Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

As speech acts in contexts, pragmemes play a major role in the pragmatics of language and they serve to illustrate what Hermann Cappellen and Ernest Lepore call speech act pluralism (Lepore and Cappellen 2005). That is, a speaker may assert many different things by using the very same sentence in different contexts. What is less clear is whether pragmemes play an important role in determining the primary semantic content of sentences.

The fundamental debate is whether or not the literal meaning of type-statements that are devoid of indexical expressions is partly determined by the context of utterance. Those who recognize a minimal semantic content conveyed by sentences not containing indexicals and who claim that their literal meaning is independent of the context of utterance can be described as minimalists. Those who maintain that the literal meaning of these type-statements is most often determined by the context of utterance are contextualists. Whether it is cast as an opposition between formal semanticists, institutionalists or literalists, on the one side, and interpretationists, intentionalists or pragmaticians, on the other side, the debate echoes the one that took place between formal semantics and ordinary language philosophy early in the second half of the 20th century. We could say that the debate has started up once again with renewed vigor, and that it is, in fact, the most important controversy to arise in the past thirty years in the analytic philosophy of language.

There are of course different sorts of contextualist debates: in epistemology, political philosophy, etc. Here I concentrate on the semantic debate. On the contextualist side, we must mention for instance Carston (2002), Levinson (2000), Recanati (2004), Searle (1983), Sperber and Wilson (1995), Stainton (2006) and Travis (2001). In the minimalists’ side, we could name Borg (2004), Lepore and Cappellen (2005), Predelli (2005), Salmon (2005) and Stanley (2005).

What does semantic contextualism consist of? To better understand this philosophical position, we must take note of the following fact. There was a time when it was believed that the meaning of an utterance was essentially a function of the meaning of the uttered sentence. Performing an assertion was nothing but the tokening of a linguistic type. But we now know that the meaning of an utterance almost always depends upon other factors that differ from those determined by the conventional meanings of words, and this is why speech act pluralism must now fully be acknowledged. But the fundamental question that remains is whether conventional meaning offers at least a nucleus onto which the far richer intended meanings of the speaker are grafted, or if pragmatic features intrude in the very determination of what is literally expressed. It is with this last claim that contextualist arguments come into play.

Semantic contextualism must be distinguished from semantic intentionalism. This last doctrine stipulates that the meaning of a sentence is determined by what the speaker has in mind. It is slightly different from the contextualist claim according to which literal meaning of sentences not containing indexicals are often determined by the context of utterance. The intentionalist and contextualist theses are of course compatible in principle, since they may up to a certain point reinforce each other in establishing the existence of a correlation between literal meaning and facts occurring when speakers utter a sentence. But these facts need not be psychological facts and affirming that two views are mutually reinforcing is one thing, and demonstrating that the two are identical is quite another. First, there are variants of semantic intentionalism that reject semantic contextualism. Grice (1975), for example, believes that in a given context of utterance, the conventional meaning of words constitutes the most important part of what is uttered, and he maintains that optional pragmatic factors only rarely disturb the delivered semantic content. Thus, he tends to minimize the importance of pragmatic factors in the determination of intuitive semantic content attached to a given utterance. The disambiguation and saturation of statements together with conversational principles are sufficient to derive, from the semantic potential of sentences, minimal truth-conditions that are intuitively associated by speakers. There can of course also be conversational implicatures, but these are an additional layer on top of minimal content. But he also argues for such a minimalist position while simultaneously adopting a conceptual reductionist approach to the subject of conventions governing the rules that give the invariant meaning of words. So in this sense, the Gricean account is all at once minimalist and intentionalist. A similar line of argument is developed by Schiffer (1972), Bennett (1976) and Lewis (1969). According to this view, the very concept of linguistic meaning is one that should be analysed in terms of intentional meaning. This conceptual reduction of conventional meaning can be described as “conceptual intentionalism” or “analytical intentionalism.” It is a case of conceptual analysis while the contextualist claim comes from the observation of empirical facts concerning meaning. Word meaning is for Griceans reducible to speaker’s meaning, but not in the contextualist sense.

Conversely, it is possible to argue for an account of “meaning as use,” interpreted as suggesting the existence of a close connection between meaning and verbal events, and thus agree with the main tenet of contextualism while rejecting psychologism. Baker and Hacker (1984) offer an excellent illustration of this particular philosophical posture. We must also take note that interpretationism, as defended by Davidson (1984, 1989, 1999), is a form of contextualism, but it entails that the interpretation of the hearer, and not the intention of the speaker, is the relevant factor in the determination of the meaning of most utterances. It is therefore also not to be confused with intentionalism. Of course, as I suggested above, it is possible to argue for a particular intentionalist variant of the contextualist thesis, but there are other instances such as ordinary language philosophy and interpretationism.

So semantic intentionalism must be distinguished from semantic contextualism. That being said, I am mostly concerned in this paper with the intentionalist version of semantic contextualism.

It is also important not to confuse semantic contextualism with the empirical facts upon which this doctrine is based. We could perhaps recognize in many cases the existence of a positive correlation between the intuitive truth-conditions (or assert ability conditions) of utterances and the intended meaning of speakers—without coming to the conclusion that intentional meaning intrudes in the literal meaning of a sentence. The empirical evidence which corroborates the contextualist hypothesis can hardly be denied, whereas the explanation in terms of the role of context in the determination of the literal meaning of the sentence is the more controversial aspect of contextualism. One can grant that the intended meaning of a speaker and/or the normal interpretation of hearers play a major role in determining the meaning of what is asserted. But does it determine what is literally said? As such the existence of a determination of context on the content of an utterance can be accounted for just as an instance of speech act pluralism.

This point illustrates that of the main questions at stake is whether one can accept the distinction between illocutionary acts of assertions and locutionary acts of saying. The empirical existence of a close link between intended meanings and assertions does have a bearing on the literal content of a sentence only if one denies the existence of locutionary acts of saying. For if what is asserted is only something done in the context of saying something, then the fact that my intention determines what I have asserted does not yield the conclusion warranted by the contextualist; that is, it does not disturb what has minimally been said.

An author such as Wittgenstein (1953) recognizes the importance of pragmatic factors in the determination of satisfaction-conditions of assertions, and would thus be inclined to acknowledge the empirical observation that serves as a basis for contextualism. Moreover, he was no doubt eager to show how the conventional meanings of sentences were community relative. But as an institutionalist philosopher, he was defending the autonomy of the institution of language, and he did therefore reject the suggestion that word meaning took place essentially in the intended meaning of a particularized context of utterance. Meaning is not what occurs in an eventful situation. Language is a practice governed by rules, understood as a system of ordinary and ostensive definitions, together with an agreement on the application of words in some paradigmatic instances. Even if the semantic rules do not anticipate all their applications, the decision made in a specific context of utterance concerning what has been said is an institutionalist decision that serves only to refine the rules (Seymour 2005). For Wittgenstein, the context of utterance plays a role in relation to the initial rules of language that is similar to the role played by jurisprudence in relation to a system of laws. Even if the laws do not have a determinate meaning, it does not mean that there are no laws. Also the decisions made by the judges concerning the application of a law are not sui generis and spontaneous, for they are made in accordance with the rules contained in the existing laws. And finally, these decisions can be interpreted as entailing nothing but a refinement of the meaning of the already existing laws. So if context play a role, it is only the institutional context, for the variations of meaning are always to be understood in relation with the institutional decisions that are made on the basis of already existing rules. Wittgenstein is not arguing for a close connection between meaning and events. As a staunch anti-psychological philosopher, Wittgenstein would certainly challenge the intentionalist version of semantic contextualism. I also believe that he would challenge the social version as well, because he rejects the connection that purportedly holds between literal meaning and events. An utterance does not all by itself determine meaning, not even an utterance made using a term or a sentence having conventional meaning. The variations created by the use of an expression in a context of utterance becomes a semantic variations only when the linguistic community, just like a judge, stipulates that this variation is semantical. Wittgenstein would consider semantic contextualists to be largely correct in the observations that they are making, since, very often, in many different contexts of utterance, the intuitive satisfaction-conditions of utterances are considerably different from the semantic content of sentences. But once again, this amounts to speech act pluralism and not to the claim that sentence-meaning is determined by the context of utterance. If ‘context’ becomes relevant for conventional meaning according to the Wittgensteinian philosopher, it is only in the sense that an institutional decision has to be made on how to apply the rules, granted that they do not anticipate all their applications. This institutional decision modifies the meaning of the rules. But it is a variation occurring within the institution of language and at no time utterance-events do intrude in the literal meanings of words and sentences. It is rather the rules and the jurisprudence involved in their application that tells us what is meant.

