Almog was Right, Kripke’s Causal Theory is Trivial

Abstract

Joseph Almog pointed out that Kripkean causal chains not only exist for names, but for all linguistic items (Almog 1984: 482). Based on this, he argues that the role of such chains is the presemantic one of assigning a linguistic meaning to the use of a name (1984: 484). This view is consistent with any number of theories about what such a linguistic meaning could be, and hence with very different views about the semantic reference of names. He concludes that the causal theory is ‘rather trivial’ (1984: 487). In this paper I argue that Almog is correct to hold that the causal theory is trivial, but, contra Almog, argue that the triviality of the existence of causal chains is not a matter of such chains having a presemantic role in assigning linguistic meanings to utterances. Instead, such triviality is due to the fact that the causal theory reflects no more than a truism about the epistemology of convention acquisition.

In ‘Semantical Anthropology’, Joseph Almog pointed out that Kripkean causal chains not only exist for names, but also for indexicals and other linguistic items (Almog, 1984: 482). He uses this fact to argue that, in fact, causal chains do not serve to explain facts about reference and semantics. Instead their role is the presemantic one of assigning a linguistic meaning to an utterance of a name, and thereby also serving as a standard that allows different uses of a name (e.g. different uses of ‘Aristotle’) to have distinct meanings (1984: 484). This view is consistent with any number of theories concerning the semantic reference of names and so he concludes that the causal theory is ‘rather trivial’ (1984: 487). In this paper I explain and radicalize Almog’s doubts about role of causal chains in semantics and reference.

In section one I explain Almog’s view in more detail. In section two I argue that the problem is worse than Almog feared. The existence of such causal chains is not only accounted for by their presemantic role of assigning a meaning to an utterance of a name. The existence of such chains also can be accounted for by the epistemic role that such chains play in the transmission of all conventions, including non-linguistic conventions. In section three I argue that the causal theory has no (non-trivial) explanatory value and in section four I defend the resulting view against some anticipated objections.

1 Almog’s Doubts About the Causal Theory

Almog considers the cause of a Martian, named ‘E.T’, who has to report on how Earthlings use the indexical ‘I’ (1984: 480–482). E.T reports that those currently on Earth use the indexical ‘I’ in accordance with a practice that originated a long time ago. This has the effect that current users of ‘I’ stand in a causal chain of transmission that connects them to those who originated the practice whereby ‘I’ gained its linguistic meaning. The same holds true for all linguistic items, current use is always causally connected to originating use (1984: 481).

Almog points out that the existence of such causal chains cannot imply the adoption of a causal theory of reference, for causal notions form no part in formulating the semantics of indexicals like ‘I’ (1982: 482). Instead the role of such chains is presemantic. Causal chains serve to assign a linguistic meaning to ‘I’ and to make it the case that all utterances of ‘I’ in such a chain have the same linguistic meaning. The relevant chains do so in virtue of correlating the syntactic shape ‘I’ with the linguistic meaning that it was historically endowed with (1984: 480–482). Almog states that the same applies to names. Causal chains preserve the linguistic meaning¹ of names and are not ‘part of the rule of reference that formulates the meaning of the name’ (1984: 482).

In dismissing the idea that causal chains form part of the rule of reference that that formulates the meaning of a name, Almog dismisses the idea that names can be treated on the model of indexicals. He calls this the ‘indexical gambit’ (1984: 484); the idea that distinct uses of ‘Aristotle’ should be seen as one word, with one meaning, but distinct referents. Instead he adopts what he calls the ‘lexical ambiguity gambit’ (1984: 484), referring to the distinct uses of ‘Aristotle’ as ‘homonyms’, i.e. multiple names with multiple linguistic meanings (1984: 483). Such homonyms are individuated in virtue of having distinct linguistic meanings, with the causal chains serving to assign such distinct linguistic meanings to distinct uses of the generic name ‘Aristotle’. Call this the linguistic assignment view of causal chains.

