Carlo Gabbani and Marc Lange 1 on Bas van Fraassen s Scientific Representation 2

Transcription

1 Carlo Gabbani and Marc Lange 1 on Bas van Fraassen s Scientific Representation 2 The Subject of Scientific Representation 3 Carlo Gabbani 1. Scientific Representation: Paradoxes of Perspective is an extensive volume made up of four parts as well as four appendices. The first part is dedicated to a general reflection regarding what a representation is; the second offers an epistemological analysis of the problems connected to scientific observation with instruments and to the coordination of mathematical formalisms to experience, paying particular attention to measurement processes; the third part presents an in-depth historical and theoretical discussion on the various forms of structuralism in the philosophy of science; the fourth and final part discusses, instead, different ways of conceiving the relationship between appearance and reality, as well as the pretence to obtain a criteria in order to explain how the appearances which we observe are produced starting from the primitives introduced by scientific theories. Many aspects of this work are worthy of discussion: from the proposal for a empiricist structuralism, to the interpretation of experimental observation outlined in chapter IV: an interpretation which, although it could be questioned, appears to be one of the most profound attempts to account for the possibilities of experience which have been opened up by modern technological instrumentation, in a way which is compatible with scientific antirealism. As it is impossible to make an overall summary of this rich and composite volume, I would like to limit myself in such circumstances to propose a preliminary reconnaissance and summary of some of the ideas that the volume offers regarding the relationship between scientific representations and the conditions, experiences, knowledge and practices of the subjects that make use of them. 1 Aldo Giorgio Gargani accepted with enthusiasm to participate in this debate on Scientific Representation. Unfortunately when he passed away on the 18th June 2009, he hadn t yet been able to let us have his text. The unbridgeable void that his absence leaves also in this small undertaking may be a tribute to the value of his presence [C.G.]. 2 * B. C. van Fraassen, Scientific Representation: Paradoxes of Perspective, Oxford: Oxford University Press, The cross-references in the body of the text refer to the volume under discussion. I thank Paolo Parrini for his reading over of a previous version of this text, of which I am solely responsible. Iris, issn , II, 3 April 2010, p Firenze University Press

2 246 Carlo Gabbani and Marc Lange As already mentioned, in the first part of the book an ample analysis of the representation is developed. It would seem important to focus, above all, on two aspects (see in particular pp ). [1] It is argued that the capacity of our constructs, abstract or concrete, to be valid as a representation can never be understood by limiting ourselves to considering only their internal characteristics, or the mere properties that they would have independently of the context and the role which is assigned to them by someone. As such, an explanation on the matter will not be adequate if it does not take into consideration, for instance, the essential importance that the factors linked to use have, as well as the cognitive or manipulative practices and the aims that, in a given context, lead someone to use a particular artifact with the role of a representation (selective and finalized) of this and that phenomena. [2] At the same time, the book underlines that the affirmation according to which a given artifact represents something would always be an abbreviated and unsaturated way to say that it represents something in that way. This is because the representative relationship can never be had in an absolute way, or simpliciter.. As such van Fraassen concludes that There is no representation except in the sense that some things are used, made, or taken, to represent some things as thus or so (p. 23). Therefore, there would not be representations if it were not for the fact that intentional subjects assume, in a specific context, that a certain construct of theirs is adequate in order to represent something else in a certain way, for certain purposes, or in the context of certain practices. Not by chance, this work highlights how representations can mean both a product as well as an activity (see p. 7). We must then also say that this context of activity in which a representation really becomes what it is, is explicitly understood as having characteristics that are genuinely and irreducibly intentional (see p. 23). We therefore understand how in this light the role of the subjects has a central importance: subjects, who make use of the representations and that, more precisely, exactly in doing so give certain