A Gricean theory of malaprops
Elmar Unnsteinsson
Preprint. Please cite published version:
Mind & Language (2017) 32(4):446–462
doi:10.1111/mila.12149
Abstract
Gricean intentionalists hold that what a speaker says and means by a linguistic
utterance is determined by the speaker’s communicative intention. On this view,
one cannot really say anything without meaning it as well. Conventionalists
argue, however, that malapropisms provide powerful counterexamples to this
claim. I present two arguments against the conventionalist and sketch a new
Gricean theory of speech errors, called the misarticulation theory. On this view
malapropisms are understood as a special case of mispronunciation. I argue that
the Gricean theory is supported by empirical work in phonetics and phonology
and, also, that conventionalism inevitably fails to do this work justice. I conclude,
from this, that the conventionalist fails to show that malapropisms constitute a
counterexample to a Gricean theory.
1 Introduction
What should Gricean intentionalists say about malapropisms or speech errors more
generally? Here, I am concerned with a group of ‘Griceans’ who share at least one
core commitment.
No say ing wi thout mean ing
If speaker S says that p by uttering sentence
σ
in language L on some occasion,
∗
(B) elmar.geir@gmail.com
∗
First, I want to thank Michael Devitt, Stephen Neale and Stephen Schier, who got me thinking
about these issues in a seminar at NYU in spring 2013. The paper has also benetted from discus-
sion, suggestions, and comments from Michael Devitt, Daniel Harris, Hrafn Ásgeirsson, Aðalsteinn
Hákonarson, Matt Moss, Thomas Hodgson, Deirdre Wilson, Robert Stainton, and Nanna Teitsdóttir.
Finally, many thanks to the reviewers for this journal for all their helpful and incisive comments.
Malapropism: Saying without meaning? 2
(i) p must be constitutively determined by S’s speaker meaning on that occasion
and, (ii) p must be compatible with the abstract linguistic meaning of σ in L.
No saying is not endorsed by all purported Griceans and is commonly rejected on
the basis of speech error data. Philosophers have tried to undermine the principle in
all sorts of other ways as well.
1
I believe No saying is true and that showing this is quite important for devel-
oping a viable theory of meaning and communication.
2
Griceans quibble about the
proper analysis of speaker meaning but that issue can be set aside here. My argu-
ments are compatible with any of the most plausible theories, all of them analyzing
speaker meaning in terms of some notion of the speaker’s communicative intention.
In what follows, however, I am only concerned with one immediate but controversial
implication of No saying, namely that speakers must always mean what they say.
I begin by explaining why malapropisms are thought to count against Griceanism
as dened by No saying. The most fully worked out argument of this kind is due
to Marga Reimer. She argues that her own ‘conventionalist’ theory of what is said by
malapropisms gives the best explanation of the data. Then I present my rst objec-
tion to conventionalism; the argument from underspecication. In the next section,
I present my own theory, or sketch of a theory, called the ‘misarticulation’ theory
of malapropisms, and show how it is supported by work done on speech errors by
phoneticians. Finally, this makes it possible to state another objection to the con-
ventionalist thesis; the argument from arbitrariness. The major conclusion is that,
contrary to common opinion, the conventionalist cannot use malapropisms to un-
dermine broadly Gricean accounts of what speakers say and mean, based on the No
saying without meaning principle.
2 Malapropism: Saying without meaning?
It is dicult to do justice to the variety of slips and verbal blunders falling under the
heading of ‘malaprop,’ so, to x ideas, I distinguish three kinds. Then I explain how
1
See, e.g., Saul (2002, 2007). Elsewhere, I focus on counterexamples to No saying where the
speaker has a conicting referential intention because of false identity beliefs—due to,e.g., Kaplan (1978),
Kripke (1977), Reimer (1992)—but these need to be distinguished from malapropisms (forthcoming).
See also my article ‘Confusion is corruptive belief in false identity’ (2016a). On my view, however, both
should be categorized as a kind of pragmatic performance error. Of course, this would constitute a
substantial revision of the traditional Chomskyan notion of ‘performance.’
2
See, e.g., Bach (1987); Donnellan (2012); Neale (2005, pp. 181–182); Strawson (1974, p. 52);
Unnsteinsson (2014, 2016b).
Malapropism: Saying without meaning? 3
these examples are thought to give rise to an objection to Griceanism. Consider the
following cases.
1.
Incidental malaprop: When John Kerry was presidential candidate for Democrats
in 2004, running against George W. Bush, he slipped while giving a speech and
uttered ‘wasabi’ instead of ‘Wahhabi.’ Kerry clearly intended to refer to a Muslim
fundamentalist sect called ‘Wahhabi’ but, due to fatigue or whatever, the similar-
sounding ‘wasabi’ stumbled through.
2.
Persistent malaprop: Reimer (2004) describes a colleague who persistently ut-
tered ‘obtuse’ when he clearly meant that something was abstruse.
3.
Intentional malaprop. Davidson (1986) cites an example where the speaker in-
tentionally utters ‘bae of wits’ instead of ‘battle of wits’ for comic eect.
I dene ‘malapropism,’ roughly, as an utterance where some target expression is re-
placed by a dierent expression that is similar in pronunciation. Only unintended
replacements, as in (1) and (2), count as ‘errors.’ Any of these three examples, however,
can then be used to make a plausible argument, along the following lines.
