“The Signifying Animal” | Open Indiana

When a diplomat says “yes”,	When a lady says “no”,
he means “perhaps”;	she means “perhaps”;
when he says “perhaps”,	when she says “perhaps”,
he means “no”;	she means “yes”;
and when he says “no”,	and when she says “yes”,
he is no diplomat.	she is no lady.
	—Voltaire

One need be neither a diplomat nor a lady to use the word perhaps to mean ‘yes’ one time and ‘no’ another. But what is the meaning of a word like perhaps if everyone can make it mean either ‘yes’ or ‘no’ as he pleases? Can one in any sense talk about a fixed word meaning here? But if not, what is it Voltaire is telling us when he maintains that the diplomat’s uttering “yes” as well as the lady’s uttering “no” means ‘perhaps’?

What is said and what is meant, coded information and its use in communication, do not seem to coincide in all cases. But, when they don’t, which content elements of an utterance should be traced back to word meanings and which to the specific use of the words in the situation in question?

In semiotic terms, I am going to deal with the delimitation of semantics and pragmatics in the description of verbal communication. In treating this problem, I shall discuss the following points:

1. Two strategies for the description of verbal communication

2. The monism of meaning

3. The monism of use

4. The identification of meaning and use

5. The use of meanings

6. Meaning versus suggestion

7. Sentence connectives: maximization of meaning

8. Sentence connectives: minimization of meaning

9. The duality of semantic and pragmatic interpretation.

1. TWO STRATEGIES FOR THE
DESCRIPTION OF VERBAL
COMMUNICATION

In uttering verbal expressions to achieve communicative goals, we must be able to use these expressions appropriately. But how do we proceed? Does every word have a fixed meaning that language users reproduce in their utterances? Or are there no fixed meanings but only rules of use that guide the language users in their formulations?

These questions delineate two competing strategies for the description of verbal communication. For linguists pursuing the first strategy, verbal communication can be exhaustively described by reference to word meaning, sentence meaning, and the meaning relations holding among verbal expressions. Linguists pursuing the second strategy seek to avoid assuming the existence of meanings; they try to describe the same phenomena in terms of language use. Both approaches have come to develop highly elaborated and finely articulated systems of terminology:

— While proponents of the first strategy speak of meaning, semantic features and the realization of semantic features, proponents of the second talk about use, rules of use and the application of rules of use;

— while one side speaks of semantic categories, of literal versus transposed meaning, of uniqueness of meaning versus ambiguity, the other is concerned with modes of use, literal versus transposed use, consistent versus inconsistent use;

— while one side speaks of hierarchies of semantic features, involving union, intersection, and opposition of features, the other discusses hierarchies of semantic rules, involving union, intersection and opposition of rules;

— while one side speaks of presuppositions, the other talks about conditions of use; constraints on the realization of semantic features appear as constraints on the application of semantic rules; feature change appears as rule change; modifications in the meaning of a word appear as modifications in the use of that word.

These two strategies I will call the monism of meaning and the monism of use.¹

Comparing the two strategies in such an abstract manner might lead one to regard them as merely terminological variants of the same theory, equally applicable and equally efficient in the description of verbal communication, were it not for the fact that they involve different empirical hypotheses and apply different methods of investigation: Monists of meaning usually assume that we have direct empirical access to word meanings and that rules of word use, if such things exist, are easily derivable from word meanings. Monists of use tend to believe that only the use of a word is empirically accessible, and that the meanings of that word, if such things exist, must be derived from its use. Common to each monism, however, is the assumption that its respective approach can give a complete linguistic explanation of how language functions.

In what follows I will show that both positions fail to satisfy some rather simple theoretical and methodological requirements for the description of language. Focusing on sentence connectives, I am going to argue against monistic approaches and in favor of a theory that assigns complementary roles to the meaning and to the use of words in verbal communication.

Since the discovery of truth-tables in the logic of the late 19th century, many logicians have offered an explication for the so-called logical particles of natural language. Nevertheless, doubts about the adequacy of such explications have never stilled. Recently, this century-long discussion has been given a new turn by the American philosopher of language Herbert Paul Grice, whose ‘Logic of Conversation’ indicates a new and promising direction to take in the description of sentence connectives in natural language.² Grice’s approach has not been received without controversy, however, and the essential arguments against it must be taken into account as well. In order to show the proper role Grice’s approach can play in empirical semantics, I shall begin by sketching the history of the monisms of meaning and of use, including some of their psycholinguistic ramifications.

2. THE MIONISM OF MEANING

At the end of the 19th century the conceptual semantics inherited from the Age of Enlightenment came under the influence of a newly developed empirical discipline, psychology. This interaction resulted in a linguistic conception that can he characterized as follows. To communicte is to transmit concepts and in order to transmit concepts one must utter words whose meanings are concepts. According to Wilhelm Wundt, “every independent conceptual word [ . . . ] elicits a certain conceptual idea that is vivid to the degree to which the meaning of the word is concrete.”³ Edward Titchener taught, “The word dog has a meaning for us because the perception of this word elicits the idea of a dog in us.”⁴ While not every word is an independent conceptual word and, as such, able to transmit an idea, every word has (at least) one constant meaning which, when combined with the meanings of other words, contributes to the meaning of the expression as a whole. Thus, a speaker transmits the intended concepts by uttering the appropriate words, with which the addressee associates appropriate conceptual ideas of his own. Associations can be interpreted and tested according to the mechanism of stimulus and response. In order to find out which ideas are associated, addressees are subjected to standard tests such as drawing the associated objects or picking out their putative qualities from a list.

Unfortunately, the association experiments did not, when carried out on a large scale, confirm this theory in the expected way. The ideas elicited by a particular word differed greatly according to the context. Thus, it gradually became a commonplace of meaning-monist word semantics to claim, “The boundaries of word meaning are vague, blurred, fluid.”⁵ Even linguists like Karl Otto Erdmann, who took such statements as the starting-point for the introduction of new distinctions in his ‘Essays from the Borders of Psycholinguistics and Logic’, were not able to neutralize their force.

3. THE MONISM OF USE

The consequences of this failure were drawn by behaviorist psychology. If the association test docs not provide reliable access to word meaning, it is useless as a linguistic method. And, if we arc forced to assume that word meanings are fluid, then the concept of meaning loses its theoretical value for the description of verbal communication as well. John Watson formulated this insight most emphatically, “From the position of a behaviorist the problem of meaning is a mere abstraction.”⁶ And Burrhus F. Skinner laconically concurred, “The speaker does not utter any ideas or images but only words.”⁷

Even Charles W. Morris prided himself on being able to do away with meanings altogether. Indeed, he thought he had renewed the connection with experimental psychology by taking dispositions into account in his semiotic program.⁸ Whereas association psychologists had put word and meaning on the same level with stimulus and response and had regarded word meaning as a directly observable variable, Morris took word usc to be a disposition that mediates between stimulus and response in the communication situation. Being dispositions, the rules of use cannot be directly observed but must be inferred from manifest behavior in the form of intervening variables. Thus, in the case of word use the hypothesis of direct empirical accessibility was abandoned just as it had been in the case of word meaning.⁹ Not even the difficulty encountered by Erdmann disappeared for monists of use; it merely became more clearly formulated. How one and the same word can play a role in fundamentally different behavioral dispositions is a question still looking for an answer.

4. THE IDENTIFICATION OF
MEANING AND USE

These empirical problems and methodological difficulties deprived the two monist positions of their initial theoretical attractiveness. To be sure, it was not only for methodological reasons that Ludwig Wittgenstein advised, “Don’t ask for the meaning, ask for the use.”¹⁰ He believed that he could also give this maxim theoretical underpinnings, claiming, “For a large number of uses of the word ‘meaning’—even if not for all its uses—this word can be explained thus: The meaning of a word is its use in the language.”¹¹ But Wittgenstein failed to develop satisfactory methods for systematically collecting and describing word uses and restricted himself to analyzing examples. For this reason, he prepared the way for an elaboration of the use-monist terminology rather than for the introduction of explicit criteria justifying its application. What survived was the slogan in its truncated version: “The meaning of a word is its use.” Because of the lack of guiding criteria it was no surprise that this identification of meaning to use was soon made to apply in the opposite direction. This has led to today’s uncritical contamination of the terminology of meaning with the terminology of use.

Of course, even such contamination can be justified, if only in the form of an analogy: are not meaning and use just two aspects of one and the same phenomenon, like temperature and particle velocity in thermodynamics or like wave and corpuscle in quantum physics? Yet, whoever holds this view, must take into account that physics has confirmed each of the two aspects experimentally so that the coexistence of the two terminologies is justified empirically.

But, can this be said of the two aspects of language?

The lack of direct experimental evidence for the existence of meanings and of rules of use has already been pointed out. Nevertheless, it must be conceded that there are phenomena in language that are more easily describable in terms of meaning and that there are other phenomena for which a description in terms of use is more convincing. For example, consider the German color words gelb and blond. If we take these words to denote those parts of the color spectrum that are reflected by yellow and blond objects, respectively, then gelb and blond have approximately the same meaning. They are, however, used in different ways; indeed, their uses in German are mutually exclusive: blond generally refers to human hair; everything else reflecting the color waves in question is called gelb or gelblich. It is true, there are certain marginal uses as in: Herr Ober, bitte ein kühles Blondes! (“Waiter, please bring me a light ale’). But even here there is no mutual substitutivity, since no one would say, Herr Ober, bitte ein kühles Gelbes! The words gelb and blond can therefore be said to have the same meaning but to be in complementary distribution.

The following example is equally instructive. When we hear little Peter say at the hairdresser’s: “Mommy, look, Annie is being mowed,” we smile, but we understand what he wants to say. Mow means the same as cut, but its use is restricted in such a way that human (and animal) hair is excluded.¹² What we must take into account here are the conditions of use that restrict the occurrence of a word beyond what its meaning would allow. In the lexicon such conditions of use normally appear as parenthetical supplements to the specification of meaning, e.g., “blond: ‘yellow (of human hair)’ ” and “mow: ‘cut (of grasslike vegetation)’.”

In recent years it has become commonplace to regard conditions of use as presupposed semantic features of a word and thus to integrate them into the meaning of the word. Correspondingly, blond would mean something like ‘having yellow human hair’ and mow something like ‘cut grasslike vegetation’. But such a solution can be maintained only with great difficulty.¹³ Once again color words demonstrate quite clearly how indispensable the specification of conditions of use can be. If the color of a cotton coat that Germans call rot (‘red’) occurs on a plastic wall, they tend to call it braun (‘brown’). Thus, the color word is changed when the area of application changes, even though the color waves have remained the same. Monists of meaning might try to cope with this situation in the same way as they did in the cases of gelb and blond; they might say: The meaning of rot includes the presupposed semantic feature of ‘having a rough surface’, and the meaning of braun includes the presupposed semantic feature of ‘having a smooth surface’. But this is plainly wrong, for it would also be perfectly acceptable to call a rough surface braun and a smooth surface rot. In contrast to the cases of gelb and blond, the transfer from one area of application to another does not affect the acceptability of use of the words rot and braun, but their meaning. If applied to smooth surfaces, rot denotes a different section in the color spectrum than if applied to rough surfaces. The same is true for braun. There seems to be a mutual dependency here between the meaning and the context of use and that makes sense only if meaning and use are independently characterizable.