constructs the value as representations of phenomena of their own experience. It is precisely for this reason, van Fraassen argues (see p. 24), that representations in nature, that is to say entities that naturally, in themselves constitute representations of something else, will not exist. Also the decisive process of coordination, that is, the process by which a system of abstract formalism is referred to physical objects and then assumes an empirical content, is considered possible and comprehensible only in relationship to a context characterized by specific conditions, procedures and finality. According to van Fraassen, indeed, this contextual (and indexical) restraint will play a central role: both as we tackle the most basic theoretical issues (see ch. XI) linked to the question: How can an abstract entity, such as a mathematical structure, represent something that is not abstract, something in nature? (p. 240); and when considering the problem of coordination more specifically in terms of institution of a particular coordination between certain theoretical structures and certain phenomena (see ch. V). In the latter case, for example, it is only within a framework in which coordinated links, experimental apparatus and procedures are already present, that even choices regarding the correlation and identification (within the theory) between certain properties and certain parameters would become possible and auspicious (see p. 137). It would

3 Bas van Fraassen s Scientific Representation 247 instead be illusive, according to van Fraassen, to expect to resolve the problem of coordination in absolute terms, by looking for justifications in natural and a-historical correlations between certain physical objects and certain abstract structures (see pp ). Therefore, an understanding of the process that we are interested in can only be obtained by moving from the idea of having a 2-place relationship between models and phenomena, to a 3-place relationship that includes also the person who makes use of such formal models (see pp. 253, 258). The crucial role of use and practice (p. 189), then, emerges and is explicitly recognized as the leitmotiv that brings together the considerations developed on the different aspects of scientific modeling. It is noteworthy how this pragmatic turn which van Fraassen articulates in relationship to the scientific representation seems to have at least some affinity with that which, over the past decades, he has defended in the field of the theories of scientific explanation. Van Fraassen had indeed proposed to interpret the explanation not as a mere dyadic relationship between a theory and facts, but rather as a three-term relation, between theory, fact, and context, 4 namely always as something which is linked to the demand, to the finality and to the interests of men. The first reviewers have, however, also pointed out a possible tension between the classic formulation of the constructive empiricism of van Fraassen in semantic terms, such as truth and empirical adequacy, and this work which emphasizes the dimension of the use of models and representations, based on different notions. 5 At the same time, it is possible that the author doesn t really intend to take leave of the previous formulations (which are still used by him in order to outline his very stance, see pp. 3, 87, 250, ), but rather to enrich, and in a certain sense reinterpret them, with particular attention from within especially regarding the contextual and pragmatic aspects. 6 Aspects which are indeed essential in order to apply those models that the theory proposes as accurate representations of certain phenomena to whom accept the theory as empirically adequate (and therefore as true if it regards only observables). But the effective composition of these two sides is a question worthy of further analysis. In any case, the existence of representations can be linked in another sense to the activities and to the context of the subjects. Indeed, van Fraassen claims, in order for an artifact to be used as a representation of empirical phenomena by a subject, the subject has also to complete the given construct with the knowledge that such a construct cannot on its own already include, yet which is essential to its very functioning. In 4 B. van Fraassen, The Scientific Image, Oxford: Oxford University Press, 1980, p. 156; see Id., The Pragmatics of Explanation, American Philosophical Quarterly, 14 (1977), pp See: R. Giere, Essay Review: Scientific Representation and Empiricist Structuralism, Philosophy of Science, 76 (2009), pp ; G. Contessa, Review of Bas van Fraassen s Scientific Representation: Paradoxes of Perspective (OUP 2008), Notre Dame Philosophical Reviews, (on-line). 6 On the theme of truth see also: B. van Fraassen, Replies, in A. Berg-Hildebrand and Ch. Suhm (eds.), Bas C. van Fraassen. The Fortunes of Empiricism, Frankfurt am Main: ontos verlag, 2006, 5.