Is it possible to separate natural language from its social use in context? The institutionalist account already in some sense recognizes the influence of the community in the determination of meaning. Meaning is social and community relative. But to say that meaning is use under this account is to say that the semantic rules are those that a community accepts in a dictionary for a certain time on a given territory, and it is also to say that they are always changing through the “jurisprudence” of institutional decisions made either by the experts, or by a critical mass of people within the community as a whole. Now this account is compatible with a fairly strong distinction between literal meaning and pragmatic meaning. So it has nothing to do with the contextualist claim according to which the background assumptions made by the members of the community influence literal meaning of a sentence in that community.

The contextualist philosopher asks: how are we to characterize the semantic rules associated with sentence type and how could we describe the semantic rules in reference to new situations? The problem is apparently that we come across new situations all the time. This amounts to acknowledge the indeterminacy of the semantic rules attached to sentences. I wholeheartedly agree with the indeterminacy thesis, but my own account is one in which the institutional rules must be accompanied by the jurisprudence of institutional decisions made concerning the application of the rules. This institutionalist model is quite different from the contextualist model in which it is claimed that speakers intentions, occurrences or interpreters determine content. This is why the institutional model is compatible with a strong distinction between semantics and pragmatics. And the institutional model allows us to cope with novelty.

Lepore and Cappelen (2005) are also eager to recognize that a very large quantity of propositions may be asserted with a sentence in the context of particular utterances. This is because that which is asserted depends upon factors that vary greatly and that may differ from the semantically expressed proposition. They accept that speech acts potentially express a very large and indeterminate quantity of propositions with a single sentence. Furthermore, they agree that in many cases we can express a full proposition only if the context of use comes into play. This happens when the sentence used contain expressions belonging to a very basic set of indexical expressions. Their semantic minimalism must be understood as the claims that (1) the meanings of words are in general not determined by the context of utterance, (2) the sentences that are determined by context are grammatically sensitive to it and (3) this sensitivity is explained by the presence of words belonging to a basic set of indexical expressions. (4) In these cases, context is understood in the limited sense implying nothing more than time, place, speaker, and proximal or distal features.

2 Secondary Sentence Meaning

Semantic contextualism stipulates that nothing or almost nothing can be meant with language independently of a conversational practice. The presence of the context of utterance is a necessary condition for sentences to express their full semantic potential. The true meanings are those that are expressed by verbal or mental occurrences. There is a moderate version held by Carston (2002) and Recanati (2004) according to which these contextualist claims apply only to sentences but not to words, since words do in general have conventional meanings apart from the context of utterance. But there are also radical versions such as those of Sperber and Wilson (1995) and Travis (2001) in which the very same claims are said to apply to words as well. Words have, at best, a “semantic potential,” but their full and complete meaning depends upon the context of utterance. Despite the important differences between moderate versions and more radical versions, proponents of these two views agree that pragmemes play a role in determining the primary meanings of sentences. In this paper, I consider only the moderate version and I ignore the issue of pragmatic intrusion into word meaning. I concentrate on sentence meaning only.

So the question we want to ask concerns the argument for the suggestion that the literal meaning of sentences is determined by the intentions of speakers in a context of utterance. Semantic minimalism sees literal sentence meaning and pragmemes as involving two different layers of meaning. Word-types and sentence-types enjoy a certain semantic autonomy relative to their occurrences in particular conversational contexts. According to this picture, a very large class of sentence-types express minimal propositions and do not require a verbal event in order to express minimal content. The only exception to this general rule is provided by sentences containing expressions belonging to the basic set of indexical expressions. According to this account, when the sentence expresses a minimal proposition, pragmemes may serve to determine a secondary sentence meaning, but not a primary one. Pragmemes behave in this case like conversational implicatures, that is, as intended meanings that add up to the minimal sentence meanings. So they do not determine primary meanings. Why do we say that the implicated content is added up to a first layer of content? The reason is that the very same act of saying could be associated with a quite different conversational implicature. To put it differently, conversational implicatures are cancelable.

Capone (2009) has argued that some particularized conversational implicatures were not cancelable, but he reached that conclusion while considering very specific conversational situations. However, if he is right this only means that conversational implicatures cannot be cancelled from a specific conversational context, and it does not imply that they could not be cancelled from a specific act of saying. So for instance, in the context of writing a letter of recommendation for a candidate to become professor in a university department, it is impossible not to infer a particular negative implicature if I merely write that the candidate has a good handwriting. There seems to be no way of suggesting anything else. So in such a case, it seems that the content expressed by the utterance of the sentence is determined by pragmemes. But in the context where the same person would be applying for a job involving essentially writing abilities, the very same act of saying could in that context of utterance become quite positive. What does this show? It shows only that context of utterance plays a role in the determination of utterance meaning. Now this can have a bearing on literal meaning only if we assume at the outset that utterance meaning determines sentence meaning. But this is precisely what is at issue. The argument purporting to show that the context of utterance determines sentence meaning cannot work if it rests on the claim that some conversational implicatures are not cancelable in some contexts of utterance. For then the argument is circular, since it has a bearing on sentence meaning only if we are presupposing the Cvlaim that we are trying to make. The fact that an implicature cannot be cancelled from a particular context of utterance does not imply that it is not cancelable per se. Showing that a particular conversational implicature cannot be cancelled from a context of utterance only shows that different contexts of utterance determine different utterance meanings. It illustrates speech act pluralism and does not necessarily have an impact on literal meaning. A conversational implicature may play a role in the primary sentence meaning only if it remains present in different contexts of utterances. The fact that a particular implicature cannot be cancelled from a particular context of use is compatible with its cancel ability within different contexts of use. Particularized conversational implicatures may be difficult to avoid in a particular context of utterance, but the very same act of saying involved in them could have been made in quite a different particularized context of utterance, and this is all we need to argue that a conversational implicature is cancelable.

3 Are There Propositional Radicals?

Pragmemes very often determine secondary meanings. These secondary meanings are on top of the ones that are expressed by the sentence itself, or by the sentence and the context if the sentence contains indexical expressions. But could they determine primary meanings of sentences that are devoid of indexical expressions? It seems that there are cases that can illustrate this thesis. Some sentence-types may perhaps express only “proposition-radicals,” as suggested by Bach (2005), although they do not contain indexical expressions, on the surface at least. Pragmemes, according to this view, serve to determine the missing ingredients in the primary meaning. However, these cases are not clear counterexamples for minimalism, and not only because of the existence of a constant propositional radical. The reason is that the recourse to contextual features may be prescribed by semantic rules attached to the sentence itself. Minimal content is indeed not clearly threatened if the requirement of context must be imposed by some semantic conditions associated with the sentence type. The requirement of context in determining content is taking place only because semantic rules tell us that we must look at the context of utterance in order to determine content.

This is perhaps the main motivation behind Lepore and Cappellen’s wish to restrict context sensitive expressions to the basic set of indexical expressions (e.g. ‘I’, ‘you’, ‘we’, ‘they’, ‘that’, ‘this’, now, tomorrow, here, there, etc.), for these have an autonomous linguistic meaning. Indeed, the semantic character of an indexical such as ‘I’ is expressed by “the utterer of the this token” or, perhaps more conveniently, “the utterer of the token that is here and now.” This does not force us to draw a contextualist conclusion, since it is an autonomous semantic rule associated with the word-type ‘I’ that prescribes recourse to the context of utterance. In Kaplan’s (1989) sense, the ‘characters’ that are attached to words themselves are not determined by context. As function from context to content, characters are autonomous and are not themselves determined by context. Since the indexical expressions belonging to the basic set do have characters, it seems that by restricting context sensitive sentences to those containing expressions belonging to the basic set, we avoid the pitfalls of contextualism.

The problem, though, is that there are sentences that do not contain expressions belonging to the basic set but that are still incomplete in some sense. A sentence like “Rowan is Mr. Bean” may itself be context sensitive although it does not contain indexical expressions. But this may simply be because of an institutionalized use of proper names according to which they are meant to vary in different contexts. Like indexicals, proper names may have a ‘character’ expressing a function from context to content. This function may often be constant but it can also determine a different content in different contexts. Proper names may partly be assimilated to the set of indexical expressions. A proper name such as ‘Rowan’ is a shorthand for the definite description “The individual named ‘Rowan’,” and the definite article may refer to a unique individual in the context. Similarly for “It is raining.” This sentence contains a covert location variable that can be saturated by means of an overt locative phrase. In all these different cases, sentences would be context sensitive despite the fact that they do not explicitly contain expressions belonging to the basic set, but it is just because they do contain implicitly such expressions.

