Can we say what we mean?: Expressibility and background

! 1! Can We Say What We Mean? Expressibility and Background Jesús Navarro-Reyes University of Seville Abstract: The aim of this paper is to discuss a basic assumption tacitly shared by many philosophers of mind and language: that whatever can be meant, can be said. It specifically targets John Searle’s account of this idea, focussing on his Principle of Expressibility (PE henceforth). In the first part of the paper, PE is exposed underlining its analyticity (1) and its relevance for the philosophy of language (2), mind (3), society and action (4). In the critical part, the notion of Background is taken into account in order to re-evaluate two basic distinctions: the one between sentence and utterance meanings (5), and the one between native and type speakers (6). PE is reconsidered in the light of the previous arguments as a methodological strategy that does not prevent uses of language from eventual semantic excesses and deficits (7), and a complementary Principle of Expression Fallibility is finally proposed (8). Keywords: Expressibility, background, utterance/sentence meaning, individual/collective intentionality, native speaker, John R. Searle. 1. The Principle of Expressibility as an analytic judgment. Back in his classic essay on Speech Acts, John Searle established what he called the Principle of Expressibility (PE), according to which whatever can be meant, can be said (Searle 1969: 19ff). Despite its explicit formulation, PE has received an astonishing little amount of criticism, probably due to the fact that it is considered unproblematic not only by Searle’s followers, but also by many of his critics. In fact, the idea is generally present wherever propositional attitudes are considered to be descriptions of real mental events, and not merely unessential strategies for behavioural prediction. PE is a key concept in Searle’s account of speech acts, because it allows him to analyse language from what he considers to be the simplest cases of meaning: “those in which the speaker utters a sentence and means exactly and literally what he says” (Searle 1975: 30). In his opinion, any other case of language use can be reduced to those, since This is the FINAL DRAFT. Please only cite the published version in Pragmatics & Cognition, 17:2 (2009), 283-308. DOI: 10.1075/p&c.17.2.04nav ! 2! for any meaning X and any speaker S whenever S means (intends to convey, wishes to communicate in an utterance, etc.) X then it is possible that there is some expression E such that E is an exact expression of or formulation of X (Searle 1969: 20). It is important to emphasise that PE is not necessarily true de facto, but only de iure—a point that makes it less interesting in practice, but also less vulnerable to conceptual criticism. Searle does not claim that, whatever the speaker intends to say, there actually is an expression in his own language that she could use to say it. There might be no expression available indeed, due to speaker’s ignorance or to lexical limitations, but Searle’s point is that those boundaries can always be overstepped: the speaker could acquire new idiomatic abilities, or language could always be extended, in order to satisfy the expression necessities of its speakers. If there are limits for expressibility, that is “a contingent fact and not a necessary truth” (Searle 1969: 20). There is no point in considering specific cases in order to prove or refute PE, since it is formulated as an analytic judgment, and analytic judgments are supposed to be immune to particular experiences.1 E.g., it would be futile to look for a married bachelor in the Civil Registry in order to falsify the analytic judgment “All bachelors are unmarried”, for all we could finally prove is that the Civil Registry was wrong. By the same token, no particular experiment could ever falsify PE, namely because such an experiment would be hardly conceivable: if I claimed that there is a particular content that could neither be expressed in fact by a particular speaker, nor in principle by anyone else, it would be a precondition for my experiment to be accepted as such, to indicate what is that content, the speaker is supposed not to be able to say. Would I not have to express it, in order to be my hypothesis considered by others? In that case, if that particular content is expressible de facto, at least for me, why shouldn’t it also be expressible in principle for that speaker? PE seems to be proved by reductio ad absurdum, since there is apparently no way to formulate the alternative possibility.2 However, if PE were analytic—had we got to believe Searle—that would only mean that it is true by definition, but would not say much about the correctness or incorrectness of our definitions. On that point, we can only make a judgment considering the fertility of our theoretical system as a whole (Quine 1950: 1-8). Being analytic does not preserve a thesis from being reconsidered, questioned, or complemented, as I will try to do in the second part of this paper. Nonetheless, we should be very conscious of the effects that a reconsideration of PE would eventually ! 3! have because, as I will show in the first part of this paper, PE plays a much more crucial role in Searle’s philosophy in general—and not only his speech act theory—than could be expected at first sight. 