THOMAS KUHN TheStructure of ScientificRevolutionsTHOMAS KUHN TheStructure ofScientificRevolutions THOMAS KUHN (1922-1996) was aprofessor ofphilosophy at MIT. In 1962 he revolutionized

THOMAS KUHNThe Structure of Scientific RevolutionsTHOMAS KUHN (1922-1996) was a professor ofphilosophy at MIT. In 1962 herevolutionized the philosophy of science with publication of his book, The Structure ofScientific Revolutions,ftom which this selection is taken. In that book he challenged theclassic view (represented by Hempel) by suggesting that we look at science as a socialenterprise, rather than as a purely logical enterprise.

THOMAS KUHN The Structure of Scientific Revolutions

INthis essay, 'normal science' meansresearch firmly based upon one ormore past scientific achievements,

achievements that some particular scien-tific community acknowledges for a time

. as supplying the foundation for its fur-ther practice. Today such achievementsare recounted, though seldom in theiroriginal form, by science textbooks, ele-mentary and advanced. These textbooksexpound the body of accepted theory, il-lustrate many or all of its successful ap-plications and compare these applica-tions with exemplary observations andexperiments. Before such books becamepopular early in the nineteenth century(and until even more recently in thenewly matured sciences), many of the fa-mous classics of science fulfilled a similarfunction. Aristotle's Physica, Ptolemy'sAlmagest, Newton's Principia andOpticks, Franklin's Electricity, Lavoisier'sChemistry, and Lyell's Geology-theseand many other works served for a timeimplicitly to define the legitimate prob-lems and methods of a research field forsucceeding generations of practitioners.They were able to do so because theyshared two essential characteristics.Their achievement was sufficiently un-precedented to attract an enduring groupof adherents away from competingmodes of scientific activity. Simultane-ously, it was sufficiently open-ended toleave all sorts of problems for the re-defined group of practitioners to resolve.

Achievements that share these twocharacteristics I shall henceforth refer toas 'paradigms,' a term that relates closelyto 'normal science.' By choosing it, Imean to suggest that some acceptedexamples of actual scientific practice-examples which include law, theory, ap-plication, and instrumentation together--provide models from which spring

99

particular coherent traditions of scientificresearch. These are the traditions whichthe historian describes under such rubricsas 'Ptolemaic astronomy' (or 'Coperni-can'), 'Aristorelian dynamics' (or 'New-tonian'), 'corpuscular optics' (or 'wave op-tics'), and so on. The study of paradigms,including many that are far more special-ized than those named illustrativelyabove, is what mainly prepares the stu-dent for membership in the particularscientific community with which he willlater practice. Because he there joins menwho learned the bases of their field fromthe same concrete models, his subsequentpractice will seldom evoke overt dis-agreement over fundamentals. Menwhose research is based on shared para-digms are committed to the same rulesand standards for scientific practice. Thatcommitment and the apparent consensusit produces are prerequisites for normalscience, i.e., for the genesis and continu-ation of a particular research tradition ....

Normal science ... is a highly cumu-lative enterprise, eminently successful inits aim, the steady extension of the scopeand precision of scientific knowledge. Inall these respects it fits with great preci-sion the most usual image of scientificwork. Yet one standard product of thescientific enterprise is missing. Normalscience does not aim at novelties of factor theory and, when successful, findsnone. New and unsuspected phenomenaare, however, repeatedly uncovered byscientific research and radical new theo-ries have again and again been inventedby scientists. History even suggests thatthe scientific enterprise has developed auniquely powerful technique for produc-ing surprises of this sort. If this charac-teristic of science is to be reconciled withwhat has already been said, then researchunder a paradigm must be a particularly

100 CHAPTER THREE What Does Science Tell Me About the World?

effective way of inducing paradigmchange. That is what fundamental nov-elties of fact and theory do. Produced in-advertently by a game played under oneset of rules, their assimilation requiresthe elaboration of another set. Mter theyhave become parts of science, the enter-prise, at least of those specialists inwhose particular field the novelties lie, isnever quite the same again.