For more controversial cases like “Mary has had enough” and “John is ready,” Montminy (2006) has argued that the best minimalist move would have been to treat them as saying something like “Mary has had enough of something or other” and “John is ready for something or other.” If that were the correct account, then “A is not ready” would be a shorthand for “A is ready for nothing at all,” which is absurd in most if not all contexts. But the correct minimalist analysis perhaps compels us to say that the sentence is to be completed by a reference to some specific element (an event, action or another kind of specific thing). We must perhaps acknowledge the fact that “A is ready” behaves almost like a proposition-radical in Kent Bach sense. But at the same time, it is available for logical inferences like

  • All those who are ready won’t be surprised

  • A is ready

  • —

  • A won’t be surprised

Does that mean that there are minimal truth conditions for “A is ready”? Or is the sentence a proposition radical? I for one would tend to argue that the sentence implicitly contains an indexical, or a demonstrative. This seems to follow from the fact that it is all at once in need of completion by a reference to something specific, and available for logical inferences. Cases like “A is ready” seem to behave like ‘proposition radicals,’ but proposition radicals are as a matter of fact sentences that implicitly contain a demonstrative expression. This is perhaps a controversial move that Bach would not approve, but it is perhaps one that minimalists must approve. The utterance of “A is ready” with the intention to mean ‘for supper’ is semantically equivalent to an utterance of “A is ready for this”, where ‘this’ is used as a demonstrative to be completed by a demonstration in which the utterer is pointing to a plate full of food served in the evening. Now in “A is ready for this,” we have a semantically constant sentence, for the demonstrative itself expresses a semantical rule that prescribes saturation. It is of course in some sense still incomplete, but just as any old sentence containing context sensitive expressions belonging to the basic set.

If someone asks you out of the blue if you are ready, you will be inclined to answer: “for what?” This shows how incomplete the sentence is. But its incomplete character is explained by the elliptical and therefore implicit presence of a demonstrative expression like ‘this’, so that the sentence should read “John is ready for this” (or “Mary has had enough of this”). Here I follow Capone’s minimalist explanation of the nature of such incomplete propositions (Capone 2008). If they could be interpreted as implicitly containing empty slots that can be interpreted as demonstratives or discourse-deictic anaphoric expressions, sentences expressing incomplete propositions would indeed be harmless for minimalism, for they would be analysed as implicitly involving expressions belonging to the basic set.

Similarly, “John is tall” and “Mary is rich” would also contain implicit semantic empty slots calling for completion by a particular reference class. These sentences should perhaps be analyzed as “John is tall (relatively to this reference class),” and “Mary is rich (relatively to this reference class).” Quantified statements like “All came for breakfast” would be an elliptical form for “All of them came for breakfast” (or “They all came for breakfast”), and would thus also be implicitly containing expressions belonging to the basic set of indexical expressions.

So it may be necessary to enlarge the set of context sensitive sentences beyond those explicitly containing words belonging to the basic set. But it would by no means ruin the main claims made by minimalists, for these sentences could be in need of saturation, not modulation. We may indeed only superficially be compelled to enlarge the basic set of indexical expressions, for the involvement of context is prescribed by semantic features such as the implicit (elliptical) presence of demonstratives. Minimalists need not be claiming that there are minimal sentence meanings associated with each and every indicative sentence of the language not containing indexical expressions. In many cases, they only have to postulate minimal ‘proposition radicals’ understood as expressions associated with the implicit semantic constraints of expressions contained in the basic set. There are empty slots involved in proposition radicals at the surface level, but they would always be counted as elliptical for sentences containing expressions belonging to the basic set of indexical expressions. So in addition to sentences expressing minimal propositions, there may perhaps also be sentences expressing proposition radicals that implicitly contain expressions requiring the presence of context. All of this is perfectly compatible with semantic minimalism.

4 Can Pragmemes Determine Sentence Meanings?

Contextualists argue that an innumerable amount of sentences not containing indexicals are dependent on context. They claim that there is an unlimited dependence of the meaning of different linguistic items on other sentential elements showing that the scope of items needing indexes is much larger than commonly accepted. A classic situation serves to illustrate the point. It concerns the use of color words. I stated at the outset that I would not discuss pragmatic intrusion concerning word meaning, but let us consider for a moment the use of color words. If someone says that red grapefruits are on sale at the department store, it seems that the truth conditions of the sentence cannot be determined without knowledge of speaker’s intentions. Is the speaker referring to fruits that are red on their surface or red inside their surface? The minimalist answers that the truth conditions of the sentence make reference to grapefruits that are red ‘period’, whether on their surface or inside their surface. Of course, the speaker may with her use of the word ‘red’, intend to refer to those grapefruits that are red inside as opposed to those that are yellowish inside and outside. But once again, this is nothing more than a determination of speaker’s meaning on the meaning of her utterance, and not a determination of speaker’s meaning on the literal meaning of the sentence itself. In the context of utterance, the sentence becomes an elliptical way of saying ‘red color inside the surface’ by using the word ‘red’. The additional information is relevant only for determining what the speaker means in the course of saying what she is as a matter of fact saying, not for determining the content of the sentence itself. A similar situation occurs in the case of the sentence “Pierre went to the gym.” For the minimalist philosopher, the minimal truth conditions suppose that Pierre went to the gym “period”, no matter whether he went inside the gym or in the vicinity of the gym. Most of the time, I may intend to refer to the fact that he went inside the gym, but this is what happens in particular utterances. My diagnosis is that contextualists very often confuse what is expressed by the sentence with what is expressed by an utterance of the sentence in the context of an illocutionary act. They take for granted that we must be looking for contextual features in order to determine the literal content of sentences, while context is as a matter of fact relevant only for determining the content of her illocutionary act.

We have seen that pragmemes do not play a role in determining primary sentence meaning. Of course, but this is perfectly compatible with the fact that pragmemes involve conversational implicatures and determine secondary meanings. We have also seen that pragmemes do determine primary meanings for very specific sorts of sentences, as long as these are interpreted as grammatically sensitive to context and as long as they do so because of the explicit or implicit presence of expressions belonging to the basic set. But in these examples, pragmatic features are not optional, since they are semantically called for. So we still do not have a strong case for semantic contextualism.

We must now ask whether pragmemes can sometimes also optionally determine primary meanings for sentences. It would then be a case of modulation and not of saturation. We are wondering whether pragmemes can sometimes involve explicatures, that is, speech acts involving a pragmatic intrusion in the determination of literal content that determine optional primary sentence meanings. Carston (2002) and Recanati (2004) both argue that there are many cases like this. Recanati’s contextualism stipulates the existence of a relation of dependence between literal meaning and pragmemes such that the literal truth conditions would all at once be (1) primary, (2) optional and (3) intentional.

(1) The pragmatic factors concerned are primary in the sense that they play a part in the determination of the literal truth-conditions of what is said with a sentence. They can be distinguished from secondary pragmatic factors that add an additional layer of (pragmatic) meaning to the literal meaning. As primary factors they are not to be confused with irony, metaphor, conversational implicatures and indirect speech acts. All of these presuppose statements that already have a literal meaning. But we are more interested in phenomena that play a part in the specification of literal or primary truth-conditions. (2) The relevant pragmatic factors are also optional in the sense that they do not stem from semantic rules associated with expressions. Indexical and demonstrative expressions, for example, have functioning rules that semantically enforce recourse to the context of utterance. The context allows us to complete the statement, which then expresses full and complete truth-conditions. This would be the phenomenon of saturation that minimalists are in a position to accept. It should not be confused with the phenomenon of modulation that contextualists are postulating, in which facultative pragmatic factors play a useful role for determining the literal truth-conditions of a given statement. In order to show that the primary truth conditions that pragmemes determine are not prescribed by the literal meaning of the sentence, Recanati argues that they are optional, which is another way of saying that they are cancelable. If indeed the intended meanings are optional, then they are not imposed by the very semantics of the sentence and they seem to offer a clear case of intrusion of pragmemes in the very content of the sentence. (3) Finally, the pragmatic factors concerned are also described as intentional in the sense that it does not suffice to refer to a limited notion of context that implies nothing more than time, place, speaker, and proximal or distal features. We must also bring in the intended meanings and the beliefs of the speakers.

