L. Lavoisier), after Priestley’s experiment of 1774, and perhaps as a result of a hint taken from Priestley, began the research that would lead him to oxygen. Early in 1775, Lavoisier said that the gas obtained by heating the red oxide of mercury was “air itself unaltered, only… purer, more fit for respiration.”3) By 1777, perhaps as the result of a second hint from Priestley, Lavoisier had concluded that the gas was a distinct chemical species and one of the two principal constituents of the atmosphere—a view Priestley could never accept.
This aspect of the discovery raises a question that is invariably asked about new phenomena that have entered the domain of scientists’ awareness. Who first discovered oxygen—Priestley or Lavoisier, if either of the two? And in either case, when was oxygen discovered? Even if there had been only one claimant to the discovery, this form of question would still arise. As a judgment about priority and the date of discovery, whatever answer is given is of little consequence to us. Nevertheless, the attempt to produce an answer will illuminate the nature of discovery, because the kind of answer being sought does not exist. Discovery is not the sort of process about which such a question can properly be asked. The very fact that the question is asked—the priority of the discovery of oxygen has been debated continuously since the 1780s—is a symptom of something amiss in an image of science that assigns discovery so fundamental a role. Let us look once more at the example of oxygen. The claim that Priestley discovered oxygen rests on the fact that he was the first to isolate a gas that later came to be recognized as a particular species. But the sample Priestley obtained was not pure. If obtaining impure oxygen is the discovery of oxygen, then everyone who has ever bottled atmospheric air should be said to have discovered oxygen. Moreover, if Priestley was the discoverer, when was the discovery made? In 1774 he thought it was nitrous oxide, a species he already knew. In 1775 he thought the gas was dephlogisticated air, which was not yet oxygen and, to a phlogiston chemist, was even a wholly unexpected kind of gas. Lavoisier’s claim has greater strength, but the same problem arises here as well.
If we deny Priestley the credit, we likewise cannot award the honor to Lavoisier on the basis of his work in 1775, when he regarded the gas as “air itself entire.” Perhaps we must wait until Lavoisier’s work of 1776 and 1777, when he not only saw the gas but also came to know what it was. Even that judgment, however, is open to doubt, for in 1777 and to the end of his life Lavoisier maintained that oxygen was the atomic “principle of acidity,” and that oxygen gas was produced only when that “principle” combined with caloric, the matter of heat.4) Are we then to say that oxygen had still not been discovered in 1777? Some may be inclined to say so. Yet the concept of the principle of acidity did not disappear from chemistry until after 1810, and the concept of caloric survived into the 1860s. Oxygen had become a standard chemical substance earlier than any of those dates.
Clearly, new terms and concepts are required for analyzing events such as the discovery of oxygen. Though it is undoubtedly correct, the phrase “oxygen was discovered,” like our ordinary—and also suspect—notion of seeing, misleads by suggesting that discovering something is a single, simple act. That is precisely why we so readily imagine that discovery, like seeing or touching, must be accomplished by one person at one sharply defined moment. But it is impossible to assign discovery to a single instant, and often it is equally impossible to assign it to a single individual. If we ignore Scheele, we may safely say that oxygen was not discovered before 1774, and perhaps we may also say that it was discovered around 1777 or shortly thereafter. But any attempt to date the discovery within such limits, or within any similar set of limits, must inevitably be arbitrary, because the discovery of a new kind of phenomenon is necessarily a complex event involving both the recognition that something exists and the recognition of what it is. For example, if oxygen were for us dephlogisticated air, then even without knowing when it was discovered we would have to insist without hesitation that Priestley discovered it. But if observation and conceptualization, fact and assimilation to theory, are closely intertwined in the process of discovery, then discovery is a process, and it must take time. Only when all the relevant conceptual categories are prepared in advance—that is, when the phenomenon is not of a new type—can the discovery of it and the determination of what it is occur together, immediately, in a single moment.
Let us now take it for granted that discovery involves an extended, though not necessarily prolonged, process of conceptual assimilation. Can we then also say that discovery involves a change in paradigm? No universal answer to this question can yet be given, but in the case of oxygen, at least, the answer must be yes. What Lavoisier published in his papers from 1777 onward was not so much the discovery of oxygen as the oxygen theory of combustion. That theory became the wedge for such a far-reaching systematic reconstruction of chemistry that it is usually called the chemical revolution. Indeed, if the discovery of oxygen had not been central to the emergence of a new paradigm for chemistry, the question of priority with which we began would not have seemed so important. Here, as in other cases, the value assigned to a new phenomenon, and therefore to its discoverer, depends on an assessment of the degree to which that phenomenon departs from predictions derived from a paradigm. Yet, because it will be important in the discussion that follows, note that the discovery of oxygen itself was not the cause of the change in chemical theory. Long before he contributed to the discovery of the new gas, Lavoisier had already realized two things: that something was wrong with the phlogiston theory, and that burning bodies absorb some constituent of the atmosphere.
