As for a pure observation-language, perhaps our hope of ever expecting such a case still depends entirely on theories of perception and mind. And modern psychological experiments are rapidly multiplying phenomena that can scarcely be dealt with by such theories. And modern psychological experiments are rapidly multiplying phenomena that can scarcely be dealt with by such theories.
The duck-rabbit experiment demonstrates that two people receiving the same retinal image can see different things. Inverting lenses show that two people receiving different retinal impressions can see the same thing. Psychology contains a great deal of evidence showing this same effect, and the doubt it raises is that no attempt to present an actual observation-language has yet come close to pure perceptual terms that can be generally applied. Yet when we examine those attempts that have come closest, a common feature appears that greatly reinforces several of the main themes of this essay. From the outset, such attempts presuppose a paradigm gathered from current scientific theory or from ordinary exchange of opinions, and then strive to eliminate from that paradigm all nonlogical and nonperceptual terms. In some areas of communication, this effort has been carried out to a considerable degree and has produced excellent results. There can be no doubt that efforts of this kind are worth pursuing. But their result is a language—which is like the languages used in science—that embodies many predictions about nature, but fails to function when those predictions go astray. Nelson Goodman stated precisely this point when describing the purpose of his book “The Structure of Appearance.” “It is fortunate that nothing is in question ‘beyond the phenomena known to exist,’ for the concept of a ‘possible’ case, that is, a case that does not actually exist but might have existed, is extremely uncertain.”19) Thus, no language confined to expressing a world already fully known can accomplish a pure, neutral, objective record of “the given.” Philosophical inquiry has not yet provided even a hint of what sort of language could fulfill such a mission.
Under these circumstances, we may at least doubt whether, when scientists treat oxygen and pendulums—and perhaps atoms and electrons—as the basic elements of their immediate experience, they are right not only in the practice of science but also in principle. As a result of racial, cultural, and finally professional experience embodied in paradigms, the scientist’s world has become filled with planets and pendulums, condensers and compound minerals, and many other objects. Compared with these objects of perception, meter-stick readings and retinal images are both sophisticated constructs to which experience has direct access only when the scientist, for the special purposes of the research he is conducting, decides that someone should do so. This does not suggest, for example, that the only things a scientist can see when looking at a swinging stone are pendulums. We have already noted that members of another scientific community could also see constrained fall. What it suggests is that a scientist looking at a swinging stone cannot, in principle, have an experience of a character more elementary than seeing a pendulum. The alternative was not some hypothetical “fixed” vision, but a vision that, through another paradigm, made the swinging stone into something else.
All this will seem more plausible if we recall once again that neither scientists nor ordinary people learn to see the world piecemeal, or element by element. Except in cases where all the conceptual and operational categories are already prepared in advance—for example, when discovering additional transuranium elements, or when a new house comes into view—both scientists and ordinary people sort out whole regions from the flow of experience. When a child first shifts the word “mama” from all people to all women, and then finally to his own mother, he is not merely learning what the word “mama” means or who his mother is. At the same time, he is learning something about how everyone but one woman behaves toward him, and also about the difference between men and women. The infant’s reactions, expectations, and beliefs—in fact, much of the world perceived by him—change accordingly. In the same way, the Copernicans who refused to call the sun by its traditional name, “planet,” were not merely learning what “planet” meant or what the sun was. Rather, they were changing the meaning of “planet” so that it could continue to make useful distinctions in a world where not only the sun but all heavenly bodies appeared in a way entirely different from before. The same can be said of the other examples presented earlier. Seeing oxygen instead of dephlogisticated air, a condenser instead of a Leyden jar, or a pendulum instead of constrained fall was only one part of the scientists’ overall transformation of vision regarding the many related chemical, electrical, and mechanical phenomena. Paradigms play a decisive role all at once across vast regions of experience.
