Prague Structural Linguistics*

Trnka and others¹

The starting point of linguistic work as done by the Prague linguists is the assumption that its true objective is the analysis of speech utterances of all kinds, both spoken and written. This subject matter of linguistics, as any other raw material of physical, psychological and other phenomena, can be grasped and comprehended only in terms of the verifiable laws which govern them. In contradistinction to the nomothetic, mechanical operation of natural laws, the validity of linguistic laws is normothetic, restricted to definite periods of time and to definite bodies of utterances. If formulated and presented in grammar, they may exert a stabilizing influence upon the speech of the community by strengthening the stability and uniformity of its linguistic norms, as prescribed by the grammarians. The normothetic character of linguistic laws does not exclude, of course, the validity of some of them for several languages or even for all languages of the world in all their historically accessible periods of development. Despite many differences all languages must have some laws in common (e. g., the general law of minimal phonological contrast, according to which some consonantal clusters and vocalic combinations are excluded from all !anguages of the world), and one of the true goals of linguistics is to discover and formulate them.

Structuralism may be defined as the trend of linguistics which is concerned with analysing relationships between the segments of a language, conceived as a hierarchically arranged whole. The question may be asked whether the segments or the relationships are primary, but this problem cannot be solved at the present stage of our knowledge, at least not by linguists alone. It is clear, however, that both relators and relations are coexisting and correlated entities which cannot appear separately from each other. The relations of a segment are recognized by its properties, and as every segment is constituted by its properties, the elimination of these results in the elimination of the segment itself, not in the discovery of a “substratum” — a concept which modern science makes superfluous. In other words, the structural linguist conceives linguistic reality as a system of sign events, i. ?., as a system of linguistic correlates to extra-linguistic reality.

The word “structuralism” is used to designate various trends in modern linguistics which came into existence between both world wars, but, apart from the school of Geneva, those associated with the Cercle Linguistique de Prague, Cercle Linguistique de Copenhagen and the name of Leonard Bloomfield are regarded as the most typical. From the historical viewpoint, these three currents of structural linguistics have at least two features in common: divergence from the Neo- grammarian methods which tended to the psychologization and atomization of linguistic reality, and a tendency to establish linguistics, looked upon by the older school as a conglomerate of psychology, physiology, sociology and other disciplines, as an independent science based on the concept of linguistic sign. Otherwise they differ considerably from one another in their principles and procedures, and it is therefore advisable to use a special designation for each of them, viz. functional linguistics (V. Mathe- sius’s term) for the linguistic school of Prague, glossematics for Hjelmslevian linguistics, and descriptive linguistics for the Bloomfieldian trends. We cannot devote our attention here to other outstanding currents of structural linguistics deviating from the above-mentioned schools in different ways, and the following remarks will thus be mainly restricted to these three.