The essential idea of Carston and Recanati is that primary, optional and intentional pragmatic factors will at times intrude between the conventional meanings of words and the actual truth conditions of sentences. The literal truth-conditions of many utterances are determined in part by the conventional meaning of constitutive expressions, but also by these sorts of pragmatic factors. According to them, the truth-conditions that stem merely from the conventional meaning of words are not always relevant. The truth-conditions that are meant by the speaker—and which are accessible to hearers—are also at times the relevant determining factor. Recanati maintains that diverse pragmatic phenomena serve to illustrate this point of view. There are cases of enrichment like for example, “he took his key and opened the door,” in which we understand that he opened the door with his key; cases of loosening like for example, “the bank machine swallowed my credit card,” in which we directly grasp what is meant without ever considering real swallowing; and cases of transfer like, for example, “the ham sandwich left without paying,” in which we directly grasp that it is the eater of the sandwich who took off without paying.

Recanati imposes an accessibility condition on literal meaning. It is the meaning that we all directly have access to in any context of utterance. So he also maintains that the intended meaning goes hand in hand with what is recognized by the addressee. In other words, he maintains that uptake must be secured. The intended meaning is also the meaning to which a normal interpreter would have access. He therefore goes on arguing that sometimes, speakers and hearers have direct access to intuitive truth conditions that do not correspond to those that seem to be expressed by the sentence itself. He also claims that speakers and hearers do not even compute the available meaning on the basis of the truth conditions supposedly determined by the sentences themselves. This is why the intended truth conditions are said to be primary. They are not parasitic upon a minimal content that would be asserted by the speaker and grasped by the hearer.

How can we reconstruct the argument for contextualism? Here is a first attempt. I shall formulate it assuming for the sake of argument that the literal meaning of a sentence is to be parsed in terms of truth conditions. It is assumed that truth conditions constitute the most important semantic ingredient of a sentence in the indicative mood. Now very often the truth conditions of utterances require for their specification the presence of pragmemes. There is therefore a close link between the literal meaning of sentences and pragmemes, and this suggests that pragmemes are the primary vehicles of meaning. We should by now know that his version of the argument is of course not valid. The fact that the truth conditions of an utterance requires context does not prove that there is a close link between the literal meaning of the sentence uttered and the context of utterance. It establishes a connection between an intention and the utterance of the sentence and not with the sentence itself. There is however a variant of the same argument that might look more sound. It is this variant that seems to be defended by Recanati. A missing premise in the first version of the argument helps establishing a close connection between the primary meaning of sentences and the utterances of those sentences. The availability principle provides the missing premise in the argument. The availability principle stipulates that the truth conditions of a sentence must be those that are available to both speaker and hearer (Recanati 2004, 20). Now the primary truth conditions that people have access to are very often the ones that are determined by pragmemes. Furthermore, this access is not mediated by the literal meaning of these sentences. That is, speaker and hearer have a direct access to these truth conditions. From these premises, it seems we can draw the appropriate contextualist conclusion. It should be stressed that this view is compatible with the suggestion that there are truth conditions expressed by the sentence itself. It is just that the semantic content expressed by the sentence itself is not relevant in determining the content that is directly available by speaker and hearer. The meaning of the sentence itself is not be directly available to those who are parsing the sentence in that context, and so it is not relevant in these contexts.

It is important to note the analogy between the argument thus reconstructed and the thesis argued for by Dummett (1993) that a theory of meaning must also be a theory of understanding. This claim played a major role in Dummett’s argument that the meaning of a sentence could not simply be correlated with a set of truth conditions. Speakers and hearers very often do not have a direct access to the truth conditions of sentences. So a theory of meaning as truth conditions falls short of determining what is understood by speakers and hearers. If truth conditions fall on the side of ‘denotation,’ there has to be a corresponding side of ‘sense’ that captures what is grasped both by speakers and hearers. Dummett thought that what is grasped is a verification procedure. Recanati’s argument is somewhat similar, for he imposes a normative constraint on what is to count as literally expressed by the sentence. This cannot merely be the proposition expressed by the conventional rules associated with the sentence. Literal meaning has to include what is directly accessible to both speakers and hearers. And then Recanati claims that the intuitive truth conditions that are associated to a sentence often do not coincide with the official ones prescribed by the sentence itself.

There is however an important difference between Dummett’s argument and Recanati’s argument. Dummet’s verification procedures are themselves associated with sentence types and not with sentence tokens. Dummett is not a contextualist philosopher. Recanati’s intuitive truth conditions are accessible by speakers and hearers in particular contexts of utterance. The accessibility constraint applies to what takes place in a conversational context. The question should then be asked: why should we map the intuitive conversational truth conditions onto the sentence itself and draw conclusions concerning its literal meaning?

Dummett’s claim that a theory of meaning must be a theory of understanding enables us to see a little more clearly what is involved in Recanati’s own theory. But it does not do all the work needed for concluding that the intuitive truth conditions are part of the primary meaning of the sentence. In order to see this, it is important to note that the accessibility constraint may in principle be satisfied in a given context by the truth conditions expressed by the sentence itself, even if these are not intuitively those associated by hearer and speaker in another context. This is a consequence that follows from the fact that the pragmatic features are optional. If they are truly optional, then they may fail to occur in a given context. In that context, speaker and hearer may directly have access to the truth conditions of the sentence itself. So if Recanati is truly committed to provide optional pragmatic features, he should consistently treat the literal meaning of the sentence as accessible both to speaker and hearer. And so according to his own account, the literal meaning of the sentence is in these contexts the primary meaning of the sentence. If pragmemes are really optional, and if there can be contexts in which the semantic content of the sentence, that is, the minimal proposition (or its minimal truth conditions), is directly accessible to all the participants in the conversation, the literal content of the sentence is in those contexts part of the normal interpretation of the sentence. So it appears that the minimal content is accessible after all, and thus could be treated as the primary content expressed by the sentence, allowing for the presence of a secondary layer of intended meaning in other contexts. But this is not what Recanati is suggesting. He is rather arguing that the primary meaning of a sentence in a given context is determined by what is accessible in that context. So if in a specific context, the intuitive truth conditions associated by the speaker and hearer are different from the ones expressed by the sentence itself, then in that context, the literal meaning of the sentence expresses a content that is determined by pragmemes. But it now appears that the notion of accessibility is context relative. The fact that the primary truth conditions could coincide in some context with those that are expressed by the sentence could be used to show that these minimal truth conditions are accessible in that other context, but this according to Recanati has no implications for determining what is accessible to both speakers in a different context. The determination of primary truth conditions should always be a function from what is accessible in a context.

Here one senses the danger of a circular argument. For we are invited to accept that primary meaning is determined by what is accessible in a context, in the course of an argument purporting to show that the primary meaning of a sentence is context relative. Why should the accessibility constraint be construed as context relative? Is Recanati not cheating here a little bit? We want to determine whether literal truth conditions are context relative, and we are told that they must be accessible. But if it is implicitly stipulated that accessibility must be context relative, then the accessibility premise in the argument is doing all the work.

I shall soon return to the circular character of the contextualist argument. For the moment, let us simply conclude that there appears at best to be contexts in which content is determined by the minimal proposition expressed by the sentence, and contexts in which the content is determined by pragmemes. But which is the content that is so determined? Is it the content of utterances of the sentence or the content of the sentence itself? If accessibility is context relative, it seems that we can only conclude that it is the content of utterances. So one may wonder whether pragmemes do indeed determine the literal meaning of the sentence. Are they not instead only determining the truth conditions of a sentence-in-a-context-of-utterance, and therefore determining the truth conditions of an utterance of the sentence, and not of the sentence itself? And if so, then where is the intrusion?

5 One Last Refuge

Recanati ran into trouble because he correctly realized that pragmatic features had to be optional. They had to be optional because they should not be prescribed by the grammatical sensitivity of the sentence to context. The problem with that solution, as we have seen, is that it implies that the truth conditions expressed by the sentences themselves will also be accessible in some contexts. It is a problem because these truth conditions look very much like the minimal truth conditions that contextualists are trying to avoid. The only way out is to contextualize the accessibility constraint. In other words, it is assumed that the meaning of a sentence is what is accessible in a certain context. From such a premise, it is then easy to draw the conclusion that meaning is context relative. But apart from the fact that the conclusion of the contextualist argument seems to be contained in such a premise, there is also the problem that if availability is context relative, contextualized intuitive truth conditions will have to be associated with utterances, not with the sentences themselves. We are back to speech act pluralism which is not an issue opposing minimalists and contextualists, and contextualists are still unable to show how to go from there to the conclusion that sentence meaning is context relative.