He had already recorded as much in a sealed note deposited with the Secretary of the French Academy in 1772.5) What the work on oxygen contributed was that it gave a much more concrete form and structure to Lavoisier’s earlier sense that something was wrong. It told Lavoisier what he was already prepared to discover: the nature of the substance that combustion removes from the atmosphere. This prior recognition of something amiss must have been an important factor that enabled Lavoisier, while performing experiments like Priestley’s, to see a gas there that Priestley could not see. Put another way, the fact that a major revision of the paradigm was required in order to see what Lavoisier saw must be the principal reason why Priestley, to the end of his long life, was unable to see it.
Two other, simpler examples will make what has just been said much clearer, and at the same time will lead us from characterizing the nature of discovery toward understanding the background from which such discoveries arise in science. In an attempt to represent the principal ways in which discoveries are made, these examples have been chosen not only to differ from one another but also to differ in character from the discovery of oxygen. The first example, X-rays, is a classic instance of an accidental discovery, a type that occurs far more often than the usual standards of scientific papers make it easy for us to realize. The story begins with a day in the life of the physicist Röntgen, who was forced to interrupt his regular research on cathode rays because he noticed that a barium platinocyanide screen, some distance from his shielded experimental apparatus, glowed during a discharge. More detailed investigation—during which Röntgen remained shut up in his laboratory in a frenzy for seven weeks—revealed, among other things, that the cause of the glow came in straight lines from the cathode-ray tube and that the path of the radiation was not deflected by a magnetic field. Before announcing his discovery, Röntgen had realized that his results were not due to cathode rays but to some agent bearing at least a partial resemblance to light.6)
Even this brief summary reveals a striking resemblance to the discovery of oxygen. Before experimenting with the red oxide of mercury, Lavoisier had performed experiments that did not yield the results expected under the phlogiston paradigm. Röntgen’s discovery began with the recognition that his screen was glowing when it should not have been. In both cases, the perception of anomaly—the oddity of a phenomenon for which the researcher’s paradigm had not prepared him—played an essential role in opening the way to the recognition of novelty. But again, in both cases, the perception that something was wrong was only a prelude to discovery. Neither oxygen nor X-rays could emerge without a further, step-by-step process of experimentation and assimilation. At what point in Röntgen’s research, for example, should we say that X-rays were truly discovered? Not at the first moment, when all that had been noticed was a glowing screen. At least one other person besides Röntgen had seen such a glow, but to his misfortune he discovered nothing at all.7) And almost certainly the moment of discovery cannot be pushed to some point in the final week of the investigation, for by then Röntgen was already determining the properties of the new radiation he had discovered. We can only say that X-rays were born in Würzburg at some point between November 8 and December 28, 1895.
However, in the third area, meaningful similarities between the discovery of oxygen and the discovery of X-rays are revealed far more opaquely. Unlike the discovery of oxygen, the case of X-rays was not hinted at in any upheaval of scientific theory for at least some ten years after that event. Then in what sense can the assimilation of that discovery be said to have made paradigm change inevitable? The case against such a change is quite strong. What is certain is that the paradigm agreed upon by Röntgen and the researchers of his era could not have predicted X-rays [until then, Maxwell’s electromagnetic theory had not been widely accepted, and the theory that cathode rays consisted of particles was no more than one of several conjectures at the time]. Yet their paradigm did not deny the existence of X-rays in any obvious sense, just as the phlogiston theory had obstructed Lavoisier’s explanation of Priestley’s gas. Rather, by 1895, accepted scientific theory and practice had come to acknowledge various forms of radiant waves—visible light, infrared, and ultraviolet rays. Why could X-rays not be accepted as merely one additional form of a well-known class of natural phenomena? Why were X-rays not regarded, for example, as the discovery of one more chemical element? The task of finding and placing new elements into the empty slots of the periodic table was still continuing in Röntgen’s time, and they were actually being discovered. The pursuit of such efforts was a standard project of normal science, and achieving success was not a surprise but a cause for celebration.
But X-rays were received not only as a surprise but as a shock. Lord Kelvin initially declared X-rays to be an elaborate hoax.8) Other scholars, though they could not doubt the evidence, were clearly bewildered by that discovery. Though X-rays were not denounced as heresy by established theory, they were violating entrenched expectations. Such predictions, I believe, were implicitly present in the design and interpretation of established experimental procedures. By the 1890s, cathode ray apparatus had been widely distributed across various laboratories in Europe. If X-rays had been produced in Röntgen’s apparatus, then numerous other experimenters must also have been producing them for considerable periods without yet recognizing them. Perhaps other such radiations, for which there might have been unknown sources, were also implicit in the behavior described above, without being associated with them. At least some apparatus that had long been in regular use must subsequently have required shielding with lead. Research already completed in standard projects must now have had to be reexamined, because previous scientists had not recognized and controlled the variables relating to X-rays. Needless to say, X-rays opened up new fields and thereby were added to the potential domain of normal science. However, X-rays also changed existing fields, in an even more important respect for the present, and in that process they stripped the former paradigmatic forms of the apparatus of their paradigmatic rights.