Hjelmslev’s glossematics introduces into linguistics the deductive method of the algebraic calculus and declares itself to be independent of any linguistic realitys His theory does not purport to be a system of hypotheses which would be found true or untrue by reference to factual linguistic evidence, but it tends to serve in analysing a given “text” by means of a strategy of assumptions which should be as few and as general as possible in order to fulfil conditions of applicability to the greatest number of linguistic data. Let us quote his own words: “This calculus, which is deduced from the established definition independently of all experience, provides the tools for describing or comprehending a given text and the language on which it is constructed. Linguistic theory cannot be verified (confirmed or invalidated) by reference to such existing texts and languages. It can be controlled on!y by tests to show whether the calculation is self-consistent and exhaustive”. It must be pointed out that the correctness of such a linguistic theory rests not only on the truth of calculation, but also on the veracity of the general linguistic assumptions which enter into the calculus. The latter in Hjelmslev’s calculus are, however, neither self-evident nor acceptable to all structural linguists (cf. his sharply-drawn dichotomy text — system, contents — expression, the forms of contents — the forms of expression; his view of sign as representing both “contents” and “expression”), and his logically consistent and conceptually well-arranged theory appears to be divorced from, and inadequate for, linguistic reality. The Prague linguists have not been influenced by Hjelmslev’s glosse- matics, and especially strong reservations are made by them concerning his theory of the phoneme as a mere “taxeme”, the identity of which is thought by Hjelmslev as consisting only in the identity of its distribution in words. Hjelmslev seems to regard both relevant (or, distinctive) and irrelevant features of a phoneme as a “substance, “ and while mixing the functional features of a phoneme with its functionless elements, he builds up an artificial wall between the sound and the phoneme. The Prague School takes into consideration not only the distribution of phonemes in words, but also their relevant features, the superposition of which determines the identity of a particular phoneme. Thus the English sounds gh and ρ in paper are phonetic implementations of the phoneme P, because both of them consist of the superposition of the same relevant (distinctive) features, i. e. bilabiality, plosion, oralness and voicelessness, aspiration being an irrelevant phenomenon determined mechanically by stress. On the other hand, Old Indian gh is a phoneme, not a positioned, variant of g, because its aspiration is not determined by its environment or its position. If two phonemes (e. g., h and r| in English) are in complementary distribution with each other and there is no environment in which both of them occur, their ever-present and potential capacity for differentiating words (cf. pen-fen-ten- den) remains unutilized, so that the only test of their phonemic status is their differences of make-up in terms of relevant features Strictly speaking, the true function of phonemes is not keeping the meanings of words (or morphemes) from each other (cf. [nait] night — knight, [frendz] friend’s — friends), but only distinguishing phonemes between each other.

Various trends of descriptive linguistics appear to be in agreement with the linguistic theory of the Prague group on some points, but owing to their behavioristic basis they differ from the latter both in terminology, definitions and details of the procedure of linguistic analysis. Both schools agree in rejecting the psychological approach to phonology and in regarding the phoneme as a unit of the phonological plane of language, but while the Prague group stresses the analysis of the phoneme into the relevant features which constitute it, the Bloomfieldians (Bloch, Trager, Harris, Hockett) seem to lay the emphasis on its distributional features in words or in utterances. As to the semantic criteria, which most American linguists exclude from their definitions of the phoneme, Prague phonologists lay stress on the capacity of phonemes to distinguish words and morphemes (cf. to ска — docka, burit’ : buril) and consequently on the morpheme and word boundaries as important factors, the disregard of which might lead to unwarrantable conclusions, e. g., in the phonological evaluation of affricates, in the formulations of neutralization rules and in ascertaining phonologica! foreignisms adopted by a given language. On the other hand, Prague linguists are aware of the fact that phonemes as such have no meaning — the word hand is something different from the mere sequence of the phonemes h, ae, n, d — and that their capacity of differentiating words need not actually be utilized in a linguistic system (cf. the zero functional load of the opposition h — ? in English cited above). Nor are phonemes the only exponents of the word and morphemic dia- crisis: homonyms, which may be defined as two (or more) words possessing identical phonemic implementation, are distinguished from each other only in that their respective meanings have no semantic elements in common. It is clear that the phoneme, as conceived by the Prague School, is an abstraction, but such a one as is made, consciously or unconsciously, by all speakers of a language and as necesasry for description of the sound level of a language as the concepts of “morpheme”, “word” and “sentence” are for that of its higher levels. Only this abstraction allows us to reduce a great number of sounds, acoustically or (and) articu- latorily different, to a comparatively small number of units capable of being represented by letters.

There are some other divergences between the linguistic theory of the Prague School on the one hand and that of the American schools of descriptive linguistics on the other. In fact, there are some minor divergences even between the views of the individual adherents of Prague functionalism. Linguistic theory is not viewed by Prague structuralists as an a-priori discipline independent of all experience, but as a theoretical framework derived from concrete linguistic materials and liable to verification, development and improvement by the use of further material and by further research work. Another characteristic of the Prague School, its structural conception of linguistic historical development, may be mentioned. At the very outset of their activity, Prague linguists emphasized the view contrary to the Saus surean theory that language is a system where “tout se tient” not only from the synchronic viewpoint, but also in its diachronic aspect. In fact, at any stage of its historical development language never ceases to be a system, as Time constantly affects all its levels and a!l its components.

