of the many conceivableexperimentsrelevant to the new ffeld doeshe elect to perform ffrstPAnd what aspectsof the complexphenomenon that then results strike him as particularly relevant to an elucidation of the nature of chemical change or of electrical affinity? For the individual, at least, and sometimesfor the scientific community as well, answersto questionslike theseare i of scientiffcdevelopment.We shall rn II that the early developmental re been characterizedby continual ber of distinct views of nature,each rll roughly compatiblewith, the dicrn and method. What differentiated thesevarious schoolswas not one or anotherfailure of methodthey were all "scientiffc"-but what we shall come to call their incommensurableways of seeing the world and of practicing sciencein it. Observationand experiencecan and must drastically restrict the range of admissiblescientiftcbelief, elsethere would be no science.But they cannot alone determine a particular bo_dyof such belief. An apparently arbitrary element, compounded of_personal and historical accident, ii always a formative ingredient of the beliefs espousedby . given scientific community at a given time. That elementof arbitrarinessdoesnot, however,indicatethat any scientiffcgroup could practiceits trade without someset of received beliefs. Nor does it make less consequentialthe particular constellation to which the group, at a given time, ii in fact committed. Effective reseatch scarcely Legins before a scientific community thinks it has acquired ffrm answers to questionslike the following: What are the fundamental entities 1

Introduction:A Rolefor History of which the universeis composed?How do theseinteract with What questionsmay legitimateeachother and with the senses? ly be askedabout such entities and what techniquesemployed in seeking solutions?At least in the mature sciences'answers (or full iubstitutes for answers) to questionslike these are ffrmly embeddedin the educationalinitiation that preparesand Iicenies the student for professionalpractice. Becausethat edu-

historic origins and, occasionally,in their subsequentdevelopment. Yet that elementof arbitrarinessis present,and it too has an important effect on scientificdevelopment,one which will be examined in detail in SectionsVI, VII, and VI[. Normal sci-

novelties becausethey are necessarilysubversive of its basic commitments.Nevertheless,so long as those commitmentsretain an element of the arbitrary, the very nature of normal research ensuresthat novelty shall not be suPPressedfor very

fhe Sfructureof ScientificRevolutions to perform in the anticipated manner, revealing an anomaly that cannot, despite repeated effort, be aligned with professional expectation.In these and other ways besides,normal sciencerepeatedlygoesastray.And when it does-when, that is, the professioncan no longer evadeanomaliesthat subvertthe existingtradition of scientificpractice-then begin the extraordinary investigations that lead the professionat last to a new set of commitments,a new basisfor the practice of science.The extraordinaryepisodesin which that shift of professionalcommitments occursare the onesknown in this essavas scientific revolutions.They are the tradition-shatteringcomplementsto the tradition-boundactivity of normal science] The most obviousexamplesof scientificrevolutionsare those famousepisodesin scientiftcdevelopmentthat have often been labeled revelutions before. Therefore, in SectionsIX and X, where the nature of scientificrevolutionsis ffrst directly scrutinized, we shall deal repeatedlywith the major turning points in scientificdevelopmentassociatedwith the namesof Copernicus, Newton, Lavoisier,and Einstein. More clearly than most other episodesin the histoqy of at least the physical sciences,these display what all scientiftcrevolutionsare about. Each of them necessitatedthe community's rejection of one time-honored scientifictheory in favor of another incompatiblewith it. Each produceda consequentshift in the problemsavailablefor scientiffc scrutiny and in the standardsby which the professiondetermined what should count as an admissibleproblem or as a Iegitimate problem-solution.And each transformedthe scientific imagination in ways that we shall ultimately need to describe as a transformationof the world within which scientific work was done. Such changes,together with the controversies that almostalwaysaccompanythem, are the deffningcharacteristicsof scientiftcrevolutions. These characteristicsemerge with particular clarity from a study of, say, the Newtonian or the chemicalrevolution. It is, however,a fundamentalthesisof this essaythat they can also be retrieved from the study of many other episodesthat were not so obviouslyrevolutionary.For the far smallerprofessional 6

lnlrodvcliontA Rolefor HistorY

an isolatedevent. only scientiftcevents that have revolutionary impact upon the specialistsin whose domain they occur. The commitments that govern normal science specify not only what sorts of entities the universe does contain,but also,by implication, those that it doesnot. It follows, though the point will require extendeddiscussion,that a discoverylike that of oxygenor X-rays doesnot simply add one more item to the populatibn of the scientist'sworld. Ultimately it has that efiect, but not until the professionalcommunity has re-evaluated traditional experimental procedures, altered its conceptionof entities with which it has long been familiar, and, in the process,shifted the network of theory through which it dealswith the world. Scientiffcfact and theory arcnot categorically separable,except perhapswithin a single tradition of normal-scientiffcpractice. That is why the unexpecteddiscoveryis not simply factual in its import and why the scientist'sworld is qualitatively transformed as well as quantitatively enriched by fundamental noveltiesof either fact or theory. This extended conception of the nature of scientiffc revolutions is the one delineated in the pagesthat follow. Admittedly the extensionstrainsctrstomaryusage.Nevertheless,I shall conNor are new inventions

fhe Struclureof ScienliffcRevolufions tinue to speakeven of discoveriesas revolutionary, becauseit is iust the pbssibilityof relating their structureto that of, say,the Copernican revolution that makes the extended conception seem to me so important. The preceding discussionindicates how the complementarynotionsbf normal scienceand of scien-

revolutionary competition between the proponents of the old normal-scientifictradition and the adherentsof the new one. It

mies is available to suggestthat it cannot properly do so. Hist_ory,we too often say, is a purely descriptive discipline. The theses suggestedabove are, however, often interpietive and

8

lntroduction:A Rolefor HidorY tion.' Can anything more than profound confusionbe indicated by this admixture of diverseffelds and concerns? Having been weaned intellectually on these distinctions and others[k1 them, I could scarcelybe more aware of their impor! and force. For many yearsI took them to be about the nature of knowledge, and I ;till suPPosethat, appropriately recast, they have ronr'"thitrgimportattt to tell us. Yet my attempts to apply them, even grooi mado, to the actual situations in which knowledgeis gained,accepted,and assimilatedhavemade them seemextraordinarily problematic.Rather than being elementary logical or methodological distinctions, which would thus be priot to the analysis of scientific knowledge, they now t"9m integral parts of a traditional set of substantiveanswersto the very q,restionsupon which they have been deployed. That_circularily doesnot at all invalidate them. But it doesmake them parts of a theory and, by doing so, subiectsthem to the same icrutiny regularly applied to theoriesin other fields. If they are to have more than pure abstraction as trheir content, then that content must be discoveredby observingthem in application to the data they are meant to elucidate. How could history of sciencefail to be a sourceof phenomenato which theoriesabout knowledgemay legitimately be askedto apply?

ll. The Route lo Normol Science In this essay,'normal science'meansresearchfirmly based upon one or more past scientific achievements,achievements that someparticular scientific community acknowledgesfor a time as supplyingthe foundationfor its further practice.Today such achievementsare recounted,though seldomin their original form, by science textbooks, elementary and advanced. Thesetextbooksexpoundthe body of acceptedtheory, illustrate many or all of its successfulapplications,and compare these applicationswith exemplaryobservationsand experiments.Before suchbooksbecamepopular early in the nineteenthcentury ( and until even more recently in the newly matured sciences ), many of the famousclassicsof sciencefulfflled a similar function. Aristotle's Physica, Ptolemy's Alrnagest,Newton's Principia and Opticks, Franklin's Electricity, Lavoisier'sChemistry, and Lyell's Geolagy-theseand many other works servedfor a time implicitly to define the legitimate problemsand methods of a researchfteld for succeedinggenerationsof practitioners. They were able to do sobecausethey sharedtwo essentialcharacteristics.Their achievementwassufficientlyunprecedentedto attract an enduring group of adherentsaway from competing modes of scientific activity. Simultaneously,it was sufficiently open-endedto leave all sorts of problems for the redefined group of practitionersto resolve. Achievementsthat share these two characteristicsI shall henceforthrefer to as'paradigms,'aterm that relatescloselyto 'normal science.'Bychoosingit, I mean to suggestthat some acceptedexamplesof actualscientificpractice-exampleswhich include law, theory,application,and instrumentationtogetherprovide modelsfrom which springparticular coherenttraditions of scientific research.These are the traditions which the historiandescribesundersuchrubricsas'Ptolemaicastronomy'(or 'Copernic"r'),'Aristotelian dynamics'(or'Newtonian'),'corpuscularoptics'(or'wave optics'),and so on. The study of l0

fhe Routefo Normol Science paradigms, including many that are far more specializedthan those named illustratively above, is what mainly preparesthe student for membershipin the particular scientificcommunity with which he will later practice. Becausehe there joins men who learned the basesof their ffeld from the same concrete models,his subsequentpractice will seldom evoke overt disagreementover fundamentals.Men whoseresearchis basedon sharedparadigmsare committed to the samerules and standards for scientificpractice.That commitmentand the apparent consensusit producesare prerequisitesfor normal science,i.e., for the genesisand continuationof a particular researchtradition. Becausein this essaythe concept of a paradigm will often substitutefor a variety of familiar notions,more will need to be said about the reasonsfor its introduction.\Mhy is the concrete scientificachievement,as a locus of professionalcommitment, prior to the variousconcepts,Iaws,theories,and points of view that may be abstractedfrom it? In what senseis the shared paradigm a fundamentalunit for the student of scientific development, a unit that cannot be fully reduced to logically atomic componentswhich might function in its stead?When \Meencounterthem in SectionV, answersto thesequestionsand to otherslike them will prove basicto an understandingboth of normal scienceand of the associatedconcept of paradigms. That more abstract discussionwill depend, however, upon a previous exposureto examplesof normal scienceor of paradigms in operation.In particular, both these related concepts will be clarified by noting that there can be a sort of scientific researchwithout paradigms,or at least without any so unequivocaland so binding as the onesnamedabove.Acquisition of a paradigm and of the more esoterictype of researchit permits is a sign of maturity in the developmentof any given scientific field. If the historiantracesthe scientificknowledgeof any selected group of related phenomenabackward in time, he is likely to encountersomeminor variant of a pattern here illustratedfrom the history of physicaloptics.Today'sphysicstextbookstell the

1l

fhe Sfruclureof ScienfificRevolufions student that light is photons, i.e., quantum-mechanicalentities that exhibit soire chiracteristics of walnesand someof particles. accordinglYtot rather accordingto $e.more Researchproceeds ^and mathematical characterization from which this elaborate usual verbalization is derived. That characterizationof light is, however, scarcelyhalf a century old. Before it was developed by Planck, Einstein, and otheri early in this century, physics texts taught that light was transversewave motion, a concePtion roo6d in a p-aradigmthat derived ultimatgly from th3 optical writings of Yo.to! and Fresnel in the early nineteenth by cintury. Nor ivas the wa-vetheory the first t9 be emb_raced During science. optical gi_gltof al-ort all practitioners !\e eenth centuiy the paradigm for this field was Provided by N:*ton's Opticki, which taught that light was material coqput-d"t. At that time physicistssought evidence,as the early -wavetheorists had notlof the pressureexertedby light particlesimpinging on solid bodies.l th.rc transformationsof the paradigmsof physicaloptics are scientiffc revolutions, and the successivetransition from one usual developmental paradigm 'pattern'ofto another via revolution is the the P-attern.charhowever, is not, 'acteristic mature science.It and that is the work, Newton's of the period before remote anbetween No period here. contrast that corrcernsus a exhibjted century the sevenGenth tiquity and the end of Illight. of nature the about sitigle generally accepted_view -n,rmber of competing schools and substeid i"h.r, *br. a schools,most of them espousingone viriant or anothe-r-ofEpiAristotelian, or PiatoniJtheory. One group togk light to ",rr""rr, be pariicles emanatingfrom m-aterialbodies; for another it was -lodifi"ation of the riedium that intervenedbetween the body "and the eye; still another explainedlight in-terms of an inter-"tt 'the from the eye; and medium with action of "*atation besides.Each modifications and there were other combinations its relation strength-from derived of the colrespondingschools as Paraemphasized, each and to someparticular metaphysic, r loseoh Priestley, The Htslallr? atd Prcse* State of Dlscooerbc RelAlng to Liglrt, ardColours (London, 17721, pp. 88L90' Visi;,

12

fhe Roufelo Normol Science digmatic observations,the particular cluster of optical phenomena that its own theory could do most to explain.Other observations were dealt with by d hoc elaborations,or they remained as outstanding problemsfor further research.2 At various times all these schoolsmade signiffcant contributions to the body of concepts,phenomena,and techniquesfrom which Newton drew the ffrst nearly uniformly acceptedparadig* for physicaloptics.Any deffnition of the scientistthat excludes at least the more creative members of these various schoolswill excludetheir modern successors as well. Thosemen were scientists.Yet anyone examining a survey of physical optics before Newton may well conclude that, though the ffeld's practitioners were scientists,the net result of their activity was something less than science. Being able to take no common

schoolsas it was to nature. That pattern is not unfamiliar in a

The history of electrical researchin the ffrst half of the eighteenth centuqy provides a more concrete and better known exampleof the way a sciencedevelopsbefore it acquiresits ffrst universally received paradigm. During that period there were almost as many views about the nafure of electricity as there were important electrical experimenters,men like Hauksbee, Gray, Desaguliers, Du Fay, Nollett, Watson, Franklin, and others. All their numerous concepts of electricity had something in common-they were partially derived from one or an2 vasco Ronchi, Hlstohede k.luml*'e, trans. Iean Tatoa (paris, 1956), chaps. l-lv.

t3

fhe Sfruclureof ScienfiffcRevolulions other version of the mechanico-corpuscular philosophy that guided all scientiffcresearchof the day. In addition, all were componentsof real scientiffctheories,of theoriesthat had been drawn in part from experimentand observationand that partially determined the choice and interpretation of additional problems undertaken in research.Yet though all the experiments were electrical and though most of the experimenters read eachother'sworks, their theorieshad no more than a family resemblance.s One early group of theories, following seventeenth-century practice, regarded attraction and frictional generationas the fundamentalelectricalphenomena.This group tended to treat repulsionas a secondaryeffect due to somesort of mechanical rebounding and also to postponefor as long as possibleboth discussionand systematicresearchon Gray'snewly discovered effect, electrical conduction.Other "electricians" (the term is their olvn ) took attraction and repulsion to be equally elementary manifestationsof electricity and modified their theoriesand researchaccordingly.(Actually, this group is remarkably small-even Franklin's theory never quite accounted for the mutual repulsion of two negatively charged bodies.) But they had as much difficulty as the first group in accounting simultaneouslyfor any but the simplest conduction effects. Those effects, however, provided the starting point for still a third group,one which tendedto speakof electricity as a "fuid" that could run through conductorsrather than as an "effiuvium" This group,in its turn, had that emanatedfrom non-conductors. difficulty reconciling its theory with a number of attractive and

r1

fhe Roufelo Normof Science repulsive effects. Only through the work of Franklin and his did a theory arisethat could accountwith immediate successors somethinglike equal facility for very nearly all theseeffectsand that therefore could and did provide a subsequentgenerationof "electricians"with a commonparadigm for its research. Excluding those ffelds, like mathematics and astronomy, in which the ffrst ffrm paradig*r date from prehistory and also those,like biochemistry, that aroseby division and recombination of specialties already matured, the situations outlined above are historically typical. Though it involvesmy continuing to employ the unfortunate simpliffcation that tags an extended historical episodewith a single and somewhatarbitrarily chosen name (e.8., Newton or Franklin), I suggestthat similar fundamental disagreementscharacterized,for example,the shrdy of motion before Aristotle and of statics before Archimedes,the study of heat before Black, of chemistry before Boyle and Boerhaave,and of historicalgeologybeforeHutton. In parts of biology-the study of heredity, for example-the ffrst universally receivedparadigmsare still more recent;and it remainsan open question what parts of social sciencehave yet acquired such paradigms at all. History suggeststhat the road to a ffrm researchconsensus is extraordinarilyarduous. History also suggests,however,somereasonsfor the difficul-

developmentmakesfamiliar. Furthermore, in the absenceof a reasonfor seekingsomeparticularform of more reconditeinformation, early fact-gathering is usually restricted to the wealth of data that lie ready to hand. The resultingpool of facts contains trhoseaccessibleto casualobservationand experiment together with some of the more esoteric data retrievable from establishedcrafts like medicine, calendarmaking, and metallurgy. Becausethe crafts are one readily accessiblesotrtce of facts that could not have been casually discovered,technology

