A Griceian Pun

Speranza

Commentary on A. Voltolini, "Puns for Contextualists"


Grice loved a pun. "Grice to the Mill" was perhaps his favourite. But he possibly couldn't deliver too many sentences in a whole discourse without relying on puns.

"He was caught in the grip of a vyse" was another of his favourites.

He understood puns perfectly, "He fell upon his sword" is a great pun he explores in the seldom quoted, "Actions and Events", Pacific Philosophical Quarterly.

Unhappy with Locke's parots which are almost rational, Grice coins the "intelligent, almost rational pirots".

And so on.

Kantotle, and Ariskant, featured as his best philosophical friends.

Language was too boring for Grice without the ability to pun about it.

"Don't strawson if you can grice". That would be disgriceful.

And so on.

In this essay, Voltolini first tries to provide a new argument in favour of the
contextualist position on the semantics/pragmatics divide. I will argue
that many puns, notably multi-stable ones, cannot be dealt with in the
non-contextualist way, i.e., as displaying a phenomenon that effectively
involves wide context, the concrete situation of discourse, yet only in a
pre-, or at least inter-, semantic sense. For, insofar as they involve
ambiguous utterances rather than ambiguous sentences, these puns
show that the wide context affecting them has a semantic role: it
provides many truth-conditions for a single utterance. Moreover, I will
try to show that the contextualist can provide a unitary account of the
general phenomenon of puns. On the one hand, this account explains
multi-stable puns as well as those puns the non-contextualist claims to
deal with successfully, i.e., the ones involving a speaker-induced
removal of a well-grounded misunderstanding. On the other hand, it
also explains zeugmatic puns, i.e., those involving an ‘impossible’
meaning.
Keywords: Contextualism, non-contextualism, wide context, puns,
multi-stability.


In this paper, Voltolini first tries to provide a new argument in favour of the
contextualist position on the semantics/pragmatics divide. The argument is
based on an evaluation of the phenomenon of ambiguity as it occurs in puns. I
will argue that many puns, notably multi-stable ones, cannot be dealt with in
the non-contextualist way, i.e., as displaying a phenomenon that effectively
involves wide context, the concrete situation of discourse, yet only in a pre-, or
at least inter-, semantic sense. For, insofar as they involve ambiguous
utterances rather than ambiguous sentences, these puns show that the wide
context affecting them has a semantic role: it provides many truth-conditions
for a single utterance. Moreover, I will try to show not only that noncontextualist
replies to the contextualist account of this phenomenon of
ambiguous utterances do not work, but also that the contextualist can provide a
unitary account of the general phenomenon of puns. On the one hand, this
account explains multi-stable puns as well as those puns the non-contextualist
claims to deal with successfully, i.e., the ones involving a speaker-induced
removal of a well-grounded misunderstanding. On the other hand, it also
explains zeugmatic puns, i.e., those involving an ‘impossible’ meaning.

1. The contextualist/non-contextualist opposition
and the role of ambiguity

As regards the divide between semantics and pragmatics, nowadays two main
options confront each other: non-contextualism and contextualism. In point of
fact, contextualism varies in type.1 Yet the opposite position lacks even a
common label.2 Nevertheless, there is a clear-cut way to cash out the main
distinction affecting these two positions à propos of the above divide. This
distinction concerns the different conceptions those options have with respect
to the so-called wide context, i.e., the concrete situation of discourse. To put it
in Perry’s (1997) terms, according to non-contextualists, wide context never
has a semantic role, that is, it never fixes the truth-conditions a sentence has in

1 Not only truth-conditional pragmatics à la Recanati (2004a) - according to which primary pragmatic processes determining contextual truth-conditional contributions of sub-sentential expressions may
even bypass the linguistic meaning of such expressions – but also broad indexicalism à la Stanley
(2007) - according to which linguistic meaning merely underdetermines contextual truth-conditions –
turn out to be forms of contextualism, a radical and a moderate one respectively. For in broad
indexicalism, linguistic meaning plus narrow context may well be unable to determine the truthconditions
of an utterance, i.e., of a sentence in wide context (for the notions of narrow and wide
context see later).
(2 To outline this position, some use the label semantic minimalism (cf. Cappellen & Lepore 2005) –
or the similar one minimal semantics (cf. Borg 2004). Predelli speaks of the “traditional approach”
(2005, p. 13). By following a proposal put forward by Recanati (2004a), one might name this position
literalism. Yet as Recanati himself admits, since according to this position what is semantically
evaluated are sentences in context, one might well be justified in taking this position as a form of
narrow indexicalism, according to which linguistic meaning automatically determines truth-conditions
given a narrow kind of context (see immediately later in the text).

narrow context, or index, i.e., a set of a limited series of parameters – typically,
agent, space, time, and world.3 Typically, for non-contextualists wide context
has the post-semantic, genuinely pragmatic, role of determining whatever
extra-truthconditional factor of significance is conveyed by means of such a
sentence in context. By a first approximation, for non-contextualists the proper
objects of semantic evaluation are therefore sentence - narrow context pairs,
theoretical representations of utterances, the concrete wide-contextual
tokenings of sentences on the speakers’ part.4 Whereas according to
contextualists, wide context may well have the semantic role in question.
Properly speaking, in fact, in order for the contextualist to win the battle
against her enemy, it is enough for her to show that contextualism is correct
with respect to the purportedly semantic role of wide context in at least one
case.5 As a result, for contextualists the proper objects of semantic evaluation
are utterances themselves, qua (roughly) sentences in wide context.6
Here is why the contextualist/non-contextualist distinction must be drawn
this way. On the non-contextualist side, since wide context never has a
semantic role, truth-conditions are assigned to sentences in narrow context by
computing the truth-conditional contributions (in narrow context) of the subsentential
components of such sentences, together with certain syntactic and
compositional principles. Since such a computation turns out to be an utterly
automatic process, even a semantic mechanism, something that is programmed
to make that computation, may perform a truth-conditional assignment.
In this respect, the fact that what is semantically evaluated is a sentence in
narrow context raises no problem. For the relevant truth-conditional
computation remains entirely automatic. This is paradigmatically shown by
indexical sentences, for which narrow context effectively is truth-conditionally
relevant. Granted, without a narrow context an indexical sentence has no truth-
(3 For this distinction between wide and narrow context cf. Perry (1997, p. 595). Predelli (1998,
2005) has shown that narrow context has to be considered a context of interpretation rather than a
context of utterance. For the narrow context’s parameters can well differ from the parameters of a
context of utterance. For a general reason as to why this must be the case, cf. Voltolini (2006)).

(4 More precisely, for non-contextualists the objects of semantic evaluations have to be clause (a
linguistic representation of a natural language sentence) – index (narrow context) pairs. See slightly
later in the text).

(5 As Recanati (2004a, p.116) claims).

(6 Kaplan (1989b, pp. 584-585) argues that sentence – (narrow) context pairs are the right objects of
semantic evaluation, for sometimes there are non-uttered sentences yet true in a context. However,
contextualists simply put this argument aside. For according to them language use, hence utterances,
is the starting point of semantic evaluation. Cf. e.g. Recanati (2004a)).


conditions. Yet once one such context is provided, that sentence again obtains
a certain truth-condition automatically.

For that sentence also has a linguistic
meaning, or better a character , which is completely determined by the
linguistic meanings, or better the characters, of its sub-sentential elements.
Character is a function mapping narrow contexts onto truth-conditional
contributions. When one such sub-sentential element is indexical, its character
specifies which parameter in narrow context has to be mobilized in order to
give that indexical a certain truth-conditional contribution in a given narrow
context. Thus, that parameter automatically determines the truth-conditional
contribution that sub-sentential element gives in that context to the indexical
sentence in which it figures.7
Consider e.g.:

1. I am an American.


Given its linguistic meaning,

(1)

conveys that whoever is the (contextual)
agent is an American.

Yet since (1) is an indexical sentence for it contains the
indexical “I”, not the sentence itself, but that sentence in a certain narrow
context, has truth-conditions.

Now, the character of “I” constituting the
sentence’s linguistic meaning is a function mapping narrow contexts onto
referents, i.e., the truth-conditional contributions of “I” in those contexts.

As
is shown by its linguistic description, roughly “the agent in context”, such a
character specifies the ‘agent’-parameter in narrow context so as to
automatically obtain the agent in one such context as the referent of “I” in that
context, i.e., as its truth-conditional contribution in that context.

As a result,
once (1) is taken along with a narrow context whose agent is Barack Obama,
then (1) is eo ipso true in that context iff Obama is an American.

Something that
even a mechanism may compute.


Now, contextualists reject this whole picture by appealing to the idea that
language functioning is a matter of intentional use, not a matter of automatic
applications of semantic operations. For them, therefore, there is no need to
represent utterances by means of linguistic representations of any kind. Since
utterances are the speakers’ concrete productions, they can be semantically

(7 For this whole picture on indexicals, cf. Kaplan (1989a,b). Recanati (2004b, p. 1) claims that the
character of an indexical is given by a token-reflexive rule. Since for each indexical its own tokenreflexive
rule specifies the narrow context parameter that has to be mobilized in order to provide the
relevant indexical sentence its truth-conditions in a given narrow context, truth-conditional
assignments for such sentences remain automatic).

evaluated in the concrete situations where they are uttered, i.e., in their wide
contexts.
As is well known, it is rather controversial nowadays whether by means of
the above account the non-contextualist manages to deal with all cases of
context-sensitivity, especially those involving so-called hidden contextsensitivity,
i.e., the contextuality that is not sententially articulated by means of
the occurrence in a sentence of a context-sensitive expression.8 Yet I want to
stress here that even if the non-contextualist managed to deal with hidden
context-sensitivity,9 the contextualist should not worry, since there may turn
out to be a very basic semantic phenomenon that preliminarily hinders any
non-contextualist account.

Ambiguity seems to be one such phenomenon, according to which one and
the same sentence of a natural language has different meanings, either because
at least one of its terms has different meanings – lexical ambiguity – or because
the sentence has different syntactic structures – structural ambiguity. At first
glance, ambiguity indeed brings grist to the contextualist’s mill. For the
contextualist may well point out that disambiguation is a matter of wide
context: a factor in the concrete situation of discourse helps one to settle
whether the ambiguous sentence is used with one or another meaning. Yet the
fact that a sentence is ambiguous is none other but the fact that it has different
truth-conditions. Thus, proceeds the contextualist’s argument, assigning to an
utterance of that sentence certain truth-conditions rather than other ones is a
matter of wide context. But this shows that in the case of ambiguity, it is
precisely wide context that has a semantic role.

More formally:

1. That a sentence is ambiguous is wide-contextually dependent

2. That a sentence is ambiguous = that a sentence has many truth-conditions

3. That a sentence has many truth-conditions is wide-contextually dependent

[from 1,2]

4. Hence, wide context has a semantic role.

Yet the non-contextualist will immediately reply to the contextualist, you have
run too fast. One may well accept the wide-contextual dependence of
ambiguity. Yet this wide-contextual dependence merely shows that neither

(8 For this label cf. Borg (2004). Sentence (9) below presents a case of hidden context-sensitivity).

(9 I myself, for one, do not believe this: cf. Voltolini (2009). Yet for many attempts at non-contextually
dealing with overall context-sensitivity, cf. e.g. the afore-mentioned Borg (2004), Cappellen-Lepore
(2005), Predelli (2005).


natural language sentences nor their utterances are the objects of semantics.
For wide context is certainly appealed to in disambiguation, but only in a presemantic
role (to get back again to Perry’s 1997 terminology).

Wide context
indeed serves just in order to select the real object of semantic evaluation,
namely an underlying linguistic representation – e.g., a Mentalese sentence10 –
that automatically has its own truth-conditions, as the non-contextualist holds.
Such a representation will represent an utterance of a certain natural language
sentence. Theoretically speaking, one and the same natural language
ambiguous sentence can be paired with different underlying linguistic
representations, each endowed with its own truth-conditions.11 Yet once one
considers the wide context in which such a sentence is uttered, this context
simply settles which underlying linguistic representation with its own truthconditions
has to be selected by the semantic mechanism as representing the
relevant utterance of that sentence, i.e., the utterance of that sentence which is
uttered in that wide context.

To be sure, for a more precise picture that accounts for indexicality as well,
the non-contextualist must reintroduce narrow context and say that the real
objects of semantic evaluations are those underlying linguistic representations
plus narrow context.

On behalf of the non-contextualist, Predelli (2005) has
formulated this idea in the most articulated way. First of all, for him one and the
same natural language sentence is to be associated, in accordance with its
being uttered in different wide contexts, with different syntactically
transparent12 linguistic representations. He calls one such representation a
clause. Moreover, he says, in order to account for indexicality, too, what really
has to be associated with one such sentence are different clause-index pairs,
namely, couples made by a linguistic representation plus a certain narrow
context. From a theoretical point of view, one such pair indeed represents an
utterance of that sentence, i.e., (roughly) such a sentence in a certain wide
context. Now, for Predelli that pair is the real object of semantic evaluation.
For that pair is automatically assigned certain truth-conditions, in conformity
with non-contextualist desiderata about truth-conditional computation.

(10 For this idea, cf. e.g. Pinker (1994, p. 79)).

(11 Since ambiguity is both lexical and structural, this formulation must be taken with a grain of salt. It
would be more precise to say that what is paired with different underlying linguistic representations
each one endowed with its own truth-conditions is a sentence as syntactically opaque. On their turn,
such representations are instead syntactically transparent for their primary task is precisely that of
presenting the different syntactic structures that sentence possesses).
(12. See the previous footnote).


As a result of this picture, ambiguity amounts to the fact that different
clause-index pairs respectively representing different utterances of one and the
same natural language sentence – i.e., respectively representing that sentence
as taken in different wide contexts – respectively have different truthconditions
(automatically as it were). Thus, wide context plays a mere presemantic
role. For it merely points out which pair has to be mobilized in order
to represent the relevant utterance of that sentence.13

Consider e.g. the following both indexical and ambiguous sentence:

2. I am taking a babe to the bank.

This sentence is indexical for it contains the indexical “I” and is ambiguous
because it contains the ambiguous term “bank”.

Given such an ambiguity, in
accordance with its being uttered in different wide contexts, (2) is respectively
associated with two clauses, clause c containing a term such as “bank1” and
clause c’ containing a different term “bank2”. Consider now a narrow context,
an index i whose agent is the former Italian Prime Minister Berlusconi. I

n
accordance with its being uttered in different wide contexts, (2) is properly
associated with two different clause-index pairs, namely C (c-i) and C’ (c’-i).
Now, C and C’ are the proper objects of semantic evaluation: C is true iff
Berlusconi takes the babe in question to a certain financial institution, while C’
is true iff Berlusconi takes that babe to a certain river edge. In this perspective,
wide context simply selects whether (2) should really associated with C or with
C’, respectively having different truth-conditions. The selected pair will thus
represent a certain utterance of (2), i.e., (roughly) (2) in that wide context. As a
result, pace contextualists wide context has no semantic role.


More formally, a non-contextualist following Predelli rejects the previous
contextualist argument, for she denies premise 2) of that argument. Instead,
she puts forward this different argument:

1) That a sentence is ambiguous is wide-contextually dependent

2*) That a sentence is ambiguous = that a sentence is associated with
different clause-index pairs respectively representing different utterances
of that sentence and endowed with its own truth-conditions

(13 In point of fact, since for Predelli an utterance is roughly a sentence in wide context, it would be
more precise to say that by selecting the right clause-index pair with its own truth-conditions, wide
context makes it clear which utterance of the sentence has been really uttered. Yet for my present
purposes I can leave this aside).


3*) That a sentence is associated with different clause-index pairs
respectively representing different utterances of that sentence and
endowed with its own truth-conditions is wide-contextually dependent
[from 1), 2*)]

4*) Hence, wide context has no semantic, but simply a pre-semantic, role.


To be sure, a non-contextualist acknowledges that the situation at stake
may be more complicated. Consider an utterance of an ambiguous sentence. It
may be the case that, in order to fix the relevant clause-index pair representing
that utterance, wide context operates not at a pre-semantic level, but so to
speak within the semantic level itself. In other terms, it may be the case that
only once that sentence is uttered, does it become wide-contextually clear
which clause-index pair represents the relevant utterance. Thus, the semantic
mechanism must be able to store all the pairs that may theoretically speaking be
associated with that sentence. Yet once the relevant pair is selected, wide
context still plays no semantic role. For that pair is automatically assigned
certain truth-conditions. That pair wears those truth-conditions on its sleeve,
as one might put it.14

2. Why puns really bring grist to the contextualist’s mill

In his perspective, puns seem to cause no particular trouble to the noncontextualist.
Granted, in order for a pun to work, it must exploit natural
language ambiguity in some way or other. Yet the way it works is simply a
speaker-induced removal of misunderstanding that is well-grounded, for it
depends on the fact that the relevant natural language sentence is ambiguous.
In an ordinary well-grounded misunderstanding, an interlocutor erroneously
believes that a given utterance of an ambiguous sentence has a certain truthconditional
interpretation. Such a belief is mistaken, for that utterance actually
has a different truth-conditional interpretation. In this respect, wide context
makes it clear that the truth-conditional interpretation of a certain utterance is
different from the expected one. In the theoretical terms Predelli provides for
the non-contextualist, wide context makes it clear that the utterance
corresponding to that sentence in such a context has to be represented by a
certain clause-index pair endowed with its own truth-conditions rather than by
another clause-index pair with its own truth-conditional interpretation. Once

(14 For these complications cf. Borg (2004, p.140-146)).


the interlocutor realizes that, she may well remove her erroneous conviction
and represent that utterance by the right clause-index pair with its own truthconditional
interpretation. Now, the only difference between an ordinary wellgrounded
misunderstanding and one prompted by a pun is that the latter
misunderstanding is due to intentional factors. A pun’s creator wants her
interlocutor, first, to be led astray by a certain possible yet incorrect truthconditional
interpretation for an utterance of an ambiguous sentence, and
second, to be led back to the correct truth-conditional interpretation of that
utterance, just in order to have fun thanks to the clash of her semantic
expectations.15

In order to see these cases, take first an ordinary case of well-grounded
misunderstanding. Suppose Wim utters:

3. Paris is beautiful.

Typically, her interlocutor will believe that by (3) Wim is talking about the
capital of France. Yet once Wim continues by saying:

4. I like the desert surrounding it

then wide context, specifically the wide-contextual factors of co-text and
background knowledge (a noun such as “desert” and the background
knowledge that the capital of France has no desert), will induce such an
interlocutor to revise her interpretation and suppose that Wim, as he was
indeed doing, was talking instead of Paris, Texas.

In Predelli’s terms, this
means that the above utterance of (3) has to be represented by a certain clauseindex
pair containing the name “Paris2”, which is true iff the city in Texas so
named is beautiful, rather than by another clause-index pair containing the
name “Paris1”, which is true iff the capital of France so named is beautiful, as
the interlocutor erroneously believed.16 Second, consider one of the most
famous puns by Oscar Wilde:

15 In such a case, the comical effect will therefore conform to the so-called ‘incongruity theory’ of
humor. On it cf. Morreal (2009:chap.1).

(16 For the purpose of this example, I take the name “Paris”, hence any sentence containing it, to be
ambiguous. In point of fact, along with some others I believe that proper names are indexicals (cf.
Voltolini 1995). But if I assumed indexicality for proper names, it would make no big difference. For
in Predelli’s terms, a non-contextualist might then account for the misunderstanding by saying that the
utterance of (3) should be represented by another clause-index pair. This pair differs from the pair the
interlocutor erroneously took that utterance to be represented by because its index, not its clause, is
different.

5. To lose one parent, Mr. Worthing, may be regarded as a misfortune; to
lose both looks like carelessness.

The Importance of Being Earnest ,
1895)

In order for such a pun to work, an interlocutor is first erroneously induced
by the presence in (5) of the noun “misfortune”, to think that in the relevant
utterance of (5) the ambiguous “to lose” means to suffer deprivation, so as to
assign that utterance a certain truth-conditional interpretation.

Yet as the
utterance proceeds, a better look at the wide context in which (5) is actually
uttered, more specifically at the fact that the wide-contextual factor of “to
lose”’s cotext in (5) contains not only the noun “misfortune”, but also the
noun “carelessness”, makes it clear that “to lose” here means to misplace, so
that its correct truth-conditional interpretation is another. In this predicament,
moreover, the interlocutor realizes that, by so construing (5), it was Wilde’s
intention to lead his interlocutor into such a misunderstanding. For he wanted
to obtain a comical effect by means of his interlocutor’s removal of the
misunderstanding.

In Predelli’s terms, instead of being erroneously
represented by the clause-index pair containing the term “to lose1”, which is
true iff to suffer deprivation from one parent’s death may be regarded as a
misfortune and to suffer deprivation from both parents’ looks like carelessness,
the relevant utterance of (5) has to be correctly represented by another clauseindex
pair containing the term “to lose2”, which is true iff to misplace one
parent may be regarded as a misfortune and to misplace both parents looks like
carelessness.17

(17 For another example of a pun mobilizing a speaker-induced well-grounded misunderstanding yet
involving structural rather than lexical ambiguity, take the famous Groucho Marx pun “I shot an
elephant in my pyjamas. How he got into my pyjamas I don’t know!” (cf. on this Borg 2004, p.
143fn.91). A similar explanation may be given for puns involving: i) indexicality rather than
ambiguity; ii) sameness of meaning rather than difference of meaning contrary to previous
expectations. Consider the joke Pinker (1994, p.80) presents: “First guy: I didn’t sleep with my wife
before we were married, did you? Second guy: I don’t know. What was her maiden name?” Given the
first sentence, an interlocutor would expect the second guy’s utterance of his second sentence to be
represented by a clause-index pair in which the theoretical counterpart of the pronoun “her” referred
to a certain individual, the second guy’s wife. This individual obviously differs from the individual, the first guy’s wife, the theoretical counterpart of the description “my wife” denotes in the representation of the first guy’s utterance. What’s punny is that, contrary to previous expectations, the clause-index pair that really represents the second guy’s utterance of his second sentence contains a theoretical counterpart of “her” yet referring to the very same individual the theoretical counterpart of “my wife”
denotes in the representation of the first guy’s utterance.

Yet, one may now immediately wonder, is this always the case with puns? As
we have just seen, from the cognitive point of view, the non-contextualist
considers puns as speaker-induced well-grounded misunderstandings, hence
as things falling into the general category of aware misrecognitions. This is to
say, for non-contextualists puns are cognitively on a par with cases of illusory
perceptions in which one discovers that what one illusorily took in her
perception to be a certain thing is in point of fact another. In the most famous
example, Carneades discovered that what he illusorily took in his perception to
be a snake was in point of fact a rope. If this is the case, puns have to be
characterized by non-reversibility of interpretation. As a matter of fact, once
we discover that what we illusorily took in our perception to be a certain thing
is another thing, we can no longer even seem to see that thing as being the
previous one. Once Carneades made the above discovery, the rope no longer
appeared to him to be a snake. By parity of reasoning, in the case of puns once
we discover the truth-conditional interpretation a certain sentence in wide
context, hence a certain utterance, has to be assigned, we can no longer take
that sentence in context, hence that utterance, as having the truth-conditional
interpretation we previously and erroneously gave to it. The utterance remains
linked with what we have discovered to be its correct truth-conditional
interpretation. In Predelli’s terms, we discover once for all that such an
utterance is represented by a certain clause-index pair having a certain truthconditional
interpretation rather than by another clause-index pair having
another truth-conditional interpretation, as we erroneously thought.

Yet cognitively speaking this does not always happen with puns. On the
contrary, for many puns reversibility obtains. In such cases, an utterance first
receives a certain interpretation, then another one, yet once it receives the
latter interpretation the former is still at play. Once we realize that we have to
do with one such pun, we continue going back and forth between the different
interpretations for that utterance, precisely because neither prevails. From the
cognitive point of view, this reversibility is easily accounted for once we find
the right explanation for such puns. As Wittgenstein (1953, 1980, 1982,
1992) originally grasped, the general cognitive category under which such
puns fall is not that of aware misrecognition, but that of seeing-as, at least the
kind of seeing-as characterized by Gestalt switches. In our case, such puns are
indeed characterized by the fact that certain linguistic configurations are
endowed with multi-, or at least bi-, stability of interpretation. Unlike cases of
aware misrecognition, in seeing-as of this kind reversibility indeed occurs. For
instance, one can pass from seeing a certain array of dots as being a twodimensional
cross to seeing them as being a rhombus and vice versa. Likewise,
one can pass from seeing a certain configuration as being (a picture of) a duck
to seeing it as being (a picture of) a rabbit and vice versa , right up to passing
from seeing an Arcimboldo painting as being (a picture of) a face to seeing it as
being (a picture of) a cluster of fruit and vegetables and vice versa. In this
respect, the cognitive upshot of facing an ambiguous figure is a mixture of
astonishment and disorientation. For although in one sense, or better at a
certain experiential level, that of direct visual perception, we are experiencing
one and the same thing, i.e., the item we actually have before us, yet in another
sense or better at another experiential level, that of seeing-as, we have an
alternate visual experience as of different things. Now, in many cases the
inventor of a pun wants to get her audience to be involved in precisely the same
predicament.

That is to say, someone who understands one such pun is in a
certain sense, or better at a certain experiential level, that of direct
auditory/visual perception, still hearing/seeing the very same utterance she
has before her, yet in another sense or better another experiential level, that of
hearing/seeing as, she is having an alternate experience of hearing/seeing that
utterance as having now one meaning, now another one.

This alternation of
meaning experiences, or more radically this ambiguity of the sentence’s
utterance rather than of the sentence itself, is what the inventor of a pun
intends her interlocutor to have.
Consider for instance the following cases:

6. Condoms should be used on every conceivable occasion

7. Without geometry, life is pointless.

8. Santa’s helpers are subordinate Clauses.

18
In all these cases, a sentence is ideally uttered in complete isolation. This may
help the interlocutor not to be led astray by other possible factors of the
specific wide context involved (co-text, gestures, salience effects, etc.) so as to
focus just on the relevant wide-contextual factor, i.e., the utterer’s intention to
convey different meanings by means of the very same utterance . Thus, by

(18 Wittgenstein (1980, I§77) gives another such example. In this example, a singer playing Wagner’sThe Rhine-gold on stage addresses another singer who had just whispered to him something about egg-cooking by singing “Weiche, Wotan, weiche!”. By that utterance, does the first singer mean the same as “Go away, Wotan, Go away!” or the same as “Soft, Wotan, soft”? Both, Wittgenstein would urge)

exploiting the ambiguity of the adjective “conceivable”, the relevant utterance
of (6) is meant to convey two different interpretations, namely not only that
condoms should be used whenever there is a risk of pregnancy, but also that
they should be used in any logically thinkable situation. By exploiting the
ambiguity of the adjective “pointless”, the relevant utterance of (7) is meant to
convey two different interpretations, namely not only that a non-geometrical
life is devoid of geometric points, but also that such a life is meaningless. By
exploiting the ambiguity of the very same plural form, “clauses”, of both the
noun “clause” and the name “Claus”, the relevant utterance of (8) is meant to
convey two different interpretations, namely not only that those who help Santa
are subordinate to him, but also that they are linguistic constructions.19


Now, the contextualist may well rely on such examples of puns in order to
reply to her non-contextualist opponent as follows.

Suppose we buy the noncontextualist
account of ambiguity. As we have seen, in order to deny wide
context a semantic role when ambiguity obtains, the non-contextualist claims
that one and the same natural language sentence in different wide contexts has
to be associated with different clause-index pairs that respectively represent
the relevant different utterances of that sentence and automatically have
different truth-conditions. So each utterance of the sentence, via its
representing pair, has just one truth-conditional interpretation. In such a
predicament, wide context simply selects which clause-index pair with its own
truth-conditions represents the relevant utterance. Yet in the present cases of
puns, we have one and the same utterance of a sentence having different
interpretations, not just one interpretation. As the non-contextualist is instead
supposed to claim, by making one such interpretation the truth-conditional
interpretation of just one clause-index pair representing that utterance. On the
contrary, for the contextualist those different interpretations amount to
different truth-conditions for that very utterance – e.g. the relevant utterance
of (7) for instance is true both iff a non-geometrical life has no geometrical
point and iff that life is meaningless. But these different truth-conditional

(19 In advertisement, many puns work this way. Cf. Tanaka (1992). Here I have chosen cases of puns in which one can say that just one utterance of one and the same natural language sentence is involved,
for the ambiguity on which the multi-stability of the utterance is involved is merely lexical. If the
ambiguity involved were both lexical and structural, as in “My girl criticized my apartment, so I
knocked her flat”, we should more properly speak of a multi-stable utterance of one and the same
syntactically opaque sentence, if not merely of one and the same string of uttered words . Whereas in
case of a pun involving mere homophony, as in “Seven days without laughter makes one weak”, we
should rather speak of a multi-stable utterance of one and the same string of uttered sounds .

interpretations for that utterance depend on wide context; notably, on the fact
that the utterer’s intention is precisely to convey two meanings at one and the
same time by means of the same utterance of a certain sentence. Thus,
contextualism is vindicated. For in the case of our puns, wide context has a
semantic role. More formally, the contextualist first rejects:

2*) That a sentence is ambiguous = that a sentence is associated with
different clause-index pairs respectively representing different utterances
of that sentence and endowed with their own truth-conditions
of the previous Predelli-inspired non-contextualist argument. Then, she puts
forward this new argument:

1’) That an utterance, i.e., (roughly) a sentence in wide context, is
ambiguous is wide-contextually dependent

2’) That an utterance, i.e., (roughly) a sentence in wide context, is
ambiguous = that an utterance, i.e., (roughly) a sentence in wide context,
has many truth-conditions
3’) That an utterance, i.e., (roughly) a sentence in wide context, has many
truth-conditions is wide-contextually dependent [from 1’, 2’]


4) Hence, wide context has a semantic role.20
Now, the claim that a contextualist account fits multi-stable puns nicely can be
further corroborated. Consider that in the case of multi-stable puns, as in all
cases in which a contextualist account suggests itself, not only is it widecontextual
for an utterance to have certain truth-conditions, but it is also widecontextual
which factor of the concrete situation of use is relevant for that
truth-conditional assignment. Contextualism involves meta-contextualism, as
some put it (cf. Bianchi 1999).
In this respect, take an utterance of:

9. Bill cuts the grass.

(20 Let me just stress once more that what the contextualist advocates here is that one and the same
utterance has more than one truth-conditional interpretation, not that it is indeterminate whether one
and the same utterance has a truth-conditional interpretation. Borg (2004, pp. 221-5) claims that
Travis’ version of contextualism, as it emerges e.g. from Travis (1997), amounts to the latter form of
meaning eliminativism. Whether or not Borg’s interpretation of Travis is correct, this form of
eliminativism is not what I am defending here. The point is not that there is no such thing as a truthconditional
meaning for a punny utterance, as Borg might claim in reconstructing Travis’
contextualism. The point is rather that there are as many truth-conditional meanings for that utterance
as its utterer intends it to have in her multi-stable pun)

which presents a typical case for which a contextualist account seems to work.
For a contextualist, that utterance has no truth-conditions in isolation; a
certain wide context, i.e., the concrete situation in which (9) is uttered, fixes
for it a truth-conditional interpretation. Yet according to a contextualist, this
means not only that in order for that utterance to have truth-conditions one has
to appeal to the concrete situation in which (9) is uttered, but also that it
depends on that situation which of its factors is relevant in order to settle which
truth-condition that utterance has. In a certain wide-contextual situation, one
has to appeal to the fact that (9) is uttered within a story one is telling about the
endings of a soccer match, hence to the co-textual factor of that wide context of
(9), in order to settle that that very utterance is true iff Bill rips the grass to
pieces. If instead (9) had been uttered in the course of garden care practice,
one would have had to appeal to this factor of the new wide-contextual situation
in order to settle that this utterance of (9) is true iff Bill trims the grass with a
lawnmower.

Now, the same holds with our multi-stable puns. If somebody uttered (6)
within some anti-AIDS campaign, that utterance would not be punny at all, for
in such an utterance “conceivable” would merely mean logically thinkable . So,
that utterance would be true merely iff condoms should be used in any logically
thinkable situation. Yet if a comedian utters (6), in this new wide-contextual
situation the prevailing wide-contextual factor is the comedian’s intention that
the new utterance of (6) be taken as a multi-stable pun. So, this utterance is
true both iff condoms have to be used in every logically thinkable situation and
iff condoms have to be used just in a pregnancy-inducing situation. Thus, it is
not only wide-contextual for that utterance of (6) to receive a truth-conditional
interpretation, but it is also wide-contextual which factor of the wide context
provides the punny, bi-stable, interpretation: the utterer’s intention.


Up to now, I have first of all presented what I take to be a new interesting
datum: a multi-stable pun counts as an utterance having different
interpretations. Moreover, I have put forward a contextualist account of that
datum: such interpretations for one and the same utterance are different truthconditional
interpretations for that utterance prompted by wide context, which
thereby has a semantic role. Yet to say that the datum may be accounted for in
contextualist terms does not yet obviously prove that it must be so accounted
for. Perhaps the non-contextualist can still shoot some arrows from his bow.


What I will try to show in this Section is that the non-contextualist’s arrows are
blunt.

To begin with, a non-contextualist might try to deny the datum altogether.
That is, she may suggest that in the case of a multi-stable pun different
utterances are involved for one and the same sentence. Perhaps these are not
physical utterances, but mental ones – typically, one in the speaker’s, the other
in the interlocutor’s, mind. Moreover, such utterances are respectively
represented by different clause-index pairs with their own truth-conditions.
Thus, the case raises no particular problem.21
To be sure, such a strategy is appealing when we want to draw a distinction
between a physical token of a sentence and its different interpretations due to
different (mental) readings, hence different (mental) utterances, of it. For
instance, this may happen with the case of a road sign such as:

10. (You) drive slowly.

In this case one and the same physical token – the road sign itself – of a certain
sentence is matched by different (mental) utterances that are differently
interpreted by each of the different drivers that pass in front of that sign. I
interpret my reading of (10) as meaning that I have to drive slowly, you
interpret your reading as (10) as meaning that you have to drive slowly, and so
on.
Yet no such account is legitimately involved in the case of a multi-stable
pun.

Let me set well aside the single outer utterance that is involved in such a
case, the concrete uttering of the relevant sentence on the speaker’s part. Yet if
we stick to inner utterances, here again we have, first of all, just one and the
same mental utterance on the speaker’s part. Yet this inner utterance has again
different interpretations, as well as the speaker’s outer utterance. Simply, we
have just moved one step back – from outer to inner utterances – so as to focus
on the speaker’s mental utterance as our relevant datum. Moreover, if over and
above that mental utterance that has different interpretations there is another
mental utterance on an interlocutor’s part that has just one interpretation, this
is just an irrelevant utterance in the situation at stake. If in such a situation we
have to choose between the interlocutor and the utterer, clearly it is the utterer
who counts.22 For as I said before, by uttering certain sentences in mere

(21 I owe this suggestion to Genoveva Marti).

(22 At least in normal cases. Sometimes, a punny effect is created in the interlocutor’s mind completely
regardless of the utterer’s intentions. Everyone remembers John Fitzgerald Kennedy’s saying in


isolation, pun creators often mean the relevant utterances as having different
meanings and thereby want their interlocutors to understand those utterances
as having such different meanings. So, taking one step back, from the outer to
the inner, is really no advantage for the non-contextualist.
Incidentally, one similar ambiguity of the utterance does not take place with
puns only. Sometimes speakers want to be interpreted as if they had uttered
something that has different meanings even if there is nothing punny about
it.23 Consider Mozart’s saying to Salieri à propòs of his compositions
(according to the movie Amadeus):
11. I never thought that such a music were possible.24

Or take a lover saying to her lover who has finally chosen someone else:

12. I wish you what you really deserve.

In both cases, two interpretations are intended to be available for one and the
same utterance: in the case of (11), both that Salieri’s music is excellent and
that it is very bad, in the case of (12), both that the errant lover has a happy life
and that s/he has an unhappy one. All in all, therefore, it seems better for the
non-contextualist to accept the datum as it occurs at the level of a physical
utterance. For it presents itself again at the level of a mental utterance.
At this point, the non-contextualist may agree to put 2*) of her original
argument aside in its overall scope – 2*), remember, states the following
identity: that a sentence is ambiguous = that a sentence is associated with
different clause-index pairs respectively representing different utterances of
that sentence and endowed with their own truth-conditions. For she may
modify 2*) in the light of the claim that one and the same clause-index pair, as a
linguistic representation of an utterance, may have different truth-conditions.
Berlin

“Ich bin ein Berliner”

 in order to say that he was to be taken as a real inhabitant of the city.


Although his German interlocutors well grasped JFK’s intention, they could not stop laughing when
thinking of the other possible interpretation of the utterance as meaning that JFK is a pastry of a
particular sort. (I thank Dan Zeman for having attracted my attention to this point).

(23 This is as it should be. For multi-stability is one thing, humor is another. If one wanted to stick to
the ‘incongruity theory’ (cf. fn.15) to also account for the comical effect of multi-stable puns, one
might say that humor here again depends on the reciprocal incongruity of the different interpretations
involved).

(24 For the present moral of this example see also Sperber and Wilson (1987, p. 751). For Sperber and
Wilson, this example indeed presents no problem for their relevance theory, as Morgan and Green
instead claim (1987, p.727).

By so doing, the pair remains the real truth-conditional bearer; simply, in
certain cases it has more than one truth-condition.
To begin with, this move seems to betray the spirit of non-contextualism.

As we have seen, non-contextualists have put forward clause-index pairs
wearing their own truth-conditions on their sleeves in order to satisfactorily
deal with sentence ambiguity. Yet if some of such pairs have different truthconditions,
doesn’t ambiguity come back in through the rear door? In any case,
this move does not seem promising. Once one and the same clause-index pair
is assigned many truth-conditions, the pair becomes not only a wheel that turns
idly, but also something on which a semantic mechanism cannot operate. It is a
wheel that turns idly, because there is no reason to assign different truthconditions
to one and the same pair rather than straightforwardly to one and
the same utterance. But it is also something a semantic mechanism can no
longer operate on. For how can a semantic mechanism not only
computationally assign many truth-conditions to a pair, but also settle to which
pair many such conditions are to be assigned? In order to do both things, the
mechanism should be able to compute not only complex meanings out of
simple ones (given a syntax) but also the utterer’s intention, which as I said
before is in this respect the decisive factor. Yet to begin with, it is not clear how
a semantic mechanism can manage to compute that intention. As I said before,
the fact that in a multi-stable pun a sentence is uttered in isolation may help the
interlocutor to grasp the intention of that pun’s utterer, hence its multistability;
such a grasp does not seem to be an automatic matter. To be sure,
there may be a mechanism for mindreading that allows one to compute the
utterer’s intentions, as some claim.25 Yet this is no semantic mechanism. As
any other mechanism, a semantic mechanism is domain-specific: it purportedly
delivers the meanings (in narrow context) of complex expressions out of the
meanings of their constituent expressions given a certain syntactical structure.

The non-contextualist may now be tempted by the opposite move. Why not
stick to 2*) while representing one and the same utterance by many clauseindex
pairs, each with its own truth-conditions?
There may be an immediate justification for this move. For a noncontextualist
may say that in a clause-index pair one may increase the index
parameters. For instance, one may add a ‘language’- parameter to an index. In
such a case, the pair would be made by a linguistic representation obviously no
25 Cf e.g. Leslie (1997).


longer of a natural language sentence, but of a string of words (or of sounds),
plus an enriched index that pairs that representation not only with agent,
space, time, and world, but also with a language. This new parameter fixes
which language the above string belongs to.26 As a result, one and the same
utterance – or at least one and the same string of uttered words, or better of
uttered sounds – may be represented by different pairs whose indexes’
parameters are differently saturated; those pairs wear different truthconditions
on their sleeves, as it were. So, why not saying that something along
these lines happens in the case of an utterance of a multi-stable pun? Take the
sentence:

13. Humpty Dumpty sat on a wall.

If you focus on the sound of it, (13) is like the French sentence:

14. Un petit d’un petit s’étonné aux Halles.27

Now, a non-contextualist may say that the same string of uttered sounds
involved both by (13) and by (14) may be represented not only by a pair C
whose index i contains English at its ‘language’-parameter, which is true iff
there is someone identical with Humpty Dumpty who sat on a wall, but also by
a pair C’ whose index i’ differs from i for it contains French as its ‘language’-
parameter, which is true iff a child’s child was astonished at les Halles.28
Conceived as an appeal to enriched indexes, perhaps this move accounts for
the first of the two problems afflicting the previous non-contextualist move,
superfluity (why not assign the different truth-conditions directly to the
utterance rather than to a pair representing it?). Yet it does not account for the
second of the two problems afflicting the previous non-contextualist move,
arbitrariness (how can the semantic mechanism not only computationally
assign many pairs to an utterance, but also settle when an utterance is multiply
represented in this way?).

Granted, by appealing to the idea of enriched indexes, hence enriched
pairs, the non-contextualist may say that this move raises no issue of
arbitrariness. First, any utterance of the relevant sounds/words is represented
26 Such an idea is envisaged by Recanati (2000).

(27 This pun can be found in a collection of English yet homophonically French nursery rhymes by van
Rooten (1967)).

(28 For an example involving the same string of uttered words , consider

I vitelli dei romani sono belli.

which in Italian is true iff Romans’ calves are beautiful while in Latin is true iff Vitellius goes to the call
of war of the Roman god).

by all the pairs that one may obtain by giving different values to the new extraparameter
in the index. Second, the semantic mechanism searches for the
extra-parameter in the index in order to compute truth-conditions for each of
those pairs; in the case presented by (13), we have at least two pairs for one and
the same utterance of sounds.

Yet the contextualist may reply that the idea of enriched indexes is rather ad
hoc, or at least is ad hoc if one wants to apply it to the case of multi-stable puns.
Theoretically speaking, one can always take a factor of wide context and put it
in a suitable index. Yet one has to provide an independent reason as to why
such a move should be performed.29 To be sure, a non-contextualist might find
a justification as to why one should add a ‘language’-parameter to an index of a
clause to be assigned to an utterance. The idea would be, an utterance of
sounds or even of words cannot be semantically evaluated until it is established
which language that utterance belongs to.30 However, all the previous cases of
multi-stable puns are precisely cases of puns in which it has already been
settled which language a certain utterance of sounds/words belongs to;
nevertheless, the utterance of the resulting sentence of that language counts as
ambiguous. So, let us assume for argument’s sake that by adding a ‘language’-
parameter the non-contextualist move accounts for bilingual multi-stable puns.

Yet in order to account for all monolingual multi-stable puns, which are the
overwhelming majority of such puns, the non-contextualist should find another
parameter over and above the ‘language’- parameter to be added to the relevant
index. Yet what parameter could it be?

Yet the non-contextualist may reply, appealing to enriched indexes is not
the right way for this move to be understood. In order to have different pairs for
one and the same utterance, one should not enrich indexes, but rather
duplicate clauses. So, one has to pair the same ambiguous utterance with pairs
that differ with respect to their different clauses, which make those pairs
possess different truth-conditional interpretations.

Yet this non-contextualist way of understanding the present move is rather
artificial. It treats utterance ambiguity in the same way as the non-contextualist
(29 For an analogous criticism of the idea that when demonstratives are involved in a truth-conditional
assignment the typical wide-contextual factors that are prima facie relevant for such an assignment,
namely either demonstrations or demonstrata, have to be put in the relevant index, cf. Recanati
(2004a, p.57).)

(30 As Borg (2004, p. 140) accurately claims, by pointing to an original example by Davidson: does one
string of uttered sounds/words count as an utterance of the English sentence “Empedocles leapt” or
of the German sentence “Empedocles liebt”?)

treated sentence ambiguity.

An ambiguous sentence is such that it must be
paired with different clauses, hence with difference clause-index pairs, that
respectively have a different truth-conditional interpretation. Likewise for an
ambiguous utterance. Yet in the case of an ambiguous sentence, the two
different clauses hence the two different pairs respectively represent different
utterances of such a sentence. But in the case of an ambiguous utterance, what
are the different things that the two different clauses hence the two different
pairs respectively represent? Mental utterances? Yet as we have seen before, in
the case of a multistable pun an inner utterance is as ambiguous as an outer
utterance. So, there are no different mental utterances the underlying clauses
hence the pairs respectively represent. Or perhaps different meaning
experiences, i.e., the experiences of hearing/reading an utterance as having a
certain meaning rather than another one? Definitely, in the case of a
multistable pun we have different such experiences. Yet a meaning experience
is not what an underlying clause, hence a pair, should represent. What should
be represented should be something that, unlike a meaning experience, has no
meaning by itself, for it should receive its meaning precisely from the
underlying clause, hence a pair, endowed with its own truth-conditional
interpretation. So, the analogy between ambiguous sentences and ambiguous
utterances the non-contextualist appeals to breaks down. In uttering an
ambiguous sentence, its utterer means it in a particular way insofar her
utterance is represented by an underlying clause – possibly, a Mentalese
sentence – having its own truth-conditions and lending them to the utterance
itself. But in thinking one and the same ambiguous utterance, its thinker means
in a particular way insofar as her clause with its own truth-conditions represent
what?

At this point, the non-contextualist may put forward another, perhaps more
obvious, move. True enough, the datum that in the case of multi-stable puns
one and the same utterance has different significances is undeniable. Yet, the
non-contextualist may go on saying, since such different significances are
precisely a matter of the utterer’s intentions, as the contextualist has pointed
out, they have to be taken as implicatures in Grice’s (1975) classical sense
rather than as explicatures (to put in Carston’s (2002) terms); namely, as far as
a sentence in (narrow) context is concerned, such different significances affect
the level of what is implicated rather than the level of what is said, hence an
extra-truthconditional level rather than the truth-conditional level. If this is the
case, wide context is certainly involved, but in its typical post-semantic role. In
this case, wide context helps in assigning an utterance, which already has (via
the proper clause-index pair representing it) just certain truth-conditions,
different implicatures at one and the same time. In this respect, a noncontextualist
may say, the comparison between the utterances of multi-stable
puns such as (6)-(7)-(8) and the utterances of insidious sentences such as
(11)-(12) is welcome. For, she would continue, each of the latter utterances is
insidious.

For by means of each such utterance its utterer intends to convey a
plurivocal significance over and above the unambiguous truth-conditional
meaning it already has. Mutatis mutandis, the same holds of each of the punny
utterances of (6)-(7)-(8).
Yet this move again seems ad hoc. For the non-contextualist has claimed
that in the case of puns involving a speaker-induced well-grounded
misunderstanding to be removed, like the one presented by (5) before,
meaning interpretation has to be accounted for at the semantic level. As a
result, the non-contextualist has insisted that in such cases wide context
operates at the pre-semantic level. Given the utterer’s intentions, the relevant
utterance has just one truth-conditional interpretation. Simply, as wide context
enables one to discover, that interpretation is not the one originally supposed
by the interlocutor. Yet why then in the case of multi-stable puns should wide
context be appealed to at the post-semantic level? In other terms, it seems
arbitrary to account for one and the same phenomenon of signification now at
the pre-, now at the post-, semantic level.

In this respect, the contextualist account is more elegant. For it appeals to a
unitary explanation enabling her to account for all kinds of puns – not only the
two previously mentioned, but also a third case of puns involving zeugmas –
precisely at the semantic level. The idea is, wherever there is a punny content
for an utterance, this content has to be accounted for at the truth-conditional
level: it is what the utterance says. Hence, the role wide context plays in the
relevant pun depends on how to ensure that the punny content is what the
utterance says. Let me describe things more in detail.
To begin with, the contextualist may even acknowledge that in the case of
puns involving a speaker-induced well-grounded misunderstanding, wide
context has a merely pre-semantic role. For she agrees that in such a case the
relevant utterance of the sentence in question is not ambiguous. Yet for her
what really counts in this move is that it saves the point of the pun, which is to
let the interlocutor realize that the utterance says a certain thing (rather than
another). For that utterance has certain truth-conditions, the punny ones. Yet
the contextualist will add that if we also want to save the point of the pun in the
case of multi-stable puns, we have to claim that wide context should play a
semantic role in such a case. For this time the point of the pun is to let the
utterance say different things. The contextualist accounts for this utterance
ambiguity by saying that in such a case wide context makes one and the same
utterance have many truth-conditions. According to the contextualist,
moreover, also in a third kind of case, namely the case of zeugmatic puns, wide
context must play a semantic role. For this time the point of the pun is to let the
utterance say an ‘impossible’ thing. On the one hand, the relevant sentence
contains an anaphoric link syntactically suggesting that in the relevant
utterance of that sentence, a certain sub-sentential token of a term yields the
same truth-conditional contribution as another token of that term to which that
token is anaphorically linked. Yet on the other hand, the anaphora
notwithstanding, in that utterance the relevant token yields a truth-conditional
contribution (possibly radically) different from the one provided by the token it
is anaphorically linked to. As wide context, notably the utterer’s intentions,
again causes the utterance to say, thereby again playing a semantic role.
Consider:

15. After two unsuccessful marriages, I find myself keeping my guard up,
along with my underpants

16. I called her a whore and myself a cab

17. John and his driving licence expired yesterday.

In all the punny utterances of those sentences, a meaning shift occurs.

Sometimes this shift is less radical, involving only different yet somehow
related polysemical senses, sometimes it is more radical, involving ambiguity.
Now, the anaphorical link subsisting in all these sentences does not prevent
that shift from occurring. In the relevant utterance of (15), an implicit token of
“keeping” refers back to a previous explicit token of the same word, yet the
implicit token has a ‘physical’ reading while the explicit one has a metaphorical
reading: underpants, not guard, are physically kept up. In the utterance of
(16), an implicit token of “called” refers back to a previous explicit token of
the same word, yet the implicit token means retrieved someone’s attention
while the explicit one means labelled. In the utterance of (17), an explicit token
of “expired” refers back to a previous implicit token of the same word, yet the
explicit token means came to an end while the implicit one means died.31 In all
these cases, the anaphorical link would prompt one to interpret the second
token exactly like the first one. Yet such a syntactically driven interpretation is
rejected. For one realizes that the utterer’s purpose was precisely to create a
punny effect in which the meaning of the second token is shifted, the anaphoric
link notwithstanding. This is precisely why the interpretation of the whole
utterance is ‘impossible’. For the contextualist, this interpretation is again a
truth-conditional interpretation of such utterances: they say something
‘impossible’.
Incidentally, once again this phenomenon – anaphorical link plus meaning
shift in one and the same utterance – does not obtain in punny cases only.32
Consider two utterances of:

18. Norman Mailer likes to read himself

19. He drank the whole bottle and smashed it to the floor.33

Contextualists would again say that also in these cases, the anaphoric link
respectively tying “himself” to “Norman Mailer” and “it” to “the (whole)
bottle” notwithstanding, those utterances of (18) and (19) are respectively
true iff Norman Mailer likes to read some work or other of his and iff the
demonstrated person drank the whole content of a bottle and smashed that
bottle to the floor. Unlike the zeugmatic cases presented by (15)-(16)-(17),
however, that the relevant utterances of (18)-(19) have such ‘impossible’ truthconditions
depends not on the utterer’s intentions, but rather on other widecontextual
factors. These factors are i) co-text – the fact that the pronoun
“himself” follows the verb “to read” in (18) and the fact that in (19) the first
token of “bottle” is linked with the verb “to drink” and the adjective “whole”,
while the second token is linked with the verb “to smash” – and ii) pragmatic
relations linking the ordinary referents with the extended referents of the
relevant expressions in those utterances, namely a pragmatic relation linking
(31 Of course, in order to fully vindicate the anaphora, one may postulate an ad hoc concept that would be the
disjunctive referent of both the first and the second token of the relevant subsentential term. Yet, as Carston
and Wilson implicitly admit in putting forward such a proposal - cf. (1996, p. 427), (1997) - this move
would annihilate the punny effect of the utterance. See also Lascarides, Copestake and Briscoe (1996, p.
47).
(32 Once again, one thing is the meaning shift, another is the comical effect. On behalf of the ‘incongruity’
theory of humor (cf. fn.15), one may again say that even here the meaning shift is comical for the ‘impossible’
interpretation is incongruous).

(33 For those examples, cf. Fauconnier (1985, p. 7) and Sainsbury (2010, p.139) respectively_

an author with his work in the case of (18) and another pragmatic relation
linking a container to its content in the case of (19).34
Clearly enough, in such cases of meaning shift meta-contextualism is again
at work. For it is wide-contextual which wide-contextual factor is relevant to
determine the utterance’s truth-conditions: utterer’s intentions in the case of
the above utterances of (15)-(16)-(17), factors i) and ii) in the case of the
above utterances of (18)-(19). Such a meta-contextualism is even more evident
if we consider different utterances of one and the same sentence, a punny and a
non-punny one. Consider a punny utterance of:
20. I’m reading a book about anti-gravity. I can’t put it down.
In this utterance, the meaning shift from literary work to physical copy of a
work involving the first token and the second (implicit) token of “book”
respectively leads that utterance to have ‘impossible’ truth-conditions. These
truth-conditions are induced by the utterer’s comical intention to generate a
zeugmatic effect. Yet if (20) were uttered without such an intention, the
resulting utterance would not have such truth-conditions. Rather, it would just
have the normal truth-conditions according which that utterance would be true
iff the agent of the context were reading a work on anti-gravity and that reading
could not be interrupted.


In this paper I have not only provided an interesting datum concerning
multistable puns, according to which punny utterances of certain sentences
have multiple readings, but I have also claimed that those readings affect the
truth-conditional level of what is said by such utterances. As such different
readings are prompted by a wide-contextual factor, namely the speaker’s
intentions, with respect to such utterances wide context plays a semantic role.
Moreover, I have claimed, this account is grounded by the idea that punny
interpretations affect the truth-conditional level of what is said also when puns
of other kinds are at stake, namely both the puns involving a speaker-induced
removal of a well-grounded misunderstanding and the zeugmatic puns. Simply,
while in the former case wide context plays a pre-semantic role – it selects a
given interpretation for an utterance of a sentence by choosing which meaning
that sentence actually mobilizes – in the latter case it again plays a semantic
)34 For this way of putting the referential distinction in question cf. Nunberg (1979)).
role, by making the utterance funnily say something ‘impossible’ in virtue of
the speaker’s intentions. Now, experimental pragmatics may further
corroborate this latter idea, if it will experimentally turn out that, as I believe,
language users straightforwardly interpret a punny utterance according to one
of the three possibilities just sketched: a) by proving an interpretation that
removes a previous interpretation; b) by providing a multistable interpretation;
c) by providing an ‘impossible’ interpretation.

ACKNOWLEDGEMENTS
This paper was originally presented at the LOGOS Seminar, Department of
Logic and Philosophy of Science, November 10, 2010, Barcelona and then at
the Institut for Logic, Cognition, Language, and Information, June 10, 2011,
San Sebastian. I also thank Manuel Garcia-Carpintero, Andrea Iacona,
Genoveva Marti and Stefano Predelli, for their important comments on
previous versions of this paper.

REFERENCES

Bianchi, C. (1999). Three Forms of Contextual Dependence. In Modeling and Using
Context. Proceedings of the Second International and Interdisciplinary
Conference, CONTEXT '99, Trento, Italy, September 9-11, P. Bouquet et al.
(Eds.), Berlin: Springer, 67–76.
Borg, E. (2004). Minimal Semantics. Oxford: Oxford University Press.
Cappellen, H., & Lepore, E. (2005). Insensitive Semantics. Oxford: Blackwell.
Carston, R. (2002). Thoughts and Utterances. Oxford: Blackwell.
Carston, R., & Wilson, D. (1996). Metaphor, Relevance and the ‘Emergent Property’
Issue. Mind and Language,21, 404-433.
Carston, R., & Wilson, D. (1997). A Unitary Approach to Lexical Pragmatics:
Relevance, Inference and Ad Hoc Concepts. In N. Burton-Roberts (Ed.),
Pragmatics, London: Palgrave, 230-259.
Puns for Contextualists 139
Fauconnier, G. (1985). Mental Spaces. Cambridge: Cambridge University Press.
**************************************

Grice, H. P. (1938). Negation. The Grice Papers, UC/Berkeley, California, Bancroft Library.
Grice, H. P. (1975), Logic and Conversation. In P. Cole, & J.L. Morgan (Eds.), Syntax
and Semantics 3: Speech Acts, New York: Academic Press, 41–58.
*************************************************************

Horn, L. R. A brief history of negation.

Kaplan, D. (1989a). Demonstratives. In J. Almog et al. (Eds.), Themes from Kaplan,
Oxford: Oxford University Press, 481-563.
Kaplan, D. (1989b). Afterthoughts. In J. Almog et al. (Eds.), Themes from Kaplan,
Oxford: Oxford University Press, 565-614.
Lascarides A., Copestake, A., Briscoe, T. (1996). Ambiguity and Coherence. Journal
of Semantics, 13, 41-65.
Leslie, Alan M. (1997). Pretense and Representation: The Origins of ‘Theory of
Mind’. Psychological Review , 94, 412-426.
Morgan Jerry L., and Green Giorgia H. (1987). On the Search for Relevance.
Behavioral and Brain Sciences ,10, 726-727.
Morreal, J. (2009). Comic Relief. Oxford: Blackwell.
Nunberg, G. (1979), The Non-Uniqueness of Semantic Solutions: Polysemy.
Linguistics and Philosophy , 3, 143-184.
Perry, J. (1997). Indexicals and Demonstratives. In R. Hale & C. Wright
(Eds.),Companion to the Philosophy of Language . Oxford: Blackwell, 586-
612.
Pinker, S. (1994). The Language Instinct. London: Penguin Books.
Predelli, S. (1998). I Am Not Here Now. Analysis, 58, 107-115.
Predelli, S. (2005). Contexts. Oxford: Oxford University Press.
Recanati, F. (2000). Oratio Obliqua, Oratio Recta . Cambridge, Mass.: The MIT Press.
Recanati, F. (2004a). Literal Meaning. Cambridge: Cambridge University Press.
Recanati, F. (2004b). Indexicality and Context-shift. Paper presented at the
Workshop on Indexicals, Speech Acts and Logophors , Harvard University,
20.11.2004. Retrieved from
http://jeannicod.ccsd.cnrs.fr/docs/00/05/36/04/PDF/ijn_00000552_0
0.pdf.
Stanley, J. (2007). Language in Context. Oxford: Clarendon Press.
140 Humana.Mente – Issue 23 – December 2012
Sainsbury, M.R. (2010). Fiction and Fictionalism. London: Routledge.
Speranza, Join the Grice Club.
Sperber, D., & Wilson, D. (1987). Precis of Relevance: Communication and
Cognition. Behavioral and Brain Sciences, 10, 697-754.
Tanaka, K. (1992). The Pun in Advertising: A Pragmatic Approach. Lingua 87, 91-
102.
Travis, C. (1997). Pragmatics. In R. Hale and C. Wright (Ed.s),Companion to the
Philosophy of Language. Oxford: Blackwell, 87-107.
Van Rooten, L. (1967). Mots D’Heures: Gousses, Rames: The D’Antin Manuscript.
New York: Grossmann Publishers.
Voltolini, A. (1995). Indexinames. In J. Hill & P. Kot'àtko (Eds.), Karlovy Vary
Studies in Reference and Meaning, Prague: Filosofia Publications, 258-285.
Voltolini, A. (2006). Fiction as a Base of Interpretation Contexts. Synthese , 153, 23-
47.
Voltolini, A. (2009). How Demonstrative Pictorial Reference Grounds Contextualism.
Pacific Philosophical Quarterly , 90, 402-418.
Wittgenstein, L. (1953). Philosophical Investigations . Oxford: Blackwell.
Wittgenstein, L. (1980). Remarks on the Philosophy of Psychology . Oxford:
Blackwell.
Wittgenstein, L. (1982). Last Writings on the Philosophy of Psychology, vol. I .
Oxford: Blackwell.
Wittgenstein, L. (1992). Last Writings on the Philosophy of Psychology, vol. II .
Oxford: Blackwell