Chapter I. Semantics at the Cross-Roads

Just over eighty years ago, a new term was introduced into linguistic studies. In 1883 the French philologist Michel Breal published an article on what he called the “intellectual laws” of language. In this he argued that, alongside of phonetics and morphology, the study of the formal elements of human speech, there ought also to be a science of meaning, which he proposed to call la semantique, by a word derived from the Greek ‘sign’ (cf. semaphore).The branch of study advocated in this article was not entirely new; yet it was mainly Breal’s genera­tion, and in the first place Breal himself, who established semantics as a discipline in its own right. In 1897 he published his Essai de semantique which saw many subse­quent editions and is still widely read. [...] Three years after its publication, Breal’s Essai was translated into English under the title Semantics: Studies in the Science of Meaning, and although the term had been used in English a few years earlier, this translation played a decisive role in the diffusion of the new science and its name.

Semantics ought to set an example to other sciences in the avoidance of ambiguity, and it is somewhat paradoxi­cal that the term itself has become highly ambiguous in recent years. Since the 1920s the philosophers have ac­quired their own brand, or brands, of semantics, which have very little in common with the homonymous science practised by philologists. Philosophical semantics in the more esoteric sense of the term is a branch of the ‘theory of signs’, dealing with relations between signs and what they stand for. In its more popular sense, philosophical seman­tics is a study of the misuse of abstractions, and of other shortcomings of language. [...]

At the time when semantics appeared on the scene, the science of language was an exclusively historical discipline. Semantics wholeheartedly accepted this orientation, and for the first half-century of its existence it remained a purely historical study. Its prime purpose was the classifi­cation of changes of meaning according to logical, psychological or sociological criteria, and the discovery of any abiding tendencies — misleadingly called ‘semantic laws’ — which governed these changes. This phase found its crown­ing achievement in Gustaf Stern’s Meaning and Change of Meaning, with Special Reference to the English Language, which was published in 1931 and contained the first scheme of classification based on an extensive collection of con­crete data. Meanwhile, however, far-reaching changes had taken place in general linguistics, as a result of which semanticists were soon faced with a dilemma which re­mains unresolved to this very day.

In the early years of the present century, linguistics underwent what has been rightly described as a ‘Copernican revolution’. This revolution, which was ushered in by the posthumous publication, in 1916, of Ferdinand de Saussure’s Cours de linguistique generale, showed itself in two main ways. Firstly, the historical bias: of nineteenth century philology gave way to a broader view which admitted the existence of two approaches to language, one de­scriptive or ‘synchronic’, the other historical or ‘diachronic’, and boldly proclaimed the primacy of the descriptive meth­od because it is more akin to the attitude of the ordinary speaker. The second great change concerned the way in which the tasks of descriptive linguistics were conceived. Language came to be viewed, not as an aggregate of dis­crete elements but as an organized totality, a Gestalt, which has a pattern of its own and whose components are interdependent and derive their significance from the system as a whole. In Saussure’s famous simile, language is like a game of chess: you cannot add, remove or displace any element without affecting the entire field of force. In the United States, thinking about the fundamental structure of language developed on similar lines; the new approach was codified with remarkable precision and rigour in Leo­nard Bloomfield’s book Language (1933), which, second only to Saussure’s Cours, is easily the most influential work on linguistics published so far in this century.

The new conception of language, which has come to be known by the name of structuralism, has sometimes been carried to unreasonable lengths. To assert, as has been often done, that language is a system where everything hangs together — ‘un systeme ou tout se tient’ — is obvious­ly unrealistic. The late Professor Entwistle was nearer the mark when he wrote: ‘I do not find language either sys­tematic or wholly unsystematic, but impressed with pat­terns, generally incomplete, by our pattern-making minds.’ Be that as it may, the idea of an underlying pattern has proved an extremely fruitful working hypothesis. It has been applied with conspicuous success to the phonological side of language where it has yielded the invaluable con­cept of ‘phoneme’ or ‘distinctive sound’. From phonology, there has been a gradual shift of interest to morphological structure where the theory has produced another useful though more controversial concept: that of the ‘morpheme’ or ‘minimum significant unit of language’ — a diverse category which comprises simple words, prefixes and suffixes, inflexions, non-independent roots and other elements, in­cluding even the intonation of the sentence. The structur­alist theory has also made an impact on syntax where it has given rise, during the last few years, to an entirely new and promising technique known as ‘transformational grammar’. Even the reconstruction of extinct languages has benefited by the advent of the structural approach. Seman­tics, too, has felt the need to align itself with the rest of linguistics by adopting structuralist viewpoints, but these attempts have so far met with less success than in other branches of language study.

