Chapter Three of Gareth Evans'
The Varieties of Reference
(Department of Philosophy, UC San Diego)
Copyright © Rick Grush 1999, 2000, 2001, 2002
Copyright holder grants permission to distribute, quote, and reproduce for any non-commercial purposes provided authorship and copyright are clearly indicated.
3.1 Going Beyond Russell: singular thoughts
- Section 3.2 Russellian sayings: the two strategies
- Section 3.3 Kripke: singular thought without discriminating knowledge?
- Section 3.4 The Photograph Model
Going beyond Russell: singular thoughts
Evans sketches a way that Russell's work has exerted influence on Evans' contemporaries: by regarding a distinction between types of natural language expressions as grounded in a distinction between ways of thinking about objects, namely i) thinking of them by way of description and ii) thinking of them by way of a direct, prototypically perceptual, link between thinker and object (for Russell this relationship was 'acquaintance'). This distinction of ways of thinking is motivated by adherence to a thesis Evans calls "Russell's Principle": thinking about an object requires knowing which object it is, and these two ways of singling out an object in thought are ways of conforming to this requirement.
Evans traces out some features of Russell's thought, features which have had a continuing influence on the topics Russell was concerned with. The first feature is a dependence of the functioning of types of natural language expression, especially singular terms, upon different ways one may think of objects. This involves a direct connection between the semantic analysis of linguistic expressions and the cognitive goings-on of those who use or understand those expressions — a connection not often explicitly recognized as the substantive assumption it is. The two types of thought recognized by Russell are i) descriptive, and ii) something more intimate, direct, and context dependent (perceptual contact being a prototypical case of this). The idea, then, is that given these two modes of thought about objects, it is possible to ground a distinction between two different kinds of referential relation between singular terms and the world; roughly, reference by description and demonstratives (or more generally some kind of direct reference based on a suitable 'acquaintance' relation). Evans claims that one of the major concerns of philosophy of language has been to try to find some principled ways to extend the range of application of the 'direct' type of reference beyond the case of simple perception-based demonstratives, which presumably would require extending the direct 'acquaintance' type of thought beyond the case of perception. Potential avenues of extension from perception involve memory and testimony, but this has not yet been done in any satisfactory way (Evans will be addressing both of these in later chapters).
Russell's motivation for restricting thought about objects to just these two varieties was what Evans will call Russell's Principle. It is this:
Russell's Principle: A necessary condition for S to be able to think about an object O (or to make a judgment about) it is that S know which object he is thinking (or attempting to think) about — he must be able to distinguish O from all other objects.
Russell took it that in order to know which object was in question, one would have to have a discriminating conception of that object capable of distinguishing it from all others, and he recognized two ways to do this — either one can pick out or locate the object in one's vicinity, or one knows some distinguishing feature(s) that only that object has. (A third possibility, recognized by Strawson, and to be taken up in greater detail in Chapter 8, is that the subject have a capacity to recognize the object.)
Evans cautions that much work in the area has been concerned not with a distinction between types of singular terms, but rather between types of belief — those reportable de re and those reportable only de dicto. The prevailing view has been that the distinction between these types of belief rests on a distinction between two kinds of a relation a thinker may have to the objects of thought: a conceptual (descriptive) relation, and a contextual (prototypically perceptual) relation. Evans' caution is that even if this distinction turns out to be a useful one for distinguishing types of belief, one should be careful about extending this distinction without argument to the case of singular terms.
3.2 Russellian sayings: the two strategies
Evans describes two ways in which one might argue that certain expressions (singular terms) are Russellian. First, one could stick to considerations about language, without resort to thought at all. Or, one could attempt to ground a distinction between Russellian and non-Russellian singular terms in a distinction between Russellian and non-Russellian thought. Employing the first strategy, one can show that proper names are Russellian. Employing the second strategy, one can show that demonstratives are Russellian.