In order to avoid circularity and an argument that has a bearing only for utterances and no impact on the primary meaning of sentences, it is perhaps important to see whether there are pragmatic features that have an impact on the primary meaning of sentences without being optional and without being prescribed by the semantic constraints of sentences. Are there pragmemes determining primary truth conditions that are not prescribed by semantic features and that are not cancelable? Burton-Roberts (2006) and Capone (2006) argue that there are. For example, “Pierre shrugged and left” means (via explicature) “Pierre shrugged and then left,” and cannot be interpreted otherwise although nothing in the sentence is grammatically sensitive to context. But it is not clear why we could not read the sentence as describing two simultaneous events. And if so, interpreting it as describing a sequence of events can be cancelled (Pierre shrugged and left all at once). Similar remarks apply concerning the example discussed by Carston (2002, 138). The explicature of “Pierre ran to the edge of the cliff and jumped” is something like “Pierre ran to the edge of the cliff and jumped over the edge of the cliff.” For Carston, this explicature can be cancelled by saying that Pierre ran to the edge of the cliff and jumped but stayed on the top of the cliff. It is simply false to suggest that Carston is playing on two different meanings attached to the word ‘jump’: a directional meaning and a transitive meaning. The fact of the matter is that the sentence initially contains an intransitive occurrence of the verb. But when the sentence is uttered, the speaker and hearer most often parse truth conditions that go beyond the literal meaning of the sentence, in accordance with a transitive use of the verb ‘jump’. But since someone may also in other contexts use the same sentence in accordance with the literal intransitive word ‘jump’, the usual explicature can be cancelled. Even if the utterance of the sentence usually make use of a transitive interpretation, the tragic interpretation need not be the good one.

A more interesting case is the following:

  1. 1.

    If the king of France died and France became a republic I would be happy, but if France became a republic and the king of France died, I would be unhappy,

This sentence ‘prima facie’ appears to be contradictory, but the contextualist philosopher argues that it is not really so. Since one cannot imagine someone intentionally accepting a contradiction, the contradictory reading, if Capone is right, is simply not in the cards. There is only one possible interpretation that avoids the contradiction and it is one in which the conjunction is read as involving an ordered sequence of events. We should parse (1) as saying that if the king of France died and then France became a republic I would be happy, but if France became a republic and then the king of France died, I would be unhappy. But consider a context in which the speaker says: although I am a republican, I like the King of France, and it is not good news if I hear that the king has died, no matter how and when he died. But I always prefer to hear the bad news first and then the good news, no matter what is the actual sequence of events. So for instance, if I am told that the king of France died and told immediately after that France became a republic I would be happy, but if I am told that France became a republic and then told that the king of France died, I would be unhappy. This might be the explicature for the initial sentence. So it is not true that the first explicature is not cancelable.

Here is a third possible context of utterance. The speaker has mixed feelings concerning the complex state of affairs involving a revolution and the death of the king. As a matter of fact, she has contradictory emotions of happiness and unhappiness. In her assertion of (1), she is expressing these contradictory feelings by referring twice to the complex state of affairs, and the order in which the facts are mentioned is not relevant. The point here is that when it comes to emotions, contradictions are quite possible. We live our lives full of contradictory emotions and mixed feelings about our personal and social environment.

A fourth possible context of utterance is the situation where the teacher in a classroom intends to show that from a false premise one could infer a true proposition as well as a false proposition. Let us assume that the conjunctive proposition “The king of France died and France became a republic” is false, and that the sentences “I am happy” and “I am unhappy” are respectively true and false. She then proceeds in uttering (1) to illustrate the point.

But no matter what is the specific context of utterance, we are not even condemned to fix the truth conditions of the sentence by appealing to contextual features in order to avoid the disturbing consequence of having to deal with the intentional acceptance of a contradiction. The logical form of the sentence expressed is:

  1. 2.

    (If P & Q then R) AND (if Q & P then −R)

or, in a more condensed form,

  1. 3.

    (If P& Q) then (R & −R)

If we construe the ‘if-then’ conditional form as a material implication, then there is a case where (3) could be true. The only case where (3) is true is the one in which the antecedent is false, since from a false premise, one can infer true and false conclusions.

But (1) raises general questions concerning the possibility that someone would use the sentence to utter what looks like a blatant contradiction. So let us rephrase it to illustrate the difficulty. Let us avoid the formulation of a material conditional and avoid referring to contradictory feelings granted that it is possible to entertain contradictory emotions toward the same state of affairs. We get something like

  1. 4.

    I am informed that the king of France has died and France became a Republic, and I believe that it is a good thing; I am then informed that France is a republic and the king of France has died, and I believe that it is not a good thing.

Here it seems hard to avoid the annoying impression that the speaker is involved in an intentional contradiction, unless we explicate the italicized conjunctives ‘and’ as directly meaning ‘and then’. In other words, we apparently here have no choice but to articulate the italicized conjunction in such a way as to reveal a sequence of events, if we are to avoid an intentional contradiction. We can then reinstate Capone’s interpretation, as in the initial interpretation and this time, we apparently have no way out. We no longer are in a position to read (4) as invoking a preference to hear the bad news first, or as involving contradictory emotions. And we do not have an explicit reference to a conditional sentence. But (4) need not be construed as involving an explicated ‘and then’ construction for the italicized ‘and’, since as the bold occurrence of ‘then’ shows that there are two occasions in which one is being presented with information concerning France and the King. So the speaker may simply be telling the hearer that he changed his mind. So he need not be ordering in a sequence of events the death of the king and the advent of a French republic. She rather means to suggest that the two events were for her initially seen as a good thing, and they no longer are. So once again, Capone’s explicatures are cancelable.

Here’s now the ultimate attempt to impose a contextual reading on the conjunctive ‘and’ and force into it an explicature that apparently cannot be cancelled:

  1. 5.

    I believe that London is pretty and I believe that London is not pretty

Granted that it is impossible to entertain contradictory intentional beliefs, the italicized ‘and’ can only be read as implying something like ‘and then’, suggesting that the speaker as changed his mind. But the very same sentence could be used in the classroom to give an example of the impossibility of contradictory intentional beliefs. In other words, the whole sentence (5) is mentioned and not used by the teacher. Nevertheless, it is a case where it is important not to read the conjunction as involving a reference to a sequence occurring in time, for that would not necessarily be intentionally believing two contradictory propositions. If this is correct, then it follows that the ‘and then’ reading of the conjunction is cancelable.

As I have shown in the case of conversational implicatures, cancel ability is a notion that should be applied with reference to all contexts of utterance and it should not be construed as indexed to a context. That is, if it’s cancelable in one context, then it’s cancelable period. I agree that some utterances performed in certain contexts are to be explained by a specific explicature and that all others should be excluded in that context. So some alternative explicatures do not even arise in these contexts. But explicatures should not be indexed to a context, unless of course we stipulate at the outset that they must be context relative. So we only need to show that an explicature is cancelable in one context in order to show that it’s cancelable at all. Otherwise, we would once again build contextualism in the very notions we are using. The substantial conclusions would already be contained in these fiats.

6 Saying and Asserting

Let me now move to considerations that diagnose the confusions leading one to embrace semantic contextualism. The first problem stems from the usual failure to distinguish between locutionary acts of saying and illocutionary acts of asserting. I believe that Searle was wrong to abandon the distinction between locutionary acts and illocutionary acts. The locutionary act is an act of saying something meaningful while uttering something. The speaker says the content expressed by the sentence uttered. She refers to objects, expresses senses, predicates properties, implies semantic consequences and presupposes semantic presuppositions. This is what is involved in her locutionary act of saying something. In particular, there is no commitment to truth involved in an act of saying (or locutionary act). Whether I assert, promise, order, or declare that p, I am in each case saying that p, so saying is not just another illocutionary act. I can consider whether p, ask whether p is true, I can wonder what would happen if p and I can ascribe to someone the belief that p. In all these cases, I am saying that p without asserting it.

What is the difference between an illocutionary act of assertion and a locutionary act of saying? There are two separate issues involved: the locutionary / illocutionary distinction and the distinction between saying and asserting. In indirect speech or in the utterance of propositional attitude sentences, the subordinate clause is said but not asserted. Consider the sentence:

  1. 6.

    Graham believes that we can save the world

If that sentence is uttered, then the subordinate clause is also uttered. So there is a phonetic act involved. And since the subordinate clause is a well formed grammatical sentence, then the phonetic act involved is also a phatic act. And since the sentence is meaningful, it is also a rhaetic act. So the utterance of the subordinate clause is a locutionary act. But if I utter (6) I perform no assertion of the subordinate clause. Of course, very often, when we perform a locutionary act on a sentence, we do also simultaneously perform an illocutionary act on that sentence. The illocutionary act is performed in the course of performing the locutionary act. So if you utter

  1. 7.

