6Understanding implicit meaning understanding
Abstract: Human communication has a unique feature: speakers do not need to encode the whole set of representations they may want to convey; rather, they can use linguistic expressions as evidence for the intended message and rely on the hearers’ inferential abilities to include some extra content during utterance interpretation. Implicatures are additional assumptions communicated by the speaker in a non-overt way; they are independent from the explicitly communicated content and cannot be predicted from the sentence meaning alone. This chapter reviews the main approaches to implicit meanings and to the inferential processes involved in implicit meaning understanding, including kinds of inference patterns and attribution of intention. The role of implicit meaning in social interaction, particularly in politeness, is also considered.
Keywords: Implicature, inference, maxims, heuristics, defeasible inference, attribution of intentions
1Human communication: Using symbols as indexes
Verbal communication is usually seen as a process in which a message is transmitted and interpreted thanks to the existence of a shared linguistic code: the sender encodes her message into a conventional signal and the receiver decodes it by using his knowledge of the same code. There are, however, a number of facts that cannot be easily explained in these terms. Consider the following situations:
| (1) | [Reading a newspaper]: – There is nothing on TV tonight. |
| (2) | [Looking at a woman that has just entered the room]: – The boss. |
| (3) | [Customer to shopkeeper]: – Is this salami good? [Shopkeeper to costumer]: – We sell only the best, madam. |
| (4) | [Talking about a new co-worker]: | – How is he doing? |
| – It’s not for me to say. |
In situation (1), what the addressee will understand is not that TV stations are not working, but rather that the programmes announced do not attract the speaker’sinterest at all; and this is in fact what the speaker wanted to communicate. The word nothing is interpreted in the more restricted sense shown in (5):
| (5) | There is nothing [INTERESTING] on TV tonight. |
Similarly, the speaker in (2) does not intend to merely produce an English noun phrase, but rather to transmit a particular, more complex message: for instance, that this woman (unknown to the hearer) is her boss, or that the boss (whom they both know) has arrived:
| (6) | [THIS WOMAN IS ] the boss. / The boss [HAS ARRIVED .] |
In (3) the costumer will understand that the shopkeeper wants to communicate that the product the lady was asking for is of first-class quality. This is not, however, what he overtly encodes: instead of a specific answer, he provides a general statement about all his products, in the confidence that the addressee will immediately reach the intended conclusion:
| (7) | We sell only the best, madam. | |
| a. | [→ WE SELL THIS PRODUCT .] | |
| b. | [→ THIS PRODUCT IS TOP-QUALITY -.] | |
Finally, the reply in (4) is not merely a refusal to answer; rather it suggests that, for different reasons, the addressee is not in a position to voice an opinion, or that he is not feeling comfortable with that question – for example, because he does not want to get the other person in trouble, etc. –, or he may even intend to imply that he has a strong negative opinion:
| (8) | It’s not for me to say. | |
| a. | [→ I FEEL UNCOMFORTABLE .] | |
| b. | [→ I DON’T WANT TO GET HIM IN TROUBLE .] | |
| c. | [→ I HAVE A STRONG NEGATIVE OPINION .] | |
These examples illustrate a phenomenon pervasive in human communication: that the overt linguistic signal need not fully encode all the speaker wants to communicate. The representations in (5)–(8) are surely closer to the intended message, but the extra contents in small capitals are not obtained by decoding. Other cognitive processes are crucially involved, such as those which combine information from a variety of sources, whether of a general or specific kind, with the linguistically decoded content. Whilst a language can be analysed as a complex code – a recursive system that allows the unlimited combination of symbols –, there is much more to communication than the use of a linguistic system: for the speaker, communication is not achieved as a result of a mechanical symbolic activity, but rather by providing clues of her intended message; and for the hearer, linguistic expressions are but a clue to establishing an inferential causal connection to reconstruct the intended message (Sperber & Wilson 1986/1995; Sperber 1994, 1995). Symbols are, then, used as indexes: this feature is unique to human communication.
2Explicit or implicit?
The representations in (5)–(8) are thus the result of processes that go well beyond linguistic decoding. In principle, one would say that the additional material in small capitals is implicit content, as opposed to what is explicit, i.e. overtly said. This is indeed the spirit in which the term implicature was first coined by Grice (1975: 24) to refer to whatever is meant but left unsaid. In his account, implicatures contrast with ‘what is said’, roughly the proposition expressed by an utterance. In determining ‘what is said’, only conventional, encoded meanings are taken into consideration (together with minor operations of disambiguation and reference resolution to obtain a truth-evaluable proposition). The gap between said and meant is filled in by resorting to some general principles (cf. § 4) that explain why and how implicit meanings are derived.
Not all researchers, however, agree with Grice’s version of the divide between explicit and implicit meanings. In fact, it is easy to notice that the examples in (5)–(8) are not all of a kind. Consider (5)–(6) first. The extra constituent in (5) is a specification, or expansion of a linguistic form; in (6), a full proposition has been built up on the basis of a non-sentential fragment. In both cases, the addressee adds new material to complete apparently “missing” constituents of what the speaker wanted to express in a direct way. The resulting enriched representations are inferential developments of the encoded meaning. These are called implicitures by Bach (1994, 2006) and explicatures by Sperber & Wilson (1986/1995), Carston (1998, 2002, 2004) and Carston & Hall (2012). Bach’s term emphasizes the non-overt nature of the added constituents, whereas the term explicature stresses their radical difference with respect to other kinds of implicit meanings.
The main argument for considering that the representations in (5)–(6) fall on the explicit side is based on the fact that it is the representations in (5)–(6), and not those in (1)–(2) that are judged as true or false. If indeed one were to disagree with the speaker, the target proposition should include the additional material, as shown in (9) and (10).
| (9) | Well, there is a documentary on volcanoes in channel 17 that could be interesting. |
| (10) | She is not the boss; she is just the manager. |
Either way, the representations in (5)–(6) are one step further from the Gricean notion of ‘what is said’ (Grice 1975; Récanati 1989), but still halfway between overtly coded meanings and other kinds of implicit meanings (namely, implicatures; see below).
The relationship between (1)–(2) and (5)–(6) is so close that some scholars (Stanley 2000, 2002) have suggested that the corresponding syntactic structures actually include tacit (i.e. non-overt) constituents, to which the addressee should assign a value. In this view, the pragmatic processes of expansion and completion are linguistically mandated. This move tries to reduce the gap between what is encoded and what is interpreted, at the expense of a more complex syntactic representation. But even in this approach, a pragmatic phase of enrichment is inevitable. Others (Stainton 2004, 2006; Barton & Progovac 2005; Hall 2009), in contrast, argue for a simpler syntactic representation and leave the bulk of the explanation to pragmatic operations that develop, complete and enrich the encoded meaning. This latter view is consistent with economy considerations that warn against positing non-overt categories when independently motivated alternatives can offer an adequate explanation.
