Reference And Sense

1 Introduction

Reference And Sense are two Fregean terms which are key to understanding his view of how we refer to objects. His picture solves a number of tricky problems arising when we seem to talk about things that don’t exist. I will illustrate the view by discussing the approach of Evans and I will raise some questions about the Evans view.

‘Reference’ is that property of a referring term by which it picks out or selects its referent. Its referent is the actual object in the world picked out.

‘Sense’ is the ‘mode of presentation’ of the object, or the way in which it is referred to. Since the introduction of sense allows us to refer to the same object in multiple ways, we can now account for the informativeness of identity statements. This is because different expressions have different senses. Different senses may have different cognitive values, even if the expression refers to the same item.

I will argue that Evans cannot use his axiom (9): (∀x) (The referent of ‘a’ = x iff [a] (x = a)) to show the sense of terms as he desires. [1, p. 38] So (9) is not that useful in understanding Reference And Sense.

Initially, I will outline the historical problem of empty terms that Evans is seeking to address, and then discuss his argument to axiom (9) and the problems with that argument. I will conclude that axiom (9) is not useful, primarily because it does not show the senses of terms or otherwise elucidate them. In addition, I will argue that there are problems with the whole category of descriptive names that Evans introduces; that he does not improve on Frege because his approach is no longer Fregean. Evans notes some logical problems but does not solve them. Finally, Evans’s approach is poorly motivated. For all five of these reasons, axiom (9) is not useful.

2 Reference And Sense: The Difficulty With Empty Terms

It is usual in probably all non-philosophical contexts and many philosophical ones to think that a word points to an object in the world. The meaning of a term is closely bound up with the object to which the term refers. This leads to a problem with empty terms though. Since they do not refer to any object, they can have no meaning on this analysis.

This runs counter to our very strong intuition that a meaning is nevertheless involved when someone fails to refer by accident. For example, imagine they gesture at a tree and say ‘that lime tree is x’ when in fact the tree they aim to refer to is not a lime. Evans correctly observes that any theory must resolve the strong impression we have on such occasions that communication has occurred.

Frege [2] was considering related problems, including how it could be possible for true identity statements to be informative. Frege approached this by introducing two components of meaning: Sinn and Bedeutung, standardly translated as ‘sense’ and ‘reference’ respectively.

Sense As Mode Of Presentation

Frege’s solution to his problem may create the space for a solution to the difficulty with empty terms. That is because now meaning has two components. Reference And Sense are both parts of meaning.

Thus, while an empty term may have no referent, it could perhaps have a sense, which would mean that it could have at least that element of meaning. Perhaps that could account for the strong impression of communication. Evans believes that while we can commend much in Frege’s treatment, the latter goes wrong in his views on empty terms.

Inaccuracy Destroys Reference

In Frege’s view, inaccuracy in a description destroys reference. If I aim to refer to a tree as a lime tree but it is a different type of tree, I have failed to refer and a sentence containing such an empty term is not truth-evaluable – it expresses no thought. Consider the situation where I aim to refer to the nearest tree to me, which is in fact an elm, but say ‘that lime tree’, when the second tree is in fact a lime. My hearers would misunderstand me, if they knew which tree was in fact the lime.

We could not call this communication. If I wished to direct their attention to a squirrel in the lime tree, they would look in the wrong place, unless they shared my error. Relying on mutual error in that way is not communication. Nor should we rely on the likely outcome that they would scan all trees in the approximate area indicated. Even if that is communication of some form, it is not communication proper of the thought as intended.

The problem with empty terms can also be glossed in terms of an obstacle in what would later become the two-step verificationist procedure for assessing whether someone understands a claim ‘a is F’. The first step is to find the object a. The second is to establish whether the predicate F applies to it.

For empty terms, there is no object a and so the first step is impossible. Clearly it would be anachronistic to suggest that Frege was thinking in this way. But we might nevertheless wish to improve on his account if we wished to combine it with a more modern verificationist approach, which does have the merit of simplicity.

Reference And Sense: What Happens In Fiction?

Frege thinks that when we fail to refer, we perhaps unknowingly enter the realm of fiction – and in this way we can have sense without a referent. So Reference And Sense would have come apart.

For Evans, this response is unsatisfactory because it fails to account adequately for the phenomenology, being the strong impression we have on such occasions that we are successfully communicating with one another without entering a fictional realm.1 Also there seems to be something that constitutes understanding a sentence containing an empty term even though, strictly speaking, we can give no procedure for its verification. So Evans will aim to retain the sense and reference elements of Frege’s approach, but allow (contra Frege, in Evans’s view2 ) that there can be sense without a referent, and provide a better account of empty terms.

