Linguistics, Literary Analysis, and Literary Translation 9781487583408

In this interdisciplinary study Henry Schogt explores the relations between linguistics, literary analysis, and literary

284 104 18MB

English Pages 192 [189] Year 1988

Report DMCA / Copyright


Polecaj historie

Linguistics, Literary Analysis, and Literary Translation

Citation preview



Linguistics, Literary Analysis, and Literary Translation


© University of Toronto Press 1988

Toronto Buffalo London Printed in Canada Reprinted in 2018 ISBN 978-1-4875-8210-4 (paper)

Canadian Cataloguing in Publication Data Schogt, Henry G., 1927Linguistics, literary analysis, and literary translation Includes bibliographical references and index. ISBN 0-8020-2649-4 1. Translating and interpreting. 2. Linguistics. 3. Literature - History and criticism. I. Title.

PJ06.2.s36 1988

418' .02

For my grandson, Alexander





Introduction 3


System, norm, usage 7

3 Literature, norm, usage, and evolution 15 4 Cost and yield


5 Formal identity and formal opposition: the linguistic sign 29 6 Field models, universals, and language-bound world-view 35 7 Arbitrariness, convention, and motivation 41 8 A stumbling-block in semantics: discreteness and gradual transition 46 9 Intentionality and relevance 52 10

Terminological confusion 59


Short of denotation and beyond it 65


What's in a name?


13 Linguistics and literary analysis: a happy alliance? 82

viii Contents 14 Linguists and literary texts 89 15

Linguistics and translation 96

16 Various options 103 17 Foreign languages and dialects 18


Answerable and unanswerable questions NOTES 125 REFERENCES








The study of the connections between linguistics, literary analysis, and literary translation may interest not only those who specialize in translation, but also linguists and literary analysts who do not deal directly with translating. Thus, a problem arose in writing this book of how to address an audience composed of experts in three different fields. Moreover, the fact that students who do not consider themselves experts in any of those fields might want to get some insight into the interdisciplinary questions that are discussed further complicated the writing of the text as did the need to maintain a balance between general accessibility and specialization. There are always many ways to read a book, but I deliberately conceived this work so as to accommodate different groups of readers. 1 First, the text can be read as a general initiation, without paying any attention to notes or cross-references. An effort has been made not to present material in any chapter that presupposes that the reader is familiar with data provided in the notes to previous chapters. 2 Rather than seeing each chapter as isolated and self-contained, one can check the cross-references, which help in the discovery of the implications for other areas of investigation of a theory or construct formulated in connection with a specific problem. Sometimes these implications are quite obvious, but in other instances they are not. 3 For those who want to go a step farther, notes are provided in a separate section of the book. Some notes are of a bibliographical nature; others provide the original text of English translations, or indicate the passage on which a remark in the main text is based. Furthermore, some contain commentaries and explanations introducing other points of view than those discussed in the text. Not all commentaries will be relevant for all readers, and each reader can make his 1 own selection.

x Preface 4 Finally, the reference list contains many titles that can serve as a starting-point for further investigations in a specific area. As only those books and articles that are mentioned in the text or in the notes are listed, it will not be difficult to find those starting-points, even though the references are arranged in alphabetical order and not according to subject-matter. Readers looking for a theoretical position will have considerable difficulty finding 'where the author stands.' They should not forget, however, that the main purpose of this book is not to give an exhaustive description of a linguistic or a literary model for text analysis, but to investigate what the implications of certain features of different models are for the translator of literary works. That some of the conclusions one arrives at for translation raise some doubts about the general validity of some theoretical positions and underline their reductive character could be seen as a rejection of theoretical models. Such an interpretation goes too far, however. In my opinion each model can make important contributions and shed new light on some aspect of linguistic or literary analysis, provided it is presented in comprehensible language or in an accessible formula. However, very often, if not always, the advantage of a theory in one area is counterbalanced by a disadvantage, be it negligence or fuzziness, in another. So, one could say that the absence of all-encompassing models excluding all other approaches represents in itself a theoretical stand, as it reflects my opinion that no theory is good enough to be pre-eminent over all other theories dealing with the same problems. Tolerance for ideas that do not coincide with one's own preferences, and a continuous effort to be as clear as possible, seem to me the most important qualities needed for a fruitful discussion. Many students, colleagues, and friends contributed to this book; I greatly benefited from working with them, exchanging ideas and asking for their opinions. It is impossible to thank them all personally, and I hope that if they read the book they will recognize discussions we had together, some recently, some many years ago. I am very grateful for their part in my work. I would like to mention the special inspiration that the collaboration with Henri Mitterand, with whom I taught a graduate course in Toronto in the autumn of 1982, gave me. Without that experience of combining a syntactic and a literary analysis of texts from nineteenth- and twentiethcentury French literature, I would probably have stayed within the boundaries of traditional linguistics. I am indebted to Larry Kerslake for

xi Preface encouraging me to give a lecture on the subject of linguistics and literature in November 1983. That lecture was the beginning of this book. However, without the help of one of my students, Kristin Collins, who compiled a great amount of bibliographical data, and especially without the help of Ed Burstynsky, who not only read the text and gave me advice on many questions, but also put everything I wrote immediately on a word-processor to facilitate reviewing what was written and making changes where necessary, I would never have been able to finish my work. As most superlatives have lost much of their strength, having been used too often and too carelessly, I do not know how to express my thanks for what Ed has done for me. I must therefore leave this linguistic problem unsolved. My family - my wife, Corrie; my daughter, Elida; my son, Philibert; my grandson, Alexander; and his parents, Barbara and Coen - created with their encouragement and their warmth the best possible conditions for my work. The University of Toronto Press made every step on the road to publication as easy as possible, and I am grateful to say that this book has been published with the help of grants from the Canadian Federation for the Humanities, using funds provided by the Social Sciences and Humanities Research Council of Canada, and from the University of Toronto Women's Association.



Quite often a translator turns for help to linguistics in order to solve a specific problem he has encountered in his work, frequently to be disappointed as he does not receive the help he expected. Instead of providing a nice, elegant solution to his particular translation problem, the linguist he consulted gives him reasons why there is no solution or offers him a choice of approximate equivalents that he could have found without any outside help. Theories of translation include many elements of a linguistic character, however, and linguistic theories are part and parcel of translation handbooks. The contributions of linguistics to literary analysis and literary theory are less visible, less numerous, and more controversial. In contemporary literary theory, linguistic terminology abounds, and structuralism and post-structuralism are closely connected to linguistic models. Whether this state of affairs reflects a real interdependence or is only the result of superficial borrowing remains a moot point. The structuralist approach is not restricted to literary analysis, but is used in other fields as well, especially in anthropology and sociology. The same term applied to different fields runs the risk of losing its precise meaning and of becoming too vague and too general, unless in each case a new definition is given. Thus, it is necessary to examine the use of any linguistic term in another discipline very closely in order to determine whether it retains its original value or the value shifts so that there is less common ground than the identical nomenclature suggests. Apart from providing ideas and techniques for translation and literary analysis, linguistics may be at the receiving end. Then the findings of translation theory or of literary text analysis are used to elaborate new linguistic theories or set up new models that incorporate the data yielded

4 Linguistics, literary analysis, and literary translation by literary texts. Grammar and syntax are often based on a body of literary texts from which examples are quoted to corroborate or illustrate the point that is under discussion. However, literature recedes into the background, if it does not disappear entirely, as soon as one leaves the domain of the description of a particular language in order to deal with general linguistic models that claim to be applicable to any language. Before examining the complex interrelationships between linguistics, literary analysis, and translation theory, it will be useful to make a few remarks about the aims of the three disciplines and to draw attention to some of their most characteristic features. It is obvious that it is impossible to cover all the aspects of each of the three fields and that my selection is the result of a personal preference that not everyone will share. 1. Linguistics deals with the function and structure of human language. In order to discover the principles underlying language as a system and to explore the ways in which speakers and possibly writers use the system that is at their disposal, a linguist may start from any language. However, he has to prove the validity of his findings for all human languages that ever existed, exist now, or will exist in the future, before he can make the claim of generality. Except for some basic features that are part of the definition of language, it is very difficult, if not impossible, to find what are called 'absolute universals' in language structure. So, linguists also look for general trends and relative universals and make predictions such as: if L is a natural language, the feature A is more likely than not to occur; or if L has feature A, it is likely to have feature B as well, whereas feature B does not allow one to make any assumption about the presence or absence of feature A. In this way one is able to find for each individual language a unique set of features that distinguish it from all other languages. As more or fewer traits are shared with other particular languages, it is possible to set up language families, not according to common linguistic ancestry (the historical approach), but based on common contemporary features (the typological approach). Besides the universal and the typological approaches there is the possibility of analysing a language without linking it in any fashion to other languages. The language is studied as an autonomous, immanent whole, the elements of which are structured in a certain way. Some descriptions focus entirely on system and norm at a given moment (the synchronic view), others take into account tensions and tendencies as well (the dynamic-synchronic view), whereas a third variety follows the evolution of the language over a certain period of time (the diachronic view).

5 Introduction Notwithstanding the considerable differences in approach and the great variety of aims that characterize their work, linguists' have one thing in common, namely their desire to give a description that is representative of the language spoken by a certain speech community at a given moment or during a certain period of time or one that has general validity without temporal or spatial restrictions. There is always an element of generalization and the possibilty of extrapolation in the work of linguists. 2. Translation theory deals with the problem of how to arrive at the equivalence of two texts: the source text and its translation, the target text. A good starting-point is the comparison of the two languages involved from a structural perspective; but that is only a beginning. It is also necessary to analyse the various components that contribute to forming the message. In some texts form is completely subordinated to content. In others, form takes on a much more important role. So, it is impossible to establish a hierarchical order of importance for the components of a text without knowing a good deal more about it than its linguistic form alone. Semantics plays a prominent part in translation theory, but as semantics operates at the point where many disciplines meet with linguistics, one can't be sure whether the implication is that linguistics is equally prominent. One of the major difficulties of a linguistic theory of literary translation, then, lies in the clash between the tendency of linguistics to stress what is general and can be extrapolated, and the character of the material that is to be or has been translated, i.e., individual literary texts, each of which has its own identity even though one may distinguish different genres and establish a list of features that all works belonging to a certain genre have in common. 3. Literary analysis and text interpretation were traditionally in the realm of aesthetic appreciation and had an overtly subjective character. However, the study of literature has become gradually more scientific and less impressionistic in its aims. Not only are many terms that originate from linguistic analysis and are borrowed from structuralism now frequently used in work dealing with literature and descriptive models, but also quantitative analyses are carried out, conveying further scientific prestige to what was once, in essence, a matter of personal taste. Very often, though, linguistic data have only an ancillary function and are overshadowed by other factors such as plot and narration. The branch of comparative literature illustrates this very clearly: works may be compared regardless of the language they have been written in, and the use of translations is standard procedure and quite acceptable.

