General Terms and Relational Modality*
2011; Wiley; Volume: 46; Issue: 1 Linguagem: Inglês
10.1111/j.1468-0068.2010.00783.x
ISSN1468-0068
Autores Tópico(s)Epistemology, Ethics, and Metaphysics
ResumoNatural kind terms have exercised philosophical fancy ever since Kripke, in Naming and Necessity, claimed them to be rigid designators. He there drew attention to the peculiar, name-like behavior of a family of prima facie loosely related general terms of ordinary English: terms such as 'water', 'tiger', 'heat', and 'red'. Just as for ordinary proper names, Kripke argued that such terms cannot be synonymous with any of the definite descriptions ordinary speakers associate with them. Rather, the name-like behavior of these so-called natural kind terms is to be explained, just as in in the case of proper names, by the doctrine of rigid designation. And as a consequence of thus extending the notion of rigid designation to general terms, he famously endorsed the claim that so-called 'theoretical identifications', i.e. statements like Water is H2O, are necessary, if true. Intriguingly, and just like that of identity statements involving two different proper names like Hesperus = Phosphorus, the truth of such statements cannot be known a priori. Rigidity thus promises a semantic explanation of the modal status of the two known kinds of a posteriori necessity. Putnam endorsed the idea of a name-like semantic category of natural kind terms. He was, however, more interested in the peculiarities of what might be called their "foundational" semantics, than that of their semantics proper. By means of the famous Twin-Earth thought experiment (cf. Putnam 1975), he brought out some widely shared intuitions presumably supporting the claim that the meaning of natural kind terms is partly determined by elements of the local environment of the speakers. The idea of "externalist" meaning determination was then generalized, amongst others by Tyler Burge (cf. Burge 1979; 1982), not only to general terms other than natural kind terms, but also to natural kind concepts. Putnam/Burge style externalism has generated a huge literature, much of which is concerned with foundational semantics, or content determination, and the various problems externalist content determination engenders (for instance, with respect to self-knowledge). More recently, however, more attention has been focused on the very semantics of natural kind terms again. In particular, it has been asked what exactly it would mean for a general term to be a rigid designator. While there is a simple and clear definition of rigidity for singular terms, the extension of the notion of rigidity to general terms is not straightforward at all. Nor is it clear whether the expressions we are most interested in here, the natural kind terms, are to be treated as singular terms, or as predicates. Different proposals for both options have been discussed, and there seems to be a rather widespread consensus at the moment that none of them does precisely the job Kripke envisaged rigidity as doing for natural kind terms. Most importantly, none of the proposals on the market is able to sustain natural kind terms as a semantic category. There are, thus, two options at this point: We can either settle for explaining as much as we can of the peculiar, name-like behavior of natural kind terms by means of one of the available notions of rigidity for general terms. Or we can try to find different semantic properties of these terms that ideally would be able to explain all of the phenomena Kripke expected rigidity to explain. In this paper, we shall argue that there in fact are such semantic properties. What we shall call 'switcher-style dual-property association' is a family of semantic properties that do for general terms what rigidity cannot do. They even single out the natural kind terms as a semantic category. Moreover, they promise to explain what rigidity, by its very nature, cannot explain: The epistemic status of both a posteriori and a priori (or analytic) necessities. Our basic idea about natural language natural kind terms is very Kripkean, indeed. It is the idea that natural language natural kind terms are associated with two properties: a manifest, stereotypical property, and an underlying physical property realizing, instantiating, and (in many cases) explaining the manifest qualities of its instances.1 Natural kind terms are peculiar in that their modal profile is governed by the underlying property. To implement this idea formally, we shall extend the 'evaluation switcher semantics' we have earlier suggested for proper names and modal operators (Glüer and Pagin 2006, 2008). This 'evaluation switcher'– or, for short, 'switcher'– semantics works with more than one semantic function, and its basic idea is that some – typically non-extensional – operators function as semantic switchers in natural language. This idea is used to explain the modal peculiarities of certain expressions, such as proper names, while at the same