1 Evans on the Fundamental Level of Thought

In fewer than 10 years, Gareth Evans produced an outstanding body of philosophical work, occupying around 800 printed pages. Brilliant, original, and wide-ranging, it opens up new ways of thinking about many central philosophical issues, and has justly earned him a place in the canon. The earliest publication, “The causal theory of names” (1985a) is still cited, more than 50 years on, in most discussions of proper names (e.g. Dickie (2015), Goodman (2021); Rami (2022)). His other papers are still rightly being mined for insights on a wide range of topics: pronouns, modality, vagueness, tense, semantic structure. His posthumous book, The Varieties of Reference (1982), is also much discussed, and ranges over many topics relating to reference and thought, including several forms of identification, special features of thinking about oneself, and the semantics of existential statements.¹ One theme of the book may strike many readers as central: Evans’s conception of “the fundamental level of thought”. I shall argue that this conception is flawed, though turning one’s back on it has remarkably few consequences for the rest of the work.

In The Varieties of Reference, Evans suggests that thought at the “fundamental level” occurs only when the thinker satisfies “Russell’s Principle” (RP), initially characterized as the principle that thinking of an object is possible only if the thinker “knows which object his thought is about” (1982, 89). The principle is subsequently refined, in ways I shall shortly describe. The upshot is that, for Evans, thought about objects is not easy to achieve.

By contrast, Evans holds that acquiring a proper name is easily achieved: one may do so “just by hearing sentences in which the name is used” (1982, 377). Having acquired a proper name, one can use it to refer to the object, if any, which is its bearer. And it’s natural to think that if one can refer to an object, one can think about it. Evans resists this last inference. He believes that one can use a proper name to refer to an object that one cannot think about. That’s because one can use a proper name to refer to something with respect to which one does not satisfy RP.

The main claim of the present paper is that this position is unacceptable. If you can refer to something you can think about it. So if it’s easy to get into a position to refer to something, for example by acquiring a proper name for it, it’s easy to think about it, and the “fundamental level of thought” falls away as misplaced.

2 Adequate and Fundamental Ideas

Evans gives content to RP by refining “knowing which”: knowing which object one’s thought is about is being able to “distinguish the object of his judgment from all other things” (1982, 90). Analyzing further: one conforms to RP iff one has an adequate or fundamental Idea of the relevant object. To think about an object using a fundamental Idea of it is to think of it “as the possessor of the fundamental ground of difference which it in fact possesses” (107). Using an adequate but non-fundamental Idea involves “knowledge of what it is for an identity proposition involving a fundamental identification to be true” (149), where this is an identity proposition one of whose terms expresses a fundamental Idea. All fundamental Ideas are adequate, but there are also adequate non-fundamental Ideas, deriving their adequacy from a connection to fundamental Ideas through identities.

The connection is via an identity proposition whose form he writes using the schema ⌜δ = a⌝. Delta marks a position held by an expression for a fundamental Idea, and “a” a position held by an expression for an adequate Idea. It is not that, in order to possess the adequate Idea a, one needs to know the truth of any such identity; one need only know what it would be for such an identity to be true.

Concerning knowing what it would be for some proposition to be true, Evans says “I am quite unable to give a general account of this notion” (1982, 106). He allows that there is a verificationist account of it, but he cannot appeal to it since he is not a verificationist. However, we can explore how the appeal to an identity works out in practice. We first need to say what a fundamental Idea is, and then see how a non-fundamental Idea can count as adequate.

A fundamental Idea of an object identifies it by its fundamental ground of difference. Evans’s examples:

• The number 3 is discriminated by the property of “being the third number in the series of numbers” (1982, 107).

• A statue is discriminated at a time by “(i) the position that it occupies at that time and (ii) the fact that it is a statue” (107).

In the examples, the fundamental Idea is introduced as a predicate. This is not surprising: we discriminate objects by discriminating their properties, as codified in Leibniz’s Law. But adequate Ideas cannot be predicative, since, as we saw, they can occupy singular term position (or its analog in thought), flanking a sign for identity. Moreover, in order to apply it as a predicate, we would first need to refer to something to predicate it of, whereas Ideas are clearly intended to do the basic referential work, not requiring some prior and more basic referential device.

