The Aim and Structure of the Semiotic
Theory of Language
1. A Semiotic Definition of Language
Linguistics is the scientific study of language. The first thing to notice about this definition is that the word language is in the singular without the indefinite article, not in the plural. Linguistics takes an interest primarily not in individual languages but in common features of all languages. The notion of language is an abstraction that characterizes the common properties of all languages.
This abstraction of linguistics can be compared with the abstraction of biology. Zoology and botany are concerned with individual organisms. Biology, however, takes interest not in individual living organisms but in life, that is, in those biological functions, such as growth, metabolism, reproduction, respiration, and circulation, that are common properties of all living organisms.
The word language is applied not only to such languages as English, Russian, French, or Chinese—which are called natural languages because they evolved spontaneously—but also to a variety of other systems of communication. For example, chemists, mathematicians, logicians, and computer scientists have constructed for particular purposes notational systems that are also called languages: languages of chemistry, mathematical languages, languages of logic, programming languages for computers. These are artificial languages. The word language applies also to systems of animal communication: one speaks of the language of bees, the language of ants, etc.
Linguistics is concerned with the notion of language primarily as it is applied to natural languages. Therefore, the word language is understood in linguistics usually in the sense of natural language, and so I will use this word.
Let us now define the notion of language. What is it?
Language is a complicated phenomenon that can be studied from many points of view. Consider the production of speech sounds. It involves different configurations and movements of the speech organs, constituting a physiological phenomenon. Speech sounds have various acoustic properties—an acoustic phenomenon. The produced speech sounds must be detected and decoded by the human auditory system, which to this end has evolved special perceptual mechanisms for detecting and decoding speech cues—a physiological phenomenon. As an instrument of thought, language is a part of the human mind—expression of thought involves psychological and cognitive processes. Language is connected with human culture—this connection constitutes an anthropological phenomenon. Finally, as a system of signs used as an instrument of communication and an instrument of the expression of thought, language is a social phenomenon of a special sort, which can be called a semiotic phenomenon. (The word semiotic is derived from Greek sēmeion, ‘sign’). A system of signs as a semiotic phenomenon has a unique ontological status, because, on the one hand, it exists only in human consciousness, but on the other hand, man is forced to treat it as an object that exists independently of him. Sign systems belong to a special world, which can be called the world of sign systems, or the semiotic world. The essential property of this world is that genetically it is the product of human consciousness, but ontologically it is independent of human consciousness.
With respect to language, we are in the position of the six blind men and the elephant. The story goes that one of the blind men, who got hold of the elephant’s leg, said that the elephant was like a pillar; a second, who was holding the elephant by its tail, claimed that it was like a rope; another, who came up against the elephant’s side, asserted that it was like a wall; and the remaining three, who had the elephant by the ear, the trunk, and the tusk, insisted that it was like a sail, a hose, and a spear.
Like the elephant in that story, language can be approached from different sides. It can be studied from the point of view of biology, physics, psychology, logic, anthropology, philosophy, and, finally, semiotics (a general theory of sign systems).
While under these circumstances the study of language requires an interdisciplinary approach, there exists a hierarchy of different approaches to language. The decisive approach has to be the semiotic one, because the semiotic aspect of language constitutes the essence of language as an instrument of communication and an instrument of thought. While all other aspects are indispensable for the existence of language, they are subordinated to semiotics because they make sense only as manifestations of the semiotic nature of language.
A question arises: What properties characterize the semiotic aspect of language?
To answer this question is to give a semiotic definition of language.
I define language as a sign system characterized by the following properties: 1) two semiotic strata, 2) sequencing, 3) use of rules, 4) structure, 5) hierarchical stratification, and 6) semiotic relevance.
1) Two semiotic strata. Natural languages differ from other semiotic systems in that they have two semiotic strata. In natural languages the primitive semiotic system of signs is overlaid by a secondary semiotic system of diacritic linguistic elements. In distinction from natural languages, artificial languages of mathematics, physics, chemistry, and other abstract sciences make do with only systems of signs.
The necessity of a secondary semiotic system in natural languages is explained by the limited capacity of the human memory. Natural languages are so richly endowed with a large number of various signs that without diacritic elements it would be impossible to remember all of the signs. Memorizing and the use of signs in natural languages are possible only because any sign can be represented as a sequence of diacritic elements, whose number is strictly limited and compassable. As for artificial languages, since they do not possess a large number of signs, they do not need a system of diacritic elements. Of course, artificial languages with two semiotic strata are possible in principle, but such artificial languages must be considered analogues of natural languages.
Let us now introduce basic concepts characterizing the sign stratum and the diacritic stratum.
I suggest two primitive concepts for the sign stratum:
1. Sign of: X is a sign of Y.
2. Meaning of: Y is a meaning of X.
These concepts refer to relations rather than to objects. Speaking of signs, we mean a binary relation ‘sign of’; speaking of meanings, we mean a binary relation ‘meaning of’.
Sign of. X is a sign of Y if X means Y, that is, if X carries the information Y. For instance, the sound sequence bed carries the information ‘bed’, it means ‘bed’; therefore, bed is a sign of ‘bed’.
A sign is not necessarily a sound sequence. It may be the change of a stress (compare cónvict and convíct), an alternation (compare take and took), a change of a grammatical context (compare I love and my love), or a change of word order (compare John killed Mary and Mary killed John). There may be a zero sign; for example, if we compare quick, quicker, and quickest, we see that er is a sign of the comparative degree and est is a sign of the superlative degree, but the positive degree is expressed by the absence of any sound sequence with quick, that is, by a zero sign.
The opposition sign:meaning is relative. There may be an interchange between these entities. For example, letter p in the English alphabet normally denotes the sound p. But when we refer to the English letter p, we use the sound p as a name, that is, as a sign of this letter. Further, the meaning of a sign may serve as a sign of another meaning. Thus, lion is a sign of a large, strong, carnivorous animal. This meaning of the sign lion can be used as a sign of a person whose company is very much desired at social gatherings, for example, a famous author or musician.
It follows from the foregoing that the proposed concept of the sign is considerably broader than the common concept of the sign.
Meaning of. The proposed concept of meaning is much broader than the traditional understanding of the term meaning. My concept of meaning covers all kinds of information, including various grammatical relations. As was shown above, the notion of meaning is relative: a meaning may be a sign of another meaning.
We must strictly distinguish between the notions ‘sign’ and ‘sign of’, and between the notions ‘meaning’ and ‘meaning of’. The notion ‘sign of’ is a binary relation between a vocal product or other elements called a sign and a concept called the meaning of the sign. A sign and its meaning are members of a binary relation ‘sign of’. For example, neither the sound sequence [teɪbl] nor the concept ‘table’ is a linguistic fact: the first one is a physical fact, and the second one is a fact of thought; only as the members of the relation ‘sign of’ are they linguistic facts: a sign and its meaning. If we denote the relation ‘sign of’ by the symbol Ʃ, the sign by the symbol s, and the meaning by the symbol m, we get the formula
(1) sƩm
This formula reads: the element s is the sign of the concept m.
In accordance with the conventional terminology for members of binary relations, we will call s the predecessor with respect to the relation Ʃ, and m the successor with respect to the relation Ʃ.
The relation ‘meaning of’ is the converse of the relation ‘sign of’. (The notion of the converse of a binary relation R is defined in logic as follows: given a binary relation R, the relation R̆, called the converse of R, holds between x and y if, and only if, R holds between y and x, and vice versa; for example, the relation longer is the converse of the relation shorter, and, vice versa, the relation shorter is the converse of the relation longer.) Introducing Ʃ̆ as the name of the relation ‘meaning of’, we get the formula
(2) mƩ̆s
where m is the predecessor and s is the successor with respect to Ʃ̆. This formula (2) reads: the concept m is the meaning of the sign s.
Linguistic signs have various degrees of complexity. Language does not offer itself as a set of predelimited linguistic signs that can be observed directly. Rather, the delimitation of linguistic signs is a fundamental problem that presents great difficulties and can be accounted for only at a later stage of our study. In the meantime, I will work with words as specimens of linguistic signs. Although the theoretical definition of the notion of the word is maybe even more difficult than the definition of other linguistic signs, the word as a linguistic sign has an advantage of being familiar as an item of our everyday vocabulary. The term word does not need introduction; rather, it needs explication. For the present purpose, it is important that the basic semiotic properties that characterize words are valid for linguistic units in general.