As for the representations in (7)–(8), they are not obtained by developing or enriching the encoded meaning; rather they contain additional independent assumptions. These new propositions, which are communicated by the speaker in a non-overt way (i.e. they are not linguistically encoded), are known as implicatures (Grice 1975; Sperber & Wilson 1986/1995; Levinson 2000; Horn 2004; Carston & Hall 2012). They are assumptions that the speaker is responsible for having conveyed; and the addressee has to retrieve (some of) them from previous knowledge or build them up specifically for the occasion, guided by the need to make sense of the explicitly communicated meaning. Implicatures can be conceived of as either premises or conclusions of a reasoning process. For instance, the shopkeeper’s answer in (3) explicitly provides the major premise of an incomplete syllogism (enthymeme), with the presumption that this will trigger the inferential retrieval of the minor premise (7a) in order to draw the intended conclusion (7b). Both (7a)– (7b) are implicitly communicated propositions, hence implicatures. In (8), the utterance in fact evades the question – which most likely will set the addressee wondering why and prompt him to induce an explanation: the propositions in (8a)– (8d) are tentative, provisional hypothesis.
3The properties of implicit meanings
From the previous considerations, it can be seen that implicatures are not intrinsic properties of linguistic expressions per se; rather, it is the speaker, not the sentence that implicates something in a context (though see § 4.2 for a possible exception). In fact, any sentence can be used to convey an infinite variety of different implicit meanings depending on the speaker’s intention and the situation. For instance, the speakers in (1)–(2) could implicitly communicate propositions like those in (11)–(12) on different occasions:
| (11) | There is nothing [INTERESTING] on TV tonight. | |
| a. | [→ I WANT TO GO OUT ] | |
| b. | [→ WE CAN PLAY CHESS ] | |
| (12) | [THIS WOMAN IS ] the boss. / The boss [HAS ARRIVED .] | |
| a. | [→ WE SHOULD CHANGE THE SUBJECT .] | |
| b. | [→ GO AND INTRODUCE YOURSELF ] | |
| c. | [→ ASK HER ABOUT YOUR PAY-RISE -] | |
Another fact about implicatures is that they can come with different degrees of strength (Sperber & Wilson 1986/1995, 2005). Consider again the examples in (7)–(8). In (7), the link between the customer’s question and the shopkeeper’s answer cannot be established unless the assumption in (7a) – the minor premise – is recovered, making it possible to obtain the implied conclusion in (7b). The question-answer pair constrains the inferential derivation of the implicatures in such a strong way that the speaker is surely responsible for intending to communicate those two precise assumptions. In (8), in contrast, the situation does not restrict the interpretation in this way, and the utterance is open to different interpretations – all of them plausible explanations for the addressee’s refusal to answer. What is at issue here is not the (in)ability of the hearer to work out the speaker’s intended meaning, but rather the fact that maybe she does not intend to convey a specific proposition, but rather to communicate a weaker, vaguer and more nebulous impression, which any (subset) of the possible assumptions would successfully convey. In fact, any utterance can carry strong and weak implicatures at the same time. Consider (3) again. In addition to the strong implicatures in (7), a wide range of weak implicatures could also be intentionally communicated:
| (13) | a. | YOU SHOULD BUY THIS SALAMI . |
| b. | OURS IS A RELIABLE BUSINESS . | |
| c. | WE ARE TRUSTWORTHY . | |
| d. | YOU SHOULD BUY HERE . | |
| e. | DO NOT BUY AT THE SHOP AROUND THE CORNER . |
In section 1 we showed that human communication typically uses symbolic expressions as evidence for the intended message. But why should this be so? The previous discussion about strong and weak implicatures provides the key to the answer. By encoding one sentence, the shopkeeper in (3) is communicating not the single proposition he overtly expresses, but a whole set of assumptions –some very strongly, others in a much weaker way. An exponential growth of information is thus obtained, which definitely contributes to the economy of the overall process: from the side of the speaker, his coding effort yields a higher number of extra communicative effects; from the hearer’s side, her interpreting effort is rewarded by returning a wide range of assumptions.
How are implicatures recognized? The most reliable and classical test to identify inferred content is cancellability (Grice 1975; Levinson 2000; Ariel 2010; see section 4.1 for further criteria; cf. Grice 1975; Sadock 1978). Unlike coded meanings and entailments, implicit content can be cancelled without contradiction. For instance, upon hearing the utterance in (1), the addressee can easily infer that the speaker meant (11a). However, this interpretation can be easily overridden by the speaker herself if she adds … but I’m staying at home anyway. Similarly, if the shopkeeper in (3) pointed out that that salami is an exception to their general rule of selling only the best products, this remark could sound conversationally odd, but surely it does not represent a contradiction. Cancellation of (easily predictable) implicit content is in fact a major strategy in humour, as illustrated in some famous quotations by Groucho Marx:
| (14) | a. | I never forget a face, but in your case I’ll be glad to make an exception. [→ I WILL NEVER FORGET YOUR FACE .] |
| b. | I’ve had a perfectly wonderful evening. But this wasn’t it. [→ THIS HAS BEEN A WONDERFUL EVENING .] |
|
| c. | From the moment I picked up your book until I laid it down, I was convulsed with laughter. Someday I intend reading it. [→ READING YOUR BOOK MADE ME LAUGH .] [→ YOUR BOOK IS WONDERFUL .] |
4Accounting for implicit meaning
In the examples considered so far implicit meanings have been listed and paired up with different situations. But why do implicit meanings arise at all? How do they get communicated? How are they understood? Pragmatic theories have as one of their main goals to provide an adequate answer to these questions. To succeed, they first have to find a solution to the problem of the one-off nature of communicative events: since every act of communication is singular (i.e. the participants and their circumstances necessarily change on each occasion), how is it possible to obtain significant generalizations, as required by scientific explanation?
Although it is true that the specific content of an implicature cannot be predicted from the sentence uttered, this does not mean that the whole process is random. We should, then, look for general principles operating on the process itself, rather than its results. This can be done in two different ways: either by determining what the specific principles that govern communication are, or by analysing how the different cognitive subsystems involved in communication work and interact with each other. These two routes define the two most influential theoretical accounts of human communication and implicit meaning understanding: the maxim-based approach stemming from Grice’s proposals, together with other post-Gricean approaches, and the cognition-oriented approach, with Sperber and Wilson’s work as its main inspiration (see also Borg 2009).
4.1Grice’s maxim-based approach
The first modern and systematic account of implicit meanings in conversation was that of Grice (1975, 1989). He argued that the gap existing in utterance interpretation between ‘what is said’ and what the speaker meant can be explained by assuming that conversation has its own logic and a specific rationality principle that governs all forms of interactive behaviour is at work. The principle reads as follows:
Cooperative Principle: Make your conversational contribution such as is required, at the stage of the conversation at which it occurs, by the accepted purpose of direction of the talk exchange in which you find yourself. (Grice 1975: 45)
This principle is further developed in four maxims:
| – | Maxims of Quantity: | |
| 1. | Make your contribution as informative as is required. | |
| 2. | Do not make your contribution more informative than is required. | |
| – | Maxims of Quality: Try to make your contribution one that is true. | |
| 1. | Do not say what you believe to be false. | |
| 2. | Do not say that for which you lack adequate evidence. | |
| – | Maxim of Relation: Be relevant. | |
| – | Maxims of Manner: Be perspicuous. | |
| 1. | Avoid obscurity of expression. | |
| 2. | Avoid ambiguity. | |
| 3. | Be brief. | |
| 4. | Be orderly. | |
| (Grice 1975: 45–46) | ||
Despite their imperative formulation, the Cooperative Principle (CP) and the maxims are not external rules that speakers should follow; rather, they represent a set of general, descriptive standards by which communicators, as rational agents involved in an intentional joint activity, are supposed to abide by default. It is precisely on the presumption that participants are being cooperative (in this specific sense) that communication works.