time allowing them to have descriptive content: Such expressions interact with (de re) modal operators in a peculiar way. Evaluation switcher semantics, thus, promises to combine the advantages of traditional descriptivism with Kripkean and externalist insights about the modal profiles of proper names and natural kind terms. We shall proceed as follows: In the next section, we review what has happened in the discussion up to now. We look at the notion of rigidity as it has been suggested both for singular and for general terms. Of special interest are, of course, proper names and natural kind terms. We also look at the consequences rigidity is supposed to have for the modal status of certain identity statements. The most important question will be whether there is a satisfactory notion of rigidity for general terms. In section 3, we shall quickly run through the basic ideas of the evaluation switcher semantics for proper names and modal operators that we have suggested. In section 4, we shall explain how to extend that semantics to general terms when they can be construed as singular terms referring to kinds, and explore the consequences for the modal status of the relevant identity statements. In section 5 we shall do the same for general terms that have to be construed as predicates. In the last section, we shall argue that the suggested semantics for dual-property associated general terms in fact singles out the natural kind terms as a semantic category. Besides explaining the modal status of identifications involving natural kind terms, switcher-style dual-property association thus also promises to explain their epistemic status. We shall add some comments on natural kind terms and reference failure, and end on a Quinean note of caution about the notions of the a priori and the analytic employed in this paper. In Naming and Necessity (1972), Kripke introduces the idea of rigid designation as follows: Let's call something a rigid designator if in every possible world it designates the same object (…). Of course we don't require that the objects exist in all possible worlds. (Kripke 1972, 48) This yields a straightforward definition of rigidity for singular terms: A singular term t is a rigid designator iff t refers to the same object o in every possible world w in which o exists. Kripke then presents a host of detailed and forceful arguments for the claim that proper names in natural language are rigid designators. If proper names are rigid designators, they cannot be synonymous with any of the (non-rigid) definite descriptions that speakers normally associate with them. Kripke presents substantial evidence to this effect. The maybe most powerful and influential argument Kripke uses to show this, is the so-called modal argument.2 Take two modal sentences for instance Aristotle might not have gone into pedagogy. The teacher of Alexander might not have gone into pedagogy. Intuitively, (3) is true. But (4) is ambiguous. It has the following two readings: The teacher of Alexander is such that: possibly he did not go into pedagogy. It is possibly the case that: the teacher of Alexander did not go into pedagogy. On one of these readings, reading (4b), (4) is intuitively false. Since (3) has no such ambiguity, 'Aristotle' and 'the teacher of Alexander' are not intersubstitutable salva veritate in all contexts, and therefore not co-intensional. Kripke concludes that proper names and the (non-rigid) definite descriptions speakers normally associate with them are not synonymous. He suggests to explain the behavior of proper names in modal contexts by construing them as rigid designators. Rigidity is also the key to explaining another notorious modal phenomenon connected with proper names. Take identity statements involving two different proper names, for instance Hesperus = Phosphorus. Let's call such identity statements 'two-name identities'. If proper names are rigid designators, two-name identities are necessary, if true. More precisely, they are true in every world in which the denoted object exists. If, as is most natural, we allow truth-value gaps where there is reference failure, two-name identities are either true or truth-valueless in all worlds, and nowhere false. Yet, two-name identities like (2) are not analytic, and their truth cannot be known a priori. Such two-name identities thus are examples of what has come to be known as a posteriori necessities.3 By contrast, a 'one-name identity' like the following Hesperus = Hesperus, has at least a certain air of being analytic, and knowable a priori. That is, it appears to be analytic, and knowable a priori, that such a sentence is not false at any world. In this sense, it appears to be an analytic necessity. Rigidity does explain why (2) is necessay, if true. But rigidity does not explain why (2) is a posteriori, or why it is non-analytic. Nor does it explain why (5) is both analytic and a priori. Rigidity, in other words, can only explain modal status, not epistemic or 'informational' status. That