    We can save the world

The act of saying goes along with a commitment to the truth conditions, but it is because in the course of saying it, you performed an assertion. If you had uttered “He believes that we can save the world,” you would also have performed a locutionary act of saying the same minimal proposition, but without commitment to the truth of the proposition expressed, because it is embedded in a larger sentential context. The locutionary act is the act of expressing a proposition and expressing minimal truth conditions, but it does not involve a commitment to the truth of the sentence.

The other issue concerns the distinction between saying and asserting. There may be a use of ‘saying’ that amounts to ‘asserting’. But here I am using ‘saying’ in a somewhat technical sense synonymous with what I take to be the locutionary act. We must coin an expression to refer to what is happening when we are using (as opposed to mentioning) a sentence without committing ourselves to the truth of that sentence. Is there a sense of ‘saying’ in ordinary parlance that captures this technical notion of a locutionary act? The answer is ‘yes’. There are at least two institutionalized uses of the word ‘say’, and I am going to use them in the next sentence. Let me just say-1 that when we do not make an assertion, but we order, promise, declare something, we are always saying-2 something. In the sentence just uttered, ‘say-1’ is the illocutionary use of the ordinary word ‘say’, and ‘say-2’ is the other use, the one that corresponds with the locutionary sense. If I utter “Alice believes that school is out for summer” and “Al believes that school is out for summer,” I say-1 two different things about Alice and Al. That is, I make two different assertions. But in both cases, I say-2 the same thing about them. To make use of Davidson’s vocabulary, there is a same saying relation between the two subordinate clauses.

We can illustrate the use of ‘say-2’ in ordinary parlance with another example. If you utter

  1. 8.

    Are you experienced?

and someone does not hear you clearly, she could ask, “what did you say-2”? If you wrongly stick with a concept of saying understood as always involving an illocutionary act committing one to truth, then you should answer:

  1. 9.

    I was not saying anything. I was just asking a question.

This sounds odd. In my sense of say (say-2), you should simply repeat your question.

Why is it so important to distinguish between locutionary acts of saying and illocutionary acts of assertions? There is a philosophical argument that can be made on the basis of the distinction that has a bearing on the issue of contextualism. Locutionary acts express minimal propositions or minimal truth conditions. Full blown assertions may come equipped with loads of presuppositions and background beliefs, and therefore often determine maximal truth conditions. We may use sentences that express minimal propositions or minimal truth conditions in order to capture, express or describe what someone is saying-2, but we also very often use them while presupposing the complex cognitive architecture of each other’s mental framework. The sentence “Pierre is cutting the grass” expresses a minimal proposition or minimal truth conditions and it is true in the minimal sense if Pierre cuts the grass, whether he is using a lawnmower or a razor blade. But in the thick, robust sense, it may be asserted with certain expectations and background presuppositions. It is in this latter sense that we are entitled to claim that the ‘truth conditions’ have not been satisfied if Pierre only used a razor blade.

Illocutionary acts are done in the course of saying something. What am I doing if I perform an assertion while saying something? In an assertion, I express a belief, I presuppose the existence of a justification for the content of my speech act, and I imply that the content of my speech act is true at least in part because of the presupposed justification. In a promise, I express an intention, I presuppose that you expect me to do the thing specified by the content of my speech act, and I imply that the content of my speech act will be fulfilled at least in part because of your expectation. In an order, I express a desire, I presuppose the existence of an expectation on my part that you do the thing specified in the content of the speech act, and I imply that you should bring about the situation described by the content of my speech act at least in part because of my expectation. In an expressive illocutionary act, I express an emotion, I presuppose the existence of a state of affairs and I imply that the content of my speech act is justified partly because of the presupposed state of affairs. Finally, in a declarative illocutionary act, I express my decision to bring it about that p by saying ‘p’, I presuppose that I have the authority to bring it about that p by saying it, and I imply that p is brought about by my utterance because of my presupposed authority.

Now all of this is done in the course of saying something. So saying p is one thing, and asserting q while we are saying p (whether or not p = q) is another thing. There are alternative ways of construing the taxonomy of illocutionary acts (viz. Searle’s notion of direction of fit), but the important point is that it does not have much to do with locutionary acts of saying things.

7 Intentional and Material Reports

I have distinguished two uses of saying: one that corresponds with asserting, that is, an illocutionary act, and one that corresponds to the locutionary act. In other words, I grant that sometimes we use ‘saying’ as synonymous with asserting, but there is still a distinction to be made because saying is also sometimes used in the sense of a locutionary act. But we should also introduce a distinction between ‘intentional’ reports and ‘material’ reports in order to capture another distinction between two different uses of saying in its locutionary sense (Seymour 1999, 1992). As we shall see, this distinction can also be used to refute contextualist philosophers.

Let me first discuss very briefly the distinction between these two kinds of report. An intentional report describes an intentional state, act or action. In the full blooded sense of intentionality, the state, act or action will have all the usual features of intentionality: directedness (intentional object), intensionality (with an s), reflexivity and first person authority. According to that last feature, if someone is in an intentional state of belief that p, she knows that she believes that p. If someone performs an intentional act of saying, the person knows what she is saying.

A material report describes a state, an act or an action by supposing much less than full blooded intentionality. It assumes the existence of a functional state of the agent. The properties of directedness and perhaps to a certain extent also intensionality are present, but not necessarily reflexivity and first person authority. In the material sense, Fido might be described as believing that there is a cat in tree because he behaves in a way that seems to take for granted the existence of a certain state of affairs. Unconscious beliefs provide another example. Oedipus believed he wanted to marry Jocasta, but he did not realize that Jocasta was his mother. So he did not intentionally believe that he wanted to marry his mother. But at the level of his unconscious states, he might have believed it.

Of course, since intentional states, acts and actions are themselves types of functional phenomena, material reports can also apply to them. If I intentionally believe that p, I also as a matter of fact believe that p. Intentional states are types of functional states. A material report is one that describes a state, an act or an action as functional without assuming full blooded intentionality. But it does not deny the presence of full blooded intentionality in what it is describing. This is why it can also apply to full blooded intentional states, acts or actions. But since it does not assume full blooded intentionality, it can be used to describe functional states, acts or actions that do not exhibit all the properties generally associated with full blooded intentionality.

Another instance of application of material reports would be concerning certain kinds of locutionary acts. When I utter some sentence, there are things that I am saying that I do not necessarily fully comprehend or entertain. I might of course know what I am saying while saying it, but I might also fail to attend or grasp all the elements involved in my act of saying. This might be because of my ignorance (not fully grasping the meaning of a word, for instance) or simply because I did not fully attend to what I was saying.

This can happen when I intend to mean something in the course of saying some other thing. There is something I intend to mean in the course of my act of saying and, precisely for that reason, I am not entirely vigilant about the actual meaning of what I am in fact saying. So when someone utters ‘she took her key and opened the door’ or ‘the cash machine swallowed my credit card’ or ‘the ham sandwich left without paying,’ she does not necessarily realize what she is actually saying. There are true material reports that could describe what she is actually saying even if they would not describe what is taking place in her mind in the course of her actual intentional assertion.

So most cases that seem to serve the cause of contextualist philosophers can perhaps be explained by using the distinction between intentional and material reports and applying it to the locutionary act of saying. Although the speaker intentionally meant that the key was used in opening the door, that the cash machine did less than swallow something and that it is a person that left without paying, she did as a matter of fact say in the material sense what is literally expressed by the sentences she uttered.

Now in their use of language, speakers also defer to others. So if someone attracts her attention to what she actually said in the material sense, she will now entertain or realize what she earlier failed to entertain. Since she defers to others, she will recognize that what she did not consider or apprehend was actually said.

To conclude on this, we must accept the distinction between what the speaker actually says and what the speaker intentionally asserts while saying it, even in contexts in which the speaker does not intentionally entertain the content of what is said. These contexts do not prove that the intended meaning of the speaker intrudes in the literal meaning of the sentence, because the speaker can be described as having said (in the material or functional sense) what is expressed by the sentence. I earlier argued for the existence of minimal content even in the case where both speaker and hearer have access only to the content of the illocutionary act. My point was that the content of the sentence used was also accessible both to the speaker and the hearer. My argument for this was that since the intended meaning was optional, there should be other contexts in which the meaning of the sentence itself was fully accessible to them; and I argued that this was all we needed in order to claim that sentence meaning met the accessibility condition, even in the hard cases discussed by contextualists. Now I claim that there are other reasons for suggesting that what is expressed by the sentence meets the accessibility condition in the three examples mentioned. This condition is met even in the context in which the speaker only has in mind what she means without knowing, or without fully attending to, what she is actually saying. In this kind of situation, even if the speaker may perhaps fail to perform an intentional act of saying, she can be described as actually saying (in the material sense) what is expressed by the sentence she is using. For if someone informs her or attracts her attention to what she is actually saying, she will then intentionally be attending to what was initially expressed by the sentence itself. As a deferring member of a linguistic community, she will herself acknowledge that she was saying what the sentence was expressing. Her disposition to defer to people will induce her to recognize that what she was in fact saying was different from what she intended, but that it was nevertheless what she said. What she said was the minimal content postulated by minimalist philosophers.

