The confusion infecting the realism debate arises from thinking of given realms of entities in two different ways at once, or more precisely, from conflating two ways of thinking.
The first way is familiar enough in its various guises as a metaphysical commitment, typically to a notion of substance; but it is difficult to state precisely. It consists in thinking about the entities in a given realm as existing in their own right, independently of other things which cause or, more weakly, sustain them in being. Such entities might be conceived as having the status of something like Aristotle's primary being, or at any rate as substances conceived as those things (or that thing, for monists) which exist, and can only be understood as doing so, in some sense in and of themselves. This thesis is sometimes been called 'metaphysical realism'. I shall refer to it as thesis A. Some of the more heroic moments in the history of philosophy consist in efforts to clarify it, for what is at issue, after all, is no less than the metaphysics of being. But I am not confident that there is clarification ready to hand. The following hints at the difficulties.
Understandings of thesis A fall into at least two broad categories. Formulations like 'existence in its own right' and 'absolute and ultimate existence' suggest a full Aristotelian concept of primary ousia. This is existence which is basic to other, derivative, existence; it at least explains itself, and perhaps indeed - as in theological employments of the notion - it causes itself. It is not obvious, without further taxonomy, whether the latter, stronger, notion is coterminous with the notion of necessary being, but apologists at any rate standardly so construe it. A reason for caution is that giving a negative answer to the question: could there not be a conception of contingently self-caused beings, in the sense that other facts about the universe do not necessitate their existence? is not obviously called for by the terms of the question itself. For there might be nothing contradictory in a description of the universe, whether false or not, in which the existence of certain self-caused beings is denied; and the reverse would have to be true if the entities in question were to have a chance of being necessary in the required fashion. But no-one need go out on an ontological limb in this way to get a notion of 'existence in its own right'; there is always an epistemological limb to venture along, which says that it is just a brute fact, to be accepted as primitive, that there are certain things which exist in this primary way. To make it metaphysical realism which the thesis thus understood amounts to, a bruteness claim has to be taken as a candidate for literal truth. [The idea of explanatory self-sufficiency is an epistemic notion, note; which shows the close connection between these ideas, as a supposed explanation of epistemic independence, and that thesis itself as discussed below.]
Or thesis A might be taken more weakly to mean that, having been caused, a thing X exists as a genuine individual whose dependence on other things, for example food supplies if X is an animal, is only necessary in the sense of physical law, and hence is metaphysically contingent. It might be a fact that X is in this way dependent on things external to itself; but the point is that any suitable external will do, so there is nothing to which X's dependence is metaphysically annexed. Of course it could be argued that X is dependent on a class or classes of things - say, foodstuffs - but this does not alter the descriptive adequacy of treating X as an individuable existent. Consider a pebble: in what sense can it be understood as a dependent individual once its causal history - its having been produced by cooling lava, say - has been discounted? In this guise the thesis aims at asserting the ontological irreducibility of X to its material and efficient causes, and the ontological definability of X in a scheme of things, with its own path through a world, say, or its own separately countable membership of it. This second aspect of the notion is a rich one, for it plays a part in a story whose other themes - individuation, identity, particularity - are familiar indeed. The idea of individual substances has its busiest employment precisely here. The connection between the two aspects is that asserting X's independence in this sense means that X is to be understood as not essentially or internally related to anything else, of which it must therefore be considered part or along with which it has therefore to be individuated. So it is a true individual.
The question of how, if at all, these two understandings of thesis A are related is a matter of controversy. One can hold some version of either without any entailment to versions of the other. The chief reason is that an ontology [a theory of what there is] can select a range of individuable, persisting and therefore reidentifiable particulars, whose status as such is entirely determined by their relation to a theory or point of view which finds it convenient to treat them as such, without concomitantly holding that these things are in any way ultimate. This is the case with the medium sized dry goods constituting the ontology of mundane experience. No-one I think would now claim that according them the status of basic particulars in our scheme (see Strawson, Individuals) means that they are the ultimately existing entities simpliciter. This would imply, for one thing, a new order of instrumentalism about physics. Given that the basic particulars of a scheme might well be epistemically but not ontically basic, and in acknowledgement of the difficulties attaching to getting to the latter from the former, some - even some contemporary (see the familiar pessimistic views of McGinn, Nagel) - philosophers hold that the veil of the former is impenetrable, and we are condemned to ignorance about how things really are.
