Abstract
Sellars’s definition of the aim of philosophy, ‘to understand how things in the broadest possible sense of the term hang together in the broadest possible sense of the term’, corresponds to my aspirations for the subject. In this article I lay out a very different view of what realism should be, in the hope that it may contribute to that inspiring aim. The difference between our two versions of realism lies in the opposition between Sellars’s picture of two ‘images’, the ‘manifest image’ and the ‘scientific image’, with the scientific image being the ‘measure’ of what is, and the picture I offer, in which there are not sharply delineated ‘scientific images’ and ‘manifest images’, but forms of human discourse that interpenetrate and depend on one another.
My 1985 Carus Lectures were titled ‘The Many Faces of Realism’.
1
Let me begin this article with the sentences that opened those lectures: The man on the street, Eddington reminded us, visualizes a table as ‘solid’ – that is, as mostly solid matter. But physics has discovered that the table is mostly empty space: that is, that the distance between the particles is immense in relation to the radius of the electron or the nucleus of one of the atoms of which the table consists. One reaction to this state of affairs, the reaction of Wilfrid Sellars [who was alive when I gave those lectures], is to deny that there are tables at all as we ordinarily conceive them (although he chooses an ice cube rather than a table as his example). The commonsense conception of ordinary middle-sized objects such as tables and ice cubes (the ‘manifest image’) is simply false in Sellars’s view (although not without at least some cognitive value – there are real objects that the ‘tables’ and ‘ice cubes’ of the manifest image ‘picture’, according to Sellars, even if those real objects are not the layman’s tables and ice cubes).
Sellars’s view was simultaneously a realism with respect to science and an anti-realism with respect to the world of common-sense objects. ‘The objects of the manifest image don’t “really” exist’ (Sellars wrote in a letter to Smart
2
). But the opposite attitude – realism with respect to the common-sense world and anti-realism with respect to the world of scientific objects – also had its defenders. In lectures given close to the end of his life, Edmund Husserl criticized Galileo for … the surreptitious substitution of the mathematically substructured world of idealities for the only real world, the one that is actually given through perception, that is ever experienced and experienceable – our everyday life-world. This substitution was promptly passed on to his successors, the physicists of all succeeding centuries.
3
In addition to realists/anti-realists of these two sorts, realists like Sellars who are anti-realists with respect to the manifest image and realists like Husserl who are anti-realists with respect to the scientific image, there are Logical Positivists who thought reality-questions were ‘pseudo-questions’ and there are ‘quietists’ of several varieties who also reject those questions (although Wittgenstein, who is often considered a quietist, uses a piece of furniture as an example of ‘a reality’ in one of his lectures on the philosophy of mathematics). But when attitudes to the exact sciences as a whole and its relation to our common-sense world are at stake, I believe that quietism and positivism are profoundly mistaken stances, although it is not my purpose to argue that in this article. (See, however, my ‘Science and Philosophy’. 4 ) What I do want to do is, first, argue contra both Sellars and Husserl, that one can and should be a realist both with respect to scientific objects and properties and to ‘common-sense’ objects and properties, and, second, to explain what I understand that to mean in connection with a number of central philosophical issues.
Before I begin, let me say that Sellars and Husserl were both great philosophers. If I disagree with them and try to go beyond their views, that is out of respect and not out of any scornful attitude. Sellars’s beautiful definition of ‘the aim of philosophy’ – ‘to understand how things in the broadest possible sense of the term hang together in the broadest possible sense of the term’ 5 – corresponds to my aspirations for the subject, and I also share both Sellars’s and Husserl’s desire to do justice to the role of value judgements and not only of scientific theories in our lives.
I Realism simultaneously with respect to both scientific and common-sense objects and properties
In the first sentence of that old Carus lecture that I quoted at the outset, I described Eddington as claiming that ordinarily people believe that tables are ‘mostly solid matter’, in a sense of ‘mostly’ that is incompatible with their consisting of widely separated particles (or, for that matter, widely separated atoms). But how many people in Britain, where Eddington published The Nature of the Physical World, 6 didn’t know that matter consists of atoms? Maybe I overrate the average person’s level of information, but I would guess that the percentage was small. But be it a minority (as I would suppose) or a majority (as Eddington may have thought) who did not know in 1928 that atoms are spatially separated, those of us who do know that do not hesitate to use the word ‘solid’ to describe many things that consist of them. There is a branch of physics called ‘solid state physics’. By failing to distinguish between how most of us visualize tables and ice cubes and what we believe about tables and ice cubes, Eddington simply overlooked (or chose to overlook) the fact that most of us see no contradiction between being solid and consisting of widely separated particles. In a terminology I used in ‘The Meaning of “Meaning”‘, Eddington (and Sellars) conflated the extension of solid (what the term is true of) and the stereotype that accompanies the term.