But the search for operational definitions or for a pure observation-language can begin only after experience has been determined in this way. The scientist or philosopher who asks what measurement or what retinal impression makes a pendulum a pendulum must already be able to recognize it as a pendulum as soon as he sees it. If he had seen constrained fall instead of a pendulum, his question could not even have been raised. And if he had seen a pendulum, but seen it in the way he saw a tuning fork or an oscillation balance, his question could not have been answered.
Therefore, although such questions are always legitimate and sometimes astonishingly fruitful, questions about retinal impressions or the results of particular experimental operations presuppose a world already subdivided in certain ways perceptually and conceptually. In one sense, such questions become part of normal science because they depend on the existence of a paradigm and receive different answers as a result of paradigm change.
To conclude this section, let us set aside the problem of retinal images here and once again restrict our attention to laboratory work, which provides the scientist with the fragmentary but concrete indices he has already seen. One way in which such laboratory work changes along with paradigms has already been observed several times above. After a scientific revolution, many earlier measurements and mechanisms become meaningless and are replaced by others. A scientist does not apply to oxygen all the same tests he applied to dephlogisticated air. But changes of this kind are not total. Whatever he then comes to perceive, the post-revolutionary scientist is still looking at the same world. Moreover, even if he used them differently before the revolution, much of his language and most of his experimental apparatus remain the same as before. As a result, post-revolutionary science, like pre-revolutionary science, continues to include many of the same mechanisms performed with the same instruments and described in the same terms. If these continuing mechanisms have changed somewhere, the change must occur either in their relation to the paradigm or in their concrete results. Now, by introducing one final new example, I wish to suggest that both of these types of change occur. Examining the work of Dalton and his contemporaries, we will find that one and the same operation, when connected to nature through a different paradigm, can become an index of an entirely different aspect of nature’s regularity. In addition, we will see that sometimes an old operational method, by virtue of its new role, produces different concrete results.
For most of the eighteenth century, and even into the nineteenth, European chemists almost universally believed that the elementary atoms composing all chemical species were bound together by forces of mutual affinity. Thus, a lump of silver held together because of the affinity among corpuscles of silver. Until after Lavoisier, these corpuscles themselves were thought to be compounded from still more elementary particles. According to this very theory, the reason silver dissolves in acid—or salt dissolves in water—was that the particles of acid attracted the particles of silver—or the particles of water attracted the particles of salt—more strongly than those particles of solute attracted one another. It was also believed that because copper-acid affinity was greater than the affinity of acid for silver, copper dissolved in a silver solution and precipitated the silver. Many other phenomena were explained in this same way. During the eighteenth century, the theory of elective affinity served as an excellent chemical paradigm and was widely, and sometimes successfully, applied to the design and analysis of chemical experiments.20)
But affinity theory distinguished physical mixtures from chemical compounds in a way that has seemed strange since the assimilation of Dalton’s work. Eighteenth-century chemists recognized two kinds of processes. When mixing produced heat, light, effervescence, or something similar, they regarded a chemical combination as having occurred. On the other hand, when the particles of a mixture could be separated or distinguished with the naked eye or by using instruments, it was regarded as merely a physical mixture.
But in numerous intermediate cases—salt dissolved in water, alloys, glass, oxygen in the atmosphere, and so on—these crude criteria were of little use. Guided by their paradigm, most chemists regarded this entire intermediate domain as chemical, because they believed the processes composing it were all governed by the same kind of force. Salt in water or oxygen mixed with nitrogen was considered an example of chemical combination just like the compound obtained by oxidizing copper. This view, which regarded solutions as compounds, was extremely tenacious.
Moreover, the formation of compounds explained the homogeneity observed in solutions. For example, if oxygen and nitrogen were not combined in air but merely mixed, then the heavier gas, oxygen, should sink to the bottom. Dalton, who saw air as a mixture, could not satisfactorily explain the fact that oxygen did not sink to the bottom.
The assimilation of his atomic theory ultimately gave birth to an anomaly where none had existed before.21)
Some might say that the only difference between the chemists who regarded solutions as compounds and their successors was a matter of definition. In one sense, that may have been so. But that sense is not one that would make definitions into mere conventional conveniences.