The explanatory power of Grice’s CP lies precisely in that it makes it possible to infer the relevant speaker’s intentions. In fact, for Grice implicatures are fully calculable at every step (criterion of calculability). The interpretation proceeds according to the following general pattern:
“He has said that q; there is no reason to suppose that he is not observing the maxims, or at least the Cooperative Principle; he could not be doing this unless he thought that p; he knows (and knows that I know that he knows) that I can see that the supposition that he thinks that p is required; he has done nothing to stop me thinking that p; he intends me to think, or is at least willing to allow me to think, that p; and so he has implicated that p.” (Grice 1975: 31)
It is then by supposing that the CP and the maxims are in operation that implicatures are calculated. It is worth noticing that this maxim-based approach defines how specific interpretive effects are obtained on the basis of the contents of both the encoded meaning and the maxims. The implicatures in (11)–(12), for instance, are obtained by taking for granted that the information the speaker provides is relevant for the addressee (to carry out some joint activity or to take a certain course of action).
Most interestingly, implicatures arise also when the maxims are apparently violated. The examples in (7)–(8) are a case in point. In (7), the connection between the customer’s question and the shopkeeper’s answer can be established only if one assumes that, though he does not provide a direct answer, he is nevertheless being cooperative and therefore his reply is actually relevant to the conversational exchange: to achieve this, some propositions have to be added. The example in (8) illustrates how an apparent refusal to offer the information requested can give rise to extra interpretive effects, as a result of the hearer wondering why the speaker is reluctant to provide a direct answer.
The implicatures considered so far are crucially dependent on the particular context and the conversational settings. In fact, it is not difficult to imagine that different implicit meanings would be conveyed on different occasions (criterion of indeterminacy). These implicatures are therefore called particularized conversational implicatures (PCIs).
Other implicit components seem to be less sensitive to the context. Consider the example in (15a), which we tend to understand as conveying (15b–c):
| (15) | a. | She met the man of her life and got married. |
| b. | [THE EVENT OF MEETING THE MAN IS PREVIOUS TO THAT OF GETTING MARRIED ] | |
| c. | [THE MAN SHE MARRIED IS THE MAN SHE MET .] |
In the Gricean account, the inferred components are obtained again by assuming that the maxims are operating. Other things being equal, the order of the propositions is supposed to reflect the order of events (4th submaxim of order); and the man she met is inferentially identified with the man she married by the maxim of relation. These inferred meanings can be cancelled (criterion of cancellability), as shown in (16):
Implicatures like the ones in (16) that depend on the operation of the CP and the maxims, but not so much on the particular discourse situation, are called generalized conversational implicatures (GCIs). Notice, incidentally, that these non-overt meanings – implicatures in Grice’s view – are part of the truth-conditional, explicit content for other researchers (Bach 1994, 2006; Carston 1998), who see them as direct inferential developments of the linguistically encoded meaning (and not as independent propositions). In this latter view, the resulting representation is rather that in (17), where some components have been added that enrich the expressed meaning:
| (17) | Shei FIRST met [the man of herI life]j and AFTERWARDS got married TO HIM. |
4.2Neo-Gricean heuristics
A great deal of recent work in Pragmatics is a revision, or a further refinement of Grice’s original insights. To avoid the methodological problems of describing one-off interpretations, most research in the neo-Gricean tradition has focussed on generalized conversational implicatures, as representing safe invariants on the implicit meaning side.
Horn (1984, 2004), for example, has suggested keeping the maxim of Quality as a general truthfulness requirement and reducing the rest of the Gricean maxims and submaxims to two more encompassing principles:
Q-Principle (Hearer-based): Say as much as you can (modulo Quality and R)
R-Principle (Speaker-based): Say no more than you must (modulo Q)
The Q-Principle conflates Grice’s first maxim of quantity and the first two maxims of manner and it is responsible of implicit meanings like the ones in (18)–(19):
| (18) | Some students passed the exam. [→ NOT ALL STUDENTS PASSED THE EXAM .] |
| (19) | John entered a house. [→ THE HOUSE WAS NOT HIS OWN .] |
In (18), the implicature arises as a consequence of the fact that the expressions <some, all> can be seen as forming a scale, so the assertion of the weaker term some implicates the negation of the stronger term all. Notice that this is not a (logical) entailment, since some is logically compatible with all: in fact, the inferred meaning can be easily cancelled by adding ... in fact they all did. The implicature in (19) arises in quite a similar way: by asserting a weaker claim (namely, that John entered a non-specific house), the speaker is implicitly negating a stronger, more detailed assertion (for instance, that this was John’s own). These implicit meanings – a subclass of generalized conversational implicatures – are called scalar implicatures.
The R-principle collects the maxim of Relation, the 2nd maxim of Quantity and the last two submaxims of Manner. It is a version of the Least Effort Law on the speaker’s side. Consider the example in (20):
| (20) | a. | John lost a book. |
| b. | [→ THE BOOK WAS JOHN ’S.] |
The use of a non-specific form (a book) counts as an invitation to the hearer to infer a more detailed and stereotypical relation – here, between an individual and a book. In this case, the use of a weaker form does not imply that a stronger meaning has to be negated, but quite the contrary: the less informative expression calls for a specification.
These two principles represent opposing, antinomic forces; however, no specific criterion is provided to decide which one will prevail on each occasion and why. In (19) and (20) the linguistic context is almost the same, but the principle at work is different for each case: what would prevent the hearer from applying them the other way and getting a non-intended interpretation?
Levinson’s (1987, 2000) work is an attempt to provide a more detailed and elaborated answer. He concentrates on GCIs as cases of preferred, default interpretations, those not depending on world knowledge or on specific contextual factors (and hence general enough to be worth scientific endeavour). The main idea is that GCIs such as the ones in (18)–(19), which seem to arise in a systematic way, are the result of general inferences made on the basis of the normal use of language. According to Levinson, in addition to the level of sentence meaning (as linguistically encoded) and speaker meaning (as determined by her intentions), there is an extra intermediate level of utterance-type meaning, where CGIs belong. This view represents a radical departure from other accounts in the sense that implicit meaning is attributed both to speakers (PGIs) and to utterances (CGIs).
Levinson’s proposal is articulated around three principles, which determine three legitimate patterns of pragmatic inference (heuristics):
Q-heuristic: What isn’t said, isn’t.
I-heuristic: What is expressed simply, is stereotypically exemplified
M-heuristic: What is said in an abnormal way isn’t normal
(Levinson 2000: 31–33)
The first heuristic is equivalent to Horn’s Q-principle, and is responsible for scalar implicatures, such as the one in (18), but also for other implicit meanings that crucially depend on the existence of scales or sets with different alternatives:
| (21) | a. | John tried to open the file. |
| b. | [→ JOHN DID NOT SUCCEED .] |
| (22) | a. | John`s shirt is blue. |
| b. | [→ JOHN’S SHIRT IS NOT BED; JOHN’S SHIRT IS NOT YELLOW …] |
From (21a) it can be inferred by default that John’s attempt to open the file was unsuccessful. This is due to the existence of a scale <succeed, try>, which licenses the implied move from try to not succeed. The implicatures in (22b) derive from (22a) by virtue of the existence of a set <red, orange, yellow, green, blue, indigo, violet> from which only one of its members has been selected, thus implicitly negating all the rest.
