Semirealism

0:00 Thank you very much for this meeting of the Sigma Club and today I'm pleased that Anand Chakrabarty, who is a fellow at King's College Cambridge, has come to produce a paper on a rather more philosophical topic. The Sigma Club tends to get involved in a lot of technical things, physics and mathematics, but we also like to keep alive the subject as well. Maybe we'll get technical and see how it goes. So the type of the paper is semi-realism. Well, thank you very much, Michael, and thanks for... I've actually been to the LSE for a few years now, if I'm not wrong. Now, as Michael said, I should really confess, actually, that this is really a talk about scientific realism. I want to talk about scientific realism that is, well, I hope at least, is particularly relevant, or should be particularly relevant to the philosophy of physics. And we may get into some of the ways in which that's the case. Certainly we'll flag some of them throughout our talk and we may come to them in a discussion. Now, of course, we're all familiar with various anti-realist arguments, the kinds of arguments that anti-realists use to put realists on the back foot.

2:30 There are worries about abduction, for example, skepticism about inference to the best explanation. There are arguments for underdetermination. And in particular, the one anti-realist argument we're going to be especially concerned with today is pessimistic induction. The idea that if we look back over the history of scientific theorizing, what we find is a kind of graveyard of past theories that we once held to be true, perhaps perfectly rationally so, and yet now we think our faults are flawed in some very significant ways. So the argument goes, by induction, it seems quite reasonable to think that our present-day theories are the same, that looking back at 50 years or 100 years' time, we're going to look back at the same sort of embarrassment that we now do when we look back 50 years or 100 years ago. I think we are embarrassed. I'm not sure that we are. Well, that's really the point of this talk, to see whether we should be embarrassed. Now, it's not obvious to me that there are convincing responses to all of these forms of anti-realist skepticism. It's certainly not obvious to me that there are knock-down arguments against various forms of skepticism. Some of these forms may be perfectly coherent philosophical positions, and ultimately, I think, some of the debate gets a bit bogged down because there may be no non-question-begging arguments, ultimately, that are going to decide the issue of the answer to realism. But that said, what I'm interested in, and certainly here, is realism. I'm interested in seeing whether it can be made a coherent, interesting, compelling account of what we're doing when we do science, so that at least if one feels pulled in that direction, Then one has recourse to a not-obviously-self-reputing or otherwise silly position. So the question is, if one wants to be a sophisticated realist, what sort of realist should one read? One general maxim that is employed by a lot of, I think, quite sophisticated realists that I respect is this. These are words that a realist can live by, and they are, don't believe everything you're told. Scientific theories, of course, can be interpreted as making all sorts of claims about the nature of reality. But we really only have good grounds, that is, we only have an epistemic warrant for believing in some of these claims. The problem, of course, as I'm sure all of you are familiar, is that none of these sophisticated realists actually agree on which bits we should actually believe.

5:00 They all seem to disagree with respect to where it is that we should draw the line between what's justifiably believed and what isn't. So I guess I should say right off the top that I like this strategy of believing in only parts of theories for which we have a limit. And I also agree with many of the insights that these various realists who proposed different drawing lines between what we should believe and what we shouldn't. And also I agree with a lot of their insights. But ultimately, I also think that none of them have drawn the line between what we should believe and what we shouldn't in exactly the right place. So the term semi-realism is really meant to convey the idea of a shared general strategy. There will be some stuff, but not all of it. Partial belief strategy. But I guess I just chose the different terms to indicate that there's a different proposal here with respect to where it is that we should draw the line between what we believe and what we don't. So, just as an outline, what I'm really going to do is just count some of my favorite sophisticated realists and pull out from their approaches the parts that I think are really onto something. And then I'm going to clarify why it is that I'm not quite happy with where they've drawn the line between what we should believe and what we shouldn't. So just to give you an example, I think a lot of what I regarded as fairly standard scientific realists, like Philip Kitcher or Stavros Silos, say quite interesting and useful things about what parts of theories we should believe in. The problem with their approach is that I find them quite vague. I don't think their advice is specific enough to actually allow us to implement it. On the other side of the coin, there are people like, on the one hand, Nancy Cartwright and Ian Hacking, and on the third face of the coin, people like John Morrill, and in a very different way, Stephen French and James Ladyman, who again have given us some, I think, very good advice on how to draw the line. And this advice is particularly good because it's not the most specific. And therefore much more useful. But again, I find in each of these cases that perhaps we're being a bit too exclusionary, or unnecessarily exclusionary, in what it is that we're going to accept. And then I'm just going to present my own view, which no doubt admits of all the same sorts of worries that I'm working on. So to begin with, perhaps I could just review very briefly some insights that I want to take from entity readings.

7:30 Entity realism, we recall, holds that under certain circumstances, we have good reason to believe that the entities referred to in scientific theories are actual inhabitants of a mind-independent external reality. It's this aspect of theory, the existence of particular theoretical entities, that we can reasonably believe to be true. Now, of course, this view is controversial for some of the anti-realist reasons that I've already mentioned. Some of these sorts of arguments are exploited by people like Manfred Rossi, who wasn't born in England. And others to raise skeptical doubts about Entity Realism. I don't want to review all of these reasons here. I think Entity Realists have some quite convincing things to say in response to those arguments, and some less convincing things to say. But what I do want to do is focus on one response in particular, which I think is crucial to responding to the worry of the pessimistic induction. So the question is, why should we believe in scientific entities? Well, the answer that Entity Realists give is that we're connected to them causally. We can use entities to manipulate, intervene in the phenomena. We can use them as tools to various ends. We can exploit their causal powers. Now, the obvious question here, in the light of that sort of commitment, is how can we believe in the existence of entities when, over time, our theoretical descriptions of entities change? First, we said an electron was one thing, and then we said it was something else. In the 20th century, we had lots of theories characterizing the nature of an electron. How can we believe in such seemingly unstable entities? Well, the entity realist person says, I think, not so fast. I don't believe in the whole list of theories. That's why entity realism is not equivalent to standard realism. There's a lot of theory that I don't believe or that I'm agnostic about. So, rather, I'm only going to believe in certain, say, core sets of properties or some low-level theory. The stuff that's required to do the causal work that I've just described. So, I think in the early days of entity realism, people complained How can you merely believe in the existence of entities? Surely, a belief in the existence of an entity minimally demands some other beliefs, beliefs that will support the existence of entities, for example, some knowledge of the properties of the entities, what the dispositions that these entities have in virtue of having those properties, etc. If you don't have those sorts of things, then you lose the causal explanations upon which entity realism is predicated.

10:00 So without those, without some sort of knowledge of properties and relations, then we can't generate the sorts of causal stories that are required by entity realism. Well, in the manner that I've just suggested, I think entity realists have really, if quietly, sort of taken this on board. Certainly, Ian Hacking accepts that there are core sets of properties that he's going to have to commit. Nancy accepts low-level theory. She doesn't accept high-level theory. She accepts phenomenological laws. So I guess it's here that my worry creeps in. Once we admit that in order to use entities causally to do interesting things, we need to have at least some knowledge of their properties and relations, does it really make sense to draw the line between what we should believe and what we shouldn't, between the existence of entities and everything else? Well, I'm not sure that it does. So, while I'm drawn for arguments, drawn by some arguments for Entity Realism, my worry is that it gives a misleading impression of where the dividing line is between what we should believe and what we shouldn't believe. The valuable insight of Entity Realism is that, of course, we're just not going to believe in a science fiction simplicity theory, but I think we need to diverge from the advice they give us, at least as it's often expressed, that we don't really believe in the existence of certain entities. We also believe in some aspects of theory, just not the whole thing. So, I'd like you to just keep that sort of message in the back of your head, because I'm going to come back to it. What I now want to do is turn my attention to another proposal for how to be a sophisticated realist, and that's structural realism. And just to give the game away, basically, what I'm going to do is attempt to describe why the same sort of ambivalence I have when it comes to entity realism also applies to the case of structural realism, and that's going to lead to the position that I'm hoping to send you in the case of. So structural realism is the view that insofar as scientific theories give us true, or approximately true, descriptions of reality, they don't actually tell us anything about its underlying nature, the underlying nature of reality. Rather, scientific theories tell us about the structure of reality. So the obvious question in this context is, what do we mean by structure? We're going to spend a bit of time learning about this. Informally, the idea of structure has to do with relations between the elements of some system of components. Structural realism focuses on the relations themselves, rather than on any punitive relata.

12:30 Now, in the contemporary literature, structural realism is coming in two flavors. There's the epistemic version of structural realism, which is associated with John, Eliezer, and there's the ontic version of structural realism. Which is associated with Stephen French and James Layman. Epistemic versions place a restriction on scientific knowledge. The advocate of such position holds that we can know structural aspects of reality, but nothing about the natures of those things whose relations define structures in the first place. The objects themselves are beyond the grasp of our cluster knowledge. So this is a kind of London, Cambridge structural realism. Quantic versions more radically, Do away with objects altogether. The proponent of this sort of position holds that, at best, we have knowledge of the structural aspects of reality, just because there is, in fact, nothing else to know. So here we find the denial of any traditional metaphysical option, as I said, associated with layman and French, so we can look on this as a kind of means-style structural realism. Now, it's going to be my contention that the most desirable form of realism is, in fact, the form of structuralism. And furthermore, it's both epistemic and ontic, although, of course, not in quite the ways that these authors intend. Unfortunately, just because of the time, I don't think I'm going to have a chance to really talk about ontic structural realism at all, which is unfortunate because I think it's very interesting and does involve recent developments in physics, but perhaps we'll come back to that in the discussion. The kind of structuralism I want to propose collapses the distinction between a knowledge of structures and a knowledge of natures. What I want to do is describe a position that stands good chance of doing some justice to both realist and anti-realist arguments in a way that, in fact, will be acceptable, I think, to the most sophisticated realists, at least realists in general, but yet in a way that emphasizes structuralism, the structuralism of scientific knowledge. Structuralism, of course, has a venerable history in 20th century philosophy of science. Poincaré champions the knowledge of relations at the expense of elusive, unknowable objects. Duhem recommends the representative parts of theories, that's the mathematical laws expressing relations, over the explanatory parts, which are the underlying entities and processes.

