by Hasok Chang
How enrichment works can be illustrated in the following true story. Driving along the scenic Route 2 in Western Massachusetts on the way from Boston to my alma mater, Northfield Mt. Hermon School, I used to spot a wonderfully puzzling road sign that announced: "Bridge St. Bridge." It seemed to designate a bridge named after the street that was named after the bridge itself. For a few years I passed this sign every so often and wondered how that name ever came to be. It finally occurred to me that I could do some research in local history to find out the story, but from that point on I could not find that sign again, nor remember which town it was in. Nevertheless, I did arrive at the following plausible historical hypothesis. Initially the town was so small that there was only one bridge (referred to as "The Bridge") and no named streets. Then came enough streets so they had to be named, and the street leading to the bridge was named "Bridge Street," naturally. Then came other bridges, necessitating the naming of bridges. One of the easiest ways of naming bridges is by the streets they connect to (as in the "59th Street Bridge" celebrated in the Simon and Garfunkel song also known as "Feelin' Groovy"). When that was done our original bridge was christened: Bridge Street Bridge! If my hypothesis is correct, the apparent circular nonsense in that name is only a record of a very sensible history of iterative town development.
We have seen iterative enrichment in action first of all in the process of quantifying the operational temperature concept, analyzed in chapter 1. Initially the judgment of temperature was only qualitative, based on the sensation of hot and cold. Then came thermoscopes, certified by their broad agreement with sensations; thermoscopes allowed a decisive and consistent comparison and ordering of various phenomena possessing different temperatures. Afterwards numerical thermometers, arising iteratively from thermoscopes, went further by attaching meaningful numbers to the degrees of hot and cold. In this developmental process temperature evolved from an unquantified property to an ordinal quantity, then to a cardinal quantity. Each stage built on the previous one, but added a new dimension to it. A very similar type of development was seen in the process of creating temperature standards in the realm of the very hot and the very cold in chapter 3. The chief epistemic virtue enhanced in these processes can be broadly termed "precision." However, this story also reveals the complexity of what we casually think of as precision. Going from purely qualitative to the quantitative is certainly enhancement of precision, but so is the move from the ordinal to the cardinal, as is the increase in numerical precision after the move into the cardinal. All three of these enhancements occurred in the invention of the thermometer as we know it.
A different aspect of iterative enrichment was also seen in the extension of the temperature scale discussed in chapter 3. After the establishment of the numerical temperature concept by means of thermometers operating in a narrow band of temperatures, that concept was extended to previously inaccessible domains by means of thermometers capable of withstanding extreme conditions. The establishment of a temperature standard in a new domain shows the familiar process of going from the qualitative (this time based on simple instrumental operations, rather than pure sensation) to the ordinal to the cardinal, in that new domain. An
end p.229
overview of the whole temperature scale gives us a picture of iterative extension, in which the concept in the initial domain is preserved but augmented into new domains. The chief epistemic virtue enhanced in this process is scope. (An extension of scope also occurred already in the replacement of sensation by thermoscopes, discussed in "The Validation of Standards" and "The Iterative Improvement of Standards" in chapter 1.) These are just some illustrative examples of iterative enrichment. I predict that many others will be seen if we examine other areas of scientific progress in the same vein.
Self-Correction
The other major aspect of iterative progress, self-correction, can also be illustrated at first with a tale taken from everyday life (with a slight exaggeration). Without wearing my glasses, I cannot focus very well on small or faint things. Therefore if I pick up my glasses to examine them, I am unable to see the fine scratches and smudges on them. But if I put on those same glasses and look at myself in the mirror, I can see the details of the lenses quite well. In short, my glasses can show me their own defects. This is a marvelous image of self-correction. But how can I trust the image of defective glasses that is obtained through the very same defective glasses? In the first instance, my confidence comes from the sensible clarity and acuity of the image itself, regardless of how it was obtained. That gives me some reason to accept, provisionally, that certain defects in the glasses somehow do not affect the quality of the image seen (even when the image is of those defects themselves). But there is also a deeper layer in this mechanism of self-correction. Although at first I delight in the fact that my glasses can give me clear, detailed pictures despite its defects, on further observation I realize that some defects do distort the picture, sometimes recognizably. Once that is realized, I can attempt to correct the distortions. For example, a large enough smudge in the central part of a lens will blur the whole picture seen through it, including the image of the smudge itself. So when I see a blurry smudge on my left lens in the mirror, I can infer that the boundaries of that smudge must actually be sharper than they seem. I could go on in that way, gradually correcting the image on the basis of certain features of the image itself that I observe. In that case my glasses would not only tell me their own defects but also allow me to get an increasingly accurate understanding of those defects.
