Misusing Language - University of Idaho

Misusing Language
Michael O’Rourke
October 16, 1996
Draft for W.S.U. Talk
Coming in to the middle of Chapter VI of Through the Looking Glass, we find Humpty
Dumpty saying, “When I use a word it means just what I choose it to mean---neither more nor less.”
To this, Alice replies, “The question is whether you can make words mean so many different
things.” Humpty responds, “The question is which is to be master---that’s all.”1
The skepticism in Alice’s reply suggests that she thinks it is language and not Humpty
Dumpty that should be identified as the master in this case. Indeed, she believes Humpty Dumpty
has misused the language when he used ‘glory’ to mean ‘a nice knockdown argument’. Normally
(although not in this case), we are good at divining the meaning of a speaker’s words in cases of
linguistic misuse---of figuring out what went wrong and correcting it so as to get at what the
speaker meant to say. However, we start our search for what went wrong only after we conclude
that something has gone wrong. This is revealing: it suggests that we apply normative standards
when communicating, and that our contributions and our interpretations of other contributions are
conditioned if not determined by what results from that application. That is, we aim to get things
right when we communicate, and to this end we rely on normative standards.
Carroll, Lewis, Through the Looking-Glass (New York: Grosset & Dunlap Publishers,
1969). Ch. VI, p. 216-217.
At this point, it would serve us well to step back and examine what it is we want to do here.
As I see it, the philosophy of language should be out to build a theory of linguistic communication.
This theory should comprise three sub-theories: a theory of linguistic structure (i.e., syntax), a
theory of linguistic meaning (i.e., semantics), and a theory of speaker meaning (i.e., pragmatics).
Before we can begin this project, however, we need to determine what phenomena our theory
should explain, what phenomena it should explain away, and what phenomena it should ignore. In
this essay I’ll focus primarily on the theory of linguistic meaning, and I’ll ask the following
question: should semantics explain, explain away, or ignore a phenomenon that I’ll call misfit
definite descriptions. In considering this question, I’ll focus on normative standards that apply
within semantics. It is important to recognize that there is a price associated with getting the
answer wrong here. If you regard it as something the theory should explain and you are mistaken,
then you will look at data that you shouldn’t and thus treat noise as evidence, a move that will
certainly lead you astray. If you fail to regard it as something the theory should explain and you are
mistaken, then the theory you’ll develop will either be merely partial, or it will be a model of a
special case of a more general phenomenon passed off as the general account.
The phenomenon I wish to focus on here is one that would appear to support Humpty
Dumpty: the case of misfit definite descriptions. Definite descriptions are noun phrases that begin
with the word ‘the’, or can be translated into such a phrase: for example, ‘the first word on this
line’, ‘the speaker’s tie’, etc. Call a definite description a misfit when it is used to mention an item
that it does not describe. The prevailing attitude toward misfit descriptions currently is that since
they are mistakes, they cannot be relevant to semantic theory, i.e., they are not phenomena that the
semantic theory should be developed to explain. These folks acknowledge that one can use misfit
descriptions and successfully communicate, but they argue that this is a pragmatic fact that should
be explained only after we have a semantic theory of definite descriptions in place. Call the
intuition at work here---viz., that semantic models for natural language should be built to explain
language used correctly---the intuition of semantic correctness. After presenting misfit descriptions
in more detail, I develop this intuition with the help of Gareth Evans, who characterizes it in terms
of linguistic convention. I then introduce an example involving a misfit description that would
appear to be semantically correct and use that case to argue for the semantic relevance of misfit
descriptions. In addition to resuscitating misfit descriptions, I hope to provide some considerations
that support Humpty Dumpty’s way of thinking.
A Semantic Role for Misfit Descriptions?
In "Reference and Definite Descriptions", his seminal paper on the semantics of definite
descriptions, Keith Donnellan argues that a definite description can be used either to refer to a
particular item, singling that item out for (possible) identification, or it can be used to mention
whatever item has the property associated with the description. The former is the referential use
and the latter the attributive use. One can say that in the referential use, the intended referent is
essential to the utterance made, whereas in the attributive use, the description is essential.