The reasons for this discrepancy are not far to seek: they lie in the very nature of the subject. Whereas the phonological and even the grammatical resources of a lan­guage are closely organized and limited in number, the vocabulary is a loose assemblage of a vast multitude of elements. The numerical contrast is striking: a recent author­ity states that there are forty-four or forty-five phonemes in English while on the other hand the Oxford Dictionary is said to contain over 400 000 words: a ratio of nearly 1 to 10 000! But there is an equally sharp contrast in cohesion and stability. The phohological and grammatical system, though subject to long-term changes, is relatively stable at a given moment, whereas the vocabulary is in a perpetual state of flux. New words are continuously formed or borrowed from outside sources to fill a genuine gap or to suit the whims of the speaker; new meanings are attached to old words... [...]

It is clear, then, that the vast, unstable and loosely organized congeries of words which we call vocabulary can­ not be analysed with the same scientific rigour and preci­sion as the phonological and grammatical system of a lan­guage. This does not mean, however, that words are not amenable to any kind of structural treatment. It is here rather than in other sectors of language that we are likely to find the incomplete patterns envisaged in Entwistle’s formula. In the search for such patterns, some linguists and other scholars have evolved, since the early thirties, a number of different techniques from which a new, structurally oriented semantics has begun to emerge. These experiments fall into three main groups: those which aim at a statistical analysis of word-frequency and other lexical features; those which seek to identify the characteristic tendencies of a language; lastly, those which are concerned with the way the vocabulary is built up, with the prin­ciples and the hierarchy of values which underlie its structure.

1. The statistical approach has become of late very pop­ular in linguistics, not only because of the precision and objectivity which it is held to guarantee, but also because language is a mass phenomenon par excellence, which seems to invite this kind of treatment. In semantics, one of the boldest attempts to introduce statistical methods was made by the late Q. K. Zipf in his book, Human Behaviour and the Principle of Least Effort (1949), and in other writings. Zipf’s analysts was based on the assumption that words work like ordinary tools and are subject to the same laws which govern the use of tools. The analogy between words and tools was by no means new, but Zipf pursued it to its ultimate implications. He argued that, just as there is a direct relationship between the frequency of the use of a tool and the diversity of the uses to which it is put, in the same way ‘we may expect to find a direct relation­ship between the number of different meanings of a word and its relative frequency of occurences’. He even found a mathematical formula for this relationship: with the pos­sible exception of the few dozen most frequent words of a language, ‘different meanings of a word will tend to be equal to the square root of its relative frequency’.

The ordinary linguist is here on unfamiliar ground; yet he cannot help feeling somewhat alarmed at the very pre­cision of the result. His experience with word-meanings has taught him that they are seldom as precise, as sharply delimited, as Zipf’s formula would suggest. He has come to view with some scepticism the dictionary method — inevitable but none the less misleading — of setting up the various meanings of a word as so many independent enti­ties; he knows that, more often than not, these meanings have no clear-cut demarcation lines but rather a kind of hazy fringe through which they imperceptibly merge into each other; they are, as Wittgenstein once put it, ‘concepts with blurred edges’. Zipf was no doubt on the right track when he discerned a correlation between word frequency and diversity of meaning; but he built on insecure foun­dations when he tried to give his findings, based on dic­tionary data, a mathematical formulation to which the material just does not lend itself.

