Guide to Chapter Two of Gareth Evans'
The Varieties of Reference

Rick Grush (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.

Chapter 2: Russell

This section begins the work of identifying distinct semantic categories within the grammatical category of singular term. Evans introduces Russell's criterion, a semantic test applied to noun phrases in order to sort them into two groups. The test is: would the noun phrase’s lack of a referent render atomic sentences employing it meaningless? If such matrix sentences can still be meaningful even if the NP is empty, then the NP is not a ‘Russellian singular term’. Russell used this test to distinguish ‘logically proper names’ from definite descriptions, and on this basis he argued that some NPs that looked like singular terms, S/Ns, were in fact, semantically, S/(S/N)s (complex quantifiers or definite descriptions). Evans explains that for Russell, almost all ordinary grammatical NPs, including all that refer to external objects and people, are not genuine referring expressions by this test (since their emptiness would not render matrix sentences meaningless), but rather are covert descriptions, which do not render matrix sentences meaningless if empty.

Russell, like pre-1890 Frege, held the following view of genuine singular sentences (the modifier 'genuine' will be explained shortly). First, the function of a genuine referring expression (aka logically proper name) is to identify an object such that if this object satisfies the predicate of the sentence, the sentence is true, and if the object identified does not satisfy the predicate, then the sentence is false. Second, if no object is identified, then the sentence is an aberration that approximated nonsense. (Russell's view diverged from that of the later Frege, however, for Russell's theory lacked anything analogous to a theory of sense for genuine refering expressions.)

From this vision of the semantic workings of atomic sentences emerges what Evans calls 'Russell's Criterion':

The grammatical subject of a proposition is a genuine proper name, i.e., a name directly representing some object, only if its failing to have a referent renders the proposition meaningless.

Russell used this semantic test as criterial for sorting NPs into two semantic categories: those that passed this test (passing means that the matrix sentence is meaningless) would be genuine referring expressions, aka logically proper names, and receive one kind of semantic analysis (specifically the analysis just described above); and those that did not were (overt or covert) descriptions, and would receive a very different kind of semantic analysis.

Evans agrees with Russell that expressions of the form 'the j', NPs that have the form of explicit definite descriptions, do not pass this test (i.e. they are not Russellian; they are meaningful even when they lack a referent). Where 'j' is a coherent description, sentences of the form 'The j is F' may be understood, i.e. convey a thought, even if nothing satisfies the description, since the description is adequate to specify the truth conditions of the sentence even in absence of a referent (provided the rest of the sentence is up to the task, of course).

But what about other kinds of NP, such as ordinary proper names (e.g. Cicero, Bill Clinton) and indexicals/demonstratives? Russell felt that these expressions were all covert definite descriptions. His reasons were based on a number of factors. The first is what Evans will later call Russell's Principle: it is not possible for a subject to entertain a thought about an object unless the subject knows which object is in question. And on Russell's view there were two ways that this requirement could be met: either the subject could think of the object as the unique satisfier of some description (corresponding to NPs in the semantic category of definite descriptions), or the subject could be 'acquainted' with the object (corresponding to NPs in the semantic category of referring expressions, aka singular terms). For Russell the only legitimate acquaintance relation was between a subject and mental sense contents, including memories of sense contents, and thus private mental indexicals and demonstratives with such entities as their accusatives would be the only 'logically proper' names, that is, the only genuine singular terms. Hence only such items in a private mental language would be amenable to the semantic analysis of proper names described above. And lacking an appropriate mental accusative, a putative mental demonstrative expression would be meaningless.

Russell felt that ordinary proper names failed this criterion, and hence that the kind of semantic analysis appropriate for singular terms could not be their correct semantic analysis. He thus held that they were covert descriptions.

Section 2.2 Radical reference-failure

In this section, Evans introduces Russell's reasons for severely restricting genuine reference, namely, Russell's unwillingness to accept the possibility that one can be mistaken about whether one is having a thought. Evans asserts the coherence of the possibility of being mistaken about whether one has succeeded, on a given occasion, in having a thought. He closes by rebutting an argument to the contrary.