In distinguishing these two uses, he considers cases involving misfit definite descriptions,
i.e., cases where a description the  is used to mention an item it does not describe. He writes that
in the attributive use, "if nothing is the  then nothing has been said to be ," whereas in the
referential use, "the fact that nothing is the  does not have this consequence."2 Adapting an
example of Donnellan’s, consider the sentence, “The man with the martini is wanted on the phone”
uttered by someone who is under the mistaken impression that there was someone in the room
drinking a martini. If that speaker uttered it without intending to refer to any particular person--that is, if the use was attributive---then there is no one person about whom she is talking; if the
speaker uttered it while intending to refer to someone in particular---that is, if the use was
referential---then even though that person is not drinking a martini, she is nevertheless referring to
him using ‘the man with the martini’ and saying of him that he is wanted on the phone. In general,
then, the description is essential in the attributive use, and so a speaker relies on the property
associated with it to serve up an item for predication; if nothing is the , then no claim is made. In
the referential use, on the other hand, the intended referent is essential, and so a speaker uses the
property merely as a way of getting at that object which she can identify in other ways; in this case,
she makes a claim about that item even if it is not the . Thus, in using misfits to differentiate
attributive and referential uses, Donnellan in effect argues that they are phenomena that must be
explained by semantic theory, and so they are relevant to it.3
The Rejection of Misfit Descriptions
The importance of Donnellan's account of the referential/attributive distinction is
undeniable, but many take exception to his use of misfit definite descriptions in developing it. For
Ibid., p. 234.
Another who makes room in his semantic theory for misfit definite descriptions is
Searle. See “Referential and Attributive,” Monist 62:190-208.
example, in describing Donnellan's reliance on misfits, Michael Lockwood argues that "Donnellan
is here quite unnecessarily riding roughshod over the common-sense distinction between what a
speaker means and what he actually succeeds in saying."4 Echoing this, Howard Wettstein contends
that "Donnellan does himself a disservice in claiming that the referential-attributive distinction can
best be brought out by considering cases in which the description fits nothing."5 Chiming in, Hector
Neri-Castaneda remarks: “The fact at the bottom of all the fuss has nothing to do with definite
descriptions. It is the fact that one can succeed in making a hearer think of something  by means
of expressions that do not in reality as the language goes correspond to .”6
While their specific diagnoses differ, these three, along with other philosophers7 believe that
Donnellan has made a mistake by relying on misfit descriptions in drawing the
referential/attributive distinction. By doing so, he confuses what a speaker can do with the
language with what that language means; that is, he confounds pragmatics with semantics. As a
result, he illicitly introduces pragmatic elements into his account of the semantics of definite
descriptions. His account may teach us something about language users, but nothing about
As these philosophers see it, a speaker can succeed in communicating her intended meaning
with a misfit description because participants in the discourse episode can compensate for the
Lockwood, Michael, “On Predicating Proper Names,” The Philosophical Review
84:471-498, p. 485, n. 21.
Wettstein, Howard, “Demonstrative Reference and Definite Descriptions,” in Has
Semantics Rested on a Mistake? (Stanford: Stanford University Press, 1991), p. 38.
See “On the Philosophical Foundations of the Theory of Communication: Reference,”
in Contemporary Perspectives in the Philosophy of Language, ed. P. French, T. Uehling, and H.
Wettstein (Minneapolis: University of Minnesota Press, 1979), p. 146, n. 7.
mistake the speaker makes by exploiting cues drawn from the context of utterance. In such a case,
the speaker does not intend to express the claim that her words convey. One may use the
description to help get a read on what the speaker might mean, but it is just one cue among many
and it is not privileged: the speaker succeeds in communicating despite her words, not because of
them. Since the words are not being used in a way that is consonant with their meaning, this type of
linguistic datum is dismissed as irrelevant to semantics. Behind this dismissal is an assumption:
semantic theories should be developed to explain sentences that conform to intuitions about the
correct use of our words, given their meanings. If a speaker makes no mistake in selecting a
sentence to convey her intended meaning in a particular case, then the resulting utterance is deemed
relevant to semantics in the sense that its significance should be explained by semantic theory; if
she makes a mistake, then it is rejected as irrelevant. That is, an utterance is semantically relevant
just in case it conforms to our intuitions of semantic correctness (i.e., if it is what I will call
semantically correct).