I5

fhe Slruclureol ScienliffcRevolulions has often played a vital role in the emergenceof new sciences. But though this sort of fact-collecting has been essentialto the origin of many signiffcant sciences,anyone who examines, for example,Pliny's encyclopedicwritings or the Baconiannatural histories of the seventeenthcentury will discover that it producesa morass.One somehowhesitatesto call the literature that results scientiffc.The Baconian"histories"of heat, color, wind, mining, and so on, are fflled with information,someof it recondite. But they iuxtaposefacts that will later prove revealirg (e.g.,heatingby mixture) with others(..g., the warmth of dung heaps) that will for sometime remain too complexto be integratedwith theory at all.' In addition, sinceany description must be partial, the typical natural history often omits from its immenselycircumstantialaccountsiust thosedetails that later scientistswill find sourcesof important illumination. Almost noneof the early "histories"of electricity,for example,mention that chaff, attracted to a rubbed glassrod, bouncesoff again. That eftect seemedmechanical,not electrical.bMoreover,since the casualfact-gathererseldompossesses the time or the tools to be critical, the natural historiesoften iuxtaposedescriptions Iike the abovewith others,say,heating by antiperistasis(or by cooling), that we are now quite unable to conftrm.oOnly very occasionally,as in the easesof ancient statics,dynamics,and geometricaloptics, do facts collected with so little guidance from pre-establishedtheory speakwith sufficientclarity to permit the emergenceof a ffrst paradip. This is the situation that createsthe schoolscharacteristicof the early stagesof a science'sdevelopment.No natural history can be interpretedin the absenceof at leastsomeimplicit body _ 1 C_g1p-e th-esketch for a natural history of heat in Bacon's Notum Organum, Vol. VIII of Tfu Works of Frarcis Bacon, ed. J. Spedding, R. L. El[s, aod D. D. Heath (New York, 1869), pp. 17$203. 6 Roller and Roller, op. cit., pp. 14, 22, 28,43. Onlv after the worlc recorded in the last of these citations do-rtpulsive efiects gain leneral recognition as unequivocally electrical. 6 Bacon, op. clt., pp. 235, 337, says, "Water slightly warm is more easily frozen than quite cold." For a partial account of the earliei history of this strange observation, see Marshall Clagett, Giooanni Marltuni atd Ldte Medb:al Fhysics ( New York, l94l ), chap. iv.

l6

fhe Roulefo Normol Science of intertwined theoretical and methodologicalbelief that permits selection,evaluation,and criticism. If that body of belief is not already implicit in the collection of facts-in which case more than "mere facts" are at hand-it must be externallysup-

difierent ways. What is surprising,and perhapsalso unique in its degreeto the fields we call science,is that such initial divergencesshouldever largely disappear. For they do disappearto a very considerableextentand then is apparentlyonceand for all. Furthelrnore,their disappearan_ce usually caused by the triumph of one of the pre-paradigrn schools,which, becauseof its own characteristicbeliefsand pre-

been discoveredby a man exploringnature casuallyor at random, but which was in fact independentlydevelopedby at least two investigatorsin the eatly 1740's.?Almost from the start of Franklin was particularly concernedto his electricalresearches,

paradigm,a theory mrtst seembetter than its competitot's,bttt ? Roller and Roller, op. clt., pp. 5f-54. 8 The troublesomecasewas the mutual repulsionof negativelychargedbodies, for which seeCohen,ry. cit.,pp. 491-94,531-43.

l7

fhe Structureof ScienfiffcRevolulions it need not, and in fact never does,explain all the facts with which it can be confronted. What the fluid theory of electricity did for the subgroupthat held it, the Franklinian paradigm later did for the entire group of electricians.It suggestedwhich experimentswould be worth performing and which, becausedirected to secondaryor to overly complex manifestationsof electricity, would not. Only the paradigm did the job far more effectively,partly because the end of interschooldebateended the constantreiteration of fundamentalsand partly becausethe confidencethat they were on the right track encouragedscientiststo undertakemore precise, esoteric,and consumingsorts of work.8 Freed from the concern with any and all electrical phenomena,the united group of electricianscould pursue selectedphenomenain far moredetail, designingmuch specialequipmentfor the task and employing it more stubbornly and systematicallythan electricians had ever done before. Both fact collection and theory articulationbecamehighly directed activities.The efiectiveness and efficiencyof electricalresearchincreasedaccordingly,providing evidencefor a societalversionof Francis Bacon'sacute methodological dictum: "Truth emerges more readily from error than from confusion."lo We shall be examiningthe nature of this highly directed or paradigm-basedresearchin the next section,but must first note briefly how the emergenceof a paradigm affectsthe structure of the group that practicesthe fteld. When, in the development of a natural science,an individual or group first producesa synthesisable to attract most of the next generation'spractitioners, the older schoolsgradually disappear.In part their disappear-

1oBacon, op. cit., p. 2f0.

l8

fhe Roulefo Normol Science ance is causedby their members'conversionto the new paradigo. But there are always somemen who cling to one or another of the older views, and they are simply read out of the profession,which thereafter ignores their work. The new paradig- implies a new and more rigid definition of the field. Those unwilling or unable to accommodatetheir work to it must proceed in isolation or attach themselvesto some other group.lr Historically, they have often simply stayed in the departments of philosophy from which so many of the special scienceshave been spawned.-As these indicationshint, it is sometimesjust its reception of a paradigm that transforms a group previously interestedmerely in the study of nature into i professionor, at least,a discipline. In the sciences(though nof in ffelds like medicine, technology, and law, of which the principal raison d'atre is an externalsocialneed), the formation of specialized iournals, the foundation of specialists'societies,and the claim for a _specialplace in the curriculum have usually been associated-with a group's first reception of a single paradigm. At Ieast this was the casebetween the time, a cJntury an{a half lgo, -whe1 the institutional pattern of scientiffc specialization first developedand the vgry recent time when the piraphernalia of specializationacquired a prestigeof their own. The more rigid deftnition of the scientific soup has other consequences. the individual scientist cin take a para{hen dign{o-r-Stqt-r$, h" needno longer,inhis maiorworks,att-empt to build his field anew,startingfrom first principlesand iustifyrr The rhe _hi historv tory of electricitv electricity provides an excellent example which could be duplicated from t}e careers of Priesdey, Kelvin, and othe?s. Franklin reoorts that l\o[er, rnar Nollet, wno who ar at nuo-cenrury mid-century was tne ihe most influential of the continental Continintal electricians, "lived to see himseli the last 9f -hi-sSect, except Mr. B.-his Eleve - J

and immediate Disciple" (Max Farrand led.l, Beniamin'Frunkfhts u*riti Calif.. l9-40J, [Berkeley. f9401. pp. op.384-86). Mor" inilresUng, interccriic h^*o.,o, ia +],[Berkeley,Calif., S8L86). Uor" ho*"u"r, is the ^-J..enduianceof ance ot whole whole schools schoolsin in increasingisolation isolationfrom from professional profelsionalscience. science.Consider, Consider. for example,the caseof astrology,-which*as on""'"r, integral part of astronomy. 3ral Or ur Or consider considerthe consider "rL";;;;:centhe continuation continuation in continuation in the i6 the late late eighteenth eishteenth and-earlv arr,l""^"|i, .i-"r..-tt'^-iearly nineteenth nineteenth centuries of a previously respected tradition of "romantic" cheriistrv. This is the tradition discussed by bi, Charles C. C. Gillispie Gillispie in "The Etrcuclmtddlc in "The oi.l th. Ercgclop&d,b_arid the rqrnhi-

Jacobin Philosophyof science: A study in Idelasand consequ"encls,"c;niA-p;;ii;r* in the \*g:V of Scietrce,ed_.MarshallClagett (Madison, Wis., lg5g), pp.255pp. 2SE_ 89; and "Ttie Formation of Lamarclc'sEiolutionary Theory,; eriiirlri iiiin twtbnales d.'histohedes scbncec,XXXVU ((fg56). 1956), g?g+9. "

r9

fhe Struclureof ScienlificRevolufions ing the use of eachconceptintroduced.That can be left to the *titet of textbooks. Given a textbook, however, the creative scientistcanbegin his researchwhere it leavesoff and thus concentrateexclusivelyupon the subtlestand most esotericaspects of the natural phenomenathat concernhis group. Ald as he does this, his researchcommuniqu6swill begin to change in ways whose evolution has been too little studied but whose modernend productsare obviousto all and oppressiveto many. usuallybe embodiedin booksadNo longet*ilIhis researches . . . on Electrinity or DarExperiments like Franklin's dressed-, win's origin of species,to anyonewho might be interestedin the subjectmatter of the fteld. Insteadthey will usually appe-ar as brief articles addressedonly to professionalcolleagues,the men whoseknowledgeof a sharedparadigm can be asstrmed and who prove to bJthe only onesable to read the papersaddressedto them. or retroToday in the sciences,booksare usually e_ither.texts scientific the of or another one asPect uPon spectivereflections life. The scientistwhb writes one is more likely to find his professionalreputation impaired than enhanced.Onlf in the earlier, pre-paiadigm,stagesof the developmentof the various iia tn. book ordinarily Possessthe same relation to scien^ceg professionalachievementthat it still retains in other creative ^fields. And only in thosefields that still retain the book, with or without the article, as a vehicle for researchcommunication still solooselydrawn that the are the lines of professionalization by reading the practiprogretl follow to layman may hope and astronomy' in mathematics Both tioners' originalieports. to be intelliantiquity in already researchrelorts hid ceased research dynamics, In audience. gible to a g:enerallyeducated recapAges, and_it Middle later in the 6u""*" similarlyeioteric sevenearly the during briefly only tured generalintelligibility had one that replacedthe new wheria teenth"century Paradigm require to begln research Electrical research. guided medieval franslationfor the laymanbeforethe end of the eighteenthcentur/, and most otherfields of physicalscienceceasedto be gerraccessiblein the ninetcenth. During the sametwo cen"r"ily 20

fhe Roufeto Normol Science ttrriessimilar transitionscan be isolatedin the variousparts of the biologicalsciences. In parts of the socialsciencesthey may well be occurring today. Although it has become customary, and is surelyproper,to deplorethe widening gulf that separates the professionalscientistfrom his colleaguesin other ffelds,too Iittle attentionis paid to the essentialrelationshipbetweenthat gulf and the mechanisms intrinsicto scientiftcadvance. Ever since prehistoric antiquity one ffeld of study after another has crossedthe divide betweenwhat the historianmight call its prehistoryasa scienceand its history proper.Thesetransitionsto maturity haveseldombeenso suddenor so unequivocal as my necessarilyschematicdiscussionmay have implied. But neither have they been historicallygradual,coextensive, that is to say,with the entire developmentof the ffeldswithin which they occurred.Writers on electricity during the first four decadesof the eighteenthcenturypossessed far more information about electricalphenomenathan had their sixteenth-century predecessors. During the half-centuryafter 1740,few new sortsof electricalphenomenawere added to their lists. Nevertheless,in important respects,the electricalwritings of Cavendish, Coulomb, and Volta in the last third of the eighteenth century seemfurther removedfrom thoseof Gray, Du Fay, and even Franklin than are the writings of these early eighteenthcentury electrical discoverersfrom those of the sixteenthcentury.I2SometimebetweenL740and 1780,electricianswere for the first time enabledto take the foundationsof their field for granted.From that point they pushedon to more concreteand reconditeproblems,and increasinglythey then reported their resultsin articlesaddressedto other electriciansrather than in booksaddressedto the learnedworld at large.As a group they achievedwhat had been gained by astronomersin antiquity

2l

fhe Sfrucfure ol ScientificRevolulions

22

lll. The Noture of Normol Science What then is the nature of the more professionaland esoteric researchthat a groupt receptionof a single paradigmpermits? If the paradigm representswork that has been done once and for all, what further problemsdoesit leave the united group to resolve?Thosequestionswill seemevenmore urgent if we now note one respectin which the termsusedsofar may be misleading. In its establishedusage,a paradigm is an acceptedmodel or pattern, and that aspectof its meaninghas enabledme, Iacking a better word, to appropriate'paradigm' here. But it will shortly be clear that the senseof model'and 'pattern' that permits the appropriation is not quite the one usual in defining 'anxo, paradigm.' fn grammar, for example, artes, amat' is a paradigmbecauseit displaysthe pattern to be usedin coniugating a large number of other Latin verbs, e.9., in producing 'l,audo, lnudns,lnudat.' In this standard application, the paradigm functions by permitting the replication of examplesany one of which could in principle serveto replaceit. In a science, on the other hand,a paradigmis rarely an object for replication. Instead,Iike an acceptediudicial decisionin the commonlaw, it is an obiect for further articulation and speciffcationunder new or more stringentconditions. To seehow this can be so, we must recognizehow very limited in both scopeand precisiona paradigmcan be at the time of its first appearance. Paradigmsgain their statusbecausethey are more successfulthan their competitors in solving a few problemsthat the group of practitionershas come to recognize as acute. To be more successfulis not, however, to be either completelysuccessfulwith a singleproblem or notably successful with any largenumber.The successof a paradigm-whether Aristotle'sanalysisof motion, Ptolemy'scomputationsof planetary position, Lavoisiert application of the balance,or Maxwellis mathematization of the electromagneticfield-is at the start largely a promiseof successdiscoverablein selectedand

23

fhe Sfruclureof ScienfificRevolulions still incompleteexamples.Normal scienceconsistsin the actualization of that promise,an actualizationachievedby extending the knowledge of those facts that the paradigm displays as particularlyrevealing,by increasingthe extentof the match between those facts and the paradigm'spredictions,and by ftrrther articulation of the paradigm itself. Few people who are not actually practitioners of a mature sciencerealizehow much mop-uP work of this sort a paradigm leavesto be done or quite how fascinatingsuchwork can Prove in the execution.And thesepoints needto be understood.Mop-

to invent new theories,and they are often intolerant of thoseinventedby others.lInstead,normal-scientificresearchis directed to the aiticulation of those phenomenaand theories that the ^paradigm already supplies. Perh-apsthese'aretifects. The areasinvestigatedby PTtl scienceire, of course,minuscule;the entelprise now under dis-

restrictions that bound researchwhenever the paradigm from which they derive ceasesto function effectively. At thalP9int scientists6egin to behave differentl)', and-the nature of their researchpro6l"*r changes.In the inierim, however,during the 1 Bernard Barber, "Resistance by Scientists to Scientiffc Discovery," Scbnce,

cxxxN (196r),59G602.