3 Reference And Sense: Evans’s Argument

Evans notes that for Frege, sense is a mode of presentation of the referent, which means that sense is a mode of presentation of the semantic value, since Frege has equated semantic value and referent. If ‘Aphla’ and ‘Ateb’3 have the same referent then we can see the equivalence of 1). and 2). below.

1. The semantic value (i.e. the referent) of ‘Aphla’ = Aphla

2. The semantic value (i.e. the referent) of ‘Aphla’ = Ateb

But we can only see the sense of the name from 1). That is because when competent users of the term ‘Aphla’ are asked for the way in which the term picks out its referent, they will indicate the object that is Aphla. They will do this while they may be completely ignorant of the term ‘Ateb’ and of the additional fact that they are also indicating the object Ateb. The term ‘Ateb’ has a different sense to the term ‘Aphla.’ We can not be shown the sense of the latter by the sense of the former even though the terms are co-referring. Again this separates Reference And Sense.

Evans argues empty singular terms cannot be accommodated in the framework of a ‘typical truth-theoretic clause’ in the form of axiom (7) [1, p. 35] below:

(7): The referent of ‘a’ = α.

Evans will continue by proposing axioms (8) and (9) as improvements. I will discuss these below, but first I will outline Evans’s argument that we need to replace (7).

Reference And Sense: Non-Referring Singular Terms

Russell [5] does not countenance non-referring singular terms, because he wishes to avoid the problems described in §2 above. So Evans introduces the expression ‘Russellian singular terms’ for those singular terms that do refer i.e. that have a referent. Axiom (7) is clearly inadequate as an analysis for non- Russellian singular terms. They have no referent and so there is no object α. We can give no truth conditions for sentences containing the term ‘a.’

Therefore, if Evans can show that there are non-Russellian singular terms, given the inadequacy of (7) to cover those terms, then (7) is inadequate since it does not cover some terms that it should.5

The Zip

Evans introduces the stipulation (3) below: (3): Let us call whoever invented the zip ‘Julius’.

Thereby he argues for the existence of non-Russellian singular terms via the claim that ‘Julius’ is one of them. Evans introduces the category ‘descriptive names’ for terms such as ‘Julius.’ This simply means that they are names whose referent is determined by description rather than by baptism or otherwise.

Photo by Pixabay on

It is unclear whether there is a single unique individual Julius. A number of people may have invented the zip. (Apparently this is in fact true in the actual world.) Or there could be no zips in a different possible world. We could indeed replace ‘zip’ by some invention which does not exist. In any case, in the scenario to be considered, ‘Julius’ may or may not have a referent. Evans claims convincingly that there is no problem understanding (4) below:

(4): Julius was an Englishman;

There is a clear sense to (4), and it is intelligible. These facts obtain irrespective of whether ‘Julius’ does or does not refer. If that is true, then we must replace (7), since we can have an empty term in a sentence but we can nevertheless understand the sentence.

Reference And Sense: A Proposal For A New Axiom

Evans’s first proposal for a replacement is (8) below: (8): (∀x) (The referent of ‘a’ = x iff x is φ)6

Evans introduces the underlining notation to indicate the uniqueness7 of x. It is the only entity which satisfies φ. Naturally, x must exist in order to so satisfy. Thus Evans is only allowing existing entities to be referents, thus ensuring that axiom (8) does not attempt to assign non-existing objects as referents of empty terms.

Reference And Sense: Classical Logic Problems

Evans now observes that classical logic has two problems which will mean we must modify (8). The first problem is a result of Existential Generalization, which licenses inferences in the following form:

( . . . a . . . ) → (∃x) (x = a & ( . . . x . . . ) )

This means allow of the inference from any sentence about an object a to the existence of some x such that x is a and also such that the same sentence is true of x. This is clearly unacceptable when we admit empty terms. We could move from a sentence referring to Zeus, who does not exist, to a further sentence to the effect that Zeus does exist.

The second problem is similar. It results from Universal Elimination, which licenses sentences about an object a to be asserted on the basis of a universal statement such as:

(∀x)( . . . x . . . ) → [ a ] ( . . . a . . . ) )

Evans introduces the square bracket notation [ a ] to indicate that the object a exists – so therefore the term ‘a’ refers. This can be read as ‘there is a unique a such that . . . ’. So the inference above means approximately ‘if for all x a sentence is true of x, then there is some unique object referred to by the term a such that the sentence is true of that object. Again, we cannot accept this because we do not wish to admit that sentences including an empty term ‘a’ can be true.

Free Logic

The solution to these two problems is to replace classical logic with a Free Logic which restricts the application of Existential Generalization and Universal Elimination. We can only employ the rule of Universal Elimination if there is in addition a premise to the effect that (∃x) (x = a) i.e. the object a must exist so that the term ‘a’ can refer to it.