In chapter 2 we saw how an initial assumption of the correctness of a certain type of thermometer could defeat itself, by producing observations that show the lack of comparability between different individual thermometers of that same type. In that episode all that was possible was falsification rather than positive correction, but in "The Validation of Standards" and "The Iterative Improvement of Standards" in chapter 1 we saw how a later standard initially based on a prior standard could proceed to overrule and correct that prior standard. This can be regarded as a self-correction of the prior standard, if we take the later standard as an evolved version of the prior standard. In chapter 4 we saw various instances of self-correction, most clearly exhibited in the Callendar-Le Chatelier method of operationalizing the absolute temperature concept, in which the initial assumption that actual gases
end p.230
obeyed the ideal gas law was used in the calculation of their deviation from it (see "Accuracy through Iteration" for further details). In his critique of Heinrich Hertz's formulation of mechanics, Simon Saunders has elucidated a very similar process for the improvement of time measurement; in short, "the accuracy of the time-piece must itself be revisable in accordance with dynamical principles." Saunders points out that: we have only an iterative process, and it is only guaranteed to yield determinate results given the assumptions of the theory, including estimates on interplanetary matter densities and the rapid fall-off of tidal effects of stellar objects. … [T]he notion of 'test' in mechanics is in large part a question of consistency in application under systematically improving standards of precision.7
I think many more examples will be found if we make a careful survey of the exact sciences.
Tradition, Progress, and Pluralism
I have articulated in this chapter a particular mode of scientific progress: epistemic iteration in the framework of coherentism. Before closing the discussion of coherentist iteration, I would like to add a few brief observations pertaining to the politics of science. Almost by definition epistemic iteration is a conservative process because it is based on the principle of respect, which demands the affirmation of an existing system of knowledge. However, the conservatism of iteration is tempered by a pervasive pluralism. There are several aspects to this pluralism.
First of all, the principle of respect does not dictate which system of knowledge one should affirm initially. One can certainly start by affirming the currently orthodox system, as the modern climate in science enc
ourages. However, even orthodoxy is a choice one makes. Nothing ultimately forces us to stay entirely with the system in which we have been brought up. In Kuhn's description of "normal science" it is assumed that there is only one paradigm given to the scientists within a given scientific discipline, but I think it is important to recognize that orthodoxy can be rejected without incurring nihilism if we can find an alternative pre-existing system to affirm as a basis of our work. That alternative system may be an earlier version of the current orthodoxy, or a long-forgotten framework dug up from the history of science, or something imported from an entirely different tradition. Kuhn has argued persuasively that paradigm shifts occur as a consequence of adhering to the orthodox paradigm and pushing it until it breaks, but he has not argued that sticking faithfully to a paradigm is the only legitimate way, or even the most effective way, of moving on to a different paradigm.
The affirmation of an existing system does not have to be wholesale, either. Scientists do often adopt something like a Kuhnian paradigm that consists of an entire "disciplinary matrix," and such an inclination may also be a pervasive trend
7. See Saunders 1998, 136-142; the quoted passages are from 137 and 140 (emphasis original). I thank an anonymous referee for Oxford University Press for pointing out this work to me.
end p.231
in professionalized modern science. However, affirmation does not have to be so complete, as we can in fact see by going back to Kuhn's own original meaning of "paradigm": an exemplary piece of work that is widely emulated, an achievement that "some particular scientific community acknowledges for a time as supplying the foundation for its further practice" (Kuhn 1970c, 10; Kuhn 1970b, 271-272). The emulation of an exemplar will not necessarily generate a comprehensive disciplinary matrix. Also the emulation of the same exemplar by different people can lead to different outcomes, in the absence of strict educational and professional enforcement. Even when there are communities that adhere to paradigms in the more complete sense, anyone not belonging fully in such a community has no obligation to adopt all elements of its dominant paradigm. And if there are competing communities that study the same subject, then the individual has the option of creating a coherent hybrid system to affirm.
It is also possible to choose the depth of affirmation. For example, if there is a sufficient degree of despair or disillusionment about all existing systems of knowledge, scientists may decide to reject all systems that are overly developed and start again with the affirmation of something that seems more basic and secure. This is how we should understand most cases of phenomenalism or positivism when it appears in the actions of practicing scientists, as it often did in nineteenth-century physics and chemistry. Disillusioned with overly intricate and seemingly fruitless theories about the microphysical constitution and behavior of matter, a string of able scientists (e.g. Wollaston, Fourier, Dulong, Petit, Carnot, Regnault, Mach, and Duhem) retreated to more observable phenomena and concepts closely tied to those phenomena (cf. "Regnault and Post-Laplacian Empiricism" in chapter 2). But even the positivists, to the extent they were practicing scientists, started by affirming existing systems, just not such metaphysically elaborate ones.