24

fhe Nofure of NormqlScience the professionwill have period when the paradigmis successful, iolved problemsthat its memberscould scarcelyhave imagined and would never have undertaken without commitment to the paradigm.And at leastpart of that achievementalwaysProves to be permanent. To display more clearly what is meant by normal or p-aradigm-basedresearch,let me now attempt to classify and illustrate the problems of which normal scienceprincipally consists. For convenienceI postponetheoretical activity and begin with fact-gathering, that is, with the experimentsand observations describedin the technical journals through which scientistsinform their professionalcolleaguesof the resultsof their continuing research.On what aspectsof nature do scientistsordinarily report? What determines their choice?And, since most scientific observation consumesmuch time, equipment, and money, what motivates the scientist to Pursue that choice to a conelusion? There are, I think, only three normal foci for factual scientiftc investigation, and they are neither alwaysnor Pennanentlydistinct. First is that classof facts that the paradigm has shown to be particularly revealing of the nature of things.By employin_g them in solving problems, the paradigm has made them worth determining both with more precision and in a larger variety of situations.At one time or another,thesesigniffcantfactual determinations have included: in astronomy-stellar position and magnitude,the periods of eclipsingbinaries 1nd of planets;in ph1'sics-the specificgravities and comPressibilitiesof materials, *aue lengths and spectral intensities, electrical conductivities and contact potentials; and in chemistry-composition and combining weights, boiling points and acidity of solutions, structural formulas and optical activities. Attempts to increasethe accuracy and scope with which facts like these are known occupy a signiftcant fraction of the literature of experimental and bbservalional science.Again and again complex special apparatushas been designedfor such purPoses,and the invention, constmction, and deployment of that apparatushave demandedffrst-ratetalent, much time, and considerableffnancial

25

fhe Structureol ScienfificRevolulions backing. Synchrotronsand radiotelescopesare only the most recent examplesof the lengths to which researchworkers will go if a paradigm assuresthem that the facts they seek are important. From Tycho Brahe to E. O. Lawrence,somescientists have acquired great reputations,not from any novelty of their discoveries,but from the precision, reliability, and scope of the methods they developed for the redetermination of a previously known sort of fact. A secondusual but smallerclassof factual determinationsis

the speedof light is greaterin air than in water; or the giganticscintillation counter designedto demonstratethe existenceof

4248.

26

fhe Nolure of NormolScience the neutrino-thesepiecesof specialapparatusand many others like them illustrate the immenseefiort and ingenuity that have beenrequired to bring nature and theory into closerand eloser agreement.s That attempt to demonstrateagreementis a second type of normal experimentalwork, and it is evenmore obviously dependentthan the ffrst upon a paradigm.The existenceof the paradigm sets the problem to be solved; often the paradigm theory is implicated directly in the designof apparatusable to solvethe problem.Without the Principia,for example,measurements made with the Atwood machine would have meant nothing at all. A third class of experimentsand observationsexhausts,I think, the fact-gatheringactivitiesof normal science.It consists of empiricalwork undertakento articulatethe paradigmtheory, resolving some of its residual ambiguitiesand permitting the solution of problems to which it had previously only drawn attention.This classprovesto be the most important of all, and its descriptiondemandsits subdivision.In the more mathematical sciences,someof the experimentsaimed at articulation are directed to the determinationof physical constants.Newton's work, for example,indicated that the force between two unit massesat unit distancewould be the samefor all typesof matter at all positionsin the universe.But his own problemscould be solved without even estimatingthe size of this attraction, the universalgravitationalconstant;and no one elsedevisedapparatus able to determine it for a century after the Principia appeared. Nor was Cavendish'sfamous determination in the 1790t the last.Becauseof its centralpositionin physicaltheory, improved values of the gravitational constant have been the objectof repeatedeffortseversinceby a numberof outstanding 3 For-two o parailax relescopes, ror fwo or of tne telescopes,_see the_paralla_x see ADranam Abraham Wolf, wolt, la A n8torv Historg ol of Jctence, Science, Technology, and Piilosophy inthe-Eighteenth Centurg (2d ed.; Loidon, 1952), pp. 103-5. For the Atwood machine, machine. see N. R. Hanson, Hanson. Patterns^ Pattqns of ol Discooeru Dis_cooery ( Cambridge, 1958 ), pp. 100-102, 207-8. For the last two pieces of special appal ratus, see see-M. M. L, FoGault, Foiri:ault, "M6thode g6n6rale pour me-surer me's,,ter Ia vitess" du e I t" a lumidre dans I'air et les milieux transparints. Viteslsesrelatives de Ia lumidre dans l'air et dans l'eau . . . ," Comptes rendus . . . de I'Acad,6mie des sciences, XXX (1850),551-60; and C. L. Cowan, Ir., et al.,"Detection of the Free Neutrino: A Conffrmation," Scdence,CXXIV (f956), f03-4.

27

fhe Struclure of ScientificRevolufions experimentalists.4Other examples of the sarnc solt of corttinuing work would include determinations of the astronomical unit, Avogadro's number, Joule's coefficient, the electronic charge, and so on. Few of these elabolate efforts would have been conceived and none would have been carried out without a paradigm theory to define the problem and to guarantee the existenceof a stable solution. Efforts to articulate a paradigm are not, horvever, restricted to the determination of universal constants. They may, for example, also aim at quantitative laws: Boyle's Law relating gas pressureto volume, Coulomb's Law of electrical attraction, and foule's formula relating heat generated to electrical resistance and current are all in this category. Perhaps it is not apparent that a paradigm is prerequisite to the discovery of laws like these.We often hear that they are found by examining measurements undertaken for their own sake and without theoretical commitment. But history offers no support for so excessively Baconian a method. Boyle's experiments were not conceivable (and if conceived would have received another interpretation or none at all ) until air was recognized as an elastic fluid to which all the elaborate concepts of hydrostatics could be applied.s Coulomb's successdepended upon his constructing special apparatus to measure the force between point charges. (Those who had previously measured electrical forces using ordinary pan balances,etc., had found no consisteut or simple regularity at all. ) But that design, in turn, depended upon the previous recognition that every particle of electric fluid acts upon every other at a distance. It was for the force between such particles-the only force which might safely be asstrmed 4 H. P[oyntingJ reviews some two dozen measurementsof the gravitational I. consiant between t7+t and l90t in "Gravitation Constant and Mean Density XII, of the Earth," Encyclopaedia Britannrca (llth ed.; Cambridge, l9l0-ll), 385-89. 5 For the full transplantation -Pascal, of hydrostatic concepts into pneumatics, see The trans. L H. B. Spicrs and A. G. H. Spiers, with an Phqsical Treatises of intioduction and notes by F. Barry (New York, 1937). Torricelli's original introduction of the paralleiism ( "We live submerged at the bottom of an ocean of the element air'r) occt,rs on p. 164. Its rapid development is displayed by the two main treatises.

28

fhe Notu re oI Normol Science

a simple function of distance-that Coulomb was looking.o Joule'sexperimentscould alsobe used to illustratehow quantitative laws emergethrough paradigm articulation.In fact, so general and close is the relation between qualitative paradigm and quantitative law that, since Galileo, such Iaws have often been correctly guessedwith the aid of a paradigmyears before apparatus cotrld be designed for their experimental determination.T Finally, there is a third sort of experimentwhich aims to articulate a paradigm. More than the others this one can resemble exploration, and it is particularly prevalent in those periodsand sciencesthat deal more with the qualitative than with the quantitative aspectsof nature's regularity. Often a paradigmdevelopedfor one set of phenomenais ambiguousin its applicationto other closelyrelated ones.Then experiments are necessaryto chooseamongthe alternativeways of applying the paradigm to the new area of interest. For example,the paradigmapplicationsof the caloric theory were to heating and cooling by mixtures and by changeof state.But heat could be releasedor absorbedin many other ways-e.g., by chemical combination,by friction, and by compressionor absorptionof a gas-and to each of theseother phenomenathe theory could be applied in severalways. If the vacuum had a heat capacity, for example,heatingby compressioncould be explainedas the resultof mixing gaswith void. Or it might be due to a change in the specificheat of gaseswith changingpressure.And there were several other explanationsbesides.Many experiments were undertakento elaboratethesevariouspossibilitiesand to distinguishbetwecn them; all theseexperimentsarosefrom the caloric theory as paradigm,and all exploitedit in the designof experimentsand in the interpretationof results.sOncethe phe6 Duane Roller and Duane II. D. Roller, The Deoclopment of the Concept of Electric Charge: Electricity from the Grecks to Coulomb ( "Harvard Case Histories in Experimental Scicnce," Case 8; Cambridge, I{irss., 1954), pp. 66-80. 7 For examples, see T. S. Kuhn, "Thc I.'trnction of Measuremcnt in Modern Physical Science,"Isis, LII (f96f ), 161-93. 8 T. S. Kuhn, "Thc Caloric Thcory of Adiabatic Compression," lsi.r, XLIX ( 1958), 139-40.

fhe Struclureof ScienfificRevolufions Iromenonof heatingby compression had been established,all further experimentsin the area were paradigm-dependentin this way. Giventhe phenomenon, how elsecould an experiment to elucidateit have been chosen? Turn now to the theoretical problems of normal science, which fall into very nearly the sameclassesas the experimental and observational.A part of normal theoretical work, though only a small part, consistssimply in the use of existing theory to predict factual information of intrinsic value. The manufacture of astronomicalephemerides,the computation of lens characteristics,and the production of radio propagationcuryes are examplesof problemsof this sort. Scientists,however,generally regard them as hack work to be relegatedto engineers or technicians.At no time do very many of them appearin significant scientificiournals.But thlse iournalsdo confaitta gt""t many theoretical discussionsof problems that, to the nonscientist,must seemalmost identical. Theseare the manipulations of theory undertaken, not becausethe predictions in which they result are intrinsically valuable, but becausethey can be confronted directly with experiment.Their pulpose is to display a new applicationof the paradigmor to increasethe preeisionof an application that has already been made. The need for work of this sort arisesfrom the immensedifficulties often encounteredin developingpoints of contact between a theory and nature. These difficulties can be briefly illustrated by an examinationof the history of dynamicsafter Newton. By the early eighteenthcentury those scientistswho found a paradigm in the Principin took the generality of its conclusionsfor granted, and they had every reasonto do so. No other work known to the history of sciencehas simultaneouslypermitted solarge an increasein both the scopeand precision-ofresearch.For the heavensNewton had derivedKepler's Laws of planetary motion and also explained certain of the observedrespectsin which the moon failed to obey them. For the earth he had derived the resultsof somescatteredobservations on pendulums and the tides. With the aid of additional but he had alsobeen able to derive Boyle'sLaw adlnc assumptions,

30

fhe Nofure of NormolScience and an important formula for the speedof sound in air. Given the stateof scienceat the time, the successof the demonstrations was extremely impressive.Yet given the presumptive generality of Newton's Laws, the number of these applications was not great, and Newton developed almost no others. Furthermore, compared with what any graduate student of physics can achievewith those samelaws today, Newton's few applications were not even developed with precision. Finally, the Principia had been designedfor application chiefly to problems of celestial mechanics. How to adapt it for terrestrial applications, particularly for those of motion under constraint, was by no means clear. Terrestrial problems were, in any case, already being attackedwith great successby a quite difierent set of tech-

saw quite how.e

point_in order- to provide a unique deffnition of pendulum length.-Mo_stof his theorems,the few e*ceptionsbeing hypothetical and preliminary,alsoignoredthe effectof air resistance. These were sound physical approximations.Nevertheless,as approximationsthey restricted the agreementto be expected

3l

fhe Sfructureof ScienfificRevolulions

To derive those laws, Newton had been forced to neglect all gravitational attraction except that between individual planets and the sun. Since the planets also attract each other, only approximate agreementbetween the applied theory and telescopic observationcould be expected.l0 The ageement obtained was,of course,more than satisfactory to those who obtained it. Excepting for some terrestrial problems,no other theory could do nearly sowell. None of thosewho questionedthe validity of Newton's work did so becauseof its limitd agreementwith experiment and obsenration.Nevertheless,these limitations of agreementleft many fascinating theo Theoretical techniques retical problemsfor Newton's successors. were, for example, required for treating the motions of more than two simultaneouslyattracting bodies and for investigating the stability of perhrrbed orbits. Problemslike these occupied many of Europe's best mathematiciansduring the eighteenth and early nineteenth cenfury. Euler, Lagrange, Laplace, and Gauss all did some of their most brilliant work on problems aimed to improve the match between Newton's paradigm and observationof the heavens.Many of theseffguresworked simultaneouslyto develop the mathematicsrequired for applications that neither Newton nor the contemPoraryContinental schoolof mechanicshad even attempted. fr"y produced, for example, an immerxe fiterature and somevery Powerful mathematical techniques for hydrodynamics and for the problem of vibrating

10wolf, op. cit., pp. 75-81, 9Gl0l; and william whewell, Ilistory of the (i6v. ed.;London,1847),II,213-71. IntluctioeSciences

32

TheNqfure of Normol Science o{y, or any other branch of sciencewhosefundamental laws are fully quantitative. At least in the more mathematical sciences, most theoretical work is of this sort. But it is not all of this sort.Even in the mathematicalsciences there are also theoretical problems of paradigm articulation; and during periods when scientific development is predominantly qualitative,theseproblemsdominate.some of []re problems, in both the more quantitative and more qualitativl sciences,aim simply at clarification by reformulation. The principia-,f-orexample,did not alwaysprove an easywork to apply, partly becauseit retained some of the clumsinessinevitable in a ftrst venture and partly becauseso much of its meaning was only rmplicit in its applications. For many terrestrial applications, in any case, an apparently unrelated set of Conti-nental techniques seemedvastly more pcwerful. Therefore, from Euler an{ in the eighteenth century to Hamilton, !3srange Jacobi, and Hertz in the nineteenth, many of Europe's most briliant mathematical physicists repeatedly endeavoredto reformulate mechanical theory in an equivalent but logieally and.aesthetically more satisfying form. Thuy wished, that is, to exhibit the e4plicit and implicit lessonsof the principia and of Continental mechanicsin a logically more eoherentversion, one that would be at onc€ more uniform and lessequivocal in its application to the newly elaboratedproblems of mechanics.rr similar reformulations_ofa paradigm have occurredrepeatedty-- all o{ the sciences,but most ofthem have produceldmore substantialchangesin the paradigm than the reiormulations of the Principda cited above.-such result from the em"hatrg"s

1#;i1"*f:ffiit#"i1x '.*",1:

nJ;"*"",T:iffi :ffiff

eqgarywethere.Beroreh"':frTi':"9"'JtrJff li:l*l#;:T;

make measurementswith it, coulomb had to emp^loielectrical theory to determine how his equipment should^be'built. The 11Ren6 Dugas, Histoire d.e ln mdcandgue (Neuchatel, lg5O), Books IV_V.

33

fhe Sfructureof ScienlificRevolufions conseguenceof his measurementswas a refinement in that theory, Or again, the men who designedthe experimentsthat were to distinguish between the various theories of heating by compressionwere generally the same men who had made up the versions being compared. They were working both with fact and with theory, and their work produced nolsimply new information but a more preciseparadigm, obtained by the elimination of ambiguities that the original from which they worked had retained. In many sciences,most normal work is of this sort. Thesethree classesof problems-determinationof significant fact, matching of facts with theory, and articulation of theoryexhaust,I think, the literature of normal science,both empirical and theoretical.They do not, of course,quite exhaustthe entire literature of science.There are alsoextraordinaryproblems,and it may well be their resolution that makes the scientific enterprise as a whole so particularly worthwhile. But extraordinary problemsare not to be had for the asking.They emergeonly on specialoccasionspreparedby the advanceof normal research. Inevitably, therefore, the overwhelming majority of the problems undertakenby even the very best scientistsusually fall into one of the three categoriesoutlined above.Work under the paradigm can be conductedin no other w&/, and to desertthe paradigm is to ceasepracticing the scienceit deffnes.We shall shortly discover that such desertionsdo occur. They are the pivots about which scientificrevolutionsturn. But beforebeginning the study of such revolutions,we require a more Panoramic view of the normal-scientificpursuits that prepare the way.