6 Linguistics, literary analysis, and literary translation So, the basic questions remain. Is there only a superficial connection between linguistics and literary analysis, or are there deeper ties? If the latter, what are these ties and is the relationship reciprocal or is one of the disciplines clearly the beneficiary whereas the other only provides ideas - concepts and models - but does not itself benefit in any way from the contact? Because literary translation and translation theory seem to be about half-way between linguistics and literary analysis it may be useful to investigate whether the study of translation can shed any light on the questions noted above. Even though clear-cut, definitive answers may be out of reach, one can try to gain insight into the links among the three fields mentioned in this introduction and thus enable oneself to distinguish between common ground and what could be pseudo-common ground where identical terminology does not correspond to identical concepts. Some of the chapters in this book will seem more relevant than others, and readers will have different orders of preference; it is hoped that in the end the justification for inclusion of whatever chapter is deemed to be least important will have become evident to each reader. More serious is the fact that not everything that has some relevance for the problems that will be discussed has been included. In order not to become a victim of the snowball effect in a field of research where everything is interrelated in some way, I had to draw the line somewhere.


System, norm, usage

Ever since Ferdinand de Saussure's Cours de linguistique generale1 was published in 1916, linguists have been trying to sort out the relationship between what Saussure called la langue, the underlying speech system shared by speakers of the same language, and la parole, the individual written or oral language utterances that are based upon that underlying system. Although not entirely identical, Noam Chomsky's dichotomy competence versus performance2 comes close to the langue!parole opposition, as it also distinguishes between an underlying virtual stratum and a level of actual, individual manifestation. Inasmuch as one can only reach the underlying stratum by observing a cross-section of individual manifestations, it is immediately clear that it is easier to accept the competence level as a theoretical construct than to indicate for a given language what belongs to that level and what does not, being an individual deviation. Even the just-mentioned crosssection raises some problems, for one has to decide to what extent geographically and socially marked samples should or should not be included. 3 There are two kinds of language descriptions: one that gives a picture of the language in question as it should be spoken and written, and one that, instead of indicating how things should be, tells us how things really are. The first kind of description is very well known and has a venerable tradition. For a long time language studies were concentrated on Latin and Greek and as the classical period of each of those languages provided at the same time the material for a description and the ideal model, description and prescription were almost identical, if they did not coincide completely. As soon as inscriptions and graffiti are taken into account, the problems of social stratification, norm, and usage come

8 Linguistics, literary analysis, and literary translation inevitably to the fore . When prescriptive, normative grammar offers an ideal form of language that is not used in everyday life, that everyday language may become an object of linguistic study in its own right. The aim of studies dealing with language in its actual use may be purely descriptive, but often other considerations introduce a special angle to the description, stressing certain data that are of particular interest. Special attention is given in some descriptions to the internal tensions, and to simultaneous use of different forms, part of which relieve or avoid those tensions. Instead of giving an uninterpreted picture of the state of the language at a given moment, and being strictly synchronic, this kind of study indicates imbalances and possible trends and has a dynamic-synchronic approach. The following examples will clarify the notions of system, norm, and usage and the tensions that may arise when they clash or when there are conflicts between the requirements as expressed in terms of phonetics, morphology, and syntax. Phonetic change has been studied intensively from the time, about a century ago, that the Neogrammarians4 formulated their ideas about the Ausnahmlosigkeit der Lautgesetzen ('the exceptionlessness of phonetic laws'). From the rather fragmented approach that still characterized historical phonetics in the period between the two world wars, to the wholly integrated views as expressed in studies on sound change by functionalist-structuralist scholars such as Andre Martinet,5 and Andre Haudricourt, and Alphonse Juilland, 6 the focus remains on internal causes. More recently, together with immanent structural factors, extra-linguistic social pressures are more and more taken into consideration. Linguistic prestige, never completely absent even in the writings of the Neogrammarians, has become a major factor in socio-linguistic studies as initiated by William Labov.7 A well-known, well-documented sound change is the one undergone by stressed vowels, except i and ii, in open syllables in post-Classical Latin and early Romance. Even though we obviously have only written sources to document the changes - a fact that has given rise to some disagreement about the precise dating of the phenomena, as spelling change tends to lag behind change in pronunciation or does not reflect the change at all - the data are well established. 8 One can explain these data by invoking influences on the syntagmatic level (sentence stress, neighbouring sounds affecting one another) or point out the consequences of one particular change for all the elements in a given paradigmatic structure and posit a chain reaction in that structure. So, the change of closed o in free position under stress into o, which occurred

9 System, norm, usage in the northern half of what is now France, may be seen as the result of diphthongization after lengthening, dissimilation of the two parts of the diphthong, and truncation (6 > oo > OI) > OI) > o), without taking into account the other stressed vowels in free position, or one can connect this change with all the other shifts that occur in the system of stressed vowels in open syllables. The impact of these changes in verbal paradigms is considerable. The identity of stressed and unstressed roots is no longer a general feature of the present tense, for example: je treuve je lieve je meurs j'aime

nous nous nous nous

trouvons levons mourons amons

However, the majority of verb stems are not affected by the vowel changes in question as their vowels are not in free position but blocked by consonants, and as stated before, free Iii and lu/ are not subject to the changes either. So, a conflict arises between the new phonological rule and the rule according to which verbal stems remain identical throughout the present tense. If one compares the situation in modem French with what should have been the outcome if the laws of phonetic change had been really exceptionless, it becomes clear that for the bulk of the affected verbs the two-stem present has been eliminated in favour of the principle of uniformity. A study by Stefan Ettinger dealing with the conflict between the tworoot paradigm and the one-root paradigm in three Romance languages - French, Romanian, and Italian9 - presents it in terms of a clash between the norm that prevails in the two-root verbs and the system of one-root verbs. The simpler system exercises great pressure to reunify the stem in the two-root verbs. As long as the more complex paradigm is protected by the prescriptive norm of what the French call le bon usage the conflict situation remains and speakers of the linguistic community either adhere to the traditional forms or use the innovations brought about by pressure from the system. Reactions of speakers vary according to social level, geographical origin, and age. Ettinger also mentions the fact that frequently used verbs stand a better chance of keeping the more complex paradigm than verbs that occupy a less prominent position. We shall come back to this phenomenon when discussing the theory of cost and yield (see chapter 4, page 26). Apart from differences in individual attitudes towards linguistic change


Linguistics, literary analysis, and literary translation

caused by system pressures, Ettinger draws attention to what one could call national differences. French grammarians are much more emotional than their Romanian and Italian colleagues. Although acknowledging the fact that even in the works of great French writers the new, and in their view non-acceptable, forms are making inroads, most French grammarians are reluctant to admit that it is difficult, if not useless, to turn back the clock. Subjective opinions about the ugliness of the new forms and the harmony of the old ones are readily given (the aesthetic appreciation of the imperfect subjunctive being an exception to this general rule) without any consideration for linguistic evolution. On rare occasions the euphonic or poetic virtues of a new form are extolled, the only reason being that a prestigious author has used it. Even a scholar such as Grevisse, 10 who has a clear sense of evolutionary trends, does not go any farther than stating that forms that are proscribed nowadays may well become the norm for our grandchildren. Yet he recommends the use of the old forms and does not want to take any reponsibility for innovation. So, one could say that Grevisse's work remains faithful to the prescriptive tradition - a fact that is hardly surprising when one considers its title, Le bon usage - whereas the functionalists, such as Martinet and his collaborators, accept new trends as they manifest themselves in the language without making aesthetic value judgments. If the acceptance of evolution is rare in France it is the rule in Romania, where the official language policy favours system-conditioned simplifications. No tears are shed about the merging of the two non-productive conjugations (the second and the third), and wherever two forms seem to compete, the one that is in line with unifying trends is recommended. So, the Romanians take full advantage of the absence of rigid norms and a long-standing tradition and try to shape a coherent form inventory where no conflicts arise between norm and system. Regional differences do exist and complicate the overall picture, but gradually the Muntenian variety of Bucharest Romanian has become the norm, and this variety happens to be the most innovative of all Romanian dialects. Linguists take an active part in the shaping of a language policy favouring the system. 11 In Italian, where local, highly developed standard varieties compete with one another for the honour of becoming the national language, regional divergencies in norm and usage are more numerous than in French and Romanian, and double forms are more often accepted as equal. Yet the same tendencies towards simplification of the verbal paradigm that are rejected in France are accepted in Italy, even if there is not as deliberate a policy of favouring them as there is in Romanian.


System, norm, usage

The same sort of conflict exists in Germanic languages between weak and strong verbs. Strong verbs form their past tense and past participle by changing their stem vowel, while in weak verbs the stem remains unchanged and special morphemes are used to create the forms in question. The latter type is productive, which means that new verbs follow the weak model. So, the strong verbs are slowly losing ground, not only because the rival group increases in number, but also through attrition as some verbs become obsolete, and through defection. This defection creates a conflict between the norm form and the system form. Many examples can be given of children using system forms such as runned instead of ran or hybrid forms like ranned, before they eventually fall in line with the prevailing norm. Sometimes the norm is not clear any more, as for the Dutch verb stofzuigen ('to vacuum-dean,' literally 'to suck dust'). Depending on the interpretation as a verb with an object or as a single verb one gets: a. ik zuig stof ik zoog stof ik heb stof gezogen

'present' 'simple past' 'perfect'

b. ik stofzuig ik stofzuigde ik heb gestofzuigd

The b series is new, and represents the system. As the new forms are interpreted as being of a single verb, one may expect the possibility of a direct object: ik heb de slaapkamer gestofzuigd (?)

'I vacuum-cleaned the bedroom'

a possibility that is excluded when stof still functions as a direct object. In the case of the Dutch stofzuigen, the pressure of the system is not resisted because of the uncertainty about the status of stof, but in a great number of strong verbs the norm is not threatened at all in adult speech. The situation is somewhat different when it is not only the signifier of a specific verbal category that is involved, but also the signified. In contemporary French the position of the passe surcompose (j'ai eu fini) is fairly well established, it seems. Prescriptive grammars indicate that the 'simple past' and the 'anterior past' (passe simple and passe anterieur) of the written language are in spoken French often replaced by the 'compound past' and the surcompose: II partit, des qu'il eut fini . II est parti, des qu'il a eu fini.