As described above, it follows from Russell's view of acquaintance that genuine reference can obtain only in a private language. Evans is confident in following Wittgenstein in thinking that the notion of a private language is 'absurd'. But Russell had a reason for his restriction: he wanted to rule out situations in which a subject thinks that he is having a thought about an object while in fact he is having no such thought. Such situations seemed to Russell incoherent. And by limiting the scope of legitimate demonstrative thoughts to those whose accusatives were private mental items, a situation could not occur in which the subject is mistaken about the presence of the object of the mental demonstrative, and hence the subject will never mistakenly take himself to be entertaining such a thought when he is not. The objects of ordinary (complex) demonstratives, such as 'that pool of water' in fact seem to allow cases where the subjects take themselves to be entertaining a thought about the demonstratively identified item, but because no such item exists, no such thought is to be had. (The reasons that there appear to be a thought even in such cases will be explored in later chapters.)

Now, of course Russell would allow subjects to be liable to a certain kind of mistake when using ordinary demonstratives or proper names (or entertaining corresponding thoughts). But for Russell such thoughts are analyzable into two components: one purely mental and another nonmental. For example, while Russell would agree that subjects might be in error concerning their own satisfaction of the predicate 'x is thinking that Bismark is F' (that is, Russell allows that subject might take themselves to be entertaining this thought when in fact they are not, perhaps because the name 'Bismark' has no referent), Russell would claim that in this case, the predicate breaks down into two components: a purely mental component of the form 'x is thinking that the j is F', and about this one can have infallible knowledge concerning whether or not one is satisfying that predicate. That is, one cannot be mistaken about whether or not one is entertaining this thought. What one cannot be certain of is what might be called an extra-mental component to the effect that 'Bismark is the j'.

Russell realized that it could not be the case that all thoughts should rely exclusively on descriptive elements. If all thought was by description, then our thoughts could fail to relate to a unique set of objects, since, for every description, it is an open possibility that more than one thing satisfies it. Russell's Cartesian ontology (whereby our own mental states were things that we had immediate and infallible knowledge of) guaranteed that there were a unique set of entities, our own mental contents, to which our system of thought was anchored. This anchoring was carried out by a set of thoughts involving reference to mental entities with which one was directly acquainted. So, for example, a thought that might be expressed in language as "The table is brown" would really be best expressed by a sentence of the form "The y is brown", where "the y" would be a description like 'the physical object that caused this" where the 'this' is a sort of mental demonstrative pointing to a sense datum, say a sense datum of brownness. So ultimately all descriptions are anchored to reference to such mental items which one knows by acquaintance.

Evans will subscribe to Russell's Principle, to the effect that in order for a subject to be credited with thought about an object, the subject must know which object is in question. And furthermore, Evans will argue that in order to understand demonstrative expressions, one must have a certain kind of thought — a demonstrative thought. But Evans wants to extend the legitimate range of possible demonstrative identification beyond Russell's boundaries to regular physical objects, and it is often the case that people take such objects to exists, and take themselves to be referring to them with demonstratives and having demonstrative thoughts about them, when no such object in fact exists. Thus, Evans is committed to the idea that subjects can in fact be mistaken about whether or not they are entertaining a thought of a certain kind on a given occasion. Evans is not arguing that one's mind will be completely empty on such occasions, just that whatever is there will not necessarily qualify as a thought even though the subject might be under the illusion that it does.

Evans, in the last two paragraphs, closes by rebutting an argument by A.N. Prior aimed at supporting Russell's view on the incorrigibility of thought entertainment. Prior's argument is that the thought 'a is F' is a component of the self-ascribing thought 'I am thinking a is F'. So if one in fact has no thought of the form 'a is F', then one will not be able to think 'I am thinking a is F'. Therefore, if one in fact is thinking 'I am thinking a is F', one must also have the thought 'a is F' — that is, one must be correct.