I know what Sellars would have replied to this accusation. I am sure he would have said that we all operate with incompatible conceptual systems. The manifest image is our everyday system, and the scientific image, or as much as we know of it, is one some of us use a lot and others (most of us) use very little, but are prepared to defer to when appropriate.
In what follows, my purpose is to summarize some of my own realist views and not to criticize Sellars’s, even though I just began by presenting my position in contrast to his. What I shall describe is what I think a sensible realism should be in a number of areas, beginning with the ice cubes and the tables, and proceeding to the issues of realism about things very different from tables, for example, physical laws and probability, and closing with the truly contentious subject of ethical properties. In the process, I will also consider some semantical issues that have become central to the discussion of realism, and defend the claim that a realist needs to recognize that the world has many levels of form, and theoretical physics is not the measure of all things. I shall not, however, even try to explain my positions in philosophy of mathematics or philosophy of logic; those of you who are interested may want to look at the articles and my replies in the Library of Living Philosophers’ volume devoted to my philosophy and the volume of my recent work published by Harvard in 2012.
Let me begin by repeating that I see no contradiction between being solid and consisting of widely separated particles, or consisting of fields and particles, or consisting of a multi-dimensional wave function in something called Fock space, or whatever physicists may discover our table and ice cubes to consist of. Science does not, on my view, discover that there are no solid objects, no tables, no pink ice cubes; it discovers what solidity is, what tables and ice cubes are, and what color is.
Obviously, some explanations are in order.
II Semantical issues
It is not only the scientific image and the manifest image that have been thought to be radically incompatible; Thomas Kuhn famously held that scientists before and after a scientific revolution ‘live in different worlds’. Using more sober language, Rudolf Carnap thought that what a scientific theory commits one to is the existence of entities that satisfy its ‘Ramsey Sentence’ – the second-order sentence that asserts that the laws of the theory have a true interpretation that assigns the intended meanings to observation predicates such as ‘red’ and ‘touches’. But for both of these thinkers, disagreement over what laws a physical magnitude (say, kinetic energy) is supposed to obey is disagreement over what the term refers to. On Carnap’s account, if kinetic energy obeys different laws according to Newtonian physics and according to relativistic physics then Newtonian physicists and Einsteinian physicists were not talking about the same thing when they spoke of ‘kinetic energy’. For Kuhn this follows from the fact that the two theories are not ‘translatable’ into each other (although in conversation he would admit to me that he did not have a clear notion of ‘translation’).
The idea that what it is to be kinetic energy is to satisfy certain laws is a special case of what philosophers of language call a descriptivist semantic theory, that is, a theory that identifies the meanings of words (say, proper names or general nouns) with descriptions. For example, a descriptivist account of that proper name might claim that ‘Moses’ is synonymous with ‘the leader who led the Jews out of Egypt and who was named “Moses”‘, and a descriptivist account of that term might claim that (prior to Einstein) ‘kinetic energy’ was synonymous with ‘one half the product of the mass and the square of the velocity’. The publication of Saul Kripke’s famous lectures on Naming and Necessity 7 convinced most philosophers that such descriptivist analyses are vulnerable to innumerable counter-examples. I myself had argued in 1962, 8 that 19th-century physics and Einstein’s physics both gave approximately correct accounts of the same magnitude, and that whatever the status of the energy definition may have been before relativity theory, in revising it, Einstein rightly treated it as just another statement about that magnitude; he did not change the reference of ‘kinetic energy’. Around 1970, Kripke and I independently arrived at the view 9 that it is causal chains connecting language uses to things and properties of things in the external world that (given appropriate referential intentions on the part of speakers) determine the reference of our terms. But the descriptions that speakers give do not fix the reference of their terms; the world is what does that, and the descriptions are often only approximately true of the corresponding substances, objects and magnitudes. That is why it makes sense to say that even if Moses existed, he may not have led the Jews out of Egypt, and sense to say that Einstein discovered that the classical energy definition required correction – but not that Einstein discovered that there was nothing that the classical physicists had been referring to by the words ‘kinetic energy’. Because the descriptions are dethroned from their reference-fixing roles, Kripke’s and my accounts were often referred to as ‘direct reference’ accounts.