However, closer scrutiny reveals a good deal of confusion. First, according to a survey published in 1981 by Henriette Walter, 12 there are regional differences in the degree of acceptance. In some areas the surcompose is


Linguistics, literary analysis, and literary translation

hardly used at all, whereas in others it is quite common. Even in the latter, however, the use of surcompose forms is mostly limited to the passe surcompose itself, and the system pressure in the direction of an extension along the lines of j'ai fini j' avais fini j' aurai fini etc.

j'ai eu fini j' avais eu fini j' aurai eu fini

seems not to have resulted in the wide use of these forms. 1 3 The system, then, allows for the extension; the norm is not very explicit; and the usage, marginal. One could say that in this instance the three factors are not really in conflict with one another. A more controversial point in the passe surcompose, where system and norm clash in such a way that most speakers of French avoid the verbal forms that are affected, goes back to the auxiliary etre, used for pronominal verbs as well as for a small number of verbs of motion (e.g., aller), or verbs indicating a change of state (e.g., mourir). 14 Whether we take Des qu'il se fut !eve, Jes autres partirent

or Des qu'il fut parti, Jes autres se leverent,

the surcompose creates a problem: Des qu'il s'a ete !eve(?), Jes autres sont partis Des qu'il a ete parti (?), Jes autres se sont !eves.

Although the correspondence a > a eu I est > a ete is solidly entrenched in French, it does not seem acceptable in this instance. Some native speakers when asked which they would say venture: Des qu'il s'est eu !eve(?) ...

thus promoting eu to the rank of general surcompose marker and introducing a new compound, est eu, into the language. In a publication of the Linguistics Section of Lausanne University, in which systems and variations are discussed, 15 one encounters ii est ete parti, a combination that bears witness to the confusion of the users of the passe surcompose rather than indicating a viable alternative to est eu parti or a ete parti. When people are in doubt about the norm of the surcompose signifiers, they can easily avoid the problem by using another tense. This policy

13 System, norm, usage cannot be followed when the problem affects frequently used forms. The French subjunctive offers a good example of a norm/system conflict that is central, in the sense that speakers cannot avoid it in the way the marginal surcompose issue may be avoided. One should keep in mind that in order to qualify as a separate category, the subjunctive should fulfil two requirements: a / it should have a formal distinct identity; b / it should have a distinct semantic function (for a complete discussion see chapter 5, pages 33-4, and chapter 13, page 87). Although many verbs do offer special forms for the subjunctive, in the very important productive conjugation in -er the formal identity of the subjunctive is questionable at best. In the paradigm of the present (some linguists prefer atemporal as the 'present' is also used for past and future events) the singular and the third-person plural are identical with the present indicative, while the first- and second-person plural forms coincide with the forms of the imperfect: je donne tu donnes ii donne nous donnons vous donnez ils donnent

que je donne que tu donnes qu'il donne que nous donnions que vous donniez qu'ils donnent

je donnais tu donnais ii donnait nous donnions vous donniez ils donnaient

As for the free choice between indicative and subjunctive where the verbal form carries the semantic opposition and all the other elements in the sentence are identical, comtemporary French offers very little of it. In most cases the use of the subjunctive is what the French call a servitude grammaticale, entirely depending on other elements, the presence of which in the sentence makes the subjunctive compulsory. This requirement is thus either purely formal as no semantic opposition such as a. je suis heureux que tu sois ici b. •;e suis heureux que tu es ici is used, where the happiness in a is greater than in b, being expressed twice, or the requirement does not make any difference at all as subjunctive and indicative formally coincide. Thus, it is not surprising that the majority of French speakers have great difficulty acquiring the norm and tend to abandon the subjunctive as a category. Some irregular forms seem to resist, but they become remnants of a disappearing breed and are fossilized . It is remarkable, however, that the norm is still enforced

14 Linguistics, literary analysis, and literary translation with great conviction and that so many speakers try so hard to master the rule for the use of a category that is semantically, as well as formally, so weak. The reward for achieving competence in this area, as expressed in terms of yield compared to cost, cannot explain this attitude unless other factors than the ones pertaining to the linguistic functional and systemic levels are taken into account. The correct use of the subjunctive has become a social-class marker, an indication that one has received the right education and belongs to the higher strata of society. So, yield is to be calculated in the amount of social prestige the subjunctive conveys to those who follow the norm. The prestige element that is part of the norm observance plays an important role in literary language. It is for that reason that using literature as a source for studying norm, usage, and system problems is not without some drawbacks. The authors may strive for purity according to the norm, or they may, as is the practice of some comtemporary writers, try to follow the usage of the majority where it is opposed to the norm. So, we are faced with a new triad: literature, norm, usage.


Literature, norm, usage, and evolution

The documentation of the tensions between norm and usage for all periods except the most recent is limited to written sources. They can be divided into the following three categories: 1 commentaries of language phenomena by grammarians, almost exclusively of a prescriptive character, but with some important exceptions; 2 writings of a non-literary character: graffiti, shopping lists, legal procedures, etc.; and 3 literary texts. Of these three categories the first and the second are often much more interesting from a linguistic point of view than the third, especially when one wants to study language evolution, norm, and usage. 1. The prescriptive, normative works in the vein of the Appendix Probi, 1 a glossary giving a list of common mistakes and the correct forms that should be used instead, provide a wealth of data especially in the column of so-called mistakes, but sometimes also in that of prescribed forms in which some of the innovations the author tries to prevent have crept in. As the Appendix was probably written in Rome in the third century, we are able to see how unsuccessful the author of the glossary was in his effort to defend the norm of Classical Latin. It turned out that weakening and eventually disappearing final consonants could not be called back and vowels that changed their pronunciation, or were diphthongized or even dropped, could not be restored in their previous splendour. The expression 'previous splendour' is not meant ironically, although seen from afar it may seem strange that the diphthong au in aurum is deemed much more beautiful than the o in oro. It becomes even more curious when one compares this example with one taken from presentday Dutch, where the relatively pure vowels of standard usage show a

16 Linguistics, literary analysis, and literary translation

strong diphthongization in popular speech, so that the beautiful o is replaced by the ugly diphthong aw. From my personal experience I can say that aesthetic evaluation is particularly loaded with emotion when one is dealing with one's native language. I know that social prejudice and the need to preserve one's identity via the stability of one's language play a very important role in judgments about what is beautiful or ugly in a language, and I would have liked not to be influenced by such linguistically irrelevant factors. Yet, when I first heard my son's diphthongs - while studying in the Netherlands he acquired the now widely accepted 'substandard' vowel patterns - I felt uncomfortable and had to make an effort to refrain from 'correcting' him. Another example, taken from French, shows even more clearly how futile these aesthetic preoccupations are. For centuries the French struggled with the problem of how to pronounce words written with oi. The standard pronunciation in the days of Louis XIV was [we], but popular Parisian showed [wa]. Grammarians, though in general prescribing [we], gave in to the pressure of [wa], accepting it for some words. The uncertainty of the situation is increased by the fact that the list of words and forms where [wa) is acceptable varies from one grammarian to the next. On the whole, however, there was a consensus that [we] was nicer than [wa]. 2 Nowadays, three centuries after the negative reaction against [wa], some Frenchmen say that the Quebec pronunciation [we], where continental French has [wa], is ugly. That Louis XIV, in all likelihood, pronounced the last word of his famous dictum L'etat c'est moi just like a Quebecker would, i.e., [mwe], will not change their opinion. It is not only pronunciation that is subject to emotional value judgments that do not take into account factors such as system equilibrium and cost-and-yield balance; these judgments also pertain to grammatical categories. Both in English and in French, conditional if-clauses do not require a special conditional marker in the verbal form: If you were here, he would be happier. Si tu etais id, ii serait plus heureux.

In Toronto English, however, If you would be here, he would be happier

is a quite common spoken variant. The same thing happens in French, where regionally (as, e.g., in Belgium)

17 Literature, norm, usage, and evolution Si tu serais ici, ii serait plus heureux

is frequently heard, if not seen in written texts. 3 In both cases the verbal forms in subordinate and main clauses now both belong to the conditional category, and this parallelism takes precedence over the avoidance of redundancy in the subordinate clause. If the now still proscribed constructions eventually become standard, nobody will find them ugly any longer. This process of gradual acceptance is already well under way for the split infinitive. One may wonder in this case what the advantage is of what is mostly seen as a journalistic mannerism. In some instances and this is rarely mentioned - the placing of the adverb between the two parts of the infinitive avoids possible ambiguities: No matter what people say about loving challenges, a. it is strange really to like impossible tasks b. it is strange to like really impossible tasks c. it is strange to really like impossible tasks

Only in c is the adverb unambiguously linked to the infinitive to like, so the negative verdict given by prescriptive grammars does not make much sense. One could object that only some of the numerous split infinitives one encounters can be justified in this way and that often there is no ambiguity whatsoever when the adverb is placed before or after the infinitive. However, once the construction serves a dear purpose in some cases, it is difficult to stop it from spreading. Whether in the area of pronunciation, of morphology, or of syntax, normative prescriptions invariably point to the existence of a competing usage. In modern descriptive works that discuss actual language usage rather than only providing information about how one should write and speak, this competing usage is amply discussed. For older periods rarely is information supplied that deals directly with usage, and it is obviously impossible to do surveys and to rely on field-work, both of which are important parts of contemporary linguistic research, but the written texts do yield some interesting data . 2 . Depending on the level of instruction of the often anonymous writers of graffiti, shopping lists, etc., the deviations from normative language will be more or less significant. Inscriptions on the walls of Pompeii and Herculaneum, well preserved under the volcanic ashes of Vesuvius, give a goodimpression of the usage of Latin in a provincial town around 79 AD. They are full of the 'mistakes' against which grammarians kept

18 Linguistics, literary analysis, and literary translation warning until those mistakes had become the norm. It would be wrong, though, to think that all non-literary documents, written by illiterate or semi-literate people, represent spontaneous, current usage not adulterated by normative preoccupations. A good illustration of what can happen to normal usage can be found in La grammaire des fautes, a book written by a Swiss linguist, Henri Frei. 4 The material the author uses for his analysis consists of letters written to the Red Cross in Geneva by relatives of missing French soldiers, who ask for help and information. Frei labels the language they use as franrais avance, a name suggesting that it is, in its evolution, farther advanced than the official language. The deviant forms are what can be expected to become the norm, if the evolution is not held in check by conservative forces such as school instruction and upper-class prestige. The latter is no doubt responsible for a great number of curious simple past tenses, subjunctives, and other forms that give away the writer's desire to use beautiful language as well as his inability to do so correctly. Sometimes the form does not exist and is created according to a more productive paradigm (simple past tense of rire: ii rit replaced by ii ria, after the -er conjugation [e.g., donner: ii donna]) . More often a prestigious category is erroneously used. It is as if the writer (speaker), who does not know the rules governing the use of the category in question, wants to be on the safe side. This hypercorrect use is just as revealing as the mistake consisting of not using the category where the norm prescribes it. One could even go one step farther and say that a hypercorrect item is more revealing because it signals not only that the writer (speaker) does not know the rules, but also that he is aware of the fact that he does not know them. 5 3. In literary texts the phenomena of analogous new formations and hypercorrect usage occur, but on the whole they are much rarer than in the sources described under category 2. Whereas those sources can rightly be considered to form the vanguard of linguistic change, literature is linguistically speaking more often than not in the rearguard, and linguists interested in evolution per se or simply in a dynamic-synchronic description (see chapter 1, page 4) prefer non-literary material. Yet, for all periods except the most recent, linguists have to rely mostly on literary texts for their data. Fortunately, tracing an evolution that has already taken place is different from trying to detect an evolution that is under way. A few examples will make clear what possibilities literary texts offer and what pitfalls the researcher may encounter, depending on the period he is concerned with and the aims he has set himself.