2. Expressibility and language: utterance and sentence meanings. Searle’s conception of meaning is constructed on Grice’s 1957 proposal: To say that a speaker S meant something by X is to say that S intended the utterance of X to produce some effect in a hearer H by means of the recognition of this intention. (Searle 1969: 43). According to Searle, although Grice’s proposal is valuable, it neglects the crucial fact that linguistic expressions, as social institutions, have literal meanings, which are independent from the communicational purposes of the particular speaker. This is a conventional feature of the speech act that the speaker must assume when she performs it: it is certainly up to her to use X in order to say something different from what X means—hoping that the hearer will guess by the context that she is not speaking literally—but, if such an indirect act is possible, it is because both speaker and hearer do previously share the knowledge of X’s meaning. Namely, they share a conventional language. Searle’s analyse is in a sense the reverse of Grice’s one. The latter tries to move from the particular meaning of an utterance—not yet conventionalised—to what he calls timeless meaning of the sentence uttered; the former, on the contrary, begins his account by conventionalised meaning and, from that starting point, tries to understand the working of non conventionalised expressions. Searle’s point is that, without a previous knowledge of linguistic conventions, no pragmatic deduction of non-literal meanings would be possible. Metaphors, ironies and indirect speech acts are in general, in this sense, parasitic on literal, direct speech acts, as they are logically dependent on them: “In each of these cases, what the speaker means is not identical with what the sentence means, and yet what he means is in various ways dependent on what the sentence means” (Searle 1975: 77).3 ! 4! We should consequently distinguish, each time the speaker made an utterance, what she meant to say from what she actually said; i.e., the meaning of her utterance from the meaning of the sentence she uttered. Any non literal feature will correspond to utterance meaning—for only utterances can be metaphoric or ironic, not sentences themselves. In fact, according to Searle (1979a: 118), “literal sentence” is a pleonasm, since sentences as such can only have literal meanings. What PE would allow us to say is that, whenever the speaker utters a sentence intending to communicate something different from what the very sentence means, there does exist (or can exist) another sentence which means—of course, literally— what the speaker is intending to say. In fact, what makes the hearer’s contextual, pragmatic deduction possible—i.e., her deduction from utterance to sentence meaning—is that there is (or could be) another sentence whose meaning would literally express the one of the utterance. What else could be the aim of her inquiry? How else could she guess what the speaker means, behind what she actually says, if it would not be eventually possible to say what she means? We cannot even speak about utterance meanings unless we identify them, in an implicit or explicit way, with particular sentence meanings, since otherwise they would not only be inexpressible, but probably unthinkable for us (I will return to this idea in the next section).4 By the title of his work, one could be led to mistake Searle’s Speech Acts for a study of what Ferdinand de Saussure called parole: the individual act, by which a speaker uses a language, in contrast to the langue, as the abstract set of signs available to the community. It would not be an abstract theory about semantics or syntax, but a study of the use of language in particular situations, considered in their varying pragmatic context. Nevertheless, Searle offers his own theory as a study of the langue, not of the parole: his object is the socially shared aspect of language that the individual cannot change, but passively accept as a pre-condition for the speech acts that she intends to perform. Searle is not interested in the particular application of the rule, but in the rule itself, which, as such, does not belong to the specific situation of a particular speech act. It is PE what allows him to take that step, in that it assures the possibility of establishing a link between utterance and sentence meanings: for each parole act we can perform, there is (or can be) a langue structure whose rules we can describe. This is what has been called the semantization of Searle’s speech acts theory (Dascal 1983, 31), ! 5! namely the idea that we can identify the rules of speech acts with the meanings of the sentences used in those acts: just as it is part of our notion of the meaning of a sentence that a literal utterance of that sentence with that meaning in a certain context would be the performance of a particular speech act, so it is part of our notion of a speech act that there is a possible sentence (or sentences) the utterance of which in a certain context would in virtue of its (or their) meaning constitute a performance of that speech act (Searle 1969: 18). The speaker may accomplish her speech act by uttering words whose literal meaning are not identical—or are even opposed—to the very act she’s doing (e.g., she could be ironical); but in that case, if PE is sound, it would always be possible to translate that particular utterance meaning into the sentence meaning of a different expression. Utterance meaning, if it does not coincide with the meaning of the sentence that is being uttered, is nothing but the meaning of a different sentence. Cases where the speaker is not speaking in a direct, literal way can be left aside for the sake of theory, since it is always possible to translate what the speaker is intending to say into words that actually say it. Perlocutive or rhetorical effects would be different (see Searle 1979a: chapter IV), what Frege disdainfully called the coloration would be lost, as well as ambiguity and vagueness (which might be part of what the speaker intended to communicate). However, thanks to PE, if the speaker was not intending to say q when she uttered q, but something different like p, then she could have just said p. Then, from a theoretical point of view, we can limit our scope to the general rules of use of expression p—langue rules that do not only govern that particular situation, but any situation of that type, in which p is literally uttered. There can be no science of particulars, magister dixit: Searle does not want to construct his solid theory on the unstable, changeable and unpredictable particularity of specific use, but on the general and socially shared structure, which supports linguistic practices. From this point of view, performance of speech acts can be analysed as the act of uttering sentences literally, in the correct situation, following socially established rules. The particularity of abnormal uses where these rules are left aside will be explained in form of corollaries (Searle 1979a), but should not be mistaken for a constitutive aspect of the theory itself. ! 