Note that there is a subtly different view that, based on his arguments, Almog could have adopted. In the linguistic assignment view, the causal chain serves to assign a linguistic meaning, and this linguistic meaning then serves to individuate the name. On an alternative view, the causal chain itself serves to individuate the name by a causal chain leading back to a baptismal event, independently of its linguistic meaning. Distinct uses of ‘Aristotle’ can then be distinct names in virtue of tracing back to distinct baptismal events, but without the notion of linguistic meaning being conceptually needed to render such use the use of distinct names. This view would, like the linguistic assignment view, also render the role of causal chains presemantic. It has the virtue that one can now coherently speak of a single word changing its meaning, whereas the linguistic change view would have to see such changes as a matter of the name itself becoming a distinct name.

Almog argues that, as the role of causal chains is presemantic, the mere existence of such chains cannot be used to settle the core dispute between the Millian and the Fregean. The Fregean can modify their view of names so that it admits the existence of such chains, but state that the linguistic meaning that a name was was originally endowed with stipulates that the semantic output of a name is its Fregean sense (1984: 486).

It would be a a radical move for the Fregean to attempt to save the notion of a sense in this manner. While Almog does not say so explicitly, it would save Fregean senses at the cost of giving up on the idea that names are idiolectical. Instead the sense of a name would be singly tied to the name via an originating event whereby the linguistic meaning, which stipulates that the semantic output of the name is a Fregean sense, becomes associated with the name. On such a view the sense of a name ceases to be individually variable, and simply becomes a function of the name’s supra-individual linguistic meaning.

The above quasi-Fregean move, radical as it may be, is a coherent option for the Fregean. No conclusion about semantics can be deduced from the mere existence of the causal chains that accompany names. This can be seen from the fact that, as Almog points out, the use of indexicals (and all other linguistic items) also generate such causal chains. If, despite their vastly different semantics, all linguistic items generate such chains, then no conclusion about semantics follows from the existence of such chains. This implies that the quasi-Fregean cannot be dismissed on causalist grounds.

Almog concludes as follows:

E.T.’s move may seem to take all the air out of the big balloon of “essential role of causal (historical) chains in semantics”. Those who put the chains into senses or reference rules have looked at it as a semantic oracle, a universal key to the understanding of reference. They have endowed the chain with unlimited powers. E.T. turns the almighty into the rather trivial. But, he says, to realize that the almighty is rather trivial is not trivial at all (1984: 487).

Almog’s claim that causal chains are, ultimately, a trivial matter when theorizing about matters of reference and semantics, is unorthodox². For Kripke is commonly credited with having proposed a substantive view about the so-called ‘mechanism of reference’³, i.e. for having offered an explanation of how names refer to their referents. Below I argue that, while his conclusion could have been arrived at via an easier route, Almog was right. Kripkean causalism does not teach us anything non-trivial about semantic reference.

My argument will also radicalize Almog’s doubts. Almog claims that such causal chains exist for all linguistic items. I will show that they exist for all conventions, whether linguistic or not. Almog further claims that the role of such chains is not semantic, but presemantic. I will argue that they have an even more basic role that that is compatible with such chains being neither semantic, nor presemantic. This also implies that their triviality is not merely a matter of not serving to decide the dispute between the referentialist and the Fregean. The problem is worse. The causal theory is trivial as the existence of such chains is a matter of epistemology, not reference.

One caveat is required before proceeding. While causal chains will exist for all linguistic items in (almost) all cases, the idea of such chains terminating in baptismal events or explicit baptisms won’t realistically capture the history of common nouns, verbs and adjectives⁴. Furthermore, these terms are subject to polysemy and other phenomena that render the idea of a single, stabilized and shared linguistic meaning, dubious at best. In real world usage, especially once we move beyond names, we can expect linguistic items to shift and change their meanings via the kinds of bottom-up, informal mechanisms outlined in Lewis (1969), with meaning adjustments constantly being made pragmatically and some such changes resulting in new linguistic conventions⁵. It is for this reason that, while I do commit to (almost) all linguistic items being passed along in causal chains, this should not be read as saying that all such items have stable meanings or are typically explicitly adopted.