The I-heuristic tries to capture the fact that minimally specified expressions tend to get maximal stereotypical interpretations. This strategy is at work in (20) and also in (17), where it makes it possible to account for both the temporal ordering of events and the preference for local coreference in the interpretation of pronouns and other nominal expressions. A further instance of the operation of this principle can be illustrated by the phenomenon of ‘conditional perfection’. As Geis and Zwicky (1971) noted, the conditional in (23a) tends to be strengthened to the biconditional in (23b), thus giving rise to the “invited inference” in (23c), which speakers derive by default unless otherwise indicated by the context:
| (23) | a. | If you mow the lawn, I’ll give you 20 A. |
| b. | [→ IF, AND ONLY IF , YOU MOW THE LAWN , I’LL GIVE YOU 20 €.] | |
| c. | [→ IF YOU DON ’T MOW THE LAWN , I WON’T GIVE YOU 20 €.] |
Finally, the M-heuristic, which is reminiscent of Grice’s maxims of Manner, can account for implicatures derived from the use of marked expressions. Consider the examples in (24)–(25):
| (24) | John stopped the car. |
| (25) | a. | John caused the car to stop. |
| b. | [→ THE CAR STOPPED IN A NON-NORMAL WAY .] |
The difference between them is that the latter conveys a default implicature about the unusual way in which John stopped the car; this inference is part of the preferred interpretation in (25), but not in (24). Of course, nothing prevents (24) from receiving an interpretation along the lines in (25b), but for this to be possible it would be necessary that other contextual factors converge; in (25a), in contrast, the implicature obtains by default.
To avoid circularity in the application of the various heuristics, Levinson suggest a ranking of the principles: Q takes precedence over M, which in turn takes precedence over I. This ordering is partially motivated by the different content of each principle: Q and M both involve linguistic forms – more specifically, the existence of alternative or contrasting expressions –, whereas the I-principle, which leads to stereotypical interpretations, is linked to world knowledge. The alternatives at the basis of Q inferences are semantic in nature (i.e. the contrast arises between semantically weak and strong forms), whilst those in M are related to form rather that to content (i.e. between different forms for the same content).
The approaches considered so far have as their main goal to identify the principles that seem to be at work in the production and interpretation of implicatures. These principles are content-sensitive in the sense that they apply to utterances depending on their formal and semantic properties and other aspects of the conversational settings. This is not, however, the only way to account for implicatures, as we will see in the next section.
4.3Cognition-oriented approaches
Cognition-oriented approaches exploit a radically different strategy: they no longer search for rules, principles or heuristics that can help to bridge the gap between what is said and what is implicitly conveyed; rather, they try to look for generalizations about how the human mind works. The starting hypothesis is, then, quite simple: evolution has endowed our species with dedicated processing devices that impose particular restrictions on how we interact with the environment, including communicative behaviour. Exploring the design architecture of the human mind is thus an adequate way to discover significant generalizations for communication.
The most well-developed and influential approach in this perspective is Relevance Theory (RT; Sperber & Wilson 1986/1995; Wilson 1994; Wilson & Sperber 2004, 2012). Their proposal is based on two specific claims:
First, or Cognitive Principle of Relevance: Human cognition tends to be geared to the maximisation of relevance.
Second, or Communicative Principle of Relevance: Every act of ostensive communication communicates a presumption of its own optimal relevance.
(Sperber & Wilson 1995: 260–261)
The first principle is a generalization about the human mind: since our processing capacities are limited, during the course of evolution our mind has developed dedicated mechanisms that tend to allocate processing resources to the inputs that seem more relevant and to process them in the more efficient way. Relevance is technically defined as a trade-off between cognitive effects and processing effort. This does not mean, however, that individuals should have a conscious, a priori representation of what is relevant, or a clear estimation of what the costs and benefits would be; nor is there any external, comparative measure for them. If so, how can our minds decide? This is precisely where the second principle comes into play. It establishes that, among all possible inputs and stimuli, communicative behaviour comes with a built-in guarantee of optimal relevance. Notice that this is not an external stipulation: after all, if the speaker wants to get her message across, it is in her own interest to make it clear enough for the audience to interpret it in the most efficient (i.e. quick and accurate) way (see Sperber, Cara & Girotto 1995; Van der Henst & Sperber 2004 for experimental work supporting this view). Of course, a certain speaker can fail to be relevant for a certain addressee on a particular occasion, but this fact does not cancel the presumption of relevance that comes with every act of communication and drives the interpretative process.
Given that linguistic expressions are but a clue to the intended message, utterance interpretation is mainly a process of inference to find the causal link between them. The hearer’ goal is to infer what message (or, more precisely, what set of assumptions) the speaker wanted to convey by means of her encoded signal. It is precisely the presumption of optimal relevance (i.e. that the stimulus offers the most efficient balance between cognitive effects and processing effort, given the preferences and abilities of the speaker) that guides the hearer during interpretation. If relevance is given, then the cognitive systems involved in utterance interpretation will do whatever is needed (though not more) to satisfy this expectation regarding cognitive effects; as for processing effort, efficiency requires minimizing costs and this minimization is achieved by considering the most available options first. Thus, any instance of communicative behaviour will systematically trigger a general comprehension procedure that guides the hearer in the process of inferring the speaker’s meaning:
Relevance-theoretic comprehension procedure
| a) | Follow a path of least effort in computing cognitive effects. In particular, test interpretive hypotheses (disambiguations, reference resolutions, implicatures, etc.) in order of accessibility. | |
| b) | Stop when your expectations of relevance are satisfied. | |
| (Sperber & Wilson 2002: 13) | ||
In this way, RT tries to set a deeper, psychologically motivated foundation for the interpretive operations carried out during utterance interpretation. The principles of relevance are descriptive generalizations, or laws about how human cognition works and are thus quite different from Gricean and neo-Gricean principles and heuristics: in RT there is nothing like following or violating the maxims, nor is there any possible conflict among them that could require establishing precedence relations. As occurs with the functioning of other biological systems, communicators cannot decide whether to activate them or opt out. Only considerations of relevance guide the hearer in the search for a consistent explanation. The principles are always in force, without the individuals being conscious of the operational details of their internal mechanisms.
From the point of view of RT, then, there is no need for specific rules for interpreting an expression with certain properties. It may well be that an utterance containing the word some implicates not all in a particular situation, but there is no reason why it should necessarily be the case, nor is this implicature dependent on the existence of a specific, default principle to this effect (Carston 1998; Carston & Hall 2012). Some experimental work has also cast some doubts on the default status of scalar inferences (Breheny, Katsos & Williams 2006: Noveck & Sperber 2007).
5Cognitive processes in implicit meaning understanding
The cognitive turn in Pragmatics has brought findings in Psychology, Cognitive Neuroscience, Evolution and Artificial Intelligence to the foreground that can offer fruitful insights when combined with linguistic and philosophical considerations in an interdisciplinary approach. Grice’s work represented a turning point after which it is commonly assumed that recovering implicit meanings is a process crucially dependent on the speaker’s intentions, on the one hand, and on the integration of the information linguistically conveyed with other pieces of (background) information, on the other. Two main cognitive abilities are then involved in utterance interpretation: that of attributing intentions to the speaker and that of selecting contextual information and combining it with the meaning obtained by decoding. The next two sections provide a brief overview of the key processes and mechanisms from a cognitive perspective.