Evans begins by abandoning the Russellian assumption that natural language is merely a mirror to human thought; specifically the notion that the meaning of an expression is no more than what the speaker intends to communicate by means of that expression and the further claim that the semantic analysis appropriate to a given linguistic expression is always mirrored by an equivalent or isomorphic analysis of the semantics of the thought involved. Rather, Evans points out that a linguistic system has an ineliminable social component in that there are norms which govern the correct use of expressions in the language, and thus supply those expressions with their standard meaning, regardless of what the individual language user intends to say by means of that expression or what thoughts or even kinds of thoughts are being entertained when the sentence is being uttered.
Among other things, this implies that there can be a difference between what a speaker intends to say by uttering some expression, and what that speaker strictly and literally says. For example, if I am under the mistaken impression that 'teal' is used for a shade of red, then I might intend to say to someone that a certain jacket is some shade of red by uttering the sentence 'I just bought a teal jacket'. Nonetheless, I would have strictly and literally said, according to the conventions of the language that I intend to be following, that the jacket is teal.
Evans goes on to argue that the speaker's intentions concerning what he wants to say play no necessary role in determining the content of what he actually says upon using some expression. That is, S's intent to say that P is not a necessary condition for S to say that P. His argument consists in defusing an argument to the contrary. This contrary argument runs as follows:
Consider the case where a speaker utters "The ship is veering to port." There are two things that this can mean, because of an ambiguity in the term 'port'. The speaker intended only one of these meanings, and it is the one intended that determines the meaning of the expression (we would be acting in a needlessly belligerent way if we maintained that the speaker's utterance meant that the ship was veering away from starboard even though he intended to say that the ship was headed toward a port). Therefore, on this line of thinking, a necessary condition for the speaker to say that P is that the speaker intends to say that P.
Evans argues that this does not follow. According to Evans, ambiguous terms are best thought of as two different expressions that happen to share a phonological (/typographical) form. Where the speaker's intentions come in is in determining which of the expressions — which of the tokens in the language game — is being employed. But once this is determined, the meaning of that expression is a matter of social practices, of the 'meaning with which that linguistic counter is endowed in the community', and not at all a matter of the speaker's intentions or possibly 'half-baked ideas and misconceptions'. To expand Evans' example, suppose that the speaker misunderstands 'port', thinking that 'port' means the side that is actually starboard. His intentions are enough to make it clear, let us suppose, that he means to be using not the 'port' which means a commercial harbor, but rather the 'port' which means a side of the ship. This is where his intentions matter, in determining which of the linguistic tokens he intends to use. But his intentions in this regard do not determine the meaning of what he says. He intended to say that the ship is veering starboard (we can suppose that he is on the ship and is in a position to know exactly which direction the ship is veering). But though he intends to say that p (the ship is veering to starboard), and thus selects for use the token of 'port' that concerns a ship's direction and not a harbor, nevertheless his selection is inadequate to his intentions, for what he actually says is not p but q: that the ship is veering to port.
That what an utterance means is not constituted by what a speaker intends is evident in situations other than when the speaker misapprehends the conventionally determined meaning of an utterance. Proper names are not typically associated with some common descriptive content (though of course each user of the name may associate this or that descriptive content with the name). The descriptions and other ancillary information that a single speaker associates with a name may have nothing to do with the conventional meaning of that name in a language community.
The possibility that follows from the above considerations is that an individual's thought and belief may be related to some object via particularities of the individual's cognitive situation (such as the entertaining of some description or other), but that when it comes to unpacking the content of what the speaker says about that object by means of expressions of a public language, this description (or other distinctive mark, such as perceptual contact) may be entirely irrelevant. Call this theory — according to which i) the individual's thought about the object she intends to refer to by using some expression E is determined by the speaker's cognitive situation (such as the entertaining of a description), but ii) the meaning of E in the language is not tied to any particular descriptive content — the composite theory.
Evans now turns to the matter of the Russellian status of expressions. A Russellian singular term is one such that atomic sentences employing it as the grammatical subject are meaningless when the term is empty — where 'meaningless' means 'nothing counts as understanding the expression', or 'nothing is said be means of the expression'. Therefore, on the composite theory, ordinary proper names are Russellian. The reasoning is as follows:
1. In order to understand a sentence 'a is F' (where 'a' is an ordinary proper name), one must think of the referent of 'a', in whatever way one is accustomed to think of it, and take the sentence to be true just in case the referent is F.
