Semantics (from the Greek semantikos, "significant,") is the study of meaning in language. The interpretation of any given word or phrase may vary greatly; for example, the term meaning itself has a number of different senses. In the phrases What do you mean by that? or What do you mean to do?, the reference is to intentions. The expression Smoke means fire refers to an inference rather than a definition. Meaning can also refer to a translation from one language to another, as in the sentence Russian nyet means "no." It can even refer to a rule in a game: Doubles means that you must roll the dice again. But as a branch of \Tlinguistics\t, semantics is not concerned with all these different senses of meaning; rather, it is limited to the study of how speakers know that a given utterance represents a given thought or idea. Thus semantics is concerned with a type of translation: the translation from thought into utterance, and vice versa.

Four Components of Language.

Linguists divide the speaker's knowledge about language into four overlapping components: the lexicon (see \Tlexicology and lexicography\t), phonology (see \Tphonology and morphology\t), \Tsyntax\t, and semantics. These four components consist of processes that make it possible for speakers to set up correspondences between sound and meaning. The lexicon, or mental stock of words, embodies elements of each of the other three components. Every word is associated with a phonological representation, or sequence of sounds; syntactic information, such as its part of speech; and a meaning. The phonological component specifies how words are to be pronounced; the syntactic component specifies how words fit together to form sentences; and the semantic component specifies how the meanings of words combine within sentences to give a general meaning, or semantic reading, for the sentence.

Semantics in Other Disciplines.

Semantics is not just a branch of linguistic theory. Anthropologists can learn a great deal about a culture simply by examining the way its vocabulary divides up semantic space. Psychologists have studied many different aspects of meaning: the acquisition of word meanings in children, the ability of chimpanzees and other primates to understand sign language, the ability to detect ambiguity, and the emotive, or connotational, meaning of words (for example, statesman has more positive connotations than politician). Literary critics also need to know something about semantics, for they study the ways in which writers and poets use words to communicate thoughts and feelings. Finally, philosophers have at least as much claim to the field of semantics as do linguists. They have been most interested in issues such as the nature of correct and valid inference, the way in which people use language to refer, and the formal properties of logical systems. The work of philosophers has recently become a major stimulus in the development of linguistic theory. Linguists possess a general knowledge about the form of linguistic expressions that can occur in natural language; to this, philosophers have added useful theories about the reference of expressions in artificial languages and have examined minute semantic distinctions in their own languages.

Word Versus Sentence Meanings.

Linguists virtually ignored sentence meaning until the mid-1960s. When the linguistic usage of the term semantics was first introduced into English in the late 19th century, it referred to the classification of historical change in word meanings. Some traditional types of semantic change are narrowing (for example, Old English mete, "food," has become meat, "edible flesh"); widening (Middle English bridde, "young birdling," has become bird); degeneration (Old English cnafa, "boy, servant," has become knave); and elevation (Old English cniht, "boy, servant," has become knight). Even at the beginning of the 20th century, when linguists shifted their attention to the description of modern languages, they still limited their conception of semantics to word meanings.

It is obvious that the meaning of a sentence cannot be regarded as the sum of the meaning of its words. The difference in meaning between the sentences The cat chased the dog and The dog chased the cat shows that word order plays a crucial role in determining sentence meaning. The hierarchical structure of word groups, or constituent structure, plays an equally important role. A classic illustration is the ambiguous phrase old men and women. One meaning correlates with the syntactic grouping (old men) and women, where old modifies only men. The other meaning correlates with the grouping old (men and women), where old modifies the constituent men and women. Another illustration of such ambiguity is to be found in the sentence They are visiting firemen. The ambiguity stems from the fact that visiting can be grouped either with the preceding word--They (are visiting) firemen--or with the following word--They are (visiting firemen). Linguists call this type of ambiguity structural ambiguity, distinguishing it from lexical ambiguity, which is traceable to homonyms. An example of the latter type of ambiguity occurs in the sentence John looked at the bank, where bank may refer either to the ground beside a river or to a building where money is kept.

Generative-Transformational Grammar.

Although the recognition of structural ambiguity is crucial to the development of semantic theory, some ambiguities are not obviously structural or lexical. The sentence Visiting relatives can be boring, for instance, can mean either that relatives who visit can be boring, or that the act of visiting relatives can be boring. Both meanings, however, are associated with the same constituent structure. This type of ambiguity could not be handled in a satisfactory manner until Noam \TChomsky\t developed his theory of generative-transformational grammar in 1957.

Chomsky showed that each sentence has two significant levels of syntactic organization: a surface structure and a deep structure. Both levels are related to each other by a set of rules called transformations. Thus a given sentence--the surface structure--can derive from more than one deep structure. The ambiguity of visiting relatives can be captured in precisely this fashion. The phrase would be transformationally related either to a deep structure in which relatives is the object of visiting (for example, visiting one's relatives . . .) or to one in which relatives is the subject of visiting (for example, relatives who are visiting . . .).

Lexical Decomposition.

Sentence meanings are determined on the basis of the words they contain, so it is necessary for semantics to take word meanings into account. Linguists generally represent word meanings as a collection of semantic properties or features. Thus if one can say John's wife broke the lamp, then it is also possible to say A woman broke the lamp and A person broke the lamp. This is because words such as wife and woman share the semantic property "female," and wife, woman, and person share the semantic property "human." It is thought that concepts such as "female" and "human" are members of a large but finite collection of properties that serve as the building blocks for all word meanings. The practice of breaking word meanings down into their semantic components is called lexical decomposition.

Selectional Restrictions.