34

lV. Normol Science os Puzzle'solving Perhaps the most striking feature of the normal research problemi we have just encounteredis how little- they aim to Sometimes, iroduce maior novelties,conceptualor phenomenal. as in a wave-lengthmeasurement,everythingbut the most esoteric detail of tlie result is known in advance,and the typical latitude of expectation is only somewhat wider. Coulomb's need not, perhaps,have fitted an inversesquare measurements law; the men who worked on heating by comPressionwere often preparedfor any one of severalresults.Yet even in cases like thesJthe range of anticipated,and thus of assimilable,results is alwayssmall comparedwith the range that imagination can conceive.And the pioject whose outcome doesnot fall in that narrowerrange is Gually iust a researchfailure, one which reflectsnot on nature but on the scientist. In the eighteenth century, for example,little attention was paid to the experimentsthat measuredeleetricalattractionwith devicesIike the pan balance.Becausethey yielded ueither consistentnor simple results,they could not be used to articulate the paradigm from which they derived. Therefore, they remained nlere facts,unrelatedand unrelatableto the continuing of progressof electricalresearch.Only in retrospect,possessed paradigm,can we seewhat characteristicsof eleci snbseqrrent trical phenomenathey display. Coulomb and his contempothis later paradigmor one that, raries,of course,alsopossessed attraction, yielded the same of the to applied problem when expectations.That is why Coulomb was able to design apParatus that gave a result assimilableby paradigm articulation. But it is alsowhy that result surprisedno one and why several had been able to predict it in of Coulomb'scontemporaries advance.Even the proiectwhosegoal is paradigmarticulation doesnot aim at the unexpectednovelty. But if the aim of normal scienceis not major substantivenovelties-if failure to come near the anticipatedresult is usually

35

fhe Slructureof ScienfiffcRevolutions failure as a scientist-then why are theseproblemsundertakerr at all? Part of the answerhasalreadybeendeveloped.To scientists, at least, the resultsgained in normal researchare significant becausethey add to the scopeand precisionwith *t i.r, the paradigm can be applied. Tliat answer,however, cannot accountfor the enthusiasmand devotion that scientistsdisplay for the p_roblems of normal research.No one devotesy"ati to, say,the developmentof a better spectrometeror the production of an improved solution to the problem of vibrating strings simply becauseof the importanceof the information that *itt be obtained.The data to be gainedby computing ephemerides or by further measurementswith an existing instrument are often i,rft as significant, but those activities are regularly spurnedby scientistsbecausethey are so largely repetitionsof proceduresthat have been carried through before.that rejection provides a clue to the fascinationof the normal research problem. Though its outcome can be anticipated,often in detail so great that what remainsto be known is itself uninteresting, the way to achieve that outcome remains very much in doubt. Bringing a normal researchproblem to a conclusionis achieving the anticipated in a new wo/, and it requires the solution of all sorts of complex instrumental,conceplual,and mathematicalpuzzles.The man who succeedsproves himself an expert puzzle-solver,and the challengeof the puzzle is an important part of what usually drives him on. The terms'puzzle' and'puzzle-solver'highlight severalof the themes that have become increasinglyprominent in the preceding pages.Puzzlesare, in the entirely standard meaning here employed,that specialcategoryof problemsthat can serve to test ingenuity or skill in solution.Dictionary illustrationsare 'jigsaw puzzle'and'crosswordpuzzle,'andit is the characteristics that thesesharewith the problemsof normal sciencethat we now need to isolate.One of them has just been mentioned. It is no criterion of goodnessin a puzzle that its outcomebe intrinsicallyinterestingor important.On the contrary,the really pressingproblems,€.8., a cure for cancer or the design of a

36

Normol Scienceos Puzzle'solving

of a solution is. We have already seen, however, that one of the things a scientific community acquireswith a paradigm is a criterion for choosing problems that, while the paradigm is taken for

trated by several facets of seventeenth-centuryBaconianism and by some of the contemporarysocial sciences.One of the reasonswhy normal scienceseemsto progressso rapidly is that its practitionersconcentrateon problems that only their own lack of ingenuity should keep them from solving. If, however, the problems of normal science are puzzles in this sense,we need no longer ask why scientistsattack them with such passion and devotion. A man may be attracted to sciencefor all sorts of reasons.Among them are the desire to be useful, the excitementof exploring new territory, the hope of finding order, and the drive to test establishedknowledge. These motives and others besidesalso help to determine the particular problems that will later engage him. Furthermore, though the result is occasionalfmstration, there is good reason

fhe Slructureof ScienlificReyolufions why motives like these should first attract him and then lead him on.l The scientiffcenterpriseas a whole doesfrom time to time prove useful, open up new territory, display order, and test long-acceptedbelief. Nevertheless,the indirsid:u,al engaged on a norrnal researchproblem is almostneDerdoing any one of these things. Once engaged,his motivation is of a rather difierent sort. What then challengeshim is the conviction that, if only he is skilful enough, he will succeedin solving a puzzle that no one before has solved or solved so well. Many of the geatest scientiffcminds have devoted all of their professional attention to demandingpuzzlesof this sort. On most occasions any particular ffeld of specializationoffers nothing else to do, a fact that makes it no less fascinating to the proper sort of addict. Turn now to another,more difficult, and more revealing aspect of the parallelismbetween puzzles and the problems of normal science.If it is to classifyas a puzzle,a problem must be characterizedby more trhan atrntid solutioir. There must also be rules that limit both the"trnature of acceptablesolutions and the steps by which they are to be obtained. To solve a iigsaw puzzle is not, for example,merely "to make a picfure." Either a child or a contemporaryartist could do that by scattering selected pieces, as abstract shapes,upon some neutral ground. The picture thus produced might be far better, and would certainly be more original, than the one from which the puzzle had been made. Nevertheless,such a picture would not be a solution.To achievethat all the piecesmust be used,their plain sidesmust be turned down, and they must be interlocked without forcing until no holes remain. Those are among the rules that govern iigsaw-puzzle solutions. Similar restrictions upon the admissiblesolutions of crossword puzzles,riddles, chessproblems,and so on, are readily discovered. If we can accept a considerablybroadeneduse of the term r The frustrations induced bv the confict between the.individual's role and the over-all pattern of scientihc development can, however, occasionally be quite serious. On this subject, see Lawrence S. Kubie, "Some Unsolved Problems of the Scientiftc Career," American Scientist, XLI (1953),596-613; and

XLII (1954),r04-r2.

38

Normol Scienceos Puzzlesofving 'established 'rule'-one that will occasionally equate it with 'preconception'-then the problems accesviewpoint' or with sible-within a given researchtradition display som,ething-much like this set oipunle characteristics.The man who builds an instrument to determine optical wave lengths must not be satisfied with a piece of equipment that merely attributes particular numbers to particulai spectral lines. He is not iust an explorer or measurer.On the contrary, he must'show,by analyzinghis apparatus in terms of the establishedbody of optical theory, that the numbers his instrument Producesare the ones that enter theory as wave lengths. If someresidual vaguenessin the theory or some unanalyzed component of his apparatus Prevents his completing that demonstration,his colleaguesmay well concludethat he hasmeasurednothing at all. For example, the electron-scatteringmaxima that were later diagnosedas indices of electron wave length had no apparent significance when first observedand recorded.Beforethey becamemeasures of anything, they had to be related to a theory that predic_ted the waveJike behavior of matter in motion. And even after that relation was pointed out, the apparatushad to be redesignedso that the experimentalresultsmight be correlatedunequivocally with theory.2Until thoseconditionshad been satisffed,no probIem had been solved. Similar sortsof restrictionsbound the admissiblesolutionsto theoreticalproblems.Throughout the eighteenthcentury those scientistswho tried to derive the observedmotion of the moon from Newton's laws of motion and gravitation consistently failed to do so. As a result, someof them suggestedreplacing the inversesquarelaw with a law that deviated from it at small distances.To do that, however,would have beento changethe paradigm, to define a new puzzle, and not to solvethe old one. In the event, scientistspreservedthe rules until, in 1750,one of them discoveredhow they could successfullybe applied.s 2 For a brief account of the evolution of these experiments,see page 4 of lecturein Les prir Nobel en 1937(Stockholm,1938). C. J. Davisson's 3 W. Whewell, Hi*ory of the luluctioe sciences(rev. ed.; London,1847),II, l0I-5, 220-i2z

39

fhe Sfructureof ScienfificRevolufions Only a changein the rulesof the gamecould have provided an alternative. The study of normal-scientiffctraditionsdisclosesmany additional rules, and these provide much information about the commitmentsthat scientistsderive from their paradigms.What can tve say are the main categoriesinto which theserules fall?. The most obviousand probably the most binding is exempliffed by the sorts of generalizationswe have iust noted. These are explicit statementsof scientiftc law and about scientiffc concepts and theories. While they continue to be honored, such statementshelp to set puzzlesand to limit acceptablesolutions. Newton's Laws, for example,performed thosefunctions during the eighteenthand nineteenth centuries.As long as they did so, quantity-of-matter was a fundamental ontological category for physical scientists,and the forces that act between bits of matter were a dominant topic for research.6In chemistry the laws of fixed and deffnite proportionshad, for a long time, an exactly similar force-setting the problem of atomic weights, bounding the admissible results of chemical analyses, and informing chemistswhat atoms and molecules,compoundsand mixtures were.8Maxwellt equations and the laws of statistical therrrodynamics have the same hold and function today. Rules like these are, however, neither the only nor even the most interesting variety displayedby historical study. At a level Iower or more concrete than that of laws and theories, there is, for example,a multitude of commitmentsto preferred types of instrumentationand to the ways in which acceptedinstruments may legitimately be employed. Changing attitudes toward the role of ffre in chemical analysesplayed a vital part in the dea I owe this qr.restio_n to W. O. Hagstrom, whose work in the sociology of sciencesometimesoverlapsmy own. 6 For theseaspectsof Newtonianism,seeI. B. Cohen, Frca/r;Ilnatd Neuston: An Inqulru lnto Speailatioe Neutonian Erperhpntal Scbrce atd Franklin's Work in nTectr*:U1i as an Erample Thercof (Philadelphia,1956), chap.vii, esp. pp.25L57, 27.-E77. 0 This exampleis discussedat length near the end of SectionX.

40

Normol Scienceos Puzzle-solving velopmentof chemistry in the seventeenthcenhrry.?Helmholtz, in the nineteenth,encounteredstrong resistancefrom physiologists to the notion that physical experimentation could illuminate their field.8 And in this century the curious history of chemical chromatography again illustrates the endurance of instrumental commitments that, as much as laws and theory, provide scientistswith rules of the game.eWhen we analyze the discovery of X-rays, we shall find reasonsfor commitments of this sort. Less local and temporary, though still not unchanging characteristics of science,are the higher level, quasi-metaphysical commitments that historical study so regularly displays. After about 1630,for example,and particularly after the appearance of Descartes'simmensely influential scientiffc writings, most physical scientistsassumedthat the universe was composedof microscopiccorpusclesand that all natural phenomenacould be explainedin terms of coqpuscularshape,size, motion, and interaction. That nest of commitrnentsproved to be both metaphysical and methodological.As metaphysical,it told scientists what sortsof entities the universedid and did not contain: there was only shaped matter in motion. As methodological, it told them what ultimate laws and fundamental explanationsmust be like: laws must specify co{puscularmotion and interaction, and explanationmust reduceany given natural phenomenonto colpuscularaction under theselaws. More important still, the corpuscular conception of the universe told scientistswhat many of their research problems should be. For example, a chemist who, Iike Boyle, embraced the new philosophy gave particular attention to reactionsthat could be viewed as transmutations. More clearly than any others these displayed the processof corpuscularrearrangementthat must underlie all z H. Metzger, Les doctrines chimiques en France du ddbut du xvlrc siccle d hfin du XVlIle siicle (Paris, 1928),.pp. 359$l; Marie Boas,Robert Boule atd. Seoenteenth-Century Chemistry ( Cam5ridge, lg58 ), pp. I l2-l5. t.L99^{gnigsberger, Hermann oon Helmholtz, trans. Francis A. Welby (Ox_ ford, 1906), pp. 6F66. 9_JamesE, Meinhard, "Chromatography: A Perspective," Science,CX ( lg4g), 387-92.

4l

fhe Struclureof ScienfificRevolufions chemical change.l' similar efiects of corpuscularismcan be observedin the study of mechanics,optics] and heat. Finally, at a still higher level, thereis anotherset of commitmentswithout which no man is a scientist.The scientistmust, for example,be concernedto understandthe world and to extend the precisionand scopewith which it has been ordered. That commitmentmust, in turn, lead him to scrutinize,either for himselfor through colleagues, someaspectof nature in great empirical detail. And, if that scrutiny displayspocketsof apparent disorder,then thesemust challengehim to a new reftniment of his observationaltechniquesor to a further articulation of his theories.Undoubtedlythere are still other ruleslike these, oneswhich have held for scientistsat all times. The existenceof this strong network of commitments-conceptual, theoretical, instrumental, and methodological-is a principal sourceof the metaphorthat relatesnormal scienceto puzzle-solving.Becauseit provides rules that tell the practitionerof a mature specialtywhat both the world and his science are like, he can concentratewith assuranceupon the esoteric problems that these rules and existing knowledge define for him. What then personallychallengeshim is how to bring the residualpuzzleto a solution.In theseand other respectsa discussionof puzzlesand of rules illuminatesthe nature of normal scientificpractice. Yet, in another waf t that illumination may be significantly misleading.Though there obviously are rules to which all the practitionersof a scientificspecialtyadhereat a given time, thoserulesmay not by themselvesspecifyall that the practiceof thosespecialistshas in common.Normal science is a highly determined activity, but it need not be entirely determinedby rules. That is why, at the start of this essay,I introducedsharedparadigmsrather than sharedrules, assumptions,and points of view as the sourceof coherencefor normal researchtraditions.Rules,I suggest,derive from paradigms,but paradigmscan guide researcheven in the absenceof rules. 10 For corpuscularism in general, see Marie Boas, "The Establishment of the Mechanical Philosophy," Osiris, X ( 1952), 412-541. For its effects on Boyle's chemistry, see T. S. Kuhn, "Robert Boyle and Structrrral Chemistry in the Seventeenth Century," I.si.s,XLIII (1952), 12-36.

12

V. The Priority of Porodigms To discoverthe relation between rules, paradigms,and normal science,considerffrst how the historian isolatesthe particular loci of commitment that have iust been described as acceptedrules. Close historical investigation of a given specialty at a given time disclosesa set of recurrent and quasistandard illustrations of various theories in their conceptual, observational,and instrumental applications. These are the community'sparadigms,revealedin its textbooks,lectures,and laboratory exercises.By studying them and by practicing with them, the members of the correspondingcommtrnity learn their trade.The historian,of course,will discoverin addition a penumbral area occupiedby achievementswhosestatusis still in doubt, but the core of solved problemsand techniqueswill usuallybe clear.Despiteoccasionalambiguities,the paradigms of a mature scientificcommunity can be determinedwith relative ease. The determinationof sharedparadigmsis not, however,the determinationof sharedrules.That demandsa secondstep and one of a somewhatdifferent kind. When undertaking it, the historian must comparethe community'sparadigmswith each other and with its current researchreports.In doing so, his object is to discoverwhat isolableelements,explicit or implicit, the members of that community may have abstracted from their more global paradigmsand deployedas rules in their research.Anyone who has attempted to describeor analyzethe evolutionof a particular scientifictradition will necessarilyhave sought acceptedprinciples and rules of this sort. Almost certainly, as the precedingsectionindicates,he will have met with at leastpartial success. But, if his experience hasbeenat all like my own, he will have found the searchfor rulesboth more difficult and lesssatisfyingthan the searchfor paradigms.Someof the generalizationshe employs to describe the communityt sharedbeliefs will presentno problems.Others,however,in-

43

fhe Sfructureof ScientificRevolufions cluding someof those used as illustrationsabove,will seema shadetoo stlong.Phrasedin iust that way, or in any other way he can imagine,they would almostcertainlyhave beenrejected by somemembersof the group he studies.Nevertheless,if the coherenceof the researchtradition is to be understoodin terms of rules, some specificationof common ground in the correspondingarea is needed.As a result, the searchfor a body of rules competentto constitutea given normal researchtradition becomesa sourceof continual and deepfrustration. Recognizingthat frustration, however, makes it possibleto diagnoseits source.Scientistscan agree that a Newton, Lavoisier, Maxwell, or Einstein has produced an aPParentlypermanent solution to a group of outstandingproblems and still disagree,sometimeswithout being aware of it, about the parthat make those solutionsperticular abstractcharacteristics manent. They can, that is, agree in their identification of. a paradigmwithout agreeingon, or even attemptingto produce, i fril interpretationor rationalizationof it. Lack of a standard interpretationor of an agreedreduction to rules will not prevent a paradigmfrom guiding research.Normal sciencecan be determined in part by the direct inspection of paradigms,,a processthat is often aided by but does not depend upon theIndeed,the existenceof formulationof rules and assumptions. a paradigmneednot evenimply that any full set of rules exists.r Inevitably, the first efiect of thosestatementsis to raiseproblems. In ttre absenceof a competentbody of rules, what restricts the scientistto a particular normal-scientifictradition? 'direct inspectionof paradigms'mean? What can the phrase Partial ans*ets to questionslike thesewere developedby the the late Ludwig Wittgenstein,though in a very different context. Becausethat context is both more elementaryand more familiar, it will help to considerhis form of the argumentfirst. what need we know, wittgenstein asked, in order that we r Michael Polanyi has brilliantly developed a very similar lh"T."t arguing that much of the Jcientist's s,tccesi dependi upon "tacit knowledge," i.e., upon knowledqe that is acquired through practice and that cannot be articulated See his Perionul Knoulidgi (Chicago, 1958), particularly chaps. v "*fti"ittf and vi.