... there are many ways in which statistical methods can be usefully applied to semantic problems. In recent years there have also been hopeful signs of cooperation between semantics and communication theory; a leading authority oh cybernetics actually went so far as to claim that there is no ‘fundamental opposition between the prob­lems of our engineers in measuring communication and the problems of our philologists’. At a practical level, electro­nic computers and other mechanical devices have already been applied with good effect to a wide variety of lexical data. At the same time, these new contacts, however val­uable, are bound to create serious problems of communication. The situation was neatly summed up a quarter of a century ago by one of the champions of mathematical linguistics: ‘...both philology and mathematics are essen­tially esoteric subjects, the latter more so than the former. This means that the mathematics will not be intelligible to someone who is not a mathematician and the philology will, at the best, be difficult for someone who is not a philologist’. Unfortunately, these difficulties are only too often disregarded by mathematicians and even by mathe­matically minded linguists. ‘Flaunting mathematical for­mulae before a linguistic audience or in a linguistic pub­lication’, writes Professor A. Martinet, ‘is either grossly misinterpreting the needs and capacities of one’s audience or readers, or else trying to bully them into accepting one’s views by claiming for these the support of a science they tend to respect as the most exact of all sciences, but whose data they are not in a position to verify. We need more and more rigour in linguistics, but our own brand’.

2. A second avenue of approach to structural seman­tics is through what is sometimes called idiomatology: the study of the unique and idiosyncratic structure of a language. As far back as 1921, one of the pioneers of structural linguistics, Edward Sapir, wrote these prophetic words:

“It must be obvious to anyone who has thought about the question at all or who has felt something of the spirit of a foreign language that there is such a thing as a basic plan, a certain cut, to each lan­guage. This type or plan or structural ‘genius’ of the language is something much more fundamental, much more pervasive, than any single feature, of it that we can mention, nor can we gain an adequate idea of its nature by a mere recital of the sundry facts that make up the grammar of a language.”

Sapir’s words refer in the first place to grammatical structure, but it seems intrinsically probable that the vo­cabulary too, if properly explored, will reveal some idiosyn­cratic tendencies, some characteristic preferences and aver­sions for certain modes of expression. These tendencies vary from language to language and may even change within the history of the same idiom. Some of them are capable of a strictly statistical formulation; others are less sharply defined, but stand out all the same very clearly. A good example in point is the ratio of transparent and opaque terms in a particular language. Greek philosophers were already divided into two camps in their views on the origin and nature of words: some regarded them as purely conventional symbols, while others believed in an intrinsic connexion between sound and meaning — a connexion which may have lost its transparency in the course of time. Modern linguists know that there are both transparent and opaque words in any language, and they seek to determine the factors which govern the dosage of the two elements. It has been found, for example, that English and French often have opaque, unanalysable names for objects and ideas which are denoted in German by transparent, self-explanatory terms. There are several symptoms of this relative opacity of English and French. A foreigner hear­ing the English word thimble or the French de for the first time will have to memorize them as they contain no clue to their meaning. But the corresponding German term, Fingerhut, is immediately understood if you know its two com­ponents, and it is also easy to remember because of the graphic metaphor which lies at its root: a hat put on a finger. The same may be said of English glove and French gant versus German Handschuh (‘hand’ + ‘shoe’), or of English skate and French patin versus German Schlittschuh (‘sledge’ + ‘shoe’). […]

German also possesses many derivatives, formed by means of prefixes and suffixes, where English and French have to use Greek or Latin formations. From the noun Stadt ‘town’, German can derive the adjective stadtisch, while English and French have the hybrid pairs town — urban, ville — urbain. [...] The prevalence of opaque terms in English and French throws an additional burden on the memory of foreigners learning these languages, though this will often be compensated by the international character of many English and French words which in German have more parochial equivalents. Among native speakers, the use of so many learned Greek and Latin terms tends to erect what has been called a ‘language bar’ between differ­ent sections of the population according to their educa­tional background, a problem which does not exist in Ger­many in such an acute form.

The contrast between transparent and opaque words is only one criterion in the study of semantic tendencies, there are other criteria which are perhaps of even greater importance: preference for particular or generic terms, the distribution and organization of synonymic resources, the various types of ambiguity and their remedies, etc. [...]

