Realism, Modality and Truths about the Past
In the fundamental sense of the word 'realism' and whenever that word is applied to the case of the past, realism about the past is the claim that, when s is a statement in one of the past tenses :
(R) (s is true · there is no evidence for s ).
Appropriately enough, anti-realism is its negation, i.e.:
(AR1) ¬ (s is true · there is no evidence for s )
or, alternatively, the positive conditional claim that:
(AR2) (s is true Æ there is evidence for s ).
I believe that anti-realism about the past is false under both construals and I shall defend realism about the past against them. I obviously take realism to be a modal thesis about the relation between truth and its recognition, a thesis which says that the truth of statements in the past tense is independent from the obtaining of the sort of evidence which we normally rely on to recognize that they are true : testimonies, reports, documents, memories, and so on. It follows from this that the most promising way of arguing in its favour is to meet the now familiar Dummettian 'manifestability challenge'. This is what I propose to do here. I shall address that challenge directly and conclude, contrary to Michael Dummett's anti-realist, that (R) is true.
I shall begin with a few words of caution regarding the content of the challenge. It is often taken for granted that anti-realism of the Dummettian variety, either negatively or conditionally construed, is inferred from the provisional claim that no one has yet managed to show, or has yet even made clear, how one could manifest an understanding of the meaning of particular instances of (R), instances such as :
(1) ("Caesar crossed the Rubicon" is true · there is no evidence for "Caesar crossed the Rubicon").
Dummett and other advocates of semantic anti-realism like Crispin Wright and Neil Tennant, have indeed repeatedly insisted that the anti-realist's challenge must be met within the theory of meaning, that theory being conceived by them as a theory of understanding. (1) It would never-theless be utterly mistaken to infer from that prescription that one could vindicate realism about the past by showing that one understands the meaning of (1), or of any other relevantly tensed instance of (R) for that matter. It is clear that we are perfectly able to understand the meaning of false modal claims, but it is equally clear that no one may be satisfied with the behavioural expression of that understanding, or with the assurance that it is possible to display it when sutably prompted, or with the indication that some complex set of practical abilities or other constitutes the understanding. What the realist about the past must show, obviously, is that realism about the past is true and not merely that we are able to understand the meaning of the statement which expresses its fundamental claim, i.e. the meaning of (R) (and of its various instances). Not only is it utterly insufficient to show that it makes sense to think, let alone to suppose, that (R), or that (1). It is, as a matter of fact, entirely beside the point to pursue that line of thought. The question is whether or not we may confidently do so, whether or not there is a non-question-begging argument whose conclusion is, say, that (1) and, more generally, that (R).
Worse still, advocates of anti-realism often claim that the question they pose is about our capacity to understand the meaning of statements whose truth may transcend their verifiability by us. (2) This is a quite misleading way of voicing their worries about the relation between truth unconstrained by verifiability, or truth tout court, and meaning. It is obvious that we do understand the meaning of statements which could both be true and unverifiable by us. We do, as a matter of fact, understand the meaning of tensed statements like "Caesar crossed the Rubicon" and that of countless other ones, which are well-formed and contain occurrences of verbs conjugated in one of the past tenses, independently of our ability to deter-mine whether or not they are true. Dummett's anti-realist cannot possibly be claiming that statements in the past tense for which there is no evidence either pro or con are meaningless. He is not a crude verificationist who would flatly deny that unverifiable statements have a meaning or could have one. His problem is definitely not with our understanding of them, or with our ascriptions of meaning to them, but with the common sense supposition that they may be true independently of our capacity to acknowledge that they are.
So, to sum up, the problem is not to show how we could manifest our understanding of the meaning of "Caesar crossed the Rubicon" provided that there was no available evidence for it in the past, that there is no available evidence for it now and that there will never be any in the future. The problem is not, either, to show how we could manifest our under-standing of the meaning of the modal statement which says that that tensed statement, "Caesar crossed the Rubicon", might be true even though there isn't anything, at any point in time, which would allow us to determine that it is true. The problem is to argue that there is a warrant for that last supposition ; or, better, that it isn't, in fact, a mere suppostion, but a plain truth.