In the eighteenth century, mixtures and compounds were not, and perhaps could not have been, completely distinguished by operational experiment. Even if chemists had sought such a test, they would have been seeking criteria that would make solutions into compounds. The mixture-compound distinction was part of their paradigm—part of the way they viewed their entire field of inquiry—and, though it did not take precedence over the accumulated experience of chemistry as a whole, it did precede any particular experimental test.
Yet while chemistry was being thought of in this way, chemical phenomena were exemplifying laws different from those that emerged from the assimilation of Dalton’s new paradigm. In particular, so long as solutions were still thought to be compounds, so many chemical experiments could not by themselves yield the law of fixed proportions. By the end of the eighteenth century it had become widely known that some compounds ordinarily had fixed proportions by weight of their constituent elements. For several kinds of reactions, the German chemist Richter had even noted more advanced regularities of the sort now included in the law of chemical equivalents.22) But, except in the case of recipes for preparation, no chemist made proper use of these regularities, and almost no one before the end of the eighteenth century thought to generalize them. Given obvious counterexamples such as glass or aqueous solutions of salt, generalization was utterly impossible without abandoning the theory of affinity and without reconceptualizing the scope of the chemist’s domain. By the end of the eighteenth century, in the famous debate between the French chemists Proust and Berthollet, the result became clear. The former insisted that all chemical reactions occurred in fixed proportions, while the latter opposed him, saying they did not. Both men had assembled persuasive experimental evidence for their views. Nevertheless, their arguments were bound to diverge from one another, and their debate held no prospect of reaching any conclusion.
For where Berthollet saw a compound that could vary in composition, Proust saw only a physical mixture.23) In this debate, neither experiment nor a change in convention by definition had any meaning. Like Galileo and Aristotle, the two men were fundamentally at odds.
This was the situation at the time when John Dalton was carrying out the research that finally led him to his famous chemical atomic theory. Until the last stage of that research, however, Dalton was neither a chemist nor interested in chemistry. Rather, he was a meteorologist dealing with the physical problems of the absorption of gases by water and the absorption of moisture by the atmosphere. Because he had been trained in a different field, and partly because of the research he himself had done in that field, he approached these problems with a paradigm different from that of the chemists of his day. In particular, he regarded mixtures of gases, or the absorption of gases in water, as physical processes in which affinity did not operate at all. Thus, for him, the observed homogeneity of solutions was a problem, but he thought it could be solved if he could determine the relative sizes and weights of the various atoms in his experimental mixtures. It was in order to determine these sizes and weights that Dalton eventually turned to chemistry, and from the beginning, within the limited range of reactions that he considered chemical, he assumed that atoms could combine only in one-to-one ratios or in other simple integral ratios.24) This natural assumption enabled him to determine the sizes and weights of the elementary particles, but it also made the law of fixed proportions a tautology. For Dalton, any reaction in which the constituents did not enter in fixed proportions was, ipso facto, not a purely chemical process. Once Dalton’s work was accepted, a law that could not have been established experimentally before his work became a fundamental principle that no set of chemical measurements could overturn.
As a result of this event, which may perhaps be the most complete example of a scientific revolution, the same chemical operations came to bear a relation to chemical generalization entirely different from the one they had had before.