These brief examples may suffice to call into question the identification of meaning and use and to show that the two kinds of terminology are not interchangeable in all areas of lexical semantics.

5. THE USE OF MEANINGS

The difference between semantic features and conditions of use has been elaborated most clearly by Grice. In his William James Lectures, held at Harvard in 1968, he succeeds in liberating himself from the simplistic opposition of “meaning versus use.”¹⁴ In contrast to Watson and Skinner, Grice is prepared to claim that in communication we deal not only with words or concatenations of words but also with word meanings and sentence meanings. What is involved here is the use of meanings. Grice strives to discover the rules according to which we use meanings in order to achieve our communicative goals.

Let us compare the following five sentences:

(1) Please take the garbage out.

(2) (a) I want you to take the garbage out.

(b) Can you take the garbage out?

(d) Have you taken the garbage out?

In everyday family communication each of these five sentences evidently can be uttered for the same purpose, that is, to express a request for someone to take out the garbage. Nevertheless, when we isolate each of these sentences from its context and describe it as such, we rightly say: (1) is an imperative sentence, (2a) is a declarative sentence, and (2b) through (2d) are interrogative sentences. But how can one, in uttering an assertion or question, get someone to fulfill a request? How can we explain that the son who hears his mother ask, “Have you taken the garbage out?” understands this as a request for him to take the garbage out? This is the sort of problem that concerns Grice.¹⁵ In order to solve it, we must take into account the context of utterance and apply certain principles from the theory of rational behavior. Let us begin with the situation in which mother and son are standing in the kitchen in front of the full garbage can. Both can see with their own eyes that the garbage can is full and can infer that it has not yet been taken out and emptied. Nevertheless, the mother asks her son, “Have you taken the garbage out?” If he takes this sentence in its literal meaning, its utterance can only be a question for information. But as such it would lack any communicative function. The son knows that the question would have to be answered in the negative and he knows that his mother knows that, too. In this situation that question seems to be absurd and its utterance by the mother irrational. Rational behavior would conform to the following maxim:

(M₁) Make your contribution such as is required, at the stage at which it occurs, by the accepted purpose of the action in which you and your partners are engaged.

This maxim is called “cooperation principle” by Grice; by appropriate specification it can also be applied to communication, understood as verbal cooperation:

(M₂) Make your utterance such as is required, at the stage at which it occurs, by the accepted purpose of the talk-exchange in which you and your partners are engaged.

What purpose might the mother be pursuing when standing with her son in the kitchen in front of the full garbage can? From the rest of his mother’s behavior the son knows, “Mother likes everything to be clean and neat and wants me to be so, too.” But does the mother’s question serve this purpose? “If taken literally, no.” Does it serve another purpose? “If taken literally, no.” Did the mother want to say something pointless? “No, this is something she normally does not do.” In order to avoid such a conclusion, the son prefers to reinterpret the sentence and to give the utterance a new meaning by relating it to the accepted purpose of the talk-exchange: “Mother likes everything to be clean and neat and wants me to be so, too. I would be neat if I took the garbage can out and emptied it. Mother’s question evidently is supposed to draw my attention to a fact which she and I already know but from which I have not yet drawn the desired consequence. Evidently, it is desired that I should now take out the garbage can and empty it. Since I would like to continue communicating with my mother and to achieve my own communicative purposes, too, I will take the present purpose of communication seriously and behave as if mother had requested me to take out the garbage can and empty it.” These considerations lead the son to the result that he has to treat the question as a request. Reasoning of this kind is normally performed automatically and remains below the threshold of consciousness. But it has to be taken into account if the behavior of the addressee is to be adequately explained.

Of course, the utterer regularly anticipates such reasoning. In fact, the mother only formulated her utterance as a question because she could assume that her communication partner was in a position to reconstruct the intended request from it. She used the literal meaning of (2d) in order to convey the literal meaning of (1). In doing so, she not only combined words with lexical meanings according to grammatical rules to produce the literal meaning of a sentence, but also used the sentence meaning according to pragmatic rules to convey a message normally conveyed with different words and different grammatical rules. She used an assertion or a question in order to suggest a request.

But in what kind of situation does one prefer to rely upon the reasoning of the addressee instead of saying literally what one wants to convey? The answer becomes obvious if we consider how an obstinate son could respond to the utterances (1) and (2) if he took them literally:

(1’) No, I won’t.

(2’) (a’) You can do it.

(b’) I can, but I don’t want to.

(c’) I will, but right now I’m doing something more important.

(d’) No, because I have more important things to do.

As these examples demonstrate, the utterances under (2) leave open to the addressee the possibility of understanding them not as requests but as assertions or questions and to respond to their literal meanings. Interpreted in this way, they can provide an opportunity for a discussion whether the conditions for a sensible request are satisfied at all. In using such a formulation the utterer makes it possible for his addressee to argue against making a request before such a request has in fact been uttered; he spares him an open conflict and avoids losing face himself. In short, the use of meanings like those of (2a) through (2d) to suggest a request is more polite than the use of the literal meaning of (1).

If this explanation of the function of suggestions proves right, it gives us a strong argument for the existence of meanings; for, the entire explanation would collapse if meaning were reducible to use.

6. MEANING VERSUS SUGGESTION

The construction of suggestions originates in the effort of the addressee to interpret the verbal behavior of his communication partner as rational behavior. If the speaker utters a sentence whose meaning, taken literally, does not contribute to the recognized purpose of communication, then the addressee asks himself if the speaker means something different from what he has said literally. He evaluates the verbal and non-verbal context of the talk-exchange, looking for supplementary information that, applied to the literal meaning, will let him infer a message conforming to the recognized purpose of communication.¹⁶ This reasoning is a heuristic operation; it follows certain rules, but its results are not strictly deducible since it often remains unclear what the purpose of communication is and which circumstances of the context are relevant to it.

In order to get such a reasoning process going, it is important to discover which maxim of rational behavior the utterance would have violated if it had been taken literally. Therefore, Grice has tried to supplement his principle of cooperation with a series of special maxims which are valid in particular for the exchange of information during a conversation.¹⁷ Grice sets up these conversational maxims by facetiously employing Kant’s table of categories:

(M3)I Maxims of Quantity

1. Make your contribution as informative as is required . . .

2. Do not make your contribution more informative than is required . . .

IIMaxims of Quality

1. Do not assert what you believe to be false.

2. Do not assert that for which you lack adequate evidence.

III Maxim of Relation

Say only things which are relevant . . .

IV Maxims of Manner

1. Avoid obscurity of expression . . .

2. Avoid ambiguity . . .

3. Be brief (avoid unnecessary prolixity) . . .

4. Be orderly . . .

The dots following these formulations indicate that the maxims of quantity, relation and manner can only be understood with respect to the purpose of communication accepted for a particular stage of a conversation.

This list of maxims is neither complete nor systematically organized in a satisfying way, and the individual maxims are neither of equal importance nor completely independent from one another (e.g., compare the relationship between I/2 and III). However, these flaws do not rule out the possibility that the maxims mentioned are actually applied in the production and interpretation of conversational suggestions. Therefore one must also take them into account when describing these processes.

How this is done may be illustrated with an example that has given logicians many headaches: A First Mate does not get along well with his Captain. The Captain is a prohibitionist and the Mate is often drunk. Therefore the Captain is looking for a pretext to have the Mate fined when the ship comes to port. One day, as the Captain has the watch, the Mate starts bellowing out a sea chantey again. The Captain can stand the Mate’s excesses no longer and writes in the log:

(3) (a) Today, March 23rd, the Mate was drunk.

A few days later, when the Mate himself has the watch, he discovers the Captain’s entry in the log and wonders what he can do about it without compromising himself any further. Finally, he also makes an entry in the log, which reads:

(3) (b) Today, March 26th, the Captain was not drunk.

This is not an ordinary conversation, but Grice’s conversational maxims are nevertheless applicable, since the institution of the log serves an accepted purpose of communication that can be realized by following the maxims. Both entries are true statements; however, there is an important pragmatic difference between them, which is revealed by the reader’s reaction. Whereas the Captain’s entry is interpreted and understood without hesitation, any reader who comes across the Mate’s entry cannot help asking, “Why is that written in here? What relevance can the statement have in a log that the Captain was not drunk on a certain day?” Once the reader has established that this entry would, if taken literally, violate the maxim of relation, the next steps of his reasoning are easy. “If the writer wanted to establish communicative cooperation with the reader of the log at all, he must have considered this entry relevant himself. A log serves to register exceptional occurrences on a voyage. Evidently the writer wanted to indicate that the Captain’s sobriety on March 26th was exceptional. Sobriety is, of course, exceptional if one is usually drunk. Under these circumstances the writer wanted to suggest with his entry that the Captain was usually drunk during the voyage.” Thus the reasoning prompted by the Mate’s entry has, on the basis of assumptions about the purpose and context of communication, turned a trivially true statement into a false statement of a rather defamatory nature. This example shows how one can lie with true statements when the utterance of those statements violates one of Grice’s conversational maxims. In our case it is the maxim of relation (III) that is involved. There are similar examples for the violation of other maxims.

What is peculiar in this defamatory suggestion is the fact that it is hardly possible to take legal action against it. When confronted with the alternative of calling the Mate to account for false statements or for disorderly conduct, any court would choose the latter.

The two examples from household and sea demonstrate that the discrepancy between what an utterer formulates and what he intends to convey can be explicated on the basis of the distinction between literal meaning (which is determined grammatically) and suggested content (which is determined pragmatically). Furthermore, our analyses of these examples have shown how the utterer proceeds to produce conversational suggestions on the basis of literal meanings and how the addressee proceeds in his efforts to reconstruct these suggestions from the literal meanings. If these analyses are not misguided, they force the linguist to postulate duality in the description of verbal communication. He must not only determine the literal meaning of the verbal expression uttered but must also examine how the utterer uses this meaning.

Illuminating as this conception may seem, it leads to theoretical and methodological questions that are hard to answer. How can a linguist determine which content elements of a given message must be considered as the literal meaning of the words or sentences uttered? Can we rightly claim that the sentence Have you taken the garbage out? is an interrogative sentence, if it can obviously be used as a request? If the negation of the word drunk can be used to indicate that the person in question was the opposite of sober, why do we continue connecting its affirmative with the opposite of sobriety?

One thing is certain: an undifferentiated treatment of all the uses we can find will not bring us any nearer to the literal meaning of a word or sentence. Rather it is necessary to select from among the many uses of a word or sentence those uses in which the literal meaning does not appear to be subjected to any modifications required by specific features of the verbal or non-verbal context.

However, by being asked in this way the question is in danger of becoming circular. We are saying: (a) the meaning of a verbal expression will not be submitted to a context-dependent reinterpretation if its utterance does not violate any conversational maxims; (b) the utterance of a verbal expression does not violate a conversational maxim if it is unnecessary to reinterpret its literal meaning in a context-dependent manner. In this perspective, the literal meaning appears to be like Wittgenstein’s beetle in a box: even if we assume that the beetle exists, we cannot tell how big it is.¹⁸

There seems to be only one way out of this dilemma, and that is the attempt to reconstruct the process of comprehension itself.