When what is at stake is not two scientific conceptual schemes such as Newtonian physics and Einsteinian physics, but the common-sense conceptual scheme (or schemes) and the current scientific conceptual scheme (or schemes), the same semantic issues are relevant. According to Sellars, for example, it is part of the common-sense conceptual scheme that the pink ice cubes are ‘continuous’; 10 if this (alleged) common-sense belief is part of the concept of being a translucent colored solid, then the description ‘colored’ semantically entails ‘continuous’. But if colors are physical properties of objects, ones which color scientists identify with types of SSRs (surface spectral reflectances), 11 and if you accept the arguments for externalist semantics, then you will not expect the nature of color to be something you can read off from something called ‘the concept of color’, whether that concept is identified with a description, or not. I see no good reason for rejecting the physicalist account of color, although that account does not pretend to conform to some a priori concept of color. Here color realism and the ‘externalist’ semantics that Kripke and I proposed fit together like hand and glove.
III Dispositions (probabilities): Do we have to find an ‘in between’ between realism and anti-realism?
So far I have spoken of realism concerning the objects and properties of the manifest image. I claimed that these are real, but obviously I reject Husserl’s claim that these alone are real, and that the objects and properties that physicists talk about, the objects and properties of the scientific image, are idealizations (‘idealities’), which physicists since Galileo have surreptitiously persuaded us to regard as the real things. My position is that both the ice cubes and tables of the manifest image and atoms and particles (including the elusive Higgs boson) are likewise real, and that the tables and ice cubes consist of atoms and whatever else science finds out they consist of. But this is not to say that all the properties of everyday objects, the objects of Husserl’s Lebenswelt, are identical with physicists’ properties. No one supposes that the property of being a book of poems, for example, is identical with some property definable in the language of quantum field theory. Having that property is supervenient on the global quantum state of the universe, but supervenience is dependent existence, not identity. Any physically possible world in which all the physical properties of the space–time continuum are the same as in our world would also be a world in which the volume of space–time occupied by the book of poems I was reading last night is also occupied by a book of poems; that is what it means to say that the property of being a book of poems is supervenient on the physical state of the world. Being a book of poems, being a human being, being an author or a reader, being the author or reader of a book of poems, are properties whose instantiation depends on the physical states of objects near and far; but they are not identical with physical properties.
There are predicates, however, which many philosophers do not regard as corresponding to genuine properties at all, but which are admittedly indispensable in certain sciences, at least in their present form. A much discussed example is the dispositional terms. Some philosophers have suggested that dispositional adjectives such as ‘soluble’ simply represent ‘inference tickets’; according to these philosophers, to say that an object is soluble is not to ascribe a physical property to it in addition to its non-dispositional properties (taking mass, charge, etc. to be such), but simple to ‘license’ one’s hearers to infer that something will dissolve if placed in water. Other philosophers have suggested that dispositional adjectives are ‘place holders’ that future science will replace with non-dispositional theoretical terms when we discover the appropriate theories. Quine famously suggested 12 that those non-dispositional replacers would be descriptions of ‘built-in, enduring structural traits’. (He had in mind speech dispositions, and the corresponding ‘structural traits’ were supposed to be states of the brain.)
Contra Quine, I see no reason to believe that a disposition to say ‘yes’ when asked ‘Is that a rabbit’ in the presence of a clearly visible rabbit is going to be explained by one and the same neural structure in all English speakers’ brains. And in the case of many disposition terms we know that the things that possess those dispositions do so for very different reasons in different cases: think of the dispositional term ‘poisonous’! In particular, the notion of a ‘reflectance profile’ that is crucial for understanding the physical nature of color does not satisfy Quine’s structuralist hopes; things have the reflectance profiles they do for a variety of different reasons. Many dispositions are extremely unlikely to be replaced by Quine’s ‘built-in, enduring structural traits’.
As for the ‘inference-ticket’ view, I don’t believe that color physics is simply issuing recommendations to adopt inference policies when it describes surface reflectance profiles. A reflectance profile is an ascertainable property of objects, one that obeys certain laws. The status of many such dispositional predicates is well entrenched in a variety of sciences. To be ‘anti-realist’ with respect to them because they do not correspond to ‘structural traits’ would be to succumb to philosophical dogmatism.