6! 3. Expressibility and the mind: urbanising the cognitive space. Previous section was focused on meaning and language and, under that light, sentence meaning had priority over utterance meaning. However, this section will focus on philosophy of mind and, from this perspective, it is the utterance or, more specifically, the speaker, who takes priority, as that is the only place where we can find what Searle calls intrinsic intentionality. Intentionality is primarily a feature of some mental states such as perceptions, beliefs, or desires, which are addressed to something else. Although language is also intentional, since words generally refer to something, they do so in quite a different way. If we consider language itself, we will only find ink on paper, or sounds in the air: just ordinary physical phenomena that are not intentional on their own. They are simply there, like sunsets or volcanic eruptions, not referring to anything else. According to Searle’s view, intentionality can only be attributed to linguistic events if they are considered as the effect of a mind—the one of a speaker or a writer. Linguistic intentionality is derived from mental intentionality, but the latter does not derive from anything else: it is just the effect of the brain, which acquired through our evolutionary history the ability to be in intrinsically intentional states (Searle 1992: 78-82). Unlike Putnam (1981) or Dennett (1987), this is considered by Searle as a brute fact that would require no further philosophical reduction. Besides the different character of mental and linguistic kinds of intentionality, Searle considers them to be articulated by one same logic, which he tries to make explicit (Searle 1983). Understanding the mind of a person is attributing intentional states to her, in the form of propositional attitudes such as believe that…, desire that…, or having the intention to…, all of them attitudes towards contents, which can be expressed in propositions. The intentionality of the mind is accessible to philosophical study given that it is essentially expressible in linguistic propositional conventions. This common articulation is indispensable if we want to understand the phenomenon in question: if the intentionality of the mind were not translatable into the intentionality of language, hardly could we explain how is it possible to say in words what we are thinking, or how can we understand what the other person is thinking when we hear what she is saying. Just as we previously established that any utterance meaning could be expressed in a sentence meaning, we will have to accept now that any mental ! 7! intentional state can be expressed linguistically. In this way, Searle seems to presuppose what Frank (1984: 399), following Derrida, called a “pre-established harmony” between intentions and expressions, which guarantees the possibility of their eventual identification. In general, a speech act with a propositional content is an expression of the corresponding intentional state, and the propositional content of the speech act is identical with the representative content of the intentional state (Searle 1979b: 192). Even if Searle does not explicitly appeal to PE in this particular point, it seems to be an unavoidable condition of possibility of the kind of philosophy of mind that he considers significant. The a priori establishment of this access to the mind is a must for Searle, whose position is epistemically controversial in a hostile environment, dominated by functionalism and its derivatives. By its side, functionalism attempts to solve the puzzle of mental inexpressibility before it arises: if language is able to express what happens in the external physical world, i.e. behavioural patterns, then it is able to express what happens in the inner mental world, since inner mental assessments are, according to functionalism, essentially linked to behaviour (functional computational states, even if unobservable in themselves, are inferred as casual explanations of observable behaviour). But the distinctiveness of Searle’s position—and the key of the appeal that it holds—is that he tries to grasp the very essence of mental states, considering their first person perspective as an essential feature, and not linking them by definition to any kind of external behaviour whatsoever (Searle 1992: 77-94). The point seems to be reasonable: we are perfectly able to entertain particular beliefs, desires or intentions, but not show any behavioural effect at all, even not in principle. Mental states are not essentially linked to behaviour: if functionalists do assume this link as a necessary condition, it is not because of the very nature of the phenomenon they are studying, but because of some unjustified empiricist prejudices they have inherited from behaviourism. Nevertheless, Searle’s view leads to a problem that does not even arise in the functionalist approach: the difficulty to articulate mental and behavioural statements, i.e., descriptions of subjective and objective events. The claim that conscious mental states are not essentially linked to behaviour is shared by Searle with authors such as Galen Strawson or Colin McGinn, but he does not want to be commited, as they are, to strong scepticism or ‘mysterianism’ about ! 