5.1Inferential processes
Utterance interpretation can be seen as a matter of combining information from different sources (linguistic decoding, general world knowledge, background information) to identify and recover the set of assumptions that the speaker intended to communicate. Explaining the way in which this is achieved is the goal of a universal theory about how information is selected, combined and processed in our minds.
In principle, there are no limits to the class or the amount of information that an individual can bring to bear in utterance interpretation. However, if this is so, the interpretive process would be an “intractable” problem from a computational point of view, i.e. one that would require a never-ending process of computing an infinity of variables. This seems far off from human processing capacities, which are known to be limited and slow. And yet we humans are usually very efficient in utterance interpretation: most of the times, we reach very accurate interpretive hypotheses in a very quick way almost instantaneously. This suggests that our processing mechanisms have found a way to circumvent that problem.
5.1.1Defeasible inference
The word usually holds the answer. We know that sometimes hearers make the wrong guess and misinterpret the speaker’s words: surely, this is something we have all experienced many times. This shows that the interpretation is, at most, usually right, but the process is not failsafe and does not per se guarantee the output (i.e. it is a non-demonstrative process). How is then spontaneous inference to be modelled? There are various ways. Sperber & Wilson (1986/1995, 1987) advocate for a view in which utterance interpretation is indeed a non-demonstrative process, but it can contain deductive chains as sub-parts. Consider again the example in (1). The explicitly communicated assumption (repeated here as (26a) combines with an implicit premise in the form of a conditional claim ‘If there is nothing on TV, then we should go out’ to yield the conclusion ‘We should go out’ by using the modus ponens inference pattern in (27):
| (26) | If there is nothing on TV, then we should go out. There is nothing on TV. Therefore, we should go out. |
| (27) | If p, then q p therefore, q |
(Conditional claim) (Antecedent) (Consequent) |
Modus ponens is a deductive form of reasoning and the argument is valid, though this does not guarantee the truth of the conclusion: for the argument to be sound, all its premises must be true. So, if the conditional claim is false, the conclusion still obtains, but it will be false as well.
An alternative account can be offered in terms of abductive reasoning. Abductive inferences are targeted towards finding an explanation for a given phenomenon; they are present both in scientific and in everyday reasoning (Hobbs 2004; Lipton 2008; Douven 2011). When the hearer of the example in (1) concludes that his partner is implicating that she would like to go out, he is trying to make sense of the utterance by establishing a hypothetical causal connection between the overt signal and the speaker’s intended meaning. This conclusion is only plausible (not logically necessary), but is compatible with what he knows about his partner and the current situation. Abductive inferences are quick, automatic, efficient and – crucially – defeasible and retractable. For instance, if the hearer in (1) comes to know that the speaker has got a headache, this new piece of information will possibly make him discard his previous conclusion (that the speaker intends her utterance to be taken as a suggestion to go out) in favour of a different one. This is so for two reasons: first, because the conclusion was merely tentative and did not necessarily follow from the premises; and second, because the conclusion was drawn on partial and incomplete information: as soon as new information is added, things can change and the hearer should be prepared to do so if enough evidence is provided. Defeasibility and retractability thus relate to cancellability (the main test for inferred meaning) in a natural way. The fact that an implicit content can be easily cancelled by the speaker herself indicates that this content is not a necessary consequence of the encoded meaning. When an implicature is cancelled the conclusion is discarded and a new inferential process is activated to find an alternative, consistent explanation.
5.1.2Bounded rationality
The inference from (1) to (11) is reached in a fast and automatic way: the hearer does not first gather all the information that could possibly be relevant to the matter under consideration and then compare all possible conclusions to select the best candidate; he merely considers the first idea that comes to his mind (the standard If there is nothing interesting here, one goes elsewhere) and uses it to infer the implicit import of the utterance. The efficiency of spontaneous reasoning derives in part from not using all the information, but a small subset of assumptions only. This avoids the problem of computational intractability, but raises a new issue: how is this small subset identified and selected?
The answer comes from the theory of Bounded Rationality (Simon 1956, 1982; Gigerenzer & Selten (eds.) 2001; Gigerenzer 2000, 2007). The idea is that human behaviour cannot be explained as a comprehensive rational activity, given the limitations of time and of processing capacity of human minds. Human agents, however, appear to behave in a nearly optimal way, given their goals and the resources available. This is so because the members of our species are endowed with “fast and frugal” heuristics (Gigerenzer & Todd 1999; Goldstein & Gigerenzer 2002), an adaptive strategy that guides information search – how and where to search for information and when to stop searching – that is simple enough to operate efficiently under limited human capacities. In informal terms, the idea is to consider only easily available information and to stop searching as soon as a certain standard is met. But then, how does one decide what counts as easily available information? And what is the standard?
The fast and frugal strategy exploits two favourable cognitive facts. The first one has to do with the structure of knowledge. The information in the environment is not stored in our minds as an array of unrelated pieces, but organized in networks that establish precise relationships among agents, objects and events. These structures (called frames, scripts, schemata) have been implemented in computational terms by researchers in the Artificial Intelligence area (Minsky 1975, 1986; Rumelhart 1975; Shank & Abelson 1977). What is important is that these knowledge structures organize the information: when processing a specific concept, the most directly related concepts are activated as well, and hence are made easily accessible to working memory.
The second factor is the existence of a threshold of satisfaction. Simon (1956) observed that humans do not aim at making optimal choices (which, incidentally, would be out of our limited time and computational possibilities), but are satisfied by choosing an option that can make us happy enough: we tend to “satisficing”, i.e. to selecting the first choice that meets our needs or expectations, without taking extra time in “optimizing”, i.e. in further considering whether there could ever be a better or an optimal choice.
Taken together, the organization of the environment in knowledge structures and the tendency to “satisficing” provide an answer to the question of how interpreters identify and select the particular set of assumptions that are used in utterance interpretation. The hearer will search first the most easily available information – the information made directly accessible by the encoded meanings (including the scripts these open up) and the communicative situation; and he will stop the search as soon as his interpretive expectations are satisfied. Only when the most accessible information does not yield the expected result, can the search space be widened. Implicatures are precisely the small set of assumptions that have been used to obtain the satisfactory conclusion.
As the reader will have noticed, this proposal goes along the same lines as the relevance-theoretic comprehension procedure presented in § 4.3. The strategy, however, is not limited to utterance interpretation; it is general enough to guide any process of decision making and in fact has been successfully applied to account for understanding and choices in areas as disparate as political administration, justice, economics or therapy.
5.2Attribution of intentions
Understanding implicit meaning crucially involves identifying the speaker’s intention and recovering the set of assumptions she wants to convey. But can we read minds like this? Consider the following situation. A man takes a key out of his pocket near a door. Under standard conditions, everyone will assume that the man wants to open that door. This conclusion is obtained as the result of an inference to the best explanation using fast and frugal heuristics that exploits the organization of our world knowledge and chooses the first matching interpretation. It is, of course, defeasible: there are many other reasons why a man can take a key out of his pocket (for instance, he may merely want to know whether the key is the right one). But what is crucial to the present discussion is that all of the possible interpretations involve attributing an intention, a mental state to an unknown individual. Moreover, it is almost impossible for any of us to escape the tendency to interpret behavior in intentional terms. Why should this be so?