Let us now turn to the second semiotic stratum of language, which can be called the diacritic stratum. In order to characterize this semiotic stratum, I will introduce the binary relation ‘diacritic of’. Just as we distinguish between the notions of the sign and of the binary relation ‘sign of’, so we must distinguish between the notions of the diacritic and of the binary relation ‘diacritic of’.
The first term of the relation ‘diacritic of’ is sounds, and the second term is a sign. If we denote ‘diacritic of’ by the symbol D, the sounds by p1, p2, . . . , pn, and the sign by s, we get the formula
(3) P1, p2, pn D s
Sounds p1, p2, . . . , pn, constituting the first term of the relation ‘diacritic of’, I call phonemes.
The function of phonemes is to differentiate between signs. A sound taken as a physical element is not a linguistic element; it turns into a linguistic element only as a phoneme, that is, as a first term of the relation ‘diacritic of’. Signs are second terms of the relation ‘diacritic of’, because they are differentiated from one another by phonemes.
Here is a concrete example of the relation ‘diacritic of’: In the word pin, the sounds p, ɪ, n are the first terms of the relation ‘diacritic of’, and the sign pɪn is a second term of this relation. The sign pɪn has a meaning, but the sounds p, ɪ, n do not have meaning: they are used to differentiate the sign pɪn from other signs, such as bɪn, pen, pɪt, hɪp, etc.
It should be stressed that while a sign has a meaning, a phoneme does not have a meaning; it has only a differentiating function—and therein lies the essential difference between signs and phonemes.
We are now ready to formulate the Principle of the Differentiation of Signs:
If two signs are different, they must be differentiated by different sequences of phonemes.
In order to understand why a diacritic, that is, phonemic, stratum is indispensable for any human language, consider a sign system that is simpler than any human language and has only two phonemes. Call them phonemes A and B. Phonemes A and B are diacritics, that is, signals that do not have meaning but are used to produce signs. Call the class of signs produced by phonemes A and B the lexicon of the sign system. If the sign produced must have one-phoneme length, this sign system will be able to produce only two signs. Its lexicon can be increased, however, if it can produce signs by combining the phonemes in pairs: AA, AB, BA, BB give four signs. If it can combine the two phonemes in triplets, eight signs will be produced: AAA, AAB, ABA, ABB, BAA, BAB, BBA, BBB. The longer the sequence, the larger the lexicon. The general rule is: m different phonemes in sequences of length n provide m n different signs.
Since the potential size of the lexicon increases exponentially as the length of the sequences increases linearly, sequencing is an efficient way to achieve a large lexicon with a limited number of different phonemes.
2) Sequencing. Auditory signs can be unfolded only in time. Auditory signs have only the dimension of time, in contrast to visual signs (nautical signal, etc.), which can have simultaneous groupings in several dimensions. The signs of human language, being auditory, are presented in succession; they form sequences. Sequencing is such an obvious property of language that it seems too simple; nevertheless, it is fundamental, and its consequences are innumerable.
An illustration of the power of sequencing was given above with respect to the diacritic stratum: a small set of phonemes can be arranged in different sequences to form thousands of different signs. No less powerful is sequencing with respect to sign stratum: thousands of signs in the lexicon of language are arranged in different sequences to form an enormous variety of sentences.
3) Use of rules. Language is a rule-governed sign system. In order to understand why such should be the case, let us focus on sequencing.
We cannot make unlimited use of the sequential strategy, because of the possibility of error. If we want to be able to recognize errors, not all possible sequences must be meaningful. Therefore, we must admit sequences longer than they would be if the sequential strategy were fully exploited. That is, we must admit redundancy in terms of communication theory.
Human languages are redundant because it is impossible for us to pronounce all conceivable sequences of phonemes. Let us illustrate this assertion in terms of written English. It has been estimated that if the sequential strategy were fully exploited, the English alphabet of 26 letters would give 26 possible one-letter words, 676 possible two-letter words, 17,576 possible three-letter words, and 456,976 possible four-letter words—a total of 475,254 words, or about the number in Webster’s New International Dictionary. But this dictionary has a great many words with more than four letters.
Redundancy exists also on the sign-stratum level: not every sequence of words is admissible. We can say John bought an interesting book, but Interesting an bought book John does not make sense and is therefore inadmissible. It has been estimated that human languages are about 75 percent redundant; that is, if some language could use all possible sequences of letters to form words and all possible sequences of words to form sentences, its book would be about one-quarter the length of the books of the existing human languages.
Redundancy is useful because it decreases the possibility of errors. We hear words more accurately in sentences than in isolation because we know that some sequences of words are inadmissible.
Our problem now is: How do we know which sequences are admissible and which are not? Can it be a matter of memory?
Memorization is of no use, because there are too many possible phonemic sequences and too many possible sentences. It has been estimated that each successive word is, on the average, chosen from 10 alternatives possible in that context. That means there are about 1010 grammatical sentences 10 words long. Because there are fewer than 3.6 x109 seconds per century, we would not have the time to memorize all of the admissible 10-word sequences, even if we could work at a rate of one second per sentence. And even if we could, we would know only the 10-word sentences, which are a small fraction of all the sentences we would need to know.1
If memorization is out of the question, we must assume that a system of rules is incorporated into any language that makes it possible to discriminate between admissible and inadmissible sequences. A grammar is a description of this system of rules.
It should be noted that the same expression can be constructed in different ways. For instance, the passive sentence “Hamlet” was written by Shakespeare can be constructed in two ways: either by directly combining the predicate was written with other parts of the sentence, or by deriving this sentence from the active sentence Shakespeare wrote “Hamlet.” In accordance with these two possibilities, two different sets of rules are used.
The equivalence of different constructions of the same expression is an important phenomenon called the polytectonic property. The polytectonic property is characteristic of natural languages, while the artificial languages of mathematics and logic are generally monotectonic; that is, every expression in these artificial languages has a unique construction (Curry, 1961; 1963: 60).
4) Structure. The question arises: How does language have the potential to produce an infinite number of sentences? The answer is: Through sentences’ having structure.
What, then, is the structure of a sentence?
To exhibit the structure of an object is to mention its parts and the ways in which they are interrelated. The structure of sentences can be illustrated by the following example. Suppose that for the words in a given sentence we substitute other words, but in a way that still leaves the sentence significant. Suppose we start with the sentence
(4) John married Mary.
For John we substitute Boris; for married, visited; and for Mary, Bill. We thus arrive at the sentence
(5) Boris visited Bill.
Sentence (5) has the same structure as (4). By substituting other nouns for John and Mary and other verbs for married in (4), we will get an enormous amount of various sentences having the same structure. The structure of all these sentences can be characterized by the following formula:
(6) Noun + Verb + Noun
We find two types of relations here: 1) relations between the parts of the sentence, in our case the relations between the first noun and the verb, between the verb and the second noun, and between the first noun and the second one; and 2) relations between words that can be substituted for one another, in our case between John, Boris, and other nouns that can be substituted for one another, and between the verbs married, visited, and other verbs that can be substituted for one another. Relations of the first type are called syntagmatic relations, and of the second type, paradigmatic relations.
The structure of a sentence is a network of syntagmatic relations between its parts and paradigmatic relations between each part and all other expressions that can be substituted for it.
Not only sentences possess structure, but also words and sequences of phonemes. Suppose we start with the word teacher; let us substitute read for teach and ing for er. We thus arrive at the word reading. The possibility of replacing teach and er with other elements shows that the word teacher possesses structure: there is a syntagmatic relation between teach and er, and there are paradigmatic relations between teach and other elements that can be substituted for teach, on the one hand, and between er and other elements that can be substituted for er, on the other.
Suppose now that we start with the sequence of phonemes man. By interchanging m and p, a and e, n and t respectively, we obtain different words, namely, pet, pen, pan, met, men, mat. This possibility of interchanging the phonemes shows that sequences of phonemes possess structure. Thus, in man there are syntagmatic relations between m, a, and n and paradigmatic relations between each of these phonemes and other phonemes that can be interchanged with it.