8! subjectivity. Therefore, he needs a guarantee that our common language is able to express what we inwardly intend to communicate, once it has been acknowledged that what we believe or desire in the inside is not tied anymore by definition to what we do on the outside. And, of course, that guarantee is PE, which states that, even if descriptions of mental states do not derive from behavioural statements, they are essentially expressible by linguistic conventions. By establishing PE, Searle seems to be urbanising the cognitive space, allowing sentence meanings to go through it. In the wild land of thought, PE assures the possibility of building roads and streets that will eventually allow the transit of social meanings; meanings that do not only belong to the individual, but to the inter-subjective community of speakers of a common language. If the mind is accessible for linguistic expression, it is not thanks to the possibility of a behavioural translation, but to the ability of the thinker, as a native speaker, to recognise her own mental states and verbalise them, i.e., to say what she believes, desires or intends, and to mean what she says. PE is the guarantee (at least a de iure one) that the mind’s state can be expressed through linguistic manifestations; namely, that the intrinsic intentionality of the mind can assume the shape of the derived intentionality of language.5 In fact, PE helps to solve a double problem in philosophy of mind: in the first place, thoughts probably do not have to be linguistic in order to be intentional, and there must be a way to express the pre-linguistic intentional states in linguistic propositional utterances.6 And in the second place, even when we do use language to entertain certain thoughts, the identity of thinker’s and speaker’s meaning should not be taken for granted. As Dascal has shown, uses of language by a solitary thinker “are not even ‘potentially’ communicative, for, ex hypothesi, there is no communicative intention whatsoever in them.” (Dascal 1983, 45). Both problems would be hard to solve without PE, since sentence meanings would not be able to express those pre-linguistic or privately linguistic kind of thoughts. Access to what is intrinsically mental would be lost, and the conceptual problem of other minds would be intractable. Philosophy of mind could not even begin to explain what Searle considers to be the very nature of its object. According to Searle, the next step in philosophy of mind will be the search for neurological correlates for mental states, for we will not find an explanation of the mind until we do not fully understand how those states are produced in the brain. On the one hand, neural states are increasingly accessible through new techniques as Magnetic ! 9! Resonance Imaging (MRI), which offer inter-subjective, observable data about them; and, on the other hand, mental events are accessible from the first person perspective, and translatable into inter-subjective language, thanks to the abilities of native speakers to perform speech acts. The articulation between both spheres, even if it is still quite far from reachable, is at least conceivable, given that PE assures a shared access to the content of the mind. Without that principle, our search for the neural correlate would be misguided, for we cannot make a rope taut just by holding one of its sides. 4. Social philosophy and theory of action. Besides the distinction between intrinsic and derived intentionality, the one between individual and collective intentionality is also very relevant here. It is introduced by Searle (1995) in his search for a philosophical foundation for social institutions such as money, marriage, civil laws, honours or political offices, which do only exist as far as we consider them to be real, and act accordingly to that consideration. What makes them effective is the intentionality of our minds and, as socially shared realities, they stem from a kind of intentionality, which is not just individually hold by each one of us, but socially shared in the form of collective intentionality. This aspect of his theory is particularly controversial since, according to Searle, collective intentionality is not reducible to any compound form of individual intentionality. In his opinion, individualistic accounts of social reality are ill-conceived, because it is not enough to describe our common beliefs as a result of putting together several individual intentional states (I believe p, you believe p, I believe you believe p, you believe I believe p, and so on); on the contrary, we need to introduce into our logic states which are collectively intentional (we believe p).7 Furthermore, most social realities cannot even exist without the intervention of language, for our ability to mention abstract entities in a common language is required in order to participate in the kind of complex collaborative actions, which are the basis of our societies. Once again, we find PE as the tacit condition of possibility for the theory to advance: if the meaning of a particular utterance, performed by an individual speaker, were not be translatable into the intentionality of the sentence she uttered—which belongs to language, as a socially shared institution—, the transit from individual to social intentionality would not be conceivable. What is more, we would not be able to ! 16! Spanish language. One more time, we would be facing a categorial mistake, or at least a categorial ambiguity. Once native and type speakers are clearly distinguished for their different logical status, it is easier to find out which of them is the best candidate to get conditions of satisfaction out of sentence meaning, not turning it into mere utterance meaning. If it were a native speaker, that is a token speaker, the difference between sentence and utterance meaning would be blurred, and we would probably be relying too much on the abilities of a particular individual, as I will try to show in the next section. However, if we consider type speaker as the one whose Background is at stake when we are trying to understand the conditions of satisfaction that PE requires from sentence meaning, its logical status would preserve the sentence from just becoming an utterance. Type speaker is simply as abstract as the sentence itself is: it just defines the class of all token speakers that would attribute the same conditions of satisfaction to the same sentences under similar situations—i.e, the class of individuals who share a language. The problem now is that Searle’s uncritical identification of all speakers of a language in the notion of the native speaker calls for further justifications. Whenever we face effective speech act situations, we always have to acknowledge