The solution is that a fundamental Idea is the analog in thought of a complex definite description in language, and so both contains predicative material and is fit to occupy subject position. This passage makes the point explicit:

a demonstrative identification need not itself constitute a fundamental Idea. It will be adequate, without being fundamental, so long as the subject knows what makes an identity proposition of the form ⌜This = the G at π, t⌝ true. (1982, 178)

The Idea corresponding to the demonstrative “this” is held to be adequate thanks to our understanding its use in the identity whose other term is the fundamental Idea expressed by “the G at π, t”. “G” marks the position for the mental analog of a discriminating predicate, and, at least for material things like statues, “π” and “t” mark positions for expressions referring to a place and a time, key elements in the fundamental ground of difference for material things.

I see a statue and mentally refer to it as this. It seems that on Evans’s account the Idea I express by “this” is adequate if I know what it would be for some identity like this is the statue in my garden now to be true (to pick an identity at random, using the structure specified at Evans (1982, 107)), where the definite description introduces a fundamental Idea. If the ‘this’-Idea is adequate, then I can use it to think of the statue in a way that satisfies Russell’s Principle.

The account has several problematic features.

1. Many people think that a demonstrative Idea, for example an Idea expressed by a token of the bare demonstrative “this”, brings one into more “direct” contact with an object than does an Idea expressed by a definite description.² Yet the form said to express a fundamental Idea, as given in the passage from Evans just discussed, is that of a definite description. Evans himself, for example, says that using “descriptive names”, names whose exercise requires one to appreciate that the bearer is whatever satisfies the associated canonical description, is not a way to satisfy RP. It is therefore surprising to find that the general form of fundamental Ideas, which serve as the basis of RP-satisfying thought, is descriptive.

2. Many people suppose that “singular” or “direct” thinking about objects requires using expressions lacking semantic complexity, whereas on Evans’s account the most fundamental referential work is done by semantically complex thought.

3. Knowing “what makes an identity proposition [involving a fundamental Idea] … true” requires knowing that the identity is true, which is implausibly demanding: I can think about a person, for example the author of a book I am reading, without knowing, for any place and time, that they occupied it then, even though, assuming people are material objects, an adequate Idea would involve such information. Perhaps it would be appropriate to shift to a later formulation: the condition is knowing what it would be for the proposition to be true, something which one can know concerning a falsehood.³ But a mystery remains. If one is to know what it would be for the identity to be true one would need to understand the adequate but non-fundamental Idea expressed by “this”, so the condition does not express knowledge available prior to possession of the relevant “this”-Idea. If the Idea was adequate before the knowledge was acquired, then the knowledge does not explain the adequacy. But if acquiring the knowledge transforms an inadequate Idea into an adequate one, it would be helpful to learn more about how this can happen.

The crucial feature of an adequate Idea is that it enables its possessor to discriminate an object. This requires an enabling source that does not presuppose that the subject has already discriminated the object. Evans mentions three sources of discrimination: perception, recognition, and knowledge of distinguishing facts (1982, 89). The last two do not qualify as enabling sources: knowledge of a distinguishing fact cannot take the form of knowledge that there is a unique so-and-so, since one could know this without, intuitively, knowing which object satisfies the description.⁴ We need also to know which thing is uniquely so-and-so, and that involves picking out the relevant thing. The “distinguishing fact” presupposes rather than enables this activity. Similarly, as Evans stresses, recognition is “re-identification as something previously encountered” (1982, 279), so it too presupposes a prior identification.

By contrast, perception supplies discrimination in a way that enables the cognitive system to single out an object. Consider Tye’s example of a moth (2010). At first you see just a tree, with an apparently consistent patterning of bark; the moth is camouflaged. Then something changes: the light shifts, or the moth on the tree moves, and then the moth stands out, clearly distinguished from the bark, and available for cognitive discrimination. Whatever one thinks about the border between the perceptual and the cognitive, or about the possibility of cognitive penetration, there is something undeniable about the example of the moth: perception delivers discrimination. No wonder perception has been for so long, and conspicuously since Russell, a paradigm source of cognitive discrimination.