Evans notes in relation to this insistence that ‘[w]ith this in mind, we can see that there is no obstacle to using a name like ‘Julius’ to state its own semantic contribution’.8 So we should replace (8) with (9):

(9): (∀x) (The referent of ‘Julius’ = x iff [Julius] (x = Julius)).

Again, the square brackets indicate that there must exist some unique satisfying object, so that (9) can be read as ‘for all x, the referent of the term ‘Julius’ is Julius iff there is some unique Julius such that x is Julius’.

Significance As Truth Value

In the Fregean theory, the significance of a sentence consists in its being true or false.9 The ‘semantic value’ of a term is its ability to affect the truth value of a sentence in which it occurs. That’s because we make up the semantic value of a sentence – a truth value – from the semantic values of its components.

For Frege, we determine the semantic value by the association of the term with an extra-linguistic entity. So axiom (9) is telling us what that extra-linguistic entity is. Namely, it is the person Julius, should he exist. Also, he must be the unique satisfier of the description we have in mind – namely being the inventor of the zip.

Alternative Semantic Values

Axioms like these ‘give truth-theoretical expression to the alternative semantic values contemplated for names in 1.7.’ Those alternative semantic values are the null set for empty terms and singletons10 otherwise. [1, p. 36] Now Evans allows that there can be sense without a referent. We see this from his plausible claim that the statement ‘Julius was an Englishman’ has a sense. We can understand it even if ‘Julius’ does not refer. He diagnoses the critical problem with a neo-Fregean theory not as being the allowance of sense to sentences containing empty terms. It is instead the ‘equation between semantic value and referent’.11 [1, p. 32]

On that view, empty terms have no semantic value. That means they have no Meaning.12. So it is hard to see how they can have sense and we can understand them.

Evans replaces this equation by holding that semantic values are sets. For empty terms, the semantic value is the null set. That means that an empty term has no referent because the null set has no members. The sense of ‘Julius’ is a mode of presentation of the semantic value of the term. The null set is that semantic value if there is no Julius.13

Truth Theory Or Sense Theory?

If we view Evans’s axiom (9) solely as a truth-theoretic axiom, we will accept it because it produces the right truth conditions for all names, including Russellian ones. But there is a problem for Evans because he is also attempting to provide a theory of sense. In fact, Evans sees the truth theoretic axioms as showing the sense of the semantic value, which means also showing the sense of the referent – because the sense is the mode of presentation of both.14 But axiom (9) does not do this i) in the empty case because there is no referent and ii) in the non-empty case as the semantic value is whatever the semantic value of a description is, which is also not a referent. So Evans’s axiom for Julius cannot show the sense.

Reference And Sense: Sense Theory Is Understanding Theory

We can see this by recalling the line taken by McDowell who reminds us [7, p. 169] that a theory of sense is meant to be a theory of understanding – we want to know how people know truths that are expressed by sentences, not how they know that the sentences are true. We can only complete the former task by examining senses, by considering the ways in which people refer to Julius. If they cannot so refer, because ‘Julius’ does not refer, we will not be able to consider how they refer.

So we have shown that Evans’s approach fails to provide a theory of sense. There is in addition a further problem with the whole category of descriptive names as well, as I will now outline.15

Geach [10] illustrates the problem with his example of a name – ‘Fifi’ – given to a particular diamond specified by description, such as ‘the diamond in a certain pendant’. ‘Fifi’ would still name that diamond even if it had fallen out of the pendant and no longer fitted the description.16 So Geach denies that there are any descriptive names at all because in his counter-example, the name continues to refer to an item which no longer meets the original description. And since the original description gives the sense, the sense of the name is also now detached from the object.

Past Achievements

Evans only avoids this difficulty because the description Evans employs relates to a past achievement of Julius. If it took place, it cannot be undone. Thus it always continues to be available to refer to Julius, if he exists. So the question becomes whether Evans can restrict descriptions to those which will always function to select a particular referent where it exists.

Timing Of Names

We can imagine that an object is named at time tname. We give it a descriptive name, and supply a particular description. That description will relate to circumstances at a time tdes . In the case of Julius, tname is later than tdes and so the name cannot subsequently become detached from its original object. Geach notes there is no reason at all for tname and tdes to occur in that order. ‘Fifi’ applies to the diamond to begin with because it meets the description but then the name stays with the diamond when it no longer does. There are other problems of this type.

I could on Thursday adopt a name for the egg that I will eat for breakfast on Friday morning (i.e. tname is earlier than tdes ). One might argue that the no effective name giving has taken place in those circumstances, because it is possible that I will in fact eat no eggs on Friday. But if I do, the name applies, surely, to that egg. So Evans needs a large number of temporal restrictions within his category of descriptive names. These restrictions appear ad hoc and there is no obvious reason to permit them to Evans.