that both speaker and listener are different token individuals, and we cannot simply assume anymore that they fall under the same type just because. Justifications for that statement will prove to be rather elusive. By appealing to the intuition of the native speaker, what Searle vindicates is the importance of first person perspective in order to explain meaning and understanding. It is ultimately there where all the system bears out: in the fact that I know what I mean when I say what I say. But first person authority has certain limitations, and one of them is particularly fundamental here: first person authority does not extend to the formulation of explicit rules. Being a native speaker of English is something very different from knowing by heart the definitions of an English dictionary, or the rules of the English grammar. There is nothing contradictory in the idea of a competent native speaker unable to give any explicit definition of the words she uses, nor of the grammatical rules she follows. Now the point is that the fungible character of native speaker’s utterances stems from the knowing-how of her habits, not from the knowingthat of explicit linguistic conventions. As Searle himself pointed out in a brilliant passage (1995, 140ff), speakers do not have to know explicitly the rules they follow in ! 17! order to follow them. What a native speaker is quite sure to know is just the convenience of each particular application. Rules can be inferred from practice, but first person authority arrives to this inference notoriously weakened. Even if explicit rules are usually not the cause of linguistic behaviour, their formulation plays an essential role in the establishment of shared meanings. We can check if our definitions are common by formulating them explicitly, expecting the acceptance of others. The problem is that we cannot check in that way whether Background is being shared: Recanati’s local PE would allow us to check specific parts of Background, maybe each one of them, but not at the same time. And that is not a method to check the commonality of Background: it would be like believing, as in Augustin’s famous legend, that we are perfectly able to empty the sea due to the fact that, firstly, we can fill a shell with sea water and throw it out of the sea and, secondly, there is no part of water in the sea that could not fit into that shell. We can check if we share definitions, but there is no way to check if we share Backgrounds, which cannot be made explicit by any definite set of conventions. It is rather a kind of fundamental assumption, which makes the acceptance of any convention possible. In this respect, we can establish definitions by convention, but we cannot establish Background by convention, since Background is the basic assumption of any convention. We cannot prove that Background is being shared: we can just expect it to be so. And that is something that cannot be verified but only—in Popper’s terms—falsified, when we notice that two speakers, that seemed to be willing to accept a common definition of a sentence, are in fact conferring to it different conditions of satisfaction (see Dascal 1983: 90). Only the commonality of explicit conventions, but not of Backgrounds, can be simply verified. 7. Saying more (or less) than what we mean to say. I will now use Searle’s well-known example of the dependence of literal meaning on Background in order to criticise the idea that Background can simply be shared. The original example is as follows: ! 18! (a) I am sitting in a restaurant, and order a hamburger. The waiter serves it to me encased in a block of concrete (or sends it to my apartment, or serves a 580 Kg. one…). According to Searle, such a response would not satisfy the conditions of satisfaction of my petition, as Background assumptions would not have been observed. The waiter would not have served what I literally ordered, just like the joker who mowed the cake did not do what I literally asked him to do, when I asked him to cut it. The argument seems to be convincing, but let’s look at the following counterexample: (b) I am sitting in a restaurant, and order a hamburger. The waiter serves a pizza to me, encased in a block of concrete (or sends a pizza to my apartment, or serves a 4 meters in diameter one…). If Searle were right, and outlandish interpretations were simply excluded of the literal meaning of the sentence, (a) and (b) would simply be two cases of the same kind of misinterpretation. But it seems to be clear that they are not so: there is something in (a) which is not present in (b): the hamburger, i.e., what I ordered. In (a) the waiter served it to me in quite an unusual way, but he did serve it, whereas in (b) he just did not. The reason why (a) is a meaningful example—whereas (b) just makes no sense at all—is that it is logically possible to include what the waiter served in the literal meaning of the petition, even if it infringes the conditions of satisfaction expected by the speaker, and his specific Background assumptions. Since it was Searle who offered the first disconcerting example, I beg the reader to let me add a new one: imagine we are sitting in an inter-planetary restaurant, where food must be served encased in concrete, in order to preserve it from long space trips. The very simple question now is: would that excess of imagination also force us to imagine a different language? Would the words “I would like to order a hamburger” mean something different? Would I be able to order a hamburger there in English, in the very same English language? I think the answer is quite plain: yes I would. I simply would be able to order that hamburger with the same sentence, in the same language and with the same literal meaning, but my particular conditions of satisfaction—if I knew this specific need of space trips—would be different, and would probably not include hamburgers served in ordinary dishes. In that imagined situation, (1) would ! 