3. Attempts have also been made to identify and de­scribe the various lexical structures into which our words are organized. These inquiries, which are still at a tenta­tive stage, are being conducted at three super-imposed levels: that of single words, that of conceptual spheres, and that of the vocabulary as a whole.

(a)At the level of single words, the most useful con­cept that has emerged so far is that of the associative field. Every word is surrounded by a network of associa­tions which connect it with other terms related to it in form, in meaning, or in both; as Saussure graphically put it, it is ‘like the centre of a constellation, the point where an indefinite number of coordinated terms converge.’ To take a very simple example, the verb to write stands at the point of intersection of three associative series: (1) derivatives formed from the same stem: writing, writer, underwrite, writ, etc.; (2) words of similar or related meaning: scribble, scrabble, scrawl; letter, script, pen, print; read, say, speak, etc.; (3) homonymous words: wright, rite, right. In (1) the association, is based on both sound and sense; in (2) on sense alone, and in (3) on a chance iden­tity of sound.

The associative field of a word is an unstable and highly variable structure: it differs from one speaker to another, from one social group to another, and possibly even from one situation to another. It has been described as a ‘halo which surrounds the sign and whose outer fringes merge into their environment’. In spite of its vagueness and its lack of sharp contours, it is a linguistic reality which can be studied by psychological as well as philolog­ical methods. [...]

(b)Between the associative fields of single words and the vocabulary in its entirety, there is an intermediate level which has attracted much attention in recent years: that of conceptual spheres or lexical fields. The concept of lexical field first arose in the 1920s and was developed by Professor Jost Trier in his famous monograph on German terms for intellectual qualities. Trier’s view of language has been described as ‘neo-Humboldtian’, and it certainly has many affinities with the ideas of Humboldt and other German thinkers, in particular E. Cassirer; but it is essentially an application of Saussurean principles to prob­lems of lexical structure. Close study of the history of intellectual terminology in Old and Middle High German convinced Trier that it was fundamentally wrong to con­sider words in isolation: they must be viewed within the context of the lexical field to which they belong. A lexical field is a closely organized sector of the vocabulary, whose elements fit together and delimit each other like pieces in a mosaic. In each field some sphere of experience is ana­lysed, divided up and classified in a unique way. In this sense, the vocabulary of every language embodies a pecul­iar vision of the universe; it implies a definite philosophy of life and hierarchy of values which is handed down from one generation to another.

How differently the raw material of experience is elabo­rated by various languages can be seen even in such a pre-eminently concrete field as the scale of colours. The spec­trum is a continuous band, without any sharp boundaries; the number and nature of colour distinctions is therefore largely a matter of habit and convention. The Greeks and Romans had a poorer palette than our modern languages; there was, for example, no generic term for ‘brown’ or ‘grey’ in Latin: modern Romance forms like French brun and gris are borrowings from Germanic. There is no single word for ‘grey’ in modern Lithuanian either; different words are used to denote the grey colour of wool, of horses, cows or human hair. Colour terms employed in other languages will often appear more differentiated, or less differentiated, than our own, although it would be more correct to say that the field is divided up on different prin­ciples. Thus Russian distinguishes between sinij ‘dark blue’ and goluboj ‘sky-blue’; conversely, the Greek wordhas a wide range of applications, some with and some with­out a notion of colour: ‘gleaming, silvery; bluish-green, light blue, grey’. [...]

It might be argued that language may have an impor­tant part to play in the analysis of a continuum, but that fields of experience where there are discrete elements will everywhere have the same linguistic structure. A glance at kinship terms in various languages shows that this is not so. It is, for example, surprising to learn that Hungarian had no term for ‘brother’ or ‘sister’ till the middle of the nine­teenth century; it had instead, and still has, separate words for ‘elder’ and ‘younger brother’ and ‘elder’ and ‘younger sister’. Malay has again a different arrangement: it has a generic term for ‘sibling or cousin’ and more specialized ones for ‘elder’ and ‘younger sibling or cousin’, the latter being further subdivided into male and female. Other fam­ily relations show the same diversity. In Swedish there is no single word for ‘grandfather’ and ‘grandmother’, only separate ones for ‘father’s father’, farfar, ‘mother’s father’, morfar, and for the two kinds of grandmother: farmor and mormor. [...]