In short, whether conscious or not, the decision to introduce a particular apparatus and apply it in a specific manner presupposes an assumption that only certain kinds of situations will arise. It involves not only theoretical predictions but also instrumental expectations, and these have often played a decisive role in the development of science. For instance, as one such prediction, part of the story of the delayed discovery of oxygen may be cited. Through the standard test for “the goodness of air,” both Priestley and Lavoisier mixed their gas with nitrous oxide in a 2:1 volume ratio, shook the mixture over water, and measured the volume of the residual gas. Past experience that had led to this standard procedure convinced them that if they experimented with atmospheric air, the volume of residual gas would be one, whereas for any other gas (or polluted air) the volume would be greater than that. In the oxygen experiments, both men confirmed the residual gas. Only long afterwards, and sometimes through accidental events, did Priestley abandon the standard procedure and try mixing nitrous oxide with his gas in a different ratio. There he learned that if he increased the volume of nitrous oxide fourfold and mixed it, there was almost no remaining gas whatsoever. Adhering to the original test method—a process sanctified by much earlier experience—meant also persisting in the belief that gases acting like oxygen did not exist.11) Cases of this kind are numerous, as in the delayed confirmation of uranium fission, for instance.
One of the reasons it was unusually difficult to uncover that nuclear reaction was that those who knew what was predicted when uranium was bombarded chose chemical tests aimed mainly at elements located at the top of the periodic table.10) Judging by the number of times such instrumental beliefs proved erroneous, may we conclude that science must abandon standard tests and standard instruments? That would result in research methods beyond imagination. Paradigm experimental methods and applications are as essential to science as paradigm laws and theories, and their influence is likewise the same. Inevitably, paradigm procedures and applications limit the phenomenological field that can be an object of scientific inquiry at any given time. By recognizing this much, we come to realize as well the essential significance of why a discovery like X-rays makes paradigm change—and therefore change in both experimental procedure and prediction—inevitable for particular factions of the scientific community. Consequently, we can also understand how the discovery of X-rays came to appear to many scientists as the opening of a strange and new world, and how thereby it could exert such a profound influence on the crisis that led to twentieth-century physics.
As a final example of scientific discovery, the case of the Leyden jar belongs to the class that can be called theory-induced. At first this term may feel paradoxical. Most of what has been discussed so far suggests that discoveries predicted in advance by theory constitute part of normal science and do not bring about new kinds of facts. Earlier I cited, as advances from normal science by such means, the discovery of new chemical elements during the latter half of the nineteenth century. But not all theories are paradigm theories. In the pre-paradigm period—and during the crisis of large-scale paradigm transformation—scientists usually develop various conjectural and unclarified theories that can themselves point the way to discovery. However, such discoveries are made only when clarified by often conjectural and tentative hypotheses, and theory is born as a paradigm.
