Transcript
Grazer Philosophische Studien 82 (2011), 329–351.
FREGE, PÜNJER, AND KANT ON EXISTENCE Tobias ROSEFELDT Humboldt University of Berlin Summary The paper tries to shed new exegetical light on Frege’s “Dialogue with Pünjer on Existence” by showing that Pünjer’s position in the dialogue is strongly inspired by Kantian claims about existence. It is argued that Pünjer’s wavering between a broadly Meinongian and a broadly Fregean view on existence can be explained by the fact that there are Kantian remarks which seem to speak in favour of each of these views. A suggestion is then made how Kant’s claims can be interpreted in such a way that the tension which they seem to entail disappears.
At some time in the early eighties of the 19th century, Gottlob Frege and the Protestant theologian Bernhard Pünjer, both professors of the University of Jena at this time, met for a discussion about the nature of existence and the logical form of existential statements. Their dialogue is documented in an incomplete transcript as well as in a subsequent summary by Frege in which he tries to elucidate the structure of the development of their discussion and explains his own position in more detail.1 While both documents make it very clear for which account of existence Frege argued in the conversation—namely that of treating the concept of existence as a second level concept—, they leave the reader somehow puzzled as to what exactly Pünjer’s position was. In the dialogue, Pünjer seems to have made several claims which were then shown by Frege to be inconsistent with each other. He also seems to have refused several suggestions made by Frege how to arrive at a coherent position. All in all, Pünjer’s remarks give the impression of originating from rather substantial philosophical confusion. A charitable reader of Pünjer’s remarks will therefore raise two questions: First, is there a coherent account of existence which covers the motivation by which Pünjer apparently has been driven? Second, why is 1. Cf. Frege (1990); I will use the translation in Frege (1979) and make slight modifications if I find them appropriate. Page numbers refer to the English edition.
Pünjer so reluctant to accept one of Frege’s suggested amendments which would make his account coherent? In this paper, I want to argue that both of these questions can be answered if one realises that, in the dialogue, Pünjer tries to defend a basically Kantian account of existence. I will show how Pünjer’s remarks originate from certain things Kant writes about existence and existential statements. The main aim of my paper will be to show that Pünjer’s hesitation to accept Frege’s amendments can be explained by a certain exegetical puzzle which arises from Kant’s claims about existence: Kant’s writings contain passages which seem to speak in favour of a broadly Fregean, and hence anti-Meinongian, account of existence as well as passages which seem to support a broadly Meinongian, and hence anti-Fregean, account of existence. I will argue that this puzzle explains Pünjer’s confusion and I will also make a suggestion how it should be solved. The methodological lesson to be learned from Frege’s dialogue with Pünjer is that it is always very instructive for a Kantian to be confronted with the conceptual rigour and insistence on clarity of a Fregean. This gives me a very personal reason to contribute a paper on this dialogue to the present volume. When I first read Wolfgang Künne’s Abstrakte Gegenstände, it was shortly after I had finished my dissertation on Kant’s theory of self-consciousness. The book made a deep impression on me and left me with two wishes. The first was to do systematic ontology myself. The second was to write about Kant in a way that does not fall short of the standards of conceptual clarity that someone like Wolfgang Künne would find acceptable. Although I have doubts that the second wish will ever become true, I am grateful to Wolfgang for keeping it alive in me. 1. Pünjer’s predicament On the basis of Frege’s summary of his conversation with Pünjer, it is reasonable to assume that their dispute took its starting point from questions about the logical form of existential statements. Pünjer seems to have held the view that existence can be expressed in general statements such as ‘There are tables’ as well as in singular statements such as ‘This table exists’. More precisely, this view could be characterised (and is so characterised by Frege) in the following way (cf. Frege 1979, 61):
330
(P1) The expression ‘there are’, in sentences of the form ‘There are Fs’, and the expression ‘exists’, in sentences of the form ‘a exists’, express essentially the same content. (P1) implies that existence is something which is predicable of individual objects, or—to put a non-Fregean claim in Fregean terms—that the concept of existence is a first-level concept and existence a property of objects. A second assumption that Pünjer seems to have held concerns the analysis of the general existence statements mentioned in (P1). It can be viewed as an explanation of how what is said in (P1) can be the case. For, it gives an analysis of statements of the form ‘There are Fs’ that implies that these statements somehow implicitly contain the first-order existence predicate. The assumption reads as follows (ibid. 61): (P2) Sentences of the form ‘There are Fs’ mean the same as the respective sentences of the form ‘Some existing things are F ’. The upshot of this claim is that it is not the particular quantifier ‘some things’ as such that carries the existential commitment implied by a use of the expression ‘there are’ but rather the particular quantifier restricted by the existence predicate, i.e. the expression ‘some existing things’.2 A third claim that is distinctive for Pünjer’s view of existence has to do with an attempt to specify further what we mean when we say that a certain thing exists or fails to exist. According to Pünjer, the exclusive function that existence statements have for us consists in making it possible for us to deny of certain things that they are objects of mere hallucination and claim that the representations we have of them count as experience (53, §§ 3, 12; 54, §18). Pünjer furthermore believes that experience can be distinguished from hallucination on the grounds that it originates from an ‘affection of the subject’ (54, § 20). The idea behind this claim is probably that representations which constitute hallucination are somehow fully created by the subject herself whereas representations that constitute experience originate in a source outside the subject through some causal impact. Driven by these considerations, Pünjer makes the following claim (53, § 12): 2. In the dialogue, Pünjer insists that the inference from ‘Sachse is a man’ to ‘There are men’ is not logically valid but needs the additional premise ‘Sachse exists’ (cf. 60).