Evans grants that 'a is F' is a component of 'I am thinking that a is F'. Evans' response to Prior, however, is that when one is deceived about one's thought, the thought that is in error is not 'I am thinking a is F' but instead of the form 'I am thinking a thought of type D'. One can be wrong about this, because no particular thought of type D is a component of the type-thought which is a component of the false belief.

Section 2.3 Russellian singular terms and descriptive names

In this section, Evans argues that there are in fact NP expressions which are non-Russellian, yet are genuinely referring expressions. Such expressions are both i) genuinely referring, in that their semantic contribution is not merely descriptive but is rather a function of the referent, and ii) able to play a role as the grammatical subject in an atomic sentence when empty without rendering the sentence meaningless. The point is to show that the class of referring expressions broadly enough so that it includes some expressions which are not Russellian (and thus that 'referring expression' and 'Russellian singular term' are not co-extensive). But it should be kept in mind that this issue is of peripheral interest to the main project of this book.

Russellian singular terms are those such that their having a sense depends on their having a referent. Though Evans recognizes Russellian singular terms, he does not follow Russell in his understanding of what Russellian singular terms are and how they are to be treated. First, Evans will want to ascribe to these terms a Sense, a feature Russell's account does not equip them with. Second, Evans does not, as Russell did, restrict the Russellian singular terms to those that refer to private mental items (and indeed Evans would think that such Cartesian bouts of pure mental reference are not intelligible).

What Evans does like about the Russellian treatment, however, is the insistence on a difference between i) Russellian singular terms and ii) expressions that are tied to descriptive content that provides conditions under which something is the expression's referent. Only the latter can be subjects in sentences that have determinate truth-conditions whether or not they are empty — the former, if empty, are meaningless and if put in the subject position of a atomic sentence this sentence will therefore be meaningless. Evans questions, however, whether Russell was correct to deny that the latter category included any semantic subcategories of expression that were referring at all. Evans postpones his discussion of definite descriptions until 2.4. In this section he focuses instead on descriptive names (discussed previously in 1.7 and 1.8, though Evans feels that there are other members of this semantic subcategory, specifically what he has called e-type pronouns, and 'deferred ostentions', e.g. 'that guy is going to be upset' said when seeing a car loaded with parking tickets, and referring to the car's owner). Descriptive names are names introduced by some reference fixing description. Evans offers that a natural example might be 'Deep Throat', used as a name for whoever in the White House was a secret source of Watergate information. Evans will argue that such names are, semantically, referring expressions, but yet are not Russellian, in that sentences in which they are the grammatical subject can be meaningful even when the descriptive name is empty. Evans is thus arguing that the semantic category of referring expressions has Russellian expressions as one subtype, but that there are, contra Russell, other subtypes.

According to Evans, reference is closely connected to the truth of atomic sentences. He offers as a simple way of defining reference in terms of truth (and satisfaction) the following principle.

(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.

Evans thus uses P to determine whether an expression is a referring expression: any expression whose contribution to the truth-conditions of sentences containing it is stated exclusively by means of the relation of reference found in (P) counts as a referring expression. Thus, referring expression is a semantic category, as its members are discerned by the kind of semantic analysis appropriate for them.

For example, clauses of the form

(1) The referent of 'Aphla' = Aphla

when used to state the contribution of an expression to sentences in which the expression occurs, as in truth theoretic semantics, do so in a way that satisfies P.

Evans argues, though, that truth-theoretic clauses for descriptive names will also be in accord with P.

First, consider that clauses like

(2) (x) (the referent of 'Julius' = x iff x uniquely invented the zip)

or equivalently,

(3) (x) (the referent of 'Julius' = x iff [Julius] x = Julius)

use only the relation of reference to state the semantic contribution that the expression 'Julius' makes to sentences containing it. The combination of (2) (or (3)), normal satisfaction clauses for atomic concept expressions, and principle (P) yields truth-conditions for sentences of the form

(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.