In the recent history of linguistics the structuralist movement represents a reaction to the atomism of the Neogrammarians, and a tendency to grasp linguistic reality without any contradiction, more fully and more simply than current methods permitted. Some elements of linguistic structuralism can be found much earlier, in the grammar of Panini, in some works of Italian and other grammarians of the Renaissance, and later on in the writings of some modern linguists, but no consistent theory of structural analysis had been undertaken before Ferdinand de Saussure and N. Troubetzkoy. A growing opposition to the neogrammarian principles of linguistic analysis, was of course, manifest as early as the 80’s of the 19th c., and took on different forms in different countries. Thus some germs of structuralism had developed in some countries independently from each other. Some linguists (as Baudouin de Courtenay, Scerba, and others) sought to deepen the mechanistic conception of linguistic development by emphasizing the psychological correlates of linguistic phenomena, others (as Winteler, Sweet, Jespersen) accentuated differentiation between the functionally relevant and functionally irrelevant features of speech, or insisted on the fact of interrelationship of all components of language. The methods of the so-called relative chronology, as developed by Meyer-Lübke and Karl Luick in Austria, M. Krepinsky in Czechoslovakia and Sachmatov in Russia in their respective fields of philology, also contributed — partly at least — to this conception. It must be pointed out that the rise of linguistic structuralism was encouraged and influenced to some extent by similar trends, taking place at the same time in logic, mathematics, psychology, ethnology, natural sciences and philosophy. Its growth and development, however, is chiefly due to the progress of linguistic thought itself. As regards Prague structuralism, its development was prepared partly by Josef Zubaty (1855-1932), whose anti-mechanistic views were shared by his younger students (B. Havránek and J. M. Kofinek, among others), partly by V i 1 é m Mathe sius (1882 -1945), and his pupils (?. Trnka and J. Vachek), who were interested in establishing new or more precise methods in both synchronic and diachronic linguistics. The Russian members of the Cercle Linguistique de Prague (R. Jakobson, N. S. Troubetzkoy and S. Karcevskij) were influenced by Sachmatov and Scerba, as well as by F. de Saussure, while their study of ethnology and of some Caucasian languages enlarged their scope of linguistic vision and strengthened them in their efforts to find a more adequate approach to linguistic materials than the traditional one, used in Indo-European comparative philology. It must be mentioned here that the extensive study of Indian languages and folklore in the United States of America was also instrumental in bringing fresh ideas to linguistics and in contributing a good deal to the development of American structuralism. As to the further progress of structural linguistics, the Prague School expects the best results of this study, provided that it will never lose sight of its true purpose to analyse linguistic reality without imposing any preconceived limitations on its materials and without excluding a study of linguistic correlates to cultural and other extra-linguis- tic realities. It is understandable, of course, that owing to the present state of structural linguistics and to the immense tasks with which it is faced, no complete structural analysis of any language has so far been achieved by any linguist. The relatively best descriptions of a number of languages have been accomplished in the field of phonology.