2 Causality and Convention

2.1 The Claim to be Defended

When I refer to the causal theory of reference I mean to refer to Kripkean causalism, i.e. the view proposed in Naming and Necessity. On this view, a name N semantically refers to the object baptized N at the beginning of the causal chain from which the user of N inherited N. Call this the coincidence claim, i.e. the claim that the referent of a downstream use of a name N is identical to the object baptized N at a causally related baptismal event. It is important to recognize that the causal theory, so stated, makes no more than a claim of correlation, i.e. it claims that baptismal reference and downstream reference coincide. The coincidence claim is distinct from reductive causalism, i.e. the view that semantic reference just is a kind of causal relation. Reductive causalism would be one potential theory as to why the coincidence claim holds true, but one can commit to the coincidence claim without committing to reductive causalism.

Kripke is not a reductive causalist; he only ever commits to what I have termed the coincidence claim. Kripke explicitly says that his theory does not eliminate the notion of ‘reference’ (1981: 97), that it may well be impossible to do, and approvingly quotes Bishop Butler’s dictum that ‘everything is what it is, and not another thing’ (1981: 94). He also says that the notion of ‘reference’ is presupposed in his theory twice, namely in its appeal to the notion of ‘intending to use the same reference’⁶ and in the notion of an initial baptism (1981: 97). Hence Kripke cannot be ascribed the view that semantic reference just is a kind of causal relation⁷.

Strictly interpreted, of course, the coincidence claim is false. As first pointed out by Evans (1982), there are cases, most famously ‘Madagascar’, where the semantic reference of a name changes from that established by an initial baptism. Such cases, however, are rare; the coincidence claim holds true for the vast majority of token utterances of names. I take it that the idea that Kripkean causalism tells us something profound about semantic reference is mostly due to the fact that the coincidence claim is almost always true. Call the claim that the coincidence claim will almost always be true the modified coincidence claim. My defense of Almog’s conclusion that causalism is trivial will here take the form of arguing that the modified coincidence claim has no non-trivial explanatory content.

Note that the claim that Kripkean causalism has no non-trivial explanatory content is not identical with, nor does it imply, the claim that Kripke’s view on names is trivial. Naming and Necessity contains a lot beyond the causal theory of reference. Kripkean doctrines concerning rigid designation, metaphysical necessity and the like are unaffected by the arguments presented here. I will return to this point again in this paper.

2.2 Name Individuation

It has already been explained that Almog views causal chains as serving to pair syntactic shapes with their linguistic meanings, and thereby to individuate such syntactic shapes. In this way two utterances of ‘Aristotle’ can be tokens of two distinct names (i.e. the philosopher and the shipping magnate), in virtue of the fact that the token utterances trace back to distinct originating events where distinct linguistic meanings were assigned (1984: 483). The idea that such chains play an individuating role is not unique to him; Kaplan also suggested that the role of causal chains is presemantic and that ‘[t]he causal theory of reference tells us… which word is being used in a given utterance’ (1989: 562).

This way, however, of accounting for the role of causal chains in the theory of reference, has two drawbacks when attempting to cast doubt on the view that Kripke’s theory has non-trivial content. The first is that, in locating the role of such chains in the individuation of names, it immediately runs into opposition from those who believe that names are generic. On such a view, the distinct uses of ‘Aristotle’ mentioned above do not serve to indicate that we are dealing with distinct names that just happen to share a phonetic and orthographic type⁸. Instead the people named ‘Aristotle’ should be said to have the same name; we may at best speak of distinct uses of a name.

The second problem is that such a view will run into opposition from those who agree that names are specific (i.e. non-generic), but deny that such specificity is to be achieved via a causal theory of name-individuation. Kripke himself, for example, explicitly denied that his theory is supposed to be read as providing a standard for the individuation of names. He mentions, but explicitly states that he does not accept or endorse, the possibility of individuating names by the baptismal event causally responsible for their use (1981: 8n). Instead he adopts a terminology in which phonetically identical names with distinct referents count as distinct names (1981: 7–8).

If Almog’s claim that the causal theory is trivial is presented as depending on adopting a defensible, yet far from universal, view on name-individuation, then the triviality claim is easy to deny for those who do not adopt the causal standard for the individuation of names. Below I will make an argument for Almog’s conclusion that renders the matter of name-individuation a red herring. For the argument for the triviality of Kripke’s causal theory can be stated in a way that is independent of the matter of name-individuation, and arises no matter which view we adopt on this issue.

