philosophy is for everyone
and not just philosophers

philosophers should know lots
of things besides philosophy

Philosophical Connections

Electronic Philosopher

Feature Articles

University of London BA

Philosophy Lovers Gallery

PhiloSophos Home

International Society for Philosophers

On Reference Failure

by Han Xiaoqiang

One of the things and perhaps the most important thing we do with language is to use it to talk about the world. We can do so because some words we use are able to be used in such a way that they, as Marga Reimer picturesquely describes, somehow 'hook on to things in the world' or 'attach to bits of reality'.[1] Proper names such as 'Marcus Tullius Cicero' and 'Barack Obama' are such words; they are often believed to be paradigmatic referring expressions, as they are used to refer to individuals in the world. But other words like 'Pegasus' and 'Zeus' cannot be so used, because there are simply no such things that actually bear the name of 'Pegasus' or 'Zeus'. What 'reference failure' describes is presumably this kind of situation: Expressions such as 'Pegasus' and 'Zeus' fails to refer to anything.

Of course, not all philosophers consider 'Pegasus' and 'Zeus' as cases of reference failure. Some think that such expressions have bearers which really exist, albeit as abstract objects.[2] Others also hold that they have bearers, but deny that the bearers of such expressions exist.[3] Now if reference is understood so broadly, then there is no such thing as reference failure, for any expression with meaning refers, that is, refers to what the expression means. So in order for the idea of reference failure makes any good sense at all, reference has to be defined as only to things in the world. It is precisely in this restricted sense of reference that 'Pegasus' and 'Zeus' may be thought as cases of reference failure, as such words cannot 'hook on to things in the world' or 'attach to bits of reality' in the way 'Marcus Tullius Cicero' and 'Barack Obama' can.

The question is, if 'Pegasus' and 'Zeus' are not referring expressions, how would the use of them cause reference failure in the first place? If an expression is really a referring expression, it cannot fail to refer, as a referring expression is precisely one that refers to something existing. Russell once said, 'If it were really a name the question of existence could not arise, because a name has to got to name something or it is not a name.'[4] But if an expression does not refer to anything, then it is not a referring expression. And a non-referring expression cannot be said to either succeed or fail to refer — 'nothing' or 'but', for example, succeeds or fails to refer to nothing. 'Pegasus' or 'Zeus' can be said to fail to refer, only if it is used, mistakenly of course, as a referring expression.

Now as most of us would agree I think, sentences that contain expressions like 'Pegasus' can be meaningful and sometimes true, as in the case of 'Pegasus does not exist'. Arguably the neatest explanation of the meaningfulness and truth valuability of such sentences is provided by the theory of descriptions: 'Pegasus' is a disguised description and hence replaceable by something like 'the winged horse'. A significance of the descriptive analysis is that it analyses away reference failure. Expressions such as 'Pegasus' or 'Zeus' cause no reference failure, as under this analysis, they have nothing to do with reference. The sentences that contain such expressions express only some general propositions which involve no reference to particular objects.

But one may wonder whether the theory of descriptions can analyse away reference failure in all its alleged instances or whether it can give a descriptive treatment to all cases that are otherwise considered reference failure. Whatever is said of ordinary proper names, it is generally accepted that indexicals, among which are demonstratives, refer directly, that is, they are not replaceable by descriptions. If that is the case, it would appear that reference failure involving the use of indexicals cannot be analysed away. In his 'Demonstratives', David Kaplan mentions three cases of using an empty demonstrative: (1) hallucination; (2) wrong demonstratum, which is, for instance, when the subject is pointing to a flower and saying 'he' in the belief that one is pointing to a man disguised as a flower; (3) too many demonstrata, as in the case where the subject is pointing at two intertwined vines and saying 'that vine.'[5]

As I wish to show, none of these is in fact reference failure. Let me first deal with (2) and (3). In (2), since 'he' has some descriptive content (indicating the object being human and male), the subject fails not in referring, but in describing or predicating. Calling a flower 'he' is simply another way of saying 'that man' or 'that which is a man'. The subject fails to correctly predicate of the object she is pointing at, but she does not fail to refer to it, inasmuch as the 'he' she uses does single out an object, a flower, which she merely wrongly thinks of as a man disguised as a flower. In (3), as long as the subject knows which of the two intertwined vines she intends to refer to, there is no reference failure either. It is really a matter of whether the audience can successfully identify what the speaker has in mind,[6] which can be solved by imposing a 'division of reference' on 'that', such as 'that dark vine' in case one of them is darker than the other.