Whereas the phonological level of a language is based on the phoneme and its relevant constituents, the morphological level is constituted by the word and the morpheme. The word may be defined as a minimal meaningful unit implemented by phonemes and capable of transposition in sentences. The morphemes are minimal meaningful segments of a word (e. g. hand-s, hand-y, might-i-er; ruk-á, rue-n-¿j, in Russian). Leaving aside the formation of words, the ro!e of structural morphology is (1) to state morphological oppositions (e. g., that of the number of substantives, that of common case — adnominal case, etc., in English) and their neutralizations (e. g., the neutralization of the opposition genitive — accusative of masculines denoting living beings in Russian and Slovak, the neutralization of grammatical genders in plural in German, the neutralization of number in the third person in Lithuanian verbs, etc.), (2) to state the phonemic means (often homonymous) implementing the morphological oppositions of a language, such as prefixes, suffixes and alternations of phonemes (so-called morphonemics). The Prague functionalists underline the necessity of describing the morphological structure, as any other, of a given language as it really is, without forcing it into the traditional categories of Latin grammar. A common basis for the morphological, phonological and syntactical comparison of languages under study cannot be created by a mechanical application of the grammatical rules of any language or any state of the same language, but only by adopting the same principles as tools for functional analysis. Thus the so-called “parts of speech” (partes orationis) of Latin grammar cannot be viewed as a-priori classes of words which must have their parallels in every language, but as word classes characterized, in a given language, by taking part in different bundles of morphological oppositions. E. g., the class of substantives in Russian includes all words that are able to participate (1) in the opposition of cases, (2) in the opposition of number (singular — plural), and (3) in the oppositions of gender, irrespective of whether they are used to denote a “substance” (stol, voda), quality (len’, zio, dobro ta) or action (bd’en’e, chod, choZd’en’ije). It is clear that the number of interrelational functions of parts of speech varies considerably from one language to another, as the classification of words into classes rests upon differences in the bundles of morphological oppositions in which they take part. Other criteria, either purely phonemic or purely semantic, cannot be applied to a linguistic analysis of “partes orationis”. Another point of deviation of Prague functionalism from the grammatical tradition is its conception of the function of morphological analogy. Whereas the Neogrammarians looked upon analogy as a disturbing factor in the phonetic development of language, the Prague functionalists regard it as a realizer of morphological oppositions which can never suspend the validity of the phonemic laws of any stage of the historical development of a language. There is no denying the fact that the phonological plane cooperates with other planes of the linguistic system in order to achieve its purpose, namely inter subjective understanding, but the roles aUotted to it are accomplished on the basis of its own specific laws. In consequence, neither new phonemes nor new combinations of phonemes can ever come into existence by the agency of morphological analogy in the historical development of a language.

As to the relationship of morphology to syntax, both discipline are viewed — apart from Hjelmslev’s theory — either as “para- digmatics” and “syntagmatics”, respectively (by the followers of the Genevan School, Karcevskij, Bröndal, and Gardiner), or as two different levels of structural analysis (by American and Prague structuralists). According to the latter morphology is concerned with the analysis of the word, whereas syntax is mainly the analysis of the sentence into its constituent relationships (e. g., subject — predicate, etc.). In contradistinction to the views of the former linguists, the Prague School holds that morphology and syntax cannot be linguistically contrasted to each other as two disciplines concerned with “parole” and “langue” respectively, because even syntax deals not only with “parole”, but also with “langue”, in attempting to discover normothetic laws,whose individual actualizations take place in utterances. In consequence, the line of division between morphology and syntax cannot be drawn as between “paradigmatics” (conceived as belonging to “langue”) and “syntagmatics” (as belonging to “parole”), and must be sought in the units in which they are concerned, i. ?., the word and the sentence. In fact, morphology and syntax represent two different levels of grammatical abstraction necessary for the analysis of linguistic material. The segmentation of a sentence (e. g., father is ill) never results in the elements of the morphological level (i. ?., in the words father + is + ill), and vice versa; the summation of morphological units can never make up a sentence, because the latter is fundamentally something different from the total of the isolated words or groups of words by which it is implemented. Of course, both morphology and syntax, as well as phonology, include both “paradigmatics” and “syntagmatics in correspondence to the logical relationships “either — or” and “both — and”. Thus phonology treats not only of the inventory of phonemes and their relevant features (phonological paradigmatics), but also of their combinations in words and sentences (phonological syntagmatics) without mixing its own units with those of the higher ranks. In a parallel way, morphology deads both with words (morphological paradigmatics) and the combination of words in sentences (morphological syntagmatics) without infringing on the domain of syntax and its higher units. The chief business of the latter is, on the one hand, the analysis of sentences into the constituent syntactic oppositions, as well as the implementation of these by morphological oppositions (syntactic paradigmatics), and the analysis of the combination of sentences (syntactic syntagmatics) on the other. The doctrine of “immediate constituents”, derived by some American linguists from de Saussure’s conception of “syntagma”, is not shared by the Prague School, since it may lead to a mechanistic analysis of speech units disregarding the basic distinction between the morphological units on the one hand and the syntactial units on the other. In fact, the Saussurean conception of “syntagma” has never become a starting-point for any scientific enquiry of the Prague linguists, if we leave aside the studies of Karcevskij, who was associated with the school of Geneva rather than with that of Prague. As another point of the Genevan “syntagmatics” questioned by the Prague School, may be mentioned the reduction of all speech elements, including the relation subject — predicate, to the relation determinant — determined.