But even if we reject such strongly anti-realist views as the ‘inference-ticket’ view, there remain important views that do not fall easily into the ‘realist–anti-realist’ classification, and we need to be aware of them.
When talk of ‘dispositions’ is made precise, it is most often done by specifying probabilities. Probabilities of the relevant sort are sometimes called ‘objective chances’; and the most serious version of the realist–anti-realist disagreement here is the debate about the nature of these objective chances between advocates of ‘neo-Humean’ views of probability, especially the ‘Best Systems’ approach advocated by David Lewis and developed further by Barry Loewer, and ‘propensity’ views of probability. 13 The Best Systems view offers not only an interpretation of probability, but also an account of the nature of natural laws. The view is classified as ‘neo-Humean’ because laws (including probabilistic laws) are explained by it as regularities with certain epistemic characteristics, or, more precisely, as regularities that belong to systems with certain epistemic characteristics (e.g. simplicity, informativeness and, in the case of probabilistic laws, statistical ‘fit’). There is no objective ‘physical necessity’ in this story, and there are no objective ‘propensities’.
Unlike the inference-ticket view, the ‘Best System’ interpretation of probability and lawfulness does not have difficulty with my example of the use of SSRs in color science. To say an object has a certain surface reflectance profile is to say light of certain wavelengths will be reflected from it under certain conditions with certain probabilities. SSRs are probability distributions, and probabilities, on the Best Systems view, are objective. They are genuinely informative regularities (ones which, however, are picked out as ‘laws’ by the epistemic virtues of simplicity and statistical fit, and not by informativeness alone; although why Humeans, ‘neo’ or not, are entitled to ‘simplicity’ is something I do not understand).
The case for the rejection of the more robustly realist ‘propensity’ view (and, correspondingly, for the rejection of non-epistemic views of the nature of laws) is summarized by Loewer in a paragraph at the end of his ‘David Lewis’s Humean Theory of Objective Chance’.
14
In this paragraph he assumes that laws are sentences in an idealized formalized language with a vocabulary for speaking of space–time points and of the fundamental magnitudes and properties that are instantiated at space–time points. True sentences that satisfy the Lewisian criteria of simplicity, informativeness, and (in the case of sentences that specify the chances with which events occur) statistical ‘fit’, Loewer calls ‘L-laws’, and the ‘Humean’ position is that being a law is the same thing as being an L-law: The main competition for L-laws and L-chances is from more metaphysical views about laws and chance. It must be admitted that these metaphysical accounts fit some of some intuitions concerning law and chance better than the L-versions do. In particular the intuition that very different systems of laws and chances can give rise to the same total history of occurrent events is quite strong. If the requirement that anything worthy of the names ‘law’ and ‘chance’ satisfy this intuition then [Best System] accounts are dead in the water. But until it can be shown that this requirement is essential if laws and chances are to play the roles that science requires of them then I think we have little reason to give such deference to the intuition. And that is especially so for those who, like myself, find metaphysical laws and chances that are at once metaphysically independent of events and yet govern their evolution and rationally guide our beliefs utterly mysterious.
IV Why I am not a neo-Humean
I certainly do not claim to have a knock-down argument against the Best System account of laws and probabilities, but I find it extremely problematic. Consider the thought experiment Loewer briefly mentions: imagining ‘that very different systems of laws and chances … give rise to the same total history of occurrent events’. It is not hard to see that two physical theories might agree on which space–time trajectories can occur, but differ in the probabilities they assign to those trajectories, and perhaps this is the case Loewer had in mind. Consider, however, a case in which two theories – both simple and informative enough to be taken seriously by scientists – differ on what would have happened if a particular experiment had been performed, and the experiment could have been performed, but it is so costly that it never got performed. Perhaps one theory predicts that a ‘super-duper-boson’ would have been detected and the other theory says the ‘super-duper-boson’ does not exist, but the theories are both compatible with all the ‘occurrent events’ that actually take place. In this case, the intuition Loewer speaks of, which is my intuition, is that there is a fact of the matter as to whether the super-duper-boson would have been created if the experiment had been funded, and performed enough times that, if the first theory is right, it should have been observed with probability very close to 1. Does this involve me in thinking of laws as ‘governing’ events in some ‘mysterious’ way? It seems to me that Loewer is forcing a metaphysical interpretation onto assertions like the assertion that either the super-duper-boson would have been created or it would not, and the correct laws of physics would, if we knew them, answer that question. If saying that is ‘metaphysical’, then it seems like a pretty good metaphysics to me.