331
(P3) For us, to say of an object that it exists means to say of it that it is an object of experience, i.e. something the representation of which originates from an affection of us. Hence, for us, ‘There are Fs’ means the same as ‘Some objects of experience are F’, and ‘a exists’ means the same as ‘a is an object of experience’. For everyone familiar with Frege’s view of existence it should be clear that he would accept none of Pünjer’s three claims. From a Fregean perspective, (P1) and (P2) suffer from the defect of treating the concept of existence as a first-level concept, i.e. a concept under which objects can fall. (P3) moreover is prone to the confusion of matters of epistemology with matters of thought and meaning, a confusion Frege is always very eager to dissolve.3 Frege’s strategy for refuting Pünjer’s view of existence is to show that there is no plausible reading of (P1)–(P3) on which these three claims can consistently all be held together. His main focus lies on a discussion of (P2). According to Frege, it would only be correct to assume that ‘There are tables’ means the same as ‘Some existing things are tables’ if the predicate ‘exist’ expressed something pleonastic and trivial (‘etwas Selbstverständliches’) and had no real content, i.e. would be true of everything, such as the predicate ‘is self-identical’ (cf. 62 and 59, § 88). The reason for this is simple: Assume that the existence-predicate is not true of everything. Because of the equivalence of ‘Not all Fs are G’ and ‘There are Fs that are not G’, that would mean that there are existing as well as non-existing things. However, if this were the case, then to say that there are tables would not be equivalent to saying that some of the existing things are tables, for only the falsity of the first not that of the second statement would imply that there are no tables at all. The falsity of ‘Some of the existing things are tables’ would leave open the possibility that there are non-existing things which are tables. Hence, (P2) is true only if ‘exists’ expresses a property that applies to everything and if it is trivial to say of something that it exists. Let us call this consequence of Pünjer’s view ‘the triviality implication’. The triviality implication has several consequences. One consequence Frege draws is that even if (P2) is true the real content of a claim such as ‘Some of the existing things are tables’ or ‘Some tables exist’ cannot lie in 3. Since (P3) is rather implausible when formulated as a claim about sameness of meaning, a more charitable way of interpreting what Pünjer says might be to put the connection asserted by (P3) in terms of sameness of conditions of justified assertion. However, the possible criticism which could thus be avoided shall not concern us here.
332
the predicate ‘exist’. Instead, it must lie in the ‘form of the particular judgment’. The reason is that when we claim that there are tables, what we say is not trivial. A further consequence of this is that (P2) cannot serve as an explanation of how (P1) can be true. The reason is that even if ‘there are Fs’ is equivalent to ‘some existing things are F’, what carries the existential commitment of ‘there are’ is the form of the particular judgement, rather than the existence predicate; and it is still hard to understand how something that is expressed by the form of the particular judgement can also be expressed by a first-order predicate. The most important consequence for Frege’s criticism is that the triviality implication shows that (P1) and (P2) are incompatible with Pünjer’s thesis (P3). If the existence predicate does not have any substantial content and is true of everything its function for us cannot consist in distinguishing between objects of experience and objects of mere hallucination, or between representations that originate from an affection of the subject and those which do not. For it is certainly not trivial to say of an object that it is not only an object of hallucination, or of a representation that it originates from an affection of the subject by some outer object. From a systematic point of view, there seem to be two possible strategies to answer Frege’s criticism. I will call them the ‘Meinongian’ and the ‘Fregean’ strategy. The first strategy owes its name to the fact that it consists in accepting a consequence of (P3) which is distinctive for the Meinongian conception of existence. The consequence is that, according to (P3), existence is a discriminating property of objects, a property that, as Meinongians assume, some objects have (e.g. objects of experience) and others lack (e.g. objects of mere imagination). Since this implies that there are existing as well as non-existing things, Pünjer would have to give up or at least modify claims (P1) and (P2). He would have to admit that there is at least one meaning of the expression ‘there are’ in which sentences of the form ‘There are Fs’ are not equivalent to the requisite sentences of the form ‘Some existing things are F ’ and do not carry existential commitment. He might furthermore explain the prima facie plausibility of (P1) and (P2) by the fact that in most philosophical as well as ordinary contexts we are prone to what Meinongians call the ‘prejudice towards the actual’ and implicitly restrict our quantifiers to existing objects. At one point of the conversation, Frege explicitly confronts Pünjer with the Meinongian consequence of his claim (P3):
333
Then it follows that there are objects of representations […] which do not exist. Now if you are using the word ‘exists’ in the same sense as the expression ‘there is’, then you have at the same time both asserted and denied the same predicate of the same subject. […] Do you grant this? Pünjer: Yes. But the word ‘there is’ is wrongly used in this context. Frege: Then put in its place another expression that will express the matter better. (59, §§ 79ff.)4 Frege:
The Meinongian response to this demand would be to distinguish between two senses of the expression ‘there is’—one in which it is equivalent with ‘there exists’ and another in which it is existentially neutral. This would make it possible to claim—to use Meinong’s own words—that ‘there are things of which it is true to say (in some other sense) that there are no such things’ (cf. Meinong 1904, 9). However, Pünjer is not willing to bite this Meinongian bullet. He answers to Frege’s challenge: Pünjer: We can’t: any other expression would again fail to say what is meant to be expressed. […] Before we deny the existence of anything whatever, we have to represent it as existing in order to go on to deny existence of it. But I don’t think that we shall get any further along these lines. (59, §§ 82, 84) Here, Pünjer seems to follow many contemporary Anti-Meinongians in finding it impossible to understand the expression ‘there is’ in a nonexistence entailing way. He seems to interpret the claim that there are things that do not exist as an unavoidable paradox of our talk about nonexistence the inconsistency of which cannot be avoided by use of some weaker reading of the expression ‘there is’ but rather only by reference to the temporal dimension of our utterances: In saying that there are things that do not exist, we first represent something as existing and later we deny the existence of what was presented as existing. However, the last remark of the quote makes clear that he did not himself believe that reference to the non-simultaneity of the utterance of the sentence-parts ‘there are things’ and ‘that do not exist’ really solves the problem Frege’s criticism had made apparent. 4. I departed from the quoted edition by translating the German word ‘Vorstellung’ by ‘representation’ rather than by ‘idea’ here. The reason is that, as I will argue below, Pünjer holds a Kantian position and for Kant, ideas are just one very special kind of representations.