should not be surprising, as rigidity, by itself, does not tell us anything about the meaning of rigid terms. Nevertheless, we would of course like to have a full explanation of the peculiar modal-cum-epistemic status of two-name identities like (2). Moreover, given the strong appearance of a connection between epistemic status and meaning, we would prefer this explanation to be a semantic explanation – just as Frege wanted a semantic explanation of the non-modal informational difference between one-name and two-name identities. For singular terms, it is thus well-understood both what rigidity amounts to, and what consequences it has for modal status. On the assumption that the proper names of natural language are rigid designators, rigidity elegantly explains their behavior in modal contexts and the modal status of two-name identities. Nevertheless, by itself, rigidity does not provide all the semantic explanation we desire for the peculiar epistemic-cum-modal properties of name identities. When it comes to general terms, the situation for a long time appeared to many as just parallel to that concerning singular terms. In Naming and Necessity, Kripke singles out a certain kind of natural language general term: the natural kind terms. Natural kind terms, he claims, are very much like proper names; they are rigid designators. Just as for proper names, Kripke argues that if a general term is a rigid designator, it cannot be synonymous with the non-rigid, stereotypical descriptions speakers normally associate with it. And again, some evidence for this is marshaled from modal contexts. Take two modal sentences like the following: Water might not have been H2O. The clear, thirst-quenching liquid flowing in rivers and from taps might not have been H2O. Intuitively, (6) is false, while (7) has a reading on which it is true. Such substitution failure indicates non-synonymy, and as in the case of proper names, Kripke proposes to explain this by means of rigid designation. It has become clear in the recent literature, however, that the supposed parallel between proper names and natural kind terms in fact is quite problematic. The examples of natural kind terms Kripke uses are of surprising variety. There are count nouns such as 'cat', 'cow', and 'tiger', mass terms such as 'gold', 'water', and 'iron pyrites', terms for natural phenomena such as 'heat', 'sound', and 'lightning', and adjectives such as 'hot', 'loud', and 'red'. All of these are, according to Kripke, terms for natural kinds. Even though these examples might prima facie look like a weird bunch, they all have something interesting in common. All these terms are such that they show a characteristic "double nature": Each of them seems to be associated with two different properties. One of these properties is specified by the descriptive stereotype normal speakers associate with the term. Thus, for instance, 'water' is associated with something like 'the clear, thirst-quenching liquid flowing in rivers and from taps', and 'tiger' with 'the carnivorous, cat-like animal with yellow and black stripes'. The stereotypical property typically is a manifest one. The second property is the physical (chemical, biological, etc.) property underlying, realizing, and (in many cases at least) explaining the manifest one. Intuitively, natural kind terms are used to talk about, or predicate, the underlying property; this is the property things need to have in order to belong to the kind, i.e. what is essential to being of that kind.4 The manifest property is merely used to 'get at' the underlying one, to 'fix' the term to it. Consequently, speakers apply these terms to actual objects on the basis of their manifest properties, but regard the modal properties of the kind itself as determined by the underlying property. Thus, nothing but H2O could possibly be water.5,6 Now, if general terms of this kind were rigid designators, the prediction would be precisely such a modal profile. Most importantly, if they were rigid designators, that should have tell-tale consequences for the modal status of sentences such as Water is H2O, Tigers are animals, or Heat is mean molecular kinetic energy. All of these should come out as (a posteriori) necessities. Sentences like these are called "theoretical identifications" since it is not altogether obvious what their logical form is; (1) and (9) might be identity statements, but (8) clearly is not an identity statement. Whatever their logical form, Kripke claims that – just like two-name identities – theoretical identifications indeed are necessary, if true. Moreover, he claims, theoretical identifications are not analytic, and not knowable a priori, either. They, too, thus seem to be a posteriori necessities. In our opinion, these predictions are born out by our modal intuitions. Natural kind terms do show a peculiar, name-like behavior both when it comes to substituting stereotypical descriptions for them in modal contexts and when it comes to the modal status of