For Evans, the cognitive activity required to make use of what perception offers does not come easily. One needs adequate Ideas, and so connections with the fundamental ground of difference. Evans is at pains to reject the sortal subsumption thesis (to the effect that thinking of an object requires bringing it under its “sortal”, conceived in a fashion going back to Aristotle and more recently espoused by David Wiggins (1980)) on the grounds that it is too demanding: one can think of an object while not bringing it under its sortal. A fundamental ground of difference will entail a sortal, but an adequate Idea may be only indirectly connected to a fundamental ground of difference via the identities we have discussed. Even so, possession of an adequate Idea is cognitively taxing. This fact returns us to the main theme: learning a proper name is not very taxing, and does not require the kind of connection to fundamental grounds of difference that Evans envisages.⁵ Yet a natural view is that proper names are a major source of singular thought about objects: thought that meets any reasonable demand on what is involved in thinking of one specific object, as distinguished from all others. To explore this tension further requires considering Evans’s views about proper names.

3 Proper Names and Pronouns

Uses of proper names are typically not one-off events, but occur within name-using practices involving several users of the same name. Such a practice is held together by deferential intentions: in each use (apart from the name’s originating use), the user aims to use the name in the way other members of the practice do.⁶ Joining a name-using practice is easy. You encounter one or more exercises of the name in a single practice, and utter the name intending to refer to what the others refer to by using it. Initiating a name-using practice is also not difficult. Certain conventions, for example those embedding religious or nautical baptismal ceremonies, facilitate name-introduction, but these are not necessary. Some names arise in the course of careful discussions of what to call the baby, even before it exists. Others, like nicknames, are just used of a person or other object for the first time, but then give rise to a convention of using that name for that referent.

Evans agrees that, having acquired a name, one can use it to say things about its bearer (if any); further, one can use it to refer to its bearer.⁷ But, according to Evans, it’s another question whether one can use it to think the thought that the name-involving sentence conventionally expresses; and another question whether one counts as genuinely understanding a name. For Evans, thinking a thought conventionally expressed by a sentence containing a proper name requires having an adequate Idea of the bearer of the name, whereas using the name merely to refer to the object does not. Hence Evans’s notion of the fundamental level of thought plays a crucial role even in the case of proper names. He agrees that it is not hard to join a name-using practice, but doing so does not require fundamental thought. Hence one may use a name, as a member of a name-using practice, without being in a position to think about its bearer.

Evans describes a view he rejects as follows:

mastery of the use of a proper name [is] an autonomous way of satisfying the requirement that one have discriminating knowledge of the objects of one’s thoughts. … any adequate introduction to a name-using practice equips one with an adequate Idea of the object which is its referent. (1982, 403)

If we replace “equips one with an adequate Idea of the object which is its referent” by “enables one to refer to and think about the referent”, this is similar to the modest picture I am proposing. For, according to this picture, using a proper name does not require an “adequate Idea”, in Evans’s demanding sense: Evans’s notion of a fundamental level of thought has no part to play in describing how proper names are used in language or thought.

Suppose you are lecturing to novices about David Hume. You might begin the lecture by saying “David Hume was an eighteenth century empiricist philosopher”. The students who are paying attention thereby acquire mastery of the name “David Hume”, and it is natural to suppose that they can use it to express thoughts. But Evans must say that some, perhaps all, students might be able to use the name to refer to Hume without being able to use it to think of him. Intuitively, there’s simply no room for this distinction.

How might one master the use of a proper name yet be unable to use it to think of its bearer? In “The causal theory of names” Evans warns against trusting initial intuitions. He offers an alleged example of easy acquisition:

A group of people are having a conversation in a pub, about a certain Louis of whom S has never heard before. S becomes interested and asks: ‘What did Louis do then?’. There seems to be no question but that S denotes a particular man and asks about him. (1985a, 6)

Given that it’s natural to think that if one can ask about someone, one can think about that person, this looks like an endorsement of the modest picture. However, on the next page, Evans warns that we may wish to retract our initial judgment when we consider how the story might develop. Consider the hypothesis that S denotes Louis XIII by his uses of “Louis”:

notice how little point there is in saying that he denotes one French king rather than any other, or any other person named by the name. There is now nothing the speaker is prepared to say or do which relates him differentially to that one king. (1985a, 7)

But, assuming S learned the name from speakers who used it to refer to Louis XIII, S also does so, for S uses the name in the same way as those from whom he acquired it. Whether he knows it or not, S is differentially related to Louis XIII. It’s another question whether S knows that Louis is a French king, or knows that his sources supposed he was a French king, or knows that he is differentially related to a French king. That question relates not to S’s understanding of “Louis” but to what he knows about his understanding, a distinction Evans is not always careful to make.