This argument is a bit dense. The basic idea is as follows. Given clauses such as (1), (2) or (3), the contribution of the expression 'Julius' to the truth or falsity of sentences such as (4) and (5) is exactly as prescribed in P. That is, take a sentence such as 'Julius was an Englishman'. This sentence will be true if (and only if) the object referred to by the name 'Julius', the person, if any, who uniquely invented the zip, is an Englishman.

Evans ends this section by claiming that 'Julius is F' and 'The inventor of the zip is F' express the same thought or belief. This is so, he thinks, even though the two sentences embed differently inside modal operators, which he will explain in 2.5.

Section 2.4 Definite descriptions

In this section, Evans argues that definite descriptions (DDs) are not referring expressions (REs), but rather quantifiers. The strategy is to run through three of Russell's own arguments to the effect that DDs are not REs. Evans argues that none of Russell's arguments is adequate. Evans then gives two arguments of his own against the claim that DDs are REs. The first is negative — if you try to treat them as REs, you end up with a clumsy and needlessly inelegant theory with some unexplained loose ends. The second is positive — if you treat DDs as quantifiers, then everything works just fine.

Evans finds in Russell's writings three arguments against treating DDs as amenable to the same semantic analyses as REs.

The first of Russell's arguments is that one can fail to know that two DDs have the same referent. This argument depends on the assumption that referring expressions are such as to render knowledge of coreference obligatory. Evans claims that it is not, and provided that referring expressions can have a sense, there can be such informative identity statements. (It will be Evans's position that that many Russellian expressions have a Fregean Sense, or are otherwise such that identity statements involving them can be informative.)

The second of Russell's arguments is this:

1. Suppose that 'The j' is an RE.
2. For any RE, the negation of 'RE is F' is 'RE is not F'
3. If 'The j' is an RE, then the negation of this sentence must be 'the j is not F'.
4. Disjoining the two gives us 'The j is F or the j is not F'.
5. By the law of the excluded middle, one of these disjuncts must be true.
6. Each disjunct entails that 'The j' refers.
7. But 'The j' may clearly fail to refer. (DDs often fail to refer).
8. Therefore, (1) must be rejected.

Clearly (2) is the crucial move, and it is what Evans will finger as the bad move. In effect, Evans' objection is that (2) is legitimate only for Russellian singular terms, and so Russell's argument assumes that all REs are Russellian.

This move is legitimate for Russellian singular terms because there are two ways to negate an atomic sentence such as 'a is F'. One way it can be false is if a is in fact not-F. The other is that a may not exist. But in the case of Russellian singular terms, this last option is not an option. Therefore, 'Not(a is F)' must be equivalent to 'a is not-F'' only for Russellian singular terms.

But for any expressions that can be meaningful if empty (non-Russellian), then there is the possibility of a wide scope negation distinct from the narrow scope negation, and hence Russell's argument fails, pending the possibility that there can be non-Russellian referring expressions. (We see now the rhetorical reason for including the discussion of descriptive names in Section 2.3, for these are the non-Russellian singular terms that provide Evans' counter-example to Russell here.)

Russell's third argument is that if DDs are REs, then they would be meaningless if empty. They are not meaningless if empty, therefore, they cannot be REs. This argument also assumes an implicit equivalence between referring expressions and Russellian singular terms. Evans insists that if DDs are REs, then they are certainly non-Russellian REs. That is, while it is true that DDs are meaningful if empty, this does not imply that they are not referring expressions. It only implied that they are not Russellian expressions. Russell assumed without argument that all and only REs were Russellian, and this is why he thought this argument worked. But Evans is trying to keep open the possibility that there are non-Russellian referring expressions, and hence, in the context of Evans' concerns, this Russellian argument is question-begging. It shows that DDs aren't Russellian. It does not thereby show that they are not referring expressions.

Though Evans has been criticizing Russell's arguments, he agrees with Russell's conclusion. So next he turns to his own arguments that definite descriptions do not belong to the category of referring expressions (but instead to the category of quantifier expressions). There are two parts to this: first, arguments to the effect that treating DDs as referring expressions can be done, but only in a very theoretically inelegant and ad hoc way, and second, that treating them as quantificational expressions does all the semantical work in a very nice and tidy way.