Needless to say, Dalton’s conclusions were attacked in various quarters when they were first published. Berthollet in particular was never persuaded. Considering the nature of the subject, there was no reason why he should have been. But for the majority of chemists, Dalton’s new paradigm proved persuasive where Proust’s had been inadequate, because Dalton’s paradigm implied meanings broader and more important than a new criterion for distinguishing mixtures from compounds. For example, if atoms combine chemically only in simple integral ratios, then a reexamination of existing chemical data would reveal not only the law of fixed proportions but also examples of the law of multiple proportions. Chemists no longer said, for instance, that the two oxides of carbon contained 56 percent and 72 percent oxygen by weight, respectively. Instead, they expressed it by saying that one weight of carbon combined with 1.3 or 2.6 weights of oxygen. When the results of past experimental operations were recorded in this way, the ratio of 2:1 immediately stood out. The same relation appeared in analyses of many well-known reactions and of other new ones. Moreover, Dalton’s paradigm made it possible to assimilate Richter’s work and to generalize it completely. It also suggested new experiments, especially Gay-Lussac’s experiments on combining volumes, and such experiments yielded other regularities that chemists had never before dreamed of. What chemists took from Dalton was not a new experimental law but a new way of doing chemistry—he himself called it a “new system of chemical philosophy”—and its usefulness was proven so rapidly that only a few of the old-style chemists in France and England were able to resist it.25) As a result, chemists came to live in a world in which chemical reactions behaved quite differently from before.
As all this proceeded, another change, typical and very important, occurred.
Here and there, the numerical data of chemistry themselves began to change. When Dalton first combed through the chemical literature in search of data to support his physical theory, he discovered several records of reactions that suited it, but he could not avoid finding others that did not. For example, Proust’s own measurements of the two oxides of copper yielded an oxygen weight ratio of 1.47 to 1, rather than the 2 to 1 prescribed by atomic theory. Proust was precisely the sort of person who might have obtained Dalton’s ratio.26) In other words, he was an excellent experimenter, and his views on the relation between mixtures and compounds were quite close to Dalton’s.
But fitting nature to a paradigm is difficult. This is precisely why the puzzles of normal science are so challenging, and why measurements carried out without a paradigm so rarely lead to any conclusion. Chemists, therefore, could not simply accept Dalton’s theory on the basis of the evidence, because much of that evidence was still negative. Rather, even after accepting the theory, they had to go through a process of bringing nature into conformity with it, a process that ultimately took nearly a generation. When this work was accomplished, even the percentage compositions of well-known compounds had changed, and the data themselves had changed. This is the final sense in which one may say that after a revolution, scientists work in a different world.
“Notes”
1) The original experiment was George M. Stratton, “Vision without Inversion of the Retinal Image,” Psychological Review, IV (1897), 341-60, 463-81. A more recent review is found in Harvey A. Carr, An Introduction to Space Perception (New York, 1935), pp. 18-57.
2) See, for example, Albert H. Hastorf, “The Influence of Suggestion on the Relationship between Stimulus Size and Perceived Distance,” Journal of Psychology, XXIX (1950), 195-217; Jerome S. Bruner, Leo Postman, and John Rodrigues, “Expectations and the Perception of Color,” American Journal of Psychology, LXIV (1951), 216-27.
3) N. R. Hanson, Patterns of Discovery (Cambridge, 1958), chap. i.
4) Peter Doig, A Concise History of Astronomy (London, 1950), pp. 115-16.
5) Rudolph Wolf, Geschichte der Astronomie (Munich, 1877), pp. 513-15, 683-93. Particularly noteworthy is how Wolf’s interpretation made it difficult to explain these discoveries as a result of Bode’s law.
6) Joseph Needham, Science and Civilization in China, III (Cambridge, 1959), 423-29, 434-36.
7) T. S. Kuhn, The Copernican Revolution (Cambridge, Mass., 1957), pp. 206-9.
8) Duane Roller and Duane H. D. Rolle, The Development of the Concept of Electric Charge (Cambridge, Mass., 1954), pp. 21-29.
9) See the discussion in Section VII, and the books indicated by the references cited in note 9).
10) Galileo Galilei, Dialogues concerning Two New Sciences, trans. H. Crew and A. de Salvio (Evanston, Ill., 1946), pp. 80-81, 162-66.
11) Ibid., pp. 91-94, 244.
12) M. Clagett, The Science of Mechanics in the Middle Ages (Madison, Wis., 1959), pp.
537-38, 570.
