JAAKKO HINTIKKA

ARE

THERE

WHY NOT?

NONEXISTENT BUT WHERE

OBJECTS?

ARE THEY?*

Our title question is important, but to my mind it is not nearly as intriguing as two responses it easily prompts. T h e first response is, Why not? T h e second is, But where are they? T h e main overall answer to the question, Are there nonexistent objects? is so obvious that it is much more interesting to ask: Why not? Why have many philosophers been m o v e d to doubt and even to deny that there are nonexistent objects? I believe that by understanding and by removing these counter-reasons we can provide important further arguments for the same affirmative answer that Parsons gives to our theme question. It seems to me as obvious as anything in philosophy that there are unrealized possibilia: that our life is - as I once expressed it intrinsically and inevitably transacted against the backdrop of possibilities (possible states of affairs and possible causes of events), most of which will never be realized. And it seems eminently plausible that there are unrealized and hence nonexistent objects in some of these unrealized possibilities, If I had to present direct arguments for these conclusions, admittedly I would be more than a little puzzled; I don't know what kinds of arguments to present. But then I am not sure whether I could produce more impressive arguments for existent objects. I am reminded here of the question (undoubtedly apocryphal ) I have heard attributed to Meneken: "Is there life before death in the Midwest?" A positive answer to this question may seem in Suitable circumstances as unconvincing as an affirmative answer to its more c o m m o n variant. It is nevertheless highly interesting to ask why philosophers have denied the (to my mind) obviously affirmative answer to our title question. T h e most conspicuous reason is undoubtedly that they have been thinking in syntactical (inference-theoretical) rather than semantical (model-theoretical) terms. Parsons's paper offers examples, both h o m e m a d e and imported, of this phenomenon. Parsons argues that Ryle's own paraphrase test fails in the case of such sentences as "Mr. Pickwick is a fiction." However, the relevance of the criterion itseff is Synthese 60 (1984) 451-458. 0039-7857/84/0603-0451 © 1984 by D. Reidel Publishing Company

$00.80

452

JAAKKO

HINTIKKA

highly dubious. Why should the possibility of paraphrasing sentences about Mr. Pickwick in this or that way prove a point about his ontological status, e.g., that he is or isn't Pickwickian? I don't see any force in the type of philosophical argument that relies on verbal paraphrase for any ontological question. It is only when we begin to ask serious model-theoretical questions that we can hope to find viable reasons for (or against) nonexistent objects. I don't think that anyone who takes these questions seriously can hope to dispense with nonexistent beings. Instead of boring you with model-theoretical technicalities, I will present an analogous but nontechnical case. In the possible world of Verdi's Tosca, the riveting question is whether there really are bullets in the soldiers' muskets in the execution scene. Is the execution a m o c k one or will the hero die in a hail of bullets? But even if there are bullets in the muskets, they are nonexistent ones, for the world of Tosca is not the world of existing objects. H e n c e a question as to whether there are certain specific nonexistent objects can be very burning indeed. T h e same goes for many other questions concerning the being of nonexistent individuals, once the plot or, more generally, the model-theoretic structures involved in our language understanding is taken seriously. T h e main weakness of Parsons's paper is that he is relying far too much on syntactic argumentation, that is, on what we would say, on what so-called "intuitions" we have about different sentences, on what inferences we feel like drawing, and suchlike. T h e r e is no future in such arguments. T h e v e r y concepts they rely on, such as intuitions, acceptability of sentences, acceptability of inferences, etc., all cry out for a real semantical (model-theoretical) foundation. Admittedly, Parsons speaks occasionally of the entities we are committed to. But he quickly moves to questions of what we would say, or can correctly say, e.g., what inference we would draw, and apparently treats the two questions as equivalent. For instance, on p. 7 he writes at a crucial juncture: " I propose that a use of a sentence involving a grammatically correct singular term commits the user to a referent for that term just in case it commits the user to the existential generalization of tile original claim." H e r e a commitment to a referent - presumably, an object - is explicated in terms of a commitment to an inference. Which is the basic kind of commitment here? Admittedly, a number of leading philosophers have rejected modeltheoretic argumentation completely. Frege, early Russell, Wittgenstein, and Quine are cases in point. However, these philosophers have had

ARE

THERE

NONEXISTENT

OBJECTS?