1. According to our initial assumption, the addressee proceeds from the literal meaning of an expression and, on this basis, establishes certain conversational suggestions corresponding to the particular features of the verbal and non-verbal context. A comparative analysis of the comprehension processes for all essential uses of an expression could thus furnish us with those content elements which are always involved, as against those elements that play a role only in certain classes of context. We may assume that the content elements involved in the comprehension of all the uses of an expression belong to the literal meaning of that expression; as to the other content elements, we may conclude that they are dependent on special circumstances of communication and are produced only in the process of special interpretive reasoning. This is the postulate of variability for suggestions.

2. Since conversational suggestions change as the situation of conversation changes, we can cancel them through the choice of certain contexts. Even simple verbal additions will do the job, and by claiming the contrary we can annul an alleged suggestion without giving rise to a contradiction. This is the Gricean postulate of cancellability for suggestions.¹⁹ If, after asking one of the questions (2b) through (2d), the mother in the kitchen had added, “But I’m not requesting you to do it now,” no request would have been suggested. Likewise, the Mate could have avoided a defamatory use of his entry without contradicting himself, if he had supplemented (3b) by the sentence, “The Captain is never drunk.” Such additions cannot, however, prevent other suggestions from arising, in case the complete utterance, understood literally, still violates some conversational maxim.

3. Finally, one cannot avoid a conversational suggestion by simply choosing another formulation with the same literal meaning. Suggestions of the relevant sort do not result from the use of special words but rather from the specific use of meanings. Therefore a suggestion generated by a particular utterance in a given situation is detachable from the words, but not from the literal meaning of that utterance. This is the Gricean postulate of nondetachability for suggestions.²⁰ In our examples, the mother’s suggestion would not have been changed if she had said, “Has the garbage been taken out?” and the Mate’s suggestion would not have been changed if he had written, “Today, March 26th, the captain was sober.”

Variability, cancellability and non-detachability are useful indications, but, unfortunately, they are not sufficient as criteria in determining which content elements have to be excluded from the literal meaning of an expression.²¹ Nevertheless, we have to work with them, as long as there are no better analytical instruments. The imperfections of this procedure only confirm once more what meaning-monists and use-monists had to discover by experience, namely that neither the meaning of a word nor the rules for the use of a word are directly accessible to the empirical linguist, but must be inferred from manifest verbal behavior.

Now if it is theoretically certain that all verbal behavior is based upon literal meanings as well as upon rules for the use of meanings it becomes necessary in the course of linguistic analysis to estimate how much of the content of a given utterance may be traceable to its literal meaning and how much is to be construed as suggestion. This is a procedural problem that gains importance in view of the fact that there is hardly any word of a natural language whose uses have been exhaustively analyzed. As long as we are in the position of creating hypotheses, we can again choose between two strategies that take up the positions of the old monists in a weakened form. Meaning-maximalists attempt to deduce as much as possible from the literal meanings of verbal expressions and tend to assume richness and ambiguity in the meanings of words. On the other hand, meaning-minimalists attribute more importance to the pragmatic rules of reinterpretation as opposed to literal meanings and tend to accept only minimal meanings and unambiguous words.²² Let us now consider the consequences of these strategies for the analysis of sentence connectives in natural language, particularly, of the word and in English.

7. SENTENCE CONNECTIVES:
MAXIMIZATION OF MEANING

When the logical particle et (written as “&”, “˄”, “.” according to the various notational conventions) occurs between two propositions, it turns them into one complex proposition that is true if and only if both constituent propositions are true. This statement, which defines the connective et of propositional logic, also seems to apply to the word and as found between declarative sentences of English. What would then be easier than to assume that this definition also characterizes the meaning of this sentence connective?

However, the truth-functional definition of and has consequences that run counter to many uses of the word and in natural language. For example, it allows sentences to be connected to one another without regard to their meaning. But any speaker of English would consider the following expressions absurd and unacceptable:

(4)(a) 2 X 2=4 and it is impossible to analyze further the concept of intention.

(b) Müller just scored a goal and eels spawn in the Sargasso Sea.

Moreover, the truth-functional definition of and places exactly the same condition on the two connected declarative sentences; so they should be interchangeable. But any speaker of English will interpret (5a) differently from (5b):

(5)(a) Peter married Annie and Annie had a baby.

(b) Annie had a baby and Peter married Annie.

Of course, these observations are not contested by anybody. What is controversial is how they should be explained.

The meaning-maximalist draws the following conclusions: The meaning of the word and is richer than the meaning of the logical connective et; it includes not only the truth-functional feature of conjunctivity, but also the feature of connexity and the feature of successivity. On the basis of connexity, the and-sentence conveys that the facts described in the second constituent sentence are part of the same situation as the facts described in the first. On the basis of successivity, it conveys that the facts described in the second constituent sentence appear at a later time than the facts described in the first.

Such an analysis of meaning, however, is liable to quite a number of objections.

1. What about the three postulates of variability, cancellability and nondetachability? It is by no means true that every use of the sentence connective and implies a temporal sequence between the facts described:

(6)(a) 2 X 2 = 4 and √4 = 2

(b)The moon revolves around the earth and the earth revolves around the sun.

This shows that successivity is variable and not fixed to the word and. The assumption of successivity can also easily be cancelled by an additional utterance of the right kind. Continuing with:

(7)But I don’t know in which sequence that happened.

after having said (5a), rules out the basis for the conclusion that the baby came after the wedding. On the other hand, one cannot get around the successivity assumption by merely reformulating the constituent sentences in such a way as to preserve their meaning. The successivity assumption is non-detachable.

These observations indicate that successivity should be regarded not as a semantic feature of the word and, but as a conversational suggestion. Whenever we use coordinate sentences to describe events in time, we relate the sequence of the sentences uttered to the sequence of the events they describe, even without the help of the word and. The utterer would volate the conversational maxim “Be orderly . . .” (IV/4), if he were not to keep the temporal sequence parallel on both levels.

However, such an objection will not deter a meaning-maximalist, since, according to him, the absence of assumptions about the sequence of the facts described in (6) should be explained in a different way. Even when we speak of chocolate hearts, paper tigers, or roses of glass, we do not imply that the hearts beat or that the tigers and roses are alive. Evidently, quite simple syntactic procedures like the addition of an attribute to a noun can lead to the deletion of semantic features of this noun.²³ This deletion occurs during the amalgamation of the semantic features, in order to satisfy a requirement to avoid contradiction. Let us apply this explanation to our example (6): If one had to assume both that the facts described in (6) are time-independent and that they occur one after another, a contradiction would arise. So the addressee, preferring a non-contradictory interpretation, will delete the feature of successivity in his interpretation, according to the meaning-maximalists.

With this procedure we now have two proposals for grasping variable content elements. These approaches impute converse operations to the process of comprehension. According to one approach, the addressee proceeds from a literal meaning with few semantic features and reaches the required interpretation with the help of conversational maxims relying on additional information specific to the verbal and non-verbal context of utterance. According to the other approach, the addressee proceeds from a rich literal meaning and deletes, according to certain preference rules, those semantic features which would come into conflict with the verbal or non-verbal context.

What is remarkable here is that both approaches are based on a similar theoretical apparatus. As shown by the requirement to avoid contradiction, even the meaning-maximalists need additional pragmatic maxims of interpretation besides grammar and the lexicon. And if even meaning-maximalists cannot manage without pragmatic rules, then we are justified in asking why they trust this instrument so little, why they apply it only to restrict meaning and not to produce new content. One might suppose that they have become victims of hypostatizing their own concept of meaning.

2. In order to support the meaning-maximalist analysis of the word and, one often compares it with the word but.²⁴ But seems to share with and the semantic features of conjunctivity and connexity. Instead of successivity it has adversativity as its third semantic feature. Someone who says but implies that the facts described in the following sentence are unexpected or contrary to the present context.²⁵

However, this parallelism is misleading. In contrast to the successivity of and the adversativity of but is not cancellable. Someone who says:

(8) (a) Annie is Martha’s daughter, but she is married to Peter.

and then goes on to say:

(9)However, I don’t mean to say that there is an opposition between the two facts.

will not be taken seriously, since there is then no way to explain his saying but in (8a). Finally, non-detachability does not hold for adversativity. In many cases the assumption of adversativity will vanish if the but-sentence is reformulated in such a way that the rest of its content is preserved. Let us compare (8a) with (8b):

(8) (b) Annie is Martha’s daughter and she is married to Peter.

In (8b) there is no longer any trace of unexpectedness or opposition.

Besides, whereas an and-sentence may or may not convey successivity, depending on the context and on the facts described in the constituent sentences, but without adversativity is unthinkable. In contrast to paper tigers and chocolate hearts, this semantic feature never disappears, even in the face of a possible contradiction:

(10) (?) 2 X 2 = 4, but 2 X 2 = 4.

We will tend to simply reject an utterance like (10) as unacceptable instead of asserting that the meaning of but is reduced to the features of conjunctivity and connexity here.²⁶

This and other observations will also make us wonder about the meaning-maximalist explication of and. It seems that the model of feature amalgamation in attributes and nouns is inappropriate as a model for the interpretation of sentence connectives.

3. There are, however, stronger objections to the meaning-maximalist description of utterances with and. Let us compare the sentences (11a) through (11g) with the versions under (11’), in which the word and has been replaced with a lengthier formulation and at least one possible interpretation of the respective initial sentence is preserved.

The reformulations show right away that successivity need not play a role at all in and-sentences, even for events bound in time. The strategy of deletion of semantic features in the examples under (11) will not help either since other relations besides conjunctivity and connexity are expressed between the facts described by the constituent sentences. And those relations can in no way be acquired on the basis of successivity.

The sentences under (11) leave the meaning-maximalist no other choice but to assume that the word and is ambiguous. He will say: There is not only a successive and (as in (d)), but also a simultaneous and (as in (b)), a local and (as in (a)), a directional and (as in (c)), and an instrumental and (as in (e)), a conditional and (as in (f)), and an explanatory and (as in (g)). He will be inclined to assume that all these different ands share the semantic features of conjunctivity and connexity and differ only with regard to the third semantic feature.

However, even this position seems plausible only as long as no further questions are asked. To begin with, one must observe that the word and also occurs in the reformulations under (11’), which are supposed to make the content of the original sentences more explicit. If the word and on the left means the same as one of the expressions listed on the right, i.e., and there, and during this time, and after that, etc., then what does the word and mean in these expressions? It is obvious that none of the seven meanings already mentioned can be considered. Now what about obtaining the meaning of this and by deleting the third semantic feature of one of the complete ands on the left? This proposal could be pursued. But from which of our seven ands should we proceed? From the successive or the simultaneous, the instrumental or the conditional and? As long as such a question cannot be answered, it would seem to be easier to postulate an eighth meaning. This one, though, would already closely approximate the truth-functional meaning of the connective of propositional logic, et.

4. An even stronger objection to the meaning-maximalist position can be found in the observation that the and in the sentences under (11) can also be omitted without involving a change of content.²⁷ Let us replace the and with a semicolon, as in the sentences under (12), or with some other punctuation mark:

(12) (a) Annie is in the kitchen; she is making doughnuts.

(b) Annie fell into a deep sleep; her facial color returned.

(d) Peter married Annie; she had a baby.

(e) Paul pounded on the stone; he shattered it.

(f) Give me your picture; I’ll give you mine.

(g) The number 5 is a prime number; it is divisible only by 1 and by itself.