To establish the structure of a sentence or of any other sequence is to mention its parts and the syntagmatic and paradigmatic relations that characterize it. It is structure that makes the number of possible sentences and other sentences of a language unlimited.
I will use the term contrast to designate syntagmatic relations and the term opposition to designate paradigmatic relations.
5) Hierarchical stratification. Let us take the Latin expression i ‘go (imperative)’. It is the shortest possible expression, and at the same time it is a complete sentence that contains a variety of heterogeneous elements. What are these elements?
1. i is a sound, that is, a physical element;
2. i is a phoneme, that is, a diacritic, a functional element that differentiates linguistic units;
3. i is a root (a lexical morpheme), that is, an element that expresses a concept;
4. i is a part of speech (an imperative form of a verb);
5. i is a part of a sentence (a predicate);
6. i is a sentence, that is, a message, a unit of communication.
These elements belong to different levels of language; in other words, i is stratified. I have chosen a one-sound expression deliberately, in order to show that the difference between the levels of language is qualitative rather than quantitative: although linguistic units of a higher level are usually longer than units of lower levels—for example, a word is usually longer than a morpheme, and a sentence is longer than a word—what is crucial is not the length of expressions but their function; language is stratified with respect to different functions of its elements.2
What are the functions of these elements?
1. Sounds per se do not have any function, but they embody phonemes and linguistic units with various functions.
2. Phonemes have a diacritical, or distinctive, function.
3. Morphemes have a function of signifying concepts:
a) root concepts, such as child-, king-, govern-, kin-, and
b) nonroot concepts of two kinds:
b1) subsidiary abstract concepts, such as -hood, -dom, -ment, -ship (childhood, kingdom, government, kinship), and
b2) syntagmatic relations, such as -s in (he) writes or -ed in (he) ended.
4. Parts of speech—nouns, verbs, adjectives, adverbs—have a symbolic function; that is, they name the elements of reality.
5. Parts of sentences have a syntactic function as elements of a message.
6. Sentences have a communicative function; that is, they are messages—units of communication.
Besides the above functions, there is an important function called deixis (which comes from a Greek word meaning ‘pointing’ or ‘indicating’). Deixis is a function of demonstrative and personal pronouns, of tense, of concrete cases, and of some other grammatical features that relate a sentence to the spatio-temporal coordinates of the communication act.
Finally, in terms of the three essential components of the communication act—the speaker, the hearer, and the external situation—to which reference may be made, a sentence has one of the following functions: a representational function, a vocative function, or an expressive function. A sentence has a representational function if it describes a situation referred to by the speaker; it has a vocative function if it serves as a directive imposing upon the addressee some obligation (a sentence in the imperative mood, interrogative sentence, etc.); it has an expressive function if it refers to the speaker’s wishes or emotions.
The foregoing shows that the notion of linguistic level is a functional notion. This notion is central to linguistic theory. There are very complex hierarchical relationships between linguistic levels. For instance, the level of sounds (for generality, this level can be considered a zero-function level) is subordinated to the phonemic level, the level of words is subordinated to the level of sentences, etc.
To discover laws that characterize linguistic levels and hierarchical relationships between them is the main concern of linguistic theory.
The above examples are meant only to give the reader a glimpse into the stratification of language. A rigorous systematic description of linguistic levels and relationships between them will be given in the course of the book.
6) Semiotic relevance. Many people think of language as simply a catalogue of words, each of which corresponds to a thing. Thus, a particular plant, say an oak, is matched in the English language with a particular sound sequence that has an orthographic representation oak. From this point of view, differences between languages are reduced to differences of designation; for the same plant the English say oak, the Germans Eiche, the French chêne, the Russians dub. To learn a second language, all we have to do is to learn a second nomenclature that is parallel to the first.
This idea of language is based on the naive view that the universe is ordered into distinct things prior to its perception by man. It is false to assume that man’s perception is passive. Far from it. Man’s perception is guided by the language he uses; and different languages classify items of experience differently. True, to a certain point there may exist some natural division of the world into distinct natural objects, such as, for example, different species of plants, but on the whole, speakers of different languages dissect the universe differently. Different languages produce different dissections of the universe.
Here are some examples of different dissections of the universe by different languages.
The solar spectrum is a continuum that different languages dissect differently. So the English word blue is applied to a segment of the spectrum that roughly coincides with the Russian zones denoted by the Russian words sinij and goluboj. Two segments of the spectrum denoted by the English words blue and green correspond to one segment denoted by the Welsh word glas. The English word wash corresponds to two Russian words, myt’ and stirat’. In English, wash occurs in both wash one’s hands and wash the linen, but Russian uses the verb myt’ with respect to washing one’s hands and the verb stirat’ with respect to washing the linen. The English verb marry corresponds to two Russian expressions: marry translates into Russian differently in Peter married Mary and Mary married Peter. English makes a difference between to float and to swim (wood floats on water, fish swim in water), while Russian uses the same word plavat’ in both cases. English makes a difference between to eat and to drink, while Persian uses the same word khordan in both cases (gušt khordan ‘to eat meat’, may khordan ‘to drink wine’). The Latin word mus corresponds to two English words, mouse and rat.
The same holds for grammar. For example, in Russian, the grammatical morpheme -l means, depending on the context, either a past event or a past event preceding another past event. It covers the area that is covered in English by two tenses—the past tense and the past perfect tense. The present tense in Russian corresponds to two tenses in English—the present indefinite tense and the present continuous tense. Modern Russian analyzes the notion of number into a singular and a plural, while Old Russian (like Ancient Greek and some other languages) added a dual. There are languages that add a trial (like most Melanesian languages) or a quadral (like the Micronesian language on the Gilbert Islands). Similar facts are well known.
Every language presents its own model of the universe. This property of languages is called linguistic relativity. The notion of linguistic relativity was advanced by Humboldt, but it was most clearly formulated by Sapir and Whorf.
Why does every language give a relative picture of the world?
Linguistic relativity can be explained as a consequence of the Principle of Semiotic Relevance:
The only distinctions between meanings that are semiotically relevant are those that correlate with the distinctions between their signs, and, vice versa, the only distinctions between signs that are relevant are those that correlate with the distinctions between their meanings.
Let us examine linguistic relativity in the light of the Principle of Semiotic Relevance.3 Consider the above examples. The English word wash has different meanings in the context of the expressions wash one’s hands and wash the linen. But the distinction between these two meanings is irrelevant for the English language, because this distinction does not correlate with a distinction between two different sound sequences: in both cases we have the same sound sequence denoted by wash. Therefore, these two meanings must be regarded not as different meanings but as two variants of the same meaning. On the other hand, the meaning of the Russian word myt’, which corresponds to the meaning of the English wash in wash one’s hands, and the meaning of the Russian word stirat’, which corresponds to the meaning of the English wash in wash the linen, must be regarded as different meanings rather than variants of the same meaning as in English, because the distinction between the meanings of Russian myt’ and stirat’ correlates with the distinction between different sequences of sounds, and therefore is relevant for the Russian language. A similar reasoning applies to the other examples. As to the relevant and irrelevant distinctions between signs, consider, for instance, the substitution of ə for ɪ in the penultimate syllables of terminations such as -ity, -ily, such as ə’bɪlətɪ for ə’bɪlɪtɪ (ability). Since the distinction between the two signs is not correlated with a distinction between their meanings, this distinction is irrelevant, and therefore they must be regarded as variants of one and the same sign. Another example: The distinction between the signs nju and nu is not correlated with a distinction between their meanings. Therefore, these signs are variants of the same sign denoted by a sequence of letters new.
One may wonder whether the Principle of Semiotic Relevance results in circularity: while the relevant distinctions between meanings are defined by their correlation with the distinctions between their signs, at the same time the relevant distinctions between signs are defined by the distinctions between their meanings. As a matter of fact, this principle does not result in circularity. The point is that the relation ‘sign of’ makes signs and meanings interdependent: distinctions between signs do not determine the distinctions between meanings, nor do distinctions between meanings determine the distinctions between signs; each kind of distinctions presupposes the other. Neither the distinctions between signs nor the distinctions between meanings should be taken as primitive. What is really primitive is the correlation of distinctions between sounds and distinctions between meanings. There is no circularity here, because both relevant distinctions between signs and relevant distinctions between meanings are determined by their correlation.