1 Levels of explanation and kinds of form
To return to an earlier example, being a book of poems is supervenient on the total distribution of physical properties over space–time points, 15 but that does make being a book of poems a physical property. Supervenience (especially global supervenience) is not identity. But even among properties that are unquestionably physical (counting geometrical properties as such, after Einstein), it is often overlooked that the properties on which a given property supervenes may not do the explanatory work that the property does. This is something I overlooked in 1958, when Paul Oppenheim and I published a paper titled ‘Unity of Science as a Working Hypothesis’; there we argued that every phenomenon that can be explained by ‘higher-level’ sciences such as psychology and sociology could in principle be explained by ‘lower-level’ sciences, and ultimately by physics. 16 Many years have passed, and I am not positive why we thought this, but my speculation is that we ‘jumped’ from the putative fact that if (as seems plausible) any event has a description in the language of fundamental physics as the behavior of certain particles and fields, or, more generally, as the time-evolution of a certain system, then the coming-into-being and the future consequences of that event could all be predicted by a mathematically omniscient super-physicist given appropriate initial data – we jumped from that putative fact – to the conclusion that, if our assumption was right, then any event has an explanation in the language of fundamental physics. Basically, we assumed that ‘if you can predict it, you can explain it’. But we overlooked two things, both of which were very much in my mind a quarter of a century later, when I wrote ‘Reductionism and the Nature of Psychology’. The two things are that a given event may have many different correct explanations, explanations that generalize to different cases, and that not all explanations speak to the questions that are of interest in a given inquiry.
In ‘Reductionism and the Nature of Psychology’, I illustrated this with a very simple example. Imagine a board with a round hole in it, say, 1 inch [2.5 cm] in diameter, and two wooden pegs (say, 2 inches [5 cm] long), one with a square cross section a little less than 1 inch [2.5 cm] to a side (‘the square peg’), and one with a circular cross section a little less than 1 inch [2.5 cm] in diameter (‘the round peg’). The round peg will go through the hole in the board without ‘forcing’ and the square peg will not.
Now consider the event of Bob’s trying to put the square peg through the round hole and failing. His try involves causing the peg to move in a certain trajectory, one that our hypothetical mathematically omniscient super-physicist could predict given appropriate initial data, and he or she would find that given those initial conditions the peg would not go through the hole.
Next consider the event of Alice’s trying to put the round peg through the round hole and succeeding. Her try involves causing the peg to move in a certain trajectory, one that our hypothetical mathematically omniscient super-physicist could predict given appropriate initial data, and he or she would find that given those initial conditions the peg would go through the hole. But what the mathematically omniscient super-physicist has explained is why the square peg did not go through the round hole and the round peg did under those precise initial conditions, and not why ‘a round peg that size and shape will go through a round hole that size and a square peg that size and shape won’t’. To explain that, one needs to realize that what particles and fields the pegs consist of, the precise forces that were applied by Bob’s and Alice’s hands, the temperature of the room, etc. are unnecessary information. It suffices to know that the pegs and boards retain their shape and rigidity and the obvious geometrical facts. Explanations seek their own level.
But why did I use this example in a paper about ‘the nature of psychology’? Well, consider a simple and useful ‘belief–desire’ explanation, say, a group of students is competing for a mathematics prize, and knows that students who don’t know any trigonometry will fail to win. So they read the trigonometry textbook. The belief–desire explanation is that they want the prize, and they believe that they won’t get what they desire if they don’t study trigonometry and that they have a reasonable chance to get it if they do. This sort of explanation does not always ‘work’, but it succeeds in a large class of cases.
According to the argument of my former self and Oppenheim (if I am right about how we were thinking), the same fact, that the students studied that textbook, could be explained by a mathematically omniscient super-physicist from initial data about the particular physical situation. And this was supposed to show that what is explained by psychology (by the belief–desire explanation) can also explained by the calculations of the hypothetical mathematically omniscient super-physicist; but it does not, because the explanations generalize to very different classes of case.