2. If a name has no referent, then nothing counts as understanding what is said by a use of the utterance (by (1)). (One cannot think of the referent if there is no referent, just as one cannot eat the pie if there is no pie. The latitude for idiosyncrasy in how the referent is thought about does not abolish the requirement that the thought be about the object.)
3. If nothing counts as understanding what is said by an utterance, then nothing is said by that utterance.
4. If an ordinary proper name has no referent, then nothing is said by someone using it in an expression; that is, it is Russellian.
Premise (1) involves a substantial departure from some of the traditional wisdom (the Strawson/Searle cluster-description theory) about proper names. This wisdom maintained that in order to understand a proper name, one had to associate the name with some sufficient sub-cluster of the cluster of descriptive content commonly associated, by the language community, with the name.
[Notice that in the case of descriptive names (which are not 'ordinary proper names' of the sort the above argument concerns) specifically because there is a reference-fixing descriptive stipulation associated with the name, the argument in 1-4 above does not work. In order to understand a descriptive name, one must understand the description with which it is associated. Hence, descriptive names are not Russellian: there is something that counts as understanding these names when they are empty: understanding the stipulated descriptive content. We can now see the overall framework within which Evans has been arguing that descriptive names are referring expressions, but not Russellian referring expressions.]
Premise (1) is also what will distinguish two different kinds of Russellian expressions: proper names and demonstratives. Proper names are such that (1) is true, and we can use the argument above to establish their Russellian status. But (1) depends on a certain view of the functioning of proper names which Evans has defended above, the view that they are not associated with any proprietary reference fixing descriptive content. Hence how one thinks of the referent is unimportant. What is important is that one think of the referent. This is the first of the two strategies mentioned in the section's title, which establishes the Russellian status of ordinary proper names given the composite theory of the functioning of proper names.
The second strategy will be to establish the Russellian status of a different category of singular term given a different account of the semantic functioning of those expressions. Ordinary demonstratives are prominent members of this other category. The ideas here are explained below:
First, Evans adopts the definition of a Russellian singular term in terms of saying (rather than thinking): a singular term is Russellian iff it is a member of a category of singular terms such that nothing is said by someone who utters a sentence containing such a term unless the term has a referent. Furthermore, to say that 'nothing is said' by a particular utterance is to say that nothing counts as understanding the utterance.
By extension, a Russellian thought will be a thought which simply cannot exist in the absence of the object(s) which it is (purportedly) about. Or to put it another way, the thought can only be successfully entertained if the object(s) which it purportedly concerns exist.
Given this, Evans outlines the second strategy for arguing that a class of singular terms is Russellian. The strategy has two parts. First, show that in order for utterances using the type of singular term in question to be understood, the audience must have a particular kind of thought. Second, show that the kind of thought that must be had is Russellian. Putting the two points together, we can see that if there are such expressions, nothing can count as understanding utterances employing them if they are empty, and hence nothing is said by means of such utterances. Hence these singular terms are Russellian.
Evans, in the last three paragraphs of this section, runs briefly through what will be his use of this strategy in several later chapters of the book to argue that demonstratives are Russellian singular terms.
At the heart of the employment of this strategy regarding demonstratives, Evans will argue that in order to understand an atomic sentence employing an ordinary demonstrative, the listener must grasp a particular kind of thought concerning the referent of the demonstrative expression, a thought that requires among other things an information link between the object and the listener. Thus: if no object, then no thought of the required kind; and if no thought of the required kind, then no possibility of understanding the sentence; hence the expression is Russellian.
3.3 Kripke: singular thought without discriminating knowledge?
In this section, Evans discusses Kripke's causal theory of names, and argues that initial appearances notwithstanding it does not directly present a challenge to Russell's Principle. This is because Russell's Principle concerns a subject's ability to think of an object, while the causal theory of names is a theory not about thought, but about the referential properties of words in a public language.