2.3 The Epistemology of Convention Acquisition

Almog makes much of the fact that utterances of all linguistic items, i.e. not only names, stand in causal relations to events whereby the specific use of a linguistic item originated. This fact, then, lead to the suspicion that these causal chains have nothing to tell us about the semantic reference of names specifically, but rather that their existence is a mere artifact of the the fact that names are linguistic items. The problem, however, is worse. It is not only the case that such chains exist for all linguistic items. It takes only a moment’s reflection to see that, except in the most recherché of cases, such causal chains are generated by all conventions, whether linguistic or not.

Consider, for example, the convention of driving on the left-hand side of the road in the United Kingdom. The current practice of driving on the left in the UK stands in a causal relationship to the event whereby the convention of driving on the left was originally adopted in the eighteenth century. It is no mystery why such a causal chain, stretching back centuries, exists. I learned that I should drive on the left-hand side of the road in the UK in virtue of having observed others doing so, the people I observed driving on the left learned it from being told to do so, or observing others doing so, and so on. This chain of knowledge acquisition terminates in the event whereby it came about that this convention was established in the UK. This implies that there is a causal chain that stretches from my driving on the left to the event whereby this convention came about. The same goes for any convention we care to name, e.g. the convention of using ‘and’ to express conjunction in English, the convention whereby the original caller calls back when a phone-call is dropped, and so on.

It is easy enough to explain the existence of such causal chains; their existence reflect no more than the fact that knowledge of the content of a convention is typically acquired causally. This fact, however, is a mere truism about the epistemology of convention acquisition. Such chains are generated by all conventions, yet no-one thinks that this teaches us anything theoretically interesting about, for instance, our practice of driving on the left-hand side of the road in the UK. Similarly, no-one has been tempted to construct a ‘causal theory’ of the correct side of the road to drive on.

The matter of name-individuation turns out to be somewhat of a red herring in the context of trying to determine whether Almog is correct in claiming the causal theory to be trivial. The existence of causal chains is explained by the epistemology of convention acquisition. This is so, independently of whatever view we care to take concerning the individuation of names⁹.

The above then, serves to explain why the coincidence claim, i.e. the claim that a name N semantically refers to the object baptized N at the beginning of the causal chain from which the user of N inherited N, is (almost) always true. It follows from the fact that a baptism is an event whereby a convention, which links an individual to a specific name, is introduced, coupled with the fact that the content of such a convention is subsequently causally learned by downstream users.

Note that there is nothing in this that tells us anything substantive about the ‘mechanism of reference’, nor does it serve to establish anything interesting about the semantic content of a name. The coincidence claim - and Kripke commits to no more than the coincidence claim - does not say anything about the nature of the link between the name and the individual named. It is for this reason that the coincidence claim is not only compatible with Millianism about semantic content, i.e. the view that the relevant convention merely pairs a name and an individual as its semantic content, but is also, as pointed out by Almog, compatible with (a modified) Fregeanism, according to which the semantic output of the name is a Fregean sense, which then serves to pick out the semantic referent of the name (1984: 486). It would also, for that matter, be compatible with a view on which semantic reference is conventionalized speaker’s reference, i.e. a view on which the conventional rule¹⁰ governing the relation between a name N and its referent o is ‘use N to speaker-refer to o’¹¹. The coincidence claim, in the final analysis, would be true on any theory that accepts that names are subject to public conventions, that baptismal events involve individuals, and that the content of conventions is learned causally. It is for this reason that the coincidence claim will typically give the correct answer to the question ‘Who does N refer to?’. Yet it does not establish that causal notions need to be constitutively involved in the determination of semantic reference.

Note that the coincidence claim itself does not award the baptism any non-historic role in the mechanism of reference. The coincidence claim is compatible with claiming, simply, that the current utterance of a name semantically identifies a specific individual in virtue of the fact that there is a current convention that somehow pairs the name and the named individual. On such a construal there is nothing concerning semantic reference that rests on past acts of baptisms in a constitutive way. The past serves to explain how the current convention came to be, but there is nothing that ‘reaches back over the centuries’ in the mechanism of reference itself. In the same way that the matter as to the correct side of the road to drive on in the UK at present, is settled by the current convention governing driving in the UK, so the coincidence claim is consistent with claiming that ‘Biden’ currently refers to Biden in virtue of the current convention governing ‘Biden’. The past serves to explain how our naming conventions - like our driving conventions - came to have the content that they do¹². But the coincidence claim does not force us to involve the past, or causal notions, in our positive theories of the nature of semantic reference itself¹³.