An analogy might be illuminating here. Say you find yourself in a room full of people
doing sums on a blackboard, but you are unaware of what it is they are doing. (Perhaps they are
using a sort of code that you do not understand.) Assume that you are unable to communicate with
these people verbally. If you were interested in modeling their behavior, you would try to discern
the structure of their writings and then assign meaning to those writings. Given enough time and
energy, this would likely succeed, but consider how difficult it would be to come up with an
adequate model of their behavior if these people were bad at arithmetic. Perhaps you would have
Among “other philosopher” can be counted Saul Kripke, Gareth Evans, David Wiggins,
and George Wilson, to name four.
some intuition about which data are corrupt: you could decide, for example, to admit only those
symbolisms that were produced with full confidence, rejecting all those that were produced with
hesitance and consternation, or were subsequently crossed out or erased. This wouldn’t be perfect,
but it would make the job easier. The situation is analogous in semantics: the goal is to model the
meanings of utterances, and to this end one should eliminate the corrupt data. As with the
arithmetic example, identification of this data is the tricky part. According to the view we’re now
considering, linguistic data that is not semantically correct is deemed corrupt and rejected; on the
other hand, data that conforms to these intuitions is admitted into the data pool.
At this point, we can produce an argument in three steps that expresses the dismissive
attitude of these philosophers toward misfit descriptions. First, an utterance is semantically relevant
(i.e., it qualifies as a datum that must be explained by semantic theory) if and only if it is
semantically correct (i.e., it conforms to our intuitions of semantic correctness). Second, if an
utterance contains a misfit description, then it is not semantically correct. Therefore, if an utterance
contains a misfit description, then it is not semantically relevant. I have examined the first premise
with the help of the analogy above. I now examine the second premise with the help of Gareth
Evans on Linguistic Correctness
In Varieties of Reference, Gareth Evans develops a theory of singular reference that is
grounded in a conception of linguistic understanding constrained in part by intuitions of semantic
correctness. These intuitions typically form a part of the background against which he develops this
theory, but he raises them into theoretical prominence when he turns to the semantics of definite
descriptions. In considering whether there is a semantically legitimate referential use, he lays down
the following condition on cases that can be introduced into evidence in support of such a use: “our
intuitive judgements of the correctness or incorrectness of utterances support the ascription to the
description-containing utterance of the referential, rather than the quantificational (i.e., attributive),
truth-conditions.”8 In searching for cases of this kind, Evans considers utterances that involve
misfit descriptions and concludes that they would not be correct and so would fail to meet this
condition. Of such an utterance, Evans remarks, “Undeniably, a mistake has been made, and the
sentence should not have been uttered.”9 Evans is clearly a member in good standing of the group
of philosophers introduced in the preceding section, and I focus on him primarily because he
supplies a way of making precise these intuitions that misfit descriptions are not semantically
correct; that is, his approach to the semantics of definite descriptions permits us to develop a more
precise version of the second premise of the argument advanced above.
As Evans understands it, these “intuitive judgements of the correctness or incorrectness of
utterances” are grounded in the conventions of language. He contends that "words and modes of
composition of a language are endowed with a significance by the conventions of a language,"
where this means that "the meanings of words are supervenient on social practices."10 He illustrates
his view of linguistic convention with an analogy involving the conventional rules that control a
board game. Linguistic conventions constrain the use of linguist expressions much in the same way
that the rules of a game constrain moves. Consider chess. The pieces have "powers and
Evans, Gareth, Varieties of Reference (Oxford: Oxford University Press, 1982), p. 322.
Ibid., p. 323.
potentialities" that are guaranteed by the rules of the game, and this constrains the way in which
those pieces can be used. Furthermore, when a player makes a move, that move has consequences
within the game that are determined by the rules "quite independently of the purpose [the player]
may have had in making the move."11 If one uses the pieces without regard to the rules or their
consequences, one is simply not playing chess. By analogy, words and expressions have meanings
secured by linguistic conventions that constrain their use, and the use of an expression (typically)
has effects that are independent of the thoughts in the mind of the speaker. If one uses words in a
way that ignores convention, one talks without saying anything.