The case of hallucination is quite different, as unlike (2) and (3), there is no object in the first place for the subject to describe or predicate of correctly or incorrectly. When Macbeth utters, 'Is this a dagger I see before me?', the demonstrative 'this' he uses to refer fails to pick out anything — he does not merely mistaken something which is really there for something else. Now according to Russell, demonstratives such as 'this' and 'that' are logically proper names and not disguised descriptions; they are therefore not subject to descriptive analysis. It would appear that the reference failure resulted from the use of it in the case of hallucination cannot be analysed away.

It must be noted that closely related to Russell's thesis of the descriptive non-analysability of demonstratives is the idea that the only genuine referents of demonstratives like 'this' and 'that' are sense data, and not ordinary physical objects, as the principle of acquaintance which is fundamental to his notion of reference dictates that apart from universals, the only objects that we can be acquainted with or directly aware of are sense data, and never ordinary physical objects, which are logical constructions out of sense data. This is why he could maintain that for a logically proper name such as 'this' or 'that', the question of existence does not arise. That is to say, the use of 'this' guarantees the existence of its referent, and therefore cannot possibly cause reference failure, precisely because 'this' can only refer to a sense datum — one cannot be mistaken about the existence of some sense datum, as that a sense datum appears to exist is no different from that it exists.

However when Macbeth utters the words, 'this... dagger', surely he intends 'this' to refer to an object, not a sense datum, for it is precisely his mistaking a sense datum for an object that constitutes the hallucination. But if we accept the sense data theory as providing a good explanation for hallucination[7], the 'this' used in this context must be understood as actually, though unintentionally, either referring to a sense datum, such that what Macbeth really states is 'this represents a real dagger', or as a shorthand for 'the object represented by this', a disguised description, where 'this' refers to a sense datum. Either way, reference failure in hallucination can be analysed away.

It may be said then that what separates empty demonstratives from empty ordinary proper names with regard to the reference failure caused by the use of them is that while sentences containing empty ordinary proper names can be rephrased into sentences that involve no reference whatsoever, but only concern some general existential claims, sentences containing empty demonstratives can still exhibit some genuine reference being made in them, once empty demonstratives are shown to be empty and genuine demonstratives are located. It seems that in any case reference failure can be analysed away, that is, re-construed either as a false existential claim or as a false predication or description.

However, as I wish to suggest in the following, there can be reference failure that cannot in any way be analysed away. Consider a situation in which someone perceives, instead of a particular sense datum, but a rapid succession of many different sense data, each of which lasts for a period of time short enough to disallow an utterance of 'this' to complete. Given the demonstrative nature of the demonstrative 'this', that is, that whatever is referred to by 'this' must exist at the time when 'this' is used,[8] 'this' cannot refer to any sense datum, if no sense datum can stay to be 'captured' by 'this'. Any use of 'this' in such a situation will then always result in a reference failure.

It does not seem possible that this kind of reference failure can be re-construed as a false predication or description, for to re-construe a reference failure involving a demonstrative like 'this' as a false predication or description, one must secure a referent of 'this', initially unintended by the referrer notwithstanding, of which the reference failure is to be re-construed as a false predication or description. However, there is nothing in the rapid succession of sense data that can ever be secured as a referent of 'this'. That is to say, in no way can one make any successful demonstrative reference, as required by the dissolution of a reference failure involving a demonstrative.

Perhaps the only way to prevent the re-emergence of the demonstrative, which always reintroduces a reference failure, is to deny the non-analysability of the demonstratives and to grant that even 'this' is replaceable by a definite description, for example, 'the thing being demonstrated at the time of demonstration'[9], such that the reference failure is to be re-construed, not as a false predication or description, but as a false general existential proposition, in much the same way a reference failure involving a empty ordinary proper name like 'Pegasus' is re-construed as a general existential proposition.