A way of viewing the fact that explanations can be on very different levels and generalize to very different classes of situation is to think of objects and events as having different sorts of ‘form’. To say that the students wanted to win a prize is to ascribe a psychological property to those students; to describe the physical make-up of their brains is to ascribe a very different kind of property. There is no incompatibility between what the mathematically omniscient super-physicist would say about the goings-on in the students’ brains and bodies and environments and what belief–desire psychology (or any other correct psychological account) would say about their actions, but very different classes of situation instantiate the two properties. The explanation of the students’ behavior we are interested in is not provided by a ‘lower-level’ science and ultimately by fundamental physics, as Oppenheim and I claimed; it is provided by the explanation at the appropriate level, which is the psychological level. Pace some publicity-hungry physicists, a theory of the fundamental particles, fields and so forth is not a ‘theory of everything’. A physical theory is not, for example, a theory of cognition or behavior. I suspect that only one who made the mistake Oppenheim and I made would think it is a theory of everything.
2 A possible misunderstanding
What I have said so far may admit of the following misunderstanding: someone could, perhaps, take what I said at the beginning of my lecture about Sellars’s pink ice cubes thus: ‘So you and Sellars agree that, as he famously put it, “in the dimension of describing and explaining the world, science is the measure of all things, of what is that it is, and of what is not that it is not”. 17 The difference between the two of you concerns what science does and does not include. You think that ice cubes (and Eddington’s table) are scientific objects as well as the particles and fields. So, for you, Sellars’s “manifest image” is not only not incompatible with the scientific image; it is, where it is not factually mistaken – e.g. neither scientists nor rational laypersons think any longer that witches exist, or that fire-breathing dragons exist – apart from such mistaken beliefs, the manifest image is a part of the scientific image.’ But that is not my view. It is true that, on my view, the scientific image is porous; the example of color science shows that colors are not just part of the manifest image, they are also part of today’s scientific image (even if their place there is still subject to some controversy). Similarly, the way in which disposition talk is frequently replaced by talk of probability distributions shows how discourse which seemed to belong essentially to the ‘manifest image’ has become a respectable part of the scientific image (and not because it uses folk-objects to ‘picture’ something of an ontologically different kind – on the theory of direct reference Kripke and I advocated, the folk and the scientists are talking about the same thing when they talk about colors, even if the scientists identify them with SSRs, and most of the folk do not know what an SSR is). Not surprising, some people (though not myself) expect that every part of the manifest image will either turn into a part of the scientific image or turn out to be just false. But turning the subjects I will close by saying something about, the subjects of intentionality and normativity, into ‘sciences’ is, at least for now, a fantasy. It is true that sometimes subjects that it seemed science could never say something about became important objects of scientific study; I do not find that a very good reason to think that every subject can or will have the same fate. But even here the boundaries between science and non-science are porous.
3 Intentionality
The case of intentionality also illustrates the porosity. The term ‘intentionality’ has been used to refer to many things, but a central case is reference. And philosophers generally take it for granted – I myself took it for granted, until I read Tyler Burge’s Origins of Objectivity – that referring requires conceptualization, and probably requires linguistic capabilities. What Burge shows – I say ‘shows’ because I find his arguments convincing – is that successful and theoretically sophisticated explanations in psychology, and specifically in vision science, presuppose the notion of perceptual representation, and that there is no good reason to demand of psychology that these explanations be recast in terms that satisfy Quine’s, or Husserl’s, or any other philosopher’s views of or qualms about representation or ‘reference’. In fact, very primitive forms of reference, of perceptual representation that has accuracy conditions, for example, representation of colors and shapes, appear in organisms far more primitive than language users. This is not to deny that conceptualization and conceptualized representation are far from being part of the ‘scientific image’ today; but the fact that, if Burge and the vision science on which he relies are right, at least some forms of reference are biologically prior to and not a consequence of or a function of conceptualization should help break the grip of the idea that reference is either (1) a mystery, or (2) something that belongs only to the ‘manifest image’ (to what Quine once called a ‘second grade’ conceptual system). There is no sharp gap between the scientific image and the manifest image, properly understood. One can be a realist about both without either assimilating one to the other or accepting an absolute dichotomy.
4 ‘Value-judgements’
I do not believe that the value judgements exemplify the porosity of the scientific image/manifest image distinction in the way that I have just claimed intentionality does. Today, only a small group of ‘Cornell Realists’ 18 believe that value judgements are in principle scientific judgements. But unless one is prepared to argue that all reasonable inquiry is in principle scientific (which, I suspect is what the Cornell Realists think), it seems clear that value inquiry is not going to be turned into a science. What I believe the study of value judgements teaches us is not that science and value judgement overlap, but that they presuppose one another. To borrow a term from John McDowell, I believe that scientific judgements and value judgements are entangled. But value judgements are of many different sorts.