453

what their followers don't have, to wit, a deep theoretical reason for their position. As unlike as the four philosophers just mentioned are, they share what undoubtedly is the most important undiagnosed assumption in contemporary philosophy of language. Or, rather, they exemplify one half of an important undiagnosed dichotomy. I have called this dichotomy a contrast between the view of language as the universal medium and the view of language as calculus. Roughly, on the former view, we cannot as it were get outside our own language and look at it from the outside. We cannot speak in language of the semantical links that tie our language to the world. In brief, semantics is ineffable on the view of language as the universal medium. We cannot think of these semantical relations as being varied, either. Hence systematic model theory is impossible on this view, f o r it relies essentially on experimentation with different representative relations between language and reality. In particular, we cannot change at will the ranges of our quantifiers. Each of them should be thought of as ranging over the domain of the correct logical type. Parsons treats this kind of absolutist view of quantifiers literally as a bad joke. In reality, it is one of the most serious and consequential ideas in the philosophy of language of Frege, early Russell, Wittgenstein, and Quine. (The case of Rudoff Carnap, to whom Parsons refers, remains to be investigated.) Even though I personally reject the view of language as a universal medium, it (and ipso facto its consequences) should be taken much more seriously than Parsons apparently does. The contrary view, that of language as calculus, maintains that we can perform all these neat tricks. It is clear that all serious model theory presupposes the view of language as calculus. But notice that a believer in language as the universal medium is not barred from having all sorts of sharp views on semantical matters. All he (or she) has to do is to admit that these views cannot be expressed in language and hence cannot be built into any real theory. It seems to me that far too many contemporary philosophers have adopted certain consequences of the view of language as the universal medium without realizing where they come from. One of these consequences is the denial of nonexistent individuals. This was certainly one of the reasons why the Wittgenstein of the Tractatus postulated his "objects" which form the substance of every possible state of affairs but which are transcendental in the sense that we cannot say that they exist or don't exist. It seems to me that a systematic discussion and systematic criticism of

454

JAAKKO

HINTIKKA

the view of language as the universal medium is the best strategy for arguing that there are nonexistent objects. A refutation of this view would remove some of the reasons for denying that there are nonexistent objects, and at the same time it would provide, in legitimizing model theory, forceful constructive lines of thought for the purpose of arguing for nonexistent objects. The rejection of language as the universal medium and the consequent model-theoretical treatment also thrust the second Of my teasing questions to the forefront of our attention. If there is not one big predetermined domain of objects for our individual quantifiers to range over, we have to ask, with a new urgency, Where are the nonexistent individuals? It will not do just to argue that natural language quantifiers like "some . . . . every" and "no" occasionally range over nonexistent objects. One has to try to say what the particular domain is that they range over. This is illustrated by the strange fate of my favorite nonexistent individuals, everybody's lover and nobody's beloved. They are both conspicuously possible, even though the first one is perhaps unlikely to exist. They have a much better claim to subsistence than, e.g., Meinong's round squares. (If I change the example slightly, both characters can come even closer to existence, e.g., "the envier of everybody" and "the one envied by no one.") The only remarkable thing about them is that they are incompatible: the former must love the latter, but the latter cannot be loved by the former. And it will not do to deny that "everybody" and "nobody" in the characterizations of these two unforgettable characters range over existing individuals only. It is precisely to allow them to range also over nonexisting individuals that Parsons and his great predecessor Meinong marshalled their arguments. The general point illustrated by this story is that in the presence of relations and relational predicates, objectively understood, we are forced to give a specific, precise answer to my second initial question. If you start with nonexistent objects, you have to partition them into several sets of possible objects, which might as well be called possible worlds. If you ask, Where are the nonexistent objects? the answer is, Each one in its possible world. The only trouble with that notorious thicket, Meinong's jungle, is that it has not been zoned, plotted and divided into manageable lots, better known as possible worlds. Observations of the same sort as mine led Leibniz and before him

ARE

THERE

NONEXISTENT

OBJECTS?

455

Duns Scotus, to divide the ill-structured mass of nonexistent individuals into neatly compartmentalized possible worlds. T o do anything else after the great Gottfried Wilhelm appears to me basically as very poor logical taste. Furthermore, other arguments can be given against pooling all nonexistent objects into one and the same big pool. For instance, I have recently argued in effect that the totality of all nonexistent individuals is an ill-formed totality. It has to be divided into smaller totalities in some way or other in any case. All this highlights further my second question: Where are the nonexistent objects? I am suggesting this answer: E a c h on its own possible world. In other words, each individual is located in some local possible world which it does not make sense to pool together. But doesn't this radically misrepresent the intentions of the defenders of nonexistent objects? Isn't their real point that there really are nonexistent objects in this world of ours, albeit some remote Platonic part of it? Of course they cannot all be there, according to what I have been arguing. But perhaps some of them are located in our world? It may seem that the possible-worlds treatment I favor is committed to a negative answer to this question. Isn't it the point of the whole exercise that nonexistent individuals are interpreted as merely possible individuals? Maybe it is, but the fact remains that when the possible-worlds approach is developed far enough, nonexistent individuals make their appearance once again in this real world of ours. How so? Let's begin at the beginning. I used to say (and to some extent still say) that possible-worlds semantics shows the triviality of questions of existence, including the question whether there are nonexistent objects. Bona fide individuals can exist in one world but fail to exist in another one. What more is there to be said of the matter except that the world in which one of them fails to exist may very well be the actual one? Existence is a much less important issue than the drawing of world lines, i.e., the notional lines which connect the embodiments of the same individual in different worlds and which help to constitute our concept of individual (very same individual). And the problem of world lines is a very real one indeed, for unless we believe in the extreme parochial theory of Leibniz and claim that no two individuals in different worlds are ever identical, it is clear that denizens of different worlds can be identical. H e n c e the question: When are they identical?