We can communicate practically the same information with (12) as with (11) or (11’). This may make us ask how the content elements explicated by the formulations of (11’) are conveyed in (12). Should we say that the semicolon itself has a meaning—or rather seven different meanings? Or should we say that the meaning is somewhere in the air and that it must be read “between the lines”? If one proceeds, like a meaning-maximalist, from seven different meanings and projects them, in (11), all on the word and, then, to be consistent, one would have to say the same about the semicolon (and about the articulatory pause between the utterances of the constituent sentences) in (12).

The only possibility left for someone who rejects this solution is to talk about contextual determination of content elements. But he must then be ready to answer the question as to whether such a solution would not be just as appropriate for the and-sentences under (11).²⁸

5. The conclusive argument against the meaning-maximalist analysis of the sentence connectives, however, is based on the fact that the list of sentences under (11) could be extended at will and thus give rise to a virtually infinite number of new meanings of and. Depending upon what the communication partners take to be the actual relationship between the facts described in the constituent sentences, one could speak about an adversative and, a consecutive and, a diagnostic and, etc., as in the sentences under (13):

In the case of ambiguous words we can usually count the number of individual meanings on the fingers of one hand. Words with three meanings are not unusual and one could even accept a word with twenty-seven meanings, but a word with an infinite number of meanings would be a contradiction in itself. Besides the practical difficulty that the lexicon cannot allow infinitely long entries, such a word would also raise theoretical difficulties: how could a language user learn to cope with a word of infinitely many meanings? The only solution here would be to assume a generative system of rules for the production of such an infinite number of meanings. Such a rule system cannot, by definition, be part of the lexicon, but would have to be assigned either to a prelexical linguistic component or to a postgrammatical component. However, after all that has been said, it is superfluous to assume that we are dealing with an infinity of word meanings.

8. SENTENCE CONNECTIVES:
MINIMIZATION OF MEANINGS

The collapse of the meaning-maximalist position now brings us back to the beginning of the last section. So let us restrict ourselves to the content elements common to all previously mentioned uses of the word and: to conjunctivity, which requires that two sentences connected by and are true if and only if the entire complex sentence is true; and to connexity, which requires the facts described by the constituent sentence to be part of the same situation. And let us attempt to reconstruct all the supposed additional content elements on the basis of conversational maxims.

A meaning-minimalist would most probably go one step further; he would ask if conjunctivity and connexity could not be explained away in the same manner as successivity. It is easy to see how this would work in the case of connexity. The examples under (4), whose absurdity had led us to postulate such a semantic feature for and, do not lose any of their characteristics when one omits the and:

(14) (a) 2 X 2 = 4, it is impossible to analyze further the concept of intention.

(b) Müller just scored a goal, eels spawn in the Sargasso Sea.

Even when formulated asyndetically, these sentences are odd, as long as the utterer cannot rely on additional information specific to the situation of utterance that would allow the addressee to establish a connection between the facts described. This shows that the construction of a relation between the facts described by coordinate sentences is not at all specific to the use of the word and. It must always be possible if the addressee does not want to assume that the utterer has violated a conversational maxim. The maxims concerned here are particularly those of manner (IV, especially IV/1 “Avoid obscurity of expression . . . ,” and IV/4 “Be orderly . . .”). Reasoning involving these maxims is available at any time; nothing prevents the addressee from also applying it in the event that the coordinate sentences are connected by and. Therefore it really is unnecessary to consider connexity to be a special semantic feature of the word and.

However, these considerations should not mislead us into trying to eliminate conjunctivity from the meaning of and. Of course, it is true that conjunctivity survives in many cases in which one omits the sentence connective and; that is especially so for coordinate sentences (cf. the examples under (12)). But what is interesting here is the cooccurrence of and with other sentence connectives in complex sentence structures. Let us consider the following conversations:

In B’s last utterance it is not possible simply to omit the word and or to replace it with a semicolon. The raison d’ëtre of this word lies in its combinatory function (not in its connecting function). In using his last utterance to elucidate the meaning of his comment, “That’s not so,” B makes it clear that he thinks one of the constituent sentences of A’s initial utterance is false, but that he does not want to specify which. It is the semantic feature of conjunctivity that enables B to do this, as can easily be seen from the following truth-table:

(The lower case letters “p” and “q” stand for the constituent propositions, “” stands for it’s not so that . . . , and the capital letters “T” and “F” stand for truth-values ‘true’ and ‘false’. For the operators It’s nice that . . . and It’s too bad that . . . we arrive at analogous results.) If one claims that the formula in the right hand box of the truth-table is true, one actually leaves three possibilities for the distribution of truth-values among the constituent propositions p and q (cf. the truth-values of p and q as noted in the corresponding lines of the left hand boxes). This example shows that in certain cases we cannot do without the word and if we want to communicate conjunctivity: when and is removed, conjunctivity is also lost. This, then, is a case where conjunctivity is detachable from the meaning of the rest of the sentence.

Conjunctivity also violates the other two criteria for the occurrence of a conversational suggestion: it is neither variable nor cancellable.²⁹ To assert:

(16) Peter married Annie and Annie had a baby; the complete sentence is true, but one of its constituent sentences is false.

is to contradict oneself.

Now that we have demonstrated that conjunctivity is a semantic feature of the literal meaning of and, we must show how it is possible to construe as conversational suggestions at least the seven other content elements discussed earlier.

re (11a): If someone explicitly states that Annie is in the kitchen and then adds without specifying another place that she is making doughnuts, then he is guilty of suppressing relevant information if he thereby wants to convey that the doughnuts are being made outside of the kitchen. This would be a violation of maxim 1/1. In order to avoid assuming such a violation, the addressee interprets the formulation of (11a) as a suggestion of identity of place (. . . and there . . .).

re (11b): If someone explicitly states that Annie fell into a deep sleep but then adds without specifying another time that her facial color returned, then he is guilty of suppressing relevant information if he thereby wants to convey that the two events took place at a completely different time. This would again be a violation of maxim 1/1. In order to avoid assuming such a violation, the addressee interprets the formulation of (11b) as a suggestion of simultaneity (. . . and during this time . . .).

re (11c): If someone explicitly states that a window is open and then adds without specifying another source that there is a draft, then he is guilty of communicating irrelevant information if he does not want to convey that the draft is coming from that window. This would be a violation of maxim III. In order to avoid assuming such a violation, the addressee interprets the formulation of (11c) as a suggestion of the source of the draft (. . . and coming from it . . .).

re (11d): If someone begins by reporting that a woman got married and then immediately adds without specifying another time that she had a baby, then he is guilty of distortive reporting if he thereby wants to convey that the wedding took place after the baby was born. This would be a violation of maxim IV/4. In order to avoid assuming such a violation, the addressee interprets the formulation of (11 d) as a suggestion of a temporal parallelism between the reporting utterances and the reported events (. . . and after that . . .).

re (11e): If someone explicitly reports someone’s action upon a certain object and then, without specifying another action, reports a result of that person’s acting upon that object, then he is guilty of communicating irrelevant information, if he does not want to convey that this result was brought about by the action mentioned. This would be a violation of maxim III. In order to avoid assuming such a violation, the addressee interprets the formulation of (11e) as a suggestion of an instrumental relation between the action and its result (. . . and thereby . . .).

re (11f) : If someone asks a favor and, in the same sentence, predicts an action of his own that can be considered as compensation for that favor, then he is guilty of communicating irrelevant information or of obscure procedures of negotiation if he does not want to make a conditional promise dependent on the accomplishment of the favor. This would be a violation of maxims III or IV, respectively. In order to avoid assuming such a violation, the addressee interprets the formulation of (11f) as a suggestion of a conditional relation between the two actions mentioned (if . . . , then . . .).

re (11g): If someone uses one and the same sentence to make two statements about a number each of which implies the other, he is guilty of prolixity if he does not use one statement to justify or explain the other. This would be a violation of maxim IV/3. In order to avoid assuming such a violation, the addressee interprets the formulation of (11g) as a suggestion of an explanatory relation between the two statements (. . . and therefore . . .).

These paradigms of the sources of conversational suggestions would not be complete without the following comments:

1. Conversational suggestions are dependent on the context of utterance. Any addition of a verbal utterance or of a detail of situation can direct the reasoning of the addressee in another direction. The seven paradigms should be read with this reservation.

2. Each reasoning process takes reference to the formulation of the sentence in question. Instead of doing this in an ad hoc way one could systematically compare the suggestion-producing qualities of sentences. On this basis, it should be possible to arrive at generalizations about the production of conversational suggestions and to approach explanatory adequacy. This can be an important methodological starting-point for progress in descriptive stylistics.

3. Conversational suggestions arise for the most part from specific qualities of the literal meanings conveyed. If a sentence manifests several such qualities at the same time, then several suggestions can arise. Thus

— (11a) is interpretable not only as local, but also as simultaneous;

— (11b) is interpretable not only as simultaneous, but also as explanatory and local;

— (11e) is interpretable not only as directional, but also as simultaneous and explanatory;

— (11d) is interpretable not only as successive, but also as explanatory;

— (11e) is interpretable not only as instrumental, but also as simultaneous, explanatory, and local;

— (11f) is interpretable not only as conditional, but also as successive.

The fact that we obtain multiple suggestions is a further confirmation of the meaning-minimalist approach, since it explains the vague and expressive character of the suggestive use of language. Which of the possible suggestions dominates in each case depends amongst other things on how the corresponding semantic dimensions are realized in the sentence in question. Thus, time plays a less significant role in (11a) than in (11b) through (11e), since (11a) is formulated in the present and not in the past tense. Place plays a more important role in (11a) than in (11b) or (11d), since (11a), in contrast to the other sentences, contains an explicit specification of place (in the kitchen). Abstract conceptual relationships play an exclusive role in (11g), since the facts described in both its constituent sentences are valid at any time and place.

4. From the structure and results of the reasoning processes sketched above we can conclude that the given suggestions are not to be added to the conjunctivity of and as semantic features of the same sort; rather, they are made possible only through the combinatory function of this word. Nor do those suggestions have the same status as connexity, since they embody rather special kinds of connection between the facts described, a connection whose existence is suggested by the contiguity of the utterances of the constituent sentences. Therefore, I propose to call them connexity-suggestions. With his “thesis of the dual control of linguistic structure” Charles W. Morris drew attention to this phenomenon as early as 1938, when he wrote: “From the inter-connectedness of events on the one hand, and the interconnectedness of [communicative] actions on the other, signs become interconnected [. . .].”³⁰

The paradigms of suggestion-producing reasoning have shown the usefulness of Grice’s maxims and have made it plausible that the other possible suggestions a speaker of English may intend in uttering and-sentences are also reconstructed pragmatically by the addressee. The truth-function defined in propositional logic has revealed itself to be the only semantic feature of the sentence connective and in English. And the hypothesis that other sentence connectives of natural languages also have a purely truth-functional meaning has gained in plausibility. A fresh start should be made in considering in detail whether it is possible to identify at least the literal meanings of the sentence connectives and, or, if, and not with the meanings of the connectives of propositional logic et, vel, si, and non, even if their use occasionally appears to be radically different.

At this point the meaning-minimalist position seems to have won the argument. However, this judgement may still be somewhat premature since here again things are not quite so easy. Therefore, I do not want to end this discussion without at least touching upon those difficulties.

9. THE DUALITY OF SEMANTIC
AND PRAGMATIC INTERPRETATION

Problems arise in the use of and in complex sentence structures. Let us consider the utterance of a conditional sentence that contains the word and in the first clause and suggests successivity:

(18) If Annie has married and has had a baby, grandfather will be happy.