334
The Fregean strategy, on the other hand, is built on the idea that we can take seriously the motivation behind Pünjer’s claim (P3) without having to assume that existence is a discriminating property of objects. This assumption can be traced back to the indeed plausible claim that not all our representations represent objects in the world. In Pünjer’s own words: Pünjer: Yes; there are two kinds of representations: those that originate from the ego alone, and those that are formed through something affecting the ego. In order to distinguish these I say: the objects of representations of the latter kind can be experienced; to representations of the former kind there do not correspond any objects that can be experienced. (54, §335, 20) Interestingly, Pünjer does not say here that representations originating from the ego refer to objects that cannot be experienced but confines himself to denying that they refer to objects that can be experienced. This would offer him the possibility to give up the distinction between existing and non-existing objects and replace (P3) by a claim that assigns to existence claims the purpose of classifying representations as to whether they represent something or fail to do so. Frege confronts Pünjer with this modification of his original position: Frege: It seems to me then as if the real subject on your way of thinking is the representation. […] When you say ‘There are men’ and ‘There are no centaurs’ you are also making a classification. But you are not classifying things […], but you are classifying the concepts ‘man’ and ‘centaur’ by assigning one to the class of concepts under which something falls, and excluding the other from this class. This is why I hold that in these sentences the concepts are the real subjects. If you say ‘This can be experienced’, where this has the sense ‘This representation of mine is not something originating from myself alone’, then you are classifying the representation. […] This is why I maintain that it is here the representation that is the real subject. Another way of putting the same thing is to say: The representation has the property that something corresponds to it. (54, §§ 21, 23) What Frege suggests to Pünjer here is a view on existence which obviously has some similarity to his own. General existence statements such as ‘There
335
are human beings’ or ‘There are no centaurs’ are analyzed as statements about the concepts of a human being and that of a centaur, saying of the first, and denying of the second, that at least one object falls under them. A singular existential statement such as ‘This exists’ or—in Pünjer’s reformulation—‘This can be experienced’ is interpreted as a statement about a particular representation, saying about that representation that at least one object corresponds to it. If Pünjer wanted to adhere to (P1) and assume that the expression ‘there are’ in general existence statements and the expression ‘exists’ in singular existence statements express the same content, he could assume that in both kinds of statements we ascribe to the property of being represented by such-and-such representations the second-order property of being instantiated by at least one object. (In the case of a general existence statement, the representation at issue would be a concept, in the case of a singular existence statement it would probably be some perceptual representation.) Although Frege certainly would not approve of the meta-representational touch of it, the resulting analysis would deserve to be called Fregean in that it sees a strong connection between existence and the second-order property of being instantiated by at least one object. (This connection would in turn leave no room for Pünjer’s claim (P2) any more.) As in the case of the Meinongian strategy, Pünjer explicitly refuses to accept the Fregean strategy as a way to deal with Frege’s objection and is not willing to interpret existential statements as implicitly meta-representational. He answers to Frege’s suggestion: Pünjer: […] Negation is possible only when something has already been posited. Hence if we say ‘Centaurs do not exist’, this is possible only because we first think of them as being outside ourselves. We have a twofold ground for denying existence: 1. a logical contradiction, 2. outside the concept or representation in experience. So properly speaking the real subject is neither the concept nor the representation. (55, § 24) Pünjer notices that, in the case of true negative a posteriori existential statements, it is not the concepts alone which make it the case that nothing falls under them but also the state of the world, of which we therefore need experience in order to find out whether the statement is true. From this he seems to conclude that such existential statements cannot be about our concepts. Now, a proponent of the Fregean strategy could agree that such
336
statements are not made true by the concepts alone but also by the world as a whole which is responsible for the fact that none of its parts falls under the respective concept, and hence that the statements are not solely about concepts but also about the world as a whole. However, Pünjer seems to think that his argument moreover shows that a true negative a posteriori existential statement such as ‘Centaurs do not exist’ is somehow about centaurs, objects which we ‘posit’ by using the concept of a centaur and of which we then find out that they do not exist. Hence he is explicitly not willing to abstain from the assumption that there are non-existing objects. Pünjer’s refusal of the Meinongian as well as of the Fregean strategy leaves him in the uncomfortable situation of not being able to present any convincing answer to Frege’s challenge whatsoever. He seems to have been forced into the predicament of having to accept a Meinongian or a Fregean view about existence and at the same time not being willing to accept either of these views. This is the situation in which we find him until the end of the conversation, or at least until the end of that part of the conversation which is reported by the transcript. The pressing question at this point is why Pünjer is so reluctant to accept one of the two proposals Frege makes to him. In the next section, I try to answer this question by showing that Pünjer’s views are heavily inspired by Kantian claims about existence and that his unwillingness to accept either the Meinongian or the Fregean strategy is due to the fact that the Kantian claims themselves seem to waver between these two approaches. 2. Pünjer’s Kantian roots According to the editors, Frege’s conversation with Pünjer took place at some time between 1880 and 1884 (cf. Frege 1990, 172). Some years before, namely in 1874, Pünjer had published a small volume called Die Religionslehre Kants (Kant’s doctrine of religion, cf. Pünjer 1874). Although this book does not contain any detailed discussion of Kant’s claims about existence, we can take it for granted that its author was well familiar with these claims because most of them appear in the context of Kant’s refutation of the ontological proof in his early writing The only possible argument in support of a demonstration of the existence of God (1763) and in the Critique of Pure Reason (1781/21787). Moreover, it seems likely that a theologian whose primary contact with philosophy was an investigation
337