Since the realist must show that we have a means of manifesting our knowledge that (1) is true and not merely that we understand its meaning, he might quite reasonably be tempted to try to meet that challenge by arguing that we can manifest our knowledge that the truth-conditions of (1) are satisfied. He may try to do this by showing that we have the capacity to determine the truth-value of (1) in some more or less mechanical way, in a finite number of steps.
There is no denying the appeal of such an endeavour and the force of the result it would yield if it could be obtained. If we could display a mechanical or quasi-mechanical ability to recognize that the truth-conditions of (1) are satisfied, then, obviously, we might immediately conclude that (1) is true. Furthermore, it would directly follow from the established modal claim that the truth of "Caesar crossed the Rubicon" is indeed independent from the obtaining of evidence in its favour. If that statement may be true even though there is no available evidence for it now, and indeed whether or not there was any in the past and there will ever be any in the future, then there is no reason to think that its truth is linked in some privileged way to the possibility of its past, present or future recognition by us. The crux of the matter is to find out how we may manifest our knowledge that the truth-conditions of (1) obtain in that particular way, i.e. mechanically, or quasi-mechanically, by exercising some specific recognitional skills.
The suggestion that we possess clear-cut recognitional skills which would automatically yield the desired result if we were to exercise them appropriately and correctly does not offer a promising perspective. The seemingly less behaviouristic contention that we possess a set of practical abilities which constitute the required implicit knowledge, although no specific behavioural output is distinctive of our possession of them, is not promising either. We cannot determine that (1) is true by the exercice of clear-cut recognitional capacities and we cannot equate the knowledge that it is true with the possession of abilities which could be appropriately, albeit diversely, exercised. Knowing that the truth-conditions of "Caesar crossed the Rubicon" could obtain undetectably - as opposed to knowing what it would be like for them to obtain undetectably - is simply not a knowledge which can be constituted by clearly identifiable recognitional and decisional skills. (3)
The following argument will be sufficient to prove the point. Suppose we are asked to manifest our knowledge that the truth-conditions of (1) are satisfied. Obviously, by the very nature of the case, these truth-conditions must be constrained by verifiability. We are asked to show that there is a warrant for (1) and not that the truth of some instance of the schema expressing the defining thesis of realism could itself be beyond our reach. So these truth-conditions may not be what some have called 'realist' truth-conditions. (4) It would be both redundant and pointless to require that some manifestability challenge be met with respect to a putative knowledge of the unconstrained truth-conditions of the modal thesis and of its instances. It would be redundant because (R) and (1) already are an expression of realism. It would be pointless because the anti-realist's complaint is precisely that, unless some argument can prove that this is not the case, realism is, both semantically and conceptually, beyond our reach. There would be no point in requiring us to be able to manifest a knowledge that, say, it is possible that (1) is true whether or not there is evidence in its favour.
What we are required to show, on the contrary, is that there is some such evidence. The crucial question, then, is whether or not there are any recognitional abilities at our disposal, the exercise of which could count as a manifestation of the knowledge that the constrained truth-conditions of (1) are satisfied, or could even constitute it. I am not merely asking, here, whether or not we can somehow manifest that we know that (1) in some yet entirely unspecified way, but, specifically, whether or not we can, either hic et nunc or at least in principle, activate some executive and recognitional capacities with respect to (1) in order to manifest that we know that its justification-conditions are satisfied. Can we even show that we could acquire these capacities in case we would not possess them? Are there, as a matter of fact, any such capacities?
I think it is quite clear that there cannot be any. The very idea that we could help ourselves to a more or less mechanical means of deciding the truth of (1) is surely a muddle. Our grasp of the possibly verification-trans-cendent truth-conditions of "Caesar crossed the Rubicon" is not a grasp which could consist in an ability to recognize that a particular state of affairs could make that statement true whether or not we are now able to recognize that it is true, or had that ability in the past, or will acquire it in the future.
