THE DYNAMICS OF EXPERIMENTATION AND ITS ROLE WITHIN A PHILOSOPHY OF SCIENTIFIC PRACTICES José Ferreirós1 1) INTRODUCTION. It’s already been twenty-five years since Ian Hacking stated, provocatively, that philosophers should start to think about the adventure that began back in the 17th century. He meant modern science, of course – what was then termed «experimental philosophy». Hacking was intimating that the whole tradition in philosophy of science (including all 20th century proposals up until Lakatos, Laudan, and the semantic and structuralist conceptions, at least) is profoundly inadequate to analyse the scientific phenomenon. And the main reason would be that philosophers have not elaborated the tools for an adequate understanding of experimentation – nor therefore of its role in theory formation. During these 25 years, a small trend of “experimentalist” authors has grown, among them names such as Hacking, Franklin, Cartwright, and also historians like Heilbron, Galison, Buchwald, Steinle – not to forget a good number of sociologists such as Collins, Schaffer, Pickering, etc. In my opinion, the emergence of new experimentalism has been one of the most exciting recent development in the theory of science, if not the most fascinating. The reaction among the community of specialists and teachers of philosophy of science still seems disappointing to me, being so scarce as it is. But perhaps we can be confident that the situation is changing. Science’s old name, «experimental philosophy», suggests already that modern science can be regarded as a hybrid of philosophy (logic, theory, argument) and experiment (intervention, technics, observation). Of course presenting it this way involves a simplification, but I believe it constitutes a useful idealisation, if our purpose is to provide a rudimentary model of how the cognitive activities of scientists are structured. (Above all, such a model hides the role of interactions among members of the scientific community, and between them and other Dynamics of experimentation 2 social actors. However, the model can integrate social factors, especially if we understand that knowledge – data and models and theories – is a social product almost by definition.2) Following that suggestion, let me propose a triadic model of scientific activity based on considering three ‘phases’ in its cognitive processes, or three broad categories of scientific practices (which no doubt would have to be subdivided into finer types): theoretical activities, experimental activities, – subsuming here the particular case of observations – and also communicational activities. This in itself cannot sound new, but the key idea is to emphasize that the experimental ‘phase’ is not reduced or subordinated to the theoretical one, and that it calls for a deeper and novel analysis. Furthermore, both types of activities are in interaction, and the complexity of those interactions still defies philosophers of science. Such a scheme is very different from the views offered by traditional authors, which by following the linguistic turn and emphasizing logic were led to something like a model where there is one principal ‘phase’ of theory formation, merely punctuated by the injection of basic statements (corresponding to what are usually called observational data). What is characteristic of experimental and observational activities remained outside the philosopher’s analysis, be it because it was considered transparent (as with empiricism) or regarded as exasperatingly swampy (as Popper liked to say). In joint work with Javier Ordóñez, we have criticized such models for their theoreticism, and we have also traced the origins of this tendency back to philosophically inclined theoretical physicists such as Boltzmann.3 No doubt, betting for a philosophy of science that is able to analyse the experimental phase complicates matters for the aspiring philosopher, because it will force her to augment her panoply of tools. A fine analysis of the factors that enter into experimental activity should include questions belonging to the cognitive sciences; it cannot be reduced to a logical scheme, nor can it be treated in the style of the linguistic turn, and it is also insufficient to speak of “paradigms” or “values”. Considered from this angle, the misery of theoreticism stems from the way it reduced the richness and complexity of scientific procedures to an affair merely of conceptual and theoretical elaboration. A philosophy of science that closes its eyes to the epistemic specificity of experimental life will thereby renounce the goal of understanding what is most characteristic of scientific knowledge. Properly considered, this already offers an explanation for the peculiar situation 1 Dpto. de Filosofía y Lógica, Universidad de Sevilla, Camilo José Cela s/n, E-41018 Sevilla. <[email protected]> 2 Which does not mean (beware the non sequitur) that the analysis of knowledge can be merely sociological. 