There have however always been more optimistic metaphysicians, whose ambition is to identify what is or has to be basic in the sense of the first understanding of thesis A, and then to explain how whatever seems basic in the second understanding of A is connected to it - perhaps, at the neatest, by being smoothly reducible to it. The denotata of referring terms in folk psychology, for example, are for some [the Churchlands e.g.] dummies, typically misleading ones, for referring terms in a future perfected science which will pick out something ultimate, deeper even than the physiological level at which attempts at explanation currently aim. [There are competing candidates: Leibniz's monads are an historical example. Only those persuaded that physics is metaphysics will think that fluctuations in the quantum field or superstrings - or whatever next - have that sort of ontological status.].
The serious difficulty with A is that there just does not seem to be a precise way of making the notion of 'ultimate existence' or 'existence in its own right' intelligible. It obscures rather than helps to speak, as we have seen, about ultimate things as uncaused, or self-caused, or necessary, or primary, or basic, or a brute fact. The argument of the metaphysicians is that because there is something, there must be something ultimate, in a sense of 'ultimate' vaguely connoted by the foregoing expressions; and they draw this conclusion because they feel that a 'ground' of being is needed for what there is, which either needs no ground itself or is its own ground. [Compare styles of cosmological argument, which employ notions of causality and contingency to just this end.] And anyway, talk of grounds keeps us firmly in the realm of metaphor. Another, allied, symptom of the problem is that efforts to give this family of notions content turn quickly into appeals to a different notion, namely, epistemological independence, discussed shortly. This is shown by ready allusions to 'explanatory' ultimacy, which at least often substitutes for the purely existential account being sought. But whereas questions of explanation mark out a logical space in a genuine problem, the metaphysical side of the ledger seems to stay blank.
As to the second understanding of thesis A, it seems a commonplace, and an unexceptionable one, that something must serve as the basic nodes of a scheme, to which reference can be made and by which its historico-geographic contours can be mapped; but as we see in the case of our dry goods world, such an ontology is determined by the schemers and their needs, and we live familiarly with any number of sometimes only partially commensurable such schemes, and their vocabularies, in our daily lives: for example, the perceptual scheme of medium sized dry goods, the explanatory schemes afforded by the biological and physical sciences, the folk-psychological scheme of interpersonal interpretation, and the sociological scheme of social explanation. It is a bold thesis which says that any one of these takes, or in a perfected state will take, reductions from any or all the others, or that all are smoothly saved, if we only knew how to recognise it, by a complete description of something else which is ontologically primary. The common sense belief, of course, is that the dry goods world imposes itself on us, rather than we on it, and our scheme carves it at its independently existing joints, because it has to: but we need only remind ourselves of Russell's remark, 'common sense gives rise to science, and science shows that common sense is false', to recognise that whatever A-type thesis we try to evolve from our epistemic needs, it had better not be precisely that one.
The second sense of 'independence', and the one which I argue is genuinely at issue in the 'realism-anti-realism' debate, is epistemic independence. Someone applies such a notion if he holds that that the entities in some realm exist independently of any thought, talk, knowledge or experience of them. Call this thesis B. Often B is expressed in terms of the 'mind-independence' of given entities. When those who discuss realism mistakenly contrast it with idealism, it is clear that they have mind-independence in mind as the chief characteristic of realism. This way of describing realism accords well, however, with more familiar statements of the position in terms of evidence-transcendent truth.
To make out my claim that B is what really matters in the realism debate, and at the same time to locate the issues with respect to more familiar ways of setting them out, I need to widen the scope of the discussion.
The question we are addressing is: what is the realism debate really about? According to the current orthodoxy prompted by Dummett, it is primarily about language, truth and logic, and much careful argument has gone into persuading us that if we think it is primarily about anything else we are misled. The claim is that what chiefly threatens to mislead us is failure to break free of traditional concern with questions about what exists. We should instead see the debate, the orthodoxy tells us, as an opposition between, on the one hand, the view that to understand a sentence is to grasp its truth conditions, where truth and falsity are understood as epistemically unconstrained properties of what we say or think, and on the other hand, the view that to understand a sentence is to grasp its assertion conditions. Much of the discussion about realism and anti-realism has accordingly focussed on these issues.