That is not, though, the end of the story and there is still room for an argument in favour of (1), but let me first explain why the suggestion that we have the ability to recognize or identify a truth-conferring state of affairs which, once acknowledged by us, would unable us to determine that (1) is true, leads to a contradiction.
This will perhaps be made more vivid if we recall that, provisionally at least, we must take the modal statement (1) to be equivalent to :
(1*) ( "Caesar crossed the Rubicon" is true · "Caesar crossed the Rubicon" is pro tempora undecidable).
For the time being, the joint supposition that "Caesar crossed the Rubi-con" could be undecidable and that there could exist somewhere, hidden from us, some document or testimony in its favour which we entirely ignore, may not be entertained. The equivalence of (1) and (1*) follows from the fact that the undecidability pro tempora of "Caesar crossed the Rubicon" cannot be based on our ignorance of some evidence which could (allegedly) exist in its favour, as if our own epistemic limitations would account for that ignorance. The reason is that the possibility of unaccessible evidence is no less recognition-transcendent than the possibility of unaccessible truth. Since both notions are anti-realistically suspect, we may not appeal to them when looking for an argument in favour of the modal thesis unless we beg the question entirely in favour of realism.
An analogy with the case of mathematics may help to bring this point across. Anyone who claims that there may be a proof of Goldbach's conjecture unknown to us qualifies as a realist with respect to that arith-metical statement. What such a person is saying or, at least, what her statement implies, is that the conjecture may be true although we are still unable to find out that it is true, the reason being that our mathematical abilities have not yet taken us far enough to unable us to grasp a proof of it. The fact that she is saying that the conjecture may be true in virtue of a proof hardly makes her an opponent of realism. On the contrary, she is putting forward a realist or platonist notion of mathematical proof, just as anyone pointing out that there could be some yet unaccessible document or testimony in favour of "Caesar crossed the Rubicon", is putting forward a realist notion of empirical and historical evidence. I think that it is precisely this latter notion which must be made respectable in order to vindicate realism about the past. It certainly cannot be taken for granted.
The moral to draw, at least provisionally, is the following. If there is no evidence for "Caesar crossed the Rubicon", then we cannot find out whether or not it is true, the reason being that there is simply nothing to rely on, or even to get started. This first contention stems from common sense. It actually is an outright expression of it. It is not, though, anti-realistically suspect as such. Both parties engaged in the debate over realism about the past will readily accept it. The converse, according to which only the lack of evidence for our past tense statement about Caesar may account for its undecidability pro tempora, i.e. for our incapacity to find out, at some particular point in time, whether or not it is true, is far from being obvious. This being noted, we cannot object that it is false by simply pointing out that there could after all be some hidden evidence which could make "Caesar crossed the Rubicon" true, unless we beg the question entirely in favour of realism. We have to argue for that possibility. Until we can make a case for that supposition, we must assume that the second conjunct of the conjunction which falls under the scope of the modal operator in (1*) is indeed equivalent to the second conjunct of the conjunction which falls under the scope of the modal operator in (1). At least provisionally, we must assume that the possibility of the unrecognizability of truth cannot differ in any respect from the possibility of the unavailability of evidence. (5)
Let me now return to the possibility of recognizing that the justification-conditions of (1) and (1*) are satisfied. It is quite clear that we cannot both find a justification for '"Caesar crossed the Rubicon" is true' and for 'There is no evidence for "Caesar crossed the Rubicon"'. It is also clear that we cannot both find a justification for '"Caesar crossed the Rubicon" is true' and for '"Caesar crossed the Rubicon" is pro tempora undecidable'. The second impossibility is striking, for no statement is both be true and undecidable, either pro tempora or otherwise. (6) The second may be less striking at first sight, but our inability to recognize a particular state of affairs which would make "Caesar crossed the Rubicon" true in the absence of evidence is nevertheless our lot for the following reason.