15:00 In neither case, however, are we given an especially clear description of how best to understand the concept of structure. And so, for clarity on this point, many structuralists, at least in the first instance, turn to Russell. So here's Russell's definition of structural identity. In fact, I think I've just I quoted it on your handout. It says, we shall say that a class alpha ordered by the relation r has the same structure as a class beta ordered by the relation s but to every term in alpha, some one term in beta corresponds and vice versa, and when two terms in alpha have the relation r, then the corresponding terms in beta have the relation s and vice versa. So, more generally, two relations have the same structure if there's a one-to-one correspondence or correlation between the members of the response set. And if and only if, for every subset of the members of alpha that has a relation of r, there are correlates in beta of a relation s, and that's the intuitive identity. The crucial point here is that on this definition of structural identity, the members of alpha and beta need bear no qualitative similarity to one another whatsoever. The only requirement is that r and s bear a purely formal similarity to one another. They need have no qualitatively similar aspect. So structural identity on Russell's account does not depend on what sorts of relations define the structures, so long as there is a relation of some sort between the elements of the system. And that's sufficient to determine the structure. Now, of course, Russell's original motive in putting forward this view wasn't to be able to counter scientific anti-realism per se, but phenomenology. So in order to make the knowledge of an external world possible while conceiving the inaccessibility of the world in itself, He proposed a structuralist epistemology, where only the structural aspects of the world could be known, and the nature of the world underlying its structure is beyond our grasp. Worldly structures, for Russell, are mirrored in the structures of our percepts, but no non-mathematical properties of the physical world can be inferred from perception. We are acquainted only with our perceptions. Nevertheless, of course, we can't have knowledge of the external world. We just don't have it directly. I'm not acquainted with it. We have that knowledge by description. We can infer and thus describe its structure. Now, in the modern context for which we're attempting to co-opt Russell here, the scientific realist, of course, opens up a structural identity not merely between reality and perception,

17:30 but also between reality and the outputs of measurement devices, instruments, and detectors. But I think we can easily co-opt what Russell is saying. This is where the controversy begins. As was documented by Demophilus and Friedman recently, in maintaining that only the mathematical, formal, or logical properties of the world can be known, Russell invites an objection. For Russell, structure is a higher-order property, a property of relations. And on this view, no first-order properties or relations of objects in the world can be known. Only the structural properties of relations between first-order properties can be inferred. So the potential difficulty with this was pointed out by Cambridge mathematician Newman. And I'll just read you the quote that comes from the novelism agreement. Newman says, no important information about the aggregate alpha except its cardinal number is contained in the statement that there exists a system of relations with alpha as field whose structure is an assigned one. Given any aggregate alpha, a system of relations between its members can be found having any assigned structure compatible with the cardinal number of alpha. Thus, the only important statements about our structure are those contained with the structure set up in alpha by a given definite condition. So the charge here is that the claim that a system has a particular structure on Russell's definition of structural identity doesn't tell us anything about the system other than the number of the elements you labeled. Because any collection of elements can be arranged so as to exemplify a given structure, so long as there are enough of them. Given any set alpha and any arbitrary structure W, it's a consequence of set theory or second-order logic that there exists some relation in alpha to the structure W, so long as W is compatible with the number of elements in alpha. So on Russell's approach, the claim that some aspect of the world has a particular structure is truly dissatisfying. That's the charge. And Grover Maxwell married Russellian structuralism to the Ramsey sentence to produce his own version of structural realism. To form the Ramsey sentence of the theory, as you know, you take the axioms of the theory, you conjoin them, you replace all the theoretical terms with predicate variables, and you essentially quantify the rhythm. Now, an unobservable entity whose place is held by a predicate variable in the Ramsey sentence is whatever it is that satisfies the relations that are specified by that sentence. Given that Maxwell, like Russell, maintains that the unobservable objects and their first-order properties, their intrinsic natures, are unknowable and that all we can know are the higher-order properties of entities that we cannot know, some have argued that this program is likewise susceptible to the Newman projection, the idea that the detour via Ramsey sentences really doesn't gain us anything.

20:00 The use of the Ramsey sentence may be useful for displaying Rossellian structures, but is no less trivial for its utility, so on this approach, an empirically adequate Ramsey sentence is trivially true, so long as cardinality constraints are satisfied. Now, a moment ago, I described this as a controversy, and it is a controversy. There are two sides to this debate. Worrell and Zahar, Don and Ellie, had recently written a paper, and Don was speaking about this at a DSPS in its last meeting last summer. To argue that this charge of triviality is in fact not properly applied to the kind of structuralism that they're interested in. And Ellie gave me a copy of the paper and I've had a look at it and completely digested it now. But to just give you the gist of what's going on there in a couple of sentences, what Ellie and John were saying is that The reason Newman is able to make this charge and make it successful is because Russell actually just made a mistake, but it's actually not a very serious mistake, if one would be correct. Russell talks about believing in only structure, believing in pure structure and nothing else. But in fact, if we make a strict distinction between observation terms and theoretical terms, we can recast the claim. But we have knowledge of structure in such a way that this charge of triviality does not apply. The Ramsey sentence is not trivially satisfied. It's not merely a cardinality constraint. Similarly, Michael has argued recently, at least in his reviews, now this is the simplest book that just came out in medicine, that the idea of the charge of triviality isn't well-founded because it's predicated on the idea that when we say that we believe in structure, We have believables and this higher-order property of relations. The charge of triviality would really only go through if we were by that statement saying that we were dismissive or ontologically dismissive of the first-order properties and their relations. But of course, we're not. No one is saying that there aren't first-order properties or things and relations.

22:30 Structural realism, based on John and Ellie's view, is an epistemic position. We're saying that there may be first-order things in relations, there are these higher-order structures, and all we have epistemic warrant for are the higher-order structures, but that doesn't say that these other things don't exist. I'm not convinced that these arguments are successful, but at the same time, I should honestly say that I'm not entirely certain that they're not successful, because I haven't quite got my head around some of the details. I'm still weighing the arguments here. But for the purposes of today, or at least the account that I want to spell out, I think that the issue of whether a human's objection is cogent or not is actually a bit of a red herring. Why is that? Well, because the issue of whether a human's objection is worrying or not turns on the question of what kinds of structures you think are actually comprising the knowledge in neuroscientific realism. And this raises again the issue of how best to think of structures in this context. So what I want to do next is outline the metaphysics structure that I think gives as much realism as is realistically defensible. So to foreshadow here, I believe that the structures to which we have access are not merely those comprised of higher order formal or logical properties. And secondly, I want to draw a lesson from Maxwell here, because there are places where Maxwell hints Admittedly, and lucidly, he doesn't say anything in detail, but he hints at least, that causal connection is an important structural library. I want to take that lesson on board. Hinting this, I think that Maxwell was on to something. And something that's of particular relevance to the construction of the defensible structuralist interpretation of science and work is. So the metaphysics and structure that I'm going to describe will feature causal relations fairly centrally in the definition of structure. So what I suggest is that we view Russell's definition of structural identity as a necessary, but not a sufficient, condition on the notion of structure and requirement. The structures we want, I suggest, are not higher-order properties, although we can have those too. The properties that we want are relations between first-order properties of objects. So let's identify structure with relations between first-order properties.

25:00 So we turn for a moment to basics, and what's the intuitive notion of a structure? To give the structure of something is to enumerate its parts and to cite the relations between those parts. What I'm suggesting is that we view structure as something that's tied to specific kinds of properties and their characteristic relations. Now, at first that may seem strange, I think, particularly for those of us who are used to approaching, at least encountering the term structure in the context of mathematics or physics, The most helpful way, perhaps, to clarify the distinction that I have in mind is to take advantage of some terminology of Michael's. Michael distinguishes between what he calls concrete and abstract structures. Concrete structure is just a relation between first-order properties of things and diseases. An abstract structure is a second-order abstraction from the concrete structures. In other words, it's what Russell had in mind when he gave his definition of structural identity. A higher-order property that can be instantiated by different kinds of concrete structures. So the way Michael puts this, he says, we can collect concrete structures into an isomorphism class. So that's the class that's given by Russell's definition of structural identity. Or if you don't like this talk of instantiating properties, then you can think of it extensionally just as the isomorphism class itself. That's what the abstract structure is. Now, I think that this distinction clarifies what sometimes otherwise remains ambiguous in a lot of everyday talk about structure. When we say, for example, that These blueprints give us the structure of a house, for example, or the Watson and Crick's demonstration model gives us the structure of DNA. These are instances of the identity of abstract structure, and this is something that I think is very common to representation. In order to represent something, a blueprint of a physical model need not instantiate, obviously, the same concrete structure as that which is represented. It need only instantiate the same abstract structure. And in fact, I think we could say something weaker than this, and we have partial satisfaction that Russell's definition of structural identity is probably sufficient for something like representation, because representations don't need to be perfect, right? Like, they're very normal. So, my point is that if we want to be structuralists and realists, then it's really concrete structures that we're after. I personally think that a knowledge of mere abstract structures, assuming that it can be defended against the Newman objection, might be defensible in that way.