44

FhePriorityof Porodigms 'chair,' 'leaf,' 'game' apply terms like or unequivocallyand or without provoking argument?2 That question is very old and has generallybeen answered by saying that we must know, consciouslyor intuitivel/, what a chair, or leaf, or gamers. We must, that is, graspsomeset of attributes that all gamesand that only gameshave in common. Wittgenstein,however, concludedthat, given the way we use languageand the sort of world to which we apply it, there need be no suchset of characteristics. Though a discussionof someof the attributes sharedby a number of gamesor chairsor leaves often helps us learn how to employ the correspondingterm, there is no set of characteristicsthat is simultaneouslyapplicable to all membersof the classand to them alone. Instead, confrontedwith a previouslyunobservedactivity, we apply the 'game'because term what we are seeingbearsa close"family resemblance"to a number of the activities that we have previously learnedto call by that name.For Wittgenstein,in short, games,and chairs,and leavesare natural families,eachconstituted by a network of overlappingand crisscrossresemblances. The existenceof such a network sufficiently accountsfor our successin identifying the correspondingobject or activity. Only if the familieswe namedoverlappedand mergedgraduallyinto one another-only, that is, if there were no natural familieswould our successin identifying and naming provide evidence for a set of commoncharacteristicscorrespondingto eachof the classnameswe employ. Somethingof the samesort may very well hold for the various researchproblems and techniquesthat arise within a single normal-scientiffctradition. What thesehave in common is not that they satisfy some explicit or even some fully discoverable set of rules and assumptionsthat gives the tradition its character and its hold upon the scientific mind. Instead, they may relate by resemblanceand by modeling to one or anotherpart of the scientific corpus which the community in question al2 Ludwig Wittgenstein, Philosophical lnoestigafions, trans. G. B. M. Anscombe (New York, 1953), pp. 3I-96. Wittgenstein, however, says almost nothinq about the sort of world neccssary to support tlrc naming procedurc hc outlineij Part of the point that follows cainot thcrcforc bc attribui6d to him.

45

fhe Struclureof ScientificRevolulions

do so,they needno full set of rules.The coherencedisplayedby the reseaichtradition in which they participate may not imply even the existenceof an underlying body of rules and assumptions that additional historical or philosophical investigation might uncover. That scientistsdo not usually ask or debate what makesa particular problem or solutionlegitimate tempts us to supposethat, at least intuitively, they know the answer. But it may only indicate that neither the question nor the answeris felt to be relevantto their research.Paradigmsmay be prior to, more binding, and more completethan any set of rules for researchthat could be unequivocallyabstractedfrom them. So far this point has been entirely theoretical: paradigms could determinenormal sciencewithout the interventionof discoverablerules.Let me now try to increaseboth its clarity and ulgency by indicating some of the reasonsfor believing that paradigmsactually do operatein this manner.The ftrst, which hasalreadybeen discussedquite fully, is the severedifficulty of discoveringthe rules that have guided particular normal-scientific traditions.That difficulty is very nearly the sameas the one the philosopherencounterswhen he tries to say what all games have in common.The second,to which the first is really a corollary, is rooted in the nature of scientificeducation.Scientists,it shouldalreadybe clear,neverlearn concepts,laws,and theories in the abstract and by themselves.Instead, these intellectual tools are from the start encounteredin a historically and pedagogically prior unit that displaysthem with and through their applications.A new theory is always announcedtogether with applications to some concrete rarlge of natural phenomena; without them it would not be even a candidatefor acceptance. After it has been accepted,those same applicationsor others accompanythe theory into the textbooksfrom which the future practitioner will learn his trade. They are not there merely as 46

The PriorityoI Porodigms embroidery or even as documentation.On the contrary, the processof learninga theory dependsupon the study of applications,including practiceproblem-solvingboth with a pencil and paper and with instrumentsin the laboratory.If, for example, the studentof Newtoniandynamicsever discoversthe meaning 'space,' and'time,' he doesso less of terms like'force,''mass,' from the incompletethough sometimeshelpful definitionsin his text than by observingand participating in the application of theseconceptsto problem-solution. That processof learningby finger exerciseor by doing continues throughout the processof professionalinitiation. As the student proceedsfrom his freshmancourseto and through his doctoral dissertation,the problems assignedto him become more complexand lesscompletelyprecedented.But they continue to be closelymodeledon previousachievementsas are the problemsthat normally occupyhim during his subsequentindependentscientificcareer.One is at liberty to supposethat somewhere along the way the scientist has intuitively abstracted rules of the gamefor himself, but there is little reasonto believe it. Though many scientiststalk easily and well about the particular individual hypothesesthat underlie a concretepiece of current research,they are little better than laymenat characterizingthe establishedbasesof their field, its legitimateproblems and methods.If they have learnedsuchabstractionsat all, they show it mainly through their ability to do successfulresearch. That ability can, however, be understoodwithout recourseto hypotheticalrulesof the game. These consequences of scientific educationhave a converse that provides a third reasonto supposethat paradigmsguide researchby direct modelingas well asthrough abstractedrules. Normal sciencecan proceed without rules only so long as the relevant scientiffccommunity acceptswithout question the particular problem-solutionsalreadyachieved.Rulesshould therefore becomeimportant and the characteristicunconcern about them should vanish whenever paradigms or models are felt to be insecure.That is,moreover,exactlywhat doesoccur.The preparadigm period, in particular, is regularly marked by frequent

17

fhe Slruclureof ScienfiffcRevolulions and deep debates over legitimate methods, problems, and standards of solution, though t'hese serve rather to define schoolsthan to produce agreement.We have already noted a few of these debatesin optics and electricity, and they played an even larger role in the developmentof seventeenth-century chemistry and of early nineteenth-century geology.sFurthermore,debateslike thesedo not vanishonceand for all with the appearanceof a paradigm.Though almostnon-existentduring periods of normal science,they recur regularly just before and during scientific revolutions,the periods when paradigmsare first under attack and then subject to change.The transition from Newtonian to quantum mechanicsevoked many debates about both the nature and the standardsof physics,some of which still continue.aThere are people alive today who can rememberthe similar argumentsengenderedby Maxwell's electromagnetictheory and by statisticalmechanics.bAnd earlier still, the assimilationof Galileo'sand Newton'smechanicsgave rise to a particularly famousseriesof debateswith Aristotelians, Cartesians,and Leibnizians about the standardslegitimate to science.oWhen scientistsdisagreeabout whether the fundamental problems of their field have been solved,the searchfor While rules galnsa function that it doesnot ordinarily Possess. 3 For chemistry,seeH. Metzger, Lesdoctrineschimiquesen Franced.uddbut du XYII' d It fi; du XVIIIe $CZte( paris, 1923), pp. 2L27 ,14G49; and Marie (Cambrilge, 1958), Boas, Robert Boyl,eand Seoenteenth-Cmtury.-ChqnAp -'The Uniformitarian-Catastrophist chap. ii. For seolbqy, seeWalter F. Cannon, De6ate," fs,is,"Lf( ibOO), 38-55; and C. C. Gillispie, Genesisatd. Geology( Cambridge, Mass.,l95l)' chaps.iv-v.

48

The Priorilyof Porodigms paradigmsremain secure,however, they can function without agreement over rationalization or witrhout any attempted rationalizationat all.

clear how they can exist. If normal scienceis so rigid and if scientific communitiesare so close-knit as the preceding discussionhas implied, how can a changeof paradigm ever affect only a small subgroup?What has been said so far may have seemedto imply that normal scienceis a singlemonolithic and uniffed enterprisethat must stand or fall with any one of its paradigmsas well as with all of them together.But scienceis obviously seldom or never like that. Often, viewing all fields together, it seemsinstead a rather ramshacklestructure with litlle coherenceamong its various parts. Nothing said to this point should,however,conflict with that very familiar observation. On the contrary, substituting paradigmsfor rules should make the diversity of scientiffcftelds and specialtieseasierto understand.Explicit rules,when they exist,are usuallycommon to a very broad scientiftcgroup,but paradigmsneednot be. The practitionersof widely separatedfields,sayastronomyand taxonomic botany, are educated by exposureto quite different achievementsdescribedin very different books.And even men who, being in the same or in closely related fields, bcgin by studying many of the samebooks and achievcmcntsmay acquire rather difierent paradigmsin the courseof professional specialization. Consider,for a single example,the qrrite large and diversc community constitutedby all physicalscientists.Each member of that group today is taught the laws of, say, quantum mechanics,and most of them employ theselaws at somepoint in

49

Ihe Sfructureof ScienfificRevolutions their researchor teaching.But they do not all leam the same applications of these laws, and they are not therefore all affected in the sameways by changesin quantum-mechanical practice.On the road to professionalspecialization,a few physical scientistsencounteronly the basic principles of quantum mechanics.Others study in detail the paradigmapplicationsof theseprinciples to chemistry,still others to the physicsof the solid state,and so on. What quantum mechanicsmeansto each of them dependsupon what courseshe has had, what texts he hasread, and which journalshe studies.It follows that, though a changein quantum-mechanicallaw will be revolutionaryfor all of thesegroups,a changethat reflectsonly on one or another of the paradigm applicationsof quantum mechanicsneed be revolutionaryonly for the membersof a particular professional subspecialty.For the rest of the professionand for those who practiceother physicalsciences,that changeneed not be revolutionary at all. In short, though quantum mechanics(or Newtonian dynamics,or electromagnetictheory) is a paradigm for many scientificgroups,it is not the sameparadigmfor them all. Therefore,it can simultaneouslydetermineseveraltraditions of normal sciencethat overlap without being coextensive.A revolution produced within one of these traditions will not necessarily extendto the othersas well. One brief illustration of specialization'seffect may give this whole series of points additional force. An investigator who hopedto learn somethingabout what scientiststook the atomic theory to be asked a distinguishedphysicist and an eminent chemist whether a single atom of helium was or was not a molecule.Both answeredwithout hesitation,but their answers were not the same.For the chemistthe atom of helium was a moleculebecauseit behavedlike one with respectto the kinetic theory of gases.For the physicist,on the other hand, the helium atom was not a molecule becauseit displayed no molecular spectrum.?Presumablyboth men were talking of the samepar7 The investigator was James K. Senior, to whom I am indebted for a verbal report. Some related issues are treated in his paper, "The Vernacular of the Laboratory," PlfiIosoplry of Science, XXV (1958), 163-68.

50

ThePriorilyof Porodigms ticle, but they were viewing it through their own researchtraining and practice.Their experiencein problem-solvingtold them what a molecule must be. Undoubtedly their experienceshad had much in common,but they did not, in this case,tell the two specialiststhe samething. As we proceedwe shall discoverhow consequentialparadigm differencesof this sort can occasionally be.

of Vl. Anomolyond the Emergence ScientificDiscoveries Normal science,the puzzle-solvingactivity we have itrst strccessexamined,is a highly cumulativeenterprise,emine_trtly of and precision scoPe ful in its aim, the steadyextensionof the great it fits with scientificknowledge.In all theserespects _pre-cisionthe most usial imageof scientlficwork. Yet one standard Normal science product of the scientificLnterprise_ismissin_g. sttccessfttl, when aud, theory fact or doesnot aim at noveltiesof rehowever, are, finds none.New and unsuspectedphenomella new radical and research, peatedly uncoveredby scientific iheotiei have again and again been inventedby scientists.History even suggeststhat the scientificenterprisehas developeda poiv:erfultechnique for producing surprisesof this ,rttq.tely 'th^is characteristicbf scienceis to be reconciledwith sort] If u,hat has already been said, then researchunder a Para{ig* must be a particularly efiective way of indtrcing paradigm change.Thai is what fundamentalnoveltiesof fact and theorydo. Pioduced inadvertentlyby a gameplayed under one set of rules, their assimilationrequiresthe elaborationof another set' After they have becomeparts of science,the enterprise,at.least of thoseipecialistsin whloseparticular field the noveltieslie, is neverquite the sameagain. We must now ask hJw changesof this sort can come about, jnconsideringfirst discoveries,oi noveltiesof fact, and then ventions,or noveltiesof theory. That distinctionbetweencliscovery and invention or betweenfact and theory will, however, immediatelyproveto be exceedi'glyartificial.Its artificialityis theses.Examirtan important-clueto severalof this essay'smain we shall this section' of rest ing sitected discoveriesin the epiextendcd btrt evertts qtiickly find that they _arenot isolated conlDiscovery strttctttre. recurretrt tlod.t with a regularly menceswith the-awar.tt.tt of anomaly,i.e., with the recognition that nature has somehowviolated the paradigm-induced 52

Anomolyond the Emergenceof ScienfificDiscoveries expectationsthat governnormal science.It then continueswith

establisha claim was the British scientist and divine, Joseph Priestley,who collectedthe gasreleasedby heatedred oiide^of

'F"", however, uno Bockluld, '4 Lost Letter from scheele to Lavoisier," _ I-ychnos. 1957-58, pp. Bg62, f i a difierent evaluatio" ol-5"t"ute's role.

53

Ihe Struclureof ScienliffcRevolulions

clusion that Priestleywas never able to accept. This pattern of discoveryraisesa question that can be asked about &uty novel phenomenonthat has ever entered the consciousnessof scientists.Was it Priestley or Lavoisier, if either, who ffrst discovered oxygen? In any case, when was oxygen discovered?In that form-the question could be asked even if only one claimant had existed. As a ruling about-priority and date, an answer doesnot at all concernus. Nevertheless,an attempt to produce one will illuminate the nature -of discovery, beciuse there is no answerof the kind that is sought.Discovery is not the sort of processabout which the question is aPPro-

claim to the discovery of oxygen is based uPon his priority in

thought he had obtained nitrous oxide, a specieshe already knew; in 1775he saw the gas as dephlogisticatedair, which is still not oxygen or even, for phlogistie chemists,a quite unexpected sort of gas. Lavoisier's claim may be stronger, bu_tit presentsthe sameproblems.If we refuse th9 plhn to Priestley, tnrecannot award i[ to Lavoisier for the work of.L775which led s I. B. Conant, The Ooetthroo of the PhlogkstonTheory: The Clwmical Reoolutiii of 1775-1789("Harvard Cise Histori-esin ExperimentalScience,"Case 2; Cambridge,Mass., 1950), p. 2s. This very usefuI pamphlet reprlntr many of the relevant documents.