Evans contends that semantic theory must respect linguistic convention. We can represent
his brand of linguistic conventionalism with the following theses:
The meaning of words and expressions of a language are determined by practices
within the community of people who speak the language. Call this meaning so determined for a
particular expression of the language the semantic potential of that expression.
Speaker intentions do not override convention; that is, one cannot make an
expression mean something that is not contained in the semantic potential of the expression.12
These theses supply the general picture, but more detail must be added to account for expressions
that can be used to refer to individual items, such as proper names, indexicals, pronouns, and
definite descriptions---their semantic potential will be tied to the range of things they refer to. In
Ibid., p. 68.
Ibid., p. 68.
While these theses apply quite generally, they do not tell the whole story. With respect
to certain types of expressions---demonstratives, for instance---meaning may be conventionally
underdetermined; that is, the semantic potential may not provide us with a determinate meaning.
When this is the case, speaker intentions determine which of the alternatives is relevant to
interpretation of the utterance.
Evans’ view, expressions such as these are conventionally associated with referential features. He
The conventions governing referring expressions are such that, as uttered in a context of
utterance, they are associated with a property which an object must satisfy if [that object] is to be
the referent of the fully conventional use of that expression in that context.13
The referential feature of such an expression is a conventionally associated property, which is to say
that the referential feature reflects the way speakers of the language use that expression. Thus,
referential features determine the semantic potential of these expressions, and so we can add a third
thesis to our list, one that accounts for the conventional character of referring expressions:
The semantic potential of a referring expression is determined by its referential
For example, the description, ‘the philosophy speaker’ is associated with a referential feature---the
property of being the philosophy speaker---by linguistic convention, and this constrains what we
can mean when we use it. The range of possible denotations this description could have, given its
referential feature, constitutes its semantic potential.
Taken together, these three theses give us the following picture of what a fully
conventional---that is, a semantically correct---use of a referring expression:
If a use of a referring expression  with the referential feature  is fully conventional (i.e.,
semantically correct) then  is used to refer to an item x that satisfies .
The referential feature of a definite description is its conventionally associated referential feature.
A misfit description is used to talk about an item that does not satisfy its referential feature. As a
result, misfit descriptions are not semantically correct, and so Evans dismisses them as semantically
irrelevant. Thus, Evans gives us a precise way of thinking about the second premise of our
argument: a misfit description is semantically incorrect because it violates linguistic convention,
which is to say that it is used in a way that ignores the conventionally determined significance of the
description in question.
An Ill-Fitting Misfit
I would now like to consider a misfit description that threatens the argument described
above. Imagine a donut shop some Monday morning. In this donut shop there is a cadre of retirees
gathered around a table, all of them regulars. You walk in, grab your cup of coffee and your glazed
twist, and strike up a conversation with these gentlemen, all of whom you know. The conversation
quickly turns to football, and when it does, the gentlemen begin making glowing remarks about the
person they believe to be the starting quarterback for the Seahawks. You divine from their remarks
that they are discussing a player who has recently been benched, but they seem unaware of this.
You are interested in finding out why they hold the views they do, but you are not particularly
interested in passing along this new information because you know that would upset them. So you
say, hoping to provoke a response, “The starting quarterback of the Seahawks isn’t really that
good.” Call this sentence S.
In uttering S, you use the definite description ‘the starting quarterback of the Seahawks’ and
you use it to talk about the player they are discussing, a player who is no longer the starting
quarterback---that is, you use it to refer to a player that is not within the semantic potential of the
description, given its conventionally associated referential feature; thus, S involves the use of a
Ibid., p. 311.