Suppose that we look for a justification of the first conjuncts of (1) and (1*). If we were to find one, we might confidently claim that "Caesar crossed the Rubicon" is true, but we could not claim, then, either that there is no evidence for it or that it is pro tempora undecidable, for we would now be in a position to decide the statement on the basis of the very evidence we would have found or retrieved and which would justify our claim that it is true.
Suppose now that we look for a justification of the second conjuncts of (1) and (1*). In the first case, we would have to check the absence of evidence for "Caesar crossed the Rubicon". This certainly cannot be done in a finite number of steps, but if, per impossibile, we could somehow manage to survey all the states of affairs which could be relevant to the determination of the truth of that second conjunct and reach a negative conclusion, we would thereby possess a justification for the negation of the first, i.e. for '¬ "Caesar crossed the Rubicon" is true'. It would thereby be impossible to justify the first conjunct. It would indeed be pointless to look for one. '" Caesar crossed the Rubicon" is true ' and '¬ "Caesar crossed the Rubicon" is true' may not both be true.
Similarly, in the case of (1*), if we could acknowledge the undecidability pro tempora of "Caesar crossed the Rubicon", the possibility of arguing in favour of the truth of the very same statement (at the same time t) would thereby be precluded. What is at stake here is the prospect of vindicating the claim that there could be some particular state of affairs which could justify the claim that a sentence in one of the past tenses could have both properties, that of truth and that of undecidability, and it is quite clear that there is just no such prospect.
Of course, it is possible for "Caesar crossed the Rubicon" to be true and it is also possible for that same statement to be undecidable or to be left without any warrant in its favour, but this is an altogether different matter. We certainly cannot infer from this that it may both be true and undecidable, or that it may be true in the absence of evidence. It is notoriously illegitimate to infer a statement of the form ' (p · q)' from one of the form ' p · q'. Although ' (p · q) Æ ( p · q)' is a theorem of modal propositional logic, its converse is not.
So far, then, we are forced to conclude that we cannot credit ourselves with a knowledge that the truth-conditions of (1) obtain. There is just no way of conclusively establishing the truth of (1) by appealing to executive and recognitional skills. The same remark may be made with respect to all instances of the modal claim (R). As a matter of fact, we are even forced to conclude that we have not yet managed to manifest an understanding of the meaning of the realist modal claim, for if, for the sake of argument, we assume with Dummett's anti-realist that only the display of executive and recognitional capacities may count as a manifestation of grasp of meaning, or may even constitute it, this is the end of realism. Not only is realism false, it also appears to be utterly unintelligible.
I shall now counter this conclusion by proposing an argument in favour of realism about the past. I shall defend the idea that statements in the past tense may be true at the time of their utterance whether or not there was, is or will be any evidence for them at that time, and that they may be true at any subsequent time whether or not there was, is or will be any evidence for them once they have been uttered.
How should we argue for this? Since we cannot detect that truths may remain undetected or verify that they may exceed all possible verification by identifying particular states of affairs which would confer truth on specific instances of (R), we must argue for these instances without appealing to recognitional and executive capacities. This has been our lesson so far : no capacities of this kind may account for our knowledge that the truth-conditions of the realist's fundamental contention are fulfilled. None may ground the modal claim. The lesson readily applies to the case of past tense statements and is in outright contradiction with McDowell's suggestion that we could meet the manifestation challenge because grasping possibly verification-transcendent truth-conditions is comparable to exercising the recognitional skills we indeed get the chance to exercise when those conditions are not transcendent (McDowell 1978: 138-140).
This must surely be a muddle. How could we ever detect that past truths may remain undetected? It isn't simply, as Wright (1987 : 102) nevertheless correctly points out, that "[...] knowing what it is for a truth-condition to obtain undetectably is not knowledge which can straightforwardly be viewed as of a type with certain clear-cut recognitional skills". What matters, since we are dealing with a modal claim, is that knowing that such a condition could obtain undetectably isn't a knowledge of that type either.