The Principle of Semiotic Relevance is an empirical principle reflecting the nature of human language, and therefore all consequences of this principle are also empirical and bear on the nature of human language.
The notion of relevance is relative: what is relevant from one point of view may be irrelevant from another. Semiotic relevance means relevance from a semiotic point of view. What is irrelevant from a semiotic point of view may be relevant from an extrasemiotic point of view. Thus, the distinction between rat and mouse is irrelevant in Latin, but it is relevant from a zoologist’s point of view. The distinction between ingesting food and drink is irrelevant in Persian, but it is relevant from a physiologist’s point of view. A distinction between some colors irrelevant for a language may be relevant from a painter’s point of view.
The meaning of a sign is an entity that has dual aspects: semiotic and extrasemiotic. The extrasemiotic aspect of meaning is rooted in the culture of people who speak a given language. There is a permanent conflict between the semiotic and extrasemiotic aspects of the meaning: what is relevant from a semiotic point of view may be irrelevant from an extrasemiotic, cultural point of view; and, vice versa, what is relevant from an extrasemiotic, cultural point of view may be irrelevant from a semiotic point of view. This conflict is resolved in a speech act: a successful act of communication presupposes that the speech situation rooted in the common cultural background of the speaker and hearer transforms the semiotic aspect of the meaning into the extrasemiotic, so that the transmitted message can be properly understood by the hearer.
Every language as a specific model of the universe is a rigid cognitive network that imposes on the speakers and hearers its own mode of the classification of the elements of reality. But cognition is not divorced from communication. Human practice, human activity rectifies the rigid cognitive structures of a language in the process of communication.
Linguistic relativity as it was formulated by Humboldt, Sapir, and Whorf is valid with respect to the cognitive structures of language, but it must be supplemented by the notion of the duality of the meaning of the linguistic sign, which explains why in the process of communication people are able to transcend a rigid cognitive framework imposed on them by the language they use.
The Principle of Semiotic Relevance has empirically testable consequences. The validity of this principle depends on whether counterexamples could be found that would invalidate it. What evidence would serve to invalidate it?
Consider ambiguity. If the Principle of Semiotic Relevance were invalid, then we could distinguish different meanings no matter whether or not they were correlated with different sound sequences. Let us take the sentence I called the man from New York. This sentence is ambiguous; it has two very different meanings: ‘I called the man who is from New York’ or ‘From New York, I called the man’. One cannot distinguish between the two meanings, because they are not correlated with different sequences of signs. Or take a case of phonological neutralization: German rat may mean either Rad or Rat, but speakers of German do not distinguish between these meanings, because they are not correlated with different sound sequences. This ambiguity can be resolved by different contexts only because there exist oppositions, such as Rade:Rate.
No explanation of puns or jokes would be possible without an understanding of the Principle of Semiotic Relevance. Consider: Why did the little moron throw the clock out the window? To see time fly. Or: Why didn’t the skeleton jump out of the grave? Because he didn’t have the guts to. The ambiguity of the expressions time flies and to have the guts to do something can be explained by the absence of a correlation between the differences in meaning and the differences between sound sequences.
In all the above examples, a suspension of a distinction between different meanings is the result of the suspension of a distinction between different signs.
2. The Principle of Semiotic Relevance and Homonymy
The Principle of Semiotic Relevance calls for reconsidering the traditional notion of homonyms.
A homonym is defined as a word or expression that has the same sound as another but is different in meaning. For example, the noun bear and the verb bear are said to be homonyms of each other, or simply homonymous. In terms of signs, homonyms are different signs that have the same form, that is, the same sound. The relation between two homonyms can be represented by the following diagram:
The notion of homonyms contradicts the Principle of Semiotic Relevance, which holds that different signs must be different in form, that is, in sound or in some other marker. A correct diagram for two different signs must be
In the light of the Principle of Semiotic Relevance, what is called a homonym is one sign rather than two (or more) signs. This sign can be represented by the diagram
The sign in (3) can be called a polyvalent sign.
Under the Principle of Semiotic Relevance, meaning 1 and meaning 2 in (1) belong to the same class of meanings, no matter how different they are; hence the ambiguity of the sign in (1). To distinguish between meaning 1 and meaning 2, we must have two different signs corresponding to these meanings, as in the diagram
The sign in (3) can be viewed as a suspension of the distinction between the signs in (4). And the ambiguity of the sign in (3) can be viewed as a suspension of the distinction between the meanings in (4), which is a counterpart of the suspension of the distinction between signs.
We can formulate the Principle of Suspension as a corollary of the Principle of Semiotic Relevance:
If a distinction between a sign X and a sign Y is suspended through replacing them with a single sign Z, then the meanings of X and Y combine to form a single class of meanings designated by Z.
Let me illustrate this principle. Consider the following sentences:
(5)a. John killed a bear.
b. Her husband is such a bear that nobody likes him.
c. These letters bear no dates.
d. What is the Russian word for bear?
The meanings of bear in (5a) and (5b) are different—in (5a) bear designates an animal, and in (5b) bear designates a man through a metaphor—but they constitute a single class of meanings since they are designated by the same sign bear. The meanings of bear in (5a) and (5c) are different and belong to different classes of meanings, because they are designated by different signs: bear + the place of a term in (5a) and bear + the place of a predicate. As was pointed out in section 1.1, a sign is not necessarily a sound sequence—it may be a context. Here we have a case of signs that are combinations of sound sequences and grammatical contexts. It is to be noted that not every change of a context means a change of a sign. Thus, the contexts of bear in (5a) and (5b) are different, but these contexts are lexical and therefore cannot constitute different signs.
In (5d), bear is taken out of the grammatical contexts of (5a, b, and c). Therefore, it must be viewed as the result of the suspension of the distinction between the two signs bear + the place of a term and bear + the place of a predicate. Hence, it is ambiguous.
In current linguistic literature, both ‘homonymy’ and ‘ambiguity’ are used as related notions. The present analysis of these notions is their theoretical explication based on the Principle of Semiotic Relevance. While ambiguity as explicated above is a consequence of the Principle of Semiotic Relevance, we have to reject homonymy as incompatible with this principle.
Since the terms ambiguous sign and ambiguity are not well defined in current linguistic literature, I propose the terms polyvalent sign and polyvalence as designating more precise concepts based on the Principle of Semiotic Relevance.4
3. Saussure’s Notion of the Sign
The notion of the sign proposed above differs from Saussure’s notion (Saussure, 1966:65-70). Saussure considers a sign to be a bilateral entity consisting of a signifiant and a signifié.
Saussure seems to have introduced his notion of the sign in order to dispel the widespread misconception according to which a language is merely a nomenclature or a stock of labels to be fastened onto preexistent things. Thus, he wrote:
I call the combination of a concept and a sound-image a sign, but in current usage the term generally designates only a sound-image, a word, for example (arbor, etc.). One tends to forget that arbor is called a sign only because it carries the concept “tree,” with the result that the idea of the sensory part implies the idea of the whole. (Saussure, 1966:67)
Saussure’s concern about dispelling the naive view of language as a stock of labels used as names for preexisting things is understandable. But it is difficult to accept his claim that arbor is called a sign because it carries the concept ‘tree’. True, the expression arbor is a sign because of its being in a particular relation to the meaning ‘tree’. Similarly, a man is a husband because of his having a wife. But it does not follow that a husband is a combination of two people.
Saussure fails to see the difference between two sharply distinct notions: 1) the notion that a thing X belongs to a class K through its having a relationship to another thing Y, and 2) the notion that the thing X and the thing Y together form a whole, that is, a member of the class K.
One might argue that by saying that “arbor is called a sign only because it carries the concept ‘tree’” in conjunction with his decision to use the term sign to designate the combination of a concept and a sound-image, Saussure did not mean that the concept is part of the sign but simply wanted to justify the introduction of a new notion into linguistics—a bilateral linguistic unit consisting of a signifiant and a signifié. One might argue that we should not cavil at Saussure’s unfelicitous wording of his thought.