Let us assume that (18) is true. The truth-functional analysis of the word if indicates that, if the antecedent clause is true, the consequent clause must also be true. The truth-functional analysis of the word and indicates that the entire sentence is true if and only if both constituent sentences are true. Under these conditions the grandfather would have to be happy if it is true that Annie has married and if it is true that she has had a baby. But sentence (18) is not normally so interpreted. Even if (18) is true, it can also happen that the grandfather will not be happy at all, if he hears that the child came before the wedding. So the truth of the consequent clause is dependent here upon the realization not only of the literal meaning of the antecedent clause, but also of its successivity-suggestion.³¹

Thus conversational suggestions arising from constituent sentences can be crucial in the evaluation of the truth of the entire sentence. In these cases one can no longer speak of a purely truth-functional use of the sentence connectives. The choice of explanations available creates a dilemma:

— whoever wants to save the truth-functionality of and by asserting that the subordinate clause of (18) is true because each of its constituent clauses is true, sacrifices the truth-functionality of if, since he must admit that the consequent clause can still be false.

— whoever wants to save the truth-functionality of if by asserting that the consequent clause in (18) is only false if the antecedent clause is false, sacrifices the truth-functionality of and, since he must admit that the antecedent clause can be false even if each of its constituent sentences is true.

In view of this dilemma we are on the verge of losing our motivation for a truth-functional treatment of the sentence connectives: it would only be of theoretical importance if it could be extended to all relevant sentence connectives.

A homogeneous treatment of the sentence connectives concerned seems possible only if we weaken the thesis that in natural language the truth-value of the entire sentence is a function of the truth-value of the constituent sentences. This thesis cannot be held in the sense that in complex sentence structures the truth-value of the entire sentence is directly deducible from the truth-values of the smallest constituent sentences. Rather, after each step in the truth-functional deduction, it must be considered whether the resulting conversational suggestions alter the derived truth-value. Each deduction in the value distribution of the complex sentence on the basis of the value distributions of two constituent sentences must be open to reinterpretation according to the context in which the sentence has been uttered.³²

This is certainly not a very elegant solution. It complicates the process of interpretation to such an extent that we might have doubts about the presuppositions of this analysis, in particular, the division of the content elements into word meaning and word use.

However, another solution is hard to find, considering the arguments against the meaning-maximalists given above. Moreover, there is a series of additional arguments that make this solution more plausible than any imaginable alternative:

1. Let us compare the following versions of sentence (18) with one another:

(18) (a) If Annie has married and has had a baby, grandfather will be happy.

(b) If Annie has married and she has had a baby, grandfather will be happy.

(d) If Annie has married and if Annie has had a baby, grandfather will be happy.

Only the subordinate clause differs each time. The number of syntactic transformations performed on the subordinate clause is greatest in (a) and is reduced progressively until (d).³³ What is significant is that in (18) the strength of the successivity-suggestion also varies. It is strongest in (a) and diminishes progressively down to (d). Obviously, the intensity of the communication of a connexity-suggestion depends on the degree of syntactic connectedness of the constituent sentences concerned. What we have here is a typical iconic relationship between content and syntactic form. The effect of this relationship is also noticeable, although to a lesser degree, where and is the only sentence connective involved:³⁴

(17) (a) Annie has married and has had a child.

(b) Annie has married and she has had a child.

(d) Annie has married. And Annie has had a child.

2. The difference between the sentences under (17) and the sentences under (18) can be generalized in the following way: the strength of a connexity-suggestion depends on the degree of embedding of the clause concerned in the entire sentence. Compare (19) with (18), (20) with (19), as well as (20), (19), and (18) with (17):

(19)If grandfather finds out that Annie has married and

(20)If grandmother finds out that Fritz has told grandfather that Annie has married

This generalization also proves valid when we reverse the sequence of sentences connected by and and formulate:

(17’) Annie has had a baby

(18’) If Annie has had a baby

(19’) If grandfather finds out that Annie has had a baby

(20’) If grandmother finds out that Fritz has told grandfather that Annie has had a baby

In all these sentences the strength of the successivity-suggestion diminishes according to the degree of embedding, and the content of the sentence connectives if and and comes progressively closer to their truth-functional meaning. In the most expanded version of the sentences under (20), it surely is irrelevant to the grandmother’s joy whether the events occurred in one sequence or another.³⁵

3. The force of embedded connexity-suggestions correlates with still other factors. Let me only mention as a last case the meaning of the higher verb:

that Annie has had a baby and has married, grandfather will be happy.

In a “report,” the sequence of the events conveyed is essential. Here, the person reporting tends to make his utterances follow the events. If he should happen to deviate from the natural sequence he would make sure that his addressees realize that. Thus, if Fritz reports something to the grandfather and the grandfather is happy about it, the sequence of the events must undoubtedly be considered among the reasons for his being happy. In a simple transmission of information, however, the sequence of the utterances can depend on any kind of accident and does not allow any conclusion about the sequence of events. Here, then, the sequence of events will not be among the reasons for the grandfather’s joy. In this way, the strength of a connexity-suggestion can be controlled by the choice of the higher verb.

These last three observations make it clear that connexity-suggestions are characterized not only by considerable variability, but also by change of intensity—a property never found in literal meanings. No grammar describes cases in which the syntactic qualities of the surrounding sentence exert an influence on the semantic value of a word such that one of its semantic features is either foregrounded or suggested with varying strength or even eliminated entirely.

All this indicates then that connexity-suggestions are not lexical phenomena but have to be accounted for by pragmatics. And since that is so, there seems no way to avoid the dual procedure previously discussed for the interpretation of complex sentences. It is not without irony that pragmatic rules should play an essential role in the interpretation of expressions the analysis of which has for decades been considered the core of semantics.³⁶

After this general conclusion it may be appropriate to summarize the individual results to which we were led by our discussion of the sentence connectives:

1. The delimitation of semantics and pragmatics in language description must follow the difference between meaning and use of words in verbal communication.

2. The meaning and the use of a word are not just two sides of the same thing, but have to be distinguished systematically. Speakers of a natural language master not only fixed word-meanings, but also fixed rules for the use of words. Both are empirically testable, even if there is no direct experimental access to them.

3. The criteria of variability, can ce liability and non-detachability can help to answer the question of which content elements of a given utterance come into play through the literal meaning and which through the use of words in verbal communication.

4. On the basis of these criteria, the meaning of sentence connectives such as and, or, if, and not in natural language may be equated with the defining properties of the connectives of propositional logic et, vel, si, and non. Corresponding to the special purpose and circumstances of communication further content elements can be acquired by a sentence connective on the basis of the formulations, the meanings and the facts described in the connected sentences. These content elements occur as conversational suggestions, more specifically, as connexity-suggestions.

5. The differentiation of semantics and pragmatics in language description, and the differentiation of meaning and use in verbal communication are theoretical distinctions; it would be false to assume that in the actual process of comprehension one begins by applying all and only the semantic rules and then continues with the pragmatic rules. Examples with a sentence connective occurring in the scope of another sentence connective show that the meaning of a complex sentence depends not only upon the meanings of its parts, but also upon their conversational suggestions, and thus, upon their use.

6. To summarize this summary:

a. The use of a verbal expression is partially determined by the meaning of this expression.

b. The meaning of a complex verbal expression is determined not only by the meanings of its constituents, but also by their specific use.

In short, in verbal communication we not only make use of meanings but this use even makes sense.

__________

NOTES

For helpful comments on earlier versions of this paper I am grateful to Jerry Edmondson, Donald Freeman, Frans Plank, and David Schwarz. I also want to express my gratitude for stimulating discussions of the material involved to students of linguistics, semiotics, and philosophy of the universities of Hamburg, Montreal, Chicago, Los Angeles, Berkeley, and Stanford. The usual disclaimers apply, of course.

1. The two façons de parler can be found in any historic-systematic presentation of linguistics, cf. Lyons (1968) and Ebneter (1973). A book exclusively applying the terminology of meaning is Schmidt (1967); presentations applying the terminology of use are Leisi (1953) and Brown (1974).

2. Cf. Grice (1968) and Grice (1975).

3. Cf. Wundt (1900, p. 596).

4. Cf. Titchener (1912, pp. 367 ff.). See also Hörmann (1970, pp. 166 ff.).

5. Cf. Erdmann (1900; 4th edition 1925, p. 5).

6. Cf. Watson (1919; 2nd edition 1924, p. 354).

7. Cf. Skinner (1937). See also Hörmann (1970, p. 165).

8. Cf. Morris (1938, pp. 43-48).

9. Cf. Kutschera (1971, 2nd edition 1975, pp. 87 f.).

10. Cf. Alston (1963, p. 84).

11. Cf. Wittgenstein (1953, I, 43).

12. Cf. Leisi (1953; 4th edition 1971, pp. 73 f.).

13. Cf. Kempson (1975) and Wilson (1976).

14. Cf. Grice (1968).

15. For this example cf. Gordon and Lakoff (1971).

16. Cf. Dascal (1976, p. 23).

17. Cf. Grice (1968), 2nd lecture.

18. Cf. Wittgenstein (1953, I, 293).

19. Cf. Grice (1968), 2nd lecture.

20. Cf. Grice (1968), 2nd lecture.

21. For the methodological value of the three criteria cf. Walker (1975, pp. 169 ff.).

22. Cohen (1971), who makes a similar distinction, talks about “semanticists” on the one hand and “conversationalists” on the other. However, he does not distinguish between the literal meaning (of a word or a sentence) and the lexical meaning (of a word).

23. Cf. Cohen (1971, p. 56).

24. Cf. Cohen (1971, p. 57).

25. Cf. Wilson (1976, pp. 118 ff.). See also Abraham (1975).

26. Cf. Lang (1977, pp. 230 ff.).

27. However, sentences (12d) and (12g) are ambiguous for many speakers of English. The fact described in the second sentence in (12d) need not be taken to occur later than that described in the first sentence, it can also be taken to have been a reason for the first to occur. The fact described in the second sentence in (12g) need not be taken to be explained by the fact described in the first sentence, it can also be given as a reason for it.

28. Naess (1961) has conducted a series of tests which show that even the decision whether the sentence connective or must be interpreted as an exclusive or inclusive disjunction depends on the facts described by the disjuncts involved. See also Seuren (1977, pp. 371 ff.).

29. Of course, we are dealing here only with the and that occurs between sentences or their transformational variants, not with the phrasal and, as occurring in Peter and Annie went to Saarbrücken. The proposed treatment can easily be applied to all cases where and is used to connect the propositional content of two sentences, even if these sentences are uttered with non-declarative illocutionary force. The treatment of the and that connects speech acts of different illocutionary force must, however, be postponed to another occasion.

30. Cf. Morris (1938, pp. 12 f.).

31. Cf. Cohen (1971, pp. 58 f.). There are a number of other interpretations possible for (18). Even if the order of wedding and birth is as it should be, it is possible that the grandfather is not happy because Annie’s husband was not the baby’s father. On the other hand, it is conceivable that the truth of all constituent sentences still does not entail the truth of the entire sentence in its intended sense because the grandfather could be happy about something else that he has learned simultaneously. These interpretations are eliminated if we formulate (18) in a more explicit way: If Annie has married and has had a baby, this is a reason for grandfather to be happy.