4 248 Carlo Gabbani and Marc Lange particular (see chs. III and XI), in order for a scientific model to be used the addition of indexical knowledge would seem to be necessitated: knowledge that the users of that model cannot find within it. What is intended by this is that a necessary condition of possibility of use (p. 257) for a scientific model is that this model is completed by the subjects who use it with knowledge regarding their own setting, the perspective and the orientation that they have. In more general terms it is, therefore, necessary to situate oneself in relationship to a scientific model, so that it may be successfully referred to our world and not remain a sort of mute map: For to use a theory or model, to base predictions on it, we have to locate ourselves with respect to it [ ] When using a model to find our way around in the world we have to be able to say, for example, that the phenomenon we are presently witnessing is classified in the theory as oxidizing, or as phlogiston escape, or the like. We have to locate our situation in the theory s logical space, in a way that is similar to our We are here with respect to a map (p. 261). We could certainly, for example, decide to create a meta-model which, in an opportune working context, would also allow us to describe the collocation of the subject in the region modeled by the representation that is used by it. But this will as such make necessary another act of self-location by the user in respect to the metamodel so constructed (see p. 79). 7 The first relevant aspect of these considerations is that the perspective that van Fraassen begins to highlight with such an approach deliberately clashes with the thesis that the scientific representation of the world could constitute a view from nowhere (Thomas Nagel). In fact it is affirmed that, if also the theoretical models used by scientific theories can be in themselves independent of every observer and of every situated determination, the things will change when we wish to make practical use of them in order to take into account the phenomena of our experience: The physical world picture in abstracto is as far removed as possible from this use, it embodies, in Eddington s words, the view of no one in particular. But to put this picture to use, something must be done by the user, and this is where choice of reference frames comes in. Hence Weyl s words are equally apt when he refers to coordinate systems as the unavoidable residuum of the ego s annihilation. (p. 71, see p. 87). These observations can also be directly connected to a second question. Indeed, what would appear to emerge here is how we always need something in addition to what science has given us (p. 83), namely how, precisely in order to effectively coordinate in practice scientific models to phenomena, we must always add some extra knowledge respect to that which they themselves can give us. In this case, it will be real knowledge concerning our positioning. And, van Fraassen adds, whatever interpretation we give of such information we will find ourselves in front of a problem of relevant epistemological interest: if we say that the self-ascription is a simple, objective statement of fact, then science is inevitably doomed to be objectively incomplete. 7 Van Fraassen also links (pp ) this point to ideas developed by Kant from the essay Concerning the ultimate ground of the differentiation of directions in space. It is not possible to discuss these parallels here. I will limit myself to adding that Alfred Schutz referred to the analogy of the map in order to analyze the use of knowledge concerning the social world in: The Stranger: An Essay in Social Psychology, American Journal of Sociology, 49 (1944), pp

5 Bas van Fraassen s Scientific Representation 249 If instead we say it is something irreducibly subjective, then we have also admitted a limit to objectivity, we have let subjectivity into science. (p. 83). In other words, the approach of van Frasseen reminds us that our use of any scientific representation of the world (no matter how evolved it is) would appear to have certain insurmountable epistemic preconditions that it could never incorporate, if not by creating the need for other analogous preconditions on which to base itself. Freely paraphrasing the noted and highly discussed political-juridical Diktum of Ernst- Wolfgang Böckenförde, it would mean that mature scientific knowledge is based on epistemic preconditions, which it can t guarantee itself. 2. It is worth dwelling a little more upon the questions of the completeness and incompleteness of scientific knowledge and upon the relationship between subjective and objective therein. This theme is indeed proposed on several occasions by van Fraassen with the question: Is there something that I could know to be the case, and which is not expressed by a proposition that could be part of some scientific theory? (pp. 210 and 261). As already mentioned, for van Fraassen the answer is that this something does exist, and it is the knowledge concerning our position in respect to the map-model, expressed by indexical propositions. At the same time, however, plausibly this is not a question of thinking that information like that relative to our position is, as far as its content is concerned, of a special kind, or that it has such a nature so as to exceed by itself the cognitive possibilities of scientific knowledge. Van Fraassen himself recognizes how in fact this gap implies no relevant incompleteness in the