theoretical identifications. These intuitions might in fact be a bit weaker than those regarding the corresponding behavior of proper names, but we take them to be strong and stable enough to be taken as data requiring (semantic) explanation.7 We take it to be clear that Kripke thinks that – just as in the case of proper names – the key to the explanation is rigidity (cf., for instance, Kripke 1972, 140; 162; see also Soames 2002, 244f). Again, such an explanation – if forthcoming – is limited to explaining the modal status of the relevant sentences. Rigidity by itself does not explain the epistemic status of theoretical identifications. Again, epistemic status appears to be connected with meaning, and if it is, rigidity can explain only part of the relevant semantic phenomena. Moreover, in the case of general terms, it is not so clear in the first place what their rigidity exactly is supposed to amount to. The most natural extension of the notion of rigidity for singular terms to general terms is this: A general term G is a rigid designator iff G has the same extension in every possible world w in in which it has a non-empty extension. But (RDG) is clearly inadequate. Most general terms, including natural kind terms, have extensions that vary from world to world. So, what does it mean for a general term to be rigid? Before we look at the most prominent, and promising, suggestions on the market, let us follow Scott Soames (2002) in listing what seem to be rather sensible, natural desiderata on a notion of rigidity for general terms: A satisfactory notion of rigidity for general terms must be a natural extension of the notion of rigidity as defined for singular terms, apply to all natural kind terms, but not to other general terms, and explain the modal status of theoretical identifications. (Cf. Soames 2002, 263) However, it has been contended in the recent literature that there is no notion of rigidity for general terms that satisfies all of these desiderata (cf. a.o. Soames 2002, 263; Haukioja 2006).8 This might not be all that surprising in the end; after all, as Kripke's examples amply demonstrate, the natural kind terms do form a syntactically rather heterogeneous bunch – including count nouns, mass nouns, and even adjectives. To deal with their semantics, it would be helpful to know whether natural kind terms primarily are singular terms or predicates. If they are primarily one or the other, it might be possible to explain what it means for a natural kind term in its primary use to be rigid, and to then treat the other use as somehow derivative. However, it might also be the case that we simply would need two notions of rigidity for general terms, one for singular terms and one for predicates. In this paper, we are not going to take a stand on the question whether natural kind terms primarily are one or the other, singular terms or predicates. Rather, we are going to proceed on the assumption that there are (uses of) natural kind terms that clearly are singular, and that there are (uses of) natural kind terms that clearly are predicative. And in the remainder of this section, we are going to look at the two main suggestions for general term rigidity, one for predicates, and one for singular terms. If natural kind terms are predicates, the main contender for rigidity on the market is what has been called "rigid application":9 A predicate P is a rigid designator iff P is such that if it applies to an object o in any possible world w, then it applies to o in every possible world in which o exists. Predicates satisfying (RD′G) have also been called 'essentialist predicates'. Such predicates specify properties that any object that has them at all cannot fail to have, properties (in this sense) essential to these objects' being the objects they are. When it comes to natural kind terms that are singular terms, the single most important suggestion is that such terms rigidly refer to kinds:10 A general term G is a rigid designator iff G refers to the same kind k in every possible world w in which k exists. These are the most important suggestions currently on the market as to how to understand rigidity for general terms, and the only ones we shall discuss here. Regarding the desiderata just listed, it is pretty clear that both of these satisfy the first one: Both are natural extensions of the notion of rigidity as defined for singular terms. But what about the other two? Let's look at the notion of rigid application first. Does this notion apply to all, and only, natural kind predicates? Predicates like 'is water' and 'is a tiger' would indeed seem to be rigid appliers. The corresponding stereotypes, by contrast, would seem not to be rigid appliers: That which is clear, thirst-quenching, liquid, and flowing in rivers and from taps is not essentially, or necessarily, so. And even though all animals might essentially be animals, those which are carnivorous, cat-like, and have yellow and black stripes would not seem to be essentially, or