32. Of course, the criteria of variability, cancellability, and non-detachability also apply to embedded sentences. E.g., a cancellation of an embedded conversational suggestion is achieved by the following context: If Annie has married and has had a baby, grandfather will be happy. But the sequence of these events will not leave him unaffected.

33. The first and second versions are generated by conjunction reduction, and the second and third by pronominalization; in the underlying structures of the first three versions if dominates and, in the last if is dominated by and.

34. Cf. Freeman (1978), as opposed to Boettcher and Sitta (1972), who consider sentences like those under (17) to be different forms of realization of the same (semantic) category and the same (pragmatic) structure. Such a characterization would make it impossible to account for the semantic and pragmatic differences of these sentences.

35. Of course, these analyses should not make us blind to the fact that the sequence of the formulations can also express other aspects than the sequence of the events described. Compare Annie now is a young mother. The fact that she has had a baby and that she has married has made grandfather very happy. Here the sequence of the embedded clauses is rather used to foreground Annie’s new role of a young mother.

36. Compare the procedures of formal logicians, who admit only a truth-functional relation between the complex sentence and its constituents and exclude all pragmatic aspects.

BIBLIOGRAPHY

Abraham, Werner. 1975, ‘Some Semantic Properties of Some Conjunctions,’ in S. P. Corder and E. Roulet (eds.), Some Implications of Linguistic Theory for Applied Linguistics, Brussels and Paris: Aimav and Didier, pp. 7-31.

Alston, William P. 1963, ‘The Quest for Meanings,’ in Mind 72 pp. 79-87.

Boettcher, Wolfgang and Sitta, Horst. 1972, Zusammengesetzter Satz und äquivalente Strukturen ( Deutsche Grammatik III). Frankurt a. M.: Athenäum.

Brown, Cecil H. 1974, Wittgensteinian Linguistics, The Hague and Paris: Mouton.

Chao, Yuen Ren. 1968, Language and Symbolic Systems, Cambridge: University Press.

Cohen, Jonathan L. 1971, ‘Some Remarks on Grice’s Views about the Logical Particles of Natural Language,’ in Y. Bar Hillel (ed.), Pragmatics of Natural Languages, Dordrecht: Reidel, pp. 50-68.

Creelman, Marjorie B. 1966, The Experimental Investigation of Meaning—A Review of Literature, New York: Springer.

Dascal, Marcelo. 1976, ‘Conversational Relevance,’ Working paper, Colloquium on ‘Meaning and Use’, Jerusalem.

Douglas, Mary (ed.). 1973, Rules and Meanings—The Anthropology of Everyday Knowledge, Harmondsworth/England: Penguin.

Ebneter, Theodor: 1973, Strukturalismus und Transformationalismus—Einführung in Schulen und Methoden, München: List.

Erdmann, Karl Otto: 1900, Die Bedeutung des Wortes—Aufsätze aus dem Grenzgebiet der Sprachpsychologie und Logik, Leipzig: Avenarius, 1900. Reprint of the 4th edition of 1925, Darmstadt: Wissenschaftliche Buchgesellschaft, 1966.

Freeman, Donald. 1973, ‘Keats’s “To Autumn”: Poetry as Pattern and Process,’ in Language and Style XI (1978).

Gazdar, Gerald. 1978, Formal Pragmatics for Natural Language, London and New York: Academic Press.

Gloy, Klaus. 1975, Sprachnormen I.—Linguistische und soziologische Analysen, Stuttgart and Bad Cannstatt: Frommann-Holzboog.

Gloy, Klaus and Presch, Gunter, (eds.). 1976, Sprachnormen III. Kommunikations-orientierte Linguistik—Sprachdidaktik, Stuttgart and Bad Cannstatt: Frommann-Holzboog.

Gordon, David and Lakoff, George. 1971, ‘Conversational Postulates,’ in Papers from the Seventh Regional Meeting of the Chicago Linguistic Society, Chicago: Chicago Linguistic Society.

Grice, Herbert P. 1968, ‘The Logic of Conversation,’ Working paper. Berkeley: University of California.

Grice, Herbert P. 1975, ‘Logic and Conversation,’ in P. Cole and J. L. Morgan (eds.), Syntax and Semantics. Vol. 3: Speech Acts, New York: Academic Press.

Grimm, Hannelore and Wintermantel, Margret. 1975, Zur Entwicklung von Bedeutungen—Forschungsberichte zur Sprachentwicklung II, Weinheim and Basel: Beltz.

Heringer, Hans Jürgen (ed.). 1974, Der Regelbegriff in der praktischen Semantik, Frankurt a. M.: Suhrkamp.

Hörmann, Hans. 1967, Psychologie der Sprache. Berlin, Heidelberg and New York: Springer. Translated into English by H. H. Stern: Psycholinguistics—An Introduction to Research and Theory. Berlin, Heidelberg and New York: Springer, 1970.

Isenberg, Horst. 1971, Überlegungen zur Texttheorie,’ in J. Ihwe (ed.), Literaturwissenschaft und Linguistik, Vol. 1, Frankfurt a. M.: Athenäum, 1971.

Kasher, Asa. 1974, ‘Mood Implicatures: A Logical Way of Doing Generative Pragmatics,’ in Theoretical Linguistics 1, pp. 6-38.

Kempson, Ruth M.: 1975, Presupposition and the Delimitation of Semantics, Cambridge: University Press.

Kutschera, Franz von: 1971, Sprachphilosophie, München: Fink; 2nd edition, 1975.

Kutschera; Franz von. 1975, ‘Conventions of Language and Intensional Semantics,’ in Theoretical Linguistics 2, pp. 255-283.

Lang, Ewald: 1977, Semantik der koordinativen Verknüpfung ( studia grammatica XIV), Berlin (GDR): Akademie-Verlag.

Lehrer, Adrienne and Lehrer, Keith, (eds.). 1970, Theory of Meaning, Engelwood Cliffs, N.J.: Prentice Hall.

Leisi, Ernst. 1953, Der Wortinhalt—Seine Struktur im Deutschen und Englischen, Heidelberg: Quelle and Meyer; 4th edition 1971.

Lewis, David. 1969, Convention—A Philosophical Study, Cambridge: Harvard University Press. Translated into German by R. Posner and D. Wenzel: Konventionen—Eine sprachphilosophische Abhandlung, Berlin and New York: de Gruyter, 1975.

Lorenzer, Alfred: 1970, Sprachzerstörung und Rekonstrukton—Vorarbeiten zu einer Metatheorie der Psychoanalyse, Frankfurt a. M.: Suhrkamp.

Lyons, John. 1968, Introduction to Theoretical Linguistics, Cambridge: University Press.

Mooij, Jan J. A. 1976, A Study of Metaphor—On the Nature of Metaphorical Expressions, with Special Reference to Their Reference, Amsterdam, New York and Oxford: North-Holland.

Morris, Charles W. 1938, Foundations of the Theory of Signs, Chicago: University Press. Translated into German by R. Posner and J. Rehbein: Grundlagen der Zeichentheorie, München: Hanser, 1972, 2nd edition, 1975.

Naess, Arne. 1961, ‘A Study of “or”,’ in Synthese 13, 49-60.

Posner, Roland. 1972 a, Theorie des Kommentierens—Eine Grundlagenstudie zur Semantik und Pragmatik, Frankfurt a. M.: Athenäum.

Posner, Roland. 1972 b, ‘Commenting—A Diagnostic Procedure for Semantico-Pragmatic Sentence Representations,’ in Poetics 5, pp. 67-88.

Posner, Roland. 1972c, ‘Zur systematischen Mehrdeutigkeit deutscher Lexeme,’ in Linguistik und Didaktik 12, pp. 268-276.

Presch, Gunter and Gloy, Klaus, (eds.). 1976, Sprachnormen II. Theoretische Begründungen—ausserschulische Sprachnormenpraxis, Stuttgart and Bad Cannstatt: Frommann-Holzboog.

Raz, Joseph. 1975, Practical Reason and Nonns, London: Hutchinson,

Richards, David A. J. 1971, A Theory of Reasons for Action, Oxford: Clarendon.

Rollin, Bernard E. 1976, Natural and Conventional Meaning—An Examination of the Distinction, The Hague and Paris: Mouton.

Sadock, Jerrold M. 1974, Toward a Linguistic Theory of Speech Acts, New York: Seminar Press.

Schmerling, Susan F. 1975, ‘Asymmetric Conjunction and Rules of Conversation,’ in P. Cole and J. L. Morgan (eds.), Syntax and Semantics. Vol. 3: Speech Acts, New York: Academic Press.

Schmidt, Wilhelm. 1967, Lexikalische und aktuelle Bedeutung—Ein Beitrag zur Theorie der Wortbedeutung, Berlin (GDR): Akademie-Verlag.

Searle, John R. 1969, Speech Acts—An Essay in the Philosophy of Language, Cambridge: University Press.

Seuren, Pieter A. M. 1977, Zwischen Sprache und Denken—Ein Beitrag zur empirischen Begründung der Semantik. Wiesbaden: Athenaion.

Skinner, Burrhus F. 1937, ‘The Distribution of Associated Words,’ in Psychol. Ree. 1, pp. 71-76.

Slobin, Dan I. 1971, Psycholinguistics, Glenview, III. and London: Scott, Foresman and Comp.

Titchener, Edward B. 1912, Lehrbuch der Psychologie, Part 2, Leipzig: Barth.

Travis, Charles. 1975, Saying and Understanding—A Generative Theory of lllocutions, Oxford: Basil Blackwell.

Ulmann, Gisela. 1975, Sprache und Wahrnehmung—Verfestigen und Aufbrechen von Anschauungen durch Wörter, Frankfurt a. M. and New York: Campus Verlag.

Walker, Ralph C. S. 1975, ‘Conversational Implicatures,’ in Simon Blackburn (ed.), Meaning, Reference and Necessity, Cambridge: University Press.

Watson, John. 1919, Psychology from the Standpoint of a Behaviorist, Philadelphia and London: Lippincott; 2nd edition, 1924.

Wilson, Deirdre. 1976, Presuppositions and Non-truth-conditional Semantics, London, New York and San Francisco: Academic Press.

Wittgenstein, Ludwig. 1953, Philosophical Investigations, ed. by G. E. M. Anscombe and R. Rhees, Oxford: Basil Blackwell,

The Bridge between
Semantics and Pragmatics

WILLIAM P. ALSTON

According to the hallowed trinity of syntax, semantics, and pragmatics, syntax is concerned with the internal structure of a system of signs; semantics with the relations between signs and their designata, referents, or meanings; and pragmatics with the use of signs as vehicles of communication. The mutual dependence or independence of these areas of inquiry has long been a thorny issue, the most visible form of which is the debate over the possibility of developing the syntax of a language without recourse to semantic considerations. In this paper I shall exhibit one mode of dependence—that of semantics on pragmatics. My concern is with conceptual rather than methodological or any other form of dependence. My explicit concern is restricted to language and to the meaning aspects of semantics, rather than to reference, truth, or anything else one may be inclined to put into that basket. My claim is that the concept of linguistic meaning is to be elucidated by reference to pragmatic concepts of the roles of meaningful units in communication—what speakers do with those units in communication. I am not suggesting that specific facts, principles, or phenomena of pragmatics must be used to work out the semantics of a language. That is not the kind of bridge I want to build. Indeed, I do not attempt here to draw any methodological consequences from my thesis. I am concerned instead with gaining a reflective understanding of the concepts we use in semantically describing a language, however such a description may be most fruitfully developed.