We must now ask how changes of thissort can come about, considering first dis-coveries, or novelties of fact, and then in-ventions, or novelties of theory. That dis-tinction between discovery and inventionor between fact and theory will, however,immediately prove to be exceedingly arti-ficial. Its artificiality is an important clueto several of this essay's main theses. Ex-amining selected discoveries in the rest ofthis section, we shall quickly find thatthey are not isolated events but extendedepisodes with a regularly recurrent struc-ture. Discovery commences with theawareness of anomaly, i.e., with therecognition that nature has somehow vio-lated the paradigm-induced expectationsthat govern normal science. It then con-tinues with a more or less extended explo-ration of the area of anomaly. And itcloses only when the paradigm theory hasbeen adjusted so that the anomalous hasbecome the expected. Assimilating a newsort of fact demands a more than additiveadjustment of theory, and until that ad-justment is completed-until the scien-tist has learned to see nature in a differentway-the new fact is not quite a scientificfact at all.

To see how closely factual and theo-retical novelty are intertwined in scien-tific discovery, examine a particularly fa-mous example, the discovery of oxygen.At least three different men have a legit-imate claim to it, and several otherchemists must, in the early 1770s, have

had enriched air in a laboratory vesselwithout knowing it. The progress of nor-mal science, in this case of pneumaticchemistry, prepared the way to a break-through quite thoroughly. The earliest ofthe claimants to prepare a relatively puresample of the gas was the Swedishapothecary, C. W. Scheele. We may,however, ignore his work since it was notpublished until oxygen's discovery hadrepeatedly been announced elsewhereand thus had no effect upon the histori-cal pattern that most concerns us here.The second in time to establish a claimwas the British scientist and divine,Joseph Priestley, who collected the gasreleased by heated red oxide of mercuryas one item in prolonged normal investi-gation of the "airs" evolved by a largenumber of solid substances. In 1774 heidentified the gas thus produced as ni-trous oxide and in 1775, led by furthertests, as common air with less than itsusual quantity of phlogiston. The thirdclaimant, Lavoisier, started the work thatled him to oxygen after Priestley's exper-iments of 1774 and possibly as the resultof a hint from Priestley. Early in 1775Lavoisier reported that the gas obtainedby heating the red oxide of mercury was"air itself entire without alteration [ex-cept that] ... it comes out more pure,more respirable." By 1777, probably withthe assistance of a second hint fromPriestley, Lavoisier had concluded thatthe gas was a distinct species, one of thetwo main constituents of the atmos-phere, a conclusion that Priestley wasnever able to accept.

This pattern of discovery raises aquestion that can be asked about everynovel phenomenon that has ever enteredthe consciousness of scientists. Was itPriestley or Lavoisier, if either, who firstdiscovered oxygen? In any case, whenwas oxygen discovered? In that form the

THOMAS KUHN The Structure of Scientific Revolutions

question could be asked even if only oneclaimant had existed. As a ruling aboutpriority and date, an answer does not atall concern us. Nevertheless, an attemptto produce one will illuminate the na-ture of discovery, because there is no an-swer of the kind that is sought. Discov-ery is not the sort of process aboutwhich the question is appropriatelyasked. The fact that it is asked-the pri-ority for oxygen has repeatedly beencontested since the 1780s-is a symp-tom of something askew in the image ofscience that gives discovery so funda-mental a role. Look once more at ourexample. Priestley's claim to the discov-ery of oxygen is based upon his priorityin isolating a gas that was later re-cognized as a distinct species. ButPriestley's sample was not pure, and, ifholding impure oxygen in one's hand isto discover it, that had been done byeveryone who ever bottled atmosphericair. Besides, if Priestley was the discov-erer, when was the discovery made? In1774 he thought he had obtained ni-trous oxide, a species he already knew; in1775 he saw the gas as dephlogisticatedair, which is still not oxygen or even, forphlogistic chemists, a quite unexpectedsort of gas. Lavoisier's claim may bestronger, but it presents the same prob-lems. If we refuse the palm to Priestley,we cannot award it to Lavoisier for thework of 1775 which led him to identifythe gas as the "air itself entire." Presum-ably we wait for the work of 1776 and1777 which led Lavoisier to see notmerely the gas but what the gas was. Yeteven this award could be questioned, forin 1777 and to the end of his lifeLavoisier insisted that oxygen was anatomic "principle of acidity" and thatoxygen gas was formed only when that"principle" united with caloric, the mat-ter of heat. Shall we therefore say that

101

oxygen had not yet been discovered in1777? Some may be tempted to do so.But the principle of acidity was not ban-ished from chemistry until after 1810,and caloric lingered until the 1860s.Oxygen had become a standard chemi-cal substance before either of thosedates.