S denotes whoever his original conversationalists denoted (or denotes nothing if they denoted nothing), for that’s what he was intending to do, and nothing has prevented the success of that intention. It’s consistent with the subject not realizing that, in using “Louis”, he was talking about a French king. This is a perfectly ordinary and familiar situation. The detective seeking the murderer learns that the murdered woman’s husband is Harry. If Harry murdered her, the detective cannot come to know that Harry is the murderer just by thinking about Harry. Likewise, the mere fact that someone learns “Louis” from people who use it to refer to Louis XIII does not entail that the learner know that “Louis” refers to Louis XIII.⁸

Evans stresses that being able to use a word to say something does not guarantee that one can think a thought whose expression involves that word. This is correct: you may be told, by a bilingual speaker of a language you yourself do not understand, on what kind of occasion it is appropriate to use a certain sentence. When you do so, arguably you count as saying whatever the sentence says, even though you do not know what the sentence says, and so did not and perhaps could not think the thought it expresses. (Evans calls such cases “mouthpiece” examples (1985a, 7).) But the modest picture should not dispute that there are such cases. It should only deny that learning a name in the usual way is a mouthpiece case, one in which a speaker can use a name as a member of a name-using practice without understanding it.

The modest picture should insist that in intentionally using a newly acquired name to refer the neophyte will typically think about the referent. That is what makes the act of using the name to refer to the referent intentional. Intending something with respect to an object is simply a way of thinking about it.

According to Evans, a theorist who sees no significant distinction between referring to and thinking about is one who holds that an information-link is sufficient to bring an object before the mind, which in turn is an application of the Photograph Model (1982, 150n15). He suggests that such a theorist might be resisted by appealing to certain uses of demonstratives. Evans gives the example of someone who is blindfolded and points in a specific direction when there are several people in the vicinity, saying “That person is F”. Intuitively, they do not know at whom they are pointing. Evans suggests that it is consistent to hold that the speaker might fail to think about anyone in particular, even if their words count as referring to the person at whom they were pointing. This speaker referred to the person pointed at without being in a position to think about that person (1982, 171).

Evans suggests that those unconvinced by this line of reasoning are in the grip of the Photograph Model, according to which a thinker engages with an object of thought in the way a photograph engages with the scene that it is of. But an example Evans uses against the Photograph Model seems to me to lend it unintended support. The subject saw both of two indistinguishable steel balls, but now remembers only one of them. The Photograph Model agrees with common sense that the subject’s ball-thoughts are about the ball he now remembers, even though he can attribute to it no property (other than being remembered) that distinguishes it from the now forgotten ball. Evans makes a revealing parenthetical comment: “I assume, as I think I am entitled to, that he would not think of distinguishing the ball he is thinking of as the one from which his current memory derives” (1982, 90). This shows that Evans is requiring not just a fact uniquely relating a thinker to an object (e.g. that this ball is the only one he now remembers) but a fact of this kind that is known to, and actively appreciated by, the thinker. I see no justification for this further demand, a demand the Photograph Model avoids. Likewise, it’s enough for the one who points (in the blindfold case) to have pointed at someone for them to be able to think about that person: did I point at Paul? (That is: is Paul the person at whom I pointed?) If so, how will he respond?

All adequate ideas have referents: adequacy is a matter of how their referent is determined. Hence the fundamental level of thought cannot make room for non-referring singular expressions, of which our thought and language have many kinds.⁹ These include the familiar kinds of empty name: fictional, like “Holmes”, error-driven, like “Vulcan” and “phlogiston”, deceptive like “Major William Martin” (the man who never was) and proleptic. The last category may be illustrated by cases like the following: a couple is planning a family, though the woman is currently not pregnant. She says “If it’s a girl, let’s call her Jane”. Subsequently “Jane” is widely used in their conversations: “Today let’s paint Jane’s room”, and so on. But the hoped-for pregnancy never occurs; Jane is the baby the couple wanted but never had, and “Jane” is a name without a bearer. The parents moved from the belief that the name had a bearer to the belief that it did not, but this change of belief did not make a change in what the name meant or referred to. It never referred to any existing thing.