theory or model itself (p. 261). That is to say, that this does not imply that scientific theories are, for this reason only, lacking in the representation of some types of phenomenon that it should contain: like a road map is not in itself lacking because it cannot furnish its user with enough information concerning his position. Simply the point is that it is always necessary to presume an act of extra-scientific self-location in order to avoid an endless regression. But, in itself, the fact of the location of a subject in a certain space is something that a scientific model is able to express, once that it has been coordinated to the phenomena of our experience. We could, however, also ask ourselves if there are not more radical and relevant forms of incompleteness of scientific-experimental knowledge. That is to say, if there aren t types of facts that are indeed both knowable, but at the same time cannot and neither could belong to the number of facts described and explained by scientific knowledge: not even when scientific knowledge has been developed and coordinated with experience. We ask ourselves, that is, if knowledge does not exist that, for the type of content it has, cannot find an equivalent formulation in terms of the type of vocabulary that does belong to the experimental sciences of nature. As is well known, this regards a question which had a very important role in the epistemology of the 20th century and then especially in the recent philosophy of mind. Our intention here is not to enter into questions regarding the philosophy of mind (even if in van Fraassen s book interesting references are made to this, precisely regarding this point: see pp and ). We will limit ourselves to pointing out how the perspective on science that van Fraassen outlines, here and in other works of his, may contribute to a significant transformation in respect to the most

6 250 Carlo Gabbani and Marc Lange common ways for considering this topic nowadays. This is also to say that, if the questions that he tackles rarely necessitate considerations concerning the philosophy of mind, it seems however that, implicitly, they can have an impact upon it (and, in this sense a significant possibility, which still needs to be developed, remains: that of articulating a philosophy of mind in the light of an anti-realist understanding of the scientific undertaking, like that of van Fraassen). We could have an initial confirmation of this, by returning to the question that has been brought up above. At the basis of the belief by which all of our effective knowledge could be part of a scientific-experimental type of knowledge there are often epistemological presuppositions that are not always discussed in an explicit way. Indeed, the thesis by which all of our knowledge would be or could be reduced to a scientific type of knowledge is not based on the exhibition of an exhaustive and self-sufficient scientific knowledge that has already been attained. Or, in an even more limited way, it is not based on the fact that all the non-scientific knowledge concerning a certain field (the conscious mind, for instance) can already be expressed in an equivalent scientific version. Rather, as has frequently been pointed out, the sustainers of this generally tend to link such a thesis to programmatic and controversial convictions, according to which the belief that science can, at least in principle, 8 arrive at providing us with a true and all-inclusive account in all fields would be intrinsic to the very logic of scientific knowledge, or connected to the only vision of reality which is compatible with what science teaches us. It is from this perspective, that it has been possible (let s consider, for example, the so-called eliminativism à la Churchland) even to arrive at assuming that either all of that which we judge to be non-scientific knowledge will end up by being absorbed by science itself, or, far from declaring science to be incomplete, we would rather have to recognize that which we considered as non-scientific knowledge, to be fallacious beliefs: and as such, if irreducible, irreducible because radically unfounded. Now, concerning this position, van Fraassen consciously develops a perspective somewhat different regarding the aims and the nature of experimental science: a perspective that leads, as such, also to different and more plausible expectations. Indeed, for example: the idea is abandoned that the adequate understanding of the process of scientific representation of the world (as well as the effective accumulation of empirical success throughout the history of science 9 ) makes it necessary (or desirable), in general terms, to assume that one among the possible theoretical alternatives in a certain phase constitutes the one true story of the world (p. 48). We may notice, rather, how every generation ends up effectively evaluating the scientific image that it inherits as open, vague, ambiguous (p. 47); the pretence is abandoned that experimental science must establish its own dominion of primitives entities, that would be the only real, and then demonstrate how all the inter-subjective appearances (the secondary qualities which we experi- 8 The reflection on the problematic questions connected to this clause has been promoted by amongst others: T. Crane and D. H. Mellor, There is No Question of Physicalism, Mind, 99 (1990), 2. 9 See on this point B. van Fraaassen, Structure its Shadow and Substance, The British Journal for the Philosophy of Science, 57 (2006), 4.