27:30 I think that's too weak. Now, perhaps this is just a question of taste. I mean, I certainly don't want to argue about a question of taste as to where to draw the line between the strength of our commitment to the way we describe the world, but what I do want to suggest is that in just the same way that arguments for entity realism support the knowledge of more than just the existence of entities, I think the kinds of arguments that are ordinarily adduced in the context of structural realism actually do support more than just a knowledge of an abstract structure. Now, why would I think it's such a ridiculous thing? Let's consider a bit more carefully the metaphysics of concrete structures. Earlier I said that the structures I'm interested in involve relations between first-order properties, and that causality is going to play an important role in understanding them. So the proposal is that first-order properties whose relations comprise concrete structures are those that we might call causal properties. There are properties that confer dispositions for behavior on the objects or the entities that have them. Let me just digress here for a minute. Why and how do objects interact? Well, it's in virtue of the fact that objects have certain properties, presumably, that they're causally applications. Let me attempt to clarify the notion of this, the nature of this in virtue of relation. Properties have things like Masses, accelerations, volumes, temperatures, and all those properties confer on the objects that have certain causal capacities. These capacities are just dispositions to behave in certain ways when in the presence or absence of other objects and their properties. So the property of mass confers, among other things, the capacity of the body to be accelerated under applied forces, for example. The property of volume on a part of a gas might confer, among other things, the capacity to become more highly pressurized under applied heat, and so on and so forth. It's the ways in which these dispositions are linked to one another, that is, the ways in which objects with various properties are disposed to act in consort with one another, that produce what we take to be causal activity. So causation ultimately has to do with relations determined by dispositions, or conferred by these properties, or objects to behave in certain ways. Now, of course, talk of dispositions, as soon as I introduce the term, it really invites the kind of detailed metaphysical account that I suppose I don't really have time to give you.

30:00 But let's suffice to say that there are, of course, numerous ways that we can spell out the security of dispositions. How exactly does a property confer a disposition to behave causally? Well, metaphysicians first describe this feature of objects in terms of powers, and it's the apparently mysterious connotations of powers that led empiricists to project it. Some who hold that dispositions are genuine, current properties, regardless of whether or not they're manifestable, speak of causal properties in terms of categorical bases of causal dispositions. Others just speak of dispositions implicitly. When I say here that properties confer dispositions to enter into certain causal relations, the ambiguity of confer is really intentional here. And I'm using confer to signal a degree of neutrality on this refined brain metaphysical picture. The current contention, the only thing that I really want to insist on here, is that properties are responsible for the behaviors of things. Objects behave in the ways that they do because they have certain causal properties, and that's enough to motivate the kind of structuralism that we want to give. Now, earlier I suggested that this kind of structuralism would collapse the distinction between structures and natures. According to epistemic structural realism, the intrinsic natures of things are different from the structures of things. Structural realism commits to the structural, but not to the intrinsic. On the view I'm proposing, however, there really isn't a distinction of that sort. I mean, perhaps there are intrinsic features of things that go beyond the structural. But there's a profound sense in which structure, by its very nature, is an intrinsic feature of the things that are related. Because structures are here identified with specific relations, and these relations are a function of the dispositions that are conferred by first-order causal properties of objects. So on this view, to say that two sets have the same structure is ipso facto to say something about the intrinsic natures of their members. The intrinsic natures of things, as is told by their first-order properties, are inseparable from the character of their relations. Speaking somewhat loosely, we might say that causal properties are not only intrinsic, but they have a kind of relational quality. That is speaking loosely, by that what I mean is that they're relational in that the dispositions things have as a result of possessing these first-order properties determine the kinds of relations that they can enter into.

32:30 So a knowledge of these relations gives us insight into the intrinsic natures of things. On this account of structure, if we know that some set of things has a certain structure, we thereby know something about the members of that set. We have a knowledge of some of their first-order properties, which confer dispositions to bring about the relations detected. Now, given that causal properties are understood in terms of dispositions for relations in the way that I've just described, then structural knowledge contains a knowledge of properties, and thus natures. Structures are, metaphorically speaking, encoded in the natures of entities. Because first-order properties convert dispositions into specific relations, those would be recognized as structures. So that's the picture. The knowledge of structures entails some knowledge of the intrinsic natures of entities. But of course that's far from an indication of the entire content of scientific theories. As I mentioned earlier, ultimately, in order to cope with anti-realist arguments such as pessimistic induction, the realist requires a principled means for differentiating parts of theories that are likely to be retained as theories move on. And we need to separate the things that are likely to be retained from those that are apt for replacement. The structures to which realists should commit, I will argue, are those involving properties and relations that are reasonably believed on the basis of our causal connections with the world. So here you can see how I'm borrowing from the entity. Let's just consider how this is going to work as a response to the pessimistic induction. Various responses to pessimistic induction have been employed by realists. For example, some argue that if one factors in various other criteria, for example, if one considers only mature theories and mature scientific discoveries, Or if one considers only theories that make novel predictions, or that don't go in for ad hoc measures, etc. If we factor in those sorts of criteria, then the data for pessimism are much reduced. These sorts of strategies may well work and may be effective, to a point. I certainly don't want to discount them. But for present purposes, what I want to do is focus in on one argumentative strategy in particular. Because even if, for the sake of argument, we grant the pessimistic induction, Realism, I think, thrives on a semi-realist interpretation, because on this approach it's quite reasonable to believe that there generally should be a significant amount of discontinuity in scientific theorizing over time.

35:00 What there is not, on the other hand, is much in the way of radical discontinuity in what is properly believed. So the trick is to separate those aspects of theory that are most worthy of belief from those from which we have less warrant. Once the distinction is made, then the realist can happily admit a pessimistic induction on the history of past science simpliciter, while simultaneously asserting an optimistic induction on the parts of theories to which she commits. Now, in recent discussions, several standard realists have echoed a common theme. The theme is that the parts of theories that realists should take most seriously are those that do the work of Those that are responsible and perhaps indispensable for doing things like making predictions, retradictions, giving explanations that are characteristic of the mature sciences. So Kitchard, for example, distinguishes between what he calls presuppositional policies, the idle parts of theories, and working policies, the parts that are really required to generate predictions and explanations. And Syllos says similar things. My worry is that this is all a big thing. So while I think it's on the right track, I'm not sure how helpful it is, and I'm going to try to replace this with something slightly less vague. I think the structuralist-type proposal that I've been sketching offers a very specific proposal for distinguishing between what works and what doesn't. We confer dispositions for particular sorts of behaviors that we detect in the form of certain relations. If this is so, then the ability to distinguish between the parts of theories for which we have warrant and those for which we don't It just boils down to an ability to distinguish between parts of theories that we think they have good reason to believe concern genuine causal properties and those about which we're less sure. So in order to elaborate this idea, I'm going to introduce a distinction here between the attribution of detection properties and the attribution of auxiliary properties. Detection properties are causal properties that we manage to detect. They're the ones that are causally linked to the regular behaviors of our detectors. Auxiliary properties are any other punitive properties that are attributed to netizens by scientific theories.

37:30 Now this is an epistemic distinction. Detection properties are the causal properties that we know, that is, the properties in whose existence we most reasonably believe on the basis of our causal contact with the world. The ontological status of auxiliary properties is something that's unknown. It's something we should suspend our judgment on. They may be causal properties. They may be non-causal properties, if there are such things that are known. Or they may be fictitious. They may be things that were attributed to theories but, in fact, were just mistaken about. Whether the attribution of a given property qualifies as a detection property or an auxiliary property may well depend on the current state of scientific inquiry. As the sciences move on, then some auxiliary properties will be retained as auxiliary, but some are converted into detection properties, and others are rejected entirely when we realize we're just parking under the wrong tree. So the little figure on the handout that I've given you kind of spells out how I see this relationship. All detection properties are, of course, causal properties. That's just because to detect is to establish a causal link with whatever it is that's under investigation. The attribution of auxiliary properties is kind of non-committal with respect to the ontological status of the punitive properties. Causal properties themselves, of course, are not exhausted by those that we attribute to entities and theories. That's why it's a much bigger set. This diagram is not to scale. So theories change over time. But if it turns out that detection content is generally retained, and that it's generally the auxiliary content of theories that's left behind, well then we would have a systematic basis for a realist analysis of the pessimistic induction. The realist could then commit to detection properties and their relations, but remain agnostic at best, skeptical at worst, about auxiliary properties. So what's needed then, obviously, is some means of demarcating detection properties and auxiliary properties. Well, here's a suggestion. The parts of theories that we have most reason to believe are those describing properties that are connected by a causal process as our means of detection. We generally describe these processes in terms of mathematical equations that can be taken to represent the relations of various properties. Thus, detection properties are just those that are required to give a minimal interpretation of those equations. I'm going to return to the idea of minimal interpretation in just a moment. For now, though, it's nice to say that anything that exceeds a minimal interpretation, including interpretations of equations that have nothing to do with putative detections, go beyond what is minimally required to do the work of science, to make predictions, to give explanations, etc.