It is clear from the foregoing exposition that the sharp Saussurean dichotomy langue — parole is no longer held to be a realistic basis of linguistic investigation by the Prague School. What F. de Saussure describes as “parole” is regarded by the Prague linguists as utterances (or parts of utterances), in which a code of inherent structural rules is to be detected.

The lexical structure of a language is its legitimate and very important part. Its connexion with morphology and syntax is clear from the fact that every morpheme (i. ?., every minimal significative part of a sentence) must needs have a meaning in order to be identified as a morpheme and that this meaning acquires many shades and variations in combination with other words. At the present stage of linguistics the lexicon of a language is merely a more or less exhaustive and alphabetically arranged store of phonological, morphological, syntactic and stylistic facts, explained by means of approximate definitions and (or) approximate equivalents of the same or some other language. An investigation of the structural features of meaning is one of the most urgent tasks of modern linguistics.

By the application of structured, principles the present-day historical grammar of a language may be converted into a description of the whole linguistic system, including features which, from the viewpoint of the Neogrammarians, seem to have been unaffected by any changes. If reconstructing any particular feature of a language, the investigator must consider the whole linguistic system and seek to synchronize its other components as far as they are ascertainable by the diachronic comparison of the written evidence. In this way he may arrive at more realistic results and at a more critical view of its possibilities than previous research workers. Both synchronic and diachronic linguistics have the final aim in common: the discovery of linguistic laws. The formulation of diachronic laws differs from that of synchronic only by stating the period of time of their validity in terms of relative chronology. E. g., the validity of Verner’s law — the neutralization of voice in the spirants _f, θ, x and the rise of the variant z, if not preceded by stress — extended from the period of time in which p, t, ? changed into f, θ, x to the dephono- logization of Indo-European stress, in Old Germanic.

As to the study of the relationship between the relevant features of language and those of society, it must be admitted that all schools of structural linguistics — with the exception of that of Edward Sapir and his followers — have failed to develop it in a satisfactory way. The Copenhagen School has been prevented by its own particular theory of language from the systematic investigation of the problems involved, and the American descriptive linguists, who often refer to Sapir as to their precursor, neither advanced to grasp the whole body of problems nor developed adequate methods for their solution, owing to non-historical trends in their linguistic thinking. As for the Cercle Linguistique de Prague, a penchant for immanentism was shared by some of its members and manifested itself in their stressing the basically correct view of the therapeutic character of many phonological and other changes. Other members of the CLP (V. Mathesius and others), however, emphasized the functional role of language as a system serving to satisfy the communicative and expressive needs of the community and liable to changes in order to meet new needs. The reflection of social factors in language has been never denied by the Prague linguists (cf. Havránek’s theory of development of the Czech literary language), although the chief interest of the Cercle was devoted to the elaboration of phonology and morphology. In the words of K. Horálek, the deficiency of the scientific enquiries of some Prague structuralists was that they were not structural enough. It must be pointed out, however, that every level of language is not affected by features of extra- linguistic structures in the same measure. Whereas the strongest influence of these factors is exerted on the lexical plane of language, it is less manifest in syntax and morphology (cf. the ousting of ME hie, hi_ by Scand. theirr; thou by you) and much less still in phonology and phonetics. The function of this plane is almost restricted to providing the linguistic community with a number of mutually distinct and combinative units (i.e., phonemes) as implements of the higher planes of language, so that its correspondence with the social needs of the community is only indirect. This subservient, though very important, role of the phonological plane is relatively constant for all linguistic systems and helps us to account for the fact that most results of the research work of the Cercle Linguistique de Prague still remain useful contributions to structural linguistics. It would be a mistake, of course, to believe that the sound level of the language has absolutely nothing to do with the social history of the speech community. Interesting examples of the influence of extra-linguistic factors on the sound level of languages may be found, e. g., in phonological changes which were suggested or at least promoted by a widespread use of loan-words containing phonological features foreign to the language of adoption. Thus the phonologization of voice distinction between _f and v, s and ζ, θ and S, which took place in Middle Engiish in consequence of the loss of final -e, the change of χ to f, and the dephonologization of consonantal quantity, was doubtless facilitated by such words as basin, mason, adopted from French, in which both voiced and voiceless spirants f and v, £ and ζ were phonemes.