7 Bas van Fraassen s Scientific Representation 251 ence), are deducible from the knowledge and the principles concerning them (see, above all, chapt. XIII); the conviction is abandoned that we can look at scientific models of our world (as much in their origin, as in their use) as if they were non-situated and non-contextualized descriptions and explanations, which refer adequately to the world by virtue of an objective and intrinsic relationship with some of its phenomenon (see especially part I); the thesis is abandoned by which certain technological instrumentations belonging to experimental enquiries would necessarily need to be understood as opening a window on the world not directly observable for us, and their outcomes would necessarily be understood as mimetic reproductions (see ch. IV). The perspective that in such a way begins to be sketched out can as such usefully contribute to defusing the risk of a generalized clash between the manifest image and the scientific image of man in the world. 10 This is a risk, which after all, is linked to the inappropriate pretence to be able to achieve, and to achieve in all fields, a scientific-experimental description and explanation that is true, objective, self-sufficient and exhaustive of our knowledge. On the contrary, on the horizon that the reflections of van Fraassen outline, we can indeed understand more easily also how there can be effective knowledge, which, due to its content, cannot, however, be expressed by a scientific theory (or meta-theory), and as such leads to the emergence of a truly relevant incompleteness of scientific theories. 3. Naturally van Fraassen s approach, like also the more general philosophical framework in the context of which it is inserted (even though it does not depend on this in all of its parts), brings several questions to the surface, as well as the need to examine this more closely. I mention, amongst other things, the conspicuous common sense realism (p. 3) that van Fraassen adopts regarding the observables. 11 That is that a non-corrective attitude is assumed regarding our way of categorizing and referring to what is perceivable for us, an attitude according to which: reference to observable phenomena is unproblematic: rocks, seas, stars, persons, bicycles [ ] (p. 3). We can ask ourselves, however, if we could not express with a less onerous and anti-revisionist formulation all of such an epistemic option (epistemic and not ontological) that is of importance for van Fraassen s discourse on science. In any case, also this aspect, however it is evaluated, contributes to the bringing about of one of the main characteristics of the entire proposal of van Fraassen: the idea according to which our scientific models and our scientific-experimental theories do not really have as their aim to go beyond the phenomena and they cannot be made completely independent of the constitutive capacities, effective practices, contexts and the finalities of the subjects who create them. In other words, we could say that we are reminded here that the experimental science of nature reveals itself as knowledge with a human face. 10 See W. Sellars, Philosophy and the Scientific Image of Man, in Id., Science, Perception and Reality, London: Routledge, 1963, pp On this topic see also E. McMullin, Van Fraassen s Unappreciated Realism, with the rejoin B. van Fraassen, On McMullin s Appreciation of Realism Concerning the Sciences, Philosophy of Science, 70 (2003), pp , and

8 252 Carlo Gabbani and Marc Lange As is well known, Hilary Putman has characterized his own position, referring to precisely this metaphor, as a Realism with a Human Face. 12 For van Fraassen we could, therefore, perhaps speak of a (scientific) antirealism with a Human Face. And not only in order to design an epistemic perspective that is faithful to the original human aspirations which it intended to achieve (instead of changing completely into its opposite), but rather in order to underline how, from this perspective, scientific-experimental knowledge cannot ever presume neither to completely encompass the human pointof-view which creates it, nor to free itself completely from it. And indeed, precisely this, far from making science arbitrary, guarantees it the possible objectivity (intended as maximal intersubjectivity, p. 266). Perhaps, therefore, the words of Wittgenstein are apt here when he notes: A curious analogy could be based on the fact that even the hugest telescope has to have an eye-piece no larger than the human eye. 13 (Translated from Italian by Emily C. Fullwood) Carlo Gabbani University of Florence c.gabbani@tiscali.it Structuralism with an empiricist face? Marc Lange Bas Van Fraassen s Scientific Representation: Paradoxes of Perspective is a rich, masterful study of a wide range of issues arising from the manifold ways in which scientists represent nature. Scientific representations or models of the world can