456

JAAKKO

HINTIKKA

All this is true, and has some bite. For one thing, it shows that Parsons's use of the viability of existential generalization as a criterion of ontological commitment is misguided. For existential generalization can fail for reasons that have nothing to do with the failure of existence. Let me demonstrate this point by means of an example. From (1)

Queen Victoria knew that Lewis Carroll is Lewis Carroll

it cannot be inferred, even though Lewis Carroll existed and was known by the good Queen to do so, that (2)

(3x) Queen Victoria knew that Lewis Carroll is x;

or that for that matter,

(3)

Someone is such that Queen Victoria knows that Lewis Carroll is him (or her).

Indeed, (2)-(3) obviously say the same as (4)

Queen Victoria knew who Lewis Carroll is,

which obviously is not entailed by (1). I cannot argue fully here for the equivalence of (2)-(3) and (4) e v e n though I believe it is completely unassailable, nor do I have time to qualify it in certain requisite ways. The main point is nevertheless patent. The equivalence of (2)-(3) and (4) is completely independent of the question whether the quantifiers are construed as ranging over existing individuals only or whether they also take nonexisting ones as their values. The reason for this failure of existential generalization is not a failure of uniquenessi in different situations compatible with what the Queen, who had been amused by Alice in Wonderland, actually knew, the name Lewis Carroll applied to different persons. Hence there is no single one who could serve as the value of " x " in (2). Hence existential quantification does not apply to "Lewis Carroll" in (1), a n d y e t it can be taken to commit its utterer to the existence of Lewis Carroll. In other words, Parsons's criterion fails. In spite of the failure of Parsons's test, there is more to be said here.

ARE

THERE

NONEXISTENT

OBJECTS.'?

457

My emphasis on cross-identification (the drawing of world lines) rather than existence leads to interesting conclusions when pushed to the bitter (or sweet) end. I said earlier that each nonexistent individual is located in its own possible world. We have already seen that this is an oversimplification. If I am not Leibniz, I have to consider my individuals as denizens of more than one possible world. This is what the Lewis Carroll examples were predicated on. (The reason why individuals can do this is that in different worlds one and the same individual can have different relations and hence be compatible with a different selection of other individuals.) Now how do we do this? H o w do we draw the imaginary world lines that connect the embodiments of one and the same individual in different worlds? All I have time to say here is: this is a highly nontrivial and complex task, so complex indeed that we may fail in it in many cases. It just is the case, I have argued at length elsewhere, that the principles that govern it fail in some cases. Sometimes we just cannot tell what the counterpart (to use David Lewis's term) to a given individual in one world would be in another. This natural observation enforces a surprising distinction. We have to distinguish from each other two kinds of failure of a world line. Case (1): Sometimes our criteria of cross-identification work and tell what a given individual, say i, would be like a given world w, while the application of those criteria to w leads to the conclusion that i does not exist in w. Case (2): However, world lines may fail in a more radical way. T h e y may fail to tell us what i would be like in w to a degree that would not even enable us to decide in principle whether or not i exists in w. T h e former is a failure of existence; the second may be called failure of weU-definedness. Now well-defined but nonexistent objects are in some natural sense in the world in question. T h e y are the best rational reconstructions of nonexistent individuals which don't merely exist in some other possible world but which enjoy some reasonable status in our real one, viz., when w happens to be the actual world. Thus, possible-worlds semantics vindicates in the end merely possible individuals even as members (albeit by courtesy) of the actual world. An attempt to dispense with them is bound to misrepresent thoroughly the true semantical and logical situation. And hence I have to conclude that, even though I reject most of Parsons's argumentation, I find myself agreeing with most of his conclusions.

458

JAAKKO

HINT1KKA

NOTE * This paper was originally written as a response to Terence Parsons's paper, 'Are There Nonexistent Objects?' which was presented at the American Philosophical Association, Pacific Division Meetings, in March 1982. Parsons's paper appeared in American Philosophical Quarterly 19 (1982), 365-371. Department of Philosophy The Florida State University Tallahassee, Florida 32306 U.S.A.