Most of those currently concerned with the semantic side of language—linguists, philosophers, and others—concentrate on applying semantic concepts to language or to particular stretches of language-definite descriptions, adverbs, tenses, token reflexives, mass terms, natural-kind terms, and so on. There is relatively little interest in making explicit what it is for a word, phrase, or sentence to have a certain meaning. When conceptual problems are addressed, it is usually by way of defining certain semantic concepts in terms of an initial set of semantic primitives. Only occasionally does anyone seek to throw light on our most basic semantic concepts or on the system as a whole.¹

As already advertised, I am attempting to do this job by exhibiting the conceptual dependence of semantics on pragmatics. This dependence can be stated in its most general form as what I call the “Use Principle”:

I. To have a certain meaning is to be fitted to play (to be used for) a certain distinctive role in communication.

I must say that the Use Principle has seemed overwhelmingly plausible, even truistic, to me from the day I first formulated it explicitly. It seems to me constitutive of our concept of linguistic meaning that the meaning of a linguistic expression is what fits it to play a distinctive role in communication. Surely it is obvious that by meaning what it does, the sentence “What time is it?” is standardly usable for one sort of communicative act rather than another, in this case, for asking someone for the time. How could we change its meaning without changing what it is standardly usable to communicate? And how could we change its standard communicative potential without changing its meaning? No doubt, the identification of meaning with communicative potential will be acceptable only if we carefully restrict the kinds of communicative roles we put into the equation. But considerations of the sort just indicated encourage us to suppose that there is some way in which this can be done.

First we must determine what kind of communicative role can constitute the meaning of what kind of linguistic unit. The best strategy would seem to be to start with the linguistic unit that is usable to perform a complete act of communication, namely, the sentence. Having identified the communicative potential that constitutes sentence meaning (SM), we can then construe word or morpheme meaning more abstractly, as the capacity of the word or morpheme to make a certain distinctive contribution to the communicative potentials of sentences in which it occurs. Thus, in the implementation of this program, the concept of sentence meaning is the first semantic concept to be elucidated, and the concept of word meaning is elucidated in terms of that.² In this paper I shall restrict myself to sentence meaning.

My basic thesis identifies sentence meaning with illocutionary-act-potential (IAP). The “IAP Thesis” can be stated as follows:

II. A sentence’s having a given meaning consists in its having a certain illocutionary-act-potential.

A few preliminary points should be made about the concepts involved in this principle. For a preanalytic demarcation of the category of illocutionary acts (IA), I rely on the familiar indirect discourse form. We have, in ordinary language, a variety of locutions for making explicit what someone said (where this does not mean what sentence he uttered), that is, what the content of his utterance is, what message it conveyed. Here is a small sample:

1. U asserted (admitted, replied, insisted, etc.) that his garden gate was open.

2. U promised to meet Jones for lunch tomorrow.

3. U asked A for a match.

4. U predicted that the strike would soon be over.

5. U remarked that the weather was warming up.

6. U assured A that everything would be all right.

7. U congratulated A on his performance.

8. U exhorted A to finish by tomorrow.

9. U expressed considerable enthusiasm for A’s proposal.

10. U expressed his intention to stay here all summer.

11. U declared the meeting adjourned.

12. U called the batter out.

Note that each of these items involves an action verb (promise, predict, advise, etc.) followed by what we might call a content-specifying phrase. We may follow John Searle³ in saying that the verb specifies the illocutionary force (IF) of the utterance, while the ensuing phrase specifies the propositional content (PC).

Although most of the examples yielded by my criterion would be recognized by other theorists, there are some differences. First, I use a third-person approach based on the ways in which we report utterances; on this approach, an illocutionary verb is not necessarily usable in the performative formula. Express, for example, is not. I do not express admiration for your performance by saying, “I express admiration for your performance.” Second, I do not share the tendency exhibited by some others to individuate IA-types solely by what I have called IF. In my scheme, predicting that it will rain tomorrow is just as much a different IA-type from predicting that Jones will come to the party as it is a different IA-type from asking Smith what time it is; albeit the first difference is only in PC, while the second is also in IF.

Now a word about the P in IAP. It is clear that if statement II above is to have any chance of being acceptable, we must restrict potential to something like standard potential, that is, usability by virtue of the constitution of the language. For if we consider private codes or special conventions for special subgroups, it is clear that any sentence (S) can be used to perform IA’s of any type. (But by the same token, given the proper arrangements, a given sentence can be used to mean anything.) Since we are interested in established meanings of S’s, we are correspondingly focused on what makes an S standardly usable to perform IA’s of a certain type.

Next I must note some counterexamples to the unqualified identification of SM and IAP.⁴ To deal with them we shall have to complicate the thesis.

SM can be unqualifiedly identified with IAP only if for each distinguishable SM there is exactly one IA-type for the performance of which the S is thereby fitted; and, vice versa, for each IA-type there is exactly one SM that fits an S for being used to perform it. If this were true, then whenever I used an S with a given meaning and thereby performed some IA, it would always be the same IA; and whenever I performed an IA of a certain type, by making a normal use of an S, I would always be uttering that S with the same meaning. But the world is not that simple.

There are, first of all, cases of the sort used by Austin to introduce the concept of IF originally, those in which a sentence can be used with the same meaning to perform acts with different IF’s.⁵ As Austin pointed out, one can use the sentence “It is going to charge” without changing its meaning either to state or to warn that a certain bull is going to charge (and, we may add, to admit, to agree, to conclude, or to announce that it is going to charge). Clearly the meaning of one’s sentential vehicle cannot be depended on to determine fully the IF of one’s IA. Austin took this to show that the IF of an utterance was something over and above meaning “in the sense in which meaning is equivalent to sense and reference.” But it is important to note that insofar as it shows this, a parallel argument also shows that PC is not determined by meaning. For just as my sentential vehicle may not make completely explicit the IF of my utterance, so may it not make explicit the PC. By saying “It will,” I may be asserting that a certain bull is going to charge, that the interest rate will go down, that a certain ladder will hold steady, and so on. And, given the appropriate stage setting, I can express any propositional content by saying “Yes,” surely a limiting case of the underdetermination of IA by sentential meaning.

Involved in all these cases is the familiar fact that if we have enough contextual clues, we need not use a sentence rich enough to carry all the details of the intended IA. This point applies equally to the IF and PC aspects; but for any IA-type it is possible (with a possible exception noted below) to construct a sentence the (or a) meaning of which does fully and unambiguously determine that type. Thus the Austinian IF ambiguities could be resolved by augmenting the sentence with a performative verb to make it “I warn you it’s going to charge” or “I admit it’s going to charge.” My first PC ambiguity can be resolved by replacing “It will” with “The bull over there is going to charge” or “The interest rate will go down.” The only aspect of IA’s that seems not fully determinable by meaning is reference (meaning by “meaning” meaning rather than Austin’s “sense and reference”). Note that my enlarged sentences did not make explicit which bull or which interest rate. But this is controversial, and in any event our present concern is with determination in the opposite direction. For any SM there is an IA-type that is fully determined by that meaning, in the sense that if someone seriously and literally utters a sentence with that meaning, then, just by knowing that, we can determine that the person intends to be performing an IA of that type. We do not need any supplementary contextual clues for that determination. This criterion gives us what we may call the matching IA-type for a given SM. Here are the matching IA-types for some of the above sentences and some others (assuming a familiar meaning for each sentence).

1. “It will”—asserting of something that it will do something.

2. “It’s going to charge”—asserting of a certain animal that it is going to charge.

3. “I agree that it is going to charge”—agreeing to the proposition that it is going to charge.

4. “The bull is going to charge”—asserting of a certain bull that it is going to charge.

5. “Jones’s prize bull is going to charge”—asserting of a certain prize bull belonging to a person named “Jones” that it is going to charge.

To construct a matching IA-type we have to find just the degree of specificity that is explicitly embodied in the SM in question—with respect to the identity of referents, the predicative aspects of the PC, and the IF. Note that an IA-type can be too poor for the sentence meaning, as well as, like all our initial examples, too rich. Thus while admitting that Jones’s bull is going to charge is too rich for “It is going to charge,” asserting of something that it will do something is too poor. Again, there are dimensions of mismatch other than specificity. The sentence “The Norwegian banners fan our people cold” can be used metaphorically to assert that the Norwegian Army is making our people afraid; but that IAP is not directly determined by any of its established meanings in the language. And second-order uses, such as ironically using “What a beautiful day” to remark that the weather is nasty, are not directly determined by meaning.

In sum, although it is not unqualifiedly true that SM amounts to IAP, we can say that for any SM there is a matching IA-type such that the SM can be identified with whatever renders a sentence usable to perform IA’s of that type.

It may help you to appreciate the distinctive character of IAP theory if I contrast it with another form of “communicative role” theory, one which identifies sentence meaning with the usability of a sentence to produce or elicit certain effects in a hearer, for example, the production of a certain belief, intention, or emotion, or the elicitation of a certain action. Using Austin’s correlative concept of a perlocutionary act,⁶ we might term this the “Perlocutionary-Act-Potential (PAP) Theory.” The PAP Theory is found in many thinkers, from Locke to Grice. Suppose, contrary to what I believe, that we could find a matching PAP for each SM, just as I have claimed we can find a matching IAP. Then each SM would be extensionally equivalent to a certain PAP. Even so I would not consider PAP theory a serious rival to IAP theory, for I would hold that an S has a distinctive PAP only by virtue of having a correlated IAP. A given SM determines a distinctive PAP through determining a distinctive IAP. Consider what makes the sentence “The house is on fire” an effective vehicle for getting the addressee (A) to realize that a certain house is on fire, or to realize that the utterer (U) believes of a certain house that it is on fire. Why should we suppose that uttering that sentence is a good way of getting one’s addressee to realize that? It is because the rules of the language and the conventions of communication are such that when U utters that sentence, A (in the absence of indications to the contrary) takes U to be performing the IA of asserting of the house in question that it is on fire. Unless A has reason to doubt U’s sincerity or reliability, A will draw the conclusion that the house is on fire, or at least that U believes it to be. But this inference would never get started if A did not suppose U to be performing the IA of asserting that the house is on fire; and, in the normal case, A has no basis for this supposition other than the fact that the sentence uttered is tagged in the language for performing IA’s of that type. Thus, even if each SM does determine a correlated PAP, that relationship is derived from the more basic determination of a correlated IAP. Even if the PAP Thesis passes the test of extensional equivalence, the equivalence will be correctly viewed as a consequence of the IAP Theory, rather than as the basis for a rival account of what meaning is.

It may be thought that the IAP Theory purchases complete coverage at the price of remaining so close to the analysandum as to be unilluminating. According to IAP theory, the sentence “The house is on fire” means what it does because it is standardly usable to tell someone that a certain house is on fire. And isn’t the concept of telling someone that a certain house is on fire too close to the concept of the sentence’s meaning that a certain house is on fire to be the basis of an illuminating account of the latter? I would agree that if we leave the concept of IAP in an unanalyzed state, we have not thrown sufficient light on the matter. To make further progress, we must develop an analysis of IAP. And to do that in an illuminating way, we must develop an account of the structure of IA’s, of what it is to perform an IA of a certain type. (An understanding of potentiality requires an understanding of the corresponding actuality, as Aristotle taught us.) This is a very large subject, and here I can do no more than present the bare bones of my account.