Unlike object motion, which is entirely dependent on external forces and physical laws, human intentional behavior is internally caused. Mental states are relations between an individual and the representation of a state-of-affairs. Positing mental states (wants, desires, beliefs, expectations, fears …) is a way to maintain mechanical links between an observable behavior and its cause (McCloskey 1983; Leslie 1994; Frith & Frith 2007). The ability to attribute mental states is called Mindreading.
Much experimental work has been devoted to analyzing the existence of a cognitive subsystem specialized in the attribution of mental states, its properties and its neuro-psychological underpinnings (Leslie 1987, 1994; Gopnik 1993; Baron-Cohen 1995; Apperly 2011). There is some debate about the most adequate way to explain our mind-reading abilities: either as a matter of theorising (Gopnik & Wellman 1994) or as the result of mental simulation (Davies & Stone 1995). Recent research has shown that this tendency is an adaptive ability of a social species like ours that increases the chances of survival of its members and makes interactions smoother and more effective (Tomasello 1999; Tomasello et al. 2005; Frith & Frith 2006; Hermann et al. 2007; Schaller et al. 2007). Autism spectrum disorders correlate with a severe deficit in all aspects of mindreading, particularly the inability to link communicative behaviour with its underlying intentions and hence the inability to identify implicit meanings (Frith, Morton & Leslie 1991; Happé 1993; Frith & Happé 1994).
Is the mechanism involved in the attribution of mental states during utterance interpretation the same that accounts for other instances of mindreading? Or rather is there a specific system exclusively devoted to communicative behaviour? From the perspective of Relevance Theory the second possibility has been clearly favoured: utterance interpretation requires inferential and representational abilities far more complex than those involved in other kinds of mindreading tasks (Wilson 2000; Sperber & Wilson 2002). This is an on-going debate, for which more empirical research is needed.
6Social aspects
6.1Politeness
Speakers use linguistic expressions as clues to their communicative intention, i.e. they leave some assumptions implicit and it is the hearer’s task to determine what his interlocutor wanted to convey. Among the reasons a speaker may have for leaving some assumptions implicit, we have mentioned above economy considerations: if she were to make explicit each and every little detail of her intended message, communication – if possible at all – would be totally inefficient. But there can be another powerful reason for not being totally explicit: politeness. This brings into the picture certain social aspects of communication.
Consider the classical example in (28):
| (28) | Can you pass the salt? |
Any native speaker of English would say that (28) is the normal way to issue a request to pass the salt. But is it? If we consider the linguistic form only, this utterance seems to be a question on the hearer’s ability to perform a certain action. However, it is true that, like many formulaic expressions, it is understood as having, in addition to its linguistic meaning, a standard or conventionalized interpretation that goes well beyond what is encoded.
Examples of this kind have been treated in the literature as indirect speech acts (Searle 1975) or as short-circuited implicatures (Morgan 1978). The idea behind the indirect speech act approach is that performing an act (in this case, asking) can count as an indirect way of performing another (requesting). The main reason invoked to account for this kind of strategy is politeness: instead of issuing a direct command (Pass me the salt!), the speaker resorts to the strategy of asking the hearer whether he would be able to do so. In this way, the hearer is left with the option of deciding whether or not to comply and the request will no longer be perceived as an imposition. There is a wealth of research dealing with the strategies that speakers can use to minimize the social risk inherent to some actions performed by means of language. The seminal work is Brown & Levinson (1978/1987), where a detailed theory is presented.
The approach in terms of short-circuited implicatures (Morgan 1978) puts the emphasis on the fact that the implicit intention, though calculable, is no longer calculated: a direct connection has been created between the linguistic form and the intended meaning, with no need for intermediate inferential steps. The interpretation has become standardized (Bach & Harnish 1979) or conventional. In fact, to interpret many of these idiomatic forms what is required is not the general knowledge of the rules of the language, but specific knowledge of the idiomatic convention (see Groefsema 1992 for critical assessment). Non-native speakers could possibly reconstruct the inferential path backwards, but this is not the way in which inferences work. It is not clear that in these cases one should still talk of implicit meaning.
6.2Humour
The examples in (14) have illustrated humorous effects related to the sudden and unexpected cancellation of easily derivable, standard implicit content. Implicatures have, in fact, a significant role in the account of humorous interpretations (Attardo 1993, 1994; Curcó 1995; Yus 2003; Higashimori 2008). The incongruity typically found in jokes can be based on conflicts between factual shared knowledge and the implicit premises the audience is driven to hold in the online process of interpretation. Consider the following example:
| (29) | Aunt (to niece): – Eat up your spinach, child, and you’ll grow up to be beautiful. Niece (to aunt): – Oh! Didn’t they have spinach in your day, Auntie? |
This joke is based on the combination of two inference patterns, in which several implicit premises have to be added. The first one is activated by the aunt’s utterance, which basically introduces a conditional claim ‘If p, then q’, as shown in (30a). The child then combines this premise with what she thinks to be a fact (30b), to draw the implicit conclusion in (30c):
| (30) | a. | If one eats spinach, one is beautiful. |
| b. | Auntie is not beautiful. | |
| c. | Therefore, Auntie didn’t eat spinach. |
This reasoning pattern is roughly an informal version of modus tollens, a rule of inference in which the negation of the antecedent licenses the negation of the consequent:
| (31) | If p, then q ~q therefore, ~p. |
The second pattern is activated by the niece’s question: she is trying to confirm her hypothesis about the fact represented by the conclusion in (30c), a fact that calls for an explanation:
| (32) | a. | Auntie didn’t eat spinach [= (30c)] |
| b. | If spinach exists, one eats it (to grow up beautiful). | |
| c. | They didn’t have spinach in Auntie’s days. |
The humorous effect arises as the result of being forced to introduce a number of rather bizarre implicated premises, namely (30b)–(30c) and (32b), to make sense of the relationship between the two conversational turns in the dialogue.
6.3Manipulation and advertising
The case of humour shows that audiences and hearers, as part of the utterance interpretation process, might be driven to construct assumptions they did not previously hold, or even they would plainly reject if explicitly presented with them. This can also be true of other communicative scenarios, from manipulative discourse to advertising.
Manipulation involves inducing the audience to inadvertently entertain false, or biased, representations. To succeed, the manipulative speaker’s actual intention must remain hidden for the audience; otherwise, they will not accept the game. As discussed in the literature (Saussure 2005; Maillat & Oswald 2011 and this volume, chapter 25), the strategy is to constrain and control the context selection process of an utterance in such a way that the hearer is inescapably led to accept some content that he would rule out under normal circumstances. This process crucially involves the implicit premises that the hearer has to use to obtain a relevant interpretation. A very simple example can illustrate this point.
In interpreting (33), the addressee is led to accept the implicit premises and the conclusion in (34):
| (33) | Only an idiot would do X. |
| (34) | a. | All people that do X are idiots. |
| b. | John has done X. | |
| c. | John is an idiot. |
Manipulation is an extreme form of persuasion. Other forms of persuasive communication, such as political propaganda and advertising, represent large scale attempts to spread some assumptions among a population, in order to fulfil the communicator’s intentions (Taillard 2004). One of the commonest strategies in persuasion and advertising is that of “massaging the message”, i.e. repeating it over and over to increase its acceptance by turning it into a piece of shared knowledge no one will contend.