In ordinary speech we would say that the parents thought about Jane when they were making their plans, even if the pregnancy never materialized. That is, we ordinarily use “thinks about” as an intensional verb: one can think about unicorns, about Pegasus, about the house one never built, and so on. Clearly this usage is not what Evans has in mind when he wishes to link thinking about with RP. No doubt philosophical theory should not cravenly follow idiom, but a stipulated extensional reading of “thinks about” is not well suited to describing our actual non-extensional use of this concept. Evans himself must recognize one aspect of the intensionality of “thinks about”, for he allows that one can think about Hesperus without thinking about Phosphorus. He should also allow that one can refer to Phosphorus (as Hesperus) without knowing that one is referring to Phosphorus. “The speaker referred to X” does not entail “The speaker knew they referred to X”; conditions for reference should not be confused with conditions for knowledge of reference.

Pronouns are grammatically singular terms that are very easy to introduce. The use of “her” in the example of Jane shows this. Indefinite noun phrases can, without appealing to other resources, give rise to perfectly intelligible dependent definite pronouns. These are not to be likened to variables, since whereas quantifier scope, as usually understood, cannot cross sentence boundaries, a definite pronoun can be supported by an indefinite in another sentence or in the mouth of another speaker. A says “I’ve invited a lady friend to dinner” and B responds “Is she a lawyer?”. B’s use of “she” refers to whoever (if anyone) A had in mind in speaking of a friend, and appreciating this is enough for understanding the exchange. As B, or as an onlooker, one needs nothing else in order to understand the pronoun.¹⁰

Examples of this kind, in which a definite pronoun depends on an indefinite antecedent, were central to the development of Discourse Representation Theory (DRT), in particular by Hans Kamp.¹¹ A central notion of DRT is that of a discourse referent: a singular mental representation, which may appropriately be introduced in interpreting certain indefinite noun phrases as well as definite ones. It is in effect an individual idiosyncratic mental name. One who hears “I’ve invited a friend to dinner” needs to introduce a discourse referent to refer to the friend, in case such a representation is required as the conversation is continued. The speaker might add “She’s a philosopher”, and the hearer needs to be ready to interpret the singular “she”. In DRT, this is possible if the hearer responded to the earlier “a friend” by introducing a discourse referent, which can now be linked by identity in interpreting “she”. Questioned about their interpretation, the hearer might respond: “Of course I understood who you meant by”she”: the friend you’ve invited to dinner”.

Discourse referents are postulated by DRT in empirical hypotheses about psychological states; in particular, they are hypotheses about thoughts. DRT was gaining momentum only towards the end of Evans’s life, and I am not aware whether, and if so how, he reacted to it. One could hardly imagine a conception of reference further removed from Evans’s own, for nothing could be more different from an adequate Idea than a discourse referent introduced in response to an indefinite noun phrase. Yet, according to DRT, such mental elements can in suitable circumstances enable the subject to think directly about objects.

Returning to the main theme of whether Evans’s account of proper names is consistent with his conception of adequate Ideas: the threat is that, whereas using an adequate Idea to think about an individual object requires some effort, using a proper name to think about an individual object seems easy. Evans addresses this very directly in the final pages of The Varieties of Reference.¹² He suggests that what is easy to acquire is the use of a proper name to refer, but that one who does so may fail to “properly understand utterances involving the name (including their own)” (Evans 1982, 400). In that case, the distinction between saying or referring on the one hand, and thinking on the other, does important work in rendering the overall position consistent. What’s easy about proper name use is the ease of using a name to refer; that's consistent with it being much more demanding to use a name to think of individual objects; thinking requires the exercise of adequate Ideas, which is not assured by being able to refer.