Let us begin with flat assertions, which, in the light of the completed account, turn out to be a case of limiting simplicity. What is it to assert that my garden gate is open? What is added (or subtracted) when I move from using a certain sentence, for example, “It’s open,” to practice pronunciation or to assert that my front door is open, to using it to assert that my garden gate is open? When I began to make a determined assault on this problem, I was surprised to find how little attention it has attracted from scholars, so the following account does not confront much overt competition.

First we must specify the requirements that an adequate answer must satisfy. The crucial condition that makes the utterance an assertion that p must contain the following elements:

1. It must involve the proposition that p. Otherwise, how could it turn the utterance into an assertion that p rather than an assertion that q?

2. It must connect the proposition that p with the utterer, U. Otherwise, how could it turn U’s utterance into an assertion that p? The most obvious way to satisfy these first two requirements would be to make the condition require a certain stance or propositional attitude of U toward p.

3. This attitude must be within U’s voluntary control. For it seems that, given the appropriate equipment, I can, at will, use the sentence “It’s open” in any of the three ways enumerated above, as well as in many others.

We can satisfy these constraints by the condition that U utter a certain sentence, S, in order to get the addressee, A, to believe that p, or, in a more complex version, to get A to believe that U believes that p. This kind of suggestion is worked out in a very complicated form in Stephen Schiffer’s book, Meaning⁷; but however complicated, it seems clear to me that it does not provide complete coverage. For example, I may have absolutely no hope of your believing what I say, or even of your believing that I believe what I say, but nevertheless I may feel an obligation to tell the truth in answer to a direct question. Hence, in the interest of finding an account that applies straightforwardly over the whole range, I am motivated to look elsewhere.

Consider the following natural ways of bringing out what U did in making a flat assertion that p.

1. He represented himself as knowing that p.

2. He purported (claimed) to know that p.

3. He committed himself to its being the case that p.

4. He vouched for its being the case that p.

5. He lent his authority to the belief that p.

The last three items indicate that when one makes an assertion one changes one’s normative status in a certain way; that one sticks one’s neck out or goes out on a limb. When I vouch for something or lend my authority to its being a certain way, I render myself liable to censure, reprimand, correction, or the like, in case things are not as I have claimed they are. I have put myself in such a position that I can be called to account provided things are not that way. (Of course, whether I am actually “called to account” in that case, and if so how, depends on various features of the social situation.) The first two items indicate that the condition that must hold if one is to escape liability to negative sanction is not simply that p (where p is what is being asserted), but that U knows that p. We can conclude that what makes an utterance of S into an assertion that p is that in uttering S, U takes on a liability to negative sanction (if otherwise appropriate), provided he does not know that p.

Let us add one more bit to the picture. If we may assume that what renders me liable to negative sanction in case that ~p is the existence of a rule that requires that p, we may reformulate the crucial condition in terms of the holding of a rule that requires U to know that p. Remembering that the crucial condition is supposed to be under the voluntary control of U, we are then forced to recognize that the crucial condition cannot be merely that the rule does hold, that it does apply to U’s utterance, but that U recognizes it to hold, utters S as subject to that rule.⁸ Putting this all together, we may formulate our crucial condition (for asserting that p in uttering S) as follows:

III. U uttered S as subject to a rule that, in application to this case, implies that U’s knowing that p is a necessary condition of U’s permissibly uttering S.

So far we have explicitly considered only one small corner of the IA terrain, namely, asserting; however, I believe that we can extract from our account of asserting a general account of IA’s. The crucial step in this extraction is the recognition that in performing an IA of whatever type we are presupposing or implying that certain states of affairs obtain. Here are some examples of what is presupposed or implied in nonassertive IA’s of several types.

1. U ordered A to open that door.

a. That door is not now open.

b. It is possible for A to open that door.

c. U is in a position of authority vis-à-vis A.

2. U promised A to meet him for lunch tomorrow.

a. It is possible for U to meet A for lunch tomorrow.

b. A would prefer U’s meeting him for lunch tomorrow to U’s not doing so.

c. U intends to meet A for lunch tomorrow.

3. U declared a certain meeting (M) adjourned.

a. M is in session at the time of the utterance.

b. U has the authority to terminate M.

c. Conditions are appropriate for the exercise of that authority.

It seems clear that the account we have given of asserting that p will apply equally well to presupposing or implying that p. For if, in ordering A to open a certain door, I am implying that the door in question is not already open, it would seem that I am purporting to know that the door is not already open, just as much as I would if I were asserting that the door is not already open. If I order you to open the door without knowing that the door is not open, my utterance is out of order as unequivocally as if I had asserted that the door is not open without knowing that it was not open. This means that insofar as we can construe these nonassertive illocutionary acts as nothing other than presupposing and implying certain things in uttering a certain sentence, we can analyze them along the same lines as assertions. To be sure, there are various difficulties which must be overcome before we can pursue this analysis to its conclusion. First, we will have to explain the difference between asserting that p and presupposing or implying that p. Second, we will have to take account of the point that associated with each of a certain range of IA-types (e.g., declaring a meeting adjourned) there is a distinctive conventional effect (e.g., the meeting is no longer in session). At this point I can only consider the second problem, and can only indicate briefly how I might deal with it.

On the one hand, it seems clear that it is a conceptual truth that declaring a meeting adjourned has as its natural outcome the conventional effect just mentioned. On the other hand, it is clear that so long as an IA concept involves only what is said (the content of the message communicated), we cannot take the actual production of such a conventional effect to be a necessary condition for the actual performance of an IA of that type. For if we contrast a case in which the speaker actually terminates the meeting and a case in which the speaker does not although the speaker had intended to do so (some essential procedures were overlooked), it is clear that what the speaker said is (or could be) exactly the same in the two cases. Hence what we may call the pure IA is not “adjourning the meeting,” but rather “declaring the meeting adjourned,” where the latter description leaves it open whether the meeting actually is terminated.

But if this is the case, how are we to accommodate the first of the two points mentioned above—that somehow the notion of the meeting being terminated is involved in the pure IA concept of declaring the meeting adjourned? I suggest that we do it by adding to the list of conditions presupposed or implied the following:

By uttering S, U brings it about that the meeting in question is terminated.

Thus we lay down as a necessary condition of U’s declaring the meeting terminated that U imply that by this utterance U is terminating the meeting, but we do not require that U actually do this. In this way we accommodate both the points made at the beginning of the previous paragraph.

One virtue of the account of IA performance in terms of rule recognition that is embodied in statement III above is that it presents us with a ready-made answer to the questions “What gives a sentence a certain IAP?” “By virtue of what is a sentence rendered standardly usable to perform IA’s of a given type?” If performing an IA of a certain type consists of uttering a sentence as subject to a certain kind of rule, then the condition that would render a sentence standardly usable for performing an IA of that type is simply that it be governed by a rule of that sort. That is, the sentence will have a certain IAP if there exists in the language community a rule that stipulates the appropriate requirement for permissible utterance of that sentence. This is, of course, only the general outline of an answer. It merely indicates the lines along which detailed specifications of IAP’s are to be developed. Actually working them out will be an enormous task. It will require us to determine what kind of rule, applied to an utterance, would imply that the utterance was out of order in the absence of a given appropriate condition, taking account of such complications as multivocality and the underdetermination of reference by meaning. For the present, I shall have to be content with having indicated in outline how sentence meaning can be thought of as constituted by IAP, and thus how semantic concepts are conceptually derived from pragmatic concepts.

NOTES

1. For some noteworthy recent attempts see: H. P. Grice, “Meaning,” Philosophical Review 66 (1957):377-88; H. P. Grice, “Utterer’s Meaning and Intentions,” Ibid. 78 (1969): 147-77; H. P. Grice, “Utterer’s Meaning, Sentence-Meaning, and Word-Meaning,” Foundations of Language 4 (1968): 1-18; John Searle, Speech Acts (Cambridge: At the University Press, 1969); Stephen R. Schiffer, Meaning (Oxford: Clarendon Press, 1972); David Lewis, “General Semantics,” in Semantics of Natural Language, D. Davidson and G. Harman, eds., 169-218 (Boston: D. Reidel, 1972); Jonathan Bennett, Linguistic Behavior (Cambridge: At the University Press, 1976).

2. One might think that the claim that sentence meaning is prior to word meaning conflicts with the universally accepted, and undeniable, point that the meaning of a sentence is a determinate function of the meanings of its constituents plus its syntactic structure. This latter principle takes sentence meaning to be derivative from, among other things, word meaning. So how can word meaning be derivative from sentence meaning? It seems we can’t have it both ways.

But it can work both ways, provided the orders of priority are different, or, more precisely, provided the terms of the opposite priority relations are not the same; and that is the case here. The fact that a given sentence has a certain meaning is derivative from facts about the meanings of its constituent words; whereas the general concept of word meaning is derivative from the general concept of sentence meaning. It is in the order of derivation (or explanation) of particular semantic facts that word meaning is prior, while it is in the order of derivation or analysis of semantic concepts that sentence meaning is prior. It is not my position that the same kind of derivation can be carried out in both directions. I do not claim that the fact that a particular word has a certain meaning can be profitably construed as a function, among other things, of the meanings of sentences in which it occurs; nor do I claim that the concept of sentence meaning is derivative from the concept of word meaning.

It is quite common in many spheres of thought to find an opposition between the order of concepts and the order of facts or things, whether the latter be a causal, compositional, explanatory, or other kind of order. For example, a popular position in the philosophy of science holds that although micro facts and micro processes (atomic structure, electron movements) are responsible for observable macro phenomena (such as rusting and the movements of electric motors) and thus are prior to the macro phenomena in the order of causality and the order of causal explanation, nevertheless our concepts of the micro phenomena are necessarily derived, at least in part, from the ways in which they enter into the explanation of the macro phenomena and so are posterior in the order of conceptual analysis to concepts of macro phenomena. This is an exact parallel to my position on the relation of sentence meaning and word meaning. On my view sentence meaning (macro phenomenon) is explained by reference to its micro (semantic and syntactic) structure; but we cannot explicate our concepts of that micro structure except by reference to the ways in which aspects of the structure enter into the explanation of particular sentence meanings.

Here are some other examples of the same duality. According to Thomism, God is the cause of all other things including men (God is prior in the order of causality), but our concept of God is derived from our concepts of men and of other creatures (creatures are prior to God in the order of conceptual analysis). According to sense-datum theory, the concept of a physical object is derivative from concepts of sense-data, even though particular sense-data are caused by physical objects.

3. Speech Acts, 30-31.

4. I have presented the IAP Thesis in an unqualified and a somewhat different form in W. P. Alston, “Meaning and Use,” Philosophical Quarterly 13 (1963): 107-24, and in W. P. Alston, Philosophy of Language (Englewood Cliffs, N.J.: Prentice-Hall, 1964), chap. 2.

5. J. L. Austin, How to Do Things with Words (Oxford: Clarendon Press, 1962), lect. 8.

6. How to Do Things with Words, lect. 8. Roughly, to perform a perlocutionary act is to produce some effect “upon the feelings, thoughts or actions of the audience, or of the speaker, or of other persons” (101) by saying something.

7. Meaning, chap. 4.

8. We will not be able to consider here the complicated issue of the extent to which it is U’s recognition that constitutes the applicability of the rule to his utterance, and the question whether this is true of different sorts of cases in varying degrees.

The Signifying Animal