19! perfectly make sense, even under the supposition that the English language is exactly the same. On the contrary, in order to imagine a context for (2) that could transform it into a meaningful situation, we would be forced to change the very conventions of the English language. The moral of the story is that each time a speaker makes an utterance, even if she is a native speaker, she can only consider the bounds and limits of her own Background expectations, but cannot foresee what the community of native speakers of her language would accept or not as a literal interpretation of her sentence. No native token speaker can know a priori the Background of the type speaker: a character that, as a kind of linguistic Leviathan, emerges from the global community of speakers, herself included. She cannot know what English native speakers would accept as a literal satisfaction of her speech act in a different situation, or even in her own particular situation, but considered from a different perspective. By my lights, Searle is right in rejecting the determination view, and holding that there are no conditions of satisfaction in sentences themselves: they do require Backgrounds in order to be really meaningful, for otherwise their connexions with beliefs, perceptions, and other intentional states would be lost (Searle 1979a: 136). But it cannot be denied that the limits of literal meaning—meaning of sentences, not of utterances—cannot depend on the particular Background of a token speaker, no matter how native she is. Sentences can mean much more than what each speaker is able to foresee; the literality of their meaning cannot be as limited as her specific expectations. At first glance, this idea seems to be demolishing for PE: it just seems that, each and every time the speaker makes an utterance, her sentence says more (or less) than what she means to say with it. Summarised, the argument is as follows: (1) the conditions of satisfaction of the sentence used cannot be determined by explicit semantic conventions, since Background assumptions are requested; (2) the Background that should be considered in order to determine the conditions of satisfaction of the sentence qua sentence is not the one of the token speaker, native or not, but the one of the type speaker; (3) we cannot know in principle the degree to which token speaker’s and type speaker’s Backgrounds are shared, since all we can do is trust their commonality, as far as it has not been falsified yet; (4) by using the sentence, the speaker is obliged to accept the conditions of satisfaction that would result from the type speaker’s Background assumptions. In conclusion, (5) each time that a speaker ! 20! produces an utterance, the sentence she utters means more (or less) than what she means to say with it (more because, as a sentence, it is also satisfied by literal interpretations which are unexpected by the particular speaker; and less because, no matter how much she tries to limit those misunderstandings, conditions of satisfaction will always remain underdetermined, since Background is globally inexpressible). Of course, in ordinary life, Background is so widely shared, or so it seems, that we usually get what we want in restaurants, and this intrinsic limitation of expressibility does not generally affect us. However, language theoreticians should not neglect it. 8. Expressibility reconsidered. With the previous argument, I was not intending to prove what Searle repeatedly tries to deny: that PE can fail in particularly marginal or irregular occasions, as there are thoughts that are intrinsically inexpressible. It has not been my aim to prove that there is something in the outer limits of ordinary language that cannot be said, whether in art, religion or other kinds of mysticism. I was rather trying to emphasise, as Scharfstein points out, that “there is something unformulated and perhaps unformulable and finally mysterious in the prosaic, everyday successes and failures of words” (Scharfstein 1993: xvii). The reason why this might happen is that, in each and every use of language, there is certain incommensurability between sentence and utterance meanings. Not being susceptible of propositional expression, the intervention of Background seems to prevent the success of expressibility in general—which would not make PE just false, in particular occasions, but impossible as a rule. It would be completely impossible simply to identify what the speaker means with the meaning of the sentence she uses to say it.15 Nevertheless, as I argued at the beginning, I am not trying to falsify a principle that, being analytic, cannot be directly confronted to facts: PE is not a particular thesis that could simply be true or false, but a methodological strategy that makes Searle’s approach possible, an a priori that provides the backbone for research in different philosophical fields. In this light, we would do well to rethink PE not as a kind of discovery about the working of language—as Searle seems to present it—but as a sort of methodological invention, and one of a kind that it would not be easy at all to go on without it, as I will try to show now. ! 21! Let’s imagine I simply reject the validity of PE, and perform a speech act strictly considering that there is the kind of inadequacy between sentence meaning and utterance meaning I have just pointed out. In that situation, I utter the sentence: (1) The cat is on the mat. My utterance does have a specific intentional meaning because I am able to determine its conditions of satisfaction. As a native speaker, I have interiorised its rules of use, which are fungible to me when I make them work with my Background. In that way, I know what I mean when I say “cat”, “on” and “mat”. However, if by uttering (1) I were just saying this: (2) What I mean by the word "cat" is in the kind of relationship I mean by the word "on" with regard to what I mean by the word "mat", I would not be communicating anything at all. I would definitely not be communicating anything about the world, since the hearer would not know what aspects of reality I am talking about, but it would be a mistake to believe that I would be communicating something about my own mind, for the hearer would not know what I am talking about either—she would have