Distinct Treatment Of Empty Terms

Evans claims to be operating in a neo-Fregean framework, modifying that approach within compass. But his treatment of empty terms is not consistent with that of Frege. For Frege, the sense of an expression is the condition that must be met by the referent.17 The natural gloss on Frege’s position is that the sense of an empty term is then the condition that would be met by the referent, were it to exist. But for Evans, the sense of an empty term is a mode of presentation of the null set [1, p. 32] which is too dramatic a departure from Frege. I have already discussed the general disagreement with Evans’s claim that Frege does not allow sense to empty terms and how this leaves a question mark over the motivation of the whole enterprise.

Moreover, Sainsbury [8, p. 66] observes that Evans is working with negative Free Logic (“NFL”) – which Sainsbury also recommends – but that Frege himself used his own different version. In NFL, all sentences containing empty terms are false. This is true even for statements of self-identity. For example, if ‘Vulcan’ does not refer, then even ‘Vulcan = Vulcan’ is false. These types of sentences might be thought the strongest candidates for true sentences containing empty terms. For Frege, sentences containing empty terms lack truth values. Evans cannot claim that axiom (9) enables him to produce a basically Fregean framework with an improved handling of empty terms if he has distorted Frege too greatly in the attempt.

Reference And Sense: Problems With Negative Free Logic

Sainsbury [8, p. 69] notes problems with NFL which Evans will naturally also need to deal with. Evans agrees that NFL will say that all sentences containing empty terms are false. We can see that Evans is on board with this because he agrees that ‘Julius = Julius’ is false just because it might not refer. This means that both of the following sentences are false.

1. ‘Vulcan is distinct from Vulcan’

2. ‘Vulcan is identical to Vulcan’

If sentence 2 is false, then ¬2 is true. But ¬2 has the same meaning as

‘Vulcan is not identical to Vulcan’, so that is also true. But that is synonymous with item 1, so we have different truth values for the same sentences. This argument relies on a Russellian scope ambiguity between the negation of (a is

F) being ¬(a is F) or (a is ¬F) because Sainsbury needs the synonymity of

¬‘Vulcan is identical to Vulcan’ and ‘Vulcan is not identical to Vulcan’, which

may be questionable. To be fair, we know Evans is aware of these problems and that he is also aware of problems arising from negation in NFL since he also uses the ‘global negation operator’ N [1, p. 52] that creates the difficulties. He observes that insisting on applying only one of the types of negation available – attaching the ¬ to the predicate – is equivalent to insisting that all singular terms are Russellian, which begs the question. But this awareness does not constitute a solution.

Problems With Motivation

There are in addition problems with the motivation for Evans’s approach. The problem of empty terms occurs if we combine non-referring terms with standard rules of logic. Frege and Russell solve this by excluding the terms in different ways. For the former, sentences containing empty terms are fictional; for the latter they are ‘nonsense’. Evans solves it by retaining the terms but adjusting the rules of logic. How do we know this is superior? Evans does indeed avoid the problem of empty terms but at the cost of adjusting logic.

It is hard to show that that price is lower than the costs he avoids. Indeed, many writers would hold that logic is more fundamental than language and that adjusting logic therefore carries a higher price. Quine [11] has suggested that there may be a core and a periphery of our web of beliefs, with the logical laws forming part of the core for revision of which we would require extraordinary persuasion. Evans has not provided such persuasion.

See Also:

Does Heidegger Establish That The Ready-to-hand Enjoys ‘Priority’ Over The Present-at-hand?

Science Is Not A Religion

Nagel And Generalisation Of The Categorical Imperative

Can Inductive Reasoning Be Justified Without Using Induction?

Reference And Sense: Footnotes

1 There may be some evidence that Evans is more sympathetic to this fictional line elsewhere, but I will consider only his views in [1] here.

2 Although Evans thinks that Frege does not allow sense to sentences containing empty terms, this is a minority view. [3, ‘Supplement: Evans on Frege’] Evans is right though, contra Bell [4], to explain Frege’s explanation of our frequent uses of empty terms in apparently communicative ways with the claim that we have lapsed into fiction. Ironically Bell accuses Evans of a mistranslation but is guilty of one himself.

Bell claims that Frege’s term ‘Gedanken, dem Scheine nach’ is best translated as ‘thoughts, to all intents and purposes’ and uses this as evidence to attack Evans’s translation of ‘Scheingedanken’ as ‘mock thoughts’. This is wrong because ‘dem Scheine nach’ means ‘after the appearances’ i.e. in name only. So Evans is right to say that Frege’s term ‘Scheingedanken’ means mock or apparent thoughts – which are not thoughts. A very large number of examples of German terms prefixed with Schein that mean ‘false’ can be supplied; for example Scheinehe – false marriage; Scheinfirma – dummy firm; Scheinbild – simulacrum.

3 These are two names used by different groups of persons for the same mountain in a scenario outlined by Frege. No thought is expressed when someone states such a sentence. We cannot understand it.