Clearly we need a new vocabularyand concepts for analyzing events likethe discovery of oxygen. Though un-doubtedly correct, the sentence, "Oxy-gen was discovered," misleads by sug-gesting that discovering something is asingle simple act assimilable to ourusual (and also questionable) concept ofseeing. That is why we so readily as-sume that discovering, like seeing ortouching, should be unequivocally at-tributable to an individual and to a mo-ment in time. But the latter attributionis always impossible, and the former of-ten is as well. Ignoring Scheele, we cansafely say that oxygen had not been dis-covered before 1774, and we wouldprobably also say that it had been dis-covered by 1777 or shortly thereafter.But within those limits or others likethem, any attempt to date the discoverymust inevitably be arbitrary becausediscovering a new sort of phenomenonis necessarily a complex event, onewhich involves recognizing both thatsomething is and what it is. Note, forexample, that if oxygen were dephlogis-ticated air for us, we should insist with-out hesitation that Priestley had discov-ered it, though we would still not knowquite when. But if both observation andconceptualization, fact and assimilationto theory, are inseparably linked in dis-covery, then discovery is a process andmust take time. Only when all the rele-vant conceptual categories are preparedin advance, in which case the phenome-non would not be of a new sort, can

102 CHAPTER THREE What Does Science Tell Me About the World?

discovering that and discovering what oc-cur effortlessly, together, and in an instant.

Grant now that discovery involves anextended, though not necessarily long,process of conceptual assimilation. Canwe also say that it involves a change inparadigm? To that question, no generalanswer can yet be given, but in this caseat least, the answer must be yes. WhatLavoisier announced in his papers from1777 on was not so much the discoveryof oxygen as the oxygen theory of com-bustion. That theory was the keystonefor a reformulation of chemistry so vastthat it is usually called the chemical rev-olution. Indeed, if the discovery of oxy-gen had not been an intimate part of theemergence of a new paradigm for chem-istry, the question of priority from whichwe began would never have seemed soimportant. In this case as in others, thevalue placed upon a new phenomenonand thus upon its discoverer varies withour estimate of the extent to whichthe phenomenon violated paradigm-induced anticipations. Notice, however,since it will be important later, that thediscovery of oxygen was not by itself thecause of the change in chemical theory.Long before he played any part in thediscovery of the new gas, Lavoisier wasconvinced both that something waswrong with the phlogiston theory andthat burning bodies absorbed some partof the atmosphere. That much he hadrecorded in a sealed note deposited withthe Secretary of the French Academy in1772. What the work on oxygen did wasto give much additional form and struc-ture to Lavoisier's earlier sense thatsomething was amiss. It told him a thinghe was already prepared to discover-thenature of the substance that combustionremoves from the atmosphere. That ad-vance awareness of difficulties must be asignificant part of what enabled La-

voisier to see in experiments like Priest-ley's a gas that Priestley had been unableto see there himself Conversely, the factthat a major paradigm revision wasneeded to see what Lavoisier saw mustbe the principal reason why Priestleywas, to the end of his long life, unable tosee it ....

To a greater or lesser extent (corre-sponding to the continuum from theshocking to the anticipated result), thecharacteristics common to the ... exam-ple above are characteristic of all discov-eries from which new sorts of phenom-ena emerge. Those characteristicsinclude: the previous awareness ofanomaly, the gradual and simultaneousemergence of both observational andconceptual recognition, and the conse-quent change of paradigm categories andprocedures often accompanied by resist-ance. There is even evidence that thesesame characteristics are built into the na-ture of the perceptual process itself In apsychological experiment that deservesto be far better known outside the trade,Bruner and Postman asked experimentalsubjects to identify on short and con-trolled exposure a series of playing cards.Many of the cards were normal, butsome were made anomalous, e.g., a redsix of spades and a black four of hearts.Each experimental run was constitutedby the display of a single card to a singlesubject in a series of gradually increasedexposures. After each exposure the sub-ject was asked what he had seen, and therun was terminated by two successivecorrect identifications.