51

Anomolyond the Emergenceof ScienfificDiscoveries him to identify the gasas the "air itself entire."Presumablywe wait for the work of 1776 and L777which led Lavoisier to see not merely the gas but what the gas was. Yet even this award could be questioned,for in L777 and to the end uf his life Lavoisierinsistedthat oxygenwas an atomic "principle of acidity" and that oxygengaswasformed only when that "principle" united with caloric,the matter of heat.aShall we thereforesay that oxygenhad not yet beendiscoveredin L777?Somemay be temptedto do so.But the principleof acidity was not banished from chemistryuntil after 1810,and caloriclingered until the 1860's.Oxygenhad becomea standardchemicalsubstancebefore either of thosedates. Clearly we need a new vocabularyand conceptsftor analyzing eventslike the discoveryof oxygen.Though undoubtedly correct, the sentence,"Oxygen was discovered,"misleadsby suggestingthat discoveringsomething is a single simple act assimilableto our usual ( and alsoqtrestionable ) conceptof seeing. That is why we so readily assumethat discovering,Iike seeingor touching, should be unequivocallyattributable to an individual and to a momentin time. But the latter attribution is always impossible,and the former often is as well. Ignoring Scheele,we can safelysay that oxygenhad not been discovered before 1774,and we would probably also say that it had been discoveredby1777or shortly thereafter.But within thoselimits or otherslike them, any attempt to date the discoverymust inevitably be arbitrary becausediscoveringa new sort of phenomenon is necessarilya complexevent, one which involvesrecogn_izingboth that somethingis and ushatit is. Note, for example, that if oxygenwere dephlogisticatedair for us, we shouldinsist without hesitationthat Priestleyhad discoveredit, though we would still not know quite when. But if both observationand co_n_ceptualization, fact and assimilationto theory, are inseparally linked in discovery,then discoveryis a processand m^ust take time. only when all the relevantconceptualcategoriesare preparedin advance,in which casethe phenomenonwould not 4 _ H. Metzger, La philosoTtlie cle la muti)re clrcz Laooisier (paris, lg35); and Lraumas, oqt.cit., chap. vii.

55

fhe Sfructureof ScienfiffcRevolutions be of a new sort, can discovering that and discovering what occur effortlessly,together, and in an instant. Grant now that discovery involves an extended,though not necessarilylong, processof conceptualassimilation.Can we also say that it involves a changein paradif? To that question,no general answer can yet be given, but in this case at least, the answermust be yes. What Lavoisier announcedin his PaPers from L777on was not so much the discovery of oxygen as t{re oxygen theory of combustion.That theory was the keystonefor a reformulation of chemistry so vast that it is usually called the chemical revolution. Indeed, if the discoveryof oxygenhad not been an intimate part of the emergenceof a new paradigm for chemistry, the question of priority from which we began would never have seemedso important. In this caseas in others, the value placed upon a new phenomenonand thus upon its discoverer varies with our estimate of the extent to which the phenomenonviolated paradigm-induced anticipations. Notice, however, since it will be important later, that the discovery of oxygen was not by itself the cause of the change in chemical theory. Long before he played any part in the discovery of the new gas, Lavoisier was convinced both that something was wrong with the phlogiston theory and that burning bodies absorbed some part of the atmosphere.That much he had recorded in a sealednote depositedwith the Secretaryof the French Academy in 1772.6What the work on oxygendid was to give much additional form and structure to Lavoisier's earlier sensethat somethingwas amiss.It told him a thing he was already prepared to discover-the nature of the substancethat combustionremovesfrom the atmosphere.That advanceawarenessof dificulties must be a significantpart of what enabled Lavoisier to seein experimentslike Priestley'sa gasthat Priestley had been unable to see there himself. Conversely,the fact that a maior paradigm revision was needed to seewhat Lavoisier saw must be the principal reasonwhy Priestleywas, to the end of his long life, unable to seeit. 6 The most authoritative account of the origin of Lavoisier's discontent is Henry Guerlac, Lanoisier-the Crucbl Iear: fhc Backgtound atd. Otigln ol n* First Erpefiments on Combustion in 1772 (lthaca, N.Y., f96l ).

56

Anomolyond the Emergenceof ScienfificDiscoveries Two other and far briefer exampleswill reinforce much that has iust been said and simultaneouslycary us from an elucidation of the nature of discoveriestoward an understandingof the circumstancesunder which they emergein science.In an effort to represent the main ways in which discoveriescan come about, theseexamplesare chosento be different both from each other and from the discovery of oxygen. The ffrst, X-rays, is a classiccaseof discoverythrough accident, a type that occurs more frequently than the impersonalstandardsof scientific reporting allow us easilyto realize.Its story openson the day that the physicist Roentgen intermpted a normal investigationof cathode rays becausehe had noticed that a barium platinocyanide screen at some distance from his shielded apparatus glowed when the dischargewas in process.Further investigations-they requiredsevenhectic weeksduring which Roentgen rarely left the laboratory-indicated that the causeof the glow camein straight lines from the cathoderay tube, that the radiation cast shadows,could not be defected by ^ magnet, and much elsebesides.Before announcinghis discovery,Roentgen had convincedhimself that his efiect was not due to cathode rays but to an agentwith at Ieastsomesimilarity to light.6 Even so brief an epitomerevealsstriking resemblances to the discoveryof oxygen: before experimentingwith red oxide of mercury, Lavoisier had performed experiments that did not producethe_resultsanticipatedunder the phlogistonparadigm; Roentgen'sdiscoverycommencedwith the recognitionthat his screenglowed when it should not. In both casesthe perception of anomaly-of a phenomenon,that is, for which his paradigm had not readied the investigator-played an essentialrole in preparingthe way for perceptionof novelty. But, again in both cases,the perception that somethi"g had gone wrong was only the prelude to discovery. Neither oxygen nor X-rays emerged without a further processof experimentationand assimilation. At what point in Roentgen'sinvestigation,for example,ought we say that X-rays had actually been discovered?Not, in any u-_!. W. Taylor,,Plrysics, the Pioneer Science (Boston, lg4l ), pp. Zg0-g4; and _ T. W. Chalnrers,Historic Rescurches(London, lg4g), pp. 218-i-g

fhe Slruclureof ScienfiffcRevolufions case,at the ffrst instant, when all that had been noted was a glowing screen.At least one other investigator had seen that glow and, to his subsequentchagin, discoverednothing at all.? Nor, it is almost as clear, can the moment of discovery be pushedforward to a point during the last week of investigation, by whictr time Roentgen was exploring the properties of the new radiation he had already discovered.We can only say that X-raysemergedin Wiirzburg betweenNovember8 and December 28, 1895. In a third area,however,the existenceof signiffcantparallels between the discoveriesof oxygen and of X-rays is far less apparent. Unlike the discoveryof oxygen,that of X-rays was not, at least for a decadeafter the event, implicated in any obviousupheavalin scientiffctheory. In what sense,then, can the assimilationof that discoverybe said to have necessitatedparadig* change?The case for denying such a change is very strong. To be sure, the paradigmssubscribedto by Roentgen and his contemporariescould not have been trsed to predict

Lavoisier'sinterpretationof Priestley'sgas.On the contrary, in f895 acceptedscientifictheory and practiceadmitted a number of forms of radiation-visible, infrared, and ultraviolet. Why could not X-rayshave been acceptedas iust one more form of a well*nown classof natural phenomena?Why were they not, for example,received in the same\Mayas the discoveryof an additionil chemicalelement?New elementsto fill empty places in the periodic table were still being soughtand found in Roentgen's day. Their pursuit was a standard project for normal Jcience,and successwas an occasiononly for congratulations, not for surprise. ? E. T. Whittaker, A History of the Theories of Aether and Electricity, | (2d' ed.; London, l95t),358, n. l-. sir George Thomson has informed me of a second ,r""r miis. Alerted by unaccountably fogged photographic plates, Sir William Crookes was also on the track of the discovery.

58

Anomolyond the Emergence of ScienlificDiscoveries X-rays, however, were greeted not only with surprise but with shock.Lord Kelvin at first pronouncedthem an elaborate hoax.8Others,though they could not doubt the evidence,were clearly staggeredby it. Though X-rays were not prohibited by establishedtheory, they violated deeply entrenchedexpectations.Thoseexpectations,I suggest,were implicit in the design and interpretationof establishedlaboratoryprocedures.By the 1890's cathode ray equipment was widely deployed in numerous European laboratories.If Roentgen'sapparatus had producedX-rays,then a number of other experimentalists must for sometime have beenproducing thoserays without knowing it. Perhapsthoserays,which might well have other unacknowledged sourcestoo, were implicated in behavior previously explained without referenceto them. At the very least, several sortsof long familiar apparatuswould in the future have to be shielded with lead. Previously completed wort on normal projectswould now have to be done againbecauseearlier scientists had failed to recognize and control a relevant variable. X-rays,to be sure,openedup a new ffeld and thus added to the potential domain of normal science.But they also, and this is now the more important point, changedfields that had already existed. In the processthey denied previously paradigmatic typesof instrumentationtheir right to that title. fn short, consciouslyor not, the decisionto employ a particular pieceof apparatusand to useit in a particular way carriesan assumptionthat only certain sorts of circumstanceswill arise. There are instrumentalas well as theoreticalexpectations,and they have often played a decisiverole in scientificdevelopment. One such expectation is, for example, part of the story of oxygen'sbelateddiscovery.Using a standardtest for "the goodnessof air," both Priestleyand Lavoisiermixed two volumesof their gaswith onevolumeof nitric oxide,shookthe mixture over water, and measuredthe volume of the gaseousresidue.The previous experiencefrom which this standard procedure had evolved assuredthem that with atmosphericair the residue of sir wi,iam rhomson BaronKetoinof r,,*tjtili,::I;,iHtfii: [#r.'tr"

59

fhe Sfruclureof ScienfificRevolufions would be onevolumeand that for any other gas (or for polluted air) it would be greater.In the oxygenexperimentsboth found a residue closeto one volume and identified the gas accordittgly. Only much later and in part through an accidentdid Priestley renouncethe standardprocedureand try mixing nitric oxide with his gas in other proportions. He then found that with quadruple the'volume of nitric oxide there was almostno residue at all. His commitmentto the original test procedure-a procedure sanctioned by much previous experience-had been of gasesthat simultaneouslya commitmentto the non-existence could behaveas oxygendid.e Illustrations of this sort could be multiplied by reference,for example,to the belated identiffcation of uranium fission.One reasonwhy that nuclear reaction proved especially difficult to recognizewas that men who knew what to expect when bombarding uranium chosechemical testsaimed mainly at elements from the upper end of the periodic table.loOught we conclude from the frequency with which such instrumental commitments prove misleading that scienceshould abandon standard tests and standardinstruments?That would result in an inconceivable method of research.Paradigmproceduresand applications are as necessaryto scienceas paradigm laws and theories,and they have the sameeffects.Inevitably they restrict the phenomenological field accessiblefor scientific investigation at any o Conant,op. cit,, pp. l&20.

with close aftliations to physics,we cannot bring ourselvesto this le_apryhich would contradict all pre'iious experienceof nucliar physics.It may be that a seriesof strangeaccidentsrenderi our resultsdeceptiie" ( Otto Hahh and Fritz Strassman,"ULer den Nachweisund das Verhalten der bei der Bestrahlungdes Urans mittels Neutronen entstehendedErdalkalimetalle," Db Naturaissensc@ten, XXVU [f939], l5).

@

Anomolyond the Emergence of ScienfificDiscoyeries given time. Recognizingthat much, we may simultaneouslysee an essentialsensein which a discoverylike X-rays necessitates paradigmchange-and thereforechangein both proceduresand expectations-fora specialsegmentof the scientificcommunity. As a result,we may alsounderstandhow the discoveryof X-rays could seemto open a strangenew world to many scientistsand could thus participate so effectively in the crisis that led to twentieth-centuryphysics. Our ffnal exampleof scientiffcdiscovery,that of the Leyden iar, belongsto a classthat may be describedas theory-induced. Initially, the term may seemparadoxical.Much that has been said so far suggeststhat discoveriespredicted by theory in advance are parts of normal scienceand result in no new sort of. fact. I have, for example,previouslyreferred to the discoveries of new chemical elementsduring the secondhalf of the nineteenth century as proceedingfrom normal sciencein that way. But not all theories are paradigm theories.Both during prep-aradigmperiodsand during the crisesthat lead to large-scale changesof paradigm, scientistsusually develop many speculative and unarticulatedtheoriesthat can themselvespoint the ryay to discovery.Often, however, that discoveryis not quite the one anticipatedby the speculativeand tentativehypothesis. only as experimentand tentative theory are together articuIated to a match doesthe discovery emergeand the theory become a paradigm The discoveryof the Leyden jar displaysall thesefeaturesas well as the others we have observedbefore. when it began, there was no single pa_radigm for electricalresearch.Inqtead,o number of theories,all derived from relatively accessibii:ph"nomena,were in competition.None of them succeededin ordering the whole variety of electricalphenornenavery well. That failure is the sourceof severalof the anomaliesihat provide backgroundfor the discovery of the Leyden jar. oni of the comp_eting schoolsof electricianstook electricityto be a fluid, a-ndjhalconception led a number of men to aitempt bottling the fluid by holding a water-filledglassvial in their-handsa,rJ touching the water to a conductor suspendedfrom an activc 6l

fhe Slruclureol ScienfificRevolulions electrostaticgenerator.On removing the iar from the machine and touching the water (or a conductor connectedto it) with his free hand, each of theseinvestigatorsexperienceda severe shock.Those first experimentsdid not, however,provide electricianswith the Leyden iar. That deviceemergedmore slowly, and it is again impossibleto say iust when its discoverywas completed.The initial attemptsto store electricalfluid worked only becauseinvestigatorsheld the vial in their hands while standing upon the ground. Electricianshad still to learn that the iar required an outer as well as an inner conducting coating and that the fuid is not really storedin the jar at all. Somewhere in the courseof the investigationsthat showedthem this, and which introduced them to severalother anomalouseffects,the device that we call the Leyden iar emerged.Furthermore,the experimentsthat led to its emergence,many of them performed by Franklin, were alsothe onesthat necessitatedthe drasticrevision of the fuid theory and thus provided the ffrst full paradig- for electricity.ll To a greateror lesserextent (correspondingto the continuum from the shockingto the anticipatedresult), th" characteristics common to the three examplesabove are characteristic of all discoveriesfrom which new sortsof phenomenaemerge.Those characteristicsinclude: the previousawarenessof anomaly,the gradual and simultaneousemergenceof both observationaland conceptualrecognition,and the consequentchangeof paradigm categoriesand proceduresoften accompaniedby resistance. There is evenevidencethat thesesamecharacteristiesare built into the nature of the perceptualprocessitself. In a psychological experimentthat deservesto be far better known outsidethe trade,Bruner and Postmanaskedexperimentalsubjectsto identify on short and controlled exposurea seriesof playing cards. Many of the cards were normal, but somewere made anoma11 For various stirges in thc Leyde'n jirr's evolution, see I. B. Cohen, Franklin_ atd Newton: An lnf,uiry into Specufuttiue Neutonian. Erperimental Scbnce and Franklin's Work in Eleitrlcltg as an Example Thereol (Philadelphia, lg56), pp. 385-86,400-406, 452-67,506-7. The last stage is described by Whittaker, oP. cit., pp.50-52.

62

Anomolyqnd the Emergenceof ScrentificDiscoyeries lous,e.9.,a red six of spadesand a black four of hearts.Each experimentalrun was constitutedby the displayof a singlecard to a single subject in a seriesof gradually increasedi*por,rr"r. A-fter each exposurethe subject was askedwhat he had seen, and the run wasterminatedby two successive correctidentifications.12 Even on the shortestexposuresmany subiectsidentifiedmost of the cards,and after a smallincreaseall the subiectsidentified them all. For the normal cardstheseidentiffcationswere usually correct,but the anomalouscardswere almostalwaysidentified, without apparent hesitation or puzzlement, as normal. The black four of heartsmight, for example,be identiffedas the four of either spadesor hearts.Without any awarenessof trouble, it was immediatelyfftted to one of the conceptualcategoriespre. pared by prior experience.One would not evenlike to say that the subjectshad seensomethingdifferent from what they identiffed. With a further increaseof exposureto the anomalous cards,subjectsdid begin to hesitateand to display awarenessof anomaly.Exposed,for example,to the red six of spades,some would say: That'sthe six of spades,but there'ssomethingwrong with it-the black hasa red border.Further increaseof exposure resultedin still more hesitationand confusionuntil finally,and sometimesquite suddenly, most subjectswould produce the correct identificationwithout hesitation.Moreover,after doing this with two or three of the anomalouscards,they would have Iittle ftrrther difficulty with the others.A few subjects,however, were never able to make the requisiteadjtrstmentof their categories. Even at forty times the averageexposurercqrrired to recognizenormal cards for what they were, lllore thirn l0 pcr. cent of the anomalouscardswere not correctlyidentificd.Ancl the subiectswho then failed often experiencedactrtepersonal distress.One of them exclaimed:"I cAn't make the strit out, whateverit is. It didn't evenlook like a card that tinre.I don't know what colorit is now or whetherit's tr sptrdeor a heart.I'nr t'I, S..-B_runer_andLco Postttutn, "On tlrc'Pcrception of Incrrrrgrrrity: A _ Paradignr," Iournal of Pusonality, XVIII (1949), gO0-:2g.