The leading idea behind the anti-realist's negative conclusion to the effect that no one has yet managed to meet the manifestation challenge is that, unless we can prove the contrary, the capacities which we possess and may exercise whenever we take past tense decidable statements into account, cannot, it seems, have counterparts whenever we turn to past tense pro tempora undecidable statements. This remark presupposes that we cannot argue for realism unless we find such counterparts. (7) The gist of McDowell's answer to the manifestation challenge reveals that some realists may even be willing to look for some. It is as if there was no other way to argue in favour of (1) than by showing that we may legitimately go from (i) the grasp of a state of affairs which is open to view and acknowledgeable at least in principle, and which confers truth on "Caesar crossed the Rubicon", to (ii) the conclusion that the very same state of affairs could confer truth undetectably on the very same statement, via the exercise of capacities to recognize the truth of (1) which should be modelled upon our capacities to recognize the truth of "Caesar crossed the Rubicon" if it were decidable.
I agree that we cannot meet the challenge in this way. Such capacities are not on a par and the corresponding attributions of recognitional skills to us are not on a par either. There is nevertheless another way to do it and I shall now turn to a positive argument in favour of (1). It is sufficient, in order to go from the first grasp to the second, to acknowledge that the relation between the truth of "Caesar crossed the Rubicon" and the availability of evidence in its favour at any time t, once it has been asserted, is a contingent matter. The difficulty in the argument from contingency is to show that it is not merely the equivalent of Dr Johnson's kicking the stone, and that it does not beg the question by presupposing what must be established, namely that there may be a gap between past truths and the possibility of their recognition by us.
Here is the argument. Any theory of nature, either folk physics or genuine physical science, is a theory of natural contingencies. We learn from these theories of the workings of nature that evidential relations between us and events which occurred in the past are contingent. They tell us, or at least allow us to infer, that the circumstances under which we may come to know that a past event occurred are naturally contingent. It follows from this that the evidence which may be available either for or against "Caesar crossed the Rubicon" at the time of its utterance and our ability, at any time t, once it has been uttered, to recognize its truth on the basis of that evidence, are both contingently connected with its truth. Since they are only contingently connected, we could very well be without means to determine whether or not "Caesar crossed the Rubicon", once it has been asserted, was true in the past, is true at the time of its utterance or will be true in the future. It follows from this that the modal claim (1) is true. Since this argument is supporting a modal claim, let me call it a modal argument from contingency. I shall now lay out its steps carefully:
MODAL ARGUMENT FROM CONTINGENCY
Dr Johnson wasn't putting forward a modal argument when - to borrow Boswell's words - he pretended to refute Berkeley's proof of the non-existence of matter by stroking his foot with mightly force against a large stone till he rebounded from it. I have neither argued nor claimed here that some event did in fact happen in the past, of which we have at the present time no inkling. Unlike Dr Johnson, I take realism to be a modal thesis which requires substantiation. My point is that, contrary to what anti-realists claim, it isn't necessary, in order to argue for (1), to rely on a truth-value link thesis.
The advocate of realism who argues from the truth-value link thesis tries to derive (1) from the existence of a systematic truth-value link between :
(a) Caesar crosses the Rubicon
(b) Caesar crossed the Rubicon.
He claims that if (a) is true at t0, then (b) is true later on at t1. He claims that because there is a truth-value link between (a) and (b), we can legitimately proceed from a conception of what it is to have evidence for (a) at t0 to a conception of what it is to have a warrant for (b) at time t1 and, from there, to a conception of what it is for (b) to be true independently of our acknowledging such a warrant at t1.