It may be seen from the above remarks that a study of culture- language correlation is one of the major tasks of linguistics and that structural methods should be extended to the problems involved, as far as the character of a particular linguistic level admits. It is especially necessary to substitute loose formulations relating to the subject by precise structural laws. This task cannot be carried through to completion, of course, without a deeper analysis of the semantic aspects of language.

A complete knowledge of linguistic reality cannot be achieved without combining the qualitative analysis of its constituents with that of their quantitative relationships. E. g., a statement as to the various formations of the plural in English would be incomplete and inadequate if it were not complemented by data relating to their productivity and frequency.

It is important to state that the quantitative approach to lin- quistic materials has no useful purpose if the mathematically minded research worker does not set out to solve a concrete linguistic problem or if he does not furnish at least any raw material for further qualitative evaluation. It is used in linguisitcs, as in other social sciences, for analysing a complicated and heterogeneous reality or for verifying the results of an analysis already obtained. The heuristic value of the quantitative analysis of a linguistic feature in terms of its productivity, contextual frequency and periodicity consists in detecting a contradiction between its anticipated quantitative relationships and the quantitative results actually arrived at. E. g., it has been ascertained that voiceless “paired” consonants jd, ?, t], s, £ in Czech and Slovak are more productive (functionally loaded) and more frequent than their voiced counterparts b, d, d’, z_, z, whereas the voiceless back fricative ch [x] is less productive and less frequent than the voiced glottal spirant h, which is in the same phonological relationship with ch as s is with z. This striking exception to the rule of quantitative relationship between voiced and voiceless consonants in both languages is to be explained in terms of qualitative linguistics. It is doubtless in connection with the fact that [x] owes its phonemic status to the phonologization of a positional variant of s in Primitive Slovonic and does not belong to the older stock of the phonemes p/b, t/d and s. Since it formed voice correlation with h after the change of Old Czech g to h, i. ?., as late as 11-12th century, its peculiar quantitative feature in Present- day Czech or Slovak must be interpreted as an interesting reflection of its earlier uncorrelated character. — The field of quantitative linguistics is very extensive. So-called linguistic typology may be regarded as one of its branches, and likewise linguistic characterology, which V. Mathesius sought to establish on the basis of the comparison of structural features between a number of European languages, cannot do without an appeal to statistics.

Notes

1. The present paper is an English adaptation of the Russian paper published in Vojarosy jazkoznanMja, No, 3, 1957, pp. 44-52. It represents the view of a group of Prague linguists associated in a special section of the Krüh moderních filologu (?. Trnka, J. Vachek, P. Trost, S. Lyër, V. Polák, O. Duchá- бек, J. Krámsky, J. Nosek, M. Rensky, V. Hofejãí, ?. Wittoch, and L. Duãková.

*From Philologica Pragensia (Prague), I: 33-40 (1958).