63

fhe Sfruclureof ScienliffcRevolulions not even sure now what a spadeIookslike. My Godl"rs In the next section we shall occasionallysee scientistsbehaving this way too. Either as a metaphoror becauseit reflectsthe nature of the mind, that psychologicalexperiment provides a wonderfully simpleand cogentschemafor the processof scientificdiscovery. In science,as in the playing card experiment,novelty emerges only with difficulty, manifestedby resistance,against a background provided by expectation.Initially, only the anticipated and usual are experiencedeven under circumstanceswhere anomaly is later to be observed.Further acquaintance,however,doesresult in awarenessof somethingwrong or doesrelate the efiectto somethingthat has gonewrong before.That awarenessof anomalyopensa period in which conceptualcategories are adiusteduntil the initially anomaloushasbecomethe anticipated. At this point the discoveryhas been completed.I have already urged that that processor one very much like it is involved in the emergenceof all fundamentalscientificnovelties. Let me now point out that, recognizingthe process,we can at Iast begin to seewhy normal science,a pursuit not directed to noveltiesand tending at first to suppressthem, shouldnevertheIessbe so effectivein causingthem to arise. In the developmentof any science,the first received paradigm is usuallyfelt to accountquite successfullyfor most of the observationsand experimentseasilyaccessibleto that science's practitioners.Further development,therefore,ordinarily calls for the constructionof elaborateequipment,the development of an esotericvocabulary and skills, and a refinementof concepts that increasinglylessenstheir resemblanceto their ttsttal leads, orl common-sense prototypes.That professiortalization the one hand, to an immenserestrictionof the scientist'svision and to a considerableresistanceto paradigm change.The sciencehas becomeincreasinglyrigid. On the other hand, within thoseareasto which the paradigmdircctsthe attentionof the ls lbiil., p. 2f8. My colleague Postman tells me that, though knowing all about the aiparatus and displai in irdvance, he neverthelcssfound looking at the incongruoui cards acutely uirc6mfortable.

64

Anomoly ond the Emergenceof ScientificDiscoveries group, normal scienceleads to a detail of information and to a precision of the observation-theory match that could be achieved in no other way. Furthermore, that detail and precision-of-matchhave a value that transcendstheir not alwaysvery high intrinsic interest. Without the special apparatus that is constructed mainly for anticipated functions, the results that lead ultimately to novelty could not occur. And even when the apparatus exists, novelty ordinarily emergesonly for the man who, knowing uith precision what he should expect,is able to recognize that something has gone wrong. Anomaly appears only against the background provided by the paradigm. The more preciseand far-reaching that paradig- is, the more sensitive an indicator it provides of anomaly and hence of an occasion for paradigm change. In the normal mode of discovery, even resistanceto changehas a use that will be explored more fully in the next section.By ensuringthat the paradigm will not be too easily surrendered,resistanceguaranteesthat scientists will not be lightly distracted and that the anomaliesthat lead to paradigm change will penetrate existing knowledge to the core. The very fact that a signiftcant scientific novelty so often emerges simultaneously from several laboratories is an index both to the strongly traditional nature of normal scienceand to the completenesswith which that traditional pursuit prepares the way for its own change.

Vll. Crisis ond the Emergence of Scientific Theories All the discoveriesconsideredin SectionVI were causesof or contributorsto paradigm change.Furthermore,the changesin which thesediscoverieswere implicated were all destructiveas well as constructive.After the discoveryhad been assimilated, scientistswere able to account for a wider range of natural phenomenaor to account with greater precision for some of those previously known. But that gain was achievedonly by discardingsomepreviouslystandardbeliefs or proceduresand, simultaneously,by replacingthose componentsof the previous paradigm with others. Shifts of this sott are, I have argued, associatedwith all discoveriesachievedthrough normal science, exceptingonly the unsurprisingonesthat had been anticipated in all but their details. Discoveriesare not, however, the only paradigm changes.In sourcesof thesedestructive-constructive the similar, but usually to consider this sectionwe shall begin of new theories. invention result from the shifts that far larger, fact sciences and theory, that in the argued already Having and not permanently invention, categorically and are discovery distinct, we can anticipateoverlapbetweenthis sectionand the last. (The impossiblesuggestionthat Priestleyfirst discovered oxygenand Lavoisier then invented it has its attractions.O*ygenhas alreadybeenencounteredasdiscovery;we shall shortly meet it again as invention.) In taking up the emergenceof new theorieswe shall inevitably extend our understandingof discovery as well. Still, overlap is not identity. The sorts of discoveriesconsideredin the last sectionwere not, at least singly, responsiblefor such paradigm shifts as the Copernican,Newtonian, chemical,and Einsteinian revolutions.Nor were they responsiblefor the somewhatsmaller,becausemore exclusively professional,changesin paradigmproducedby the wave theory of light, the dynamicaltheory of heat, or Maxwell'selectromagnetic theory. How can theories like these arise from normal 6

of Scienfiffcfheories Crisisond fhe Emergence science,an activity even less directed to their pursuit than to that of discoveries? If awarenessof anomalyplays a role in the emergenceof n_ew sortsof phenomena,it should surpriseno one that a similar but more piofound awareness is pierequisite to_ all accep!1bfe of theory. On this poinl historical evidence is, I think, change^s entirely unequivocal. The state of Ptolemaic astronomy was a scandalbefoie Copernicus'announcement.lGalileo'scontributions to the study bf motion depended closely uPon difficulties discoveredin Aristotle'stheory by scholasticcritics.2Newton's new theory of light and color originated in the discovery that none of the existing pre-paradigmtheorieswould account for the length of the spectt,t*, and the waye theory that replaced Newtoi's was anno-uncedin the midst of growing concernabout anomaliesin the relation of difiraction and polarization effects to Newton's theory.s Thermodynamicswas born from the collision of two existingnineteenth-centuryphysical theories,and quantum mechanicJfrom a variety of difficulties surrounding black-body radiation, specific heats, and the photo_electric efrect.aFurthermore,in ill these casesexcept that of Newton the awarenessof anomalyhad lastedso long and penetratedso deep that one can appropriatelydescribethe fields affectedby it aJ in a stateof growing crisis.Becauseit demandslarge-scale paradigm destruction and major shifts in the problems ""i iechniques of normal science,the emergenceo{ new theoriesis precededby a period of Pronouncedprofessionaling"t "tuily (London, 1954), P. 16. 1A. R. Hall, The ScientificReoolution,$0f18N in the Middle.lg?r lMadison, 2 MarslrallClagctt,Thc Scienceof_L[echanics of medievalelementsin a Wis., l9S9), parii II-III. A-.Koy16-displays ""T!gt C;iil"r'r tlrought in his EhtdesGalitieniei ( Paris,1939), particularly Vol. I. 3 For Newton, seeT. S. Kuhn, "Newton's Optical Pa-pers,"in lsaac Neu;ton's paoerco6 asfters in Natural PliIosophy, ed. i. B. Cohen (Cambridge, Ivlass, seeE. T. Whittaker,A i9's3l- o1,.27-45.For the prelude -(2d 'Aaher to'thL wave theor/, ed.; Lond_on,1951), and. Electricity,| ihe fheortes of ilriii'it (rev. ed.; London, b;-105, J"J w. Whewell, Ilistory of tlrc lnclucti;e Sciences 1847),II,39M66. 4 For thermodynamics,see SilvanusP. T}-ompso\ Life of William Thonxon nnrii Xelotnof Largs ( London,l9l0), I, 26&81. I.l th" quantumtlreory,scc i.iiJ n"i"h* , Thc Qiuntutn Tlrcory,trans.II. S. Ilatfteld and II. L. Brose( London, 1922),chaps.i-ii.

67

fhe Sfructureof ScienlificRevolulions s-ecurity.As on-e _might-expect,that insecurity is generatedby the persistentfailure of thi puzzles of normal sciJnce to com'e out a_s $ey should. Failure of existing rules is the prelude to a searchfor new ones.

Ptolemy's system.Given a particular discrepanc/r astronomers w_ereinvariably able to eliminate it by making someparticular adiustment in Ptolemy's systemof compoundedcircles. But as time went on, a man looking at the net result of the normal researcheffort of many astronomerscould observethat astronomy's complexity was increasingfar more rapidly than its accuracy and that a discrepancycorrectedin one place was likely to showup in another.s Because the astronomical tradition was repeatedly interrupted from outside and because,in the absenceof printing, communicationbetween astronomerswas restricted,thesedifA History of Astronony from Tlulec to KepLer(2d ed.; --oI.__L..E, D-1eyer, New York, 1953), chaps.ri-xii.

68

of ScienfificTheories Crisisond fhe Emergence ficultieswere only slowly recognized.But awarenessdid come' By the thirteenth centuiy Alfonso X could proclaim that if God had coosultedhim when creating the universe,he would have receivedgood advice.In the sixteenthcentury, Copernicus'coworker, domenico da Novara, held that no system so cumbersome and inaccurateas the Ptolemaichad becomecould possibly be true of nature. And Copernicushimself wrote in the Preiace to the De Reoolutionibis that the astronomical tradition he inherited had finally created only a monster. By_the early sixteenth century an incre,asingnumber of Europe'sbest astronomerswere recognizingthat the astronomicalparadigm was failing in application to its own traditional problems.That recognitioo *a-t prerequisite to Copernicus' reiection of the PtolJmaic paradifm attd his searchfor a new one. His famous preface stil provides one of the classicdescriptionsof a crisis state.o Breakdown of the normal technical puzzle-solvingactivity is not, of course,the only ingredientof the astronomicalcrisisthat faced Copernicus.An extended treatment would also discuss that made the for calendarreform, a Pressur_e the socialltess*" -precession a fuller addition' Inparticularly urgent. puzzle of -account the rise Aristotle, of would considermedieval criticism historical signiffcant of RenaissanceNeoplatonism,and other elementsbesides.But technical breakdown would still remain the core of the crisis. fn a mature science-andastronomyhad becomethat in antiquity-external factorslike thosecited above are principally signiffcantin determititg the _timing_ofbreakdown, thJease with which it can be recognized,and the area in which, becauseit is given particular attention, the breakdown ffrst occurs.Though immenselyimportant, issuesof that sort are out of boundsfor this essay. If that much is clear in the caseof the Copernicanrevolution, let us turn from it to a secondand rather different example,the crisisthat precededthe emergenceof Lavoisier'soxygentheory of combuslion.In the 1770'smany factors combinedto generate 6T. S. Kuhn, The CopernicanReoolutfon(Cambridge,Mass.,1957), pp. 135-43.

fhe Structureof ScienfificRevolulions a crisis in chemistry,and historiansare not altogether agreed about either their nature or their relative importance.Bui two of the-mare generallyacceptedas of first-rati significance:the rise of pnzumatic chemistry and the question Jf weight relations. The history of the ffrst beginsin the seventeentlicentury np and its deploymentin chemi-

:'';'1",,:TffiJ,:?;Ji chemical reactions. Butwithil*t:-::ff:'Jr"-[%ffijJi

that they may not be exceptionsat all-chtmists contiirued to believe that air wry lhe olly rott of gas.until r7s6, when Joseph B-1":\ showedthat ffxed air (cor) was consistentlydistinguishablefrom normal air, two samplesof gaswere thbught to be distinct only in their impurities.t After Black'swork the investigationof gasesproceededrapid^and Iy,_most n_otablyi_nthe handJ of Cavendisli, priestley, sche_ele, w-hotogethe-rdeveloped a number of new tectrniques c-apableof distinguishingone sampleof gas from another.-All thesemen, from Black thlough scheele,bJlieved in the phlogiston theory and often employedit in their designand intiqprCtution of experiments.scheeleactually ffrst producedoxygin by an elaborate chain of experimentsdesigned to dephlogi-sticatl heat. Yet the net result of their experimentswas a varief, of gas samplesand gas properties so elaborate that the phiogist-on theory proved ingeasingly little able to cope with iaboiatory elperien-ce.thgugh none of thesechemistsiuggested that thL theory_shguld_be replaced, they were unable to apply it consistently.By the time Lavoisier beganhis experimenti on airs in the early 1770's,there were almost as many versionsof the phlogiston theory as there were pneumatic chemists.sThat stnrtH*toryof chen*mry lg51),pp. Qd ed.;London, nrJi,l*tttfl'f a Though their main conc€rn is with a slightly later period, much relevant material is scattered throughout J. R. Partingtori and Douglas McKie's "Historical studies on the Phlogiston Theory," eirwls of sclerc6,II (rggz),961404; III (1988), l-58,837-7t; and tV (tggg), gg7-7t.

70

of Scientificfheories Crisisond the Emergence proliferation of versionsof a the crisis.In his preface,Copernicur The increasingvaguenessanc giston theory for pneumatic chemis only sourceof the crisisthat confror .,rth concernedto explainthe gain in weight that most bodies experiencewhen bnrtted or roasted,and that againis a problem wiih a long prehistory. At least a few Islamic chemistshad known thal Jome meials gain weight when roasted. In the seventeenthcenfury sett"tal investigatorshad concluded from this same fact that a roastedmetal takes up some ingredient firomthe atmosphere.But in the seventeenthcentury that conto most chemists.If chemicalreacclusionseemedunnecessary color, and texture of the ingrevolume, the tions could alter weight as well?-Weight was not alter they dients, why should of quantity of matter. Bemeasure not alwayt t"k"t to be the an isolated phenomeremained roasting sides,*iigttt-gain on weight on roasting lose (e.g., wood) non. Mosinatrrral bodies they-should' to say later was as the phlogistontheory During the eight""nth century, however,theseinitially adequate reiponsesto the problem of weight-gainbecameincreasiirgly difficult to maintain. Partly becausethe balancewas incrEasinglyused as a standardchemicaltool and partly because the dev-eiopmentof pneumatic chemistry made it possille 1nd desirabletb retain tfie gaseousProductsof reactions,chemists discoveredmore and more casesin which weight-gain accom' the gradual assimilationof panied roasting. Simultaneou_sly_, to insist that gain in chemists i.lewton'sgravitationaltheory led weight must mean gain in quan' did not result in reiection of theory could be adiustedin maI negativeweight, or perhapsfir, ter-edthe roasted body as phlo explanationsbesides.But if the lead to rejection,it did lead to studiesin which this problem bulked large. One of them, "On

ffie Sfructureof ScienlificRevolufions

consider now, as a third and ftnal example, the late nineteenth century crisis in physics that ptepared the way for the emergenceof relativity theory. one root of that crisis can be

0 H. Guerlac, LaooisW-the Cructal Year (Ithaca, N.Y., 196l). Ttre entire book doctnents the evolution and ffrst recognitim of a crisis. For a clear statement of the situation with respectto Lavoisier, seep. 85. _.toM"TJaqmgl, Cl*qq_oJ .SWe: Tfu Hlstoty of Theottcs of Spce tn Physbs (Cambridge,Mass.,l9$t),-pp. ll+24.