The coincidence claim effectively states that the semantic referent of a downstream use of a name is identical to the semantic referent established at the causally related baptism. One way in which it can fail to be the case that downstream use is co-referential with baptismal use is if the content of the convention has been changed from the one established at the causally related baptism, without such a change taking place in virtue of a new baptism¹⁴. This can happen if there somehow occurs a situation whereby an unwitting change in the semantic referent of the name takes place. Such cases, however, i.e. cases like Evans’s ‘Madagascar’ (Evans, 1982), are extremely rare, and hence the modified coincidence claim, as defined earlier, will hold true.

The view that Kripke’s causal theory is a truism about the epistemology of convention acquisition also implies that rare cases where we have knowledge of a baptism, but without prior causal contact with such a baptism, should not trouble us. Consider the case, adapted from Searle (1983: 239), of an architect who, when drawing up plans for a new city, always uses a consistent numbering scheme to name streets, and always does the layout in a specific way. In such a case my knowledge concerning the architect’s naming practices may allow me to have knowledge about the street named ‘5th avenue’, without being in causal contact with the event whereby 5th avenue was named. Such cases present headaches to those who wish to present the causal theory as a substantive theory of semantic reference. On the view that the causal theory is no more than a truism about how we come to have knowledge about baptismal events, however, such cases are no longer puzzling. These cases reflect no more than the fact that, while knowledge of baptisms is typically acquired causally, there can be odd cases where such knowledge is acquired without causal contact with the event whereby the thing is named. Once the causal theory is recognized as epistemic, nothing about this is surprising or theoretically troublesome.

My claim, then, is that Almog’s conclusion as to the triviality of the causal theory is correct, though such triviality does not arise from the presemantic matter of assigning a linguistic meaning to the utterance of a name. Kripke makes no more than the coincidence claim, and the truth of the coincidence claim is a mere artifact of a truism concerning the epistemology of convention acquisition.

I do not, of course, doubt that the causal theory is surprising. There is something delightful about the fact that there is a causal chain stretching back from my use of ‘Aristotle’ to the actual Aristotle. Almog’s claim, as defended here, however, is not the claim that the modified coincidence claim is immediately obvious. Instead it is the claim that, upon reflection, we should see that the causal theory has no non-trivial explanatory content.

I discuss the matter of explanatory value below.

3 The Illusion of Explanatory Value

Kripke’s causal theory is usually read as explaining or illuminating the phenomenon of semantic reference in some interesting way¹⁵ . We may well think that, while Kripke’s causal theory is not a version of reductive causalism, it claims more than mere coincidence between baptismal reference and downstream reference. It is generally thought to claim that causal chains are explanatory, i.e. that downstream reference obtains in virtue of a causally linked, baptismal reference.

I do not mean to deny that Kripke is generally interpreted as giving an explanation of why names have the referents that they do. My claim is that he is interpreted in this manner, yet the impression of explanatory value is an illusion.

Kripke’s explanations of his view goes as follows:

Someone, let’s say, a baby, is born; his parents call him by a certain name. They talk about him to their friends. Other people meet him. Through various sorts of talk the name is spread from link to link as if by a chain. A speaker who is on the far end of this chain, who has heard about, say Richard Feynman, in the market place or elsewhere, may be referring to Richard Feynman even though he can’t remember from whom he first heard of Feynman or from whom he ever heard of Feynman. He knows that Feynman is a famous physicist. A certain passage of communication reaching ultimately to the man himself does reach the speaker. He then is referring to Feynman even though he can’t identify him uniquely (Kripke, 1981: 91).

If we read passages like the above, and Naming and Necessity in general, as explaining how names refer to the individuals that they refer to, then, stripped to its bare essentials, the underlying view is as follows:

• 1. The user of a name stands in a causal chain which leads all the way back to a baptismal event.