Kripke, in Naming and Necessity, criticizes the Strawson-Searle cluster description theory of proper names according to which a language user must associate with a name a cluster of descriptive content that is a sufficiently weighty subset of the descriptive content conventionally associated with the name in order for the language user to be able to understand the name. This cluster theory at least offers the promise of the language user's conforming to Russell's Principle because the sub-cluster of descriptive content might pick out the referent uniquely.
Kripke's criticism of this view has several aspects. The first, crucial, but subtle and inexplicit aspect is the focusing of the question not on a subject's capacity to think of an object, but on the subject's capacity to use a word in public language and thereby refer to an object. (The cluster-description theories tended to conflate these issues, and Kripke does not explicitly separate them as clearly as one might like.) The second aspect is a positive account of how it is that a word comes to be able to refer to an object (or better: how a language user comes to be able to refer to an object by using the word). The account is as follows: all objects that have proper names acquire them at some point in time via what Kripke calls an initial baptism. In the case of people this typically happens near birth, and is effected by the parents. The people that are present at this initial baptism are then able to use the name to refer to the object. Other people can be made competent users of the name by being introduced to the name by those who are part of the initial circle, and this passing of competence is a causal link, occurring as a result of some sort of causal interaction between someone competent with the name and someone not yet competent with it, often during an introduction ('This is Smith'). By continuing a chain of such transactions, more and more people can come to learn the name — the chain of referential competence potentially reaching far beyond the referent's spatial travels and temporal lifespan.
On this view, what makes it the case that S can use a name N to refer to R is not that S associates much or any or even correct descriptive content with the name, but rather that S is introduced to the use of the name by someone whose own competence is grounded in having been introduced to the name by someone .... (and here a causal history is traced, perhaps quite long and involving printed matter, books, testimony or reports of witnesses, to the initial baptism).
So far, Kripke's causal theory and Russell's Principle can both be maintained, with the surely plausible consequence being that one can use a singular term of a language to refer to some object without being able to think of that object. This consequence is not surprising to anyone who recognizes the distinction, which Evans has argued for, between what a subject can say, and what a subject is thinking, or intending to say.
Evans then goes on to discuss the claim that a subject might actually be able to think about these objects — the objects to which she can refer simply because of the right causal bloodline in her use of the name — and hold beliefs about them. He says that Kripke mentioned this possibility a few times, but that it was not essential to his purposes. The only evidence in its favor was that in some cases it would be natural to say that such a subject has a belief about the referent. For example, if a child in a history class hears the name 'Socrates' in the sentence 'Socrates was snub nosed' and repeats the sentence, it might in some circumstances seem natural to say that the child believes that Socrates was snub nosed.
Evans has two arguments against the position that a person who can use a name to refer to an object in absence of discriminating knowledge (as per Kripke's theory of reference) can actually think about the object.
The first relies on the generality constraint, which Evans will not introduce by name until section 4.3. In short, the objection is that in order for a subject S to be able to think of object a that it is F, the subject must be in a position to be able to entertain (though not necessarily judge to be true) the thoughts that a is G, a is H, and so forth for all predicates G, H, ..., not semantically anomalous with a. And he claims that this capacity is not possible for someone who is merely exploiting a link in a reference preserving causal chain. This argument will be expanded at various places later in the book.
The second response is the claim that we cannot without argument use 'untutored linguistic intuitions' to establish theoretical claims about thought, reference, belief, etc. The untutored linguistic intuitions in question are the ones which make it seem natural to say that such a person (e.g. the child in the above example) believes something about the referent of the name. (Evans will later, in sections 5.3 and 11.5, explain these untutored intuitions in a way which defuses their challenge to Russell's Principle.)
3.4 The photograph model
In this section, Evans discusses a theory of mental representation which purports to be a genuine challenge to Russell's Principle, and which was, in some sense, inspired by Kripke's causal theory of names. Evans calls this the photograph model. Evans argues that the photograph model is quite distinct from Kripke's theory, enough so to render suspect any argument in its favor from analogy to Kripke's theory.
Evans points out that (when it counted) Kripke himself was very careful to limit his conclusions to the referential properties of words, as opposed to the representational properties of thought and belief. Nonetheless, Kripke's work has inspired many theorists of mental representation to adapt the form of his theory, and apply it to the content of belief and thought.