necessarily, so. Moreover, paradigm cases of non-natural kind predicates, like 'is a bachelor' or 'is a hunter' do not seem to be rigid appliers, either. It does not seem essential to (most) bachelors that they are bachelors. And the same for hunters. Trouble starts when it comes to predicates like 'is hot', 'is loud', and 'is red'. These are included in Kripke's list of examples, but they clearly are not essentialist predicates. It is not essential to most (maybe all) red things that they are red. Nor is it essential to hot things that they are hot, or to loud ones that they are loud. In fact, properties like these seem rather paradigmatically accidental. Now, at this point one might of course argue that these terms should not be counted among the natural kind terms. Kripke was wrong to include them, you might say; for surely, the red things do not form a natural kind? For one thing, the natural kind designated by the corresponding natural kind terms arguably is not so much the kind of, for instance, red things, but simply the color red itself (cf. Soames 2008). For another, as we said above, what seems to unify the examples Kripke gives is the specific way they are associated with two different properties. Whether terms for color, sound, and temperature should be included among the natural kind terms depends on whether they in fact are so associated. Clearly, in all these cases, two different properties of the right kinds are in play: In all these cases, objects are characterized in terms of their manifest, sensible qualities. And in each case, there naturally are underlying physical (or chemical, etc.) properties responsible for the objects' sensible qualities. Moreover, Kripke is very clear on the further, crucially important claim that in all of these cases it is the underlying property that forms the essence of the kind. According to him, the relevant question is thus not whether it is essential to a kind's instances that they are of that kind, but whether the kind itself is such that its essence resides in its physical (or chemical, etc.) nature. Whether that is actually the case when it comes to color, for instance, is of course debatable. In the case of heat, on the other hand, intuition seems rather clear: The essence of heat is in its physical nature, not in the way it affects sensible creatures like us. Consequently, even if we disagree with Kripke on the particular case of 'red' and other color terms, there do seem to be clear examples of natural kind terms that are not rigid appliers. Moreover, there clearly could be any number of such terms in natural language, terms showing exactly the characteristics Kripke was after. Since the notion of rigid application would exclude these terms from the category of natural kind terms, this shows that rigid application does not capture the nature of the terms Kripke was interested in.11 Anyone accepting the reasoning just presented will have to conclude that the notion of rigid application does not succeed in applying to all, and only, natural kind predicates. While it clearly draws a distinction between different kinds of general terms, it is far from clear that it is the distinction between natural kind terms and other predicates.12 What might be worse, the notion of rigid application does not explain the modal status of theoretical identifications, either. Let's take (1) as our example: Water is H2O. Since 'water' and 'H2O' are construed as predicates, (1) has the form ∀x (water(x) ↔ H2O(x)). Then, the question is: Does rigid application explain why □ ∀x (water(x) ↔ H2O(x)) is true if (1′) is? And the answer seems to be negative. As Soames (2002, 257ff) has argued, rigid application explains why ∀x □ (water(x) ↔ H2O(x)) is true, if (1′) is, but not why (10) is. That is, rigid application has the consequence, and explains, why all objects that exist in the actual world and are both water and H2O, are both water and H2O in every world where they exist. But that does not rule out the possibility of a world w in which there is something that does not exist in the actual world and that is water in w, but not H2O. And this generalizes, of course. Consequently, the notion of rigid application does not satisfy the third desideratum on a notion of rigidity for general terms, either.13,14 Let us next look at the notion of rigid designation of kinds. The idea would be that general terms designate universals, properties, or kinds, and that rigid general terms designate the same universal, property, or kind in every possible world (where they designate anything at all). Natural kind terms would seem to be rigid general terms in this sense, but does that distinguish them from other general terms? The answer seems to be clearly negative. General terms for paradigmatically non-natural kinds like 'bachelor', 'hunter', or 'pencil' would seem to rigidly designate those non-natural kinds just as much as 'water' or 'tiger' would designate their natural ones (cf. Schwartz 1980; 2002; Soames 2002; 260ff). With López de