3 J. FERREIRÓS & J. ORDÓÑEZ, “Hacia una filosofía de la experimentación,” Crítica 34 (2002), nº 102, 47–86. Dynamics of experimentation 3 we saw towards the end of the 20th century, when the rationalism and faith in progress of philosophers was confronted head-on by strong sociological approaches. The views originating in theoreticist and “logicistic” approaches to the philosophy of science were, malgré lui, feeding the sociologism of the 1980s and 90s. That is because of the way they promoted losing sight of the processes by which data are obtained (produced?) in science. They promoted excessive simplification of our models of scientific practice, and also rigidly formalistic conceptions of human rationality.4 In my opinion, the “third way” that can take us out of that bog consists in a reflection upon scientific practices, understood not as an attractive yet void formula, but rather as the decision to fully consider the epistemic and cognitive specificity of scientific activities, and in particular experimental activities. The plural form is of the essence: there is not scientific practice in the singular, but a plurality of coexisting practices, and the crux of the analysis has to do with their heterogeneous cognitive roots and their complex interactions. It is for this reason that, as a first step and to counter their traditional oblivion, we must consider the roots and the dynamics of experimentation.5 The way opened by studies of the philosophy of experiment opens a promising course for navigating the waters between the Scilla of theoreticism and the Caribdis of sociological reductionism. 2) EXPERIMENT AND ITS INTERPRETATION: THE BASIC STRUCTURE. In traditional speech about data and experimental results, these are presented as ready-made elements emerging from some black box. The possible structural and dynamical complexities in the modus operandi of that black box, what I shall call the «processes of data formation», are not a focus of attention. Indeed, the traditional idea of empiricists is that we are actually talking of a «white box»: a transparent process of reception of impressions, which ends up in an automatic and infallible disposition to formulate basic statements. On a completely different line, we find Popper’s peculiar idea that basic statements are like pillars introduced from above (theories rule) into the “swamp” of observational and experimental work, pillars 4 The importance of formalistic rationalism as an stimulus for sociological conceptions is clear in the work of a central author like Harry M. Collins: see in particular his classic COLLINS, H. M. Changing order. Replication and induction in scientific practice (Univ. Chicago Press, 1992; original edition 1985). It is also abundantly known how simplified versions of the theses promoted by Popper, Lakatos, or Quine have been appropriated by specialists in STS or sociology of science. 5 Note that this formulation, properly understood, involves the theoretical phase, since both ‘phases’ are in almost constant interaction. But it emphasizes that which is still the least known and understood. Dynamics of experimentation 4 ultimately justified “by convention”.6 Although Popper never extracted the radical conclusions that this position is calling for, others (e.g. Lakatos) did. It will be worthwhile to pause for a note on terminology. Here, as in previous work, I follow the scientist’s usual way of talking when it comes to experimental and observational data. An alternative terminology has been proposed by Bogen & Woodward, who contrast “data” and “phenomena” with connotations that are fundamentally different from mine, as their “data” mean the fluctuating outcomes of particular experimental trials, while “phenomena” are the stable constructs which are theories are meant to predict and explain.7 So the reader should beware: data in my sense are the “phenomena” of Bogen, Woodward and others, which is why I emphasize the need to speak about processes of data formation. In detailed considerations, it is customary to think that the production of an experimental result involves at least three elements – a material procedure, an instrumental model, and a phenomenic model:8 A. The material procedure is a complex of objects and actions, or interventions, performed practically in the material world: arranging the apparatus and the specimens, and making them function properly (i.e., setting them to work in the proper sequence and controlling their performance). B. The instrumental model expresses a certain conceptual understanding on the side of the experimenter about how the apparatus works; this is central to the design, realization, and interpretation of the experiment. Such models can be of a highly theoretical and mathematical nature, but sometimes they depend on a modest amount of low-level theory. C. The phenomenic model codifies basic elements of the way in which the experimenter understands conceptually aspects of the phenomenal world that are under study; without it, the results would lack sense and meaning and could not be interpreted. And again, phenomenic models do not always depend on high theory.9 6 K. R. POPPER, Logik der Forschung, Tübingen, Mohr, 1935, chap. 5; English edition Logic of scientific discovery, Hutchinson & Co, 1959. Spanish version by V. Sánchez de Zavala, La lógica de la investigación científica, Madrid, Tecnos, 1962. 7 BOGEN, J. & J. WOODWARD, “Saving the Phenomena,” The Philosophical Review, vol. 97 (1988), pp. 303– 352. A good number of other philosophers, for instance Maricio Suárez, have adopted this peculiar terminology. 8 See e.g. pages 276–77 of A. PICKERING, “Living in the material world,” in GOODING, D., PINCH, T. J. & SCHAFFER, S. (eds), The uses of experiment. Cambridge Univ. Press, 1989. 