themselves be physical objects, as in the case of scale models, but more characteristically they are mathematical models. Under what conditions can such an abstract entity represent concrete objects, events, or processes? What does it take for a mathematical model to succeed in fitting the world? This question runs as a motif through Van Fraassen s book. Consider, for instance, the equation (derived from Fourier s law of heat conduction) specifying how an insulated metal bar with no internal heat sources, having some initial internal temperature distribution, arrives over time at a uniform equilibrium temperature. Each model of the theory consists of a mathematical function TEMP(x,y,z,t) that solves the heat equation. Each model represents one way that such a bar s internal tem- 12 H. Putnam, Realism with a Human Face, in Id., Realism with a Human Face, edited and introduced by J. Conant, Cambridge-London: Harvard University Press, 1990, pp. 3-29, see: p. XV. 13 L. Wittgenstein, Vermischte Bemerkungen Culture and Value, ed. by G. H. von Wright, Oxford: Blackwell, , p. 17.

9 Bas van Fraassen s Scientific Representation 253 perature could vary as a function of spatial location in the bar and time elapsed. How does the model succeed in capturing the temperature behavior of metal bars? The model does not represent a bar by having the same temperature as the bar. The model does not have any temperature; it is an abstract object. The model s success must involve various features of the model standing for various features of the bar. For example, the t coordinate stands for the elapsed time, whereas TEMP stands for the bar s temperature at a given location after a certain period of time has passed. Various relationships among the model s properties (such as TEMP s approaching uniformity as t increases) must be the same as various relationships among a bar s properties (such as its temperature s approaching uniformity as time passes). The model and the world must share some structure. How can this condition for a model s success be cashed out? Apparently, if we lack an interpretation of the model connecting it to the world identifying which features of the model are supposed to represent which features of the world then the model tells us nothing about the world. (I just suggested one such interpretation: t stands for the elapsed time, TEMP stands for a metal bar s temperature, and so forth.) The model itself, absent an interpretation, does not specify whether TEMP(x,y,z,t) stands for a bar s temperature or the reciprocal of its temperature or its electric charge density, or its mass density, or some quantity not pertaining to a metal bar at all. The model taken alone fails to specify what it is supposed to be a model of. How an interpretation can latch the model onto the world is the problem of coordination, which Van Fraassen traces through the work of philosophers including Ernst Mach, Moritz Schlick, Rudolf Carnap, and Hans Reichenbach. An interpretation specifies how a model relates to the world, but how does an interpretation relate to the world? Here we run the risk of reproducing exactly the problem that we were trying to solve. To specify an interpretation, we must specify various features of the world (together with which features of the model represent them). But then we need to represent those features of the world and thus we apparently again need an interpretation. One way to try to avoid this problem is to return to the thought that a model is successful when it and the world share some structure. Rather than a scientist s having to have an interpretation that picks out which features of the world are purportedly represented by which features of a model, the scientist generates empirical predictions from the theory as Ramsified. So formulated, the theory does not have terms that purport to represent various unobservable features of the world. Instead, the theory consists of Ramsey sentences to the effect that there exist various unspecified properties possessing various structural properties and standing in various structural relations to one another and to observable features. The theory represents the world as having some properties filling these various structural roles, but the theory does not purport to specify what those role-filling properties are in themselves. According to such structuralism about science, all we can know (or, at least, all that science cares about knowing) about these properties is that a certain structure is thus instantiated. Van Fraassen traces this view in thinkers including Heinrich Hertz, Pierre Duhem, Henri Poincaré, and Bertrand Russell. It is easy to sympathize with structuralism, since it is easy to find oneself thinking (to paraphrase Lord Kelvin) that from Maxwell s equations, we understand everything about electrodynamics including when charges flow and what accompanies their flow except what it is that is then flowing.