Evans' first misgiving is that many philosophers have not carefully distinguished the referential properties of words and the capacity of thought to be about objects, and have thus assumed without argument that Kripke's theory has shown Russell's Principle to be false. That is, it has been assumed that Kripke has undermined Russell's Principle when in fact he has not.
Evans then, in preparation to stating his second misgiving, runs through a synopsis of Kripke's causal theory of names (which was recapped in the notes to the previous section, and so won't be repeated here). The attempted generalization tries to show that the representational properties of mental states, such as belief and thought, depend on their causal ancestry as well. Roughly the idea is that mental state M represents a's being F just in case a's being F was an appropriate causal antecedent of (that is, played a causal role in the tokening of) M. This is what Evans calls the photograph model.
Evans' second misgiving is that the sort of causation required in both cases is quite distinct. For in the causal theory of names, the 'causal links' are uses of a name, and in the photograph model the 'causal links' are links are of a quite different nature. The way causality operates in both cases is different, and as Evans points out, when both are in operation they might give different results. (Chapter 11 will have examples of this sort of thing.)
It must be admitted that Evans' argument here is a bit lacking. He has not shown that the photograph model, apart from any mileage it gets from it association with Kripke's theory, is unable to challenge Russell's Principle. Much of this argument will come in later chapters. Two things to note about this: first, Evans did plan to expand this chapter, and specifically to continue his discussion of the photograph model. McDowell attempts to fill out this projected material in the appendix. Second, arguably Evans' argument works, insofar as the only reason to think that the photograph model undercuts Russell's Principle derives from its similarity to the causal theory of names. If this is so, then showing how this similarity is merely superficial, and too thin to support the conclusions, is tantamount to showing that there are no considerations in favor of the photograph model, and hence no viable challenge to Russell's Principle.
The appendix is part regurgitation of some of the things that have already been covered in the chapter, and part argument against the claim that the photograph model poses a serious threat to Russell's Principle. The argument is, in effect, that the photograph model can make no room for an analog of sense, and since any adequate account of representation must make such room, the photograph model is not an adequate account of mental representation, and hence cannot pose a challenge to Russell's Principle. The appendix closes with some remarks about the general strategy of the book.
Evans [I am here continuing the practice of referring to the author as 'Evans', though of course in the appendices, more so than the main text, McDowell is behind the pen] opens by recapping the two strategies for arguing for the Russellian status of singular terms. [Note that he discusses the two strategies here in the opposite order of that in which they were discussed in section 3.2.] Recall, the first sort of strategy establishes that a kind of expression is Russellian by showing (i) that expressions of that kind require, in order to be understood and hence to be meaningful, that the audience have a certain kind of thought; and (ii) that the kind of thought in question is Russellian. Thus, no object, no thought; no thought, no possibility of understanding; no possibility of understanding, no meaning. Russell took this as the model for mental demonstratives, and Evans will adopt this strategy to argue that ordinary demonstrative expressions are Russellian.
The second sort of strategy does not involve considerations of thought about objects, but rather tries to show that in absence of a referent, the mechanisms that endow a kind of expression with a meaning are not operative. Evans thinks that ordinary proper names are Russellian, and will employ this strategy in his arguments for that conclusion.
He goes on to recap the point that Kripke's work has been commandeered in an effort to establish a theory of mental representation, the photograph model, which would, if viable, flout Russell's Principle. The photograph model aims to extend the class of Russellian singular thoughts beyond the cases Russell recognized (sense data) — this is because if a thought's content is determined by the object that causes it, then in absence of a causing object there will be no content, hence the photograph model is Russellian. But it extends the reach of Russellian thought by failing to respect Russell's Principle, thus relaxing the constraint that the subject know which object is in question. The text at lines 15-16 of page 81 says that the Photograph Model holds out the promise of expanding the class of Russellian singular terms. One may be puzzled by this, since the Photograph Model is concerned not with language, but mental representation. But the best way to interpret this is that the Photograph Model could play a role in the expansion of the class of Russellian names, via an argument of the first type: in order to understand an utterance using an expression of semantic category S, one must have a thought of kind P, a thought underwritten by the Photograph Model. Since such thoughts would be Russellian, then so would the expressions piggy-backing on them.