• 2. The user of a name acquires the ability to refer in virtue of standing in such a causal chain, as.

• 3. The downstream use of a name refers to the person baptized at this baptismal event.

• 4. Hence we can see that causal chains are the mechanism of reference for names.

The above explanation mixes triviality with mystery. (1) is a mere truism. It is true of all words that the user of a word stand in a causal chain which leads back to the event where the word acquired its meaning, and in the case of a name such an event will typically be a baptism. Causal chains will obtain for all words subject to public standards; (1) reflects the fact that the user of a name could typically not have acquired the semantic beliefs that lead him to use the name if the name had not been causally learned. (2) is generally true, but we need to be careful about which ‘ability’ we are talking about. If it is merely claimed that the user needs to be in such a chain in order to acquire semantic beliefs, then this is mostly true, but only reflects a trivial matter of the epistemology of convention acquisition. If (2) is supposed to make more than the merely epistemic point, then (2) cries out for an explanation as to what is meant when it is claimed that causal chains confer this ability. Some authors develop claim (2) by saying that downstream users ‘inherit the reference’ of upstream users, but this is of little help. If they merely mean that downstream and upstream users use a given name with the same reference, this is merely the coincidence claim. If they mean that one needs to be in such a chain in order to acquire semantic beliefs, then this, again, is fairly trivial. If ‘inherit the reference’ is supposed to mean more, I confess I have no clear grasp of what that would be.

Claim (3) is generally true, but is merely the coincidence claim. It does not serve to explain why the name refers to the baptized individual. Given what has come before, the ‘hence’ in (4) is unjustified, as no explanation has been given. Also note that there is nothing in (1)–(3) that justifies the use of the phrase ‘mechanism of reference’ in (4), or that serves to give an indication of what the phrase means.

Kripke’s theory, as portrayed above, does not succeed in giving an answer to the question of why names have the referents that they do. It merely states the coincidence claim and a claim about how semantic beliefs are acquired, and then states that these claims constitute an explanation of the ‘mechanism of reference’. Hence my claim that, despite the general interpretation of the causal theory as an explanatory theory, it has no non-trivial explanatory content.

Below I respond to some anticipated objections to this view.

4 Potential Objections Considered

Objection 1

The causal theory cannot be trivial as it is essential to Kripkean doctrines like rigid designation and metaphysical necessity.

The claim that the causal theory is trivial is not the claim that all of Naming and Necessity is trivial. In fact, it is not even the claim that much of Naming and Necessity is trivial, for nothing concerning the doctrines of rigid designation, metaphysical necessity and the like logically depend on the causal theory. These doctrines, instead, rest on the implicit Millianism in Naming and Necessity.

I say implicit, as Kripke never explicitly commits to Millianism. Furthermore, remember that the causal theory itself cannot force any such commitment, for, as pointed out by Almog, it is compatible with both Millianism and alternative semantic theories. I follow Soames (2002) in viewing Millianism as the semantic theory most easily compatible with Kripke’s general views.

The adoption of Millianism immediately vindicates rigid designation. Kripke views possible worlds as stipulated, not found (Kripke, 1981: 44). Hence the stipulated content of a possible world is a function of the semantic content of the sentence(s) whereby the possible world is stipulated; we cannot speak about a possible world independently of the semantic content of the utterances that stipulate it. This implies that, if a possible world is stipulated by uttering a sentence containing a name N, then the counterfactual referent of N will be identified in virtue of the semantic content of N. Millianism holds that the semantic content of N will be its referent, and so the counterfactual reference of N will not vary across possible worlds.

Matters of metaphysical necessity (and the necessary a posteriori) are similarly independent of the causal theory of reference. Millianism about names implies that names will not be epistemically transparent to their users, i.e. that competent users of a name can use co-referring names without knowing that they co-refer¹⁶. This, coupled with the doctrine of the necessity of self-identity, implies that a claim like ‘Hesperus is Phosphorus’ can be both a posteriori and metaphysically necessary.