Evans suggests that opponents fail to do justice to the distinction between merely joining a name-using practice and being in a position to aim one’s thought at an object:

The notion of thinking of an object, and the notion of the object of thought, when proper names are involved, is confused, because of a failure to distinguish two notions of the intended referent of the use of a name: one in which the intended referent is determined by which name-using practice a speaker manifested the intention of participating in (the intended referent is the referent of the name as used in that practice); and one in which the intended referent is the object which the speaker is aiming at with his use of the name. Full understanding of a use of a name requires that the referent of the name be an object of the subject’s thought in the second sense. (Evans 1982, 402)

It is hard to regard “aiming at” as more demanding than “referring to”: one might miss a target aimed at, whereas one cannot miss referring to that to which one has in fact referred. “Aiming at” seems less demanding and less intimate than referring to, though Evans was trying to show just the opposite. To sustain the contrast between referring and thinking, Evans would need to say that your students in the Hume course might learn to refer to Hume without learning to think any thoughts about him. How difficult it would be to tell the essays apart, distinguishing those who expressed thoughts about Hume from those who merely referred to him without thinking about him. In my view, one could not make the distinction because there is none to be made. When one uses a proper name to refer and succeeds in referring, one thereby thinks about the bearer.

4 Name-Using Practices; Evans Against Kripke

One feature of Kripke’s account of names in Naming and Necessity (1980) is its appeal to name-using practices: groups of users of a name whose name-uses share a causal origin. It is the ease with which such a practice can be joined that explains the ease with which a proper name can be learned. The critical mental state required of new members of a practice is deference: in coming to use the name, they defer to the way in which it is used by others. This basic attitude is hard to describe in detail (obviously the neophyte does not have to believe everything they hear in which the new name is used); but it is an attitude that plays a central role in the dispersion of a name-using practice.

Evans saw Kripke as threatening his position, by advancing an account of names and name-using practices not requiring any heavy-weight theses like RP. Evans reacts thus:

It must surely be agreed that, had Kripke genuinely demonstrated Russell’s Principle to be false, he would have shown something of even greater importance than the unacceptability of the Description Theory of Names, or than any of the many other fascinating conclusions contained in the lectures. But I think it fair to say that Kripke did not refute Russell’s Principle. (Evans 1982, 74)

One of Kripke’s most telling examples is intended to suggest that someone could refer to Feynman by name while possessing no uniquely identifying information about him, being able to offer only something like “He’s a famous physicist”, knowing full well that there are many famous physicists (1980, 81). On Kripke’s view, as Evans interprets it, the subject in the example says something about Feynman, and refers to him, without satisfying RP. Evans responds by pointing to the distinction between saying and thinking. Kripke’s subject may well be able to say something using the name “Feynman”, but cannot use it in thought, thanks to his lack of discriminating knowledge of Feynman. RP applies only to thought, and not to saying, so, according to Evans, Kripke’s example does not even address RP.

Suppose Kripke’s example were elaborated with an Evans-style opponent in mind. The opponent must say that one using “Feynman” to think about Feynman must have an adequate Idea of Feynman, which is lacked by the speaker Kripke envisages. This will be either a fundamental Idea, or, failing that, an understanding of what it would be for an identity relating a non-fundamental Idea to a fundamental one to be true. Suppose the example is of someone who, as Kripke’s story requires, does not know any individuating information concerning Feynman. Then they cannot have a fundamental Idea of him. Might they have an adequate but non-fundamental Idea? If they begin by asserting “Feynman is a famous physicist”, could they know what it would be for something of the form ⌜Feynman = the G at π, t⌝ to be true? By one standard, it seems easy. Let the form on the right of the identity be realized by “the person standing in front of me right now” (“G” is replaced by “the person”, “π” by “standing in front of me” and “t” by “right now”). Our speaker is not required to know that “Feynman is the person standing in front of me right now” is true (we can assume it is not, and that the speaker knows this), but only to know what it would be for it to be true. Surely he can know this. He could continue his remarks, so that they amount in full to: “Feynman is a famous physicist. If he were the person standing in front of me right now I would ask him if he’d mind if I attended some of his lectures”.

Taken in this way, Kripke’s example fails to be a counterexample to RP, since the subject is presented as satisfying the principle. But the example, as extended, could be used to make a different point: on a certain understanding, RP is at most a very modest restriction. It is met even in cases in which Evans himself took it not to be met, cases to which he responded by saying that the subject was not able to use the name to think about its bearer. Whereas I’ve mostly been regarding RP as making very heavy demands on a thinker, the suggestion just made suggests it is a trivially satisfied requirement.