Even on the shortest exposures manysubjects identified most of the cards, andafter a small increase all the subjectsidentified them all. For the normal cardsthese identifications were usually correct,but the anomalous cards were almost al-ways identified, without apparent hesita-

THOMAS KUHN The Structure ofScientijic Revolutions

tion or puzzlement, as normal. The blackfour of hearts might, for example, beidentified as the four of either spades orhearts. Without any awareness of trou-ble, it was immediately fitted to one ofthe conceptual categories prepared byprior experience. One would not evenlike to say that the subjects had seensomething different from what theyidentified. With a further increase of ex-posure to the anomalous cards, subjectsdid begin to hesitate and to displayawareness of anomaly. Exposed, for ex-ample, to the red six of spades, somewould say: That's the six of spades, butthere's something wrong with it-theblack has a red border. Further increaseof exposure resulted in still more hesita-tion and confusion until finally, andsometimes quite suddenly, most subjectswould produce the correct identificationwithout hesitation. Moreover, after do-ing this with two or three of the anom-alous cards, they would have little furtherdifficulty with the others. A few subjects,however, were never able to make therequisite adjustment of their categories.Even at forty times the average exposurerequired to recognize normal cards forwhat they were, more than 10 percent ofthe anomalous cards were not correctlyidentified. And the subjects who thenfailed often experienced acute personaldistress. One of the exclaimed: -r can'tmake the suit out, whatever it is. It didn'teven look like a card that time. I don'tknow what color it is now or whether it'sa spade or a heart. I'm not even sure nowwhat a spade looks like. My God!" In thenext section we shall occasionally see sci-entists behaving this way too.

Either as a metaphor or because itreflects the nature of the mind, that psy-chological experiment provides a won-derfully simple and cogent schema forthe process of scientific discovery. In sci-

103

ence, as in the playing card experiment,novelty emerges only with difficulty,manifested by resistance, against a back-ground provided by expectation. Ini-tially, only the anticipated and usual areexperienced even under circumstanceswhere anomaly is later to be observed.Further acquaintance, however does re-sult in awareness of something wrong ordoes relate the effect to something thathas gone wrong before. That awarenessof anomaly opens a period in which con-ceptual categories are adjusted until theinitially anomalous has become the an-ticipated. At this point the discovery hasbeen completed. I have already urgedthat that process or one very much like itis involved in the emergence of all fun-damental scientific novelties. Let menow point out that, recognizing theprocess, we can at last begin to see whynormal science, a pursuit not directed tonovelties and tending at first to suppressthem, should nevertheless be so effectivein causing them to arise.

In the development of any science,the first received paradigm is usually feltto account quite successfully for most ofthe observations and experiments easilyaccessible to that science's practitioners.Further development, therefore, ordinar-ily calls for the construction of elaborateequipment, the development of an eso-teric vocabulary and skills, and a refine-ment of concepts that increasinglylessens their resemblance to their usualcommon-sense prototypes. That profes-sionalization leads, on the one hand, toan immense restriction of the scientist'svision and to a considerable resistance toparadigm change. The science has be-come increasingly rigid. On the otherhand, within those areas to which theparadigm directs the attention of thegroup, normal science leads to a detail ofinformation and to a precision of the

104 CHAPTER THREE What Does Science Tell Me About the World?

observation-theory match that could beachieved in no other way. Furthermorethat detail and precision-of-match havea value that transcends their not alwaysvery high intrinsic interest. Without thespecial apparatus that is constructedmainly for anticipated functions, the re-sults that lead ultimately to novelty couldnot occur. And even when the apparatusexists, novelty ordinarily emerges onlyfor the man who, knowing with precisionwhat he should expect, is able to recog-nize that something has gone wrong.Anomaly appears only against the back-ground provided by the paradigm. Themore precise and far-reaching that para-digm is, the more sensitive an indicatorit provides of anomaly and hence of anoccasion for paradigm change. In thenormal mode of discovery, even resist-ance to change has a use that will be ex-plored more fully in the next section. Byensuring that the paradigm will not betoo easily surrendered, resistance guar-antees that scientists will not be lightlydistracted and that the anomalies thatlead to paradigm change will penetrateexisting knowledge to the core. The veryfact that a significant scientific novelty sooften emerges simultaneously from sev-erallaboratories is an index both to thestrongly traditional nature of normal sci-ence and to the completeness with whichthat traditional pursuit prepares the wayfor its own change ....