misfit description. But is use of this misfit description a mistake? That is, is it semantically
Intuitively, the answer to this seems to be “no”. There is certainly something amiss here,
and of this you are well aware. Given your superior knowledge of the world, you realize that,
strictly speaking, the description you use cannot be used to talk about the benched quarterback. The
description does not denote the object of their affection, but that does not necessarily imply that S
is semantically incorrect. First, you are correctly oriented to both the world and to the
circumstances surrounding this particular communicative episode, and this enables you to select a
sentence that will work in this context based on its meaning. You are merely reacting to the context
of communication as you find it: relative to the constraints that are operative in this context, S is
the right tool for the job. Second, S is the just the right sort of sentence to use given your goals in
this communicative episode (viz., to avoid correcting the gentlemen and to inquire about their
views of the quarterback) : it will elicit the response you desire, given its meaning. Selection of S
is a way to cancel out their confusion and keep it from undermining your communicative goal. As
we remarked above, misfit descriptions are typically dismissed as mistakes even if they succeed,
since they succeed in spite of their meaning and not because of it; in this case, though, the situation
is very different: you succeed in accomplishing your communicative goal because of the meaning
of the sentence you utter, and this accounts for the intuition that there is no mistake here.
I find these intuitions compelling, but before embracing them let’s consider whether this
example passes muster with the more rigorous account of semantic correctness developed in the
preceding section. Does S meet the three conditions on semantic correctness borrowed from
Evans? Since it is a misfit, the immediate answer has to be no: you do not use the description to
refer to an item that satisfies the referential feature. However, you do use the description in a way
that depends on the referential feature, and you don’t attempt to override linguistic convention in
using it. In this case, you use the description because of the referential feature it has. If the
description were associated with another referential feature, it would not serve your purposes, given
the context of utterance. Furthermore, you do not attempt to override linguistic convention in this
case: your success depends on the ability of your audience to successfully identify the conventional
meaning of the description, viz., its referential feature, and so you use S in complete conformity
with linguistic convention. While it isn’t fully conventional, and so not fully correct from the
semantic point of view, it is certainly partially correct and so should be explained in part by
semantic theory.
Therefore, S qualifies as semantically correct relative to both our intuitive judgments of
correctness and relative to the more precise conditions grounded in linguistic convention. We have
identified a misfit description that is not semantically incorrect, so the argument above is not sound
because premise 2 is false. Furthermore, assuming that premise 1 remains true, this example and
other examples of this type establish the semantic relevance of misfit descriptions.
Should we accept the conclusion of the preceding section, viz., that misfit descriptions are
semantically relevant? At the very least, the argument establishes that not all misfits are created
equal. It is a mistake to sweep all the misfits together into one pile and then slide that pile under the
rug. There are two broad types of discourse situations in which a speaker might produce a misfit
Speaker Confusion : The speaker is confused about the referential feature
associated with the definite description, or she is confused about the world; in either case, she has
false beliefs that incline her to select a definite description that is in conflict with her
communicative goals.
Audience Confusion : The audience is confused about the world but the speaker is
not; in this case, the speaker would select a definite description that can elicit the response she aims
at given the false beliefs of the speaker.
Misfit descriptions produced as a result of speaker confusion will be mistakes that should not
influence semantic theory, but those produced as a result of audience confusion would appear to be
data that should be explained by the semantic theorist.
But the original question remains: assuming my analysis of the example is correct, does it
establish the semantic relevance of misfit descriptions? There are two responses that could be
offered: yes, and no. Let’s take the negative answer first. Someone might respond by insisting that
instead of establishing the semantic relevance of misfit descriptions, the argument in fact
establishes the semantic irrelevance of the intuitions of correctness. According to this view, these
intuitions fail to provide an adequate foundation for semantic theory. Instead of relying on these to
determine what does or does not fall within the province of semantics, one should just draw a neat
line between semantics and pragmatics with the help of truth conditions: semantics is about
determining the truth conditions of sentences in the language, and pragmatics is about aspects of
linguistic significance that are not truth conditional. Thus, the semantic account of S will focus on
its actual truth condition as determined by linguistic convention, which involves a quarterback
different from the one mentioned by the gentlemen and any success enjoyed by utterance of S in
that context will be considered pragmatic in character.