4 Evans’s own gloss here on the problem is that for empty terms, no axiom such as (7) can be truly stated. Sainsbury [6, Ch. 1] alternatively describes the problem as being that axioms in this form entail the existence of the referent. This would follow from a ban on empty terms. Sainsbury also claims that the decision at this point about how to handle empty terms drives the later choice between classical and free logic, as I will outline later.

Further Footnotes

5 An alternative approach would be to allow that different theories could cover empty and non-empty terms. Intuitions will differ here, but many will have sympathies with Evans’s general view against such ad hoc procedure.

6 Some might argue that the problems I outline in this essay are in fact already present in (8). While this may be so, I retain the reference to axiom (9) in the title of this essay since (9) is Evans’s finished product, as it were.

7 There is also a small vertical dash on the underline that points to x meaning that it is x that is unique.

9 This applies only to sentences which can be true or false, and so not to imperative sentences for example.

8 Various writers including McDowell [7], Evans himself and Sainsbury regard this as a ‘standard’ ‘homophonic’ approach, wherein the sense of Hesperus is shown by stating the referent as follows: The term ‘Hesperus’ refers to Hesperus. This is an ‘austere’ methodology, to adopt McDowell’s term, which is appropriate since semantic theory ‘should not aspire to provide detailed analyses of the meanings of individual words’ according to Sainsbury. [8, Ch. 2] This means that there is no better way of explaining the semantics of a term than to employ the same term and step back a meta-linguistic level. This Evans will now do for terms which may be empty.

10 Evans actually says that these semantic values are ‘formally adequate’ which some claim is consistent with this not actually being his view. That seems strained, but in any case, this is the only line he gives.

Further Footnotes

11 Note that for Evans such an allowance is not a problem for Frege’s actual theory because Evans believes, as previously noted, that Frege does not allow sense to empty terms. Morris [9, p. 37] though goes so far as to claim that Frege used sense specifically to handle the problem of empty names. This is a significant problem for Evans as his motivation for denying the equation is to make it possible to allow sense to empty terms. So he is solving a problem for Frege that Frege does not have.

12 Evans’s preferred translation for Bedeutung, elsewhere translated as ‘reference,’ is ‘Meaning.’ Evans can correctly claim normal German usage as a justification here, but opponents could urge that it is a term of art for Frege. I judge the opponents correct here because ‘meaning’ is really what Frege is trying to elucidate with his theory; it is not an aspect of that theory.

13 Note that Evans is not vulnerable to Sainsbury’s criticism [8, p. 53] that it is not appropriate to model failure of reference on successful reference to a special entity because Evans continues to hold that empty terms have no referent even though they do have a semantic value.

14 Alternatively, Evans wishes only to show the sense in his terms as being the mode of presentation of the semantic value and not the mode of presentation of the referent. These are the same for Frege since Frege equates semantic value and referent. Since Evans denies that equation, he can restrict his claim to sense in his terms. But then he is not explaining Fregean senses.

15 Someone could say that a problem with descriptive names is just that. It is not necessarily also a problem for axiom (9). Be that as it may, descriptive names are key to Evans’s argument for axiom (9). So we are entitled to question the latter if we can question the former.

16 Note that ‘Fifi’ is not a term equivalent to ‘whichever diamond is in the pendant’; it is a ‘descriptive baptism’, so to speak.

17 Bell [4, p. 275] takes this line.


[1] G. Evans and J. McDowell, The varieties of reference. Clarendon Paper- backs Series, Clarendon Press, 1982.

[2] G. Frege, “Sense and reference,” The Philosophical Review, vol. 57, no. 3, pp. pp. 209–230, 1948.

[3] G. Fitch and M. Nelson, “Singular propositions,” in The Stanford Encyclo- pedia of Philosophy (E. N. Zalta, ed.), Metaphysics Research Lab, CSLI, Stanford University, spring 2009 ed., 2009.

[4] D. Bell, “How ‘Russellian’ was Frege?,” Mind, vol. 99, no. 394, pp. pp. 267–277, 1990.

[5] B. Russell, “On denoting,” Mind, vol. 14, no. 56, pp. pp. 479–493, 1905. [6] C. Macdonald and G. Macdonald, McDowell and his critics. Philosophers and their critics, Blackwell Pub., 2006.

[7] J. McDowell, “On the sense and reference of a proper name,” Mind, vol. 86, no. 342, pp. 159–185, 1977.

[8] R. Sainsbury, Reference Without Referents. Oxford University Press, USA, 2007.

[9] M. Morris, An introduction to the philosophy of language. Cambridge introductions to philosophy, Cambridge University Press, 2007.