Other techniques exploit the recovery of implicit content. For example, many products advertise near-total effectiveness with the ‘99% formula’, as in handsanitizers, birth-control methods, pregnancy tests, and network coverage, among others. Apart from the fact that the advertised figures may obtain under very specific lab testing conditions only, the major problem from a communicative point of view is that consumers tend to infer that their particular case will fall squarely under that favourable 99 %, without ever considering how significant could be the remaining, unfavourable 1 %. In the case of germ-killers, for instance, people tend to accept the total effectiveness without asking themselves whether what they want to kill is a germ, or rather a different kind of microbe, or one of the viruses the product does not kill (such as clostridium difficile, a gastrointestinal scourge, and the hepatitis A virus), or even to what extent the application procedure (including the application time) is relevant to the result.
These few examples thus show that the implicit assumptions brought to bear on utterance interpretation and the processes to identify them are important for a theory of verbal communication, but can also have far-reaching consequences for many aspects of our daily life.
References
Apperly, Ian. 2011. Mindreaders. The Cognitive Basis of ‘Theory of Mind’. Hove: Psychology Press. Attardo, Salvatore. 1993. Violation of conversational maxims and cooperation: The case of jokes. Journal of Pragmatics 19. 537–558.
Attardo, Salvatore. 1994. Linguistic Theories of Humor. Berlin: Mouton de Gruyter.
Bach, Kent. 1994. Conversational impliciture. Mind and Language 9. 124–162.
Bach, Kent. 2006. The top 10 misconceptions about implicature. In Betty Birner & Gregory Ward (eds.), Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn, 21–30. Amsterdam: John Benjamins.
Bach, Kent & Robert M. Harnish. 1979. Linguistic Communication and Speech Acts. Cambridge: MIT Press.
Baron-Cohen, Simon. 1995. Mindblindness. An essay on autism and theory of mind. Cambridge, MA: MIT Press.
Barton, Ellen and Ljiljana Progovac. 2005. Nonsententials in Minimalism. In Elugardo Reinaldo & Robert Stainton (eds.), Ellipsis and Nonsentential Speech, 71–93. Dordrecht: Springer.
Borg, Emma. 2009. On Three Theories of Implicature: Default Theory, Relevance Theory and Minimalism. International Review of Pragmatics 1(1). 63–83.
Breheny, Richard, Napoleon Katsos & John Williams. 2006. Are generalised scalar implicatures generated by default? An on-line investigation into the role of context in generating pragmatic inferences. Cognition 100. 434–463.
Carruthers, Peter & Peter Smith (eds.). 1995. Theories of Theory of Mind. Cambridge: Cambridge University Press.
Carston, Robyn. 1988. Implicature, explicature and truth-theoretic semantics. In Ruth Kempson (ed.), Mental Representations: the Interface between Language and Reality, 155–81. Cambridge: Cambridge University Press.
Carston, Robyn. 1998. Informativeness, Relevance and Scalar Implicature. In Robyn Carston & Seiji Uchida (eds.), Relevance Theory: Applications and implications, 179–236. Amsterdam: John Benjamins.
Carston, Robyn. 2002. Thoughts and utterances. The pragmatics of explicit communication. Oxford, Blackwell.
Carston, Robyn. 2004. Relevance theory and the saying/implicating distinction. In Laurence Horn & Gregory Ward (eds.), The Handbook of Pragmatics, 633–656. Oxford: Blackwell.
Carston, Robyn & Alison Hall. 2012. Implicature and explicature. In Hans-Jorg Schmid & Dirk Geeraerts (eds.), Cognitive Pragmatics, Vol. 4 of Handbooks in Pragmatics, 47–84. Berlin: De Gruyter Mouton.
Curcó, Carmen. 1995. Some observations on the pragmatics of humorous interpretations. A relevance-theoretic approach. UCL Working Papers in Linguistics 7. 27–47.
Davis, Wayne. 2012. Implicature. In Edward Zalta (ed.), The Stanford Encyclopedia of Philosophy. http://plato.stanford.edu/archives/fall2012/entries/implicature/ (accessed 18 March 2015).
Davies, Martin & Tony Stone (eds.). 1995b. Mental Simulation: Evaluations and Applications. Oxford: Blackwell.
Douven, Igor. 2011. Abduction. In Edward Zalta (ed.), The Stanford Encyclopedia of Philosophy. http://plato.stanford.edu/archives/spr2011/entries/abduction/ (accessed 18 March 2015).
Frith, Chris D. & Uta Frith. 2006. How we predict what other people are going to do. Brain Research 1079. 36–46.
Frith, Chris D. & Uta Frith. 2007. Social cognition in humans. Current Biology 17(16). 724–732.
Frith, Uta, John Morton & Alan M. Leslie. 1991. The cognitive basis of a biological disorder. Trends in Neuroscience 14. 433–438.
Frith, Uta & Francesca Happé. 1994. Autism: beyond ‘theory of mind’. Cognition 50. 115–132.
Gibbs, Raymond W. & Jessica F. Moise. 1997. Pragmatics in understanding what is said. Cognition 62. 51–74.
Gigerenzer, Gerd. 2000. Adaptive thinking: Rationality in the real world. New York: Oxford University Press.
Gigerenzer, Gerd. 2007. Gut feelings: The intelligence of the unconscious. New York: Viking Press.
Gigerenzer, Gerd & Reinhard Selten (eds.). 2002. Bounded Rationality. Cambridge: MIT Press.
Gigerenzer, Gerd & Peter M. Todd. 1999. Fast and frugal heuristics. The adaptative toolbox. In Gerd Gigerenzer, Peter M. Todd & ABC Group. Simple heuristics that make us smart, 3–34. New York: Oxford University Press.
Goldstein, Daniel G. & Gerd Gigerenzer. 2002. Models of Ecological Rationality: The Recognition Heuristic. Psychological Review 109(1). 75–90.
Gopnik, Alison. 1993. How we know our minds: The illusion of firstperson knowledge of intentionality. Behavioral and Brain Sciences 16. 1–14.
Gopnik, Alison & Henry M. Wellman. 1994. The theory theory. In Lawrence A. Hirschfeld & Susan A. Gelman (eds.), Mapping the Mind: Domain Specificity in Cognition and Culture, 257–293. Cambridge: Cambridge University Press.
Grice, H. Paul. 1975. Logic and conversation. In Peter Cole & Jerry L. Morgan (eds.), Syntax and Semantics, Vol. 3: Speech Acts, 41–58. New York: Academic Press.
Grice, H. Paul. 1989. Studies in the way of words. Harvard University Press.
Groefsema, Marjolein. 1992. ‘Can you pass the salt?’: A short-circuited implicature. Lingua 87. 103–135.
Hall, Alison. 2009. Subsentential utterances, ellipsis, and pragmatic enrichment, Pragmatics and Cognition 17(2). 222–250.
Happé, Francesca. 1993. Communicative competence and theory of mind in autism: A test of relevance theory. Cognition 48. 101–119.