of Kantian writings would try to bring in the Kantian position into a discussion like the one with Frege about existence. This suspicion is confirmed by the fact that much of what Pünjer says in the conversation can be traced back to Kantian remarks in either The only possible argument or the Critique of Pure Reason. The idea behind (P2)—namely that claims of the form ‘Fs exist’ or ‘There are Fs’ are not equivalent to the respective claim of the form ‘Something is an F’ but rather only with ‘Some existing thing is an F’—is probably inspired by passages from The only possible argument in which Kant speaks about the logical form of existential statements, such as the following: The expression “A sea-unicorn is an existing animal” is not […] entirely correct. The expression ought to be formulated the other way round to read “Some existing sea-animal has the predicates which I collectively ascribe to a unicorn”. (The only possible argument, AA II 72f., cf. also 745)6
(P3), which asserts a strong connection between existence—or at least existence we can make sense of—and the property of being an object of experience (and hence of representations which originate from an affection of the subject), obviously mirrors Kant’s remarks in the chapter on the “Postulates of Empirical Thinking” in the Critique of Pure Reason about empirical useful application conditions of the categories of modality, among them the category of existence. Kant writes there: In the mere concept of a thing no characteristic of its existence can be encountered at all. […] that the concept precedes the perception signifies its mere possibility; but perception, which yields the material for the concept, is the sole characteristic of actuality. (Critique of Pure Reason A 225 f./B 272 f.)
For Kant, perception entails an affection of the subjects by some outer object. Moreover, already in The only possible argument, Kant had claimed 5. In this paper, I follow, by and large, the translations in Kant (1998) and (2002). The page numbers refer to the Academy Edition of Kant’s writings (Gesammelte Schriften, hg. von der Preußischen [später: Deutschen] Akademie der Wissenschaften, Berlin 1900 ff.) and, in the case of the Critique of Pure Reason, to the pagination of the first (A) and second (B) edition. 6. Kant would also agree with Pünjer’s assumption that a singular statement such as ‘Sachse is a man’ does not have existential import and hence its truth does not imply the existence of men (cf. fn. 4). This is obvious from two passages. Firstly, Kant claims that judgements about impossible objects, such as ‘The God of Spinoza is subject to continuous change’, can be true, but he would certainly deny that there exists some Spinozean God (cf. AA II 74). Secondly, he writes: ‘The proposition “God is omnipotent” must remain true even for someone who does not acknowledge the existence of God […]’ (ibid.).
338
that to say that sea-unicorns exist ‘simply means: the representation of a sea-unicorn is an empirical concept’ (AA II 74), i.e. a concept that is derived from sensible experience. Now, if Kant shares the core of Pünjer’s views on existence (or rather the other way round), he would be confronted with the same problems that Frege’s objections had made apparent. Hence, one might wonder whether there are any hints in the Kantian text that he has chosen either the Meinongian or the Fregean strategy to make his views coherent. As I will show next, the answer is that for both of these strategies we can find passages in Kant which seem to imply that he would have disapproved of them. This is the reason, I think, why Pünjer, trying to be a good Kantian, is so reluctant to accept any of the amendments of his conception which Frege offered to him. Let us start with the Meinongian strategy according to which we should accept that there are existing as well as non-existing objects. It is noteworthy that, when Pünjer and Frege speak critically about the view that there are non-existing things of which we can find out that they are not objects of experience, they use an expression which is characteristic for Kant’s talk about existence, namely the term ‘determine’ (‘bestimmen’): Pünjer: The statement ‘A cannot be experienced’ is not possible. […] Neither does it make sense to deny that a thing can be experienced. […] Frege: To say of a thing that it can be experienced is not to determine it in any way. (Durch die Aussage der Erfahrbarkeit wird dasjenige, von dem sie ausgesagt wird, nicht irgendwie bestimmt.) Pünjer: No. That is the difference between this statement and the others. (53, § 10; 54, §§ 15, 16) The claim that by saying that a certain thing exists or—which according to (P3) amounts to the same—that it can be experienced, we do not determine it in any way, is reminiscent of Kant’s most famous dictum about existence: […] the illusion [on which the ontological proof rests] consists in the confusion of a logical predicate with a real one (i.e. the determination of a thing [der Bestimmung eines Dinges]) […]. Being is obviously not a real predicate, i.e. a concept of something that could add up to the concept of a thing. (Critique of Pure Reason, A 598/B 626)
339
Kant claims here that the concept of existence is only a logical predicate, i.e. a concept that can be used at the predicate position of a categorical judgement such as ‘God exists’, but not a real predicate, i.e. a concept by which objects are ‘determined’. This claim could be understood as a disclaimer of the Meinongian strategy if it was interpreted in the following way: Real predicates are predicates that express discriminating properties of objects, i.e. properties that divide the class of all objects in those which have it and those which fail to have it. To deny that the concept of existence is a real predicate would then mean that existence is not a discriminating property of objects and that it is not the case that, as the Meinongian strategy assumes, there are existing as well as non-existing objects. I think it is very likely that Pünjer interpreted Kant’s dictum in this way and for that reason was not willing to accept the Meinongian answer to Frege’s challenge. Would Pünjer have found support in Kant for subscribing to the Fregean strategy? At first sight it looks as if there are passages in which Kant sympathises with a Fregean account of existence.7 As I have characterised the Fregean strategy, its distinctive features are (i) that it interprets existence not as a first-order property of objects but rather as a second-order property of properties, namely the property of being instantiated by at least one object, and (ii) that it reconstructs existence-claims as implicit meta-representational in the sense that, in these claims, the property of being instantiated by at least one object is ascribed to the property of corresponding to such-and-such representations. Of the following three quotations from Kant, the first two seem to support the first, and the third the second of these two aspects of the Fregean strategy: [When I say] “God is”, or “There is a God”, then I add no new predicate to the concept of God, but only […] posit the object in relation to my concept. (Critique of Pure Reason, A 599/B 627) Existence cannot, therefore, itself be a predicate. If I say: “God is an existent thing” it looks as if I am expressing the relation of a predicate to a subject. But there is an impropriety in this expression. Strictly speaking, the matter ought to be formulated like this: “Something existent is God”. In other words, there belongs to an existent thing those predicates which, taken together, we designate by means of the expression “God”. (The only possible argument, AA II 74) 7. Many interpreters read Kant as a proto-Fregean with regard to existence. (E.g. Bennett 1974, 228 ff.; Forgie 2000; Reed 2007, 169f.)