[10] P. T. Geach, “The varieties of reference by Gareth Evans edited by john mcdowell oxford: Clarendon press, 1982, xiii + 418 pp., £15.00, £5.95 paper,” Philosophy, vol. 61, no. 238, pp. 534–, 1986.

[11] W. V. Quine, “Main trends in recent philosophy: Two dogmas of empiricism,” The Philosophical Review, vol. 60, no. 1, pp. pp. 20–43, 1951.


Theories Of Reference: Evans On Russell

Introductory: Russell’s Criterion

The stated aim of the chapter [1, Ch. 2] is to examine Russell’s theory of reference. These views were only held by Russell at one particular point (around 1918). [2] Russell’s model is basically that of Frege, but again at only one point — before 1890. Evans’ basic theme will be the claim that Russell is often right for the wrong reasons.

Referring expressions have exactly one function — to identify an object in such a way that if the object satisfies the predicate, the sentence containing the expression is true; and if not, it is false. If the term does not refer, the sentence is not truth-evaluable. The person saying it has said nothing; it is `nonsense’ (Russell). [`Nonsense’ is a very interesting term of course; it specifically denies what Frege says can be there even for non-referring terms — Frege says this in order to account for our clear view that sentences containing non-referring terms can be significant.] Russell’s model did not follow Frege in Frege’s introduction of the distinction between sense and Meaning; [recall that Evans is using Meaning for what is normally translated from Bedeutung as reference].

Russell’s argument for the non-sensicality claim relies on the claim that name is a means only; it does not occur in the sentence. We as it were look through it to what it points to in the same way that we will rarely consider a signpost as such; we will be more interested in the location to which it directs us. And if the only significance of a signpost is what it points to, then we don’t care about the difference between signposts — so long as they point to the same thing.

Photo by Artem Bali on

This leads directly to what Evans calls `Russell’s Criterion’ which states that if you can imagine that the subject of a proposition does not exist, and the proposition does not become meaningless, then the subject is not a proper name: it does not represent some object.

All definite descriptions like `the φ’ fail this test. Evans agrees with Russell on this point — cf. his `Julius’ argument. [Recall: Julius was the inventor of the zip.] So definite descriptions are not names. However, we can still understand (and truth-evaluate) `the φ is F’ irrespective of whether `the φ’ refers. Evans does not agree with Strawson [3, `On Referring’] that nothing has been said by a sentence containing an empty term.

Russell also holds that you cannot think about something unless you know which one it is. There are two ways of knowing that: by acquaintance or by description. Different persons will therefore be effectively using only approximately the same names because their way of knowing about objects will differ.

Radical Reference-Failure

A corollary of this last point is that there can only be private languages (of full precision). This is held to be absurd — it is constitutive of languages that they not be private.

If you are acquainted with an object, it must exist. Thus if you think of an object (successfully refer to it) you know by acquaintance, it must exist. [We must guard here against any kind of reification error { there is no suggestion that thinking makes it so. Just that if acquaintance was the route, then you must be acquainted with something.] Russell wants to rule out the possibility that you can be mistaken if you think you are thinking of an object. The infallibility relates only to the mental components of predicates.

Evans wishes to deflate the problem that Russell has: Evans thinks it is not incoherent for a subject to have a thought which phenomenologically is as of object x, but that there may be no physical object x, and then `he may fail to have a thought of the kind he supposes himself to have’. [Is this enough? Is a thought `of a different kind’ the same as `no thought’? A thought of a hallucinatory object does appear in one sense to be a thought about nothing, but isn’t Russell worried about it being a thought about nothing at the as it were — mental end of the thought? Bear in mind that Russell is a sense data theorist and so intentional objects are mental for him.]

Evans concedes that a full defense of his position would require an account of the nature of demonstrative identification of material objects.

Evans also considers Prior’s view that someone must be able to think `a is F’ because it is part of the compound thought `I am thinking that a is F’. But Evans denies that those who claim that someone can be mistaken that they are thinking `a is F’ are committed to them also thinking that they are thinking that thought.

Russellian Singular Terms And Descriptive Names

Define: Russellian Singular Terms are those whose sense depends on them having a referent. Evans wishes to improve Russell’s model in two primary ways.

The first is to revert to Frege in allowing sense to empty singular terms. The second is to avoid the private language problem; Evans aims to do this by abandoning the `Cartesian’ requirement that you cannot think about an object x unless x exists. But — retain Russell’s opposition between proper names that refer and definite descriptions.

All sentences have truth conditions. They can only be true if they do not contain empty terms. Sentences containing empty terms are useless rather than nonsensical. [But even this is too strong. The objections that can be brought against the claim of nonsensicality are as valid against the claim of uselessness. We can and do communicate using claims about lime trees that do not refer and such communication is neither nonsense nor useless. What Evans means by useless here is that if we have a name for something which turns out not to exist — like the 10th planet — we will stop talking about it. But not if the name is `doing duty’ for something that does exist. Further — Evans does not explicitly state here that all sentences which are not true are false; one position would be to hold that sentences containing empty terms fall into a truth-value gap.]