Herrmann, Esther, Josep Call, María Victoria Hernández-Lloreda, Brian Hare & Michael Tomasello. 2007. Humans have evolved specialized skills of social cognition: The cultural intelligence hypothesis. Science 317. 1360–1366.
Higashimori, Isao. 2008. New perspectives on understanding jokes: A relevance-theoretic account. 龍谷大學論集 (Ryūkokudai Gaku ronshū-Ryukoku University Manabu) 471. 52–69.
Hobbs, Jerry R. 2004. Abduction in Natural Language Understanding. In Laurence Horn & Gregory Ward (eds.), The Handbook of Pragmatics, 724–741. Oxford: Blackwell.
Horn, Laurence. 1984. A new taxonomy for pragmatic inference: Q-based and R-based implicature. In Deborah Schiffrin (ed.), Meaning, Form and Use in Context (GURT ‘84), 11–42. Washington: Georgetown University Press.
Horn, Laurence. 2004. Implicature. In Laurence R. Horn & Gregory Ward (eds.), The Handbook of Pragmatics, 3–28. Oxford: Blackwell Publishing.
Leslie, Alan M. 1987. Pretense and Representation: The Origins of “Theory of Mind”. Psychological Review 94. 412–426.
Leslie, Alan M. 1994a. ToMM, ToBY, and Agency: Core architecture and domain specificity. In Lawrence A. Hisrchfeld & Susan A. Gelman (eds.), Mapping the Mind. Domain Specificity in Cognition and Culture, 119–148. Cambridge: Cambridge University Press.
Levinson, Stephen. 1987. Minimization and conversational inference. In Jef Verschueren & Marcella Bertuccelli (eds.), The Pragmatic Perspective, 61–129. Amsterdam: John Benjamins.
Levinson, Stephen. 2000. Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press.
Lipton, Peter. 2008. Inference to the best explanation. In Stathis Psillos & Martin Curd (eds.), The Routledge Companion to the Philosophy of Science, 193–202. London: Routledge.
McCloskey, Michael. 1983. Naive Theories of Motion. In Dedre Gentner & Albert L. Steven (eds.), Mental Models. Hillsdale, NJ: Erlbaum.
Maillat, Didier & Steve Oswald. 2011. Constraining context: a pragmatic account of cognitive manipulation. In Christopher Hart (ed.), Critical discourse studies in context and cognition, 65–80. Amsterdam: John Benjamins.
Merchant, Jason. 2004. Fragments and Ellipsis, Linguistics and Philosophy 27(6): 661–738.
Merchant, Jason. 2010. Three Kinds of Ellipsis. In François Récanati, Isidora Stojanovic, and Neftalí Villanueva (eds.), Context-Dependence, Perspective, and Relativity. Berlin: De Gruyter Mouton.
Morgan, Jerry L. 1978. Two Types of Convention in Indirect Speech Acts. In Peter Cole (ed.), Syntax and Semantics: Pragmatics, Vol. 9, 261–280. New York: Academic Press.
Nicolle, Steve & Bill Clark. 1999. Experimental pragmatics and what is said: A response to Gibbs and Moise. Cognition 69(3). 337–354.
Noveck, Ira A. & Dan Sperber. 2007. The why and how of experimental pragmatics: The case of ‘scalar’ inferences. In Noel Burton-Roberts (ed.), Pragmatics, 184–212. Basingstoke: Palgrave Macmillan.
Récanati, François. 1989. The pragmatics of what is said. Mind and Language 4. 295–329.
Saussure, Louis de & Peter Schulz (eds.). 2005. Manipulation and Ideologies in the Twentieth Century. Discourse, Language, Mind. Philadelphia: John Benjamins.
Searle, John R. 1975. Indirect speech acts. In Peter Cole & Jerry L. Morgan (eds.), Syntax and semantics. Vol. 3, 59–82. New York: Academic Press.
Shank, Roger & Robert Abelson. 1977. Scripts, Plans, Goals and Understanding. Hillsdale: Lawrence Erlbaum.
Simon, Herbert A. 1956. Rational choice and the structure of the environment. Psychological Review 63(2). 129–138.
Simon, Herbert A. 1982. Models of Bounded Rationality. Cambridge, MA: MIT Press.
Sperber, Dan. 1994. Understanding verbal understanding. In Jean Khalfa (ed.), What is Intelligence?, 179–198. Cambridge: Cambridge University Press.
Sperber, Dan. 1995. How do we communicate? In John Brockman & Katinka Matson (eds.), How Things Are: A Science Toolkit for the Mind, 191–199. New York: Morrow.
Sperber, Dan, Francesco Cara & Vittorio Girotto. 1995. Relevance theory explains the selection task. Cognition 57. 31–95.
Sperber, D. & Deirdre Wilson. 1986 [1995]. Relevance. Communication and cognition. Oxford: Blackwell.
Sperber, Dan & Deirdre Wilson. 2002. Pragmatics, modularity and mind-reading. Mind and Language 17(3). 23.
Sperber, Dan & Deirdre Wilson. 2005. Pragmatics. In Frank Jackson & Michael Smith (eds.), Oxford Handbook of Contemporary Philosophy, 468–501. Oxford: OUP.
Stainton, Robert. 2004. The Pragmatics of Non-Sentences. In Laurence Horn & George Ward (eds.), The Handbook of Pragmatics, 267–287. Oxford: Blackwell.
Stainton, Robert. 2006. Neither Fragments nor Ellipsis. In Ljiljana Progovac, Kate Paesani, Eugenia Casielles & Ellen Barton (eds.), The Syntax of Nonsententials: Multidisciplinary Perspectives, 93–116. Amsterdam: John Benjamins.
Stanley, Jason. 2000. Context and logical form. Linguistics and Philosophy 23. 391–434.
Stanley, Jason. 2002. Making it articulated. Mind & Language 17. 149–168.
Taillard, Marie-Odile. 2004. Adaptive persuasion. UCL Working Papers in Linguistics 16. 247–263.
Tanaka, Keiko. 1999. Advertising Language. London: Routledge.
Tomasello, Michael. 1999. The Cultural Origins of Human Cognition. Harvard: Harvard University Press.
Tomasello, Michael, Malinda Carpenter, Josep Call, Tanya Behne & Henrike Moll. 2005. Understanding and sharing intentions: The origins of cultural cognition. Behavioral and Brain Sciences 28. 675–735
Van der Henst, Jean-Baptiste & Dan Sperber. 2004. Testing the cognitive and the communicative principles of relevance. In Ira A. Noveck & Dan Sperber (eds.), Experimental Pragmatics, 141–169. London: Palgrave.
Wilson, Deirdre. 1994. Relevance and understanding. In Gillian Brown, Kirsten Malmkjaer, Alastair Pollitt & John Williams (eds.), Language and Understanding, 35–58. Oxford: Oxford University Press.
Wilson, Deirdre. 2000. Metarepresentation in linguistic communication. In Dan Sperber (ed.), Metarepresentations, 411–448. Oxford: Oxford University Press.
Wilson, Deirdre & Dan Sperber. 2004. Relevance theory. In George Ward & Laurence Horn (eds.), Handbook of Pragmatics, 607–632. Oxford: Blackwell.
Yus, Francisco. 2003. Humor and the search for relevance. Journal of Pragmatics 35(9). 1295–1331.