Finally, the affirmation of an existing system does not fix the direction of its development completely. The point is not merely that we do not know which direction of development is right, but that there may be no such thing as the correct or even the best direction of development. As noted in "Making Coherentism Progressive," the desire to enhance different epistemic virtues may lead us in different directions, since enhancing one virtue can come at the price of sacrificing another. Even when we only consider the enhancement of one given epistemic virtue, there may be different ways of achieving it, and more than one way of achieving it equally well. We are often led away from this pluralistic recognition by an obsession with truth because mutually incompatible systems of knowledge cannot all be true. The achievement of other virtues is not so exclusive. There can be different ways of enhancing a certain epistemic virtue (e.g., explanatory power or quantitative precision in measurement) that involve belief in mutually incompatible propositions. Generally speaking, if we see the development of existing knowledge as a creative achievement, it is not so offensive that the direction of such an achievement is open to some choice.
All in all, the coherentist method of epistemic iteration points to a pluralistic traditionalism: each line of inquiry needs to take place within a tradition, but the researcher is ultimately not confined to the choice of one tradition, and each tradition can give rise to many competing lines of development. The methodology
end p.232
of epistemic iteration allows the flourishing of competing traditions, each of which can progress on its own basis without always needing to be judged in relation to others. This pluralism should be distinguished clearly from any reckless relativism. In an amorphous type of coherentism, any self-consistent system of knowledge is deemed equally valid; in contrast, the coherentism I advocate is driven by the imperative of progress, so each tradition is continually judged by its record of enhancing various epistemic virtues. Even Feyerabend (1975, 27), with his "anarchism" in epistemology, was concerned with assessing systems of knowledge in terms of progress made, though he declined to define "progress" explicitly. In the coherentist framework of iterative progress, pluralism and traditionalism can co-exist happily. It is doubtful whether the intellectual and social constraints governing specialist communities of professional scientists can allow a full functioning of the freedom inherent in epistemic iteration. However, not all scientific activity is subject to those constraints. This last insight provides the basis of my kind of work in history and philosophy of science, as I will elaborate further in chapter 6.
The Abstract and the Concrete
The abstract insights summarized and developed in this chapter arose from the concrete studies contained in earlier chapters, but they are not simple generalizations from the few episodes examined there. It would be foolish to infer how science in general does or should progress from what I have seen in a small number of particular episodes, all from the same area of science. The problem of generalization continually plagues the attempts to integrate the history of science and the philosophy of science, so I think at least a few brief remarks are in order here outlining my view on how that problem can be dealt with.
I agree with Lakatos (1976) as far as saying that all historiography of science is philosophical. Abstract ideas emerge from concrete studies because they are necessary ingredients of narratives. Abstract ideas are needed for the full understanding of even just one concrete episode; it is a mistake to think that they can be eliminated by a conscientious avoidance of generalizations. We cannot understand actions, not to mention judge them, without considering them in abstract terms (such as "justified," "coherent," "observation," "measurement," "simple," "explanation," "novel," etc., etc.). An instructive concrete narrative cannot be told at all without using abstract notions in the characterization of the events, characters, circumstances, and decisions occurring in it. Therefore what we do when we extract abstract insights from a particular episode is not so much generalization as an articulation of what is already present. It may even be an act of self-analysis, in case our episode was initially narrated without much of an awareness of the abstractions that guided its construction.
In each episode studied in this book I have been asking abstract questions. Most generally put, the recurring questions have concerned the building and justification of human knowledge: How can we say we really know what we know? How can we come to know more and better than we did before? The asking of these questions requires an abstract conception of "justification" and "progress," even if the intended answers are only about particular episodes. The questions retain their
end p.233
abstract character even when they are narrowed down to particular epistemic values, because those values are also abstract.
In this chapter I have advanced
one central abstract idea: epistemic iteration is a valid and effective method of building scientific knowledge in the absence of infallible foundations. But I have not proposed anything resembling "the scientific method." The ideas I advanced in this chapter are abstract, but they are not presumed to have universal applicability. What I have attempted is to identify one particular method by which science can progress, and discern the circumstances under which that method can fruitfully be applied. That does not rule out other methods that may be used as alternatives or complements. It is important not to confuse the abstract and the universal. An abstraction is not general until it is applied widely.
An abstract idea needs to show its worth in two different ways. First, its cogency needs to be demonstrated through abstract considerations and arguments; that is what I have started to do in this chapter for the idea that epistemic iteration is a good method of progress when there are no firm foundations to rely on. Second, the applicability of the idea has to be demonstrated by showing that it can be employed in the telling of various concrete episodes in instructive ways. The idea of epistemic iteration helped me in understanding the possibility of scientific progress in each of the historical episodes explored in this book. Much more concrete work is needed, of course, for getting a good sense of the general extent of its applicability.
end p.234
6. Complementary Science—History and Philosophy of Science as a Continuation of Science by Other Means
Abstract: This chapter presents an in-depth discussion of the aims and methods of complementary science. Complementary science is presented as a productive direction for the fields of history and philosophy of science, without denying the importance of other directions. It can trigger a decisive transformation in the nature of scientific knowledge by adding a body of knowledge that combines a reclamation of past science, renewed judgment on past and present science, and exploration of alternatives.