9 Remember e.g. the «taxonomies» of Kuhn’s late work. Dynamics of experimentation 5 To present these ingredients concretely, giving a clear and simple example that we shall continue using in the sequel, – Newton’s famous experiments on the decomposition of sunlight – the three elements are as follows: A’. The material procedure includes the prisms (made of some or another kind of glass, sometimes filled with water), screens, procedures to modify the incident light (from simple holes on a window shutter, to lenses employed to colimate the light), etc. B’. The instrumental model is built upon an interpretation of the material procedure in terms of an antecedently established theory, geometrical optics, so that in this case it depends on high theory. (The model did not consider details about possible differences between different kinds of glass, and this was historically important.) C’. The phenomenic model is again formulated by means of geometrical optics, concretely by using the concepts of a ray of light and ideas about its behaviour upon reflection or refraction. The model assumed idealisations that are typical of geometrical optics, like ignoring the fact that shadows have fuzzy edges. On this last point I should add a clarification. You know of course that Newton was a corpuscularist, believing light to consist in tiny corpuscles travelling at great speed, and that he opposed the wave theories that had been formulated at the time. However, in his optical writings he made an effort to establish key theses – in particular the principle that simple light rays are associated with colours and have a characteristic refrangibility – on a basis that was neutral with respect to the physical theories in dispute. This is why his phenomenic model does not presuppose corpuscularism and is based on geometrical optics, by then a classic theory, well established at least among “mathematicians”. Traditional images of experimentation would suggest that, at the stage of justification, the material procedure and the instrumental model remain fixed and unaltered. Their features would be relatively natural and uncontroversial, both for the particular scientist who first proposed them, and for the scientific community that must replicate the experiments and judge the results. Meanwhile, the phenomenic model would be more flexible or “plastic”, since of course one allows for the possibility of competing theories defended by different scientists. Moreover, in what was traditionally (since the 19th century) presented as the prototype experiment, the main goal would be to measure in great accuracy some data to be contrasted with theoretical predictions, or perhaps some parameter fixed by theory (e.g. a physical constant, as a result of which the phenomenic model would be refined and specified to a greater level of precision). Dynamics of experimentation 6 But sociologists of science have challenged those assumptions, studying in detail cases where one finds the scientist showing almost no flexibility as regards the phenomenic model, but treating the other two components as very flexible indeed. Famous in this regard is Pickering’s work on what he called «the hunting of the quark», some experiments performed by Italian physicist Morpurgo during a period of 15 years. The studies of Harry Collins on the search for gravitational waves are also well known and have been celebrated.10 Pickering concludes that the three structural elements A. B. and C. are equally «plastic resources» that, far from being fixed and determined, can be modified at will until a result of coherence is attained. He believes that experimental work begins in such a way that there no definite relation between the structural ingredients: «incoherence and uncertainty are the distinctive seals of experiment», as shown abundantly by studies of laboratory life. But at the end of the day, some form of non-trivial coherence is obtained, a stabilization such that «material procedures, […] when interpreted through an instrumental model, produce facts within the framework of a phenomenic model».11 The analysis of such processes of interactive stabilization between the three structural elements constitutes what, in the sociologist’s perspective, would correspond to our dynamics of experimentation. Any conclusion we may finally extract about experimental activity, be it about its epistemic relevance, or say its dependence upon contextual factors, will obviously hang on the characteristics attributed to the structural ingredients, and to their interrelations. If Pickering’s position is somehow typical, the main point in dispute today would no longer be the «social construction» of experimental results, a conception of sociological reductionism that has been superseded by many promoters of social studies of science. But there remain the hot problems of the epistemic reliability of experimental data, the extent to which they provide information on natural processes, as opposed to the possibility of vicious circles,12 or a mere coherentist stabilization such as described by Pickering.13 All of this depends on whether the structural elements are “equally plastic” or not. 10 For discussion and questions about the details of the case studies offered by Collins y Pickering, see FRANKLIN, A., “Experiment in Physics,” in E. N. ZALTA (ed.), The Stanford Encyclopedia of Philosophy (Summer 2003 Edition), URL = http://plato.stanford.edu/archives/sum2003/entries/ physics-experiment/. See also the abridged Spanish version: FRANKLIN, A., “Física y experimentación,” Theoria 17 (2002) nº 44: 221– 242. An interesting exchange between Franklin and Collins took place in Studies in History and Philosophy of Science, 25 (1994) no. 3. 11 PICKERING, “Living in the material world,” in GOODING, D., PINCH, T. J. & SCHAFFER, S. (eds), The Uses of Experiment. Cambridge Univ. Press, 1989, 277–78. 12 COLLINS, H. M., Changing Order. Replication and induction in scientific practice, Univ. Chicago Press, 1992. 