Confronted with this development of Kripke’s example, Evans has two options. One is to deny that one can so easily and trivially meet the demands of the fundamental level of thought. This is the better option for his overall dialectic, and is the one I shall consider him as having taken; but I am at a loss to construct, on his behalf, a plausible argument for it. The other is to admit that the speaker can indeed think about Feynman, and not merely refer to him. This resolves the tension currently under discussion by making the fundamental level of thought as easy to attain as is the use of a proper name to refer, but it is clearly not an option Evans would be happy with.

An impressive theory enforcing a distinction between saying and thinking was presented by Russell in Problems of Philosophy (Russell (1997)). It would perfectly suit Evans’s conception of the fundamental level of thought, treated as something pretty demanding. According to Russell, we must distinguish “the thought in the mind of a person using a proper name correctly” (1997, 30), which typically involves a definite description and so, for Russell, is quantificational, from “the proposition we should like to affirm”, which is singular (1997, 31). If we utter “Bismarck was an astute diplomatist” the thought guiding our utterance is descriptive (perhaps that the first chancellor of the German Empire is an astute diplomatist), for that is the only way in which we can think about Bismarck, given that we are not acquainted with him. By contrast, what we assert is singular, a proposition containing Bismarck himself. We know this proposition exists, because we know that Bismarck (i.e. the first chancellor) exists, and the truth or falsehood of what we say depends on whether or not Bismarck himself is as we state him to be. But we cannot think the proposition, since we are not acquainted with Bismarck.¹³

This is a beautiful example of just what Evans needs in order to maintain a significant contrast between saying and thinking. However, it is of no real help to him, since no one nowadays accepts the conjunction of Russellian assumptions on which it depends: that we can be acquainted only with sense data (and so we are not acquainted with Bismarck) but can understand a name only if we are acquainted with that to which it refers (and so cannot understand any name for a material object).

Evans insists that a question as important as whether one can usefully distinguish saying from thinking “is not to be decided by evidence as to the pattern of ordinary English usage with respect to propositional attitude words” (1982, 76). Instead, the decision must be made in the light of theory, which may correct “the deliverances of untutored linguistic intuition”. Perhaps unsurprisingly, Evans regards the relevant theory as containing RP, and this indeed suggests a distinction between saying or referring to on the one hand, and thinking about on the other. By contrast, Kripke’s theory of proper names makes them easy to acquire: one just needs to be in the right kind of causal contact with an existing practice of using the name. So if using a name enabled one to think about its bearer, thinking about objects would be easy, and hence would not require RP.

Perhaps with this dialectic in mind, Evans criticizes Kripke’s theory of names in The Causal Theory of Names (Evans 1985a), where he suggests that Kripke’s view of names is mistaken, because it does not allow a name to change its referent. Yet such change is what we seem to have in two cases Evans asks us to consider: “Madagascar” and “Jack”. I shall argue, contra Evans, and indeed contra most discussions,¹⁴ that it is impossible for a name to change its referent, so that there cannot be such counterexamples to Kripke.

Marco Polo learned the name “Madagascar” from people who used it for a town on the mainland. Through error, he applied a similarly spelled and pronounced name to an island in the Indian Ocean,¹⁵ and this became the standard use of the word (at least by Europeans). Evans assumes that the name used by Marco Polo’s sources is the same name as that used by Marco Polo. Under this assumption, we have a change of referent, from a town on the mainland to an island out at sea.

Kripke partly characterizes his conception of a name-using practice in these words, which are the target of Evans’s criticism:

When the name is “passed from link to link”, the receiver of the name must, I think, intend when he learns it to use it with the same reference as the man from whom he heard it. (Kripke 1980, 96).

Marco Polo no doubt did intend this. But Kripke continues: “refinements must be added to make this even begin to be a set of necessary and sufficient conditions”. An obvious refinement, perhaps one that Kripke was taking for granted, is:

the novice should know that their source is using the target name referentially and should successfully intend to use it to refer to what their source referred to.

We all know examples of success. Marco Polo provides an example of failure: he tried but failed to conform to the practice of those from whom he wished to acquire the name.¹⁶ Hence the example does not undermine Kripke’s theory as slightly modified above.