40:00 The excess is auxiliary. The parts of theories to which rules should commit are structures involving detection problem. Any other putatively structural content, together with any other auxiliary supposition, is properly met with, I think, suspended judgment or skepticism. So perhaps we could just recall John's and Poincare's example of the transition in theories of light from Fresnel to Maxwell. Fresnel had a particular idea of what light was. It was a disturbance in the ether, an elastic, solid medium, alternating medium. And this particular case study of John's looks at equations that Fresnel developed to describe how light behaves when it moves from one medium into another medium of a different optical density. Some of the light's reflected, some of it's refracted. These equations, Fresnel, give you a way of calculating the intensities of these various beams of light based on their directions of propagation or what's present. The insistence of certain properties, I argue, is required to make sense of these mathematical equations. Now, the interesting bit of John's story, of course, is that ultimately, Maxwell's theory was accepted in the context of a non-inferior physics, so a different kind of metaphysical picture about what light was. If not him, then people who followed him accepted his theory, but the equations that Fresnel came up with can of course be derived from Maxwell's theory, so what we have here is a kind of continuity of mathematical formalism, but we seem to have a very different metaphysical picture of what that is. What I'm suggesting is that in order to give the bare minimum in terms of interpreting these equations, we need to attribute certain properties to that. And I think that these properties are shared across the two theories. Namely, there are things like intensities and directions of propagation. So these are properties of light, but do they make reference to an ether? Well, no. Do they for that matter make reference to an electromagnetic field? I don't think that they do. So in the very limited context of these specific equations, things like ethers and fields are auxiliary positives. Our theories incorporate such entities as important heuristic devices, no doubt. They help to fill out our conceptual pictures of the phenomena. But the semi-realist advice here is, look, we should believe in certain structures, the relations of detection properties, and treat anything that exceeds these structures with suspicion.

42:30 And on that basis, the realist has some principled grounds for believing in some, but not all of those principled grounds. Now, of course, there are a number of questions that need to be addressed here if this is going to provide anything like a compelling response to pessimistic induction. Why should we think, for example, that descriptions of detection properties, identified by giving a minimal interpretation of certain mathematical equations, why should we think that that's likely to be retained over time? Now, the answer to this question might come, in part, from lots of case studies. It could be brought to bear for and against, and some people have recently done some interesting case studies. But I think that there's an a priori reason to think. Or to suspect that detection properties in our relations are likely to be retained in some form in later theory. The reason is that it's these structures that are the ones that are required to explain, and thus ex-psychology are responsible for, the regular behaviors of our detectors. It's these structures to which we have the closest epistemic access. Only these structures are causally connected to our means of detection. Descriptions of causal properties to which we have the best epistemic access Should remain relatively stable as theories are modified and improved. In fact, I think the realist may actually want to assert something stronger here, and that's that, in some cases, we may need to retain specific structures involving detection properties in some form or another if we're actually going to retain the ability to make certain kinds of predictions. So, here's an example. This was actually something Stephen French suggested to me. He said, think about electromagnetism. No matter what form later incarnations of the theory take, One might argue that we would just lose the result if the speed of light or the speed of electromagnetic radiation is C if we didn't retain something like Maxwell's equations as a component of a later theory. So, on this view, we have reason to expect both, I think, an optimistic and a pessimistic induction on the history of science. We should expect that many contemporary theories are going to jettison some of their content, but they're not going to be in the same form at a later date. But at the same time, descriptions of relation between detection properties are likely aspects of theories to be retained as auxiliary components are in place.

45:00 But given the important heuristic role, and I think this is quite important, played by auxiliary properties, I think we should nevertheless expect and in fact commend their presence in scientific theory. If further investigations don't rule them out or convert them into detection properties, then they'll simply remain in a theory as auxiliary content. So, this proposed account of structuralism suggests that the relation between detection properties are generally the best confirmed parts of theories and thus, to echo people like Hans Post, are most likely responsible for their empirical successes and light ways to be obtained. If this is correct, then the problem with the pessimistic induction is really that it's painting with too broad a stroke. It's not the case that later theories generally incorporate the rules of their predecessors, everybody knows that. That's, of course, partly impetus for scientific change and rejection of aspects of the mistake. Let's take an example of Loudoun's. Loudoun claims, in just one of many similar claims, that there can be no continuity of the sort that posts between, say, classical epoch theory and relativistic mechanics, because although some of the laws of classical mechanics are limiting cases of laws of relativistic mechanics, others can't. Why is that? Well, because they invoke an entity that just isn't counted as one of those specific entities, that is, the ether. But of course there are things in later theories that aren't carried over. I mean, that, in addition to incorporating new knowledge, is the whole point of scientific change. And that's hardly surprising or injurious to the semi-realists. The ether is part of the auxiliary content of earlier theories. It's not required to give a minimal interpretation of the equations representing relations between detection properties. So the structuralist has nothing invested in them. So thus restraining her belief, the realist is immune to various charges that are made in connection with the pessimistic election. Now, it would be disingenuous in the extreme if I were to suggest that the solution I've just offered is going to be easy to apply. I mean, semi-realism in practice is going to be substantially more difficult than it is in theory. In particular, the notion of a minimal interpretation of mathematical equations, on which the whole identification of detection properties depends, bears a great epistemic weight. In many cases, the determination of a minimal interpretation, I think, will prove quite a challenge.

47:30 If structures are retained in some form across different theoretical frameworks, it's crucial that the meanings of these terms that are used to describe the relevant properties and their relations likewise extend. But this would seem to require a theory of meaning with a referential component, construed at least partly in terms of a causal theory of reference. So I think there are two especially pressing concerns here. And in the time that I have left, I'm just going to say something brief about those. The first complaint that one might have is that the task of providing a minimal interpretation of mathematical formalism is actually impossible in any McMulligan retrospect. Where hindsight makes the identification of detection properties and their relations trivial. One problem in life. The second problem is that one might hope that the appeal to some version of a causal theory of reference makes it impossible that the unobservable terms in most scientific theories can fail to refer. I think these two worries are linked, and forever deter in the sort of structuralist position that I've described into an empty position, one that's trivially satisfied and uninformed. So let me just examine these a bit more carefully. The first charge is that hindsight is 20-20. It's easy to claim that we've identified past aspects of theories that do the real work of science after the fact, because then we find, surprise surprise, that these aspects have been retained subsequently. So the charge here is really a kind of post hoc rationalization. We're looking back from the perspective of the present, we're bound to identify the aspects of past theories that have been retained as those that do the real work. And if this chart is correct, then it renders the realist's commitment to these aspects somewhat empty, for then one might argue that the realist believes in the retained elements just because they're retained, not because they're more likely to be true. If this is the case, then it would be impossible to know at any given moment in time what new structures posited by current theories should be believed. That kind of determination is one that you can only make in retrospect. A commitment to concrete structures, however, It doesn't merely identify belief-worthy parts of theories with parts that are preserved. That would be to suggest a correlation without any explanation for why the correlation or tendency. Rather, I've offered a formula for identifying the belief-worthy parts of theories, and this formula is recommended on the basis of its putative epistemic value.

50:00 When a realist determines which detection properties and relations are minimally required to interpret the mathematical formulas in their theory, Her belief in them stems from the fact that these are the structures that cannot be denied if one believes that the theory is to any extent successful in representing parts of the world and their relations to our detectors. These are the aspects of theory for which we have the greatest epistemic warranty. But these are the elements that cannot be done without making predictions, forming generalizations, and so on. So as I said, I grant that this would-be recipe for success is likely to prove difficult in many cases. Theories often describe quite general causal frameworks, and the portion of theories to which the realist should commit is going to be embedded in a larger framework at one time. So consider again the Maxwell, Fennell-Maxwell case study. Though Fennell had a particularly general causal framework involving the behavior of climate, not all of this understanding is required to give a minimal interpretation of this equation. In Fresnel, there are the labels of disturbance in the ether. But the structuralist says, look at the equations. My version of the structuralist is attempting to say, look at the equations of what properties do we find there. Well, the variables represent altitudes, intensities, and angles, directions of propagation. But, you might say, are these not intensities and directions of propagation in the ether? Well, the structuralist is lamooned by this appeal to the greater causal framework. To suppose that the direction of propagation is furthermore a direction of propagation in the ether is to move beyond what is minimally required to give an interpretation of this particular set of equations. So, the structuralist thus commits to relations of intensities and directions as described by the theory, but remains agnostic about any further embellishments. Now, I think the reason that this prescription is going to seem difficult to follow is that it asks us, it asks realists at least, to withhold a commitment from parts of theories that play near-explanatory roles and commit instead only to those causal structures that are minimally demanded by mathematical equations describing detected And that's going to be a challenge. So, for Fresnel and other ethotheists, we might say, the causal story told by their equations is ipso facto part of the causal story of all the meaning of it. The structuralist here is kind of asking for a bit of a radical suspension of belief. We're being asked to separate out various aspects of an overarching story.

52:30 Perhaps for Fresnel or other ether theorists, for psychological or professional or theological or other reasons, they might not actually be able to achieve. But the point is, the point that I'm trying to make, or would like to make, is that it's not impossible. In fact, we know it's not impossible, because Fresnel's equations were ultimately accepted as part of Maxwell's theory in the context of non-inherial physics. So we know that it can be done. What about the second charge? Well, the worry here is that on this account of realism, the reference to desired structures can't be fixed purely by description. The theoretical descriptions, of course, are refined as we move on from theory to theory, and yet we continue to refer to the same properties and relations. It seems that what's required is an account of reference that makes at least some appeal to a causal theory. It may not be entirely a causal theory. It may not be a causal descriptive theory or something like that. But it's going to involve a causal element. Now, it's a common complaint that causal theories of reference trivialize reference. If we can ensure that we refer by assigning the reference of terms vague enough causal worlds, then that would seem to make error impossible. And of course, no self-respecting realist couldn't claim that her commitment to a scientific discourse is invalid or silly. I think that so long as the structuralist is wary of this charge of triviality, then she need not succumb to it. Just to illustrate this, consider Hartman-Rosenberg's response to the accusation that the central terms of many past theories are non-referring. One of the suggestions they put forward is that, well, we can construe past theory terms as referring on the basis of causal roles attributed to their reference by the theories in which they occur. If the role assigned by the past theory is the same as that associated with an entity in the current theory, then we can reasonably conclude that the terms refer to the same thing. So this is essentially a variation on Pudlin's principle of the benefit of the doubt. Now again, I think this advice is far too vague. The question facing the realist, again, is where to draw the line between reasonable applications of this principle and ones that clearly go too far. And ones that would attribute approximate truth to arcane theories merely because their ontologies were positive to account causally for some of the same things that we're interested in today.