340
When existence occurs as a predicate in common speech, it is a predicate not so much of the thing itself as of the thought which one has of the thing. For example: existence belongs to the sea-unicorn but not to the land-unicorn. This simply means: the representation of a sea-unicorn is an empirical concept; in other words it is the representation of an existent thing. (The only possible argument, AA II 72)
A closer look at these passages, however, shows that they are far less Fregean as they may appear at first sight. According to the Fregean strategy described above, the statement ‘God exists’ should be analyzed as equivalent to the statement ‘There is at least one object that corresponds to the concept of God’. Hence, existence would be expressed by the quantifier ‘there is at least one object’, which Frege would interpret as a second-order predicate expressing a property of properties. What Kant suggests as an analysis of ‘God exists’ in the second quotation, however, is different from this approach. His analysis of ‘God exists’ is ‘There is at least one existing object that corresponds to the concept of God’, and it seems as if the existential import of the analysans is not carried by the quantifier alone. For God to exist it is not sufficient that there is some object corresponding to the concept of God, but that object has to be an existing object. In this analysis, however, the verb ‘exist’ is used as a first-order predicate expressing a property of objects. This difference between the Kantian and the Fregean account of existence is also apparent in the fact that for Frege existence is a quantity and existential statements are concerned with the number of objects falling under a certain concept. (To say that Fs exist is to deny that the number of Fs is zero. Cf. Frege 1988, § 53, 64; 2002, 18 f.) For Kant, however, the concept of existence does not appear among the categories of quantity but is a category of modality. For him, to exist seems to amount to being an actual not just a possible object (cf. Critique of Pure Reason, A 80/B 105). The most striking textual evidence for the assumption that Kant would not approve of the Fregean strategy are passages in which he explicitly says that there are existing as well as non-existing objects. Here is one from the The only possible argument: Take any subject you please, for example, Julius Caesar. Draw up a list of all the predicates which may be thought to belong to him […]. You will quickly see that he can either exist with all these determinations, or not exist at all. The being who gave existence to the world and to our hero within that world could know every single one of these predicates without exception and yet
341
still be able to regard him as a merely possible thing which, in the absence of that Being’s decision to create him, would not exist. Who can deny that millions of things which do not actually exist are merely possible […]. Or who can deny that in the representation which the Supreme Being has of them there is not a single determination missing, although existence is not among them, for the Supreme Being cognises them only as possible things. (AA II 72, my emphasis)
I will return to the question of what the point about Julius Caesar and God’s complete representation of him is later. At the moment we simply have to acknowledge the fact that Kant not only seems to find it consistent to speak about non-existing merely possible objects—a way of talking that a Fregean would have to dismiss as being ill-formed—, but also claims that there are millions of such objects. If Pünjer was influenced by these claims, which I think he was, he must have thought that centaurs, for example, are such non-existing possible objects. And hence he must have found it plausible that, when we say that centaurs do not exist, we speak about these non-existing objects—objects of which we think as ‘being outside of ourselves’ (55, § 24), and not only about our concept of them. Now, you might think that it is one thing to assume that Pünjer was inspired by Kant’s critical philosophy and another thing to hold that he was convinced by every thing Kant claimed in one of his early pre-critical writings. It is important to note, however, that the distinction between existing and non-existing objects has not disappeared from his critical philosophy at all. Beside Kant’s remarks about the ‘1000 possible thalers’ which do not improve my financial situation in the context of Kant’s refutation of the ontological proof in the Critique of Pure Reason (cf. A 599/B 627) there is a highly instructive passage at the very end of the “Transcendental Analytic” in which Kant’s commitment to non-existing objects becomes apparent. Kant claims there that it is important to note that the real objects—i.e. those objects he was concerned with throughout the Critique so far—can be distinguished from objects of another kind, so that the class of ‘objects in general’ must be divided into the class of ‘objects that are something’ and the class of ‘objects that are nothing’. He then distinguishes different kinds of objects that are nothing, among them so called ‘objects of mere thought’ (‘Gedankendinge’ or ‘entia rationis’). An ‘object of mere thought’ is an ‘object of a concept to which no intuition corresponds’ (A 290/B 347). Objects of mere thought are distinguished from so-called ‘Undingen’ (impossible objects, another kind of ‘objects 342
that are nothing’) in that they are objects of non-contradictory concepts. And they are distinguished from real existing objects in that there is no intuition of them. According to this view, the concept of a centaur would represent an object of mere thought. Hence it would not be empty at all, but just fail to represent an existing object. One might not like what Kant is writing in the passages about nonexisting objects but it is undeniably something Kant himself seems to have found unproblematic. And this is the reason, I think, why someone like Pünjer, who tries to be a good Kantian, gets into a predicament if he is confronted with the alternative Frege offers to him: He cannot accept the Meinongian strategy because it seems to conflict with Kant’s dictum that existence is not a determination; and he cannot accept the Fregean strategy because it seems to run counter to Kant’s view that all representations, even those not constituting experience, represent objects, albeit some of them only represent non-existing objects. In the next section I will briefly sketch which of the two alternatives Pünjer should have chosen if his intention had been to adhere, as closely as possible, to the original Kantian view about existence. That is, I will try to solve the exegetical riddle that Kant’s remarks about existence and non-existing objects pose to his readers. 