Some value judgements describe. If I say that someone is cruel, most of us would agree that I have made a value judgement, but in some contexts my aim may be primarily descriptive. When a historian describes someone who lived long ago as cruel, the aim is frequently not to condemn, but simply to state a fact. And value judgements do not always convey praise or blame. If I say a scientific theory ‘lacks elegance’, I am not blaming it. Nor need I be blaming the people who thought it up; they may have good reason to discuss it. Similarly, when contemporaries described the 1755 Great Lisbon Earthquake as a terrible event, they were evaluating but not condemning or blaming. And a pure norm like ‘It is wrong to oppress widows and orphans’ does not describe. And there are epistemic value judgements – at least they have all the characteristic features of value judgements – that describe and others that state pure norms, e.g. ‘other things being equal, prefer the explanation that is simpler and more informative’. (Please note that I reject the neo-Humean assumption that we possess a metric of simplicity defined over all possible empirical theories. Simplicity is not a single ‘parameter’, but has many dimensions, some of which may be purely formal, but many of which involve the content of the theories and their coherence – another epistemic value! – with background theory or lack thereof.) For present purposes, I understand the term ‘value judgement’ as including all these different sorts. Moreover, the class of value judgements, in this sense, contains both some judgements that belong to Sellars’s ‘logical space of reasons’ and some that do not, although they also do not belong to the scientific image.
Although various forms of entanglement may involve values and facts of almost any kind, perhaps the most universal kind of entanglement involves epistemic values. Judging the credibility of an empirical claim, ranging from the kind of claim involved in a legal trial (‘The car that hit so-and-so was a Honda Civic’) to the kind involved in a cutting-edge scientific experiment (‘The particle was a Higgs boson’), and ranging from claims about individual events to ambitious theories, involves judging coherence with data, ranging from personal memories to elaborate records of data, as well as coherence with putative background knowledge (which may itself be called into question), and, of course, ‘coherence’, ‘plausibility’ and the like are value terms. High marks for these epistemic virtues are not enough by themselves to justify accepting a theory – not in a deliberately skeptical modern science, anyway – but low marks can justify the decision not even to test it. As Jacob Bronowski once said to his friend Popper (or so Bronowski told me in a conversation), ‘Karl, you wouldn’t claim that scientists test all testable theories if as many crazy theories crossed your desk every week as cross mine!’ The claim that a theory is implausible, or ad hoc, or otherwise epistemically flawed is not a scientific claim, yet science depends on such claims. And, conversely, what counts as simplicity, coherence, plausibility and the like, changes as successful science changes. If all value judgements were merely expressions of subjective attitudes then epistemic value judgements should lack all authority; but the very idea of objective knowledge loses its coherence and plausibility apart from that authority. Contrary to what is sometimes claimed, successful prediction is not the only virtue a scientific theory has to demonstrate – and even demonstrating that virtue requires appeal to judgements that involve the other epistemic virtues. The mutual dependence of claims about epistemic values and claims about empirical facts illustrates the way in which science and the vast terrain of non-scientific common sense presuppose each other.
5 Ethical values
In view of all this, what does it mean to be a realist about value judgements? In order to avoid biting off more than anyone could chew, I shall confine attention to realism about ethical values. It is simpler to say what it means to be an anti-realist than to say what it means to be a realist here. The grandfather of anti-realism about ethical statements, Charles Stevenson, 19 held that ethical sentences only look as if they stated claims that could be true or false; according to him, in reality they only ‘express attitudes’. This was presented as a claim about ethical language, and not surprisingly opponents argued that there is no linguistic evidence that the claim is right. Paul Ziff’s Semantic Analysis was a particularly powerful rebuttal. 20 Other, more sophisticated, ‘linguistic’ versions of ethical anti-realism were offered by Hare 21 and Reichenbach, 22 and are subject to the same criticism. J. L. Mackie’s Ethics: Inventing Right and Wrong avoided this controversy by holding that ethical sentences do (linguistically speaking) purport to state facts, but that this is because speakers make a metaphysical error: the error of supposing there really are such properties as right and wrong, when in reality there are no such properties. Simon Blackburn’s ‘quasi-realism’ 23 is a particularly subtle version of the linguistic version of anti-realism; Blackburn not only concedes that ethical sentences purport to state facts; he claims that it is perfectly legitimate to say that an ethical sentence (say, ‘It would be terrible if Donald Trump became president of the United States’) ‘states a fact’, ‘is true’, etc. if you agree with it. But ‘true’ and ‘fact’ bear both seriously realist and quasi-realist understandings, according to Blackburn. When a scientist says that there are black holes at the centers of all or almost all galaxies, his statement must be understood as trying to ‘fit’ the world. If it does not, it was not true in a seriously realistic sense. But if someone utters or writes the sentence about Donald Trump, his statement does not ‘fit’ or ‘not fit’ the world; the appropriate ‘direction of fit’ for ethical sentences (and certain other sentences, including mathematical sentences, according to Blackburn) is not from sentence to world but from sentence to speaker’s attitude (Blackburn even uses Stevenson’s term ‘attitude’.)