19 Literature, norm, usage, and evolution

By extrapolating what we observe as the relationship between literary language and common usage for our own time we may assume the following general characteristics, regardless of the century from which we take a sample of literary texts. a From the point of view of evolution the text is likely to represent a stage of the language that is anterior to the one prevailing at the moment the text was written. b Depending on the strength of norm pressure the text more or less conforms to its requirements; it is unlikely that it reflects the completely spontaneous usage of the author. c Norm pressure not only varies from one period to another, it also varies from one literary genre to another. So, depending on the genre a literary text is more or less subject to the influence of what is seen as the prestige variety of the language. d Substandard and regional forms may or may not be deliberate. If they are deliberate - used to indicate the social or regional background of the person who is quoted directly - they are not always a reliable source for the language variety they are supposed to represent. a. The introduction of new spellings, reflecting changes in pronunciation, lags inevitably behind. It takes some time before the change is sufficiently general to impose itself- at first the old and the new coexist, the former constantly losing ground to the latter. Even when it has become general, orthography may remain unchanged as, for instance, in the case of Latin [u] that became [ii] in French, or in that of the great vowel shift in English. So, orthographical changes give only a partial picture, at best, of what happens to pronunciation. In modern times a problem that arose when spelling based on the leading dialect became the standard for people who spoke a similar but yet noticeably different dialect has become particularly acute, as regional susceptibilities and national pride manifest themselves very strongly. For French it all started when Parisian Francien became the model for the written language throughout Northern France. 6 For most writers the written form represented a deviation from the oral. This majority had an obvious interest in keeping the spelling as they had learned it instead of substituting a new spelling for the old one, neither of which corresponded to the language they heard spoken around them and spoke themselves. It was during the period that Paris standards were accepted that the orthography was more or less frozen and ceased to follow the evolution of Parisian pronunciation. The long-standing tradition of a spelling that is divorced from pronunciation makes it difficult to reunite


Linguistics, literary analysis, and literary translation

the two elements. The unity of the Francophone world would be threatened if spelling were adjusted, as each variety of French would have its own spelling. The same is true for English, where North America [wad;)r] differs significantly from British [wot;)].7 b. General education and mass media have put their mark on normative language use. The norm tends to become less elitist and proscribed forms may find their way into standard language more easily. An impression of greater individual freedom may be created, but the greater tolerance is offset by the strong levelling influence of radio, television, and the print media. Authors have never been completely shielded from outside influences, and some linguistic features of their texts are the result of subconscious or conscious borrowing, motives in the latter case not being important here. c. Already in medieval French literature the standard forms were more scrupulously used in what one could call high literary genres than in more popular types of texts. Adam de la Halle wrote standard Parisian French in his serious works, but in his humorous poetry local substandard usage is frequent. 8 d . Regional and substandard forms belonging to a sociolect without prestige do occur, at times, in literary texts. Often the subject-matter is sufficient explanation for their presence in the text, but unless used in direct quotation of the words of a protagonist, those forms pose a problem. They may be part of the normal usage of the author, in which case his intention would be misinterpreted if one assumed that their function is to convey, as well as the semantic content, an element of couleur locale. If these forms occur in the same text as the semantic equivalents of the standard language the couleur locale interpretation is most likely to be the correct one, but in the absence of standard equivalents the question of authorial intention remains open. For the reader of such a text the dilemma of how to interpret the deviant use may be of little importance, but for the translator the problem is neither simple nor irrelevant (see the discussion of English translations of French-Canadian novels, chapter 17, pages 116--19). The deliberately used forms often belong to a language variety that the author knows very well and that he has spoken during part of his life, even if he does not use it any longer. However, his representation may be selective in that it gives what seems to the writer to be the most salient features, leaving out what does not strike him as such. If he is an outsider, the rendering is necessarily incomplete, and in some cases incorrect. A very curious literary phenomenon is that of a phonetic


Literature, norm, usage, and evolution

notation of 'deviant' forms that when pronounced aloud turns out to be the notation of the standard language (see below, chapter 9, page 56). In summarizing we can say that literary texts form an important source for the linguist who studies the dynamics and evolution of language, but that they should be approached with circumspection because not everything can be accepted at face value, even when the text is completely authentic, and not, as is the case for many medieval manuscripts, changed or corrupted by copyists from a different region and a later century. For the past the texts provide material that allows us to reconstruct and trace back an evolution, the outcome of which we know by observing the present. As soon as the research loses its retrospective character, however, and turns to the present or even the future, literature's basically conservative language practice makes it a cumbersome source to use. In Dans un mois, dans un an, one of the elegant short novels that made Fran~oise Sagan famous, there is one instance of a subjunctive after apres que, 9 and in Henri Barbusse's L'enfer an imperfect subjur..ctive is replaced by a simple past, 10 but one isolated example may pass unnoticed and even if one spots it, it only corroborates what was known already from other sources, namely that the subjunctive past or present and the simple past are trouble-spots in the French verbal system. As we stated already, non-literary sources yield a much richer harvest in hypercorrect or otherwise deviant forms. The confusion between I and me in sentences such as John and me were first or People like John and I don't do that is a widely attested, though widely deplored, phenomenon of contemporary English. It will probably take some time, however, before the confusion penetrates into literary language to the point where the norm will shift and usage and norm will coincide. All one can say at present is that the overreaction against substandard me in the first sentence gives rise to the hypercorrect I in the second one. Documenting the process of change that is under way at the moment, relying exclusively on literary texts, would be premature.


Cost and yield

When talking about linguistic change, system pressures, and analogies, where a frequently used form influences a form that occurs in association with it but is not as frequent, linguists often invoke the cost-and-yield theory to explain the phenomena under discussion. This theory constitutes an improvement on, and a refinement of, the law of least effort (la loi du moindre effort). According to that law intrinsic human laziness is at the root of change. Dropping of final consonants, voicing of unvoiced consonants between vowels, and a slurred articulation resulting in a change of the sound that is being pronounced are among the most quoted examples of how the law manifests itself in phonetics.1 It is obvious, however, that least effort gives only half of the picture. If the only force at work were the one aiming at reducing the speaker's effort as much as possible, the ultimate result would be a speaker opening his mouth and proffering a vowel sound whose pronunciation requires a minimum of tension of the speech organs. This sound, then, would signal the desire to communicate a message (or maybe to express a feeling without any particular desire for communication), and the onus to interpret the message or expression would rest entirely on the listener. Reduction never reaches this extreme, of course. But even a partial reduction will in most cases add to the burden of the listener and reduce the chances of a successful communication. So, the tendency to reduce the pronunciation effort is held in check by the desire to be understood. One only needs to think of an occasion where someone, because of extreme fatigue or excessive drinking, spoke indistinctly or with a slurred voice, to remember how much more effort it took to decode the message. The same is true for handwriting: it is infinitely more difficult to decipher a sloppily written letter than one that is neatly presented. Especially

23 Cost and yield

when the subject-matter is difficult or unfamiliar - or when the language in which the message is formulated is not the listener's/reader's native language 2 and not as easily understood by him - the yield will go down rapidly if the speaker/writer reduces his cost. So, the cost-and-yield theory takes not only the encoder into account, but also the decoder. In order to have a successful communication, or as it is also called, a felicitous speech act, the speaker must ensure that the cost does not become too high for the listener, for otherwise he is at risk of losing his audience. He should not overdo it, however, for too emphatically pronounced language, even if it is used with the best intentions and only to alleviate the task of the listener, may create the impression of condescension. The listener feels that the speaker underestimates him and is annoyed and offended. What has been said about pronunciation applies also to other levels of language. The denotative contents (for a discussion of denotation, connotation, and linguistic sign see chapter 10, pages 63-4) of a message vary greatly in precision. Some texts - spoken or written - are easy to interpret because the elements that compose them allow for one interpretation only. If generic terms 3 are avoided and a term is chosen that is more restricted in its applications there is little room for misunderstanding and if the terms of the message are known to the listener, he also knows what the speaker refers to. If, on the contrary, the speaker uses what the French call mots-passepartout, words with such a wide range of applications that the choice for the listener as to how to interpret the message is almost limitless, the message he receives does not have the same built-in precision. Instead of a picture with neatly drawn contours, he gets a blurred view of he does not exactly know what. Context and extra-linguistic situation will in many cases help him to focus correctly. Thus, he will not confuse the facilities of 'This university has excellent facilities for its students' with those of the road-sign 'Rest-area without facilities,' but in the former message he might like somewhat more precise information and in the latter he will either understand what is meant or be puzzled. Toilets or even the euphemistic bathrooms would have conveyed the same message more directly and clearly. It is possible, of course, that many facilities are unavailable and everybody is free to make his own list depending on his own frustrated expectations. A somewhat different case arises when somebody speaks in cliches. The cost for the speaker is rather low, and the same is true for the effort that the listener has to make. However, the yield may be so low that the latter thinks that his effort is too high in comparison with what he