3. Kant’s view of existence 8 Although many people read Kant’s remarks about existence as protoFregean, I think there is no way of interpreting away the fact that he speaks of existence as a discriminating property of objects, a property that some objects have (e.g. objects of sensible experience) and others, such as merely possible objects, lack. It may be that such non-existing objects are objects in some other, weaker sense than the ones Kant usually talks about when he considers the question of how our representations refer to objects. But it is undeniable that he also makes use of a concept of an ‘object in general’ (‘Gegenstand überhaupt’) which allows him to distinguish between existing and non-existing objects. This makes his view of existence Meinongian avant la lettre.9 8. I presented an early version of the following on the 10th International Kant-Congress in Sao Paolo in 2005 (cf. Rosefeldt 2008). 9. This is far less surprising than it may first seem if we take into account the fact that many philosophers of the pre-Kantian philosophical tradition were Meinongians in this sense. Descartes—just to mention one famous example—found it completely unproblematic to speak
343
This diagnosis leaves us with two pending exegetical questions: (i) If, for Kant, existence is a discriminating property of objects, what is meant by his claim that the existence-predicate is not a ‘real predicate’ and existence not a ‘determination of a thing’? (ii) If, for Kant, existence is a discriminating property of objects, what is meant by those of his remarks that seem to speak in favour of a meta-representational analysis of existential statements like the one in which he writes that existence ‘is not a predicate of the thing itself but rather of the thought one has of this thing’ (cf. AA II 72)? In the remainder of this paper, I will try to present the sketch of an answer to both of these questions. Let me quote again the passage from The only possible argument in which Kant speaks about non-existing possible objects: Take any subject you please, for example, Julius Caesar. Draw up a list of all the predicates which may be thought to belong to him […]. You will quickly see that he can either exist with all these determinations, or not exist at all. The being who gave existence to the world and to our hero within that world could know every single one of these predicates without exception and yet still be able to regard him as a merely possible thing which, in the absence of that Being’s decision to create him, would not exist. Who can deny that millions of things which do not actually exist are merely possible […]. Or who can deny that in the representation which the Supreme Being has of them there is not a single determination missing, although existence is not among them, for the Supreme Being cognises them only as possible things. (AA II 72)
Kant speaks here about the scenario of God’s creation of the world, and his description of this scenario is obviously Leibnizian in spirit. The general idea is that God chooses the actual world among different possible worlds and the actual individuals among different possible individuals. God can think of possible individuals—of those which he then decides to make actual and of those which he leaves uncreated—by means of Leibnizian complete individual concepts. Once we have accepted that those complete individual concepts represent objects, albeit maybe merely possible ones, it seems plausible to accept the following principle which I will call the ‘comprehension principle for possible objects’:
of objects that only have objective being but lack formal existence, a view that also seems to have been motivated by the fact that ideas which do not represent objects in the world can nevertheless be ideas of something and hence represent some intentional objects that do not really exist.
344
(CP) For every complete individual concept, there is a possible object falling under it.10 Now, in the quoted passage, Kant apparently wants to stress that a concept can be a complete individual concept although it neither contains the concept of existence nor that of non-existence: God’s representation of a possible object misses ‘not a single determination’, although this representation neither represents the thing as existing nor as non-existing. The question is, then, what the point of this claim is and why Kant thinks that nobody would deny it. The answer, it seems to me, lies in a problem that, rather obviously, arises from (CP) together with the assumption that concepts have to contain either the concept of existence or its negation in order to be complete. Let us suppose that someone proposed the following ‘simple’ definition of complete individual concepts: (SD) A concept C is a complete individual concept iff, for every predicate P, C has either P or its contradictory counterpart P as a part (and does not contain predicates which are incompatible with each other).11 Now, let C be a concept which is almost complete and just indeterminate with respect to the predicate of existence. Let C + be the concept which we get if we add to C the predicate of existence, and C – the concept which we get if we add to C the predicate of non-existence. According to (CP), there would then be an object falling under C + (an existing C), and an object falling under C – (a non-existing C). This would have two unacceptable consequences: Firstly, God would have no choice as to whether to bring a C into existence or not, for solely on the ground that there is a concept of an existing C would there be an existing C. Secondly, and much more disastrously, there would be an existing and a non-existing variant of any 10. Regarding the fact that, in his critical philosophy, Kant distinguishes between real possibility and merely logical possibility, we have to read ‘possible object’ in the sense of ‘logically possible object’ here because, for Kant, real possibility is never established by the noncontradictoriness of a concept. 11. Kant does not use the term ‘predicate’ to refer to linguistic expressions. He wavers between speaking of predicates as of concepts (e.g. when he says that predicates can be parts of other concepts) and as of properties (e.g. when he says that predicates can be ‘determinations’ of things). I will use the term ‘predicate’ to refer to (non-individual, i.e. general) concepts and call what is expressed by them ‘property’ or ‘determination’.