55:00 So the sort of structuralism that I've been describing, I think, provides the beginnings of an answer here, for it seems reasonable to apply the principle of the benefit of identity in cases where not just general causal roles are retained, but where quant-specific dispositions conferred by particular causal properties are preserved. So on this view, it would be unreasonable to apply the principle in such a way as to identify with one another the putative reference of significantly different systems of properties. About to run the risk of trivializing reference. So to just illustrate this, I was just thinking on the train of a particular example. I thought this one might do nicely. It's the example of the one that's often sought as a serious problem for the realist. A transition from the dominant theory of chemical theory of the 18th century, Stahl's theory of phlogiston, to Lavoisier's theory of oxygen. Priestley is of course celebrated for his lifelong commitment to phlogiston theory. Which accounts for various phenomena like combustion, calcination, respiration. In terms of removal, oxygen theory accounts for the same phenomena but in terms of the absorption of oxygen. Now, on a Harden and Rosenberg type strategy, you could say that deep logisticated air refers just to the same thing that oxygen does. It refers because deep logisticated air and oxygen are playing the same causal roles in the two respective theories. The greater the extent to which there is low logistic, the greater the extent to which it's rich in oxygen, and vice versa. Now, of course, this is precisely the kind of conclusion that those who condemn the realist legal to the causal theory of reference dislike. It seems to trivialize reference to say that Priestley was actually talking about oxygen all along. Now, of course, there's a sense in which he was talking about oxygen all along, in the assumption that there is oxygen and there isn't logistic. But realism appears ridiculous if it has the consequence, perhaps, that logistic theory and oxygen theory are, in effect, just different versions of the same theory. From a structuralist perspective, though, at least the one that I've described, I don't think there's any question of identifying Priestley's descriptions with Lavoisier's, although some specific causal roles described by their theories for deep logistic air and oxygen might be the same. It's plainly not the case that the supposed causal properties of dephlogisticated air and those of oxygen are generally coextensive.

57:30 So, for example, it's partially auxiliary understanding of oxygen that it has a fixed chemical composition, but dephlogisticated air, ex-hypothesis, does not have a fixed chemical composition. Different combinations of gases could all be likely to be logistical. And, of course, different... Different combinations of gases that are lacking in fluidism would all have quite different causal capacities. The same can be said of oxygen. So clearly here we're dealing with different punitive sets of detection properties. And realism shouldn't license sameness of reference in that case. So I think structuralism at least begins to provide a way of giving an antidote to the threat of triviality. Alright, just to conclude. I think it's a major undertaking for the realist, merely to spell out the details of the coherent position that does some justice to scientific knowledge, and that offers a response to any of the anti-realist arguments, let alone all of them. I've attempted to outline an intuitive account, I think, of scientific realism, that is interpreted as a form of structuralism. Now, I'm hesitant to call it structural realism, but that's just because of the very strong connotations that I take it. The term comes with an act that implies that structural realism cannot be a commitment to anything other than higher order properties and things. So I hope it's clear by now why it is that I'm calling it semi-realism. And then when I do that, what I really do hope to do is unite what I take to be the most convincing parts of entity realism and structural realism. So I want to endorse the general strategy that we shouldn't believe everything we're told. For entity realism, I do think that we can know that particular kinds of entities exist. Why? Well, on the basis of the knowledge, the relations of first-order properties. But against entity realism, precisely because we have a knowledge of the existence of such concrete structures, precisely because the existence of entities, the claims about the existence of entities depends on knowledge of concrete structures, that it isn't right to say that we only have a knowledge of the existence of entities. That doesn't draw a line between what we should read and what we shouldn't read. For structural realism, I think that we can know that particular structures exist

1:00:00 by an interpretation of the things that we typically take to represent those structures. That's mathematical equations. But again, structural realism, as I sometimes put it, our interpretation of these equations is, I think, too weak if all we think that we can know are abstract structures. I actually think that by looking at the mathematical forms, we can know more. Because I think they tell us about relations between first-order properties. In other words, they tell us about concrete structures. So the crucial insight is that structures do not under-determine the natures of entities. Rather, they under-determine the auxiliary contents of theories. Theories offer explanatory frameworks which incorporate both detection and auxiliary properties. And this combination constitutes a causal understanding of the phenomena we investigate. In order to distinguish these properties, we just have to turn to the equations that we use to represent the world and ask, what would the truth of these equations minimally demand? Now, in answering that question, I think we should consider not what is possible metaphysically, not what possible metaphysical pictures are consistent with the equations, but rather what sorts of property attributions are essential to their satisfaction. Not what's possible, but what's required. We can use them to construct the math of natural phenomena in terms of mathematical equations. If we understand these detection properties as dispositions for relations, first-order properties of things, then it seems we can have a knowledge of both entities and structures. And I think that means plenty of auxiliary content about which we can be as agnostic or skeptical in life, and which we can point to when we need something to blame when it comes to things like pessimistic induction. So that we then have grounds for both an optimistic and a pessimistic induction on the history of quantum physics. Things sort of come behind it. But let me just, as part of the discussion, just run through the miscellaneous things in no particular order. First, I just want to get clear about it. When you use the word semi-realism, it's supposed to be an epistemological, nevertheless, great source of information.

1:02:30 Sure, yeah. I want to actually see The distinctions that I'm making are constitutive of semi-variables and epistemic distinctions, so I do want to say that there may be all sorts of things that are described in our theories that may very well be the case, it's just that we're not in a position to confirm or deny those things. In other words, we don't have epistemic warrant for some of the things that are typically described in some So, the distinction really is an epistemic one. I certainly don't want to go the ontological route and say that if it's only structures of a certain kind that we know, that there really is nothing else. I mean, that as a paraphrase of what the ontic structure was up to is a bit unfair. I mean, that's an argument for ontic structure. I was going to ask about the stratification. Yeah. Well, I mean, can we not do what you said about what we can abstract the structure of the circle? I don't think that was so useful. That's the approach. It is not only physical concrete structures, but also concrete mathematical structures and so forth. And this is how you do it. Well, it's one way of understanding how mathematics gets locked onto physics because they both have the same structure. It's only translated from mathematical language to order of calculation, and then translated back into physics. So it's that possibility in that it involves all types of opposite parts. You've got all the concrete mathematical structures in physics. There are such, but that's again, the issues that cross the mathematics. I mean, that would seem to be an important point. Well then, the next thing in this video is about Newton. Now, it's simply that Newton is not a human. I've got a trivial logical point here. It's so trivial. A, you can't understand how a bus is seeing.

1:05:00 And secondly, was it really worth the human mind to do this? I don't think so. We actually need to do this. What he's really saying is that we've got a relational, a system of relations which has a structure X that we've existentially quantified, that exists in R such that R has a structure X. Well, then there's a terribly weak thing, a miserable weak thing, so weak that it's bound to be true. Well, you know, causality is having what you said and you didn't say. But it seems to me that what the answer to that is sort of, well, I want to impress you a little bit, but you agree with me or don't agree with me. But if you think, if you actually put four on the table, as Paul and I would say, and they're on the table about the actual system of relations, and then say that that system of relations has to start from this, Now, it's not logically much stronger that you may have a structure S, you may not. It may be, when you investigate the situation, you find that it does have a structure S, that's important for the model to apply, but it's certainly not inquiring for the topological knowledge you get from that terrible, miserable, existential classification, you know, that Newton goes on about. I mean, I agree with this point about the deal of Syllops' book, but Syllops wrote back and said, well, you can't really be a structuralist because, you know, what are these relational materials and all that sort of thing, because you can't, I think it's probably called the structural terminology because, of course, I'm trying to admit that, but it seems to me that you're also really on that aspect of saying this is why it's a question about the universe and the elements and so on. I mean, yes, there are these relations, I'm not saying there aren't any such relations, And then the question of whether they have the structure or not is something we might learn to discover, although we couldn't learn other things about the relation. I'll come on to this question of the other things, like the inner work. And so that was, it seems to me, a kind of trivial answer to you, but I've never ever been able to persuade anybody to say, yes, Richard, you're right. I mean, I've never actually had that response. I mean, technically, it's a hog. Well, John Paul was a hog. So, people always say, oh, well, maybe, and no, I don't really agree, and so I want to put you right on the spot, and do you think Ritter is right or not, and that's trivial, but we don't get paid to say that.