24 Linguistics, literary analysis, and literary translation gets for it, and stops paying attention to what is said. This situation also applies to readers. A text in which almost everything is predictable is easy to read. The success of series such as the Harlequin Romances is based on the need for easy reading material4 that provides a non-strenuous pastime and an alternative to passive television watching. The more difficult a text is, either from the point of view of morphology and syntax or as far as content is concerned, or both, the harder the decoding becomes. Some literary texts are so hermetic that only a well-rested reader is able to tackle them, and a person who is a sensitive reader of poetry when ideal conditions make him receptive may revert to detective stories when he is tired or preoccupied. In this case the cost-and-yield theory is more relevant for the receiver than for the sender, as it is very difficult to measure the cost of an artistic creation. The question of cliche versus original text is closely linked with the problem of information and entropy. The first term is rather confusing in that it is used to convey two different meanings that slightly overlap so that it is not always clear which interpretation is intended. The common meaning of providing information is the imparting of knowledge of some particular fact or occurrence. In that sense a text that does not contain much information is one that does not impart much knowledge to the reader. As well as this common use, one finds a more technical and precise definition according to which the informational load of a linguistic term is inversely proportional to its degree of predictability. To a certain extent non-predictability also goes together with greater imparting of knowledge as, for instance, when the expected nice in 'I hope you'll have a nice evening' is replaced by awful: 'I hope you' ll have an awful evening.' In some instances the knowledge factor is not relevant at all. When the French poet Paul Eluard writes in a poem that the earth is blue like an orange,5 orange carries a heavy information load as it is totally unexpected as a term of comparison but the information, in the more common sense of the word, is dubious. The question of what texts without a message that can be directly related to some real-world event or situation do convey is a difficult one and yet it has to be answered somehow when such texts are translated (see further chapter 15, page 101). Turning back to the question of information in its technical meaning, we should keep in mind that predictability cannot be calculated with equal precision for all levels of linguistic analysis, and that at the lexical level, for instance, predictions become the easier to make, the more familiar are the style and content-matter of the text.6 Only for elements

25 Cost and yield

belonging to a closed inventory can precise calculation of probability of occurrence be carried out. Thus, in English a two-consonant cluster in word-initial position consisting of Isl and ltl rules out all consonant phonemes but lrl for the third phoneme slot, and admits all vowel phonemes. For each language a separate set of probabilities exists, so that even when the phonemic inventories of two languages are very similar the probabilities of occurrence of each phoneme in specific positions may differ greatly from one language to another. If the probability of occurrence of an element is 100 percent, that element is considered to be redundant. Some redundancy is built in in the language system as, for instance, the verbal ending -ons in French when the pronoun nous has already unambiguously indicated the first-person plural category and even the most efficiency-oriented speaker or writer cannot avoid it.7 In linguistics the notion of entropy8 is sometimes introduced. Entropy represents the degree of uncertainty about the appearance of each signal. Thus, for a given number of possible responses, maximum entropy occurs when all responses have the same frequency. In the case of absolute redundancy, the element in question (in the example just given, the morpheme -ons) is the only one that can occur. If redundancy is only relative, the following formula can be used: relative redundancy = absolute redundancy - entropy (degree of uncertainty). 9 This concise and exact formula is of limited usefulness, though, for all those cases in which the degree of uncertainty cannot be calculated, and when entropy is mentioned in spite of this impossibility we should not forget that we are dealing with an intuitive approximation and not with a precise entity. The greater the closed inventory of elements, the more cumbersome entropy calculations become, and as soon as one is faced with open inventories, precise calculations turn out to be impossible. Precise calculations are also excluded in another area where cost and yield are referred to in order to explain the prevailing state of affairs. In spite of the absence of measurability, the cost-and-yield principle provides some useful insights into the phenomena of irregularity and normalization. The cost element is fairly easy to grasp. For the speaker it is less cumbersome to internalize one morphological pattern that is applicable without restrictions, than to memorize a whole series of semantically equivalent but morphologically different forms. If one of the latter categories does not come to mind quickly enough, many speak-

26 Linguistics, literary analysis, and literary translation

ers have the tendency to substitute a formation, arrived at by applying the general rule of the first category, for the form they cannot produce immediately or have never been able to produce correctly. There are, of course, refinements to be made when more than two categories are competing; the most numerically strongly represented category with the most general applicability will grow at the expense of less frequently used categories with a smaller range of applicability. The cost of using one general model is lower, obviously, than that of having a variety of models, or even unique forms occurring only once in the inventory. While the cost factor is fairly straightforward, yield is much more elusive. The question that immediately arises when it is pointed out that one general model is less costly and more efficient than a variety of formation patterns is why some nominal and verbal paradigms are not simplified and normalized according to the principle of least effort, but instead keep formal identity that is perceived as irregular. The answer to this question has to take into account a number of factors that are anything but homogeneous. First, there is in some forms an element of directness that may be conceived of as efficient. Instead of an analytical form composed of a lexical stem and a grammatical morpheme, there is only a synthetic whole that signals lexical and grammatical content at one and the same time. The only first-person plural in French without the grammatical morpheme -ons is nous sommes. The signifier Isom/ cannot be analysed into smaller meaningful constituent units. The strong verbs in English, Dutch, and German offer a past with vowel gradation (ablaut) - sang, zong, sang (from to sing, zingen, and singen), as opposed to the past (with suffixes) of weak verbs lived, leefde, lebte (from to live, leven, and leben). If the efficiency of a synthetic form were the only factor involved it would be enough to say that, in the case of frequently used forms, this efficiency outweighs the cost of formal deviance. However, there is more at stake, and yield cannot be expressed in terms of systemic efficiency alone. One has also to keep in mind that once a form is established in the minds of the members of a speech community, the introduction of a change may adversely affect the chances of a successful communication. This risk partly explains conservative attitudes towards innovations. In an ideal situation all members of a speech community would use the same forms and attach the same meanings to those forms and no changes would occur. Although such a situation obviously does not exist anywhere, many people tend to react as if the status quo - mostly the status quo at the time that they acquired the language- is intrinsically good and that change is bad. When sacred language and religious texts

27 Cost and yield

are involved conservatism may take a violent tum but even when religion is not involved tempers often flare in language discussions about right and wrong. Thus, conservatism accounts for the retention of irregular forms in the prescribed norm even when those forms are not frequently used. It is in this way that non-efficient forms of a high cost and a low functional yield often are invested with prestige and become social markers. The higher the level of education the more likely the active use of these social markers is. As their occurrence is rare, the risk of hypercorrect or otherwise deviant forms is great. The yield of the correct forms is to be found in the realm of social prestige, but eventually the conservative influence of upper class and upper-middle class may not be strong enough to protect the costly form. A curious reversion may take place, turning the prestige marker into a negative element linked with snobism, with old-fashioned stuffiness, or with both. In the discussion of prestige and yield, one should not overlook the fact that two opposite forces are at work. On the one hand we find the conservatism that was just mentioned; on the other hand there is the desire to follow the trend, to be up to date and to project a modem image. In literary texts both attitudes manifest themselves. Whereas conservatism used to be more widely represented than modernism, in present-day literature the latter has a strong showing. The author's choice - if there is a choice and if we do not deal with a writer who has only one language va~ety at his disposal - often reflects his political and social beliefs. Special language varieties are frequently used in direct speech, reported in novels, in order to provide social and/or geographical information about the speaker. Some authors comment on the language of their heroes or have them reflect themselves on language use. These comments about language - the technical term is metalinguistic comments - represent, as a rule, attitudes of at least some segment of the speech community. In spite of the great differences between the various branches that can be distinguished in the cost-and-yield theory, what they all have in common is that formal complexity results in either additional or more precise or more directly perceived information (used in the non-technical way). Unless extra cost gives extra yield, people are not likely to make the additional effort. The appreciation of the yield is subjective, and whereas some may think the niceties of a complex morphology a waste of energy, others cling to the old ways either because they have the impression of more clearly formulating what they want to say, or simply because they want to make their social status known. Apart from the cases where, as in nous sommes, no alternative is avail-

28 Linguistics, literary analysis, and literary translation

able, the interpretation of the form requires that it be compared with another form that could have been used instead. In other words, dialect is only perceived as dialect if opposed to standard language. If somebody has only a dialect or a substandard code at his disposal, he will interpret a dialectal or a substandard form used by someone else without attaching any social value to it. This principle of formal opposition is at the basis of linguistic thought. Even the just-mentioned form nous sommes, of which it was said that it was the only option available in the absence of morphologically more analytical forms, finds itself in opposition to other forms, be it not at the cost-and-yield levels that have been discussed so far. (One might object that nous sommes is socially or stylistically opposed to nous on est, but for the sake of simplicity this possibility was not taken into consideration here.) This principle of formal identity and formal opposition has far-reaching implications not only for linguistic theory, but also for literary analysis and for the theory and practice of translation. It is for this reason that a brief discussion of what could be called the comer-stone of structural and functionalist linguistics will form the next chapter.


Formal identity and formal opposition: the linguistic sign

If there were one term to indicate both a pear and an apple, let us say papple, the two fruits would not be distinguished in the way Englishspeaking people keep them apart now. This papple would be different from other fruits such as oranges, but the possibility of differentiating between apples and pears would be lost. One could, of course, use adjectives to create subcategories such as oblong pappies and round pappies, just as there are red apples and green apples, but the general term would still be papple. So, we have to conclude that the availability of two different terms rather than one general term has an impact on how people structure their messages. 'Could I have a papple?' is not quite the same as 'Could I have an apple or a pear?' for in the latter two kinds of fruits are mentioned, whereas in the former there is no specification of the subcategories of papple. The situation described in this hypothetical example occurs with a certain number of variations in natural languages. The most common, or at least the most widely known variety, is the one in which language A has one term to denote that for which language B has two terms. 1 One can introduce a further refinement in the description by distinguishing for language B between two terms that are completely separate and two whose meanings overlap to a certain extent. Both varieties will be further examined when translation theory and practice are discussed (see chapter 15, page 97). Another case arises when A and B divide a certain area in the same way, but whereas A also has a term that covers the whole area, B cannot but be specific, not having what is called a 'generic term.' In A each specific term is opposed to the other specific terms and to the generic term, while in B only the oppositions between specific terms are found .2

30 Linguistics, literary analysis, and literary translation

Until now everything that has been said ties in with Ferdinand de Saussure's ideas presented in his Cours de linguistique generale. According to him language is composed of linguistic signs. Those signs have a form and a content. The form is called the signifier (signifiant), the content is labelled the signified (signifie). For the signifier the Cours also uses acoustic image (image acoustique), and the term concept alternates with signified. Saussure states that the two components are indissolubly linked together, the one calling forth the other. Moreover, it is stressed that the link is conventional, which means that in order to know what a signifier 'signifies' one has to be familiar with the convention, for there is no natural connection between signifier and signified. In the opposite direction the same is true: the character of a concept does not give any indication about the signifier that is sufficient to establish its form. If one does not know the signified belonging to the signifier table, he will be unable to deduce the content from the acoustic image, and it is equally impossible to deduce from the characteristics of a concept - provided that one can draw up a list of characteristics - what signifier is linked with it. Saussure denies that there is an inner motivation for the connection between a given signifier and its signified and stresses the arbitrary character of the conventional link. This arbitrariness is a major point of contention, and especially in literary analyses one often finds remarks about the harmony between aesthetic sensations triggered by the signifier and the specific character of the signified (see chapter 7, pages 41, 44-5; chapter 10, page 62; and chapter 11, pages 68-70). Apart from the issue of motivation, there are some other problems connected with the saussurean linguistic sign. First, the sign can only function within a speech community whose members recognize it as such, because they all observe the same conventions. In the ideal speech community all members have the same lexical and morphological inventory, use the same rules, and have the same pronunciation. Such a community does not exist in reality, and the question arises as to what exactly the internalized system that Saussure called la langue represents, as opposed to the individual speech acts (spoken or written) that fall under la parole. The next question is how much individual deviation from the langue inventory is tolerated in a person without his being excluded from the speech community. Within one speech community situations can arise reminiscent of those that have been discussed when two different languages were compared. When the individual inventories of speaker A and speaker B are examined, we may find again cases where one element in one inventory corre-