But this is a mistake. First of all, as I have already remarked, one cannot argue for the modal thesis from conceptions. One has to show that there is a ground for them. Furthemore, the problem with this suggestion is that the idea that (b) is as a matter of fact true independently of any warrant, either at t1 or a later time t2, cannot possibly help. It is a mistake to claim, as Wright (op. cit. : 14) does, that one must understand the truth-conditions of "s is [emphasis mine] undetectably true" in order to grasp "the contingent tie between s's truth and the availability of evidence" which is essential to realism. I agree that if (b) is indeed true beyond all verification, then (1) follows. If Caesar crossed the Rubicon and everyone ignores it at t1 because there is no available evidence for it at that time, then it is possible that Caesar did exactly that and that everyone ignores it at t1 because there is no available evidence for it at that time. The problem is that we could deny that (b) has both properties, namely truth and undecidability, without thereby denying that it may have both, i.e. without denying the defining thesis of realism.
The position I have defended does not rely on the truth-value link thesis and is therefore immune from that objection. On the contrary, the argument from contingency tells in favour of a truth-value gap thesis according to which (b) may very well not be true at t1 even if (a) is true at t0. The argument from contingency allows us to conclude that (a) may be true whether or not there is any evidence for it at t0 and whether or not there is any evidence that there will be evidence for it by the time we get to t1. The same reasoning applies, mutatis mutandis, with respect to (b) at times t1 and t2.
It is very clear that I have proposed to ground realism on natural possibilities, in the sense that the possibility of the combined truth and pro tempora undecidability of past tense statements follows from our knowledge of the workings of nature. Epistemic modality is at stake when we claim that it is possible that a state of affairs obtained in the past of which we had, at the time of its occurence, no inkling, of which we have no inkling now and of which we shall perhaps never have any even at the end of ideal Peircean scientific inquiry. The realist could be looking for a stronger, logical or metaphysical possibility and object that, although what is naturally possible is thereby logically possible, the realist is required to ground the metaphysical or logical possibility captured in (R) directly, without relying on epistemic possibility. I agree that I have not provided such a direct argument here. On the other hand, I do not see why something of that sort should be required.
(1) Dummett's anti-realism is expounded at lengh in the articles collected in Dummett (1978), "The Reality of the Past" being reprinted as chapter 21, at pages 358-374. It is espoused most remarkably by Crispin Wright in Wright (1987) and by Neil Tennant in Tennant (1987). On the theory of meaning being conceived as a theory of understanding, see Dummett (1976). On the need to meet the challenge within such a theory, see Loar (1987).
(2) See e.g. Dummett (1978 : 17, 225, 362-363), Wright (1987 : 22-25, 102-103), Tennant (1987: ch. 3, 11, 12) and Loar (1987 : 82).
(3) Knowing what it would be like for an a to be an F is obviously different from knowing that it is possible for an a to be an F.
(4) The expression is Devitt's in Devitt (1984 : ch. 12).
(5) The same remarks may be made, mutatis mutandis, about any appropriately tensed instance of (R).
(6) There is the case of Gödel's undecidable formula which is both true and proven to be undecidable in Principia Mathematica and related systems, but I shall leave this special case on the side.
(7) See in particular Dummett (1978 : 358, 362, 364), Wright (1987: Introduction, sect. II, especially at pages 13-14, 20, 22-23 and also at pp. 72, 102) and Tennant (1987 : ch. 11). Wright (op. cit. : 22) claims that these capacities can necessarily have no counterpart.
Devitt (Michael), 1984, Realism and Truth, Princeton U. P., Princeton, New Jersey.
Dummett (Michael), 1976, "What Is a Theory of Meaning (II)", Truth and Meaning - Essays in Semantics, G. Evans and J. McDowell, eds., Clarendon Press, Oxford, pp. 67-137.
- 1978, Truth and other Enigmas, Harvard U. P., Cambridge, Mass.
McDowell (John), 1978, "On 'The Reality of the Past'", Action and Interpretation, C. Hookway and P. Pettit, eds., Cambridge U. P., Cambridge, pp. 127-144.
Tennant (Neil), 1987, Anti-Realism and Logic, vol. 1: Truth as Eternal, Clarendon Press, Oxford.
Wright (Crispin), 1987, Realism, Meaning and Truth, Basil Blackwell, Oxford.