10 254 Carlo Gabbani and Marc Lange However, in 1928, the mathematician M. H. A. Newman offered a now-classic objection to structuralism (or structural realism, as Grover Maxwell later termed it) 14. Newman showed that any domain of entities, as long as it is large enough, exhibits any structure whatever. Hence, if a theory s observational consequences are true, then (as long as the world contains sufficiently many entities) there is certain to exist some interpretation of the theory under which the world makes its Ramsey sentence formulation true. That the world exhibits a certain structure, specified by the theory, was supposed to be a substantial claim about the world. (As Newman said, the question of whether or not matter is atomic is a real question to be answered by consideration of the evidence. ) On structuralism, however, it amounts only to the claim that there are enough entities to exhibit that structure, since as long as there are enough, it is guaranteed that they will somehow exhibit it. Structuralism portrays a theory s truth as coming far too cheaply. Let me put this point in another way, so as to better exhibit how devastating it is. Structural realism is supposed to be a variety of scientific realism, since it holds that in accepting a scientific theory, scientists believe something over and above its empirical adequacy namely, that the world (perhaps in its unobservable features) possesses a certain structure. One supposed advantage of this view (emphasized especially by John Worrall) is that it can embrace the argument that a scientific theory s success in predicting novel sorts of phenomena is probably no miracle (that is, no mere coincidence, as it would be if the theory were utterly mistaken about the world) since the theory s success is likely a result of its having correctly described some feature of the world s structure. However, Newman has shown that for a theory to represent the world s structure amounts to nothing more than the world s having enough role players to fill all of the places in the structure that the theory posits. Accordingly, structural realism fails to do justice to the idea that it would probably be no miracle that a theory makes accurate novel empirical predictions. Structural realism tried to avoid needing an interpretation to fix the way a theory represents the world; it construed theories as representing only the world s structure and thereby contented itself with the mere existence of some interpretation under which the theory represents the world. But as Newman showed, this form of structuralism is undone by its liberality in allowing the existence of any interpretation to suffice. We could try to steer a middle course between needing an interpretation and allowing the existence of any interpretation to suffice by imposing some nonstructural but relatively liberal constraints on allowable interpretations that is, on the respects in which the world has to realize the posited structure in order for the model to count as accurately representing the world. For instance, David Lewis suggested roughly that a theory purports to represent the structure exhibited by the world s natural properties and relations. That the world includes enough entities to fill out the structure posited by a theory fails to guarantee that the structure is realized in a unique way by natural properties and relations. It is thus no longer too easy for the model to count as an accurate representation of the world. However, Van Fraassen rejects Lewis s gambit on the grounds that Lewis s constraint on allowable interpretations is empty unless we have some grip on what would 14 M. H. A. Newman, Mr Russell s Causal Theory of Perception, Mind, 37 (1928), pp

11 Bas van Fraassen s Scientific Representation 255 make a property natural apart from its ability to serve in an allowable interpretation. Of course, I agree that Lewis s constraint presupposes that naturalness has some independent content. Indeed, Lewis puts the notion of a natural property to work in other parts of his metaphysics, such as in his account of natural law. Nevertheless, Lewis is prepared to countenance naturalness as a metaphysical primitive. Although there are some motivations for construing science as aiming to identify the world s natural properties and relations, Van Fraassen sees the formulation of comprehensive, empirically adequate theories as the unique goal of science. Van Fraassen offers a different way around the problem of coordination. As Newman showed, it is easy for there to exist an interpretation of some model under which it is true and it should be easy, since as long as we do not know which interpretation will do, the model tells us nothing about the world. For the model to be informative, we must give it an interpretation, and to specify an interpretation, we need to pick out features of the world (and then specify the features of the model that purport to represent them). Van Fraassen accepts that to pick out features of the world requires using a language that is already coordinated. But in making observation reports, compiling them into what Van Fraassen calls data models, and smoothing