The above reasoning should make it clear that Kripkean doctrines like rigid designation and metaphysical necessity do not depends on his causal theory of reference. In fact, the argument would go through even if we did not ascribe an implicit Millianism to Kripke. The argument only depends on recognizing that the the mentioned doctrines crucially depend on matters of semantics, not matters concerning the theory reference.

Despite the close association of Naming and Necessity and the causal theory, nothing much of substance, whether it be doctrines like rigid designation or metaphysical necessity, is lost of we accept that the causal theory is trivial. Millianism does the real work in securing these doctrines and, as pointed out earlier, the causal theory is neither helpful, nor needed, in order to support Millianism. Similarly, Kripke’s criticisms of the traditional descriptivism of Frege and Russell retain their power independently of the arguments presented here.

Objection 2

Kripke’s causal theory cannot be trivial as it denies the traditional descriptivism of Russell and Frege.

My claim is that the modified coincidence claim no non-trivial explanatory content. This is consistent with believing that the modified coincidence claim refutes traditional descriptivism, as theories can clash with trivialities without rendering such trivialities interesting explanatory claims. Kripke, properly interpreted, pointed out that descriptivism, as he interprets it, clashes with the coincidence claim. This only shows that traditional descriptivism implies a falsehood; it does not turn the coincidence claim into an explanatory account of reference. Of course, if Kripke had also claimed that semantic reference can be reduced to a type of causal chain, such a claim, whatever we may think of its merits, would have explanatory content. But, as pointed out before, Kripke makes no such claim¹⁷.

Objection 3

Kripke never claims to put forward a theory, but explicitly states that he is merely proposing a ‘picture’.

Kripke claimed that a name N semantically refers to the object baptized N at the beginning of the causal chain from which the user of N inherited N. My claim is that this view is commonly, but falsely, credited with having some non-trivial explanatory content. While it is true that Kripke does state that he is proposing a ‘picture’ (1981: 94), nothing depends on whether we talk about this view as being a ‘theory’ or ‘picture’. As the locution ‘causal theory of reference’ has become standard, I will continue to talk of Kripke’s ‘theory’.

Objection 4

The argument portrayed Kripkean causal chains as merely stretching back to an act of baptism. Kripke, however, is concerned with causal chains that stretch back beyond that, to the baptized object itself.

There does exist some textual support for the view that Kripke is thinking of causal chains that stretch back beyond the act of baptism, and to the baptized thing itself¹⁸. On balance though, I do not think that these are the chains that Kripke has in mind. Kripke, as already pointed out, states that his use of the notion of a baptism presupposes the notion of reference (1981: 97). Kripke also never commits to the view that baptisms will include such causal relations between the baptizer and baptized. He is happy to treat cases like Neptune’ as cases where a successful baptism happens in virtue of descriptions (1981: 96) and goes even further when he writes that ‘[t]he case of baptism by ostension can perhaps be subsumed under the description concept also. Thus the primary applicability of the description theory is to cases of initial baptism’ (1981: 96). Kripke seems to think that his own positive view would be unaffected if he concedes the construal of baptisms to the descriptivists. Hence I take it that the causal chains he is concerned with are the ones stretching from the act of baptisms to downstream use, and not to the baptized object itself.

However, even if Kripkean chains are supposed to be those that terminate in the baptized object, this would not make the resulting view non-trivial. The causal chain from the baptized to the act of baptism is no less trivial than the causal chain from the act of baptism to downstream use. The existence of a causal link between the baptism and the baptized object merely reflects the fact that we typically baptize individuals when we learn of their existence, and we typically learn of their existence via some or other causal interaction with such individuals. The fact that we name people after learning of their existence via some causal interaction with them is not theoretically interesting, and suffices to account for the existence of such causal links.

Objection 5

Kripke’s theory is non-trivial as it explains how people can refer using a name despite having wildly false beliefs about the referent of the name in question.

The above objection credits Kripke with having shown that we can safely ignore the cognitive constraints imposed by the descriptivists. As such it is essentially a variant of the second objection, and fares no better. It is in the nature of public conventions that they do not depend on the beliefs of any specific party to the convention. If I have the mistaken belief that the UK has a convention of driving on the right, then it remains the case that the convention in question advises me to drive on the left. The same goes for naming: if I have the mistaken belief that ‘Krugman’ semantically refers to Lucas, this does nothing to upset the fact that the conventional, semantic referent of ‘Krugman’ is Krugman. This again, is no deep truth about naming, but simply a triviality on any view that portrays our naming conventions as communal.