no idea what my “beetle in the box” is (Wittgenstein 1953: §293). If she is able to identify what I mean by those words, that is because their meanings are not only mine, but ours; i.e., their content is able to assume the shape of collective intentionality, and therefore be named in our common language. But then, when I say (1) I seem to be saying something similar to this: (3) What a type speaker means by the word “cat” is in the kind of relationship that a type speaker means by the word “on” with regard to what a type speaker means by the word “mat”. That is to say: if I am effectively performing a communicative speech act, the words that I use do not only mean what I mean by them, but what we mean by them. But, by the same token, the meanings of my words are not determined by my Background, but by the one of the type speaker. Now, if I reject PE, and do not accept that it is always possible to identify native and type speaker’s Backgrounds, I seem to ! 22! be lead to a dilemma: either I am encased in a private language, like in (2), or I will just never be able either to say what I mean nor to mean what I say, loosing control of my own utterances, as in (3). It is obvious that a simple rejection of PE would be quite disturbing for social philosophy, since the commonality of content would be unconceivable; but it would be a mistake to believe that those devastating consequences would be restricted to the collective aspect of the problem. We would also have to pay an excessively high price in philosophy of mind if we just decided to give up PE, since that would prevent us from studying mental intentional states through propositional attitudes. If individual thoughts were not expressible in sentences, we would simply not be able to articulate mind from language, not even from the individual perspective. The problem would not be restrained to the issue of communication, because language does also accomplish an essential role in the development of our cognitive inner abilities. In other words: it is not only a matter of identifying utterance and sentence meanings, but also speaker’s and thinker’s ones. What do the words that I use in my mind mean, not when I utter them on the outside, but when I use them in inner speech, putting my thoughts in order for myself? If they only mean what I want them to mean, (2), how can I say what I think? And if they only mean what the type speaker means by them, (3), how do I know what I think? The attribution of propositional attitudes, not only to others but also to oneself, would be extremely problematic. The possible identity of native and type speakers’ Backgrounds seems to be unavoidable, as it is a necessary condition for me to be able to tell, even to myself, what I am thinking—and maybe even to think it. But we should be aware that this is only a presupposition, and the kind of presupposition that cannot be checked to be true, but only falsified in particular occasions. I do not think then that we should just reject PE, but I think we should acknowledge that the equation it tries to postulate will always remain suspicious, since it cannot avoid the possibility of being affected by semantic excesses or deficits.16 In that sense, the opposite alternative should be acknowledged as equally possible: a sort of Principle of Expression Fallibility (PEF) that would be as follows: ! 23! For any Expression E and any speaker S whenever S utters E as an exact expression or formulation of meaning X, it is possible for E to mean something different from what S means (intends to transmit or communicate) by uttering it. Once again, it would be a de iure principle, not a de facto one: if PE claims that it is always possible to say what you mean, PEF claims that this expression could always fail, and is essentially subject to further revisions. PE and PEF are not incompatible principles, since both are modal judgements of possibility that do not exclude each other.17 On the contrary, they complement each other for, when we do not find a particular expression, PE assures that it can exist whereas, when we think we have found it, what PEF assures that we can be wrong. If we do not find it, it could exist; if we think we found it, it could not be. Under this light, expressibility should be regarded as a process, instead of as a product: the kind of task that could not be definitely accomplished, but must essentially remain in progress, and always to be done. At this point it could be alleged—as Kannetzki (2001b)—that PE should just be abandoned, and substituted by a sort of “principle of explicability”, but this step would be inadvisable at least for two reasons: in the first place, because without PE—as I have tried to show in the first part of the paper—the backbone that articulates speech acts, mental intentional states, social collective institutions and reasons for action would be broken, and that would certainly shake the foundations of the whole system. And, in the second place, because it is not clear if explicability would be understandable at all if we didn’t keep expressibility as a kind of regulative principle—for the same reasons that Popper did not abandon the notion of truth, even if falsability became unpreventable.18 In conclusion: if PE allows a rational approach to language, mind and action, PEF limits the scope of this approach, preserving the elusiveness of human rationality, and advocating us to confront in the theory the unavoidable indefiniteness that constitutes, in fact, processes of communication, expression, translation, or even understanding, that are, and will always be, matters of degree. The acceptance of the tandem PE/PEF would help us to grasp the peculiar character of speech acts, and to be aware of the limits of systematisation in a scientific approach to the nature of thought and language.19 ! 24! Notes 1. For this reason, I find Searle’s defence of expressibility more challenging than Katz’s one, as Katz seems to consider his “Principle of Effability” as an empirical truth, alleging observational evidences for it, and expecting for empirical corroboration (Katz 1972: 18-23). 