31 The linguistic sign

sponds to two elements in the other. If Buses two terms where A uses only one, we can for B's two terms again distinguish between separation and overlap. Whatever the division may be like in B's inventory, the communication between A and B is hampered and the outcome may not be entirely felicitous. When B uses a term that is unknown to A, no communication takes place (unless A guesses what B wants to convey, by relying on the context or the extra-linguistic situation). However, when B uses the only term that A knows, A may not grasp the specific meaning that B attaches to that term, as opposed to the other term of his inventory. Similarly, when A uses the term in question, B may interpret it according to his own inventory, and lend it a specific character that it does not have in A's mind. The risk of misunderstanding is a consequence of the differences in inventory. As each concept is defined in relation to other concepts belonging to the same notional field, a difference in the number of concepts will automatically entail differences between more or less corresponding concepts in A's and B's languages. (For notional field theories, see the discussion, chapter 6, page 35.) In many instances, however, the differences do not prevent A and B from communicating successfully, as context and/or extra-linguistic clues help in arriving at the correct interpretation of the message. Each sign is defined not only paradigmatically - i.e., in the inventory - but also syntagmatically - i.e., by the other elements of the utterance in which it occurs. Whereas, for most practical purposes, the double delimitation on the paradigmatic and the syntagmatic levels is adequate, serious problems may arise in literary texts. Literary theory and literary analysis do not always explicitly mention the problem of different inventories of writer, reader, and possibly critic, but if they do, approaches to the problem vary a great deal, as will be seen later (chapter 16, page 105). One of the aspects of the inventory question is linked with the incorporation of non-denotative elements in language description. The saussurean linguistic sign is, as was said before, based on the conventional link between an acoustic image and a concept. Very little is said about the concept and how the division into concepts is arrived at. The examples Saussure provides in the text of the Cours do not give a clear indication as to how the opposition between two concepts should be interpreted. Is there room for overlap as the example craindre, redouter, avoir peur suggests, or is each concept totally independent, not sharing any meaning element with the concepts it is opposed to? The second way of representing the concept is in accordance with the image of a mosaic used by Jost Trier (see notional fields, chapter 6, page 37), but

32 Linguistics, literary analysis, and literary translation

even then there ought to be a common element: although the meanings of apple and orange do not overlap in the way the meanings of to fear and to be afraid do, they share the element fruit . So far the only types of content that have been discussed have had an intellectual and objective character, in accordance with the term concept used by Saussure. Even that restriction does not help very much in the task of breaking down a concept into its minimum meaningful components because these components do not fall under the rule that governs the linguistic sign and according to which the signified expresses its identity formally by a signifier. The smallest meaningful components that are generally called semes 3 do not have a conventional and necessary link with specific elements of the signifier, and this leaves the door open to a proliferation of semes. Yet the difference in seme inventory between terms such as to fear and to be afraid or the saussurean triad craindre, redouter, and avoir peur has never been established satisfactorily. This absence of complete seme inventories raises the question of synonymy. If, on the one hand, complete identity of two signifieds, each linked with a different signifier, exists, the one-to-one relationship posited by Saussure as the corner-stone of his linguistic sign model is no longer without an exception. If, on the other hand, the signifieds are very similar without being identical, one has to accept two concepts that are very close to each other and to determine what makes them different. A simple opinion poll carried out by asking acquaintances and friends if there is a difference between redouter and craindre, and if so, of what it consists, will show that apart from denotative elements, the answers also suggest a difference in register or level of language, redouter being not only stronger than craindre but also more formal. Qualities such as formality, vulgarity, and poeticalness do not belong in the denotative seme inventory, and it seems odd to include them in the concept. A further argument against inclusion can be found in the fact that often non-denotative elements (the term element is deliberately vague here) are expressed by pronunciation alone, the same signifier + signified of the language level being manifested differently at the speech (parole) level as far as the pronunciation of the signifier is concerned. The strictly synchronic point of view adhered to by the Cours has been softened over the years by the gradual introduction of dynamics. 4 This notion solves the problem of how to deal with archaic forms still used by elderly people at the moment of description and it can be used as well in dealing with trendy innovations. The price to be paid is that one has to face an even greater number of non-denotative elements that do not belong to the concept, yet play a very important role in the communication process.

33 The linguistic sign Synonymy is not the only phenomenon that creates problems. Homonymy and polysemy also invalidate the principle of bi-uniqueness of the tie between acoustic image and concept. In the case of homonymy two or more unrelated concepts share the same acoustic image. For polysemy the situation is in theory different as the unique concept can be subdivided into mutually related subconcepts. In practice it is not always possible to tell whether the concepts, or rather subconcepts, are related or not. If the divergent meanings of a signifier represent a case of polysemy the concept in which all the subconcepts should be incorporated becomes very vague. So, all the different meanings of the French acte (1: 'written deed'; 2: 'act of a play'; 3: 'action') do not fit easily in a more general concept. In literary texts homonyms and polysemes are particularly cumbersome to translate when, apart from the more or less expected denotation, the signifier calls forth other meanings by a process of association (see chapter 10, pages 63-4). These associations, unless made explicit by the author, are not given directly by the text and form a source of heated discussion between text critics whether they are translators or not. Until now only lexical infringements on the bi-uniqueness principle have been quoted, but they are exceptions to the rule. When we are faced with the semantic description of grammatical categories, however, the exceptions become the rule. In English, French, and Russian, for instance, the concept of plurality is expressed in combination with nouns and verbs, and takes on a great variety of forms depending on the lexical item it is combined with. For the overwhelming majority of instances the language does not allow for choice and all the different signifiers are in complementary distribution. 5 Although the one-to-one relationship does not apply to the plurality concept as such, each form expressing the plural has its own identity that is ratified by convention, so the infringement is not very serious. Formal identity is much more questionable when Roman Jakobson 6 introduces the different functions of language: in addition to the referential (denotative) function he distinguishes a poetic function, a conative function (when language is used to exhort or to persuade), an emotive function (where apart from denotative reference the emotion of the speaker is expressed), and finally the phatic function (where language is used in order to establish a contact but without the desire to transmit a specific message) and the metalinguistic function (where language is used to comment on itself). In spite of the fact that some formal criteria exist, especially at the oral level where intonation can be taken into account, there is a good deal left to the subjective interpretation of the decoder

.34 Linguistics, literary analysis, and literary translation (listener or reader), and in many cases there is overlap so that different interpretations are equally justified by the signifier level. So, it is not surprising that when linguistic concepts are introduced in literary analysis, the principle of formal identity is not always maintained. Even when the original text from which the concept is taken makes it clear that formal criteria are used for setting up different notional categories, this formal requirement is sometimes neglected. A clear example of this phenomenon can be found in the application of Emile Benveniste's ideas about the function of the middle voice 7 to languages where there are no formal criteria for setting up the middle-voice category (see further chapter 1.3, page 87).


Field models, universals, and language-bound world-view

In the previous chapter notional fields were briefly mentioned in connection with the question of formal identity (see chapter 5, page 31). As notional fields are sometimes used to compare two or more languages, and may thus provide some useful material for translation, it is important to give a few more details about field theories, their possibilities, and their limitations. First, it should be pointed out that the original terms used to describe a field were the German Sprachliches Feld or Sprachfeld ('language field') and Bedeutungsfeld ('meaning field') and that other terms which one encounters, such as Wortfeld, lexical field, semantic field, conceptual field, are all of a later date. 1 These different names do not serve to indicate one and the same concept. It is even impossible to pair off the names with the different types of fields one can distinguish, one name being used for more than one kind of field and one variety of field having more than one name. For this reason we shall enumerate the different types of fields without attaching too much importance to the names they may have in the writings of specialists. The nomenclature is a secondary problem and for the purpose of the present survey it is enough to make the reader aware of the fact that there is some terminological confusion. The general principle of a field, whether it is formed by lexical items or other linguistic meaningful units, is derived directly from the teachings of Ferdinand de Saussure. 2 Although Saussure does not use the metaphor field, his example craindre, redouter, avoir peur - about which verbs he states that, if one of them disappeared, the remaining two would occupy the space that has become vacant - is a perfect illustration of the field principle. Another example taken from the Cours deals with a grammatical category: if a language drops the dual as a number category, singular and plural occupy the entire number space. 3

36 Linguistics, literary analysis, and literary translation The fundamental difference between these two examples stems from the fact that the latter is exhaustive and closed, which means that apart from the three or two elements that are mentioned no other number category is available in the language in question. The former series formed by the three verbs is open and not exhaustive, as other verbs containing an element of fear may be added. The structuralist theses about language, that an element is defined negatively by all the other elements in the same category and that language is a system where everything is interrelated, are easily illustrated with examples where the elements belong to a closed, limited inventory. Phoneme systems, grammatical inventories such as verbal tenses or moods, personal pronouns, and number categories are often used as illustrative material. However, the term field is, as far as we know, only used in connection with lexical groupings. As the lexicon is open, consisting of an unlimited number of items, the impact of a single unit on the value of all others is impossible to demonstrate. In many cases it is for all practical purposes non-existent. The semantic value of table is not affected by the introduction of the neologism smog. However, fog and smoke do experience the influence of the new term. It is, therefore, not surprising that most linguists, when trying to structure the lexicon, limit themselves to selected subsections of the vocabulary rather than tackling the unwieldy whole.4 For those partial fields two approaches are used. 1. The diachronic approach deals with a very limited number of items, and studies the consequences of the disappearance or addition of an item for the interrelationships prevailing between all items under study. This kind of study is in line with the saussurean hypothetical example of the three verbs (of fearing), where two would share the area left vacant if the third had disappeared. 5 For translators of older texts it would, of course, be extremely helpful if diachronic studies of important lexical items were available. As we have no access to complete data about the lexicon that was used in the various speech registers in older periods, the constructing of diachronic lexical fields turns out to be very difficult and involves a great many assumptions and a good deal of guesswork. So, it is not surprising that not much work has been done diachronically, as far as lexical fields are concerned. 2. The synchronic approach focuses on the network of relations within a subgroup at a given moment, this moment being, almost without exception, the present. Different criteria are used to select the lexical items or to delimit the field. The most common method is to take a