Many words, particularly verbs and adjectives, impose constraints on the meanings of other words in the sentence. Such constraints are called selectional restrictions. For example, the three words kill, murder, and assassinate overlap in meaning, and their objects must at least be animate. The object of murder, however, must also be a sentient being. It makes little sense to say John murdered the elephant unless the elephant is somehow personified, as in a children's story. The object of assassinate must be a person of political importance. Thus the sentence John assassinated the janitor also sounds funny, unless one thinks of the janitor as a political person, like the head of a secret terrorist organization. These words also say something about their subjects. Murder imposes the interpretation that its subject acted willfully and with malice. One can say John accidentally killed his brother to describe a situation in which John never intended to kill anybody; but the sentence John accidentally murdered his brother means that he murdered the wrong individual, not that he never intended to murder anybody. Similarly, the verb assassinate implies that the subject acted out of political motives. One would never say John assassinated the king if John committed the act because the king had insulted his wife.

Anomaly, Contradiction, and Analyticity.

Sentences that violate selectional restrictions produce semantic anomaly, or incoherence. Thus the sentence I stubbed my stomach makes little sense, because stub selects protruding objects such as toes and fingers. Harry broke the water is anomalous because break selects objects that are brittle.

Another type of sentence that exhibits semantic incompatibility between words is a contradiction. Contradictions are always false by virtue of the word meanings they contain. The sentence My wife is unmarried is a contradiction because it attributes the semantic feature "not married" to a word containing the feature "married." The sentence My wife is married is the opposite of a contradiction, because it can never be false. Sentences that are always true by virtue of the meanings of the words they contain are called analytic.

Presupposition.

One aspect of meaning that has given linguists and philosophers particular difficulty is presupposition. A presupposition is a proposition that must be true in order for a sentence to function properly as a statement, question, request, or other act of speech. Consider the classic joke in the form of a question: Have you stopped beating your wife? The addressee finds it difficult to respond without accepting the presupposition that he beat his wife. A famous example from the philosophical literature is the sentence The present king of France is bald. The presupposition here is that there is a king of France. Philosophers have debated whether such statements ought to be considered false or merely lacking a truth value.

Communicative Intent.

Another aspect of meaning that has proved difficult to represent in semantic theory is the communicative intent of the speaker. Every act of speech is performed for the purpose of achieving some goal in the mind of the speaker. That goal may be to inform, request, question, promise, bet, pass judgment, and so on. The philosopher J. L. Austin labeled this kind of intention on the speaker's part the "illocutionary force" of the utterance, and he discovered a class of verbs that actually perform the speech act that they describe. For example, one performs a bet by saying I bet you $5 that the Mets will win, a promise by saying I promise you that I'll clean this mess up, a request by saying I request that you leave your galoshes in the hall, and so on. Austin labeled this class of verbs "performatives," and he gave a few interesting properties that differentiate performatives from other verbs.

One such property is that performatives can occur with the word hereby, but other verbs cannot. One can say I hereby request that you stop smoking (a request), but not I hereby wish that you would stop smoking or I hereby understand that I am under arrest. Wishing and understanding are simply not speech acts.

It is possible, of course, to perform a given speech act without using a performative verb to do it. One can promise to be home on time by saying I promise to be home on time, or simply I'll be home on time. The latter type of speech act is said to have an implicit performative associated with it. It sometimes becomes difficult, however, to determine how a given utterance should be classified as a speech act. For example, a sentence like Can you carry this package? is, on the surface, a simple request for information. In reality, speakers use expressions like this as a way of requesting the listener to carry the package, and a response of Yes, unaccompanied by an action, would be considered inappropriate in many situations.

Pragmatics.

It seems reasonable to include the speaker's communicative intent as part of the meaning of a sentence, particularly because such intentions affect the linguistic form of expressions. Yet it is still necessary to differentiate a meaning from some inference based on that meaning. It is possible to produce sentences like My arm is sore or Somebody has to carry this package with the intention of getting the hearer to carry the package, but such sentences would not be called requests or orders. The study of nonlinguistic factors that affect the interpretation of expressions is called pragmatics, and in recent years the study of pragmatics has become an important part of semantics.

Meaning and Reference.

Just as it is possible to get confused over the difference between meaning and inference, it is also possible to fail to perceive the difference between meaning and reference. That this difference exists is demonstrated by the following example. The two expressions evening star and morning star refer to the same thing--the planet Venus--and yet no one would claim that the sentence Venus is Venus means the same thing as The morning star is the evening star. The former statement is analytically true and fails to be informative; the latter statement is not at all trivial. Moreover, some expressions have no reference at all--for example, the phrases female presidents of the United States and Albanian astronauts. If the meaning of an expression were its reference, then these two expressions would mean the same thing--nothing.

Examples such as the above led the German philosopher Gottlob Frege to distinguish between the "sense" of an expression and its "reference," the sense being what allows us to assign reference to an expression. The distinction is one of the most important principles underlying theories of logic and semantics, and it has enabled scholars to make steady progress in their investigations of the way the mind handles meaning. \TRichard\t WOJCIK

Bibliography: Allwood, J., Andersson, L.-G., and Dahl, O., Logic in Linguistics (1977); Austin, J. L., How to Do Things with Words (1962); Davidson, Donald, and Harman, Gilbert, eds., Semantics of Natural Language, 2d ed. (1973); Fodor, J. D., Semantics: Theories of Meaning in Generative Grammar (1977); Katz, Jerrold, Semantics Theory (1972); Leach, Geoffrey, Semantics (1974); Lyons, John, Semantics, 2 vols. (1977); Palmer, F. R., Semantics (1976); Rosenberg, J. F., and Travis, C., eds., Readings in the Philosophy of Language (1971).