What all these forms of anti-realism have in common is best brought out by the error theory, I believe. At the end of the day, they all claim that there are words that look as if they signified ethical properties, but there are in reality no such properties to signify.
But, just as Barry Loewer’s ‘neo-Humean’ attack on realist views of probability and of causal laws depends for its force on attributing to realists metaphysical views about causality and probability that I do not think they need to hold, so likewise does Mackie’s attack on ethical realism. I do not think, for example, that someone who sincerely believes that it is wrong to do something must thereby be motivated not to do it (most over-eaters are a counter-example to this claim), or that someone who believes it would be very good to do X must thereby be motivated to do X. Yet the claim that ethical properties, if they existed, would be such that knowing that something has them is intrinsically motivating is essential to Mackie’s argument that their existence would be ‘queer’. Indeed, everything that the British empiricists and the logical positivists could not stomach has been called ‘queer’ or ‘mysterious’ by some philosophers, while fantasies of formalizing total science, induction, ‘simplicity’, etc. seem never to get called ‘metaphysical’ or ‘queer’. Given the many glaring weaknesses of empiricist arguments (not to mention empiricist metaphysics and epistemology), I see little reason still to be worried about offending empiricist sensibilities.
Still, it is reasonable to ask for an account of how ethical properties can fit into a naturalist view that is not reductionist or scientistic. Here is a very brief sketch of the sort of account that I favor.
I believe no one kind of fact or value is the basis on which all ethical values rest, although much ink has been spilled trying to show the reverse. The values that have been taken as fundamental – concern with the happiness of the great majority, concern with individual flourishing (and concern with trying to determine what that consists in), equality, justice, liberty, Thomas Scanlon’s ‘Kantian’ concern with providing reasons for our moral rules that others with the same concern could not reasonably reject 24 – obviously can and do come into conflict in many circumstances. Nevertheless, in a healthy democratic society, all of these concerns are important, and the attempt to find ways of reconciling them when they conflict is the source of as much social progress as there has been.
Obviously, none of this is a priori. Our conceptions of these values are historical products. But that does not imply ethical relativism, because historical processes need not be disguised power plays and rationalizations, even if they frequently are that. Something that American pragmatism rightly emphasized is that given certain human rights – in particular, freedom of expression (including freedom to protest and advocate) and freedom of inquiry – there can be and there have been learning processes in history. More enlightened values always depend on more enlightened view of the facts as to how things are and as to how things have been in the past, and particularly facts about injustices and their consequences. On a pragmatist view like Dewey’s, seeing ethical problems as requiring answers that are historically contingent without being arbitrary goes with accepting a more modest role for philosophy itself. Answers to ethical questions can and should be suggested by philosophers, among other people, but philosophers’ suggestions, once they enter the public arena, become just that: suggestions to be discussed, not ‘expert testimony’. In this picture, the ‘property’ corresponding to an ethical truth is the property of being a good compromise between desiderata which are themselves changing historical products, but not merely historical products because the needs they serve are ones whose satisfaction is essential to the flourishing of human communities and the individuals in them. Realism does not require that we regard our concepts, either in ethics or in science, as unchanging Platonic forms.
Conclusion
Early in this lecture I said that Sellars’s definition of the aim of philosophy, ‘to understand how things in the broadest possible sense of the term hang together in the broadest possible sense of the term’ corresponds to my aspirations for the subject. If I have laid out a very different view of what realism should be, it is in the hope that it may contribute to that inspiring aim. The difference between our two versions of realism lies in the opposition between Sellars’s picture of two ‘images’, the ‘manifest image’ and the ‘scientific image’, with the scientific image being the ‘measure’ of what is, and the picture I have offered, in which there are not sharply delineated ‘scientific images’ and ‘manifest images’, but forms of human discourse that interpenetrate and depend on one another.