Starting from rather different premises, the French lin­guist G. Matore has evolved a field concept closely akin to Trier’s, but marked by a strong sociological bias. In his book, La Methode en lexicologie (1953), he has outlined a technique for studying the structure of the vocabulary as a reflection of the structure of society. His conceptual spheres are mapped out on the basis of sociological crite­ria, and organized around two types of important words: mots-temoins (‘witness-words’), which occupy a prominent place in the hierarchy, and mots-cles (‘key-words’), which epitomize the leading ideals of each generation. M. Matore has given an illustration of his procedure in his remarkable monograph on vocabulary and society in the age of Louis-Philippe.

(c)Some linguists believe that the structural approach, which has been tried out so successfully at the level of single words and conceptual spheres, can be extended to embrace the entire vocabulary of a language. To this end, R. Hallig and W. von Wartburg have devised a general classification of concepts which, in their view, is both broad and flexible enough to be applied to any idiom. In this scheme, which springs basically from the same idea as Roget’s Thesaurus, concepts are divided into three groups, each of them with numerous subdivisions: the Universe, Man, and Man and the Universe. The aim is to provide a uniform framework for lexicological studies of different languages and different periods of the same language, so that the results should be readily comparable with each other. Without claiming any special virtues for this scheme, which is only one of many possible arrangements, the adoption of a common framework offers obvious practical advantages, and a beginning has already been made in applying it in lexicological inquiries.

It can be seen from the foregoing that considerable pro­gress has been made, during the last three decades, in the introduction of structural view-points into semantics. It is indeed symptomatic of current interest in these problems that structural semantics appeared on the agenda of the last two international congresses of linguistics, held at Oslo in 1957 and at Cambridge, Massachusetts, in 1962. All this makes it surprising and regrettable that many structuralists should still feel disinclined to handle prob­lems of meaning. As a leading American linguist once put it, ‘for many linguistic students the word meaning itself has almost become anathema’. It has been suggested that language could be defined in such a way that seman­tic problems would lie outside the purview of linguistics proper. As Professor W. S. Allen pertinently remarked in his recent inaugural lecture at Cambridge: “Meaning, as at least one linguist has expressed it, has become a ‘dirty word’; but if the name tends to be avoided, there is no doubt that every linguist employs the concept, though some would be unwilling to admit to such improper thoughts.” At the London congress of linguistics in 1952, the term crypto-semantics was coined to describe this paradoxical attitude.

Some of this reluctance to deal with semantic problems undoubtedly started as a reaction against the indiscriminate use of the term meaning and other ‘mentalistic’ abstrac­tions; but this is surely no sufficient reason for excluding the semantic side of language from the field of linguistics. […]

Another reason for the avoidance of semantics by many linguists is the widespread belief that structural viewpoints are inapplicable to problems of meaning. In the light of recent developments in semantics, this position is, as we have seen, no longer tenable — unless, of course, the term structure is equated with ‘formal structure’, as is only too often the case.

But there is an even deeper cause for the structuralists’ refusal to tackle problems of meaning. For the reasons mentioned earlier on in this chapter, semantic phenomena cannot usually be described with the same scientific rigour as the formal elements of language, and to many linguists scientific rigour is the supreme test of scholarship, even where the subject-matter would invite a different method of approach. This attitude explains why, not so long ago, semantics was virtually ostracized by extreme structuralists. The last few years have witnessed a spectacular change of climate in linguistics, but one still has the impression that many structuralists are merely paying lip-service to a study which has become more respectable. If this formalistic bias were to be perpetuated, linguistics would develop into a strangely unbalanced discipline and would lose much of its humanistic content. It would become an esoteric study, unable to contribute to the solution of the great problems of our time, some of which are closely bound up with the nature of our words. In this sense, not only semantics but linguistics at large is at the cross-roads, and the direction it will take may determine its future for a long time to come.