345
object one could think of. Using a Quinean slogan, one could say that (SD) together with (CP) would imply that everything exists, meaning, in a completely un-Quinean spirit, that for any kind of object there really exists an object of that kind. This is obviously absurd. It is the insight in this absurdity, I think, that motivated Kant’s claim that the predicate of existence is a merely logical, not a real predicate and does not express what he calls a determination of a thing. Any theory of possible objects which allows that there are contingently existing possible objects and entails a comprehension principle such as (CP) has to restrict the definition of complete individual concepts to a certain class of predicates of which existence is excluded. Kant calls those predicates ‘real predicates’ and claims that only real predicates are—or rather express— determinations of objects. This claim is not identical to the claim that a merely logical predicate such as the existence-predicate expresses no property of objects altogether, for we have seen that Kant does conceive of existence as a property of objects. Determinations rather form a subset of the set of properties. They are the properties which determine what an object is independently of whether it is an existing object or not. In more traditional terms, we could say that determinations make up the ‘realitas’ of a thing, and this, I conjecture, is the reason why Kant calls predicates that express those determinations ‘real predicates’. Building on the distinction between real predicates and merely logical predicates, we can formulate the following refined definition of complete individual concepts: (RD) A concept C is a complete individual concept iff for every real predicate P, C has either P or its contradictory counterpart P as a part and C does not have any non-real predicate as a part (and also does not contain predicates which are incompatible with each other). On the assumption that the concept of existence is not a real predicate we can avoid the absurd consequences that the simple definition of complete individual concepts together with the comprehension principle for possible objects have created. It is interesting to note that Kant’s distinction between real and non-real predicates and between determinations and properties that are not determinations is very similar to a distinction by which Meinong and contemporary Meinongians try to avoid certain unacceptable consequences of their
346
theory of objects (for the following cf. Meinong 1904; Parsons 1980). As is well known, Meinongians do not only accept that for every consistent set of properties which form the content of a complete individual concept there is an object that has all these properties, but moreover want to assume that for every set of properties whatsoever—no matter whether it is complete or not, consistent or not—there is an object having exactly the properties in the set. That has the consequence that besides existing objects there are not only possible objects but also such things as incomplete objects and impossible objects. One might try to spell out this assumption in form of the following ‘naive’ comprehension principle for Meinongian objects: (NCP) For every set S of properties, there is an object that has exactly those properties which are elements of S. However, (NCP) leads to a problem very similar to the one which was caused by the conjunction of (CP) and (SD) above. If we allowed the property of existence to enter into the sets of properties (NCP) speaks about, we would end up with lots of false existence claims which follow from (NCP). For example, if you take the set of properties definitory for centaurs and add existence to this set, (NCP) would tell you that there is an object that has all this properties, i.e. an object that is a centaur and exists. In order to avoid this absurd consequence, Meinongians introduce the distinction between so-called ‘nuclear’ and ‘extra-nuclear’ properties (or ‘konstitutorische’ and ‘außerkonstitutorische Eigenschaften’ as Meinong himself called them). The general idea behind this distinction is that nuclear properties have a certain discriminatory potential while extranuclear properties lack it. Only nuclear properties P are such that we can distinguish by them, for any given set of further properties, two non-empty sets of objects which have these further properties, namely those that, in addition, have P and those that do not have P. Existence is an extra-nuclear property because we cannot distinguish by it centaurs that exist from those that do not exist, simply because there are no centaurs that exist. (In contrast, humility is a nuclear property, and among the non-existing centaurs there are humble ones and non-humble ones.) The distinction between nuclear and extra-nuclear properties allows Meinongians to restrict their comprehension principle in such a way that the absurd consequence of (NCP) are avoided, namely in the following way (cf. Parsons 1980, 19):
347
(RCP) For every set S of nuclear properties, there is an object that has exactly all those properties which are elements of S. Since existence is not a nuclear property, we cannot use (RCP) in order to define all kinds of things into existence any longer. Although neither Kant nor the Meinongians give a precise definition of ‘real predicates’ or ‘determinations’ on the one hand and ‘nuclear properties’ on the other, it is clear that these two notions play a very similar functional role in their accounts. As soon as a theory assumes that objects in general can be divided into existing and non-existing objects, and claims that for objects in general it is true that there are certain kinds of such objects just because there are such-and-such individual concepts or such-and-such sets of properties, the theory has to mark certain properties as inappropriate for forming the content of these concepts, or for being elements of these sets, in order to avoid inconsistencies. The claim that existence is not a real predicate or that it is not a nuclear property is the claim that it is a property that suffers from this shortcoming. The second open exegetical question was this: If, for Kant, existence is a discriminating property of objects, what is meant by those of his remarks that seem to speak in favour of a meta-representational analysis of existential statements? One of these remarks occurs in the following passage: When existence occurs as a predicate in common speech, it is a predicate not so much of the thing itself as of the thought which one has of the thing. For example: existence belongs to the sea-unicorn but not to the land-unicorn. This simply means: the representation of a sea-unicorn is an empirical concept; in other words it is the representation of an existent thing. (The only possible argument, AA II 72)