This response is problematic, however, for several reasons. First, a sentence like S can be
associated with different propositions each of which can be called its truth condition.14 In
particular, the sentence can be associated with a general proposition---the quantificational
proposition that gets so much attention in Russell’s work---as well as a singular proposition that
includes the intended referent in this case. Which truth condition is the truth condition that matters
here? The answer to this is crucial, since the first of these is clearly involved in the speaker’s
decision to use this description, and the success of the utterance in this context depends on that truth
condition. Secondly, the audience is confused about the singular proposition associated with S but
they are not confused about the general proposition---that is, they are confused about the world but
not about the conventionally associated referential feature. Finally, you do not flout linguistic
convention here---it is out of respect for linguistic convention that you select the description you do.
Thus, utterance of S is grounded in linguistic convention, which is the foundation insisted upon by
those who advocate this line of response.
Finally, there is the positive response: yes, this case does establish the relevance of misfit
descriptions to semantic theory. Let’s say this is true---so what? Well, acceptance of this would
appear to entail some revision in the semantic approach to definite descriptions. First, it suggests
that we must attend to more than just the proposition expressed by a sentence containing a definite
description if we hope to account for its meaning, since at least two propositions associated with S
Cf. Perry, John, “Individuals in Informational and Intentional Content,” in The Problem
of the Essential Indexical (Oxford: Oxford University Press, 1993), pp. 279-300.
are required to account for the details of the discourse context within which it is uttered, viz., the
singular and general propositions mentioned above. Second, we look to ground our semantic
models in the discourse setting, taking into considerations the beliefs and expectations of the
participants. Third, we must take seriously the intentions with which the participants produce their
utterances. Linguistic convention is important, but not overriding---speaker intentions can at times
be given priority over a full-blown linguistic conventionalism. Thus, the positive response seems to
lead to what you might call a Constrained Humpty Dumptyism: the view that semantic theory
should accommodate certain utterances even when those utterances contain expressions used in
conflict with linguistic convention. Humpty Dumpty went too far, to be sure, but in certain cases it
is we who are the masters and not language.
This view of the relation between linguistic convention and semantic investigation
underpins the dismissive attitude adopted by many toward misfit descriptions. They insist that
semantics is the study of linguistic meaning, which consists in the conventionally supported
associations made with linguistic expressions by those who speak the language. This they contrast
with speaker meaning, which consists in the wide variety of things a speaker might succeed in
communicating with those expressions. Evans agrees with this, and he argues that these can be
distinguished in terms of our intuitions about the correctness of utterances---call this linguistic
correctness. An examination of linguistic meaning should be based on correct utterances, since
incorrect utterances might introduce meaning-related errors into the account. An examination of
speaker meaning, on the other hand, should focus on successful communication, and this will
include cases where the speaker is mistaken about the meaning of the expressions used. Misfit
descriptions are incorrect by definition, and so they can form no part of semantics, although they
will figure into a pragmatic account of speaker meaning.
Evans believes that linguistically correct utterances provide the foundation for semantics
understood as the study of the conventional meanings associated with linguistic expressions. An
utterance is linguistically correct on his view if the speaker makes no mistakes in uttering it.
Mistakes that render utterances incorrect come in two types: (1) violation of linguistic convention,
and (2) using an expressionAn utterance is correct if the speaker uses it in a way that is consistent
with the linguistic conventions associated with the sentence used; that is, the utterance is correct if
the semantic potential of the sentence contains the intended meaning.
In dismissing misfit descriptions from semantics, these philosophers are in agreement:
misfit descriptions are mistakes and so should not provide the foundation for a semantic
This reaction to Donnellan's use of misfits stems from the belief that if one is to account for
the semantics of natural language, one must focus on linguistic convention, or on the patterns of
common knowledge and expectation that account for the association of meanings with strings of
symbols. In particular, definite descriptions are associated with their properties by linguistic
convention: for example, I can use the description "the antlers on my desk" as I do because I know
that my listener will associate the property of being the antlers on my desk with it, because he will
know that I know this, and so on.15 As we have seen, though, a description is a misfit if it is used to
talk about an item that doesn't have the property conventionally associated with it. In focusing on
misfits, Donnellan and his ilk have turned away from linguistic convention and therefore fail to
develop a semantic account of definite descriptions. In the next section, I consider a particular
development of this view in the work of Gareth Evans.
See David Lewis, Convention (Cambridge, Mass.: Harvard University Press, 1969) for a
detailed development of this view, which is grounded in the work of Paul Grice.