2. There is some resemblance between Searle’s PE and a central idea of Wittgenstein’s Tractatus Logico- Philosophicus: that “What can be said at all can be said clearly” (Preface). Nevertheless, even if Wittgenstein’s idea of clarity were the same as Searle’s idea of literality—which is not the case—, an important difference should be pointed out: Wittgenstein assures that what can be said, can be said clearly, whereas Searle’s point is that whatever the speaker intends to say, can be said. Searle’s claim is stronger than Wittgenstein’s, since the former assumes that everything can be said, whereas the latter only claims that what can be said, can be said in a particular way. The author of the Tractatus would probably not accept PE, as it would abolish the distinction between saying [sagen] and showing [zeigen] (4.1212): for him, there is something we cannot say, something we can only show, the mystical, “Whereof one can not speak” (7). And obviously, in his opinion, it is possible to intend to communicate what cannot be said, since he recommends us not to do it. 3. Most griceans would probably not accept this dependence, since it would imply a kind of circularity in the analysis of meaning—unless we assume that analysis must not be reductive to be enlightening (see Avramides 1989). Furthermore, Grice does not restrict communication to the utterance of sentences. 4. Criticism of literal meaning is sometimes directed towards its empirical lack of evidence, as people in fact do not seem to go across literal meanings in order to grasp utterance meanings; see for instance the Dascal-Gibbs debate (Dascal 1987, 1989 and Gibbs 1989, 2002). In any case, what is at stake here is not the psychological role accomplished by literal meaning, but the logical necessity of literality in order to provide a philosophical account of speech acts. In fact, many of those who explicitly criticise literality from an empirical point of view, offer argumentations that are logically dependent on it. Once again, analytical principles seem to be immune to empirical approaches. 5. According to Searle, mental intentionality seems to be logically previous to its expression in a shared language, which leads to a puzzling dilemma (Kannetzky 2001a): how should we consider this priority, other than by the postulation of a private language? And how are we supposed to establish its connections and translations with public languages? If we deny the existence of private languages, the development and enlargement of language seems to be a mystery (where do new meanings come from?). But if we accept that they do exist, and we therefore consider the speaker’s intention as something given before it is expressed in a public language, what we are unable to explain is mutual comprehension. Kannetzky is right when pointing out the relationship of this puzzle with PE, but he does not seem to notice that it is precisely Searle’s intention to avoid the appearance of the dilemma a priori. Blinkey (1979) also ! 25! considers PE incoherent by its pretensions to compare expressed meanings of utterances with unexpressed meanings, which would not be identifiable on their own. 6. In Scharfstein’s words (1993: 11), “we generate ideas from a substratum that is opaque to consciousness and resists transmutation into concepts. It is true that part of our more abstract thinking may be done in snatches of inner speech, but external speech is not ordinarily generated from preceding speech of any kind”. 7. The refusal to introduce collective intentions in gricean analyses of meaning seems to be the reason why they cannot avoid to keep a kind of vanishing point, and always seems to leave us “waiting for the next ingenious counter-example” (Avramides 1999: 73). 8. I am not saying that only literal speech acts do lead to collective intentionality. On the contrary: metaphors, ironies, or even non-conventionalised ways of communication are perfectly able to fulfil that task. My point is that, whether in conventionalised or non-conventionalised situations, at least from the point of view of the analyst, literality must be acknowledged as a possibility, since that is the only final move which could make the collectivization of content explicit: the possibility of literalizing what is being said and, in that way, discarding any ambiguity in the utterance. In this sense, Grice’s analysis of meaning (Grice 1957), which attempts to give an account of the rise of literality from non literal uses of language, might also presuppose PE, for it accepts that, whatever the speaker intends to communicate (literally or not), it can assume the shape of a particular intention that can be expressed by the first clause of the analysans—a clause expected to be literally understood by the community of scholars who read his paper. Even if literality is absent from the analysandum, and we can get rid of the code-model in order to explain some communicative acts, Grice has to accept it as a precondition for the use of the metalanguage that paraphrases it. From this point of view, code-models and inference-models of communication are not so opposed as it was originally thought, a point that facilitates their articulation in complex theories of communication—as Sperber and Wilson (1986: chapter 1.5) or Dascal (2003: 507- 520) have pointed out. The reason why they are not incommensurable might be that they do share a basic common assumption: that expressibility is, in the end, at least possible. 9. For a seminal view on this issue, see Ryle (1949, chapter 2). Brandom (1994) has taken this as a key point in his account of explicitness. 10. Meaning 3.b of the Merriam-Webster Online Dictionary (www.merriam-webster.com), last accessed January 30, 2009. 11. The idea of Background should be complemented with the one of Network: intentional states are not completely independent from each other; on the contrary, many of them require other intentional states in order to be effective. In this way, most beliefs and desires are articulated in a Network of other beliefs and desires (Searle 1983: 65). But I will not focus on this idea here because, unlike Background, Network