How can Kant conceive of existence as a property of objects, and, at the same time, treat existential claims as claims about our own representations? The answer to this question lies in the fact, I think, that, in the case of merely possible objects, there is a very close ontological connection between these objects and the concepts we have of them. In the Critique of Pure Reason, Kant explains the difference between merely possible thalers and actual thalers in the following way: In the case of actuality the object is not merely included in my concept analytically (der Gegenstand ist bei der Wirklichkeit nicht bloß in meinem Begriffe analytisch enthalten), but is added synthetically to my concept […]. (A 599/B 627)
348
This passage implies that a merely possible object has being only ‘within’ the concept by which it is represented, while actually existing objects have being ‘beyond’ the concepts we have of them. In my opinion, the best way to spell out the spatial metaphor in this characterization is the following: In the case of merely possible objects, there is a very strong ontological connection between these objects and the individual concepts by which they are represented. As we have already seen, Kant’s comprehension principle (CP) assures that all it takes for such objects to have being is that there is a certain individual concept, hence individual concepts could not exist without representing possible objects. The quoted passage suggests that the dependence also holds the other way round: A possible object could not have being unless there was a concept by which it was represented. In the context of The only possible argument, this claim is confirmed by the fact that all possible worlds and all the possible objects which inhabit them are represented in God’s intellect before he chooses which world to make actual (cf. AA II 74). In the Critique of Pure Reason, the strong ontological connection between non-existing possible objects and the representations we have of them is documented by the fact that Kant calls these objects ‘objects of mere thoughts’ (‘Gedankendinge’). All this is evidence, I think, that for Kant, non-existing possible objects are essentially intentional objects. They are the internal representata of concepts and have being only insofar as someone represents them. In contrast, existing objects have being independently of the concepts we have of them. (They are ‘added synthetically to my concept’, as Kant puts it.) This means two things: On the one hand, it is not enough to investigate our own concepts in order to find out whether there are such objects, but we need some other source of knowledge, namely intuition, by which they are given to us. On the other hand, existing objects are not merely intentional objects because they would not cease to exist if nobody had thought about them.12 If non-existing and existing objects can be distinguished by being more or less ontologically dependent on the concepts by which they are represented, it is clear in what sense the existence-predicate ‘is not a predicate of the thing in itself but rather of the thought one has of this thing’. To say that sea-unicorns exist does not simply mean to say that there are sea-unicorns, but rather to say that there are sea-unicorns which have a certain relation to the concepts we have of them, namely that of 12. This is the case if we allow for a somehow ‘realistic’ reading of Kant’s transcendental idealism. I argue for such an interpretation in Rosefeldt (2007).
349
being ‘outside of ’ or independent of these concepts. As a consequence, we can reformulate the statement ‘Sea-unicorns exist’ by one about the concept of a sea-unicorn, namely the statement that this concept represents something ontologically independent of the concept itself. For Kant, the only justification for a statement like this would be that the objects which are claimed to be independent of our concepts are given to us by some other form of representation, namely intuition. This, in turn, would make the concepts applicable to objects of intuition and hence empirical concepts. It is for this reason, I think, that Kant writes that to say that sea-unicorns exist ‘simply means: the representation of a sea-unicorn is an empirical concept; in other words it is the representation of an existent thing’.
REFERENCES Bennett, Jonathan 1974: Kant’s Dialectic. Cambridge: Cambridge University Press. Forgie, J. William 2000: “Kant and Frege: Existence as a Second-Level Property”. Kant-Studien 91, 165–177. Frege, Gottlob 1979: “Dialogue with Pünjer on Existence”. In: G. Frege, Posthumous Writings, trans. Peter Long and Roger White, ed. Hans Hermes, Friedrich Kambartel, & Friedrich Kaulbach, Oxford: Blackwell, 53–67. — 1988: Die Grundlagen der Arithmetik. Eine logisch mathematische Untersuchung über den Begriff der Zahl. Ed. by Christian Thiel, Hamburg: Meiner. — 1990: “Dialog mit Pünjer über Existenz”. In: Schriften zur Logik und Sprachphilosophie. Aus dem Nachlaß. Edited by Gottfried Gabriel, 3rd edition. Hamburg: Meiner, 1–22. — 2002: “Funktion und Begriff”. In: Funktion, Begriff, Bedeutung. Edited by Mark Textor, Göttingen: Vandenhoeck & Ruprecht. Kant, Immanuel 1998: Critique of Pure Reason. Translated and edited by Paul Guyer & Allen W. Wood, Cambridge: Cambridge University Press. — 2002: Theoretical Philosophy 1755–1770. Translated and edited by David Walford, Cambridge: Cambridge University Press. Meinong, Alexius 1904: “Über Gegenstandstheorie”. In: Untersuchungen zur Gegenstandstheorie und Psychologie. Leipzig: Barth, 1–50. Parsons, Terence 1980: Nonexistent Objects. New Haven: Yale University Press. Pünjer, Bernhard 1874: Die Religionslehre Kants. Jena: Maucke’s Verlag. Reed, Delbert 2007: The Origins of Analytic Philosophy: Kant and Frege. London: Continuum.
350
Rosefeldt, Tobias 2007: “Dinge an sich und sekundäre Qualitäten”. In: Jürgen Stolzenberg (ed.), Kant in der Gegenwart. Berlin, New York: de Gruyter, 161–203. — 2008: “Kants Begriff der Existenz”. In: Recht und Frieden in der Philosophie Kants. Akten des 10. Internationalen Kant-Kongresses. Edited by Valerio Rohden et al., 5 vol., Berlin: de Gruyter, vol. 2, 669–678.
351