13 PICKERING, A., The Mangle of Practice, Univ. Chicago Press, 1995. Dynamics of experimentation 7 The coherentist thesis has been formulated again by Hacking, who speaks of a “self- vindication” of laboratory sciences, and presents the idea as a kind of expanded Duhem thesis.14 If correct, the thesis of Pickering and Hacking would have noteworthy consequences. It would be definitive confirmation of the «theory-ladenness» of results, certainly in the company of their correlative «technics-ladenness» and «social-ladenness», but forcing us to abandon as elusive or noumenical – to abuse of Kant’s terminology – any possible «nature- ladenness». Maybe the business of science would have its continuity and production of technological effects guaranteed (albeit one could not quite understand why), but from an epistemic point of view it would lack any special justification. Points like those are thus crucial to any conclusion with respect to the epistemic reliability of the whole scientific enterprise, hence to the project of a philosophy of science. After all Einstein, even during his period of greatest enthusiasm for the theoretical and mathematical components of science (and although he was willing to grant that “the creative principle resides in mathematics”), emphasized that “experience remains, of course, the sole criterion of the physical utility of a mathematical construction”. And some years later, Feynman would begin his lectures saying: The principle of science, the definition, almost, is the following: The test of all knowledge is experiment. Experiment is the sole judge of scientific ‘truth’.15 So, is that “metaphysical” idea, the concept of Nature, totally foreign to this game? 3) THE «EXPERIMENTER’S REGRESS». Let us come back to Newton’s famous experiments. Contrary to common lore, historians have established that his work on the composition of light, and in particular his experimentum crucis, were by no means an immediate success. Indeed, Schaffer has turned this case into another argument for the decisive influence of sociological factors in science’s decision making. The experimentum crucis was contested during some fifty years, mainly – but not only – due to the difficulty of replicating its quantitative results.16 A quick reading of the controversies suggests that Newton was arguing as follows: «simple» rays of light behaved 14 HACKING, I., “The self-vindication of the laboratory sciences,” in A. PICKERING (ed), Science as Practice and Culture, Univ. Chicago Press, 1992. 15 Quoted by A. FRANKLIN from the famous Feynman Lectures on Physics (Addison-Wesley Pub Co, 1963) in note 1 of “Experiment in Physics,” in E. N. ZALTA (ed.), The Stanford Encyclopedia of Philosophy (Summer 2003 Edition), op. cit. 16 See SCHAFFER, S., “Glass Works: Newton’s prisms and the uses of experiment,” in D. GOODING, T. J. PINCH & S. SCHAFFER (eds), The Uses of Experiment. Cambridge Univ. Press, 1989. Dynamics of experimentation 8 according to his statements, but this could only be detected using «good» prisms, and «good» prisms were those which produced «simple» rays. Thus the case is presented as a clear illustration of what Collins has termed the «experimenter’s regress», that menaces the epistemic reliability of experimental results. The experimenter’s regress of Collins consists in a vicious circle that stems from severe problems with the replication of experiments and the calibration of scientific instruments. The main problem is that correct results are only obtained using apparatus that functions properly, while the apparatus is functioning properly only if it provides correct results.17 So in the last analysis the outcomes of a scientific controversy do not depend so much on what “Nature” has to “say,” or on any special use of methods with some epistemic virtues, but on who is the experimenter in a social position of dominance, that enables her or him to determine what is correct and what functions properly. Collins offers as prototypical the case of J. Weber’s experiments on gravitational waves,18 but the example of Newton’s experiments, in Schaffer’s interpretation, is by no means worse. In this case, Schaffer argues that it was well into the 18th century (around 1715), when Newton enjoyed a position of extraordinary influence as President of the Royal Society, that he displayed diplomatic activities ending up in the promotion of his scientific views in France. I believe that this interpretation is incorrect, which incidentally shows that good history of science (such as Schaffer’s) is still not sufficient for an in-depth philosophical analysis. I shall now offer my own revision of the case, where the third kind of practices mentioned above (practices of communication) play an important role. * In good measure, the polemics generated by Newton’s work and his experimentum crucis were caused by himself, by what we might call a youthful error in his strategy of argument. Retrospectively one can locate the main error, not in anything Newton did while investigating the matter, but in the way he wrote his first published paper on natural philosophy.19 The ‘error’ was motivated by Newton’s great experience with mathematical texts, and his lack of experience in physical controversies. The young Newton believed that he could solve the 17 COLLINS, H. M., Changing Order. Univ. Chicago Press, 1992, chap. 4 and 5. 18 See COLLINS, Changing Order (1992) and FRANKLIN, “Experiment in Physics” (2003). Collins makes a lot of the statement that «there are no formal criteria» that could be applied to decide whether the instruments are functioning properly. On this topic, see footnote 5 above. 