and massaging those models in turn (under the influence of some theory that we already accept) into surface models, we are using a framework that we have presupposed to be coordinated. In other words, we regard ourselves as already having the means of picking out features of the world. What a theory confronts directly is itself a representation that we consider successful; we already have ways of picking out those representations (since we did so in constructing the surface model) and we have already recognized them as accurately representing the world (namely, in making our observation reports and turning them into our data models and surface models). That we can interpret a model only by using representations whose relation to the world is not in question might seem to launch an infinite regress. But as Van Fraassen says, we already have a language that we live in. Van Fraassen characterizes his solution as Wittgensteinian, and his use of the term interpretation emphasizes this kinship. An interpretation of a rule is just another rule, so doesn t it need an interpretation itself, just as an interpretation of a model can latch it onto the world only if its components are already latched onto the world? As Wittgenstein famously replied (in Philosophical Investigations, 201), there is a way of grasping a rule that is not an interpretation. In making our observations and measurements, we have stabilized a practice that we take as yielding representations of some features of the world. In using them for that purpose, we do not need to specify an interpretation of them. Of course, we may at some other time find ourselves with reason to doubt whether they do indeed represent the world accurately, and we will then need evidence to support some interpretation of them. But the evidence to which we then appeal we will thereby use without interpretation. As Wilfrid Sellars put the point, empirical knowledge, like its sophisticated extension, science, is rational, not because it has a foundation but because it is a selfcorrecting enterprise which can put any claim in jeopardy, though not all at once W. Sellars, Empiricism and the Philosophy of Mind, in Id., Science, Perception and Reality, London: Routledge, 1963, 38.

12 256 Carlo Gabbani and Marc Lange In short, meaningful questions about a model s capacity to represent the world (such as Does TEMP(x,y,z,t) in this model capture the behavior of that bar s temperature? ) presuppose meaningful ways of talking about the world (e.g., talking about that bar s temperature). Van Fraassen concludes that structuralism finds its proper articulation only in an empiricist setting. But I don t see his response to the problem of coordination as available solely to empiricists in some narrow sense. The representations whose status is not being called into question could just as well concern unobservables, it seems to me. Scientific realists like Sellars are keen to emphasize the presuppositions about the world that lie behind making observations, data models, and surface models especially when the theory used to massage data into surface models is justified only in the context of a certain conception of the unobservable world. It seems to me that structuralist ideas can make important contributions for instance, toward capturing the way science uses structural analogies among the fundamental laws governing physically dissimilar processes to explain why the phenomenological regularities in those processes exhibit various similarities that would otherwise have to be miraculous coincidences. Take these two phenomenological regularities: Consider a cylinder of length L, radius a, generating heat that keeps it at constant temperature T 1, surrounded by a uniform layer of thermal insulation, thickness (b-a), the outside of which is kept at temperature T 2. We find experimentally that in all such cases, the rate at which heat is generated inside the cylinder, thence to pass through the cylinder s surface, is proportional to L(T 1 T 2 ) / ln(b/a). Consider a cylinder of electrically conductive material of length L, radius a, held at a constant voltage (electrical potential) V 1, surrounded by a uniform layer of electrical insulation, thickness (b-a), the outside of which is kept at voltage V 2. We find experimentally that in all such cases, the charge on the cylinder is proportional to L (V 1 V 2 ) / ln(b/a). Is it a coincidence that these two results are analogous? Physically, yes: after all, the electric field is not heat. But mathematically, it is no coincidence. The more fundamental laws of electrostatics and thermodynamics have precisely the same structure. (For instance, heat flow plays the same role as the electric field, temperature as electric potential, and heat generated as electric charge.) That is why there is a single mathematical proof deriving both of these phenomenological results from their respective more fundamental laws. It is no miracle that these two phenomenological laws take the same form not because heat and electric charge share key intrinsic properties, but because they are structurally alike. Marc Lange University of North Carolina mlange@ .unc.edu