8 Circular Contextualist Arguments

I am concerned about the implications of ignoring the locutionary act of saying in the argument that leads to contextualism. If we have methodologically decided to consider only illocutionary acts of assertions, we will then perhaps inevitably be led to think that sentences can only be meaningful in contexts. Cappellen and Lepore admit speech act pluralism, and this relates to what takes place at the illocutionary level. They agree that we can use a sentence in many different ways in order to perform different kinds of illocutionary acts: that is, acts with the same illocutionary force (assertion) but with different contents. We should not be surprised about this since, as we saw, illocutionary acts of assertions come with expressed beliefs, pragmatic presuppositions and pragmatic implications. If we begin by ignoring at the very outset the legitimacy of the distinction between saying and asserting, this can play a major role in the argument for contextualism. I am wondering whether we are not ruling out from the very beginning the alternative conclusion, that is minimalism.

This would be a somewhat circular argument in favor of contextualism. First, we reject the distinction between locutionary acts of saying and illocutionary acts. We then proceed to consider illocutionary acts only. We then note that many different intentional illocutionary acts of assertions can be performed on a single sentence even if it is devoid of indexicals and demonstratives. Primary pragmatic features in the context seem to explain this pluralism. Therefore, literal meaning is strongly dependent on context, and pragmemes become the primary vehicle of meaning. But in this argument, the rejection of the distinction between locutionary and illocutionary acts of assertions in favor of intentional illocutionary acts is an indication of a possible circular argument. If our objects of study are intentional acts of asserting many different things in different contexts, then obviously, meaning is context relative and pragmemes are the primary vehicles of meaning.

Of course, the contextualist philosopher may be willing to accept the distinction between locutionary acts of saying and illocutionary acts of asserting. At first, granting this point might be seen as granting the truth of minimalism, because it seems like she is willing to accept the existence of minimal propositions or minimal truth conditions. But just like the radical contextualist philosopher might be willing to accept that words have a ‘semantic potential’, the moderate contextualist philosopher might be willing to accept that sentences only have a semantic potential, but it is one that is not relevant for determining the intuitive truth conditions that both speaker and hearer associate to the sentence. But I still have worries about the circularity of the argument, even when the distinction between saying and asserting is granted. For in the argument that purport to establish the importance of the intuitive content accessible to both speaker and hearer, the contextualist philosopher is assuming that what is important is located at the level of occurrences, and this also influences the conclusion of the argument which is precisely that meaning is relative to contextual features in the context of utterance. I have expressed this worry in my critical study of François Recanati’s Literal Meaning (Seymour 2006). Recanati replied that he is perfectly willing to admit that, in some sense, we may be saying the same thing when we assert two different things. And he thinks for this reason that he is not vulnerable to a criticism of circularity in his argument. There may be something minimal that we say even when we assert many different things, but that does not prove minimalism. On the contrary, Recanati insists that the content of what is said, even if it exists, plays no important role in the process of a normal interpretation, which is after all the only game in town. But it is here that I locate the most important danger of a circular argument. The real problem concerns the emphasis on interpretation or actual processing. If we are all interpreters struggling to decipher particular inscriptions, and if what is crucial concerning primary meaning is to capture what is going on in utterance events, well then of course, we must concede immediately victory to the contextualist philosopher. Linguistic inscriptions are events, or tokens, and it has already been granted that we could make very different illocutionary acts of assertions with the same sentence, even if it is devoid of indexicals and demonstratives. Now if doing this is the only game in town, then of course, pragmemes are at center stage and they have won the day.

There may of course be many different contexts of utterance in which the literal, minimal, primary content does not appear to coincide with the content of the interpretation. But this has a bearing on literal meaning only if, from the start, we assume that literal meaning is what is taking place at the illocutionary level.

There is an important discrepancy between what is actually said and what is interpreted in cases of enrichment (she took her key and then unlocked the door), loosening (the cash machine swallowed my credit card) and transfer (the ham sandwich has left without paying). What is ‘literally’ said may be interesting, but according to the contextualist, it is not relevant to what is actually asserted and interpreted. But the minimalist philosopher is eager to reply: so what? What is the bearing of these observations on the issue of literal meaning? It is here that the contextualist philosopher is forced right from the start to answer that what is important in meaning is what goes on at the level of speech acts. But wasn’t it precisely the conclusion the contextualist was looking for?

So I am afraid that contextualist conclusions are very often implicitly contained either in the accessibility constraint, if it is interpreted as accessibility-in-a-certain-context, in the rejection of the locutionary/illocutionary distinction, or as suggested above, in the methodological principle that interpretation or information processing is the only game in town. We saw also that another instance of circularity is bound to occur in arguments that assume that one could diagnose pragmatic intrusion if some pragmatic features cannot be canceled-in-the-context-of-utterance. If what is important occurs at the level of assertions, interpretations and accessible truth conditions in a context, well then of course, it is hard to resist the contextualist conclusion. But these methodological assumptions make almost all the work in the argument, and they already presuppose the truth of contextualism. If we insist that what is important concerning meaning is what takes place at the level of intended meaning, interpretation or illocutionary acts, then what is important is automatically related to occurrences, tokens and events along with their contextual features. We should not then be surprised to be in a position to conclude, along with Recanati, that meaning is relative to illocutionary acts and that illocutionary acts are the primary vehicles of meaning. Recanati’s argument is that even if there is a minimal proposition expressed by a locutionary act, it doesn’t necessarily play a part in occurrent on-line processing, but he is assuming from the start that what is important to meaning theory is “occurrent on-line processing,” and this is why he is in a position to conclude that meaning is context relative.

A similar debate has been raised with Rob Stainton in a private exchange. He asks: “Are we trying to model natural languages, understood as systems of expressions, or are we modeling human psychological processing of language in context?” If we are assuming from the start that meaning theory is “modeling human psychological processing of language in context,” well who will be surprised about the conclusion that meaning is relative to context?

9 A Coherent Theory?

One could also question the very coherence of the contextualist argument if the moderate contextualist philosopher is willing to grant the locutionary/illocutionary distinction. Can contextualism be coherently defended with the locutionary/illocutionary distinction if primary pragmatic features are optional? For if they are truly optional, then there may be contexts of utterance in which they are not taking place. That is, the options are cancelable. In these other cases, the speaker does not perform the same pragmemes and her intended meaning might in these contexts coincide with the content of what is said. In accordance with the accessibility principle, this minimal content should therefore also be accessible to the hearer. The conclusion would then be that the truth conditions that follow from the sentence itself are accessible and should be primary truth conditions. Now since for any given meaningful sentence of the language, there are minimal truth conditions or minimal propositions expressed by the sentence itself and accessible to the speaker and hearer. But this is exactly what the contextualist philosopher is supposedly denying.

According to Recanati’s theory itself, optionality entails that for each sentence there must be at least one use made in accordance with its locutionary meaning, and so the theory itself is committed to acknowledging the existence of two relevant levels of meaning: a sentence (or literal or locutionary) meaning (that is, minimal propositions or minimal truth conditions) and a pragmatic meaning. Both are accessible if primary pragmatic features are truly optional. So aren’t we stuck with two levels of meaning, as the syncretic view suggests? But the syncretic view is a particular version of minimalism, not of contextualism. The cancel ability or optional character of the pragmatic features signals the presence of a minimal content that could have been expressed without these additional features. This minimal content must therefore also be somehow always present, even in the context of utterance in which pragmemes are said to intrude in the determination of primary truth conditions. Now if minimal proposition are acknowledged, they must determine minimal truth conditions. And so the intuitive truth conditions associated by speaker and hearer in a context become secondary. So it seems that Carston and Recanati cannot have it both ways. Pragmemes cannot all at once determine primary truth conditions and be cancelable. The only way out of this dilemma is to construe cancel ability and accessibility as context relative notions.