Let us then accept the term sign as an arbitrary label for designating the bilateral linguistic unit consisting of a signifiant and a signifié. And let us consider whether linguistics needs this bilateral unit.
The least one can say about the sign in the sense of a bilateral linguistic unit consisting of a signifiant and a signifié is that this notion is gratuitous. Actually, since the traditional notion of the unilateral sign, if correctly explicated, involves reference to meaning, linguistics does not need the notion of the sign as a bilateral linguistic unit. What linguistics needs is a correct explication of the traditional notion of the unilateral sign. As was shown in section 1.1, neither the sign nor the meaning must be taken as a primitive notion. The primitive notions are to be the relation ‘sign of’ and its converse ‘meaning of’. The sign and the meaning are the terms of these binary relations.
The relations ‘sign of’ and meaning of’ are asymmetric: the sign communicates the meaning, but the meaning does not communicate the sign. Semiotically, the meaning is subordinate to the sign. Therefore, taking the sign as a unilateral unit, we do not exclude the reference to meaning but view the sign and the meaning in the right perspective. One proof that the notion of the bilateral sign unnecessarily complicates linguistic theory is that both Saussure himself and the linguists who adhere to his terminology constantly fall back into the unilateral-sign terminology.
Under the present explication, the sign remains unilateral, and we do not need the notion of a bilateral linguistic unit.
The notion of the bilateral sign unnecessarily complicates linguistic terminology and precludes the use of the good term sign in its traditional sense, which is quite useful for linguistics, if correctly explicated.
There might be other objections against the notion of the bilateral sign. One is this: If we conceive of the word as a unilateral sign, then such a word as bear with different meanings will be characterized as a polyvalent unilateral sign. But if we conceive of the word as a bilateral sign, then we have to say that there are two homonymous signs bear1 and bear2, which have a common signifiant. And, as was shown in section 1.2, the notion of homonymy is untenable on semiotic grounds.
By introducing some new distinctions and definitions, we probably could resolve the above difficulty. But that would add new complications to the unnecessary complication involved in the notion of the bilateral sign.5
4. Linguistics as a Part of Semiotics
As was shown above, language can be studied from different points of view, but a semiotic approach is crucial for comprehending the essence of language. Since natural languages are only a subclass of possible sign systems, a study of natural languages in a context of a general theory of possible sign systems is indicated. This theory is called semiotics. The principles and laws of semiotics apply to the study of natural languages, as well as to other sign systems. Therefore, linguistics must be regarded as a part of semiotics. We can define linguistics as a part of semiotics that deals with natural sign systems, that is, with natural languages.
Given the definition of linguistics as a part of semiotics, we have to look for the specific features of linguistics as a semiotic discipline. To this end we must consider the question, What is the specific difference between natural languages and all other kinds of sign systems?
Any natural language must be an adequate means of expressing the conceptual network of the human mind; that is, any natural language must be a sign system, which can have as many distinct signs as there are distinct concepts in the conceptual network of the human mind and which can create as many signs as are necessary to express new concepts in the processes of human cognition and communication. This feature of any natural language can be called the property of cognitive flexibility of natural languages. Being cognitively flexible, any natural language is an extremely rich, variable, and productive sign system. This feature of natural language makes it possible always to translate a message expressed in some sign system into a message expressed in a natural language, whereas the reverse is not true: not every message expressed in a natural language can be translated into a message expressed in some other sign system.
As a direct consequence of its cognitive flexibility, any natural language has a system of diacritics. To be cognitively flexible, a natural language must have the capacity of producing a potentially infinite number of signs. The system of diacritics serves as a means for the production of signs. Owing to a system of diacritics, any sign can be produced by combining diacritics; so any sign is nothing but a combination of diacritics.
Cognitive flexibility is a fundamental property of any natural language. Owing to this property, any natural language is a remarkably rich, complex, and infinitely variable and productive system of signs. From the point of view of cybernetics, natural languages belong to a type of systems that are called in cybernetics very large systems.
What, then, is a very large system, as understood in cybernetics?
In cybernetics the concept ‘large system’ refers to a system with a large number of distinctions. The larger the system, the larger the number of its distinctions. A very large system will have a very large number of distinctions. The words very large are used in cybernetics to imply that, given some definite observer with definite resources and techniques, the system beats him by its richness and complexity: so that he cannot observe it completely, or control it completely, or carry out the calculations for predictions completely.
Natural languages belong to such ‘very large systems’ because each natural language has a very large number of distinctions.
How, then, can a child learn a natural language, which is a very large system? How can a child acquire a system that beats him by its richness and its complexity?
The acquisition of a natural language by a child will remain a mystery unless we assume a hypothesis whereby there is available a simple sign system that underlies natural languages and controls their functioning. I call this simple sign system the linguistic genotype. The linguistic genotype is a common semiotic basis underlying all natural languages. Natural languages, in this sense, are embodiments of the linguistic genotype, and the functioning of every natural language simulates the functioning of the linguistic genotype. We can assume that a child acquires a natural language through the linguistic genotype, which, of course, does not exist independently of natural languages but is, so to say, built into them. Thus, the process of acquiring a natural language through the linguistic genotype must be conceived of as an unconscious process. But it is essential to assume that the linguistic genotype has an objective existence as a common semiotic basis of natural languages.
To study the linguistic genotype is to study the common semiotic properties of natural languages, that is, to study the basic semiotic laws of the functioning of natural languages.
The treatment of linguistics as a part of semiotics contrasts with a fairly widespread view that linguistics is a part of psychology.
True, linguistic processes in the human mind involve psychological phenomena. But logical processes in the human mind also involve psychological phenomena; nobody, however, views logic as a part of psychology. Although logical processes involve psychological processes, logic is independent of psychology, because logic has a specific subject matter for which psychological considerations are irrelevant: rules of deduction and the construction of deductive systems.
Although psychological phenomena accompany linguistic processes in the human mind, linguistics is independent of psychology, because psychological considerations are irrelevant for understanding sign systems. Apart from some marginal cases of onomatopoeia and related phenomena, the signs of human language are arbitrary and conventional. Language is a tacit social convention shared by the members of the linguistic community; the infringement of this social convention makes the speaker run the risk of being incomprehensible or ridiculous. Language as a network of social conventions is what linguistics is all about. Social conventions are supraindividual; they are logically independent of the psychological processes that may be connected with them in the minds of individuals.
Linguistics is completely independent of psychology. The psychology of speech is not even an auxiliary science of linguistics. The investigation of linguistic phenomena by methods of psychology is, of course, possible, and it is important. But a necessary prerequisite for such investigation is the previous establishment of linguistic facts: psychology of speech presupposes linguistics as its basis.
5. The Goals of Linguistic Theory and the Semiotic Basis of Abstraction
Observing natural languages makes obvious the many differences among them; not only are genetically unrelated languages, such as English and Chinese, very dissimilar, but languages that have a common origin, such as English and German, also differ from one another in many important ways. And yet, one can also discover important similarities among languages. Thus, the grammar of every language includes a system of obligatory syntactic functions. For instance, not every language differentiates between nouns and verbs, but every language must differentiate between the two basic components of a sentence: predicates and their terms.
On the other hand, although languages may vary greatly one from another, the possibilities of variation among languages are not unlimited: there are regular patterns of variation among languages that are limited by intrinsic functional and structural properties of signs. For instance, languages may vary in word order patterns, but these patterns can be reduced to a limited number of types determined by the intrinsic linearity of sign sequences in human speech. Language typology is possible only because there are functional and structural constraints on possible differences among languages.
Linguistic similarities and differences seem to be determined by some unknown factors that constitute the essence of natural languages. Therefore, linguistic similarities and differences must be recognized as significant phenomena that provide clues to the understanding of what a natural language is. These phenomena must be explained in terms of principles that account for the essence of natural languages.
The basic question of linguistic theory must be: What factors contribute to the similarities and differences between natural languages?
To answer this question, linguistic theory must achieve the following goals:
First, it must define the essential properties of natural language as a special kind of sign system.
Second, it must state linguistic universals, that is, all necessary or possible consequences of the properties of natural language as a sign system.