37 Field models, universals, and language-bound world-view broad, general concept such as kinship, colour terminology, kitchen terms, seats, domestic animals, 6 and structure all the terms that qualify into a field . Depending on the character of the 'guiding notion' it will be more or less easy to decide which terms to accept for the field and which to reject. In all cases knowledge of the outside world is required, which means that non-linguistic criteria are used. If the structuring of the outside world is controversial, arbitrary decisions that are open to criticism will have to be made. It is undoubtedly to avoid this arbitrariness as much as possible that lexical fields often cover an easily defined and circumscribed part of the outside world, such as kinship or colour terminology. 7 These two fields have the added advantage that each item can be described according to a set of objective parameters (for colours, taken from physics; for kinship, taken from biology). The various structuring patterns one finds for these fields in different languages are compared and discussed extensively in anthropological studies. Some field studies follow what has been called the 'mosaic pattern,' using only terms belonging to the same grammatical category (nouns, verbs, and adjectives being the main targets), whereas others include all categories as long as the semantic connection with the guiding notion is clear. The links between the different items in the field are even looser when associative relations come into the picture. The cultural fields set up by Georges Matore, 8 where for a given period a few key terms are chosen that then serve as the centre for an intricate network of derivations, associations, and oppositions, snowballing towards more and more peripheral regions, are probably the extreme one can reach in synchronic fields. The field idea has, however; an even looser application in what is called the 'morpho-semantic field,' a creation of Pierre Guiraud.9 The organizing principle lies in expressiveness of sound (see chapter 7, page 41, and chapter 11, pages 69-70), word association, and word creation. No limitations as far as dialect, register, or period are imposed, regional and standard forms taken from medieval or from contemporary sources all having equal rights. The approach is neither synchronic nor diachronic and could be called 'panchronic.' One may wonder whether the field metaphor is still justified in a panchronic study. Even if one does not accept the morpho-semantic structures as fields, it is worth noting that formal, verifiable criteria are used. Formal verifiable criteria also play a role in series based on the identity of central lexemes, on the identity of morphological structure, or on sound similarity (word, wordbook, wordy, to reword, foreword, etc.; haste, hasty, hastily I health, healthy, healthily; rage, stage). More interesting from

38 Linguistics, literary analysis, and literary translation a structural and functional point of view are endeavours to correlate a certain grammatical behaviour with specific semantic elements. 10 These studies, carried out for Russian, are more promising for languages with a well-developed flexional system (declensions and conjugations) than for those with a rudimentary morphology and relying more on word order and function words (see for these concepts the section on analytical and synthetic languages, chapter 15, pages 99-100). Although phonic, morphological, lexical - that is, based on lexicalroot identity - and even syntactic criteria play an important role in some field studies, semantic considerations far outweigh formal ones. Without semantic criteria that are based on knowledge of the outside world very few field studies could be carried out. This reference to the outside world, combined with the schematic representation of each unit in a field whose surface is completely covered by all items together, creates the impression that, provided we know the position of an item of language A in the field it belongs to, it must be possible to find the term belonging to that position in language B, even if the area the two terms cover is not identical. This argument is obviously a non-linguistic one, but there are also linguistic features that in their universality seem to argue in favour of the possibility of translating from one natural language into another. It is not only for translators, but also for all those who are interested in comparative literature or who simply read literature in the original language not coinciding with their mother tongue, that the issue of universals versus uniqueness is of some interest. Linguistic universals are features that all languages have in common. Some linguists also recognize relative universals that are different in nature and will be discussed after the absolute universals, 11 as the real universals are called in order to distinguish them from the relative ones. Apart from some basic principles the search for absolute universals has not yielded anything. It is evidently important that all languages have a conventional character, that all languages have the means to express actors and actions and can qualify these elements, and that all languages have meaningful units that articulate human experience into discrete elements, the acoustic images of which are composed of discrete non-meaningful units called 'phonemes.' So, one can say, in ordinary words, that in each language the speakers abide by the conventions of that language (calling a table a table and not using another word of their own choice), that all languages can describe an action that takes place and say who or what is the subject of those actions, and that it is possible in all languages by some means to say what French and English say with the help of adjectives

39 Field models, universals, and language-bound world-view and adverbs. The last principle that was mentioned is called the 'double articulation principle,' a corner-stone in the teachings of Andre Martinet. 12 What it means is that larger conceptual wholes are cut up into smaller independent elements. Temperature is thus divided into hot, warm, tepid, cold, icy; time is divided up into verbal tenses. The units resulting from the first articulation are carriers of meaning. On the level of expression (signifier level) each unit of the first articulation is expressed by one sound unit or by a specific combination of sound units: eye lay!, icy /aysi/, 13 hot /hat/, etc., that have no meaning in isolation. That these sound units of the second articulation, or phonemes, do not have any meaning in themselves is easy to see when one examines eye and icy: the lay! combination does not carry any meaning in itself, otherwise eye and icy would have a meaning element in common. Relative universals are more numerous than the very few absolute universals that have a clearly linguistic character. Apart from the case in which a feature A is more likely than not to occur in any natural language, the general formula that applies in a more or less complex form to all cases runs as follows: if a language expresses A, then it is (almost) certain that that language has B. In this formula B may stand for a positive feature or the absence of a feature, the first occurring much more often than the latter. Gender distinction in third-person personal pronouns may serve as an example. If a language distinguishes between masculine and feminine in the third-person plural (A), it most likely also makes that distinction in the singular (B). French elleslils (A), almost implies ellelil (B). From other languages such as English and Dutch it is clear that B may occur without A: she/he (B), they (-A); and zij!hij (B), zij (-A). Russian has fairly recently shifted from A/B to -A/B: one, oni (A) / ona, on (B) to oni (-A) / ona, on (B). For translators this sort of universal is not so helpful, because they may have to deal with two languages, one of which has A and B, while the other has only B, or even neither B nor A. The absolute universals are taken for granted by those who adhere to the idea that each language is unique and that it is impossible to translate from one language into another without severe loss and distortion. The most extreme proponent of the uniqueness idea is Benjamin Lee Whorf. His views combined with those of Edward Sapir, whom Whorf knew and whose lectures he attended at Yale, are known as the Sapir-Whorf hypothesis. 14 According to this hypothesis there is a link between language, thought, and reality. Different realities engender different languages, but it is also true that different languages shape different realities.

40 Linguistics, literary analysis, and literary translation

People are used to seeing the world divided and shaped according to the vocabulary of their language and to experiencing time and space in terms of the grammatical and lexical categories that are at their disposal. So, persons in whose language two terms exist, one for apple and one for pear, are more likely to perceive the general differences between apples and pears than those who have only one term at their disposal for all the fruits falling under 'apple' and 'pear' in the two-term language. As most people have experienced situations in which two persons speaking the same language do not observe the same divisions (e.g., an urban dweller and a farmer talking about cattle or crops), second- or foreignlanguage learners are usually not surprised when they come into contact with these kinds of inter-language differences. The same is not true, however, in the case of grammatical categories. It turns out to be very difficult to imagine another system of representation of time, for instance, than the one prevailing in the mother tongue. The perception of the outside world and the structuring of abstract notions is, then, according to the Sapir-Whorf hypothesis, determined by the native language. Thus, people are prisoners of their mother tongue. Even those who think that they have mastered a second or a foreign language keep the value system of their first language, which continues to shape their thoughts and to function as a frame of reference. Even those who do not go as far as Sapir and Whorf and accept the possibility of really mastering another language in addition to the maternal one do not deny that each language is unique. The clash between the Sapir-Whorf hypothesis and the view that there are universals is more apparent than real, as they do not concern themselves with the same level of examination. Universals are to be found in the domain of basic functions, whereas the Sapir-Whorf hypothesis deals with more refined analyses. The uniqueness of the structuring of each language in regard to the lexicon and grammatical categories appears in the way in which the universals manifest themselves. There is some common ground to start from but from then on each language goes its own way.


Arbitrariness, convention, and motivation

Although the morpho-semantic fields that were mentioned in the previous chapter (page XX) are not very helpful when one wants to give an illustration of how fields are set up and what kind of criteria for setting up a field and structuring it, are used - they are, as was pointed out, without the temporal and linguistic homogeneity that characterize all other types of fields - their organizing principle (expressiveness of sound) needs some further discussion for other reasons. Expressiveness of sound manifests itself in those cases where the signifier has sound qualities that are in harmony with the character of the signified. Sometimes there is direct sound imitation (to crack, to squeak), sometimes the association is carried over from sound to dimension (weeny), to weight (French leger, lourd), to degree of light (German licht, dunkel) and then to mental states (sometimes by metaphors based on one of the previous categories: gloomy, bright). The range goes from straight onomatopoeia to subtle associations, but in all cases there seems to be a link between sound and meaning that refutes the idea of arbitrariness on which the saussurean model of the linguistic sign is founded. Before examining the validity of the refutation we should explain, however, what is involved in saussurean arbitrariness. In the model of the linguistic sign the signifier (acoustic image) evokes the signified (concept), and the signified calls forth the signifier. Thus, as was already discussed (see chapter 5, page 30) a manifestation of the acoustic image table /teb;:,1/ evokes in the mind the idea of a table; we deliberately avoid the term 'representation' because it does not cover abstract notions. When we see a referent that fits into the conceptual category of table for abstract notions the process of evocation of the signifier is less easily described as there are no concrete referents - the signifier presents itself

42 Linguistics, literary analysis, and literary translation

whether it is manifested aloud or not. This solid connection between the two elements of the linguistic sign has to be learned inasmuch as the acoustic image in itself does not reveal anything about the concept and the concept does not give any indication as to what to expect as an acoustic image. In other words, if someone hears the word table without knowing what it means, the sound combination does not provide him with any clues, and similarly, if by seeing a great number of different tables someone ends up forming a conceptual category, nothing in the features of table imposes a special sound combination. If we do not know the name, the characteristics of the concept will not reveal it. A further argument in favour of arbitrariness can be found in the fact that very similar concepts such as pomme in French and apple in English have completely different signifiers, and that very similar signifiers belonging to two different languages are linked with completely different concepts, as for instance /bor/ in American English ('bore') and in French /bor/ (bord: 'side' or 'edge') in French. The absence of intrinsic motivation does not imply that individual speakers are free to choose their own combinations at random. If they did, they would no longer be part of the speech community to which they used to belong, and unless they managed to have their choices accepted by others nobody would understand them, and communication would become impossible. So, the combinations of signified and signifier, arbitrary as they may be, have to be respected by the members of the same speech community whose very existence is based on the fact that its members all respect the same conventions. Yet, motivation plays an important part in literary text analyses and, especially in works about poetry, remarks about inner harmony between sound and meaning abound. The complex question of motivation in poetry will be dealt with later (see chapter 11, pages 69-71); we shall limit ourselves here to a general discussion of motivation from a linguistic point of view. Although motivation covers more than the sound phenomena of onomatopoeia and expressivity and also applies to metaphors and to the morphology of word formation, only sound motivation seriously calls into question the saussurean arbitrariness. The intrinsic difference between metaphorical and morphological motivation on the one hand and sound motivation on the other lies in the fact that the first two categories do not start from scratch, but presuppose the existence of arbitrary forms with which to operate. Thus, the metaphorical use of old hat for 'stale, not original' is motivated, but old and hat are each in themselves arbitrary elements; peach can be used meta-