For the rst two types there is some proposition p such that the speaker actu-
ally said and asserted that p, without meaning that p. In the rst example, Kerry said
something about wasabi without meaning anything about wasabi. In the second the
colleague said someone’s writing was obtuse without meaning that anyone’s writing
was obtuse. The third type of case is somewhat more complicated, but here is how
the argument could go. In intentionally uttering a sentence containing the expres-
sion ‘bae’ the speaker said something involving a bae without meaning anything
about a bae, because there is no such thing. The speaker may have meant, however,
something involving a baing battle.
All three examples thus appear to support the idea that, in cases of malapropism,
speakers can say something without actually meaning it. For now, I focus only on the
rst two types, leaving intentional malaprops until the very end. In fact, on Reimer’s
(ibid., 325) conventionalist account, only the rst two types will provide counterex-
amples to No saying, as she believes that the speaker does say and mean the same
thing in the third kind of case.
This leaves us with a question: Why should we accept the description of the rst
two cases as involving saying without meaning? Citing the ‘obtuse’-example, Reimer
(ibid., 322) writes that “[t]he speaker of a malaprop, upon being informed of the fact
that his use was non-standard, would likely agree that what he actually said was
dierent from what he intended to say.” A few sentences later, she gives a fairly direct
answer to my question.
Malapropism: Saying without meaning? 4
[I]f we are going to develop a philosophically sound notion of saying, it
would presumably be best to build it upon a pre-theoretical notion that is
sensitive to a distinction that has clear explanatory value: the distinction
between saying and (speaker) meaning. This is a distinction that allows us
to explain (inter alia) the coherence of claiming that one doesn’t always
mean what one says.
According to Reimer, then, we should posit a distinction between saying and meaning
to explain why it is coherent for speakers to describe malapropisms in terms of the
distinction.
3
But how is what is said determined, then, if not by way of communicative
intentions? On Reimer’s view, it is determined by linguistic convention; in making the
malaprop the speaker simply says and asserts what the words ‘conventionally mean’
in the language in question. Without assuming a full-blown theory of conventionality,
Reimer takes this to imply,at a minimum, that by engaging in the rule governed activity
of speaking English, the speaker tacitly agrees to have their utterances interpreted in
accordance with the conventions of the language. As they say, speaking a language is
like playing a game. It follows, then, that Reimer’s colleague asserted that someone’s
writing was obtuse while meaning that it was abstruse.
More specically, Reimer—and those who share her general outlook
4
—speaks of
‘contextually relativized conventional meaning’ in order to “accommodate indexicality
and ambiguity” (ibid., 333n2). Assuming that this is clear enough let us state the
Reimerian thesis as follows:
Conventionalist theory of malapropisms
When speaker S utters malaprop-sentence
µ
, the content of what S says is de-
termined by the contextually relativized conventional meaning of
µ
, which is,
say, the proposition that p.
Additionally, S may mean and implicate all sorts of things other than what S strictly
says by uttering
µ
. Further, at least when dealing with unintended malaprops, that p is
no part of what S means or intends. A malaprop-sentence is any sentence containing
an expression that ‘replaced’—in a sense to be made precise below—another similar-
sounding expression in the speaker’s utterance.
3
Michael Devitt (2013: 88) also describes speech errors, spoonerisms in particular, as cases of
unintentionally saying one thing and meaning another.
4
Searle (1969: ch. 2) is a classic statement of this type of conventionalism.
The argument from undersp ecication 5
3 The argument from underspecication
Now I will argue that the conventionalist fails to correctly describe the cases at hand
because the conventionally encoded meaning of a linguistic expression—even if rela-
tivized to context—always underspecies what is said by the speaker in uttering that
expression on a given occasion. This idea is not new and was spelled out in some
detail by Dan Sperber and Deirdre Wilson in Relevance (1986/1995) and other work.
5
Gricean theorists propose to bridge this gap with communicative intentions, so that
successful communication is always explained by means of the hearer’s capacities for
intention-recognition and mindreading.
6
In light of this tradition in philosophy of language, let’s take a closer look at so-
called ‘incidental’ malaprops. Assume Kerry uttered,
(1) Wasabi is a dangerous sect,
but his plan was to utter,
(2) Wahhabi is a dangerous sect.
Now, what exactly, according to the conventionalist, did Kerry say by uttering (1)?
What is the contextually relativized conventional meaning of (1)? The simple disquo-
tational answer would be that Kerry said (3).
(3) that wasabi is a dangerous sect.
But this is far from clear. ‘Wasabi’ is clearly polysemous or ambiguous.
7
It can refer to
(i) a condiment popular on sushi, or (ii) a plant of the Brassicacae family, from which
the condiment is produced. The conventionalist has three options, it seems: either
Kerry said something about (i), or (ii), or (iii) the meaning is somehow indeterminate
between the two. But all three options are problematic.
5
Cf. Wilson & Sperber (2012). I argue for this view of underspecication in my (2014), but related
ideas are to be found in, e.g., Bach (1994a, 1994b); Bezuidenhout (2002); Carston (2002); Neale (2016);
Perry (1986, 2012); Recanati (2010); Reimer (2002); Searle (1978); Travis (1989, 1997); Weiskopf (2007);
Wettstein (1984).
6
To be clear, the notion of what is said at play in No saying is a notion that incorporates illocu-
tionary force. This is as it should be, as competent hearers must recognize this part of the speaker’s
intention if communication is to succeed. Dierent theorists have of course proposed dierent ways of
cutting the saying-pie. For example, Bach (1994a) denes a notion of a proposition type relativized to
narrow context, which explicitly excludes creatures of illocution. This stipulated notion is not at issue
in this paper.
7
The dierence between polysemy and ambiguity seems to be mostly a matter of degree. See
Sennet (2011) for discussion.