The same holds for Evans’s example of “Jack”. Baby A is christened Jack, and sleeps in a cot marked “Jack”. Through error, baby B is placed in the cot, and no one realizes the mistake. Thinking they are continuing the practice of calling baby A Jack, people call baby B Jack (and baby A gets baby B’s name). Their intention to use the name “Jack” as their sources used it failed. So it is not a case in which the later speakers joined the name-using practice they intended to join, and so not a case of a single name shifting its referent. Rather, the adults inadvertently initiated a new practice, creating a new (specific) name that is pronounced and spelled just like the old one.

This claim presupposes a certain view about the identity conditions for names. Kaplan (1990) asks whether he and Hume have the same name ‘David’, or whether they are different Davids. His solution is to say there are two kinds of names. I shall mark the distinction as that between “generic” and “specific” names.¹⁷ Hume and Kaplan share a generic name “David”, but they have distinct specific names of this genus. Likewise, the inland town and the island share the generic name “Madagascar”, and baby A and baby B have the same generic name “Jack”. But one specific name “Madagascar” refers to the inland town, and another specific name, spelled the same way, refers to the island. And one specific name “Jack” referred to baby A and another specific name of this genus came to refer to baby B. The names are as different as “David” applied to Hume and “David” applied to Kaplan. Clearly, semantic theorizing about names needs to focus on specific rather than generic names. Generic names do not have semantic properties.

Generic names may be thought of as name-templates. They are not really names,¹⁸ as they are not used to refer, though they can occur in lists of suitable names for your baby or pet, and they come closer than specific names to being individuated by spelling or pronunciation. But there is no clear ruling: some say that “Ann” and “Anne” are the same (generic) name differently spelled, others that they are different generic names because of the difference of spelling. On the other hand, most people say “Paris”, referring to the French capital, is the same name whether spoken in France or in the US despite the different pronunciation. The main point is that the fact that the same generic name is used in name-using practice A and also in name-using practice B does not entail that A and B involve the same specific name. The names may have different referents, in which case the practices are different.

It is natural to see name-using practices as individuated by their specific names, and specific names individuated by their referents. These principles are enough to ensure that no specific name changes its referent, or gets used in distinct name-using practices. They ensure that to the extent that Evans’s alleged counterexamples to Kripke’s theory of names rely on reference change, they are based on a mistake.

In the closing pages of The Varieties of Reference, Evans gives us a positive picture of proper names, involving a distinction between producers, who initiate the use of a name, and consumers, who extend extant use. Producers “do more than merely use the name to refer to x [its bearer]; they have dealings with x from time to time, and use the name in these dealings” (1982, 376). By contrast, consumers “are not able to inject new information into the practice, but must rely upon the information-gathering transactions of the producers” (1982, 377). Consumers are here characterized as always deferring to a specific kind of testimony, and having an “overriding intention to conform” (1985a, 21).

Producers are well-placed to satisfy RP, for “having dealings with” a name’s referent presumably enables them to think about it. Consumers might not be so fortunate. This combination is possible only if the deference exercised by consumers does not enable them to think the thoughts of those to whom they defer. In the light of standard views about deference, this limiting presupposition requires defense. In Burge’s original example (1979), deference is characterized not just in terms of what the subject referred to when using “arthritis”, but also what they thought about (namely, arthritis). In a full account, Evans would need to explain why the deference we exercise in joining a name-using practice does not automatically enable us to use the name to think of its referent, which is the analog of what deference achieves in other cases.

Behind this discussion, there looms another issue, which I will mention but not discuss. Evans assumes that producers of a name will “have dealings with” its referent. This is possible only if there is a referent whereas, as already noted, there are many kinds of name lacking a referent. It may be good policy to start one’s account of the semantics of names by considering the “good” cases, those in which the names have a referent. But one should not ignore other cases, in which non-referring names are used, in apparently intelligible discourse, to express thoughts that are apparently true or false.¹⁹

Acknowledgements

This paper stems from my Gareth Evans Memorial Lecture, given in May 2024, organized by Bill Child. My thanks to Bill for inviting me and for many helpful comments. I would like also to thank other members of the audience on that occasion for their comments, and two very helpful referees for this journal.