72

Crisisond the Emergenceof ScienfificTheories them during the early decadesof the eighteenth century to_be resurrectedbnly in the last decadesof the nineteenth when they had a very different relation to the practice of physics. The technical problems to which a relativistic philosophy of spacewas ultimately to be related began to enter normal sciencewith the acceptanceof the wave theory of light after about 1815,though they evokedno crisis until the 1890's.If light_is wave motion propagatedin a mechanicalether governedby Newton's Lawi, then both celestialobservationand terrestrial experiment become potentially capable of detecting drift thiough the ether. Of the celestial observations,only those of aberration promised sufrcient accuracy to provide relevant information, and the detection of ether-drift by aberration measurementstherefore becamea recognizedproblem for normal research.Much specialequipment was built to resolveit. That equipment, however, detected no observable drift, and the problem was therefore transferredfrom the experimentalists and observersto the theoreticians.During the central decades of the century Fresnel, Stokes,and others devised numerous articulations of the ether theory designedto explain the failure to observe drift. Each of these articulations assumedthat a moving body dragssomefraction of the ether with it. And each was sufficiently successfulto explain the negative results not only of celestialobservationbut alsoof terrestrialexperimentation, including the famous experimentof Michelsonand Morluy.tt There was still no conflict excepting that between the various articulations.In the absenceof relevant experimental techniques,that conflict never becameacute. The situation changedagain only with the gradual acceptance of Maxwell's electromagnetictheory in the last two decades of the nineteenth century. Maxwell himself was a Newin general tonian who believedthat light and electromagnetism of mechandisplacements of the a to variable particles were due electricity and ical ether. His earliestversionsof a theory for rr Joseph Larmor, Aether and Matter . . . lnclud.ing a Discussion of the lnfluence ol the Earth's Motion on Optical Phenomena (Cambridge, 190O), pp. 6-20,32V22.

73

fhe Struclureof ScienliffcRevolufions m_agnetismmade direct use of hypothetical properties with which he endowedthis medium. Thise were dropped from his final version, but he still believed his electromagnetictheory compatiblewith somearticulationof the Newtonianmechanical

Maxwell'stheory, despiteits Newtonian origin, ultimately produced a crisis for the paradigm from which it had sprung.tg Furthermore,the locus at which that crisisbecamemost acute was provided by the problemswe have just been considering, thoseof motion with respectto the ether.

thereforewitnesseda long seriesof attempts,both experimental and theoretical,to detect motion with respectto the ether and to work ether drag into Maxwell'stheory.The former were uniformly unsuccessful, though someanalyststhought their results equivocal.The latter produced a number of. promising starts, particularly thoseof Lorentz and Fitzgerald,but they also disclosedstill other puzzlesand finally resultedin just that proliferation of competing theoriesthat we have previously found to be the concomitantof crisis.laIt is againstthat historicalsetting that Einstein'sspecialtheory of relativity emergedin 1905. Thesethree examplesare almostentirely typical. In eachcase a novel theory emergedonly after a pronouncedfailure in the 12 R. T. Glazebrook, Iamg1 Clerk Marwell and Modern Physics (London, 1896), chap. ix. For Maxwell's final attitude, see his own book, A Treatise on Electricity and, Magrctisrn (3d ed.; Oxford, 1892), p.470. rB For astronorrry's role in the development of mechanics, see Kuhn, op, cit., chap. vii. 1r Whittaker, op. cit.,I, 38G-410; and II (London, 1953), 27-40.

71

of Scienfificfheories Crisisond the Emergence normal problem-solvingactivity. Furthermore, except for the caseof Copernicusin which factorsexternalto scienceplayed a particularly large role, that breakdownand the proliferation of theoriesthat is its sign occurredno more than a decadeor two before the new theory'senunciation.The novel theory seemsa direct responseto crisis.Note also,though this may not be quite so typical, that the problemswith respectto which breakdown occurredwere all of a type that had long been recognized.Previouspracticeof normal sciencehad given every reasonto consider them solvedor all but solved,which helps to explainwhy the senseof failure, when it came, could be so acute. Failure with a new sort of problem is often disappointingbut never surprising.Neither problemsnor puzzlesyield often to the first attack.Finally, theseexamplesshareanothercharacteristicthat may help to make the casefor the role of crisisimpressive:the solutionto eachof them had been at leastpartially anticipated during a period when there was no crisis in the corresponding science;and in the absenceof crisis those anticipationshad been ignored. The only completeanticipation is also the most famous,that of Copernicusby Aristarchusin the third century n.c. It is often said that if Greek sciencehad been less deductive and less ridden by dogma,heliocentricastronomymight have begun its developmenteighteen centuriesearlier than it did.15But that is to ignore all historical context.When Aristarchus'suggestion wasmade,the vastly more reasonablegeocentricsystemhad no needsthat a heliocentricsystemmight even conceivablyhitve fulfflled. The whole developmentof Ptolemaicastronomy,both its triumphs and its breakdown,falls in the centuriesafter Aristarchus' proposal.Besides,there were no obvious reasonsfor taking Aristarchusseriously.Even Copernicus'more elaborate proposalwas neither simpler nor more accuratethan Ptolemy's system.Availableobservationaltests,aswe shall seemore clear15 For Aristarchus' wor\ see T. L, Heath, Aristarchus of Samos: The Ancient Copernicus (Oxford, l9l8), Part II. For an extreme statement of the traditional position about the neglect of Aristarchus' achievement, see Arthur Koestler, ?he Sleepualkers: A History of Man's Clwnging Yision ol the l|nioerse (London, 1959),p.50.

75

- ff:{'l;fr^nn ^9d,.'Kftr5*-r t\-^ fheTlructureof ScienfificRevolulions

J

lv below, provided no basisfor a choicebetweenthem. Under one of the factorsthat led astronomersto thosecircumstances, Copernicus(and one that could not have led them to Aristarchus) was the recognizedcrisis that had been responsiblefor innovationin the ffrst place.Ptolemaicastronomyhad failed to solve its problems;the time had come to give a competitor a chance.Our other two examplesprovide no similarly full anticipations.But surely one reasonwhy the theoriesof combustion by absorptionfrom the atmosphere-theoriesdevelopedin the seventeenthcentury by Rey, Hooke, and Mayow-failed to get a sufficienthearingwas that they madeno contactwith a recognized trouble spot in normal scientiftcpractice.loAnd the long neglect by eighteenth- and nineteenth-centuryscientists of Newton's relativistic critics must largely have been due to a similar failure in confrontation. Philosophersof sciencehave repeatedly demonstratedthat more than one theoretical constructioncan always be placed upon a given collection of data. History of scienceindicates that, particularly in the early developmentalstagesof a new paradigm,it is not evenvery difficult to invent such alternates. But that invention of alternatesis just what scientistsseldom undertake except during the pre-paradigmstage of their science'sdevelopmentand at very special occasionsduring its subsequentevolution.So long as the tools a paradigm supplies continue to prove capable of solving the problems it defines, sciencemovesfastestand penetratesmost deeply through conffdent employment of those tools. The reason is clear. As in manufactureso in science-retoolingis an extravaganceto be reservedfor the occasionthat demandsit. The signiffcanceof crisesis the indication they provide that an occasionfor retooling ha.sarrived. 1oPartington, op. cit., pp. 78-85.

76

Vlll. The Responseto Crisis Let rrs then assumethat crisesare a necessaryprecondition for the emergenceof novel theoriesand ask next how scientists respondto their existence.Part of the answer,as obviousas it is important, can be discoveredby noting ffrst what scientists never do when confronted by even severeand prolonged anomalies.Though they may begin to losefaith and then to consider alternatives,they do not renouncethe paradigm that has led them into crisis.They do not, that is, treat anomaliesascounterinstances,though in the vocabulary of philosophy of science that is rryhatthey are. In part this generalizationis simply a statementfrom historic fact, based upon exampleslike those given above and, more extensively,below. Thesehint what our later examinationof paradigmrejectionwill disclosemore fully: once it has achievedthe statusof paradigm,a scientiffctheory is declaredinvalid only if an alternatecandidateis availableto take its place. No processyet disclosedby the historical study of scientiffcdevelopmentat all resemblesthe methodological stereotypeof falsification by direct comparisonwith nature. That remark doesnot mean that scientistsdo not reject scientiftc theories,or that experienceand experimentare not essential to the processin which they do so. But it doesmean-what will ultimately be a central point-that the act of iudgment that leadsscientiststo reiect a previouslyacceptedtheory is always comDarisonof that theory theorv with the based upon uDon more than a comparison world. The decisionto reject one paradigm is always simultaneouslythe decisionto acceptanother,and the judgment leading to that decisioninvolvesthe comparisonof both paradigms with nature an^dwith each other. There is, in addition, a secondreasonfor doubting that scientists reject paradigms becauseconfronted with anomaliesor In developingit my argumentwill itself forecounterinstances. shadow another of this essay'smain theses.The reasonsfor doubt sketchedabove were purely factual; they were, that is,

77

fhe Sfructureof ScientificRevolufions themselvescounterinstancesto a prevalent epistemological theory. As such,if my presentpoint ii correct,th-eycan at-best help to create a crisis or, more accurately,to reinforce one that is already ve1y much in existence.By themselvesthey cannot and will not falsify that philosophical theory, for its defenders will do what we h-avealready seenscientisfsdoing when con-articulations fronted by anomaly. They will devise numerous and ad hoc modiffcationsof their theory in order to eliminate any apparent conflict. Many of the relevant modificationsand qualificationsare, in fact, already in the literature. If, therefore, these epistemologicalcounterinstancesare to constitute more than a minor irritant, that will be becausethey help to permit the emergenceof a new and different analysisof sciinceiithin which_theyare no longer a sourceof trouble. Furthermore,if a typical pattern, which we shall later observein scientiffc revolutions, is applicable here, these anomalieswill then no longer seemto,be simply facts.From within a new theory of scientific knowledg.,$ry may insteadseemvery much hkl tautologies, statementsof situations that could not conceivably have i'."r, otherwise. It has often been observed,for exampre,that Newton's second law of motion, though it took centririesof difficult factual and theoretical research-to achieve, behaves for those committed to Newton'stheory very much rike a purely logical statement that no amount of observationcould tif,rt".t hisection x we shall see that the chemicallaw of ftxed proportion, which before Dalton was an occasionalexp_erimenfal finding'of very dubious generality, became after oiltont work a' iigredient of a definition of chemical compound that no expeiimental could by itself have upset.sbmethingmuch lik'e that will lork also happen to the generalilation that scLntists fail to reject paradigms when faced with anomalies or counterinstances. They could not do so and still remain scientists. _ Though history is unlikely to record their names,somemen have undoubtedly been diiven to desert science because of I see particularlv the discussion in N. R. Hanson, pattcrns of ( Cambridge, lg58 i, pp. 9S-t05.

78

Discooerg

fhe Response lo Crisis their inability to tolerate crisis.Like artists, creative scientists must occasionallybe able to live in a world out of joint-elsewhere I have describedthat necessityas "the essentialtension" implicit in scientificresearch.2But that reiection of sciencein favor of anotheroccupationis, I think, the only sort of paradigm rejection to which eounterinstancesby themselvescan lead, Once a first paradigm through which to view nature has been found, there is no such thing as researchin the absenceof any paradigm.To reject one paradigm without simultaneouslysubstituting anotheris to reiect scienceitself. That act reflectsnot on the paradigmbut on the man. Inevitably he will be seenby his colleaguesas "the carpenterwho blameshis tools." The samepoint can be made at least equally effectively in reverse: there is no such thing as researchwithout counterinstances.For what is it that differentiatesnormal sciencefrom sciencein a crisis state?Not, surely,that the former confronts no counterinstances. On the contrary,what we previouslycalled the puzzlesthat constitutenormal scienceexistonly becauseno paradigmthat providesa basisfor scientificresearchever completely resolvesall its problems.The very few that have ever seemedto do so (e.g.,geometricoptics) haveshortlyceasedto yield researchproblems at all and have instead becometools for engineering.Excepting those that are exclusivelyinstrumental, everyproblem that normal scienceseesas a puzzle can be seen,from anotherviewpoint, as a counterinstanceand thus as a sourceof crisis.Coperrricussaw as counterinstances what most of Ptolemy'sother successors had seen as puzzlesin the match between observationand theory. Lavoisier saw as a counterinstance what Priestleyhad seenasa successfully solved puzzlein the articulationof the phlogistontheory.And Einstein saw as counterinstanceswhat Lorentz, Fitzgerald, and others had seenas puzzlesin the articulation of Newton's and Max2 T. S. Kuhn, "The Essential Tension: Tradition and fnnovation in Scientiftc Research," in The Third (1959) unhsersity of utah Research conference on thc ldentification of crcatioe scientfic Taleht, c"luit w. Taylor 1s"lt L"k" City,_ 1959), pp !62:77.-F9r the-comparable"d. phenomenon among artists, see Frank Barron, "The Psychology of Imagination," Scicntifw Amuican, CXCIX ( Scptenrber, 1958), 15l-66, esp. 160.

79

fhe Sfruclureof ScienlificRevolufions well's theories.Furthermore, even the existenceof crisis does not by itself transform apuzzle into a counterinstance. There is no suchsharpdividing line. Instead,by proliferatingversionsof the paradigm,crisis loosensthe rules of normal puzzle-solving in ways that ultimately permit a new paradigm to emerge. There are, I think, only two alternatives:either no scientiffc theory ever confronts a counterinstance,or all such theories confront counterinstances at all times. How can the situationhave seemedotherwise?That question necessarilyleads to the historical and critical elucidation of philosoph/, and those topics are here barred. But we can at leastnote two reasonswhy sciencehasseemedto provide so apt an illustration of the generalizationthat truth and falsity are uniquely and unequivocally determined by the confrontation of statementwith fact. Normal sciencedoesand must continually strive to bring theory and fact into closeragreement,and that activity can easilybe seenas testingor as a searchfor conffrmation or falsiffcation.Instead, its object is to solve a puzzle for whose very existencethe validity of the paradigm must be assumed.Failure to achievea solutiondiscreditsonly the scientist and not the theory.Here,evenmorethan above,the proverb applies: "It is a poor carpenterwho blameshis tools." In addition, the manner in which sciencepedagogyentanglesdiscussion of a theory with remarkson its exemplary applieationshas helped to reinforcea confirmation-theorydrawn predominantly from other sources.Given the slightestreasonfor doing so, the man who readsa sciencetext can easilytake the applicationsto be the evidencefor the theory, the reasonswhy it ought to be believed.But sciencestudentsaccepttheorieson the authority of teacherand text, not becauseof evidence.What alternatives have they, or what competence?The applicationsgiven in texts are not there as evidencebut becauselearning them is part of learning the paradigmat the baseof current practice.If applicationswere set forth as evidence,then the very failure of texts to suggestalternativeinterpretationsor to discussproblemsfor which scientistshave failed to produce paradigm solutions 80

lo Crisis fhe Response would convict their authors of extreme bias. There is not the slightest reasonfor such an indictment. How, then, to refurn to the initial question, do scientistsrespond to the awarenessof an anomaly in the fft between theory and nature? What has iust been said indicates that even a discrepancy unaccountably larger than that experiencedin other applications of the theory need not draw any very profound response.There are always somediscrepancies.Even the most stubborn ones usually respond at last to normal practice. V"ry often scientistsare willing to wait, particularly if thereare many problems available in other parts of the fteld. We have already noted, for example,that during the sixty years after Newton's original corhputation, the predicted motion of the moon's perigee remained only half of that observed.As Europe's be_st hathematical physicists continued to wrestle unsuccessfully with the well-known discrepancy, there were occasionalpro' posalsfor a modiffcation of Newton's inversesquarelaw. But no one took these proposals very seriously, and in practice this -ilot aoo*ily proved iustifted. Clairaut in patience with 1750was able to"-show that only the mathematicsof the applica' tion had been wrong and that Newtonian theory could stand as before.sEven in caseswhere no mere mistake seemsquite possible (perhaps becausethe mathematicsinvolved is simpler or of a familiar and elsewheresuccessfulsort), persistent and recognized anomaly does not always induce crisis. No one seriously questioned Newtonian theory because of the longrecognizeddiscrepanciesbetweenpredictionsfrom that theory and both the speedof sound and the motion of Mercury. Thc first discrepancywas ultimately and quite unexpectedlyresolved by experimentson heat undertaken for a very different purpose;the secondvanishedwith the general theory of relativity after a crisis that it had had no role in creating.aApparentsW. Whewell, History of the Inductioe Sclences (rev. ed.; London, 1847), II,22U2L. a For the speed of sound, see T. S. Kuhn, "The Caloric Theory of Adiabatie Compression,'fsds, XLIV (1958), lg&37. For the secular shift in Mercury's perilielion, see E. T. Whittakcr, A Flistonl of thc Thcories of Aahu and El,octri