This section will deal only with `descriptive names’ like `Julius’; definite descriptions are postponed to the next section. [So Evans may be disagreeing with Russell: terms like `Julius’ can refer. Alternatively, we can say that `Julius’ is not a definite description and therefore Evans is not disagreeing with Russell.]

Understanding a descriptive name involves understanding what it will be for an object to fit the description. [By contrast, proper names just refer to whatever they have been introduced to refer to; you cannot fail to understand a proper name because even failing to know what it refers to does not constitute failing to understand it: there is nothing to understand.]

Even when most people know only one thing about a named individual –(Homer wrote the Odyssey) — that is just an piece of data about Homer rather than a signal that one intends to use the name to refer to whoever wrote the Odyssey. This is Kripke’s line. Examples of descriptive names include `Jack the Ripper’ (the person committing murders in the East End) and `Deep Throat’ (the person leaking in Watergate). And we can introduce new ones, like `Julius’, by stipulation.

Principle P connects reference with truth:

(P): If S is an atomic sentence in which the n-place concept-expression R is combined with n singular terms t1 . . . tn, then S is true iff < the referent of t1 . . . the referent of tn > satisfies R.

(P) defines both reference and satisfaction simultaneously in terms of truth. [Bold.] Any expression refers if its contribution to the truth-condition of sentences containing it is given by P.

Returning to `Julius’, Evans excludes non-referring uses, producing the axiom (2) below.

(2): (∀x) (The referent of `Julius’ = x iff x uniquely invented the zip.

Thus `Julius’ does not refer if no single person invented the zip.

This is held to be equivalent to (3) below.

(3): (∀x) (The referent of `Julius’ = x iff [Julius] x = Julius)

[The square brackets indicate the uniqueness condition; and `Julius’ is used — not mentioned — as being the inventor of zips if there is one. So we get (4) or equivalently (5) below.]

(4): `Julius is F’ is true iff the inventor of the zip is F

(5): `Julius is F’ is true iff [Julius] Julius is F

— which can be read as `Julius is F’ is true iff there is one unique Julius such that Julius is F. Again, nothing is said about occasions when either there is no unique Julius or there is one but Julius is not F other than `Julius is F’ is not true in both cases.

So if there is a unique inventor of the zip, `Julius’ refers to that person. The argument for the equivalence of (4) and (5) is that they are belief states and the evidence for (4) would be the same as the evidence for (5). It is admitted that the two sentences after `i’ in (4) and (5) embed differently in modal operators, and that this must be explained — such discussion is postponed to section 2.5. [By this, Evans means that `necessarily, the inventor of the zip invented the zip’ is true while `necessarily, Julius invented the zip’ is false.] But Evans denies that this is sufficient to make them separate thoughts.

Photo by Pixabay on

Definite Descriptions

[This is by far the largest section of this chapter at nine pages.]

Russell gives three arguments against admitting that definite descriptions refer.

1. George IV did not know that Scott = the author of Waverley but he did know that Scott = Scott.

2. The negation of `The φ is F’ is `The φ is not F’; the disjunct of the two must be true by the law of excluded middle; both disjuncts entail that the φ exists; this is absurd.

3. Finally, the argument we have been considering i.e. that if definite descriptions are treated as referring expressions then sentences containing empty terms will be meaningless.

On the first, Evans thinks Russell should have seen that Frege’s concept of sense allowed for the differing cognitive significance of co-referring terms. On the second, Evans observes that there is no difference between wide-scope and narrow-scope negation, because : (a is F) is meaningless for Russell if `a’ does not refer instead of being true. But in a Free Logic, there is a distinction between (7) and (8) below.

(7): ¬ [a] F(a)

(8): [a] ¬ F(a)

The difference is that between (7) `it is not the case that there is a unique a such that a is F’ and (8) `there is a unique a and it is not the case that a is F’ because (7) and (8) take different views on the existence of a. The truth of (7) allows but does not require that a may not exist while the truth of (8) requires a to exist.

Russell should have chosen wide-scope negation which means that the existence of the would not be entailed by the negated disjunct. [There is a fair amount in this chapter of `Russell should have seen x’ when in fairness, it is highly likely that Russell saw x and chose ¬ x for specific purposes.]

We come to (12) below, which is argued for by noting that taking the narrow-scope assumes that all singular terms are Russellian, which begs the question.

(12): ¬ (a is F)

— is intelligible and true if a does not exist.

A theory treating definite descriptions as referring expressions would enable the derivation of truth conditions for sentences containing a definite description, and would include (14) below.