I agree that metaphysical issues are not primary, but neither it seems to me are semantic issues. In my view what is primary is the epistemological question of the relation between thought and language, on the one hand, and on the other hand the entities or realms of entities over which they range. To demonstrate this I shall use as a foil Dummett's thesis that realism turns on a commitment to a truth-conditional theory of meaning where truth is understood as evidence-transcendent.
Dummett began by thinking that commitment to the principle of bivalence is the hallmark of realism, and its rejection therefore as the hallmark of anti-realism. He later came to hold that bivalence is not sufficient for realism, although it is necessary; what is additionally required is acceptance of a semantic theory setting out the particular classically-based manner in which the statements in a given class are determined as true or false . Rejection of bivalence remains, however, characteristic of anti-realism.
This change of emphasis results from what are now, thanks to the detailed debate, familiar points about why a rethinking of the connection between realism and bivalence is necessary, which show that commitment to recognition-transcendent truth does not entail commitment to bivalence, so that if for other reasons it proves desirable to abandon bivalence, a notion of recognition-transcendent truth nevertheless survives. One can be a realist, in short, whether or not one thinks that there are exactly the two jointly exhaustive truth-values 'true' and 'false'. Dummett's earlier equation of realism with acceptance of bivalence was motivated by the converse relationship, namely the apparent fact that bivalence entails recognition-transcendence; which, whether or not it is right, at least appears plausible. An implication of the detachability of bivalence and recognition-transcendence, however, is that one can have a view of the relation between the truth-values of sentences and the presence or absence of their asserters' capacities to determine what they are, which does not follow automatically from one's choice of semantic principle. This raises questions about the nature of the relationship as Dummett has it, for in his view choice of semantic principle determines the nature of the relationship between the truth-values of sentences and the capacities of language-users to identify them.
Dummett claimed to have noticed that what are often described as two quite different debates, the realism-nominalism and the realism-idealism controversies, in fact share a certain form ; and then to have noticed that they share this form with other debates also, for example those which concern the reality of the future and past, mathematical objects, and values . The common feature seemed to be that a realist in any of these different subject areas is committed to bivalence. Dummett argues that to take this commitment as the mark of realism is 'preferable' to treating realism as an ontological thesis, in which the commitment is to the existence of entities of certain sorts, because some species of realism, for example those about the future or ethics, 'do not seem readily classifiable as doctrines about a realm of entities' . On this ground Dummett concludes that 'in every case we may regard a realistic view as consisting in a certain interpretation of statements in some class' (my emphasis) .
The interpretation in question is described by Dummett in terms of a classical two-valued semantic theory specifying how the semantic values of statements are determined by the values of their parts together with their arrangement. Any theory of meaning based on such a semantics will be a truth-conditional theory which is 'objectivist' about truth, that is, which is committed to a sharp distinction between questions about the truth of a statement and questions about anyone's having grounds for taking it to be true. This means that the plausibility of the semantic theory can be tested by assessing the plausibility of the theory of meaning based upon it. Again familiarly, it is Dummett's contention, as supported by, among other arguments, the 'challenges' over acquisition and manifestation of linguistic competence, that the objectivist-truth-conditional theory of meaning will not do .
What is pivotal in this train of reasoning is that it identifies classical objectivist truth as the key to realism. But this, it seems to me, is a mistake. For the notion of truth in play crucially depends on a pair of prior commitments, one metaphysical and the other epistemological; it is these, and especially the latter, which really do the work in what Dummett describes as realism. And when one recognises this, one sees that Dummett has run together quite different issues as 'realisms'. His blanket definition arguably obscures rather than clarifies what is at stake in each of the different debates.