Third, it must explain facts of individual languages; that is, it must subsume these facts under classes of phenomena characterized by the principles and laws it has stated.
Fourth, it must construct insightful typologies and grammars of individual languages.
In accordance with the distinction of two semiotic strata in a natural language—the system of signs and the system of diacritic elements—linguistic theory consists of two parts: phonology and grammar. Phonology is the study of the system of diacritic elements, and grammar is the study of the system of signs.
In pursuing its goals, linguistic theory faces the following problem: What basis could we provide for justifying our abstractions? How can we distinguish between correct and incorrect abstractions? What abstractions have cognitive value, and what abstractions, far from having cognitive value, distort linguistic reality and hamper progress in linguistics? For example, what basis could we provide for such abstractions as deep structure in transformational grammar, underlying phonological representations in generative phonology, the universal notion of subject in various linguistic theories, and so on? A linguistic theory that lacks a reliable basis for its abstractions is built on thin ice.
The problem of justifying abstractions is central to any linguistic theory. As a matter of fact, all important discussions and controversies in contemporary linguistics turn around this basic problem, which so far has not been resolved satisfactorily.
Having come to grips with the above problem, I propose to constrain abstractions by tracing out the consequences of the semiotic principles, such as the Principle of Semiotic Relevance explained above and related principles, which will be presented in the following chapters of the book. If the assumption that the semiotic aspect of language constitutes its essence is correct, then tracing out consequences of the semiotic principles will be a necessary basis for a principled choice between conflicting abstract hypotheses advanced for the explanation of linguistic phenomena.
6. Synchronic Linguistics and Diachronic Linguistics
Languages change constantly. The evolution of a language is a succession of its different states in time. Evolution implies at least two different states: an initial state and a final state. Evolution has structure: any two successive states of a language are interrelated with each other by some regular correspondences. These correspondences can be stated as a series of rules. Two types of facts are distinguished: 1) paradigmatic and syntagmatic oppositions between linguistic units coexisting at a given state of a language—synchronic facts, and 2) correspondences between the successive states of a language—diachronic facts.
We want to find out, of course, how synchronic and diachronic facts are connected to each other. The answer is paradoxical: Diachronic facts are responsible for and at the same time irrelevant to the synchronic structure of a language.
Here is an example of the interrelation of the synchronic and diachronic phenomena (Saussure, 1956:83-84). In Anglo-Saxon, the earlier forms were fōt:fōti, tōp: tōpi, gōs.gōsi, etc. These forms passed through two phonetic changes: 1) vowel change (umlaut): fōti became fēti; tōpi, tēpi; gōsi, gēsi; and 2) the fall of final -i: fēti became fēt; tēpi, tēp; gēsi, gēs (Modern English: foot: feet; tooth: teeth; goose: geese). A comparison of the two states of the evolution of English—the Anglo-Saxon state and the Old English state—shows that the synchronic rules of the formation of the plural were different in both instances: in Anglo-Saxon the plural was marked simply by the addition of an -i; in Old English the plural was marked by the opposition between vowels.
Although the plural formation in Old English evolved from the plural formation in Anglo-Saxon, the former was logically independent of the latter. The synchronic statement of the rule of plural formation in Old English has nothing to do either with the above phonetic changes or with the rule of plural formation in Anglo-Saxon.
This example shows that synchronic facts are logically independent of diachronic facts. Hence, linguistics bifurcates into two branches: synchronic linguistics and diachronic linguistics.
A language as an instrument of communication is a synchronic system. Therefore, synchrony is logically prior to diachrony. The synchronic study of languages is the fundamental task of linguistics and a necessary prerequisite for the diachronic study. Synchronic linguistics takes precedence over diachronic linguistics.
Synchronic should not be confused with static. Any synchronic system contains archaisms and innovations consisting of newly coined words, new metaphorical uses of words, etc.; deviations from the received phonetic, grammatical, and lexical standards occur that may represent both remnants of the past and germs of the future development. These phenomena have nothing to do with static: they characterize the dynamic aspects of the synchronic system. Therefore, the term static linguistics, which is sometimes used as a synonym of synchronic linguistics, is inappropriate.
In accordance with the distinction of the two semiotic strata in language, synchronic linguistics consists of two parts: 1) phonology—the study of the diacritic stratum of language; and 2) grammar—the study of the sign stratum of language. I take the term grammar in a broad sense. It covers not only the notion of grammar in the narrow sense, but also semantics and everything pertaining to the study of the linguistic units that constitute the sign stratum of language.
Language changes not only in time but also in space and with respect to the stratification of a linguistic community into social classes and ethnic groups.
In accordance with the variation of a language in different geographical territories, we distinguish different dialects of a language. The term dialect applies to every regional form of a given language without any implication that a more acceptable form of the language exists distinct from the dialects. This use of the term dialect does not imply a judgment of value. For example, every American speaks a dialect—New England dialect, Southern dialect, etc. Speakers of these dialects never feel that they speak anything but a form of American English that is perfectly acceptable in all situations.
The term dialect also has quite a different use. When applied to languages such as Russian, German, or Italian, the term in current use implies a judgment of value: dialects are contrasted with a national language—for example, German has different dialects, but there is a form of German that is not a dialect but a language. There are Germans who do not speak any German dialect but the German language. In this sense, dialects are contrasted with a language not as particular forms of a linguistic system with its general form but as particular forms with a particular form that is accepted nationwide.
Dialects arise as a result of a decrease in the frequency and intimacy of contact between various sections of the population separated from one another by different geographical localization. But the diminution of the frequency and intimacy of contact may be a result of a social and ethnic stratification of a linguistic community. Therefore, it is appropriate to use the term dialect to distinguish not only varieties of a language in space but also varieties conditioned by a social and ethnic stratification of a linguistic community. In current linguistic literature, the terms social dialect (sociolect) and ethnic dialect are used in this sense. Examples of social and ethnic dialects are the working-class dialect in England (cockney) and black English in the United States.
Dialects of a language can be viewed not only in synchronic but also in diachronic perspective. If we combine the space perspective with the time perspective, we can distinguish four types of linguistic systems reflecting different kinds of abstraction from time or space:
1) a linguistic system with time and space excluded (monochronic monotopic linguistic system);
2) a linguistic system with time excluded and space included (monochronic polytopic linguistic system);
3) a linguistic system with time included and space excluded (polychronic monotopic linguistic system); and
4) a linguistic system with both time and space included (polychronic polytopic linguistic system).
The first and second types of linguistic systems relate to synchronic linguistics, and the third and fourth types of linguistic systems relate to diachronic linguistics. In this book I will deal exclusively with the first type.
8. The Semiotic versus Generativist Notion of Language
Noam Chomsky, the founder of a linguistic theory called generative-transformational grammar, once defined language as follows:
From now on I will consider a language to be a set (finite or infinite) of sentences, each finite in length and constructed out of a finite set of elements. (Chomsky, 1957:12)
Defining the goals of linguistic theory, he wrote:
The fundamental aim in the linguistic analysis of a language L is to separate the grammatical sequences which are the sentences of L from the ungrammatical sequences which are not sentences of L and to study the structure of grammatical sentences. (Chomsky, 1957:12)
Furthermore, Chomsky defined the grammar of a language as follows:
The grammar of L will thus be a device that generates all of the grammatical sequences of L and none of the ungrammatical ones. (Chomsky, 1957:12)
What strikes one in these definitions of language and grammar is Chomsky’s complete disregard of the fact that language is a sign system.
As a supposedly superior alternative to the semiotic notion of language as a sign system, Chomsky suggested a notion of language as a set of sentences.
As an alternative to the notion of grammar as a system of rules that constitutes an integral part of language, Chomsky suggested a notion of grammar that is not a part of language but is an external device for generating a language understood as a set of sentences.
Let us not argue about definitions. After all, every linguist, like every other scientist, has the right to define his terms in his own way. What matters is not definitions in themselves but the empirical consequences of such definitions. So, let us consider the empirical consequences of Chomsky’s notions of language and grammar.
If we accept the notion of language as a sign system, we cannot investigate grammar independently of meaning, because linguistic units are signs, and a sign as a member of the binary relation ‘sign of’ cannot be separated from its meaning. A sign separated from its meaning is no longer a sign but merely a sequence of sounds—a purely physical phenomenon.