The crucial issue is that of accessibility. For if the potential truth conditions are accessible in certain contexts, then they are accessible period, unless of course, we arbitrarily decide to constrain the accessibility principle to the context of utterance. This, I believe, is what is taking place in most if not all contextualist arguments. Since optionality implies that the potential minimal content is accessible in certain contexts, accessibility has to be construed as a contextual feature in order to avoid the conclusion that there are minimal literal contents in each context of utterance. Minimal content will be accessible in certain contexts but it won’t be accessible in other contexts. So it appears that the only way out of incoherence for the contextualist is to say that there is no such thing as accessibility divorced from context. Literal meaning must intimately be related to what is accessible in a context. The contextualist philosopher must assume from the start that the medium in which literal meaning is determined is the context of utterance. So a circular argument seems to be the only way out not only for proving contextualism but also for avoiding an internal incoherence in the theory.

10 The Final Blow

For the sake of argument, let us avoid the problems of circularity and internal coherence of the theory. Let us suppose that the contextualist philosopher can make the controversial and circular methodological claims according to which the most important level of meaning is language processing. Let us ignore the fact that this claim contains in a way the essential ingredients involved in the conclusion of the argument. Let us instead consider the debate concerning the claim that what is said, the minimal proposition (or minimal truth conditions), is not relevant in this process.

The minimalist philosopher could still argue against contextualism that even if language processing were crucial, minimal propositions or minimal truth conditions would still be essential to the language processing of sentences involved in cases of enrichment, loosening and transfer. Without them, interpretation would never get off the ground. This is especially so in the case of Ms Malaprop. When Ms Malaprop is interpreted as asserting that this is a nice arrangement of epithets, her actual act of saying that this is a nice derangement of epitaphs does not figure in the net result of the process of interpretation. But the interpreter must first consider the literal meaning of the sentence uttered, that is (“this is a nice derangement of epitaphs”), and find it odd in the context of a poetry class, for instance. But then the interpreter notices that another sentence, phonetically similar to the one uttered, has a literal meaning that is relevant in the context (“this is a nice arrangement of epithets”). So the interpreter is led to conclude that Ms Malaprop asserted the latter. Now even if the net result of the interpretation does not include what was said, it is hard to claim that what was said did not play an important role in the process of interpretation. As Wittgenstein would put it, capturing what Ms Malaprop wanted to say is moving from a form of expression to another form of expression (Wittgenstein 1953 § 334).

I acknowledge that Recanati’s arguments are compatible with admitting that there are locutionary acts. He simply argues that, very often, speaker and interpreter do not process the content expressed by the sentence, and that it is useless precisely for that reason. But there are problems with his examples, similar to the problems I just raised concerning Ms Malaprop. To give an obvious illustration, if someone utters “she took her key and unlocked the door,” who will deny that the minimal proposition expressed by the sentence was of some use? The information that she took her key and unlocked the door is surely contained in the information that she took her key and unlocked the door with her key! I would argue the same thing concerning many (all?) cases of loosening (“the cash machine swallowed my card”). The first time that we hear such a sentence we imagine a strange deglutition process. We then very rapidly get used to the secondary meaning, but the fact that we do get used to it does not turn it into a primary meaning. I would argue in a similar vein against Recanati concerning transfer, for when we say that the ham sandwich left without paying, we are first processing the proposition literally expressed by the sentence, in order to arrive at the conclusion that it is the client that ate the ham sandwich that left without paying. You first hear what has been asserted as very strange indeed, and then you get the point. According to Recanati, we never reach the stage where the minimal proposition is processed. We merely feel the incompatibility between the subject and predicate, and then shift the process of interpretation by taking into consideration primary pragmatic features. The first proposition that we grasp is the one determined by these pragmatic features, and not the one literally expressed by the sentence. For the minimalist, on the contrary, we do reach the stage of processing the minimal proposition, for feeling the incompatibility between the subject and the predicate amounts to an attitude toward the sentence containing these subject and predicate. In order to arrive at a final interpretation, we must first consider the strange character of what is being said, and then consider “secondary pragmatic features.” The minimal propositions or truth conditions may not end up in the final interpretation, but they are occurring at one point in the process of interpretation.

One may think that the debate might be settled by considering “reaction experiments.” Perhaps these can show that the speaker who hears about the ham sandwich leaving the restaurant automatically computes that it is a person that left. Metaphors and idioms investigated in those reaction experiments may perhaps show that interpretations are made without considering literal meaning first. But if the experiment has been performed on someone who is accustomed to a secondary meaning, the fact that she no longer computes the primary meaning does not turn secondary meaning into primary meaning. Of course, one could reply that reaction experiment in a particularized situation allow us to draw conclusions concerning literal meaning, but this is problematic for reasons that were already stated. We should not draw any conclusion from such a result unless of course we stipulate at the outset that literal meaning is determined by pragmemes in a context, for this would be precisely asserting the conclusion of the argument, and not providing an argument.

11 Conclusion

In this paper, I have been looking for an argument in favor of semantic contextualism that would show how sentence meaning is determined by the intended meaning of speakers in contexts. As we now know, thanks to the contributions of those working in the fields of pragmatics, an intentional act of assertion may give rise to all sorts of things: pragmatic presuppositions, conversational implicatures, metaphors, irony, indirect speech acts and so on and so forth. There may be some minimalists like Stanley and Szabo (2000) who would deny that a wide variety of speech acts can be performed in intentional acts of assertion. But I tend to favor an enlightened version of minimalism that acknowledges speech act pluralism.

Nevertheless, while we may be asserting all sorts of things in the context of a particular utterance, we may at the same time be saying the very same thing if we use the same sentence and if the sentence contains no indexicals and is not ambiguous. Our locutionary act remains the same unless it contains indexicals and it is ambiguous. What we are saying is the minimal proposition (or minimal truth conditions). I have argued that conversational implicatures were a secondary level of (pragmatic) meaning and that they were cancelable. I have also suggested that we could harmlessly increase the number of context sensitive sentences beyond those that explicitly contain expressions belonging to the basic set, as long as we construe them as implicitly containing these indexicals. I then investigated the possibility of primary, optional and intentional pragmatic features determining literal truth conditions. It was shown that if one accepts the distinction between locutionary acts and illocutionary acts, then the notion of accessibility appealed to by Recanati could be used against him. For if pragmatic features were really meant to be optional, then there is at least one context of utterance in which what is said is accessible and this is all that we need in order to defend the syncretic view.

Instead of referring to what Recanati calls the ‘syncretic’ view, I would use ‘bifurcationism’, because there are two levels of meaning: the minimal proposition expressed by the sentence (what is said by the speaker) and the additional intentional pragmatic meaning conveyed by the full illocutionary act (what is intentionally asserted by the speaker). Recanati does not deny in principle the existence of such a double level of meaning, but he argues that in the case of enrichment, loosening and transfer, the so called literal meaning of the sentence is not cognitively relevant for the speaker and normal interpreter. So the literal truth conditions are in these cases determined by speakers’ intentions and normal interpreters in the context of an intentional act of assertion. Recanati rejects the syncretic view (or bifurcationism) not because it is not possible to distinguish in principle different levels of meaning, but because there are apparently cases where no one computes (parses, cognitively entertains) the minimal proposition or the minimal truth conditions. So even if we admit the existence of minimal propositions or minimal truth conditions, they should according to Recanati be ignored in many cases and replaced by propositions or truth conditions that are determined by what the speaker meant in the context, which also happens to coincide with what the normal interpreter understands. Therefore, since pragmatic features like enrichment, loosening and transfer determine primary propositions or truth conditions, they are primary pragmatic features, and this refutes bifurcationism.

It is possible for Recanati to accept the distinction between saying and asserting and still be arguing for his idea that the so called minimal proposition or minimal truth condition is irrelevant. Recanati could perhaps claim that in certain contexts of utterances, the speaker is not intentionally saying the thing that is literally expressed by the sentence. Simultaneously, the normal interpreter is not processing the minimal content of the sentence in order to achieve his interpretation. So this is why the literal, minimal proposition (or truth conditions) is irrelevant. And this is why Recanati believes that the syncretic view is false.

I have argued that this argument presupposed methodological assumptions that turn contextualism into a circular argument. I also suggested that this circularity was needed in order to avoid incoherence. I then challenged Recanati on each of the examples he has provided in his attempt to refute bifurcationism. I would want to say that in all the examples discussed, the minimal proposition is in different ways relevant for the normal interpreter and even for the speaker. It seems that I am cognitively entertaining the minimal proposition when I am intentionally saying a sentence that expresses such a minimal proposition. But even if there were good examples where it appears that neither the speaker nor the hearer really considers the minimal proposition in their language processing, it would not affect the general criticism of circularity.

I suggest instead a general twofold approach in which bifurcationism and speech act pluralism form a sophisticated and enlightened version of semantic minimalism.