19 NEWTON, I. “A Letter of Mr. Isaac Newton, Professor of the Mathematicks in the University of Cambridge; containing his New Theory about Light and Colors,” Philosophical Transactions of the Royal Society 80 (1672), pp. 3075-3087. Available on the web, see URL = http://www.newtonproject.sussex.ac.uk/prism.php?id=47. Dynamics of experimentation 9 question in great brevity and full precision by writing more geometrico: two carefully planned experimental «demonstrations», together with a series of definitions and propositions, would suffice to convince his readers. (The reader should notice that modern protocols for doing and reporting experimental research only consolidated during the 19th century, while the millenary Euclidean style of writing mathematics remained paradigmatic for scientists throughout the 17th and 18th centuries.) Newton based all of his argument on two experiments, of which the first was extremely rudimentary, even though he offered some variations of the theme (one prism projecting an elongated image onto a screen), and the second was quite sophisticated, being offered in a single purportedly definitive version. This «experimentum crucis» employed two prisms and two screens by means of which a monochrome ray of light was isolated, with results that were meant to stamp the key proposition that sunlight consists in a mixture of coloured rays with different specific refrangibilities. The famous “crucial experiment” was thus made to support all of the weight of proof, single-handedly. Both experiments turned out to be difficult to replicate, again in large measure because of their “mathematical” character, i.e., and perhaps surprisingly to the reader, precisely because the results were quantitative. The elongated image that Newton obtained in Cambridge was a palette of colours (an artificial rainbow) enlarged by a factor of 5, but the Jesuit Antoine Lucas working in Liège only obtained an elongation by a factor of 3. Was this perhaps because sunlight is different in both locations? When Lucas published his discrepancy in the Philosophical Transactions, Newton took it very seriously as an offence to his honesty as a gentleman and his reliability as a reporter of observed physical phenomena. The discrepancy was relevant in the context of discarding alternative explanations of the result, by an argument relying on Snell’s law of refraction. Therefore Newton felt an imperious need to attack his opponent and annihilate him. Quite unfortunately, it never occurred to him that they could be confronting a real problem caused by differences in the nature of the prism’s glass. As the crystalline composition of glasses produced in different places and factories differed greatly, it was plainly naïve to expect standardized quantitative results as an outcome! Another source of difficulties was made manifest by a highly reputed French experimenter, Edme Mariotte, founding member of the Académie des Sciences. Around 1680, Mariotte set out to reproduce the supposedly “crucial” experimentum crucis, finding results that he interpreted to contradict and even refute Newton. Having isolated a “simple” violet Dynamics of experimentation 10 beam of light, he obtained after the second refraction tones of red and yellow colouring both ends of the violet image.20 What the English was inclined to consider an understandable imperfection of the experimental setting, was interpreted by his much more empiricistic colleague as a very clear contrary result, a vindication of the old theory of the modification of light by the prism (that Newton was intent on refuting). It was “evident” that in this experiment a ray of light of the kind that Newton called «simple» had been modified or shown to be complex. Given Mariotte’s deserved reputation as an experimenter, this episode brought as a result a very long delay – almost 40 years – for the acceptance of Newton’s theory in France and other places. This time the discrepancy between both actors can be located in their instrumental and phenomenic models, or more precisely in what we may term – following Hempel – the «bridge principles» necessary for Newton’s interpretation of the results. Their discrepancy measures the conceptual distance between the simple ray promoted in geometrical optics, hence in Newton’s models, and the concrete beam of light that the experimenter was able to isolate. In a sense, the epistemic character of modern science was at stake: whether it was to be crudely empirical, based directly upon the observed in the style of Mariotte, or inextricably linked with mathematical idealisations, as Newton advocated. In the latter’s opinion, the study of Nature had to be mathematical, and the narrowest beam obtained by an experimenter was, self-evidently, very far from the «simple» ray in the model. The corresponding adjustments were more than enough to explain away Mariotte’s observations. Such incidents show the enormous difficulties encountered by scientific research in its infancy, and make us wonder how it was possible to obtain any clear advances given all the material and technical difficulties: inexistence of standardized instruments, lack of experimental protocols, unreliability of the practical and intellectual training on the side of the savants. Little wonder that, if you wish to look for rhetorical elements in writings and letters from the time, you will find plenty of material that can be used for the conclusion that Collins’ regress was fully in action, that the dispute was impossible to close except by an appeal to politico-diplomatic operations. A clear example of social construction and negotiation, it seems. 20 See GUERLAC, H. Newton on the Continent, Cornell Univ. Press, 1981, pp. 98–99. Mariotte’s work appeared as a book: E. MARIOTTE, De la nature des couleurs (Paris, 1681), which can also be found in his Oeuvres, vol. 1 (Leiden, 1717).