(14): (∀φ)(∀x) (The referent of the φ) = x iff Satisfies (x, φ)

— i.e. `the φ’ refers to x if x uniquely satisfies the description. However, Russell was right that descriptions are quantifiers and Evans says he will show that (14) is not adequate for for all modal contexts.

Evans replaces modal operators with quantification over possible worlds and revises P to P’ in a possible worlds version employing truew i.e. true at world w, and satisfiesw. But there is a problem combining (14) and P’ if definite descriptions are treated as referring. This is that only one reading of certain modal sentences is captured. An example is `the first man in space might have been American’. This may mean either i). Gagarin could have been American or ii). someone else who was American might have been the first man in space.

Only the first reading is accommodated but the second does appear in some possible worlds.

(14) is now relativized to (15) by allowing for reference in a possible world w, and P’ becomes P” which is now relativized both to truew and referentw to allow for both readings. This has been done [4] but comes at the price of relativizing reference in all cases. This seems expensive just to handle definite descriptions [so perhaps Russell was right to be wary of it.] And names get this relativized reference as well, even though they never use it. [This is an anticipation of the Kripkean points of the next and final section.]

Similar ambiguities occur temporally and in relation to binding by higher quantifiers, so we end up replacing `refers to (t, x)’ with `refers tow,t,π (t, x)’ — and this seems complex and suggests that definite descriptions do not belong with referring expressions as a natural kind.

So Evans considers alternative approaches, one of which is to notice the similarity of `the’ in definite descriptions to a quantifier, since all of the following share a form: (∀φ) (φ is ψ); (∃φ) (φ is ψ); ¬(∃φ) (φ is ψ) and The (φ is ψ). This has the merit of simplicity; especially so when compared with Russell’s version (16) of `The φ is F’ below.

(16): (∃x) (φx & ∀y (φy → x=y) & Fx)

— which is a format Russell allegedly required for the Principia.

Another proposal would be to regard natural language quantifiers as binary rather than Russell’s unary ones. These quantifiers would be pairs of functions from objects to truth values — to truth values. This would represent `The φ is F’ as Ix [φ(x); F(x) ] i.e. there is exactly one x which is φ and that x is F.

Chomsky suggests treating `the’ as a universal quantifier with existential import. There is an implication of uniqueness in singular forms; whether the `the’ is being used in singular or plural forms is given by the associated concept. For example, the `the’ in `the man who broke the bank at Monte Carlo’ is singular while that in `the men who robbed your wife’ is plural. Evans himself does not consider plural reference.

`Rigid Designations’ And Fregean Sense

Evans relies on the claim that it is just a feature of how we use language that in truth-evaluation of sentences, we are solely interested in whether the object referred to satisfies the predicate and not in whether the singular term refers. He says we would not accept sentences like (20) below.

(20): `If Haldeman had released the information to the reporters, he would have been Deep Throat’.

[This seems open to question however. Deep Throat was the name the reporters Woodward and Bernstein used for their secret informant to avoid disclosing his identity. Speculation was rife for many years until finally Deep Throat was unmasked as being Mark W Felt. Before that unmasking, speculation considered a number of candidates, and Haldeman was among them. The general public knew only that Deep Throat was the source for Woodward and Bernstein. If they had eventually announced that Haldeman was their source, and that was true, then we would in fact accept (20). What Evans is relying on here is that Woodward and Bernstein knew that Deep Throat was Felt and that was their name for him and not for anyone else. But could they not have dubbed Haldeman with the same name if he had been the source?]

[Evans is on much stronger ground with (21) which seems to attempt to ascribe causal power from leaking to seniority — when the other direction seems more likely; or at least, a senior position would have enabled the leaking if not caused it.]

Evans also says we would not agree with (22) below.

(22): if you had invented the zip, you would have been Julius.

[But again — clearly there can be no claim that your given name would have been different if you had invented the zip. But if we have agreed that `Julius’ is our term for whoever did invent the zip, and that turns out to be you, doesn’t that make that another name for you?]

Evans wishes to avoid Kripke’s term `rigid designators’ because if it is useful, there must also be non-rigid designators. And that re-introduces the relativization of reference to possible worlds which Evans thinks is unnecessarily complex.

See Also:

Quine And Fine on Reference and Modality

The Structure And Content Of Truth For Davidson

Why Kids Are Robots

Why Does Epicurus Think That His Radical Views Will Be Persuasive To The Average Person?


[1] G. Evans and J. McDowell, The varieties of reference. Clarendon Paperbacks Series, Clarendon Press, 1982.

[2] B. Russell and D. Pears, The philosophy of logical atomism. Classics Series, Open Court, 1985.

[3] P. Strawson, Logico-linguistic papers. Ashgate, 2004.

[4] R. H. Thomason and R. C. Stalnaker, “Modality and reference,” Nous, vol. 2, no. 4, pp. pp. 359{372, 1968.