1:07:30 Yes and no. Oh, no, no. You're off again. Can you amplify this? I have to say, yeah, I'll certainly amplify that answer. I certainly am with you, to the extent that, you know, I feel this toll as well, and I feel Green, John, and Rose take it, but of course, the claim, Newman's objection, isn't obviously trivial when you agree that, of course, there are other things at play, there are observation terms at play, there are other things that we have access to, but it's not entirely clear to me that merely saying that we think that these things exist, but we have no knowledge of them, will actually do the kind of work that we want to do. I'm a little bit sympathetic to the response he gave to you throughout. So let me just outline it for those people who haven't read it. I mean, Michael was saying, look, this worry, human's worry, really only goes through if we're denying that there are first-order relations and abstracting our structural knowledge. Is the higher order property of something that, you know, we're denying in some sense, or we don't know, but of course, Michael says, of course, we think that these things are there, these first order things are there, it's just that we don't have knowledge of them. And Stathis' response to you, if I'm remembering this correctly, was to say, well, isn't that a bit like wanting to have your cake and eat it too? Because what you're saying is that we can have the knowledge of the abstract structures and consider the first order stuff as kind of an ontological composite out of which we're not going to draw any substantive conclusions. But really, the epistemic warrant you get for the abstract structure It's really only conferred to the structure if you do have some knowledge of the first-order properties and their relations. And I take it that that was the sort of gist of his point. And I am sympathetic to that. Well, I cannot respond to that. I think the synopsis is not clear. Because you see, I take John Watt's original insight, Watt, for the things that really survive. It's the mathematics, you know, things come and go where the mathematics survive. Remember the sort of, the equations that John put. Well now, if you take that view, what survives in the equations, or the equations of the structure, now I'm perfectly happy to believe that you can't have a relationship without a structure, without a relationship. It's all back there. But if you say, well, what actually survives as physics develops, as structural contiguous, we'll come on in a moment to the question of the discontinuity of development, of course you're right.

1:10:00 But on the topic of continuity, well then, I think that all that you get is that that's the right equation, it's in the right equivalent of what we're talking about, and if you say, oh well, what about the extra bits of knowledge about the relations? It might be an electric field or something like that. You say, well, no, I actually don't really know anything about the liquid field. I don't really understand those at all. But I can tell you they have very poor partial differential equations. So I'm not there to ask you that, but to say, what is a field? Well, now, of course, part of the what-is question, and again, I do agree with you, because distinction between structure and nature is a silly one that I've brought to the table, and I don't want to say what-is-of-me. I'm very likely to say, well, it's something that's satisfying, but I'm not going to say what-is. You know, so I'm totally happy to bring in my sort of structural knowledge into answering the what-is question, but my point is that to understand many of these equations as a result of mathematics, you don't need to know any more. Of course, you've also got to satisfy other equations to be able to detect the fields of other kind of networks and that sort of isolation. So that's my feeling. Could I now press you on this question about... Would you like me to say something about it? Yeah, go ahead. Yeah, okay. Because, again, I agree with you, certainly, up to a point, and that is, I agree with you to the extent that, I mean, it is going to be just what we can know on the basis of the fact that we have equations over time and that we have proof of reasoning to say, you know, this is what we're really going to commit to. The rest is something that, you know, we flow and we change our minds on it when we come up with different theories. I certainly agree that far. I guess where the point of difference is really occurring here is that I think that just by looking at the mathematics, we actually get a little bit more than what we think we get. So I think that by looking at the equations, we get more than just the abstract structure.

1:12:30 We actually get a partial story on the concrete structure. Well now this is where I think they're going to bring the detection properties and that kind of thing in. Well now this is what I want to really press you with, because that story about when you divide the things you know into the things that you don't have, the detection properties that you do know, I think that works quite nicely in many episodes in classical physics. But if you look at what's happened since the beginning of this century, classical physics being replaced on one hand by relativity and on the other hand by quantum mechanics, then the type of explanation, the type of property becomes much more structural. And I think it's just an example of what I mean by that. I mean, for example, in relativity, the thing we discover about relativity is that space-time is what it is a metric to make. It's not a metric. It's a pseudo-Romanian metric. Well now that seems to me really much a structural property, it's not a kind of causal thing, it's more like geometrical theorems, but if I say the angle of a triangle and the angle meets the point, you wouldn't want to say that's a cause of a triangle, it's some sort of structural property of a triangle. It seems to be going on in those insights. And when you have the quantum mechanics and things from the examples, it seems to be not causal at all. I mean, I know Jim Bush is sitting in the back there going, tell us what you can tell us. But if you just take bits of information from the political parties, This seems to me just a kind of structural property about the possible states we have in the same class and so on. There doesn't seem to be any sort of cause in this topic, certainly not this electron repelling that there. I mean, in popular accounts it looks as though it's the same electron in the same sense, they're all pushing each other apart. But it's not the electric charge which is producing the propulsion that's coming from, but such a structural feature. I mean, there are lots and lots of other examples, but again, in which the type of explanations you get in quantum mechanics and relativity, because they tend to be, these days, pretty much geometrical, they tend to have this sort of structural character, and so I don't really recognise these sort of causal detection properties in modern physics. Philip happens to go along with most of what you said in classical physics. Why?

1:15:00 It's the only bit highlighted to that. Why would you go home? I mean, Russell argues for an eliminability of cause, I think, completely. It's hard for me to understand how one could deny it. Are you using classical examples? No, no, I'm perfectly happy to go along with Schubert and Kuhn were the famous articles that had the notion. I mean, I think it's even more dramatic in the 20th century cases. I was trying to give you the most, I mean, don't give him anything to speak of. I was giving you half, half of that, and you're trying to take half of that away. So we can argue that the next problem is that it operates passively, but it certainly doesn't seem to operate in anything like a certain way, and most orthodoxy can't do that. So that's my problem with your overall scheme. My idea is to try and redefine bits, you know, protection bits and so forth. Should I say something? Yeah, well then I'll shut up and everybody else can get through. You have a lot of reports and a lot of learning points. Well, I mean, let's take the, you know, objection straight off the top. Of course, I haven't said anything about causation here, and that's a possible worry, I guess, since I'm relying quite heavily on the notion of decisions for behavior and taking those to be causal. I actually think that When I employ this language, I'm actually intending a very, how should I say this, the kind of causal relation I'm requiring is very weak. So when Russell, for example, argues that we have no need for the philosopher's notion of causality because we have these differential equations and they tell us what sorts of things can exist at what sorts of times, and blah, blah, blah, I think you can give a thoroughly causal account of what's happening in those differential equations. Whenever you have an equation that says when one parameter is varied, the other parameters must also be varied in a certain way, I take that as sufficient for giving a very weak account of causality. By that, what I mean is these properties Convert dispositions such that when one parameter has one value, there are only certain values that other parameters can take on, and that's a relation between those properties, and that's all I'm describing. Pardon me? You have to be back with causation. With? Back with causation, that's it. Because everything's reversible. Right, right. Well, there may be more to be said. I don't know.

1:17:30 This changes in the future, then, something that... Sure, exactly. Yeah, I mean... On your account, that would be equally causal. Well, I don't think I've spelled out much on the count of causation. I've just said that, I mean, minimally, right, that's the kind of relation that I would, or it's causal, whenever you have a mutual dependence of properties in that case, it's not an accident. It's written more as an accident. That's all I mean by causality. So, now when it comes to various other examples that you might cite in contemporary physics, well, I'm not sure about some of them. Others, I think, if they're a part of the story that you need to tell, right, In order to explain what's happening when you do actually get testable predictions that may verify or falsify your theory, then there's a sense in which those relationships are causal in this very weak sense. But I'm not sure whether, I mean it may be that some of these relations, if they're purely structural, are a consequence of the properties that things have. Which can be described in causal terms. I think there's a big difference between the experimentalists and the theoreticians in modern physics. The experimentalists tend to really, you know, spread out from the round cabin, sort of that kind of thing. They're technically involved in what they're spread out. Sharpening diatoms and that sort of thing. In some sense, I mean, most experimentalists are really being castigated. There's a little bit of complementarity in this, let's say, paradigm. But the theoreticalists, you know, I mean, they're the ones that go deep into the structural, geometrical type of people. It's a little sort of sea change in how the whole subject gets done, I think, is kind of the geometrical, structural approach. And the main thing is that they're bad at doing science and tools and all that. I'll get back, and I'm sure Jim in a moment will tell us how to get back to the formal story or something. I think I've spoken quite a lot. Jim, John, you're the no-bun person. Do you want to pursue this no-bun? I think, to me, if you interpret cause in such a weak way, as Andrew was indicating, there's not really any dispute between the no-bun and the half-bun. Between you and him, I mean, I've not seen any real... If it's just your cause, it just means some connection that's betrayed in mathematics.

1:20:00 If you change one thing, then you change something else, and it can be forward as well as backward, then it's not contrasted with your means, you know, it's not something that's a correlation, it's not an accident, whichever way you like to put it. The point about this is actually made between connection properties and auxiliary properties. You mentioned that auxiliary properties sometimes turn into detection properties provided that we discover these causal regularities, but you didn't mention something that you mentioned in your 1998 paper, and I just wanted to know whether you changed your mind, and if not, give us an example. This is just one sentence on page 404. You say, in the process of theory, properties are converted into detection properties. That's what we just said. And then you say, conversely, when we rule out particular relations or detection properties, we suggest others that have not previously been considered. So you seem to be suggesting that detection properties can themselves be eliminated. Sometimes. So, do you still think that's the case and can you give us an example? Sure, but they can or they cannot be, sorry. Yes, but they can be eliminated, ruled out, thrown away after they... So, you think that some are detection properties, but then at some point you get some of these detection properties discovered. I'm sorry I've led you astray there. I mean, I certainly didn't mean to give the impression that it's likely that detection properties are going to be things that we change our minds about. However, you know, I don't want to say that with absolute certainty because ultimately I want to be validist about this, right? I mean, there are instances where we just don't have much to our chagrin. Continuity of mathematical performance. Sometimes there are upheavals. And in those sorts of cases, perhaps those are cases in which we're going to want to say, actually, we just got it wrong. The kind of set of detection properties that we postulated to explain what's going on here perhaps needs a serious rejigging. But in any case where you do have kind of continuity in that mathematical mechanism, I don't want to suggest that that's going to occur at all. I mean, those are the properties that I think we can be reasonably sure about.