However, 'the thing being demonstrated at the time of demonstration' cannot be accepted as really a definite description, and therefore capable of replacing 'this', for a definite description at least purports to by itself fix the thing it applies to, which that phrase clearly does not. The phrase may better be understood as only expressing what Kaplan calls the character of the demonstrative 'this'. The character of such an expression, according to Kaplan, merely specifies the semantic rule that dictates the correct use of the expression and by itself does not fix the thing the expression can apply to. In order for the expression to apply to anything at all, or in Kaplan's terminology to have a content, it must be supplied with a context. In other words, the referent or the content of an expression like 'this' is determined by the character of the expression together with the context which, however, is no part of the expression.[10]

Now if 'this' is analysed as 'the thing being demonstrated at the time of demonstration', the reference failure caused by a use of it would be re-construed not as a false general existential claim, but only as the denial of a semantic rule, which surely cannot be considered as analysing away the reference failure.

While such a reference failure should be recognized as ineliminable, I don't think that talk of reference failure necessarily requires a sense data theory, or that reference failure occurs only when a demonstrative is used to refer to a sense datum. Reference failure can occur when 'this' is used to refer to physical objects and their qualities. Although the sense data theory is indeed needed, in my view, to account for hallucination, it does not have to be accepted as explanatorily necessary and even useful for understanding the way how language is related to the external world, and more specifically how words 'hook on to things in the world' or 'attach to bits of reality' and fail to do so.

Nevertheless, an analogy can be made between reference failure that occurs when 'this' purports to refer to a sense datum and reference failure that occurs when it purports to refer to a physical object or a quality: The rapid succession of objects or qualities, much as the rapid succession of sense data, makes any successful reference involving demonstratives like 'this' impossible. This understanding of reference failure is by no means original. In fact it was well articulated by Plato to serve his characterization of the phenomenal world. In a number of occasions[11], Plato describes a situation in which demonstratives like 'this' (tode) or 'that' (touto), when used to refer to things in flux, fail to refer. The constant and incessant transformation between the phenomenal stuffs, fire, water, earth and air, makes it impossible to say that any one of them is really one thing (e.g., water) rather than some other. In a well-known passage in the Timaeus, Plato remarks,

Now then, since none of these [fire, water, and etc.] appears ever to remain the same, which one of them can one categorically assert, without engrossment, to be some particular thing, this one, and not something else? One can't. Rather, the safest course by far is to propose that we speak about these things in the following way: what we invariably observe becoming different at different times — fire for example — to characterize that, i.e., fire, not as 'this', but each time as 'what is such', and speak of water not as 'this', but always as 'what is such'. And never to speak of anything else as 'this', as though it has some stability, of all the things at which we point and use the expressions 'that' and 'this' and so think we are designating something. For it gets away without abiding the charge of 'that' and 'this', or any other expression that indicts them of being stable.[12]

If reference is defined as necessarily involving a relation between an expression and some particular thing which exists in time, then no doubt what Plato says here can be understood as arguing that any reference or any use of referring expressions in the context of flux is fundamentally inadequate. According to Plato, in order for something to be referred to by an expression such as 'this' or 'that', it must have some sort of stability. Since nothing in flux is stable, any expression used to refer will always fail.

The sort of stability may be understood as merely a minimal stability required by the use of demonstratives like 'this'. A minimal stability, as Russell once noted, is that anything referred to by 'this' (which for him is a particular sense datum, not a physical object) must last for at least a minute or two, long enough for anyone who uses 'this' to finish talking about it.[13] Quite certainly, the time can be much shorter for a successful reference, as there is no difficulty to demonstratively refer to a flash or a bang that lasts as briefly as only a second or two. But below that minimum, perhaps 'this' or 'that' will no longer be able to perform its role. Demonstrative reference is perception-based, and is therefore temporally constrained. Surely one cannot use 'this' or 'that' to demonstratively refer to something that has already disappeared.[14]

If nothing in the world of flux has a minimal stability or permanence, there is no way language could ever 'hook on to' anything in the world or 'attach to' any bit of reality. What I have been trying to show is that there is a necessary ontological condition for language use, if language is to be used at all to describe the external world. Reference failure, understood in the sense as just described, reflects, not the deficiency on the part of language or language use, as in the case of false existential propositions and that of false description or predication — one cannot blame the lack of application of 'the present king of France' or the emptiness of 'Pegasus' and 'this' used in hallucination on reality for failing to supply the corresponding items. Reference failure reflects, rather, the 'deficiency' on the part of reality, namely, the lack of sufficient stability and permanence.