If, on the other hand, we do not include the notion of the sign in the definition of language and base this definition on some other set of notions, as Chomsky used to do, then we are free to consider grammar to be independent of meaning. According to Chomsky, “grammar is autonomous and independent of meaning” (Chomsky, 1957:17). As a special case, he considered syntax to be an autonomous component of grammar distinct from semantics.
To support his claim that the notion ‘grammatical’ cannot be identified with ‘meaningful’, Chomsky devised an example of a sentence that was allegedly nonsensical but grammatically correct:
Colorless green ideas sleep furiously.
As a matter of fact, the nonsensical content of this sentence has no bearing on the question of whether or not its grammatical structure is meaningful. Chomsky confounded the notion of grammatical meaning with the notion of lexical meaning. But we must distinguish between lexical meaning and grammatical meaning. No matter whether or not from the standpoint of lexical meaning a sentence is nonsensical, if the sentence is grammatically correct, it is grammatically meaningful. So, the above sentence contains the following grammatical meanings: the noun ideas signifies a set of objects, the verb sleep signifies a state ideas are in, the adverb furiously signifies the property of sleep, and the adjectives colorless and green signify two different properties of ideas. Grammatical meanings are categorial meanings, that is, the most general meanings characterizing classes of words and other linguistic units. If this sentence did not have grammatical meanings, we could not even decide whether it is nonsensical or not. We consider this sentence nonsensical because of the conflict between the grammatical and lexical meanings: the grammatical meanings of the adjectives colorless and green and the verb sleep assign contradictory properties and a highly unlikely state to the object denoted by the noun ideas; the adverb furiously assigns a strange property to a state denoted by the verb sleep.
Compare the following expressions:
(1) round table
(2) round quadrangle
The meaning of (2) is nonsensical, because the grammatical, that is, categorial, meanings of its words conflict with the lexical meanings: the grammatical meaning of round assigns a contradictory property to the object denoted by the noun quadrangle. Expression (1) makes sense, because its lexical and grammatical meanings are in keeping with each other.
Consider the following verses from Alice in Wonderland by Lewis Carroll:
’Twas brillig, and the slithy toves
Did gyre and gimble in the wabe;
All mimsy were the borogoves,
And the mome raths outgrabe.
Do these verses make sense? Yes, of course, they do, for although they contain made-up nonsense words created by Lewis Carroll, we understand that did gyre and gimble signifies some actions in the past, in the wabe signifies a localization in some object, slithy signifies a property of the set of objects called toves, etc.
What, then, are grammatical meanings?
Grammatical meanings are morphological and syntactic categories. These categories are represented in the above verses by the plural suffix -s, the preposition in, the auxiliary verbs did and were, the conjunction and, the article the, and the word order.
Affixes, prepositions, conjunctions, etc. all have meaning because they are signs, and signs presuppose meaning. The notion of the meaningless sign is no better than the notion of the round quadrangle.
An analysis of a sentence into immediate constituents is impossible without an analysis of meaning. Consider the sentence
(3) The mother of the boy and the girl will come soon.
This sentence admits of two analyses into immediate constituents:
(4) The mother (of the boy and the girl) will come soon.
(5) (The mother of the boy) and the girl will come soon.
We can analyze these two sentences differently because they have two different grammatical meanings.
Let us now have one possible complete analysis of sentence (3) into its immediate constituents:
(6) (((The mother) (of ((the boy) (and (the girl)))))((will come) soon)).
If we disregard the meaning of (3), (6) is not the only possible way of analyzing (3) into immediate constituents. We could have constituents such as
(7) | (mother of) | ((boy and) the) | (and the) |
((mother of) the) | (come soon) | ((and the) (girl will)) | |
(girl will) | (boy and) | (of the) |
An analysis of a sentence into immediate constituents without an analysis of the meaning of the sentence admits of any arbitrary bracketing.
Why do we not analyze the sentence as shown in (7)? Because this analysis contradicts the semantic connections between words.
Any analysis of phrases into immediate constituents presupposes an analysis of semantic connections between words. A syntactic analysis presupposes a semantic analysis.
It is clear from the foregoing that an autonomous grammar independent of a semantic analysis is impossible unless we are resigned to doing a sort of hocus-pocus linguistics.
Here a further question arises: How did Chomsky manage to avoid unacceptable constituents, such as (mother of) or (of the) in the above example?
He did the trick by tacitly smuggling an analysis of meaning into an analysis of immediate constituents. But, smuggling semantic analysis into syntax cannot be an adequate substitute for a straightforward, consistent analysis of meaning as a part of syntax.
It should be noted that in the first version of generative-transformational grammar (1957), Chomsky was not concerned about semantics at all. However, he introduced a semantic component into the second version of his grammar (1965). That did not mean a change of his conception of an autonomous grammar independent of meaning: his grammar remained autonomous because the semantic component was conceived of as a component interpreting syntactic structures established independently of meaning. Clearly, the perverse idea that syntactic structures can be established without recourse to an analysis of meaning has persisted in all versions of generative-transformational grammar.
As a matter of fact, Chomsky inherited the idea of autonomous grammar from the distributionally oriented type of American structuralism, in particular from the works of his teacher Zellig S. Harris, who was concerned in grammatical description primarily with specifying patterns of occurrence and cooccurrence of elements. Harris worked without reference to meaning. His aim was to develop a method for representing grammatical structures of sentences without reference to semantic criteria; it was assumed that semantic statements would follow from purely formal syntax constructed independently of meaning.
Generative-transformational grammar is essentially a recasting of American distributional structuralism into a formal system. The new idea introduced by Chomsky was generation. He declared that structuralism was merely taxonomic, and he opposed his generative system to it as an explanatory model.
In order to evaluate the methodological significance of the notion of generation, let us consider some other important notions used in Chomsky’s works.
One fundamental factor involved in a speaker-hearer’s performance is his knowledge of grammar. This mostly unconscious knowledge is referred to as competence. Competence is distinct from performance. Performance is what the speaker-hearer actually does; it is based not only on his knowledge of language but also on many other factors—memory restrictions, distraction, inattention, nonlinguistic knowledge, beliefs, etc.
Chomsky uses the term grammar in two senses: 1) on the one hand, the term is used to refer to the system of rules in the mind of the speaker-hearer, a system that is normally acquired in early childhood; 2) on the other hand, it is used to refer to the theory that the linguist constructs as a hypothesis concerning the actual internalized grammar of the speaker-hearer.
Grammar in the sense of a linguistic theory is called a hypothesis, because the internalized grammar in the mind of the speaker-hearer is not available for immediate observation.
Chomsky assumes that the grammar in the speaker-hearer’s mind is not an ordinary grammar but a generative grammar. He constructs his theoretical generative grammar as a hypothesis about the real grammar in the speaker-hearer’s mind.
Chomsky assumes further that since the generative grammar in the speaker-hearer’s mind is not available for immediate observation, the only way to draw conclusions about it is from the results of its activity, that is, from the properties of the set of sentences it has generated. Under this assumption, only those aspects of a generative grammar in the speaker-hearer’s mind are relevant that cause generation of a particular set of sentences under consideration.
By analyzing all available sentences produced by this allegedly “generative” grammar in the speaker-hearer’s mind, Chomsky constructs his theoretical generative grammar, which serves as a hypothesis about the generative grammar in the speaker-hearer’s mind. Since only those aspects of generative grammar in the speaker-hearer’s mind are considered relevant that cause it to generate a set of sentences, the only thing that is required from theoretical generative grammar is a capacity for generation of the same set of sentences that is available for immediate observation. To verify a theoretical generative grammar therefore means to establish that it is capable of producing this set of sentences.
The idea of a theoretical generative grammar as a hypothesis about the generative grammar in the speaker-hearer’s mind looks very attractive, but it is actually a mistaken idea. If nothing is required of a theoretical generative grammar except that it generate correct sentences for a given language, then it must be considered unverifiable as a hypothesis about the real grammar in the speaker-hearer’s mind.
But what is wrong with generative grammar as a theoretical hypothesis?