If awareness of anomaly plays a role inthe emergence of new sorts of phenom-ena, it should surprise no one that a sim-ilar but more profound awareness is pre-requisite to all acceptable changes oftheory. On this point historical evidenceis, I think, entirely unequivocal. Thestate of Ptolemaic astronomy was a scan-dal before Copernicus' announcement.Galileo's contributions to the study ofmotion depended closely upon difficul-

ties discovered in Aristotle's theory byscholastic critics. Newton's new theoryof light and color originated in the dis-covery that none of the existing pre-paradigm theories would account for thelength of the spectrum, and the wavetheory that replaced Newton's was an-nounced in the midst of growing con-cern about anomalies in the relation ofdiffraction and polarization effects toNewton's theory. Thermodynamics wasborn from the collision of two existingnineteenth-century physical theories,and quantum mechanics from a varietyof difficulties surrounding black-bodyradiation, specific heats, and the photo-electric effect. Furthermore, in all thesecases except that of Newton the aware-ness of anomaly had lasted so long andpenetrated so deep that one can appro-priately describe the fields affected by itas in a state of growing crisis. Because itdemands large-scale paradigm destruc-tion and major shifts in the problemsand techniques of normal science, theemergence of new theories is generallypreceded by a period of pronounced pro-fessional insecurity. As one might ex-pect, that insecurity is generated by thepersistent failure of the puzzles of nor-mal science to come out as they should.Failure of existing rules is the prelude toa search for new ones ....

Philosophers of science have repeat-edly demonstrated that more than onetheoretical construction can always beplaced-upon a given collection of data.History of science indicates that, partic-ularly in the early developmental stagesof a new paradigm, it is not even verydifficult to invent such alternates. Butthat invention of alternates is just whatscientists seldom undertake except dur-ing the pre-paradigm stage of their sci-ence's development and at very specialoccasions during its subsequent evolu-

THOMAS KUHN The Structure of Scientific Revolutions

tion. So long as the tools a paradigmsupplies continue to prove capable ofsolving the problems it defines, sciencemoves fastest and penetrates most deeplythrough confident employment of thosetools. The reason is clear. As in manufac-ture so in science-retooling is an ex-travagance to be reserved for the occa-sion that demands it. The significance ofcrises is the indication they provide thatan occasion for retooling has arrived ....

How, then, do scientists respond tothe awareness of an anomaly in the fitbetween theory and nature? What hasjust been said indicates that even a dis-crepancy unaccountably larger than thatexperienced in other applications of thetheory need not draw any very profoundresponse. There are always some discrep-ancies. Even the most stubborn onesusually respond at last to normal prac-tice. Very often scientists are willing towait, particularly if there are many prob-lems available in other parts of thefield. . .. [D Juring the sixty years afterNewton's original computation, the pre-dicted motion of the moon's perigee re-mained only half of that observed. AsEurope's best mathematical physicistscontinued to wrestle unsuccessfully withthe well-known discrepancy, there wereoccasional proposals for a modificationof Newton's inverse square law. But noone took these proposals very seriously,and in practice this patience with a majoranomaly proved justified. Clairaut in1750 was able to show that only themathematics of the application had beenwrong and that Newtonian theory couldstand as before. Even in cases where nomere mistake seems quite possible (per-haps because the mathematics involvedis simpler or of a familiar and elsewheresuccessful sort), persistent and recog-nized anomaly does not always inducecrisis. No one seriously questioned New-

105

tonian theory because of the long-recognized discrepancies between pre-dictions from that theory and both thespeed of sound and the little motion ofMercury. The first discrepancy was ulti-mately and quite unexpectedly resolvedby experiments on heat undertaken for avery different purpose; the second van-ished with the general theory of relativ-ity after a crisis that it had had no role increating. Apparently neither had seemedsufficiently fundamental to evoke themalaise that goes with crises. They couldbe recognized as counterinstances andstill be set aside for later work.