Dummett himself states that the realist conception of statements in some class turns on the idea that their truth-values are settled by knowledge-independent states of affairs. 'The very minimum that realism can be held to involve,' he says, 'is that statements in the given class relate to some reality that exists independently of our knowledge of it, in such a way that that reality renders each statement in the class determinately true or false, again independently of whether we know, or are even able to discover, its truth-value' . The immediate interest in this for Dummett is that such a commitment shows - leaving aside problems of vagueness - that the statements in question are bivalent, because they are selected as determinately true or false by an independent reality which settles the matter without reference to any cognising subject. And this is why he describes realism as a semantic thesis: it is 'a semantic thesis [because it is] a doctrine about the sort of thing that makes our statements true when they are true'. But he goes on to unpack the expression 'sort of thing' in a way which shows that its being a semantic thesis comes courtesy of something else: 'the fundamental thesis of realism, so regarded, is that we really do succeed in referring to external objects, existing independently of our knowledge of them, and that the statements we make about them are rendered true or false by an objective reality the constitution of which is, again, independent of our knowledge' .
What this characterisation immediately shows is that the conception of truth at work cannot be understood otherwise than in terms of the logically antecedent metaphysical and epistemological theses which determine its content. The theses are simply stated. First, there is a determinately charactered reality. Secondly, truth-values are properties of statements which they possess as a result of standing in certain definite relations - the usual candidate is some sort of 'correspondence' - to that reality, the relations being external ones as required by the independence constraint; which is, thirdly, that both the reality and the truth-values of what is or could be said about it are independent of any knowledge of either. So the semantic theory (the theory of truth and reference) presupposes the existence of a determinately charactered reality (the metaphysical thesis) which is independent of our knowledge and confers, independently of our means or even our capacity for getting such knowledge, truth or falsity on whatever could be said about it (the epistemological thesis).
This epistemological thesis is in essence a negative one. It says that our conception of reality is in no way constrained by our capacities to know anything about it. More precisely, it says that the knowledge relation is external, contingent and limited; it states (a) that the objects of knowledge can and for the most part do transcend our powers of access to them, and (b) that the sense of remarks about the existence and character of these entities or realms is not governed by considerations relating to our epistemic powers. In the anti-realist view it is the incoherence of (b) which underlies the incoherence of realism, as we shall see later. Among other things it makes realists construe (a) as saying that the independence of objects of knowledge from acts of awareness of them entails the independence of objects of knowledge from knowledge tout court. There is no such entailment: which is the starting point for a story to be told elsewhere .
What is claimed or denied about the relevance of epistemic constraints is neutral with respect to finer-grained accounts of what those constraints are. At very least they embody a demand that whatever is required for a conception of some object of discourse, it should lie within the ability of discoursers to get it - with it being allowed that the results of epistemic co-operation, yielding resources which can be shared and distributed courtesy of language, count among discoursers' possessions. Given inherent limitations on individuals' powers of perception, reason and memory - a finitary predicament which imposes narrow boundaries even on what the community of discoursers can do co-operatively, as a whole - that demand is an austere one. It is what identifies our problem in the theory of knowledge: the problem of trying to understand belief-acquisition and justification, given our strategic need for beliefs whose content often exceeds the empirical grounds there can be for holding them.
The negative epistemic thesis has it that we can attribute possession of certain concepts to ourselves without having to provide, or even to possess, grounds for that attribution. It is natural to express this in terms of the meaning of the expressions we use in applying such concepts, not least because the most tractable - and often the only - way of specifying the content of a concept lies in inspecting what we say. But what a theory which takes this route turns on is the prior commitment to there being truth-conferring (and so: meaning-conferring) states of affairs whose existence and character is independent of our knowledge of them; which is why a realist holds that language understanding is not constrained by the terms of some epistemological story.
On Dummett's order of exposition, if one accepts a commitment to bivalence and knowledge-independence of truth-value, one is thereby committed to holding that there is a knowledge-independent reality which makes statements determinately true or false. In this way the thesis about truth and its semantic embedding appears to be the decisive factor. But the logical order of dependence among these commitments is, as the foregoing remarks suggest, the reverse of his order of exposition. The crucial commitment is to there being knowledge-independent states of affairs, for without this view already in place for the semantic thesis to presuppose it, that thesis is empty: we have no other way of characterising the concept of truth required.