13) [Jacques] Hadamard, Subconscient intuition, et logique dans la recherche scientifique (Conference faite au Palais de la Devouverte le 8 December 1945[Alencon, n,d.}), pp. 7-9. Although limited to new mathematical inventions, a far more detailed account is The Psychology of Invention in the Mathematical Field (Princeton, 1949), by the same author.
14) T.S.Kuhn, "A Function for Thought Experiments", in Melanges Alexander Koyre, ed.
R.Taton and I.B. Cohen, Herman(Paris).
15) A. Koyre,Etudes Galileennes (Paris, 1939), I, 46-51; "Galileo adn Plato", Journal of the History of Ideas, IV(1943), 400-428.
16) Kuhn, "A Function for Thought Experiments", in Melages Alexandre Koyre; see the quotation in note 14).
17) Koyre Etudes...., II, 7-11.
18) Clagett, op .cit, chaps.iv, vi, and ix.
19) N. Goodman, The Structure of Appearance (Cambridge, Mass. 1951), pp. 4-5. This passage needs to be quoted at greater length: “For example, if among the residents of Wilmington in 1947 only those people weighing 175-180 pounds, and all of them, have red hair, then ‘red-haired residents of Wilmington in 1947’ can be combined, in structural definition, with ‘residents of Wilmington in 1947 weighing between 175 and 180 pounds.’..... The question whether there ‘might be’ some person to whom only one of these predicates would apply has nothing to do with the matter.... Once we have determined that there is no such person.. fortunately, nothing further is at issue.
For the notion of a ‘possible’ case, a case that does not actually exist but might exist, is a very uncertain one.”
20) H. Metzger, Newton, Stahl, Borehaave et la doctrine chimique(Paris, 1930), pp.34-68.
21) Ibid., pp. 124-29, 139-48. On Dalton, see: Leonard K.Nash, The Atomic-Molecular Theory("Havard Case Histories in Experimental Science", Case4; Cambridge, Mass., 1950),pp. 14-21.
22) J. R. Partington, A Short History of Chemistry (2d ed. ; London, 1951), pp.
161-63.
23) A. N. Meldrum, "The Development of the Atomic Theory;(1)Berthollet`s Doctrine of Variable Proportions", Manchester Memoirs, LIV (1910), pp. 1-16.
24) L.K. Nash, "The Origin of Dalton`s Chemical Atomic Theory", Isis XLVII(1956), 101-16.
25) A. N. Meldrum, "The Development of the Atomic Theory: (6) The Reception Accorded to the Theory Advocated by Dalton", Manchester Memoirs, LV(1911), 1-10 26) On Proust, see "Berthollet`s Doctrine of Variable Proportions", Manchester Memoirs, LIV(1910), 8. A detailed history of the gradual changes in the determination of chemical composition and atomic weights remains to be written, but material helpful toward that end is contained in Partington,op.cit.
XI. The Invisibility of Revolutions
The Invisibility of Revolutions
We must still ask and answer how scientific revolutions come to an end. But before doing so, it seems necessary to make one final attempt to strengthen our conviction concerning the existence and character of revolutions. I have thus far tried to reveal revolutions by means of examples, and those examples could be multiplied ad nauseam. But clearly, since most of them were deliberately chosen because of their familiarity, they have ordinarily been regarded not as revolutions but as additions to scientific knowledge. Adding further examples would have made no difference, for whatever examples were added would lead to the same thought. I believe there are good reasons why revolutions prove to be almost invisible. Both scientists and laymen derive most of their image of creative scientific activity from authoritative sources that systematically disguise__sometimes for important functional reasons__the existence and significance of scientific revolutions. Only when the nature of that authority is recognized and analyzed can historical examples be expected to become fully effective. Furthermore, though this point can be fully developed only in the final section concluding my argument, the analysis required here will perhaps begin to present one of the characteristics that most clearly distinguishes scientific activity from all other creative intellectual pursuits, with the possible exception of theology.