It follows that if an anomaly is toevoke crisis, it must usually be more thanjust an anomaly. There are always diffi-culties somewhere in the paradigm-nature fit; most of them are set rightsooner or later, often by processes thatcould not have been foreseen. The scien-tist who pauses to examine every anom-aly he notes will seldom get significantwork done. We therefore have to askwhat it is that makes an anomaly seemworth concerted scrutiny, and to thatquestion there is probably no fully gen-eral answer. The cases we have alreadyexamined are characteristic but scarcelyprescriptive. Sometimes an anomaly willclearly call into question explicit andfundamental generalizations of the para-digm, as the problem of ether drag didfor those who accepted Maxwell's the-ory. Or, as in the Copernican revolution,an anomaly without apparent funda-mental import may evoke crisis if the ap-plications that it inhibits have a particu-lar practical importance, in this case forcalendar design and astrology. Or, as ineighteenth-century chemistry, the devel-opment of normal science may trans-form an anomaly that had previouslybeen only a vexation into a source of cri-sis: the problem of weight relations had a

106 CHAPTER THREE What Does Science Tell Me About the World?

very different status after the evolutionof pneumatic-chemical techniques. Pre-sumably there are still other circum-stances that can make an anomaly par-ticularly pressing, and ordinarily severalof these will combine. We have alreadynoted, for example, that one source ofthe crisis that confronted Copernicuswas the mere length of time duringwhich astronomers had wrestled unsuc-cessfully with the reduction of the resid-ual discrepancies in Ptolemy's system.

When, for these reasons or others likethem, an anomaly comes to seem morethan just another puzzle of .normal sci-ence, the transition to crisis and to ex-traordinary science has begun. Theanomaly itself now comes to be moregenerally recognized as such by the pro-fession. More and more attention is de-voted to it by more and more of the field'smost eminent men. If it still continues toresist, as it usually does not, many of themmay come to view its resolution as thesubject matter of their discipline. Forthem the field will no longer look quitethe same as it had earlier. Part of its dif-ferent appearance results simply from thenew fixation point of scientific scrutiny.An even more important source of changeis the divergent nature of the numerouspartial solutions that concerted attentionto the problem has made available. Theearly attacks upon the resistant problemwill have followed the paradigm rulesquite closely. But with continuing resist-ance, more and more of the attacks uponit will have involved some minor or not sominor articulation of the paradigm, notwo of them quite alike, each partiallysuccessful, but none sufficiently so to beaccepted as paradigm by the group.Through this proliferation of divergentarticulations (more and more frequentlythey will come to be described as ad hocadjustments), the rules of normal science

become increasingly blurred. Thoughthere still is a paradigm, few practitionersprove to be entirely agreed about what itis. Even formerly standard solutions ofsolved problems are called in question ....

Confronted with anomaly or with cri-sis' scientists take a different attitude to-ward existing paradigms, and the natureof their research changes accordingly.The proliferation of competing articula-tions, the willingness to try anything, theexpression of explicit discontent, the re-course to philosophy and to debate overfundamentals, all these are symptoms of atransition from normal to extraordinaryresearch. It is upon their existence morethan upon that of revolutions that thenotion of normal science depends ....

In learning a paradigm the scientistacquires theory, methods, and standardstogether, usually in an inextricable mix-ture. Therefore, when paradigms change,there are usually significant shifts in thecriteria determining the legitimacy bothof problems and of proposed solutions.

That observation returns us to thepoint from which this section began, forit provides our first explicit indication ofwhy the choice between competing para-digms regularly raises questions that can-not be resolved by the criteria of normalscience. To the extent, as significant as itis incomplete, the two scientific schoolsdisagree about what is a problem andwhat a solution, they will inevitably talkthrough each other when debating therelative merits of their respective para-digms. In the partially circular argumentsthat regularly result, each paradigm willbe shown to satisfy more or less the crite-ria that it dictates for itself and to fallshort of a few of those dictated by its op-ponent. There are other reasons, too, forthe incompleteness oflogical contact thatconsistently characterizes paradigm de-bates. For example, since no paradigm

KARL POPPER Science: Conjectures and Refutations

ever solves all the problems it defines andsince no two paradigms leave all the sameproblems unsolved, paradigm debates al-ways involve the question: Which prob-lems is it more significant to have solved?Like the issue of competing standards,

107

that question of values can be answeredonly in terms of criteria that lie outside ofnormal science altogether, and it is thatrecourse to external criteria that most ob-viously makes paradigm debates revolu-tionary.