The crux of the problem is that, according to Evans, if definite descriptions are to be included in the category of referring expressions, then referring expressions must have the reference relation relativized to times, possible worlds, and empty singular term assignments, because DDs behave semantically in ways that would require such relativization. All of this is overkill for referring expressions like names, pronouns, and demonstratives since their reference, unlike the reference of definite descriptions, does not vary from world to world, or time to time.

If, however, one assimilates definite descriptions to the category of quantifiers, then everything is a lot simpler.

From page 57 to the close of this section, Evans discusses competing versions of how to express definite descriptions in quantificational terms. The details are unimportant for understanding anything in the remainder of the book.

[Note: there is a typo on page 59. The sentence in the middle of the first full paragraph should read: "This is to regard 'The' not as an S/( (S/N), (S/N) ), but as a 'unary quantifier former' (an (S/(S/N))/(S/N) )." The text has an incorrect number of parentheses.

Section 2.5 'Rigid designation' and Fregean sense

In this section, Evans makes brief remarks about the relation of the views expressed in this chapter to some terminological distinctions employed by Saul Kripke and David Kaplan.

According to Evans, our evaluation of the truth of sentences containing names, pronouns, and demonstratives is exclusively concerned with whether the referent of the term satisfies the relevant predicate. And this holds even when those expressions have their reference fixed by description as is the case with descriptive names.

The sentence

If you had invented the zip, you would have been Julius,

is infelicitous because 'Julius' was introduced by the agreement "let us call whoever invented the zip 'Julius'". If someone did utter the aforementioned sentence, their intent would be to use 'Julius' as an abbreviation of a description, not to refer to a particular person.

Kripke marked the difference between i) (effectively) the use of names whose reference is fixed by description (descriptive names) and ii) the use of names as abbreviated descriptions by calling the former 'rigid designators'. As should be expected given the previous section, Evans dislikes the commitment to a relativized referential relation that the notions of rigid and non-rigid (flaccid?) designation bring with them. Thus, Evans describes the result of 2.4 as showing the consistency of a singular term's being associated with a clear descriptive criterion for something's being a referent, and its functioning as a rigid designator.

Evans closes this section, and thus, this chapter, by remarking on his disagreement with David Kaplan who asserts that a singular term cannot possess a Fregean sense if it behaves as a rigid designator. Kaplan draws a distinction between direct and indirect reference in which the former is devoid of Fregean sense and the latter, e.g., has a sense constituted by a description. The idea here seems to be that if there is a sense associated with a term, then this sense would be an intermediary between the thinker/language user and the object picked out by the intermediary. On this conception, it would seem that this intermediary might latch onto different objects in different circumstances or possible worlds, in much the way that a description might be true of different objects in different possible worlds.

Evans claims that this is misguided for two reasons. First, Fregean senses need not be thought of as mediating between terms and referents. Recall that on Evans' reading, sense is just a way of thinking about a referent. 'Way' need not be cashed out as a description nor as anything else that would serve to mediate between an expression and a referent. Evans regards as absurd the supposition that the mere fact that an object is thought about in a particular way should render relation to the object indirect. This absurdity is described as being on a par with supposing that if I gave you something in a particular way (with my left hand instead of the right) the giving would be rendered indirect.

Second, Evans diagnoses Kaplan's position as perhaps implicitly assuming that Fregean senses are basically Carnapian intensions — functions form possible worlds to extensions. Of course, if this is so, then such terms cannot be REs, because they will, by definition, have different extensions in different worlds. But as the case of descriptive names shows, even when a name has a clear descriptive reference-fixing component, this need not allow it to behave in this way. It can still be a rigid designator. (Consider: Julius might not have invented the zip (had he died in his youth, for example).)

[End of Guide to Chapter Two]

Copyright holder grants permission to distribute, quote, and reproduce for any non-commercial purposes provided authorship and copyright are clearly indicated.

Chapter 2: Russell

Table of Contents: