69

4 Particularized Conversational
Implicatures: Why There Are
Conversational Implicatures

4.1 INTRODUCTION

In Chapter 1, we introduced the Gricean distinction between

generalized conversational implicatures (GCIs) and particularized
conversational implicatures (PCIs). Given the Gricean approach,
where implicatures are mostly derived from a comparison between
what the speaker actually said (her utterance) and what she might
have said but did not say (the alternative(s) to her utterance), the dis-
tinction is grounded in how the set of alternatives is determined. In
GCIs (of which scalar implicatures are the epitome, see Chapter 6),
the set of alternatives is lexically determined. In PCIs, the set of
alternatives is contextually determined. In both cases, however, the
Gricean mechanism of implicature derivation implies that the hearer
will reason from the actual utterance and the set of alternatives,
under the premise that the speaker complied with the Principle of
Cooperation. This means that attribution of mental states to the
speaker (e.g. beliefs and intentions) will be crucial to the process. This
part of the Gricean account has raised some doubts as this would
make the mechanism very cumbersome, while discourse processing
is very fast (people produce ten to twelve utterances per minute,
which means that each utterance takes about 6 seconds to produce
and, one would presume, to interpret; see Bickerton 2014: 36). This
makes it rather unlikely that anything as complicated as the mech-
anism described by Grice can be operational in implicature deriv-
ation. It is noteworthy that Grice himself insisted that his ‘logic of
conversation’ was not intended as psychologically realistic. Rather
the Principle of Cooperation and the attendant maxims are supposed
to be norms that apply to rational communication (and not even spe-
cifically to linguistic communication).
Thus, the criticism that the logic of conversation is unlikely
to account for what happens in human communication is not

69
Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
70

70 T Y P E S O F I M P L I C AT U R E

unexpected. While the criticism applies to the derivation of all

implicatures, the alternative accounts that have been proposed have
generally maintained the distinction between GCIs and PCIs on
the basis that the first are context independent (and depend on lin-
guistic, that is, lexical or grammatical mechanisms, see Chapter 6).
The exception is, of course, relevance theory, which claims that, as
a matter of fact, the interpretation of all implicit communication
is context dependent and that the distinction is thus spurious (see
Chapter 3). It has to be said that, with the exception of relevance
theory, PCIs have been largely neglected, as they are clearly not
amenable to linguistic treatment, and people have been content to
stay with the Gricean account.
The present chapter will concentrate on PCIs, comparing the
Gricean with the relevance-theoretic accounts. It will also deal
with two important questions relative to implicatures, questions
on which PCIs highlight possible answers: whether one needs
Gricean, alternative-based reasoning to access implicatures, and,
most centrally, why there are implicatures at all. Relative to this
second question, one possibility, which we will now explore, is that
implicatures are used because there is no other way for the speaker
to communicate what she intended to communicate.

4.2 METAPHORS AS IMPLICIT COMMUNICATION

Grice (1989) discussed metaphors, which he considered to be a form

of PCIs. On the Gricean account, metaphors, taken literally, are false
and hence flout the Principle of Cooperation and the maxim of Quality
that enjoins the speaker to say only what she believes to be true. To
restore the notion that the speaker did comply with the Principle of
Cooperation, the hearer accesses an implicature, corresponding to
a true proposition. This true proposition corresponds to what was
known in classical rhetoric as the figurative meaning of the metaphor.
Relevance theory gave two successive accounts of metaphors, but
we will only be interested here in the first one (it is not clear that
these two accounts are compatible, see Reboul 2014, 2017a, and the
second is not relevant to our current purpose, i.e. explaining the use
of implicatures). The relevance-theoretic account is based on a stark
criticism of the Gricean one. First, while Grice supposed that his
view would explain how all metaphors work, this cannot be the case,
given that all metaphors are not false. Let us look at the following
examples:

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
71

Particularized Conversational Implicatures 71

(1)
Bill is a bulldozer.

(2)
No man is an island.

While clearly (1) is false, as Bill, however unpleasant an individual

he is, is a human being, (2) is clearly true: no human being is an
island. Additionally, if falsity was a central feature of metaphor, one
would expect that negating a metaphor would change it into a literal
utterance, given that negation reverses the truth value of a sentence.
As (3) shows, this is not the case:
(3)
Bill is not a bulldozer.

This criticism means that Grice’s account in terms of metaphors

flouting Quality is not tenable (as was already noted, with a similar
argument, by Searle 1979).
There is more, however. Grice’s analysis supposes that, for any meta-
phor, there is a single implicature that will be accessed. This impli-
cature, as said above, corresponds to the figurative meaning of the
metaphor, in rhetorical parlance. This view falls victim to one central
objection to the rhetorical notion of figurative meaning. A consen-
sual observation regarding metaphors, notably creative metaphors1
like (2) rather than dead metaphors like (1), is that metaphors cannot
be paraphrased without loss of content. One reason for this may be
that creative metaphors give rise to non-propositional sensory effects
(see Davidson 1978, Guttenplan 2005) that are lost in any kind of
paraphrase. And, clearly, it is hard to see how one can understand
figurative meaning if not as a paraphrase of the metaphor. Given
that (creative) metaphors are not susceptible to paraphrase, Grice’s
analysis is less than satisfactory. The question, however, is whether
the sensory effects that accompany metaphors are enough to explain
why metaphors cannot be paraphrased. The first relevance-theoretic
account claimed, convincingly, that this is not the case.

1
Creative metaphors are metaphors that are not lexicalized, contrary to dead
metaphors, which have one strongly lexicalized meaning. For this reason, the
interpretation of creative metaphors is much more open and variable between
hearers compared to the interpretation of dead metaphors.

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
72

72 T Y P E S O F I M P L I C AT U R E

Relevance theory pointed out that in fact, in creative metaphors,

identifying a single implicature as the figurative interpretation
intended by the speaker is far from obvious. Indeed, in quite a few
cases, it is not even clear that there is a single interpretation the
speaker intended. To take a classic example, Romeo’s speaking about
Juliet in the Shakespearean play, what exactly would the figurative
meaning of (4) be?
(4)
ROMEO: But soft, what light through yonder window breaks? It is the East, and
Juliet is the sun.

There does not seem to be a single implicature that either Shakespeare

(or Romeo) intended to communicate. Rather, it seems that what is
communicated is a set of implicatures and it is up to the reader to
take his or her pick.
As we have seen, this encouraged Sperber and Wilson (1986/1995)
to make a distinction between strongly communicated implicatures
and weakly communicated implicatures (see Chapter 3). A weakly
communicated implicature is an implicature that does not have to
be recovered for an utterance to have been correctly interpreted. In
other words, when a speaker produces an utterance that gives rise
to weakly communicated implicatures, her informative intention is
not entirely determined, and this can be true in two ways. First, it
seems that recovering any of the weakly communicated implicatures
would be enough to satisfy the informative intention of the speaker.
Second, there is no fixed set of weakly communicated implicatures
in the sense that the hearer might come up with an interpretation
that the speaker had not foreseen, without the speaker’s informative
intention failing. By contrast, in a case where the speaker intended to
strongly communicate an implicature, if the hearer does not recover
the set that is the object of the speaker’s informative intention, then
her communicative intention is not satisfied. The relevance of this dis-
tinction to the question of why people use indirect communication is
that, in cases where the speaker weakly communicates implicatures,
there is no way she could have communicated what she intended to
communicate by using direct means. Her only possible way of sat-
isfying her informative intention was to communicate it implicitly.
Thus, we have an explanation for why speakers use implicit communi-
cation: because they cannot communicate the same content explicitly.
So far so good, but clearly this is not an explanation that can
apply across the board. Strongly communicated implicatures are not

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
73

Particularized Conversational Implicatures 73

susceptible to such an explanation. In their case, it seems clear that
there was an explicit way of expressing the same content. So, the
question of why speakers will use implicit communication when
explicit communication would do as well remains open.

4.3 NEITHER THE GRICEAN NOR THE RELEVANCE-THEORETIC ACCOUNTS

CAN EXPLAIN THE EXISTENCE OF CONVERSATIONAL IMPLICATURES

Suppose that during a dinner party, John is serving wine. He asks

Gwendolyn whether she wants any wine. She answers:

(5)
I am a teetotaller.

Clearly (5) is an answer to John’s proposal and the answer is indirect

in the sense that an implicature is needed to ascertain whether it
is a positive or a negative answer. The question is how John will go
about recovering the implicature. Under a Gricean account, he must
construct a set of alternative utterances that Gwendolyn might have
(but did not) utter(ed). These alternatives, given that (5) is an answer,
would probably correspond to explicitly positive or negative answers
to John’s question, e.g.

(6)
No, thank you, I don’t want any wine.

(7)
Yes, please, I’d like some wine.

Given that Gwendolyn chose to utter neither (6) nor (7), but (5),
and on the hypothesis that she is complying with the Principle of
Cooperation, John will have to reach an interpretation. Prima facie, (6)
or (7) would have complied better with the Principle of Cooperation
and more precisely with the maxim of Quantity than does (5):

Quantity:
1. Make your contribution as informative as is required (for the
current purpose of the exchange).
2. Do not make your contribution more informative than is
required.

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
74

74 T Y P E S O F I M P L I C AT U R E

Given that (6) and (7) are more informative relative to John’s query
than (5), how does John recover the negative implicature of (5)?

Gricean interpretation:

a. Gwendolyn is a teetotaller.
b. By saying that she is a teetotaller, Gwendolyn intends to answer
John’s offer of wine.
c. Someone who is a teetotaller does not drink any alcoholic
beverage.
d. Wine is an alcoholic beverage.
e. Gwendolyn does not want any wine.
f. By saying that she is a teetotaller, Gwendolyn means that she
does not want any wine.
g. By saying that she is a teetotaller, Gwendolyn intentionally
informs John that she will not accept any offer of alcoholic
beverage.
h. John should not offer Gwendolyn any alcoholic beverage.
Thus, arguably, by choosing (5) as an answer to John’s proposal,
Gwendolyn has complied with the maxim of Quantity and hence
with the Principle of Cooperation.2
On a relevance-theoretic account, the reasoning is simpler:

Relevance-theoretic interpretation:

a. Gwendolyn is a teetotaller.
b. Teetotallers do not drink any kind of alcoholic beverage.
c. Wine is an alcoholic beverage.
d. Gwendolyn does not want any wine.
One obvious question is why Gwendolyn answers John’s query with
(5) rather than (6) or (7). Granted, she does manage to communicate
that she does not want any wine (and thus complies with Quantity
in the Gricean account), but she could have done that just as well by
saying (6), and it would have spared her hearer the rather complicated
kind of reasoning illustrated above. Now, the answer to that, both in
the Gricean and in the relevance-theoretic accounts, would seem to
be that, by using (5), she could also communicate the ulterior infor-
mation that she not only does not want wine, she does not want any

2
Note that in the Gricean reasoning outlined above, some of the premises have to
do with the speaker’s mental states.

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
75

Particularized Conversational Implicatures 75

kind of alcoholic beverage. But, as we will now see, it is doubtful
that this explanation is available on a Gricean account. Given that
the Principle of Cooperation says that the speaker should ‘make [her]
conversational contribution such as is required at the stage at which
it occurs, by the accepted purpose and direction of the talk exchange
in which you are involved’, and that Quantity enjoins her ‘not [to]
make [her] contribution more informative than is required’, one may
wonder whether by volunteering that she is a teetotaller, Gwendolyn
is not being over-informative. In other words, she is communicating
information far in excess of what is ‘required at this stage’ in the con-
versation. Thus, the Gricean account fails to explain how the implica-
ture (Gwendolyn does not want any wine) is recovered (as Gwendolyn
flouts Quantity) and also fails to explain why Gwendolyn chose to use
(5) as an answer to John’s query.
Let us now turn to the relevance-theoretic account. As described
in Chapter 3, the pragmatic mechanism at work in interpreting
utterances in relevance theory is constrained by a principle of least
effort: least costly interpretations are accessed first and the process
stops as soon as an interpretation consistent with optimal relevance
is reached. Optimal relevance itself is defined as the balance between
interpretative cost and cognitive effects (see Chapter 3 for a detailed
presentation of the mechanism involved). Regarding cases like (5) as
an answer to John’s question, it clearly is costly and thus this entails
that (5) must be communicating information that will balance this
additional cost. This is done through the fact that (5), in addition to
implicitly communicating a negative answer, also conveys the add-
itional information outlined in (8):

(8)
Gwendolyn does not drink wine.
Gwendolyn does not drink beer.
Gwendolyn does not drink vodka.
Gwendolyn does not drink whisky.
Etc.

A straight answer, such as (6), would only convey one piece of infor-
mation (that Gwendolyn does not want any wine), while an implicit
answer, such as (5), communicates all the information in (8) in add-
ition. Thus, (5) is used because it communicates more information
than (6). More generally, implicatures occur as a result of the search
for optimal relevance.

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
76

76 T Y P E S O F I M P L I C AT U R E

All of this, however, is based on a comparison between a less inform-

ative and less costly answer, (6), and a more informative and more
costly answer, (5). In other words, in such a comparison, both factors
in optimal relevance, i.e. cost and information, vary. The question is
whether this is the right comparison. One might argue that a more
relevant comparison would be between utterances where one factor
in optimal relevance is kept constant while the other varies. For
instance, comparing (5) with (9):
(9)
No, thank you, I am a teetotaller.

In this case, clearly (9) is as informative as (5), and less costly as it is a

direct answer to John’s question. In other words, while the principles
of least effort and optimal relevance explain why the hearer draws
the implicature in (5) (both the negative answer and the propositions
in (8)), it does not explain why Gwendolyn chose to answer with
(5) rather than with (9). This seems to subject John to unnecessary
costs.
Thus, neither the Gricean account nor relevance theory seems
to explain why speakers use indirect ways of communicating their
meaning. Note that this conclusion applies to implicatures in gen-
eral, that is to GCIs as well as to PCIs. Indeed, regarding GCIs, there
would seem always to be a direct utterance that might have been
used instead of the indirect one. For instance, to return to an example
already used in Chapter 1, a speaker who wants to communicate
that either Peter or Paul but not both will come could have used the
explicit (11) rather than the implicit (10):
(10)
Peter or Paul will come.

(11)
Peter or Paul, but not both, will come.

So the question is why the speaker chooses to use (10), subjecting her
hearer to more interpretive costs while not communicating anything
in addition to what she would have done by using (11). What is more,
by communicating implicitly rather than explicitly, the speaker runs
the risk of being misunderstood, that is, she runs the risk that the
hearer will not draw the implicature. This adds to the mystery of why
speakers choose to communicate implicitly rather than explicitly.

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
77

Particularized Conversational Implicatures 77

4.4 RECOVERING STRONGLY COMMUNICATED IMPLICATURES

DESPITE THE PRINCIPLE OF COOPERATION

As we have just seen, neither the Gricean nor the relevance-theoretic

accounts of implicatures account satisfactorily for why speakers use
implicit rather than explicit communication. We will now show that,
in some cases at least, implicatures are accessed even though it is
clear that the speaker does not respect the Principle of Cooperation.
In other words, we will examine a case where the recovery of the
implicature is not governed by the hearer’s trying to come up with
an interpretation that would restore the premise that the speaker is
being cooperative.
The relevant example is borrowed from Solan and Tiersma
(2005: 213) and is an authentic example. Solan and Tiersma were
interested in what might be called ‘forensic pragmatics’, the use
of pragmatics in law enforcement and the law courts. In that spe-
cific case, taken from an official transcript of a court procedure, the
defendant, Mr Bronston, is accused of fiscal evasion. Here is the rele-
vant exchange:
(12)
Q. Do you have any bank accounts in Swiss banks, Mr Bronston?
A. No, sir.
Q. Have you ever?
A. The company had an account there for about six months, in Zurich.

The relevant utterance is the second answer, which is reproduced

in italics. It implicitly communicates the false information that
Bronston did not have a (personal) account in Swiss banks, while, as
a matter of fact, he did. Bronston clearly did not communicate this
information explicitly in his second answer. In his first answer, he
communicated explicitly that, at the moment of utterance, he did
not have such an account, which was true. However, he had had one
before. And he communicated implicitly through his second answer
that he had not. Thus, clearly, Bronston did not comply with the
Principle of Cooperation and notably with the maxim of Quality.
However, Bronston clearly intended to communicate that he had not
had a Swiss bank account.
Thus, Bronston did not intend to comply with the Principle of
Cooperation. But the most notable thing is that his hearer (the judge)
did not believe that Bronston complied with Quality. Nevertheless,
the judge did recover the implicature that Bronston had not had a

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
78

78 T Y P E S O F I M P L I C AT U R E

Swiss bank account. Indeed, Bronston was later put on a further trial
for perjury on the basis of that answer. So, the judge did recover the
implicature, and, given that he clearly had no illusion that Bronston
was being sincere, he can hardly have recovered it from the premise
that the answer was sincere (as it should be, given the Principle of
Cooperation). Basically, what happens in (12) is that the judge recovers
the implicature, but does not thereby believe that Bronston did not
have a Swiss bank account. It would be erroneous, however, to think
that the problem could be solved by dropping the maxim of Quality.
This is because it goes deeper than dishonesty on the speaker’s part.
As noted above, the problem here is rather that his hearer does not
believe him to be complying with Quality. Thus, he cannot recover the
speaker’s implicature by relying on the premise that the speaker
respected it. To put it in a nutshell, there does not seem to be any
room in the Gricean account for the hearer discounting the speaker’s
sincerity and recovering the speaker’s implicature. This means that,
in any case where the hearer does not believe the speaker to be sin-
cere, he should not be able to recover any implicature whatsoever, as
we will now see.
One might argue that, in such cases, the hearer uses a higher-order
type of reasoning, where he reasons from a hypothetical premise
that the speaker does respect the Principle of Cooperation. As the
premise is hypothetical, the hearer does not have to believe it to
be true but only to draw the conclusions that would derive if it
were true. Supposing this were the case, clearly, in (12), Bronston’s
answer is not only flouting Quality, it is also flouting Quantity. Being
interrogated about his own personal bank accounts, he answers
relative to the company’s bank accounts. Here, obviously, deriving
the implicature that he did not have a personal bank account
(which entails assuming that his answer is exhaustive, i.e. that he
is mentioning all the Swiss bank accounts there were) does restore
the premise that Bronston is complying with Quantity. The premise
on which the reasoning rests, i.e. assuming that Bronston is being
exhaustive about Swiss bank accounts in his answer, is only oper-
ational if the hearer takes it that Bronston complied with Quality. Is
it available if, rather than believing that this is the case, the hearer
only entertains hypothetically (but does not believe) the notion that
Bronston is sincere? It is not clear that it is. While it makes sense to
believe that Bronston is exhaustive about Swiss bank accounts if he
is sincere, there is no reason to believe so if one is doubtful that he
is. If he is insincere, then, clearly he is not being exhaustive in his
answer.

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
79

Particularized Conversational Implicatures 79

In fact, the only way to recover the implicature is by assuming that
Bronston is not only insincere, but is actively engaged in deception.
On the assumption that Bronston is trying to deceive him, the judge
can assume that Bronston intends him to think that his answer is
exhaustive (and, hence, that he had no personal Swiss bank accounts).
So, it is only by assuming that Bronston did flout both the maxim
of Quality and the maxim of Quantity that the judge can actually
recover the implicature. In other words, this is clearly a case where
the Principle of Cooperation cannot be the basis from which the
implicature is recovered.
Regarding relevance theory, the question does not arise at all in
the same terms. Bronston’s answer is most economically interpreted
(according to the principle of least effort) as exhaustive, as it is the
only way in which it is relevant, given the judge’s question. This is
in no way dependent on the hearer’s view regarding the speaker’s
sincerity. In other words, on a relevance-theoretic account, there
is no problem with both believing that Bronston is insincere and
recovering the implicature.
However, the interest of example (12) is not only in showing that
the Gricean account does not work across the board, it also indicates
a direction that could lead to an explanation of why people some-
times prefer to communicate implicitly rather than explicitly. Indeed,
Bronston could have answered the judge’s question as in (13):
(13)
I never had any account in a Swiss bank.

So why did Bronston answer in such an oblique fashion? The end

of his legal story gives an interesting insight. After his initial trial,
Bronson underwent a further trial for perjury. He was acquitted from
this charge, however, because he had not actually claimed not to have
had any Swiss bank account. We will now follow this trail.

4.5 THE POSSIBILITY OF DENIAL

Pinker (2007) diagnosed what is wrong with the Gricean account as

the assumption, implicit in the Principle of Cooperation, that in com-
munication the speaker’s and the hearer’s interests are aligned. This,
however, as he points out, is an unwarranted assumption. Pinker
suggests that the reason why people use implicit communication
is that, by communicating implicitly, they can deny having had the

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
80

80 T Y P E S O F I M P L I C AT U R E

Table 4.1 Pay-offs in the traffic-fine situation (explicit communication)

Action Dishonest officer Honest officer

Do not bribe Traffic fine Traffic fine
Bribe Go free Arrest for bribery

intention to communicate the implicature. His proposal is based on

three elements:
1. the logic of plausible denial;
2. relationship negotiation;
3. language as a digital medium.
Example (12), as we will see below, illustrates both the logic of plaus-
ible denial and language as a digital medium. Pinker’s examples,
however, are not examples of deception via indirect communication,
but rather examples of indirect proposals, e.g. offering illegal bribes
or sexual innuendos.
Beginning with the logic of plausible denial, in Pinker’s first
example, a driver is stopped by a police officer for excessive speed.
The driver’s options are either to accept the fine or to offer the officer
a less costly bribe. Pinker analyses the situation in game-theoretic
terms, with pay-offs indicated in Table 4.1.
Clearly, if the driver bribes a dishonest officer, this brings him the
best outcome, as he gets off with a bribe less costly than the fine
would have been. However, if he bribes an honest officer, he goes
to jail, and that is a much worse outcome than paying a fine. The
problem is that he cannot know in advance whether the officer is or
is not honest. Thus, bribing the officer is hardly an optimal solution.
However, as Pinker notes, this is only true if the bribe is proposed
explicitly, because in such a case, there is no way for the driver to deny
having proposed a bribe. If, on the other hand, the driver proposes
the bribe implicitly, he can deny having done so if the officer turns
out to be honest. The pay-offs are now as indicated in Table 4.2.
In this case, clearly, implicating a bribe is an optimal solution, as
it does not impose a worse pay-off than not bribing, and potentially
offers a better outcome.
Obviously, the use of indirect communication is not limited to legal
situations, and Pinker acknowledges this. This, in fact, is where rela-
tionship negotiation, the second element in Pinker’s theory, comes in.
He offers other examples, for instance bribing the maitre d’ in a busy

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
81

Particularized Conversational Implicatures 81

Table 4.2 Pay-offs in the traffic-fine situation (implicit/explicit

communication)

Action Dishonest officer Honest officer

Do not bribe Traffic fine Traffic fine
Bribe Go free Arrest for bribery
Implicate bribe Go free Traffic fine

restaurant to avoid a long wait for a table, or sexual innuendos. In both

types of case, using indirect communication allows the speaker to deny
having had the intention to make the proposal and thus to avoid social
embarrassment. Pinker limits relationship negotiation to cases where
the speaker wants to preserve her social status by avoiding embarrass-
ment. But, clearly, it is also possible to use indirect speech to preserve (or
make as if to preserve) others’ social status and avoiding embarrassing
them. This is obviously what happens in polite communication. While
Pinker does not acknowledge this, and dismisses politeness theory on
the same grounds that he dismisses the Gricean Cooperation Principle
(i.e. that the speaker’s and the hearer’s interests might not coincide),
we will quickly outline so-called politeness theory, which, after all,
seems quite clearly to relate to relationship negotiations.
Politeness theory (see Brown & Levinson 1987) adopted the
notion of face proposed by Goffman (1982). The face of a person is
the social value that a person thinks she deserves. Given the satis-
faction conditions of some speech acts, such as orders or threats,
which suppose that the speaker is in a status of dominance over her
hearer, they seem to threaten the hearer’s face. Indirect communi-
cation solves the problem. According to Brown and Levinson (1987),
the communication involved in politeness can be ranked into a hier-
archy, ranging from the less to the more polite:
• positive politeness (exposing sympathy);
• negative politeness (adopting a deferential stance);
• indirect speech acts;
• off-the-record indirect speech acts.

Indirect speech acts are illustrated in (14), while off-the-record

indirect speech acts are illustrated in (15):
(14)
Could you pass the salt (please)?

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
82

82 T Y P E S O F I M P L I C AT U R E

(15)
The soup is delicious, but it could do with a bit more salt, I think.

While (14) is standardly used to ask someone to pass the salt, (15) is
not. Thus, the hearer of (14) should understand very easily that the
speaker wants him to pass the salt. On the other hand, the hearer
of (15) might not get the message. While Brown and Levinson (1987)
do not use the terminology of plausible denial, it seems clear that
plausible denial is the difference between indirect acts and off-the-
record indirect acts. While off-the-record indirect speech acts can
plausibly be denied, this is much less plausible in the case of indirect
speech acts. Thus, one might legitimately conclude that Brown and
Levinson’s distinction is based on what is the core of Pinker’s analysis,
i.e. plausible denial. What is more, they predict that politeness, just
as is the case in the examples given by Pinker, depends on plausible
denial. Thus politeness theory predicts that indirect speech should
be seen as more polite than direct speech, and off-the-record indirect
speech as more polite than indirect speech.
This prediction is not, however, borne out by the empirical evi-
dence, as Pinker (2007) reports. While indirect speech acts are indeed
taken to be more polite than direct speech acts, it is not the case
that off-the-record indirect speech acts are perceived as more polite
than indirect speech acts. On the contrary, they are perceived as less
polite. This strongly suggests that plausible denial is not central to
politeness. Up to a point, this is readily understandable: while direct
speech acts of order or threat, having in their satisfaction conditions
the dominance of the speaker over the hearer, threaten the hearer’s
face, this is not the case for indirect speech acts. Thus, indirect speech
acts do not threaten the hearer’s face. On the other hand, off-the-
record indirect speech acts do not add anything to indirect speech
acts, and indeed plausible denial does not either. Just avoiding the
expression of dominance through direct speech acts by using indirect
speech acts is enough for politeness. Plausible denial is an unneces-
sary addition.
But how exactly do indirect speech acts avoid indicating domin-
ance? It is here that the last ingredient in Pinker’s theory comes into
play: language as a digital medium. The necessity for this third compo-
nent is directly related to what happens in cases of indirect speech
when, nevertheless, denial would sound implausible. Pinker claims
that even in such cases, the implicature is still not certain, as cer-
tainty would imply a 100 per cent certainty. This is enough to allow

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
83

Particularized Conversational Implicatures 83

denial, and the possibility of denial is what counts, even if plausibility
is unlikely.
The notion of an off-the-record indirect speech act, indeed, throws
light on the idea of deniability, plausible or otherwise. The idea is that
information transmitted through indirect speech is not immediately
put into common or mutual knowledge (or in Sperber and Wilson’s
(1986/1995) terminology, into mutually manifest assumptions). This
would be the case only for information transmitted through direct
speech. Thus, the distinction between off-the-record speech acts and
indirect speech acts is in fact more similar to the distinction between
dead and creative metaphors (see Section 4.2). Neither indirect
speech acts nor off-the-record indirect speech acts go automatically
into common, mutual or mutually manifest knowledge (in other
words, neither are ‘on-the-record’). Rather, the distinction between
weakly communicated and strongly communicated implicatures is
appropriate here. Indirect speech acts such as (14) above correspond
to strongly communicated implicatures, while off-the-record indirect
speech acts correspond to weakly communicated implicatures. But
all are still implicatures. And here is where the so-called language
as a digital medium ingredient of Pinker’s theory comes into its
own. Given that all indirect speech acts (or at least all those Pinker is
interested in) are conversational implicatures, they are all cancellable
(see Chapter 1), or in other words, they can all be denied without
contradiction, as shown by (16) and (17):
(16)
Could you pass the salt? I am asking because I just want to know about your
motor capacity.

(17)
The soup is delicious, but it could do with a bit more salt. But, as my doctor
forbids salt to me, I will not add any.

Regarding politeness and avoiding assuming dominance, given

cancellability, the speaker is not, by his indirect speech act in (14),
claiming superiority. This will in fact be done by her hearer, if he
passes the salt, acknowledging the legitimacy of the demand in the
implicature. But the use of the indirect speech act allows the hearer
to choose not to obey by ignoring the indirect order/request.
We will now return to example (12) and discuss it in the light of
Pinker’s theory. As said in Section 4.4, Bronston was intending to
deceive the judge into believing that he had never had any account in

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
84

84 T Y P E S O F I M P L I C AT U R E

a Swiss bank. While he could have asserted that he did not by saying
(13), he chose to produce the indirect speech act in (12). The judge,
as we have seen, was not deceived and Bronston underwent a further
trial for perjury. But he was acquitted through language as a digital
medium. While denial of an intention to deceive is clearly implaus-
ible in this case, in spite of the implicature being cancellable, (12) is
clearly not an assertion that Bronston did not have an account in a
Swiss bank. This mere possibility of denial, however implausible, was
enough to acquit Bronston.
Pinker’s theory of conversational implicatures has been vindicated
empirically (see Lee & Pinker 2010) and successfully modelled (Pinker
et al. 2008). Let us now turn to speaker’s commitment and hearer’s
epistemic vigilance.

4.6 SPEAKER’S COMMITMENT AND HEARER’S EPISTEMIC VIGILANCE

Both speaker’s commitment (as expressed by the choice of explicit

or implicit communication) and hearer’s epistemic vigilance are
shaped by Pinker’s point of departure, i.e. the notion that, contra
the Principle of Cooperation, the interests of communicators are
not always aligned. Quite aside from such obvious deception as that
involved in (12), it may still be the case that the speaker’s intentions
are ‘manipulative’. As was pointed out by Krebs and Dawkins (1984),
manipulation does not necessarily involve deception. What it does
involve, however, is the fact that the communicative act produces
in the recipient a behaviour that is beneficial to the communicator
and that would not have occurred without her communicative act.
Note that, on this definition, manipulation may be, but need not
be, costly to the hearer. It may be costly, neutral (neither costly nor
beneficial) or even beneficial to him. However, this does not mean
that hearers welcome manipulation, especially given that it will, at
least occasionally, involve deception. What this means is that one
would expect hearers to develop defence mechanisms. As Sperber
et al. (2010) propose, this is indeed the case, and human cognition
encompasses a number of mechanisms dedicated to calibrating trust
and evaluating arguments as well as the plausibility of claims, which
they called epistemic vigilance.
Additionally, in a series of recent publications, Mercier and Sperber
(Mercier 2009, Mercier & Sperber 2011, 2017) have argued that
human reasoning is mainly dedicated to argumentation, that is per-
suading others that one’s opinion is better than theirs. This implies

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
85

Particularized Conversational Implicatures 85

mechanisms both for the production and for the evaluation of reasons.
This suggests that epistemic vigilance is present from the outset in
human communication, as argumentation implies sophisticated
communication, i.e. presumably, linguistic communication. There
are other mechanisms targeting the source of information (is it reli-
able and competent?) and the communicated content (notably, its
compatibility with previously held beliefs). Additionally, there is a
bias, the myside bias (see Mercier & Sperber 2017), which, without
being strictly speaking part of epistemic vigilance, complements it.
This corresponds to a preference for one’s own opinions over the
opinions communicated by others. Obviously, one’s own opinion may
be erroneous, but preferring it over others’ opinions avoids being
manipulated by them.
The myside bias has a few interesting consequences (for a fuller
discussion, see Reboul 2017b).

• People will tend to hold more strongly to conclusions that they

have reached themselves.
• People will tend to be less epistemically vigilant towards infor-
mation on the truth of which speakers do not seem to commit
themselves, that is, information that is not asserted.
• People will tend to be less epistemically vigilant towards infor-
mation that is not presented to them as a reason to change their
own opinions or decisions, that is, information which is not
perceived as communicated with manipulative or argumenta-
tive intentions.

Now, implicit communication, and specifically conversational

implicatures, seems to be exactly what is needed. Given that it can
only be accessed through an inferential process, its recovery is the
result of a process of reasoning done by the hearer, and, as such, is a
conclusion that the hearer reaches himself. Given cancellability and
the logic of plausible denial proposed by Pinker, it is clear that the
speaker is not committed to the truth of an implicature, as she would
(inevitably) be to the truth of an asserted information. In other words,
implicit communication, and especially conversational implicatures,
is the best way for speakers to avoid epistemic vigilance mechanisms.
This leads to an additional benefit to speakers who communicate
implicitly. Given that epistemic vigilance targets the source (the
speaker) as well as the content of a message, it is in the interest of a
speaker attempting to manipulate and/or deceive her hearer to pre-
serve her reputation for trustworthiness even when the content of the

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
86

86 T Y P E S O F I M P L I C AT U R E

message has been shown to be deceptive or manipulative. One way

to do that is precisely to use implicit communication, notably impli-
cature. Given plausible denial, hearers will tend to give speakers who
communicate a suspect (deceptive or manipulative) content impli-
citly the benefit of the doubt, while they would be much less ready to
do so had they communicated the same content explicitly. Mazzarella
et al. (2018) have investigated this experimentally. They found that,
even after the content of a message has been falsified, participants
were significantly more likely to trust a speaker who had implicated
this content than a speaker who asserted or presupposed it.

4.7 SUMMARY

Before we turn to a summary of the contents of this chapter, we

want to emphasize that, while the arguments in this chapter mostly
concerned PCIs, they also apply to GCIs. In other words, the notion
that implicit communication is mainly used by speakers to avoid or
escape epistemic vigilance, facilitating manipulation and argumen-
tation, does not only concern PCIs, but also GCIs. We will present
generalized conversational implicatures in detail in Chapter 6.
Turning to a summing-up of the contents of this chapter, we first
noted that the Gricean account is not psychologically realistic, as the
mechanisms Grice proposed, involving attributions of mental states
to others, as well as comparison between the actual utterance and
alternatives to it, are much too costly to take place fast enough for lin-
guistic communication. Our main goal in this chapter was to account
for the very existence of conversational implicatures, using PCIs as an
example. They are a good starting point, as there does not seem to be
any way of accounting for them in linguistic terms. A first possibility
is that speakers use PCIs because there is no way to communicate
explicitly what it is they want to communicate. Based on the distinc-
tion between weakly and strongly communicated implicatures, this
would apply to weakly communicated implicatures (for instance,
creative metaphors), but clearly not to strongly communicated
implicatures. The next question is then whether either Grice or rele-
vance theory can explain why speakers use strongly communicated
implicatures rather than explicit utterances. It turns out that nei-
ther can. Additionally, the Gricean account fails to account for the
recovery of implicatures when there is no way to assume that the
speaker complied with the Cooperation Principle. In a case where
a deceptive implicature is produced without securing the hearer’s

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004
87

Particularized Conversational Implicatures 87

belief, the Gricean account fails. This kind of example has the added
interest of indicating a potential answer to the question of why
speakers use implicit rather than explicit communication: in such a
case, the speaker can (legitimately) deny having lied. We then turned
to Pinker’s analysis, which is based precisely on the idea that the
reason why speakers use implicatures is to be able to deny having had
the intention of communicating them. In other words, the speaker
of an implicature does not commit herself in the way she would if
she had communicated the same content explicitly. This is precisely
what allows her to avoid epistemic vigilance, the set of mechanisms
hearers developed to eschew deception, manipulation and argumen-
tation. Additionally, she preserves her reputation for trustworthiness
at the same time.
Thus, implicit communication arises either when the speaker has
no way of communicating the same content explicitly, or when the
speaker wants to escape her hearer’s epistemic vigilance or both. This
is true for GCIs as well as for PCIs.

DISCUSSION QUESTIONS

• Why is Grice’s analysis of metaphors as blatant violations of the

maxim of Quality problematic?
• What is the difference between a strongly communicated and
a weakly communicated implicature, and why is an analysis of
metaphor based on this distinction a better account of creative
metaphors?
• What are the advantages of including the notion of plausible
denial in a theory of implicit communication?
• What is the difference between off-the-record and on-the-record
indirect speech acts?

SUGGESTED READINGS

Sperber and Wilson (1986/1995), Chapter 4, is the main source

regarding relevance theory’s criticism of Grice’s analysis of metaphor.
Regarding Pinker’s theory of implicit communication, see Pinker
(2007), Chapter 8. One of the most influential accounts of politeness
in pragmatics remains Brown and Levinson (1987). The notion of epi-
stemic vigilance was first introduced by Sperber et al. (2010).

Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 15:18:00, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316410875.004