1. Marga Reimer, 'Reference', The Stanford Encyclopedia of Philosophy, (first published 20 Jan, 2003) internet, April 2nd, 2007, available:

2. See van Inwagen 'Creatures of Fiction,' American Philosophical Quarterly, Vol. 14, No. 4, (1979) 299-308, Zalta Abstract Objects, (Dordrecht: Reidel, 1983) and Salmon 'Nonexistence', Nous 32 (1998) 277-319.

3. Reimer, 'The Problem of Empty Names', Australian Journal of Philosophy 79 (2001): 491-506.

4. Russell, Logic and Knowledge, (London: George Allen and Unwin Ltd, 1956) 243.

5. Kaplan, 'Demonstratives', Themes from Kaplan, eds. Joseph Almog, John Perry and Howard Wettstein (New York and Oxford: Oxford University Press, 1989) 490-491.

6. The reference I am considering here may be what Kent Bach calls 'linguistic or semantic reference', as opposed to 'speaker's reference' (Kent Bach, Thought and Reference, Oxford: Clarendon Press, 1987: 52), although the phrase sounds a bit misleading as if it has nothing to do with the speaker or the speaker's intention, which is not the case. What distinguishes speaker's reference from linguistic reference is that the former does and the latter does not involve the audience. Referring to something, understood in the latter sense, does not require the audience to identify it. When I utter 'Marcus Tullius Cicero was a Roman orator', the name 'Marcus Tullius Cicero' I use refers or is intended to refer to a particular individual in history, regardless whether or not my audience is unable to identify him, and regardless even whether or not I intend them to identify it.

7. The reason that the sense data theory has not been completely outmoded is that it provides a good explanation for hallucination, among other things. However, we need not and should not extend the application of the theory to normal perception, that is, think that what we directly perceive are always sense data, and never physical objects themselves.

8. Pointing at a picture of something in order to refer to it is, strictly speaking, not a demonstration.

9. Russell later came to hold, as did Quine, that even 'this' can be so analysed.

10. This is true of all indexicals, of which demonstratives are a special kind. For Kaplan, while every expression has a character, indexicals differs from non-indexicals in that the character of indexicals is sensitive to the change of context whereas that of non-indexicals is not. Cf. Kaplan, 'Demonstratives', 506.

11. Timaeus 49a6-c7, Cratylus 439d and Theaetetus 182c1-183b5.

12. Timaeus, 49c7-50a4.

13. Russell, Logic and Knowledge, 203.

14. One may point to the fact that at least 'that', which is unlike 'this', can be used to refer to something that was, but is no longer perceived. However this use of 'that' is not a demonstrative use, as quite simply, there is nothing or nothing any more to demonstrate. Perhaps it should be more adequately called a memory-based referring expression, or even a disguised description such as 'the noise I just heard'. In fact, even 'this' can occasionally be so used.


Bach, K. Thought and Reference, Oxford: Clarendon Press, 1987.

Kaplan, 'Demonstratives'. Themes from Kaplan. Eds. Joseph Almog, John Perry and Howard Wettstein. New York and Oxford: Oxford University Press, 1989. 481-563.

Plato. Timaeus, Cratylus and Theaetetus, in Plato's Complete Works. Ed. John M. Cooper, Indianapolis and Cambridge: Hackett, 1997.

Reimer, Marga. 'The Problem of Empty Names'. Australian Journal of Philosophy 79 (2001): 491-506.

______. 'Reference'. The Stanford Encyclopedia of Philosophy. Online. First published 20 Jan, 2003. Internet. April 2nd, 2007. Available:

Russell, Bertrand. Logic and Knowledge. London: George Allen and Unwin Ltd, 1956.

Salmon, Nathan. 'Nonexistence'. Nous 32 (1998): 277-319.

Van Inwagen, P. 'Creatures of Fiction,' American Philosophical Quarterly, Vol. 14, No. 4, pp. 299-308. 1979.

Zalta, E. Abstract Objects, Dordrecht: Reidel. 1983

© Xiaoqiang Han 2008


Department of Philosophy
Queen's University at Kingston