Generative grammar aims at constructing a mathematically consistent system of formal rules. But mathematical consistency does not guarantee a correct description of reality.
Using a mathematical formalism, we can posit a system of rules for deriving sentences from certain basic linguistic objects. Granted that these rules work, does it mean that they present a reasonable model of the real rules of a language we are describing? No, it does not. From the fact that a mathematical design works, one cannot conclude that language works in the same way. Real rules of real language are empirical dependencies between truly basic linguistic objects and sentences that are derived from them because of an empirical necessity. But empirical necessity should not be confused with logical necessity. In accordance with the laws of logic, true statements can be logically necessary consequences of both true and false statements. Let me illustrate that with two examples.
We can deduce the true statement Butterflies fly from two false statements by constructing the following syllogism:
(1) Cows fly.
(2) Butterflies are cows.
(3) Butterflies fly.
In accordance with the rules of logic, the deduction of (3) from (1) and (2) is a logical necessity. But the logically necessary connection between (1) and (2), on the one hand, and (3), on the other hand, conflicts with empirical necessity.
Another example: Suppose we construct a calculus in which we posit some false initial statements such as 2=5, 3=7, and so on. Suppose, further, that this calculus has the following derivation rule: If x=y, then x can be substituted for y and y can be substituted for x. By applying this rule, we can derive true statements from the initial false statements, for example: 2=2, 3=3, 5=5, 7=7, and so on. The logically necessary connection between these true statements and the initial false statements from which they are derived conflicts with empirical necessity.
If a linguist claims that his mathematical design is a model of the grammar of a real language, it is not enough for him to show that grammatically correct sentences can be derived by applying formal rules to certain initial objects. He also bears the burden of proving that the initial objects are not fictitious, that logically necessary derivations in his formal system correspond to empirically necessary derivations in the real system of a real language.
A linguist who claims that his formal system corresponds to the real system of a real language can validate his claim by basing his argumentation on an analysis of semiotic properties of language.
Generativism is unacceptable as a methodological postulate because it confounds logical necessity with empirical necessity. Granted that generative-transformational grammar is able to generate only true linguistic objects of the surface structure of a language (actually, it is far from being able to do so), this fact in itself does not guarantee that this grammar is not fictitious. Generative-transformational grammar based on the fictitious notion of deep structure and fictitious phonological entities conflicts with the functional properties of language as a sign system.
Fictionalism and generativism are two sides of the same coin. There is nothing wrong with the mathematical notions of algorithm and generation; rules of algorithmic type, when properly applied to particular domains, can be an important mathematical aid in empirical research. But generativism is a different story. Generativism as a methodological postulate is an attempt to justify fictitious entities in linguistics by devising mechanistic rules that convert fictitious linguistic entities into observable linguistic objects. Inventing and manipulating mechanistic rules are the only way to justify fictitious entities, but all of that has nothing to do with explanation—rather, it all is akin to reflections in curved mirrors.
The only right alternative to generativism is the semiotic method with its concept of semiotic reality. The semiotic method does not reject mathematical notions of algorithm and generation as useful tools of linguistic research. Rather, it rejects generativism as a methodological postulate. The choice of a mathematical tool is not crucial; what is crucial is to associate our mathematical tool with a correct hypothesis about language, and that can be done only by applying the semiotic method.
We can construct different systems that will generate the same set of sentences. If so—which system is the right system?
There is no way to answer this question if the only thing we require of a generative model is that it generate correct sentences for a given language. The only way to solve our problem is to study the properties of language as a sign system. Then and only then will we be able to make a right choice between different ways of constructing sentences of a given language. The correct system of rules must respect linguistic stratification; it must respect functional properties of linguistic units; it must respect the distinction between synchrony and diachrony, and so on. In other words, we must use the semiotic method, which provides the necessary criteria for an evaluation of linguistic models.
Generativism is unacceptable because it aims at constructing arbitrary mechanistic rules that either distort linguistic reality or at best have zero explanatory value.
Generativism distorts linguistic reality in the following ways:
1) It confounds the phonological level with the morphological level.
2) As a result of 1), it rejects the phonological level.
3) It uses fictitious entities called deep structures and fictitious phonological representations.
4) It confounds the constituency relations with linear word order, which involves an inferior formalism that is inadequate for a study of linguistic relations and formulating linguistic universals.6
Chomsky writes a lot on the philosophy of language. A discussion of his philosophical views is outside the scope of this book. Suffice it to make a few comments on his claim that his linguistic theory relates to the ideas of Humboldt. He claims that his notion of competence as a system of generative processes is related to Humboldt’s concept of free creativity. Chomsky writes:
We thus make a fundamental distinction between competence (the speaker-hearer’s knowledge of his language) and performance (the actual use of language in concrete situations). . . . This distinction I am noting here is related to the langue-parole distinction of Saussure; but it is necessary to reject his concept of langue as merely a systematic inventory of items and to return rather to the Humboldtian conception of underlying competence as a system of generative processes. (Chomsky, 1965:4)
The works of Humboldt are not easy reading. But anyone who is familiar with these works and is able to understand them can see that Chomsky’s notion of competence as a system of generative processes has no relation whatsoever to Humboldt’s ideas.
The most often quoted passage from Humboldt’s main linguistic work contains a succinct characterization of the essence of language as he understood it:
In itself language is not work (ergon) but an activity (energeia). Its true definition may therefore be only genetic. It is after all the continual intellectual effort to make the articulated sound capable of expressing thought.* (Humboldt, 1971: 27)
Obviously Chomsky confounds his notion ‘generative’ with Humboldt’s notion ‘genetic’. If one does not base one’s speculations on the etymological affinity of the words generative and genetic but concentrates his intellectual powers on understanding what he reads, he will see that the above passage, in the context of the whole of Humboldt’s work, has the following meaning.
By stating that language is not work (ergon) but an activity (energeia), Humboldt really meant that language is a constantly changing instrument of expressing thought and that expressive activity is a constant struggle by the individual to adapt the meaning of linguistic form for the thought he wants to express.
Humboldt conceives of the word as a bilateral unit—a combination of sign and meaning. Here is what he writes about the notion of the word:
By the term ‘words’ we mean the signs of individual concepts. The syllable forms a sound unit, but it only becomes a word when there is some significance attached to it; this often requires a combination of several such units. Therefore, in the word two units, the sound and the idea, coalesce. Words thus become the true elements of speech; syllables lacking significance cannot be so designated.* (Humboldt, 1971: 49)
Humboldt’s conception of the word as a bilateral unit consisting of sign and meaning is in keeping with the conception of the word in modern semiotics and has nothing in common with Chomsky’s conception of autonomous grammar independent of meaning, with its unilateral linguistic units.
Humboldt’s conception of language as an activity (energeia) has nothing in common with the mechanistic rules of generative grammar. Rather, it is in keeping with the conception of language in modern semiotics as a dynamic conventionalized conceptual system that is in a state of constant flux as a result of the constant struggle of individuals to adapt linguistic form to the thoughts they want to express.
Generativism must be abandoned. By saying that generativism must be abandoned, I do not mean to say that linguistics must return to one of the earlier varieties of structuralism—we have to attain a higher level of comprehension by reconstructing the old concepts in the framework of new formal systems based on semiotics. That will be genuine progress, not a return to worn old concepts.7
______________
* Original text: “Sie [die Sprache] ist kein Werk (Ergon), sondern eine Thätigkeit (Energeia). Ihre wahre Definition kann daher nur eine genetische sein. Sie ist nämlich die sich ewig wiederholende Arbeit des Geistes, den articulirten Laut zum Ausdruck des Gedanken fähig zu machen” (Humboldt, 1836: 41).
* Original text: “Unter Wörten versteht man die Zeichen der einzelnen Begriffe. Die Silbe bildete ine Einheit des Lautes; sie wird aber erst zum Worte, wenn sie für sich Bedeutsamkeit erhält, wozu oft eine Verbindung mehrer gehiöirt. Es kommt daher in dem Worte allemal eine doppelte Einheit, des Lautes und des Begriffes, zusammen. Oadurch we rden die der Bedeutsamkeit ermagelenden Sylben nicht eigentlich so genannt werden können” (Humboldt, 1836: 74).