The discovery of the Leyden jar exhibits all these characteristics, as well as the features examined earlier. When that discovery began, no single paradigm existed in electrical research. Instead, numerous theories derived from relatively accessible phenomena were competing with one another. Not one of those theories had succeeded in governing all electrical phenomena well. This failure is precisely the source of the various anomalous phenomena that provided the background for the discovery of the Leyden jar. One school of electricians treated electricity as a fluid, and because of this notion many people tried to capture the fluid in a bottle by holding a water-filled glass jar in their hand and bringing the water into contact with a wire attached to an operating electrostatic generator. When they removed the bottle from the machine and touched the water (or the wire connected to the water) with the other hand, all the experimenters felt a severe electric shock. But those initial experiments did not immediately provide electricians with the Leyden jar. Such a device took shape only gradually, and here too it is difficult to say precisely when the discovery was completed. The initial attempts to store electrical fluid succeeded only because the experimenters had their feet on the ground and held the glass jar in their hands. Electricians had not yet realized that the jar required not only an internal conductive coating but also an external one, nor did they know that it was not actually fluid that was being stored in the bottle. Only at some point in the course of investigation that proved this fact to them and showed them several other anomalous results did the so-called device called the Leyden jar appear. Moreover, the experiments leading up to its appearance—many of which were performed by Franklin—also made comprehensive revision of the fluid theory inevitable, and thereby provided the first complete paradigm for electricity.11) Though the degree differs (corresponding to the continuum from shock to expected results), the characteristics common to the three cases cited above are the characteristic nature of all discoveries from which new aspects of phenomena emerge. These characteristics include prior awareness of anomaly, the gradual and simultaneous emergence of observational and theoretical recognition, and the resulting change in paradigm categories and procedures, often accompanied by resistance. Evidence has even been obtained that these same characteristics are inherent in the very nature of perceptual processes. In a psychological experiment for which there is ample reason to be far better known outside psychology, Bruner and Postman showed subjects a briefly altered deck of playing cards and asked them to identify it. For example, they made the six of spades red and the four of hearts black. In one run of the experiment, cards were shown to each person one at a time, with the number of exposures gradually increased. Each time a card was shown, the subject was asked what he had seen, and the run ended when he identified it correctly twice in succession.12)
Even with very brief exposure, most subjects recognized almost all the cards; with increased exposure, they were usually correct about normal cards, but anomalous cards were identified as normal ones almost without exception, and with no outward sign of hesitation or confusion. For example, the black four of hearts was answered as the four of spades or the four of hearts. Without any reservation, it immediately fit into one of the conceptual categories that prior experience had provided. None of the subjects tried to say that they had seen a different card from the one they had answered. As anomalous cards were shown increasingly often, the subject began to hesitate and started to reveal a sense of anomaly. For instance, after being shown the red six of spades several times, some would say: “That’s the six of spades, but something is a bit odd about it; has it been bordered in red on black?” Prolonged exposure caused more prolonged hesitation and confusion until finally, at some point—sometimes quite suddenly—most subjects began to identify them correctly without hesitation. Moreover, after conducting this test with two or three anomalous cards, they no longer showed much difficulty with other anomalous cards. But some individuals could not properly adapt their categories. Even after encountering cards forty times more than the average needed to identify normal cards correctly, more than 10 percent of the anomalous cards could not be properly identified. Those who failed in this way often felt considerable confusion. One person even exclaimed: “I can’t make out the cards at all. That one didn’t even look like a card then. Now I don’t even know what color it is, or whether it’s a spade or a heart. I’m even bewildered about what a spade looks like now. My God!”13) In the next section, we shall often see that scientists behave in this manner as well.
As a metaphor, or because it reflects the very nature of the mind, these psychological experiments provide a remarkably simple and persuasive schematic account of the process of scientific discovery. As in the playing-card experiment, in science a striking novelty emerges only after breaking through difficulties made conspicuous by resistance to what had been expected. Even in situations later arranged so that anomalies may be observed, at first only the anticipated and the customary are experienced. With deeper perception, however, one comes to realize that something is wrong, or else to connect the result with something that had previously gone awry. This recognition of anomaly opens the period in which conceptual categories are adjusted, until at last what was initially anomalous becomes what is expected. At that point, the discovery is complete. As emphasized above, such a process, or one very much like it, is involved wherever fundamental novelty appears in science. What I wish to stress here is that, by recognizing that process, we can at last begin to see why normal science, though it is an inquiry not directed toward novelty and indeed inclined to suppress it, is nevertheless so effective in bringing about innovation.
In the development of any science, the first accepted paradigm usually seems to account quite successfully for most of the observations and experiments readily accessible to its practitioners. Further development, therefore, ordinarily requires the construction of elaborate apparatus, the development of an esoteric vocabulary and techniques, and the refinement of concepts that increasingly diminish their agreement with common sense. Such specialization, on the one hand, greatly restricts the scientist’s field of vision and acts as a considerable resistance to paradigm change. Science becomes increasingly rigid. On the other hand, in those areas to which the paradigm directs the group’s attention, normal science leads to a detailed articulation of information and to a precision of observation-theory match that could not be achieved in any other way. Moreover, such detail and precision-of-match possess a value that exceeds their intrinsic interest, which is not always very great. Without special apparatus constructed chiefly for the functions expected of it, the results that ultimately led to novelty would never have occurred. And even when the apparatus is in place, novelty reveals itself only to the person who knows exactly what to expect and can recognize that something has gone wrong. Anomaly appears only against the background provided by a paradigm. The more precise and influential the paradigm is, the more sensitive an indicator it provides of anomaly, and thus of the possibility of paradigm change. In the normal pattern of discovery, even resistance to paradigm change has its usefulness, as will be examined in greater detail in the next section. By ensuring that the paradigm will not collapse without force, resistance makes certain that scientists will not be shaken by trifles, and guarantees that the anomalies leading to paradigm change penetrate to the very core of existing knowledge. The fact that significant new discoveries in science often appear simultaneously in several laboratories is itself an indication both of the intensely traditional character of normal science and of the completeness with which such conventional inquiry prepares the way for its own transformation.