1:22:30 And it's the auxiliary properties that are kind of up for grabs. And, you know, the reason they can be converted into detection properties or ruled out will just, will follow as a matter of what kinds of detections we become good at performing. If we think that we can isolate properties of these things and test for their existence, then we might have a good reason to think that, you know, they exist in just the same way that, because we haven't come up with really convincing experiments to prove that the ether exists. Ultimately, we've given that up as part of the metaphysical picture of binary theories that we no longer accept. So, I think I believe the same thing then, but I believe now, which is detection properties are going to remain relatively stable, and it's their relations that comprise the structures that are going to form the sort of backbone of our realism, and it's auxiliary properties that, you know, Relatively stable, not completely stable, which means that we may actually need to have another distinction or try to find a way in which we can say which aspects of those protection properties we want to keep and which ones we don't. Well, I mean, you know, I don't think there's a lot of room. We could come up with another distinction if you wanted, which is absolute certainty. But since I'm like, I never thought that it's not a problem, you know, it's the kind of distinction in which I'm happy to say, you know, a lot of atheists are actually really agnostic, right? They think a lot of atheists think that they can't be absolutely sure, but they just think that the problem is there, right? I mean, it's in that sense that I think I'm pretty sure that the tension properties aren't going to change. I don't want to say absolutely, but I think we're in pretty good hands. I think Jim, you're doing a good job. Oh, thank God. Very good. It's helpful for me to talk on realism because I was a duper typical. But one of the things you mentioned at the end was, the question was, if I had these mathematical equations and relations, what sort of the, what's minimally required? That just seems like a sort of amorphous beast at the present. It seems to me, and I think it relates to something Michael used as an example, it sort of relates to what Do these equations, do these relations require any further explanation, or do they just stand there on their own? I think one's response to that isn't, there's a great deal of variation, I'm sure, in that response.

1:25:00 Michael's example from special relativity about this metric structure of space and all, I mean, if you want to play intellectual hardball with your criterion, all I'm committed to then from the equation of special relativity is covariance of predictions. I'll just stop right there and talk about space, so talk what you want. I mean, it seems we all pick and choose what we want to make our case, and I don't know how you get a sense of it. So, just a specific question. If we have a theory, match those equations, set of relations, do those require further explanation, or are those brute facts of nature, and they need a causal explanation, and if so, How do you get away from the ontology? These structures come from the ontology. I mean, do we lust after an axiom in terms of the ontology, or are we intellectual enough to reciprocate the structure? Right, right. Well, I think that it isn't enough just to present the equations and say, this is what I believe. I don't think that's explanatory enough to constitute much of a scientific realism. Now, again, as I hinted at during my talk, I mean, that might be a matter of taste. If someone's happy to say, This equation is kind of an ancestor of equations that have resembled it in form over time, and I'm quite happy just to present it ostensibly and say, this is the content of my realism. Well, you know, I think perhaps you're welcome to, but I think that is not saying enough. I do think that the equation is the inside explanation. So the question then is, what sort of explanation, John is shaking his head vigorously, what sort of explanation do they need in order for us to think Do you think that we have some sort of, something that we might reasonably call communities? Well, I was suggesting something, I think, intended to be quite minimal. That is, when we have relations of this sort, each of the variables represents a property, or a kind of property. What do we know about those properties? Well, I was suggesting that we can think of them in terms of dispositions for relations, precisely those relations that are displayed in front of you, that are represented by the mathematical equations. So we know, by looking at the variables, that there are properties that these things have, which confer dispositions such that these sorts of relations will obtain. Now, that's pretty minimal. It's pretty vague, too. I mean, it might seem to me because, I mean, they're somehow related to something. I mean, I said, for example.

1:27:30 Oh, no, they're related to one another. Okay, but, I mean, variables in a wave function or something, do they represent properties of the system? There's nothing definite there. I mean, there's no definite. There's no definite configuration there. There's all these possibilities. Right. Well, see, that's why I was careful to describe it in terms of dispositions, right? I mean, that's what I think these properties are. So, I mean, I could... I'm not sure whether... I mean, tell me whether this is getting out of your way or not. I mean, I suppose you could say, look, you know, we have... Various, you know, Fresnel's equations describe what's going on, when light moves from one dispersal medium to another, and, you know, describing light as a sort of classical waveform, but in, you know, now we think of, we give, we talk about light in terms of quantum photons and other excitation states in a quantum field, but the idea of a classical waveform is going to go on by the wayside. So, is there any sort of real continuity there at the level of ontology? And what I'm suggesting is that when we say that the properties that are named on these simulations are dispositions, what I mean is that light has first order properties. The conferred dispositions, such that when they're measured in certain ways, we're going to detect wave-like behavior, as is, for example, approximated in classical physics. And, of course, you know, wave-like behaviors, when light is subjected to detection, haven't gone anywhere. They're still with us today, just over 200 years ago. Despite the fact that I'm doing these things in terms of the quantum field. Could I ask you just to put it slightly, but aren't you close to all of these virtues? Oh, I'm not close to all of these virtues. Pardon me. I mean, these are dispositions. And so this is one aspect of the topic. One of the things that I didn't really get to go into very much was the idea that, I mean, this view really comes for me with a realism about dispositions, and so in some ways it does borrow from, although there are things that I disagree with in sort of Nancy's ontology, it borrows from the fact that the first-order properties of things that we trust in are ultimately dispositioned. I had another question. Earlier on, this whole thing about emerging properties and layers, where do they come from? So you have first-order properties.

1:30:00 I mean, in other words, collective phenomena. I mean, what do you take collective phenomena for? Are they sort of an incipient piece of the collection compound system? I'm not entirely sure what you mean by Well, the properties of the systems exist only when there are lots of them. Okay, right, the emergent properties. Right. First of all, I should just say as a throwaway, Mark, that I think the idea of calling them dormant virtues is, I think, pejorative. If you're a realist about dispositions, then of course you don't think they're dormant. I mean, dormants, you think that they're actually real properties. So, I mean, there is a kind of, if you're a realist about dispositions, then of course, I guess you don't adopt it in ontology. You view them as occurring properties, whether or not they're manifest. So, I mean, I shouldn't sort of lay that on the table, but with respect to emergent properties, I see no reason to think that if there are emergent properties, that they cannot be related in precisely the same sorts of ways that the underlying properties, which when, you know, when they reach a certain level of organization, they give rise to emergent properties, I think they could admit of exactly the same sorts of relations. Some people seem to find a real problem with that, right? Yeah, on this subject of the dispositions, In a different debate, which I've engaged in a lot, which is the debate about the realist interpretation of space and time, the metric field in general relativity, there are people who say that if you believe in general relativity, you have to believe that space-time is a thing, and so there's properties that are described in the theory, and then there's people like Paul Teller. We said, well, no, no, we can still be relationists and say that space on a thing isn't there, and instead we just talk about the relations between objects and the dispositions, there's a bunch of dispositions that objects have to behave in certain ways when they're in certain relations. And I pressed him on this because I found the whole story dating. And what he ended up retreating to was the position that the dispositions are dispositions not of the objects but of the world. The whole world is the bearer of the disposition. And I think you'll be pretty much forced into that same position because you've got ontologies changing over time with such a degree of... Well, between Fresnel to Maxwell to Einstein to quantum theory, but anything you want to describe the dispositions to has fluctuated so often that if you've got a constant set of dispositions that you want to talk about and anchor your detection properties to some set of dispositions that really haven't changed over all these upheavals, then they just have to be dispositions of the world. And are you happy with that?

1:32:30 The first point is, I'd like to accept some of what you say, but resist the conclusion. I don't think that it's obvious that by adopting this sort of picture, we end up just with dispositions of the world, if you say. And that's because, I mean, you're absolutely right to say that ontologies are changing over time. But that's because, when we talk about ontologies generally, we're, in constructing the ontology of a particular theory, I'm going to be building that ontology by citing both the detection and the auxiliary properties of the theory, so when I want to construct my kind of minimalist ontology, it's going to exclude the auxiliary bits, and I'm just going to be talking about the dispositions. So now the worry is, what's going to anchor those dispositions? I mean, this is, you could cast this in the terms of a very old debate. I mean, what anchors any object? Is it just a bundle of properties or is it a bunch of properties that's in here in a substratum? I mean, I don't think there's a peculiar question that applies specifically to the account that I'm giving here because An object is, I mean, I'm claiming knowledge of an object or an entity, which is given purely in terms of some of its properties, and then whether those properties can cohere together as a bundle, and that's enough to describe its ontological status as an entity, or whether it actually requires something else like a substratum. I think that's actually a separate question. That isn't to say that... It's really obvious to me how we should solve that question. I mean, what is the ultimate metaphysical status of these entities that I'm describing basically as bundles of dispositions? I think you can go different ways on this. I mean, I think perhaps this is part of the worry that Part of the inspiration for people like Stephen French and James Layton, when they say that, look, if we're just characterizing these things structurally, then we should just say that all there is is structure. In talking about Maxwell and so on, you tended to use the word light as what has the dispositions. That's a nice observational term, so that's a good example. But of course, light means the things we can see, and then there's the other bits of the electromagnetic spectrum. And once you sort of pile in The whole electromagnetic creation spectrum, then you've got a much more theoretically entangled idea, don't you?