Starting at mid-p.81 and running to the bottom of p. 84 is the argument to the effect that the Photograph Model is not an adequate model of mental representation, and it can be broken up into three parts.
The first part of this argument (from mid-81 to near the bottom of 82) is just a reminder to the effect that, Russell's opinion notwithstanding, one can recognize the possibility of Russellian terms which also have a sense. There can be two Russellian terms, with the same referent, which are nonetheless not synonymous because they differ in sense. The belief to the contrary arises from thinking of (monadic) Russellian thoughts along the lines of the 'ordered-couple' conception of thoughts, the idea that the object literally is a component of the thought. Of course, this is one way a thought could be Russellian, but not (Evans argues) the only way.
The second part of this argument (which is the paragraph bridging pp. 82 and 83) is an argument to the effect that the photograph model is in its essentials an ordered couple model, at least to the extent that it makes no room for a notion of sense. Evans gets there by means of the following circuitous route:
1) if one tries to extend the class of Russellian singular thoughts while observing Russell's Principle, then one has the tools for a theory of sense. This is because the extension will take the form of exposing the non-descriptive means of identification the subject uses to identify the object in thought — that is, the subject's way of knowing which object is thought about. But these very means of identification will play a role in the content of the particular thought. In effect, identifying the same object in two different ways amounts to different senses for the same particular thought.
2) Because the Photograph Model flouts Russell's Principle, claiming that no means of identification of the object is necessary, it does not have the tools to construct a notion of sense.
[Both (1) and (2) rely, in the text, on the assumption that modes of identification are open to awareness, while causal ancestry of mental states is not, and that being open to awareness is what allows something the play a role in the content of a thought. But it is not clear that the argument requires this, and it is not clear that a reference to what is available in awareness is a good idea anyway. All the argument needs is the plausible claim that the photograph model cannot supply a difference is content between two mental representations if those representations are caused by the same object. This is plausible since the content just is the object which is the cause, and this will be the same in both cases, and hence the content will be the same in both cases.]
The third (final) part (the bottom third of p. 83 to the very bottom of p. 84) argues that no notion of Russellian thought which lacks an analog of sense can be adequate, yet alone coherent. Consider as an example a man who can see a long ship from two windows in his room. Though it is a single ship the man sees, he takes it to be two different ships. It might easily be the case that the man is prepared to believe 'that ship was built in Japan,' (looking at the ship through the first window) but believe 'that ship was not built in Japan' (looking at the ship from the second window). According to the Photograph Model, each mental tokening of 'that ship' has the same content, because it has the same causal antecedent (both are caused by the same ship). The subject will then have two directly contradictory beliefs. And we can presume cases in which both beliefs are concurrently entertained: "That ship was built in Japan, but that one was clearly not." The photograph model thus requires us, or so it seems, to be prepared to suppose normal subjects to be quite irrational in some very ordinary circumstances. And with that, the argument against the Photograph Model comes to a close.
The final paragraph on p. 84 is quite interesting. Evans explains that two common, and (Evans thinks) mistaken, ideas go together. The ordered-couple conception of Russellian thought, and the idea that Fregean senses do not require a referent. This way of looking at things takes it that if there must be a referent, then there is no room for sense (because 'must be a referent' is taken to entail 'ordered couple conception'). Conversely, if there is room for sense, this can only be because the object is not itself a part of the thought, and hence it cannot be required for the thought.
The final page is straight-forward, but important enough to merit further mention. Evans points out the importance that adherence to Russell's principle will have in the account of Russellian reference in the remainder of the book. The closing paragraphs constitute a useful orienting map of the territory to be traversed in this account. But since all of the issues Evans mentions here will be expanded in later chapters, I will not comment on them separately here.
[End of Guide to Chapter Three]
Copyright © Rick Grush 1999, 2000, 2001, 2002
Copyright holder grants permission to distribute, quote, and reproduce for any non-commercial purposes provided authorship and copyright are clearly indicated.
[Material here based on previous material Copyright © Rick Grush and Pete Mandik 1997, 1998.]