Sa, we can call this the 'over-generalization problem' (2008b). The only significant contrast that could plausibly be drawn by means of the notion of rigidly designating kinds is one between all such general terms and certain complex descriptive phrases. Examples for the latter would be the explicit stereotypes associated with 'water' and 'tiger': 'clear thirst quenching liquid flowing in rivers and from taps' and 'carnivorous cat-like animal with yellow and black stripes'. Other, much discussed examples include 'Mary's favorite color' (Martí 2004) and 'the color of the sky at noon' (Salmon 2005; Linsky 2006). All of these descriptive phrases arguably designate different properties in different possible worlds.15 There seems to be growing consensus that this contrast suffices to shield rigid designation of kinds from utter triviality.16 But even though the contrast drawn is an interesting one, abandoning 2) has an air of making a virtue out of a necessity. After all, Kripke does single out the natural kind terms, and he clearly thinks that their characteristic double structure of manifest stereotype and underlying physical property is of semantic significance. Even though the notion of rigid designation of kinds does not satisfy desideratum 2) and, thus, cannot capture this semantic significance, to us at least it still seems very worthy of capture. What about explaining the modal status of theoretical identifications, then? Take (1) again. It now has the form of a simple identity statement: Water = H2O. Moreover, □ Water = H2O, would clearly seem to be true (where it has a truth-value) if (1″) is. The notion of rigidly designating kinds thus does account for the modal status of theoretical identifications. These are here construed as statements about property identity. The same pattern – necessary, if true – does of course also hold for sentences like the following: Water = Water. Bachelorhood = unmarried-manhood. No distinction is drawn between these a priori, analytic necessities and the aposteriori (1). And again, even though that was not to be expected of a notion of rigidity anyway, it shows rigidity's limitations as a semantic explanans. To summarize: Both rigid application and rigid designation of kinds are natural extensions of the notion of rigidity for singular terms. Rigid application, however, does not satisfy desiderata 2) and 3). Rigid designation of kinds satisfies 3), but fails 2). Here is the conclusion Soames once drew from this discussion: [T]here is no reason to think that any notion of rigidity for [general terms] is capable of satisfying all of these demands. (…) Thus, we need to look to other semantic properties of names that may be shared by natural kind [terms], and that may play an important role in explicating the modal properties of sentences containing such predicates. (…) Even in the case of proper names, it can be argued that their rigidity is the result of other, more fundamental, semantic properties that they possess. More specifically, the doctrine that names are rigid designators may be viewed as a corollary of the more central thesis that they are nondescriptional, together with an account of how their reference is fixed in the actual world. (Soames 2002, 263f, emph. ours) It seems clear that there is no notion of rigidity for general terms that satisfies the desiderata we considered. We tend to think that this is bad; it would be good to find semantic properties that actually do the work described by these desiderata. Settling for something like rigid designation of kinds is settling for second best. We might have to do that in the end, but only if there are no "other semantic properties of names that may be shared by natural kind [terms], and that may play an important role in explicating the modal properties of sentences containing such predicates" (ibid.). In what follows, we shall suggest that there are such properties. As the quote just given shows, Soames himself thinks of the rigidity of proper names as a mere corollary of their non-descriptionality, and of non-descriptionality as one of the key factors underlying the analogies in the behavior of proper names and natural kind terms. It is here that we part company. In Glüer and Pagin 2006 and 2008, we have suggested what we call an 'evaluation switcher semantics' for proper names and modal operators. This semantics allows names to be descriptive while interacting with the modal operators in the intuitive way. In what follows, we shall apply the basic idea of the switcher semantics to general terms. Elaborating on a rather tentative suggestion in Glüer and Pagin 2007, we shall show that the extended switcher semantics can be used for explaining the behavior of natural kind terms without having to rely on either a notion of rigidity for general terms, or their alleged non-descriptionality. Moreover, such a semantics draws some rather interesting distinctions between a priori and a posteriori necessities, thus promising a full semantic expl
Referência(s)