1:35:00 So, do you intend to usually, to use a term like light, anchor things to the directly observational, direct, anchor your detection properties into the directly observational? Not necessarily, no. And so this is, I think, a place where John and I, one of the several places where we're going to diverge, because John might want to I'm not going to anchor some of the structuralist, the knowledge that's going to characterise in structuralist terms, in the observable. Is that correct, John, because you want a strong theory of observation distinction, whereas I'm not actually going to be thinking about that problem. But you are, I should talk, I mean, I take it detection is... But another good term would be observation. I'm not like Elie in thinking that the only observation terms we can subtly answer, but I do think we need to think of certain predicates as directly anchored in reality somehow or other, and the obvious candidates are And stuff at the Poincaré crude factual level, you know, the needle points to 5, the computer print out is 0, 1, 1, 0, 0, and all the rest is interpretation, but that, I think you've got, what's that? I think it's, I'm afraid I'm one of the people who don't accept Michael's resolution and the Newton problem. That's the main first step, to see that it's not trivial, the Ramsey sentence, which I do think the properly constructed Ramsey sentence does capture that. So the observational terms have got to be thought of as tied to reality independently of any further description if you want to take this line at all and then of course you can prove straightforwardly that the Ramsey sentence is co-observationally All of these terms are very powerful with the original theory, so it's automatically can't be trivial, because it's got the same observational consequences, so it's refuted by any observation that refutes the original theory. So it's certainly not trivial. Now, you might want to argue, and I think that's what Demopoulos and Friedman are hinting at without getting it quite straight, that not that it's trivial and that it reduces to... There is a commonality constraint, which is also not true, and I don't know how to analyse it properly, but it's trivial.

1:37:30 Over and above, I mean, it's the claim that if you identify the cognitive concepts of a theory with a dramatic sentence, you are in effect reducing it to its observational consequences, because what you're saying over and above the observational consequences is somehow or other trivial. The first thing, that sounds plausible, but it's not terribly easy when you're trying to think it through to see exactly what that means. I mean, after all, you know, the structure of this is just, it's not trivial that there are these observational, it's a massively complex theory that you're going to get, but interrelationships between theoretical and observational terms. I think the trouble, the trivial, the assumption that it's always going to be trivial comes from thinking about simple cases. Where you might indeed want to say, you know, things like that, T1x, for all x, T1x, then L1x, for all 1x, O1x, then T2x. Well, you might well want to say in the end that that's not a genuine theory. It doesn't have any, well, in those sorts of cases you can prove that it's fully, I think, fully equivalent. In the Ramsey sentence, it's fully equivalent to observational query. But that's not true in general. The whole picture is that... The rest is all known by description and so ultimately when we unpack the theory in terms of its theoretical primitives, what we know about those primitives is given by the overall theory. And so what on earth reason could there be not to existentially quantify them? I mean, it's just like, it would be word magic to think that there was, that you were losing any cognitive content by thinking of it in an existentially generalized way. But see, when you put it that way, I think, you know, that sounds completely uncontroversial to me. So, I guess... That's right, at least I'm not supposed to just know that he accepts it. But then, but then see, my question is... My worry is that, why is there this fixation on the idea of higher order, to use Michael's terms, abstract structure? Because I could come up with a Ramsey sentence, give descriptions of all the first order of properties I want, and then I would be happy with that. It would be exactly the sort of structuralism I'm arguing with. I want to say something about the abstract and the concrete as well. There will be cases. I mean, you pretend that the cases that Michael rightly draws attention to where you've got two

1:40:00 First-level structures only prove that they're isomorphic, interesting, isomorphic, and resourceful. Now, in a way, that's not... what the structural realism is directly saying is that just that you can't expect your theory to distinguish between structures that are isomorphic with respect to all its properties. You aren't going to know. You've got to admit the possibility that there are, I completely agree with Michael, that there are, in the world, when you get to the pearly gates and look down and have Rossellian knowledge by acquaintance of everything, so you just directly intuit the electrons and the quarks and the superstreams just in the same way that I directly equated with Carl, If there is such a thing. But when you get to that, you'll be able to see that you'll see that there were two sorts of things that are different now that you couldn't be acquainted with everything, that even the latest theory and before everybody got wiped out couldn't distinguish between them. Because, you know, there are two isomorphic models of your, of your, you can't, I mean, it just, it is a logical triviality. Your theory can't distinguish between, by definition, between isomorphic structures, isomorphic with respect to the properties and relations specified by the theory. End of story, it seems to me. That, that's now, you see, I think I might be in the waters as far as anybody reads this stuff, but it might be the waters by Going along with thinking of Poincaré, a saying, which sometimes he does seem to be saying, when he has this famous quote about ruin, the man of the world sees ruins, harm upon ruins, but the skepticism is superficial, the real man of science sees that there's a continuity going through this set of ruins. I think what he's really saying is that when you view it properly, there aren't any ruins, there aren't any substantial ruins. Because it's only when you kid yourself into thinking that there's more to the theory than you should, because he is thinking in the same terms that you talk about, which is minimal mathematical representation of the mathematical theory, that there aren't any real.

1:42:30 Because what you never thought of for an analyst talking about an ether in the sense of something that, I don't know, that if you squashed it all together and bounced it on the floor it would bounce back, but just in the sense that it's something or other, disturbances in which operate according to certain mathematical equations. And then there's no ruin at all, because that's exactly what Maxwell tells you as well. Well, I mean, I completely agree with you, which is a nice thing to know. I mean, there is a sense in which there are rubens. They're not substantive rubens. They're rubens of theories that were interpreted in correct faith, perhaps, right? I mean, once you take it on board, once you take it on board Poincare's advice, then you can see how there might be rubens. But once you take it on board Poincare's advice, then perhaps the rubens just dissipate because you realize that you never should have interpreted it in the first place. I mean, yeah, I think that's... I'd like to bring Chris in here. Chris has a particular thesis related to the Ether model, and so it is, you know, if this corrects the view about Brunel, you know, in particular, the function of the Ether. Oh, he did. No, no, he didn't. No, no, no. I didn't say it, so I don't know. Yeah, well, no, I was going to say, I mean, I really think that the sort of, certainly the views over the 19th century do change massively, and, you know, young when, because, I mean, apart from the whole question of, There's a whole extra question about, for instance, during the first couple of decades, about whether the vibration is transverse and longitudinal or whatever, and Young has a major issue when he realises that vibrations have to be transverse for him to believe it has to act as an acetolic rather than a fluid, and he finds this immensely difficult to conceive. In fact, it says that it should be taken as a mathematical postulate, but after that, certainly for now, it is quite definite that it needs to be a real thing. But I actually take issue with a few things that John says, because certainly I think before Poincaré, there was certainly Maxwell and some of the other people at that time, Fitzgerald, and I think even if you go

1:45:00 Even further back to the mid-century, Airy is definitely trying to distinguish between the mathematical theory and the physical theory. I actually would be tempted to look at the ether field transition not so much as a shift in entity realism in itself, but actually a shift from I suppose entity realism to something akin to structural realism, so that people are then saying, no, hang on a second, what we really want is to take the equations as primary anyway, which is not terribly helpful for you, because you want to use example as a shift between entity realism. Well, not entity, but I think we agree. I've expressed it in a bit of an apparently different way. One of the problems in this area, and this is a problem that, for example, I have to find, makes a meal of when he says that realism is dead, is attaching realism, the view of scientific realism just basically says that you should read off your best present description of the world from this developed scientific theory, tying it instead to some specific, already well-established, less physical view. Now, of course, there were people, Maxwell himself firmly believed that you couldn't just postulate as primitive the electric and magnetic field vectors at every point in space. There had to be an underlying medium, which, the contortions in which, well, there's a whole massive research program held in other people who were desperately trying to work on that. But what happened, of course, was that they failed, just like Newton failed to produce an even theoretical account of gravity, although he was completely convinced that the issue was a lot smaller. They went back to the view of... I don't see that there's any shift in... It's just which theory you think is likely to be true. Can we start off thinking that we would just like to have... All of these have to be in motion and not have these weird alleging of magnetic fields as a sui generis entity, but then it turns out the best theory you could get just takes them as primitive entities and you finish up reconciling yourself to that, just in the same way that 18th century Newtonians reconciled themselves to actions of distance, temporarily, because it was the best account that you could give and that then for the time being became

1:47:30 But what was realist about? When you become realist about it, then you take a view of Maxwell's theories, Maxwell's economists, which was always true, what they were really saying was that there might be some still deeper theory of what you're looking at, what you're looking at, of course you can always produce a model, you can just say that there is an underlying medium which happens to have the property of that. I don't want to drag the conversation into too much of a historical alleyway, because I know that's not what we're here to discuss. Personally, I see it as a shift from what is quite definitely an entity-realist situation around the middle of the century, where people are definitely trying to... You mean an entity-realist, like a real mechanical medium? Yes, something you can sort of squash up and see what they did, except that they didn't. To then, around the 1870s, 1880s, even 1890s, a situation where people basically said, no, we're going to be agnostic about what the thing is, we're going to simply use the equations. Then, back after the 10th century, towards something which is then rather more sort of energy realistic in terms of there being a real field. So it goes from sort of ether to the equations and back to the field, not from one to the other. But they're always agnostic at the primitive level, except they don't realize it because they've got used to it. I mean, going back to the Newton case again, it's always been amazing to me that people thought that action by contact was somehow more naturally primitive than action at a distance. It's only because people are used to doing it. It's immensely mysterious. And all the others do it. It would have been just as mysterious if they'd been... Mechanical medium that does all these things as it is. They're just sui generis electric and magnetic field strengths at every point in space. I think that's driven by a false metaphysics. That's what structuralism is. Well, let's run out of time. Did you want the last word? I don't think there was anything more to say. Well, thank you very much indeed for these two minutes.

1:50:00 Thank you for your attention.