Objection 6

The argument in defense of Almog presupposes that facts about conventions are analytically prior to facts about semantic reference.

It is true that the argument here presupposes that our conventions concerning names determine the semantic reference of names. It is, in fact, hard to see how one could doubt such a claim.

More importantly, note that the Kripkean could not object in such a way. Kripke has explicitly stated that “[t]he notion of what words can mean, in the language, is semantical: it is given by the conventions of our language” (1977: 263, my italics).

Also consider:

If a speaker has a designator in his idiolect, certain conventions of his idiolect (given various facts about the world) determine the referent in the idiolect: that I call the semantic referent of the designator’ (1977: 263, my italics).

The above quotations make it clear that Kripke does view semantic reference as being conventionally determined. Hence we cannot defend the view that his theory is non-trivial by portraying him as somehow inverting the relation between facts about conventions and facts about reference.

Note, however, that such priority only relates to matters of semantic reference. The arguments made here would be compatible with a view on which speaker’s reference is prior to both conventions and semantic reference, as Kripke (1977: 263) seems to suggest. This would also, for example, be the case on the view that semantic reference is conventionalized speaker’s reference.

5 Conclusion

I have argued that the causal theory of reference is a truism about the epistemology of convention acquisition, not a revelation about reference. If this is correct, then the question arises as to how the confusion on this issue came about. Here my remarks will necessarily be a somewhat speculative attempt at inference to the best explanation, but I think that the story below fits the facts tolerably well.

The impact of Kripke’s causal theory must be understood with reference to the theory it aimed to replace, i.e. the descriptivism of Russell and Frege. Russell (1910: 114) and Frege (1948: 210; 1956: 297) portray the semantic reference of names as ‘idiolectical’, i.e. as dependent on the descriptive condition that the individual utterer of a name associates with the name. Such a view, whatever its virtues may be, is profoundly contrary to common sense¹⁹. Nothing seems more obvious than that names are subject to public conventions, i.e. that there is a public, communal convention that links Biden to ‘Biden’. The communality of this convention implies that the content of this convention is independent of the beliefs that any particular speaker may have about the name ‘Biden’. In this way, those who use ‘Biden’ to (non-ironally) try to refer to Trump are simply wrong, and we correct such mistakes when we encounter them.

It has already been explained that Kripke’s causal theory is an artifact of the fact that names are governed by communal, causally acquired conventions. This means that the causal theory - i.e. the coincidence claim - is an artifact of the view denied by Russell and Frege, as they deny the common sense view that names are governed by communal conventions. In this way, then, Kripke’s proposal of the causal theory effectively amounted to a partial reintroduction of the common sense view that the conventions governing names are communal conventions. This reintroduction of the common sense view, however, occurred in an oblique way, under a banner (‘causal theory’) that made common sense sound much more exciting than it is²⁰.

The reintroduction of the common sense view under a different guise, coupled with the fact that it is oddly difficult to notice that Kripkean causalism affirms no more than a correlation between baptismal and downstream reference, makes it easy to make the mistake of thinking that Kripkean causalism is more than a truism. Relatedly, it is surprising, at first glance, to realize that one is connected by such chains to all the people whose name one knows. Furthermore, and although Kripke does not suggest reductive causalism, the very idea that the relation of reference may be explicable in terms of causal relations is philosophically exciting in that it offers the hope of naturalizing semantic reference. Lastly, the causal theory is presented in the context of a series of doctrines - rigid designation, metaphysical necessity - that are anything but trivial.

Given the above factors it is perhaps not surprising that Kripkean causalism was thought to be a revolutionary theory of semantic reference. When examined on its own terms, however, it becomes plain that it contains no positive, explanatory content concerning the nature of semantic reference. Instead of being a revelation about reference, the causal theory turns out to be a truism about the epistemology of convention acquisition.

To my mind, Almog’s charge of triviality is vindicated.