43 Arbitrariness, convention, and motivation phorically, but the motivation of this metaphor, however easy to grasp it may be, does not diminish in any sense the arbitrariness of the link between signifier and signified. The same is true for the morphological category, where motivation depends on regularity of formation patterns: house and boat are arbitrary but houseboat and boat-house are not, although the relations between the components offer some room for variations: a houseboat is a boat that serves as a house whereas a boat-house is a house for boats and not a house that serves as a boat (compare also a treehouse, that is, a house in a tree, offering yet another pattern of dependency). Sometimes the components differ from the element in isolation, as in French pomme ('apple') and arbre ('tree'), but pommier ('apple-tree'). The regularity of the formation enables somebody who knows pomme, pommier, and poire ('pear') to arrive at the meaning of poirier ('pear-tree'), but even though the term 'motivation' is justifiably used in such cases, the components are all arbitrary in the saussurean sense. In the case of what we have called 'sound motivation' the initial unmotivated signifiers on the basis of which further, this time motivated, signifiers are formed are lacking because it is in the initial signifiers themselves that the motivation is found. The most straightforward examples of sound motivation are the onomatopoeias, where the signifier imitates the sound that is connected with the signified. Thus, the acoustic image 'to gargle' (French gargariser, Dutch gorgelen, German gurgeln) is closely connected with the sound that is produced when the action of gargling takes place. A slightly different type of sound imitation occurs when the signifier reproduces the sound that is occasionally produced by the referents of the linguistic sign. This somewhat laborious formulation is necessary in order to account for the fact that the signifier cuckoo does not imitate the bird, but the sound the bird produces. The 'cuckoo' is frequently mentioned in connection with sound motivation and often a number of signifiers taken from various languages that are not related to one another are quoted to show the general character of the phenomenon.1 Sometimes when discussing the imitative aspect of onomatopoeias, linguists draw attention to the fact that the signifier has to comply with the rules governing the sound system of the language it belongs to, and that it is composed of phonemes taken from the inventory of that language. As phonemic inventories and distribution rules, 2 as well as stress patterns, vary from one language to another and depend on convention, one can say that, at least to the extent that onomatopoeias are subject to the requirements of phonology and stress, they are arbitrary.

44 Linguistics, literary analysis, and literary translation There is a second argument in favour of the saussurean arbitrariness that is not mentioned as often as the first. If we take the name of a wellknown European meadow bird, the Latin name of which is Vanellus vanellus, it becomes clear that, whereas some languages resort to sound motivation, others do not. In English the generic name 'plover' is used and along with it 'lapwing,' based on the way in which the bird flies (the motivation for the second name is of the type that combines unmotivated elements into a motivated compound), or to make the picture even more confusing, a mixture of the morphological type and the sound type as 'lap' goes back to 'flap' which is motivated). The French vanneau goes back to the Latin vanellus and is completely arbitrary. However, the German Kiebitz, the Dutch kievit, and the Russian cibis are clearly in the same category as cuckoo. Thus, the onomatopoeia is not imposed by the special features of the signified and some languages do not use the procedure in instances where other languages offer clear examples of sound motivation. (Compare all the gargle forms and Russian poloskat' .) Conventions differ from one language to another, and although there is an undeniable direct link between the sound-imitating signifier and the signified, the choice to use this direct link is free and arbitrary. Apart from direct sound imitation, motivation is seen in a very large part of the vocabulary and also in some grammatical morpheme changes. The link between signifier and signified is, as in the case of onomatopoeia, a direct one, but it is more elusive and many speakers may not be aware of it. At the beginning of this chapter (see page 41) some examples of expressivity were given without further comments about the sound involved. Regardless of which language we are dealing with, some tendencies manifest themselves in phoneme frequency for certain categories of linguistic signs. Closed unrounded vowels, /if, le!, are more likely than others to occur in words denoting smallness, whereas largedimension words tend to have rounded back vowels, /u/ and fol. In general small is then linked with lightweight and light colour, and, further, cheerful mood and positive moral quality are often included. In the opposite group, heavy, dark, sombre, and morally negative go together. As for the consonants, liquids are overrepresented in words denoting softness and smoothness while stops are connected with hard and rugged.3 Because it is no more than a tendency or a statistical probability and not a steadfast rule that applies without exception, counter-examples abound in every language one examines. Moreover, the distribution of expressive signifiers is not the same for all languages, and an expressive acoustic image in language A often corresponds with a totally arbitrary

45 Arbitrariness, convention, and motivation one in language B. So, here again, in spite of undeniable motivation, the saussurean principle of arbitrariness is not undermined. A special case of sound motivation is found in the morphology of strong verbs in Germanic languages. 4 The present tense shows a preference for front unrounded vowels, the past tense and the past participle, for back vowels: I sing, I sang, I have sung; German, ich singe, ich sang, ich habe gesungen; Dutch, ik zing, ik zong, ik heb gezongen . Here a connection between visible (light) present and not so visible (dark) past for the two groups discussed on page 44 may be assumed, but again the vowel distribution of some strong verbs does not follow the dominant pattern: German ich fange an, ich fing an, ich habe angefangen ('to begin'), ich bleibe, ich blieb, ich bin geblieben ('to stay, remain'), where the past tense has the vowel that one associates with the present tense. Finally, some languages have certain phoneme combinations linked up with special notional areas. English has Isl/ in many words that possess the qualitys of being 'smoothly wet'; slush, slime, slip, slide, slobber, slop. Such combinations are called 'phonaesthemes' to underline their aesthetic expressive function or 'meaningful submorphemic elements.'5 Although it is interesting to note that if one accepts this name it becomes inevitable that one must reassess the definition of the morpheme as the smallest meaningful element, the main problem lies not in the relation between the morpheme and the submorphemic element, but in the meaning of the submorphemic element. It is extremely vague, and individual speakers may identify it in different lexical units, one rejecting the analysis that singles out the submorphemic element, and another accepting it without the slightest hesitation. Sound expressivity, then, is not rigorously codified and is often controversial. Yet, it is a very important element in literary language, especially in poetry, and preserving sound expressivity in a translation is a major problem for which there are no easy solutions. When discussing translation we shall mention yet another aspect of the relation between sound and meaning, namely the psychoanalytical interpretation of phonemes (see chapter 15, page 101). This approach raises the question of intentionality and lies outside the domain of functional linguistics. It is for that reason that it has not been included in the survey of types of sound motivation given in this chapter.


A stumbling-block in semantics: discreteness and gradual transition

One of the basic features of the elements with which structuralists are used to operating is their discreteness. At each level of analysis welldefined finite units form a structured inventory. By choosing a particular unit the speaker excludes all other units that could occupy the same position in the chain of speech. By selecting /r/ in the Ir/at is under the table, one discards /kl (cat), /bl (bat), /f/ (fat), /n/ (gnat), etc. Similarly one can select a past-tense morpheme he [want]ed, and by doing so exclude the present and the future. Illustrations of the structuralist theory are mostly provided by phonology and by morphology, where inventories are limited and closed. In the case of the lexicon, examples are often taken from fields (see chapter 6, page 37) where the constituent units do not overlap. In most instances, however, lexical items are not clearly and neatly separated from one another but share some common area with other items. 1 It is clear that the either/or choice of phonology and grammatical morphology does not apply here. Whereas a change of vowel results in a different lexical item, a change of lexical item does not always result in a different message: meat, moat, moot, mit, mate are different lexical items, but whether the substitution of loatlisome for repulsive in 'His behaviour is repulsive' changes the message is a matter of personal interpretation. Such interpretations may even vary for the same individual according to context and circumstances. For those who distinguish between the two messages, it will not be easy to find a common answer as to what the difference consists of. The method of breaking down each lexical item into semes does not yield an indisputable result for the following reason: semes more often than not do not have a recognizable counterpart at the level of the signifier, so their number and their identity are not fixed;

47 Discreteness and gradual transition therefore, the analysis into semes may vary from one person to another and there is no possibility of telling who is right and who is wrong. One could, provided a large enough number of persons are willing to take part in the experiment, try to find some statistical feature by asking questions such as: Check which of the following is or are acceptable to you: ] ] ] ]

a. b. c. d.

His His His His

behaviour is behaviour is behaviour is behaviour is

not repulsive but loathsome. not loathsome but repulsive. loathsome and even repulsive. repulsive and even loathsome.

If a and b are rejected by a majority one could conclude that the opposition between the two elements is not strong enough to warrant the use of the construction 'not x but y.' If a majority accepts c, repulsive should be considered to be stronger than loathsome; if d is preferred, loathsome would be the stronger of the two. This rather laborious procedure for discovering at least some of the semes that make up the semantic content of a lexical unit has unfortunately some flaws and does not guarantee a result that reflects the actual state of affairs. 1 When asked to choose, people tend to think that it is better to make a choice than to say that they are unable to distinguish between two or more options. 2 In the same vein, people do not like to admit that one or more of the elements between which they are asked to choose is unknown to them. 2 In the case of repulsive and loathsome, for instance, the latter adjective may be unknown to some speakers, who group repulsive with disgusting, disgraceful, abject, awful, or with only some of these. Many variations in inventory are possible and a term that is marked 'strong' in one inventory may not be as strong in some other grouping. Thus, the principle of discreteness is not very relevant for the semantics of some parts of the lexicon, it seems, and rather than presenting sharp contours the units have blurred edges and large overlapping areas. The continuum is not always neatly divided into discrete units and each unit is not always neatly divided into semes. It is for this reason that discussions about the meaning of lexical elements - in a specific context or not - have to take into account the lexical inventories of those who take part in the discussion in order not to become a dialogue of the deaf. If the denotative level presents serious problems for an exhaustive division into semes, the situation becomes even more difficult when

48 Linguistics, literary analysis, and literary translation non-denotative elements are taken into account. First, it has to be decided whether to include these elements at all in a semantic description. If the answer is negative, the next question will concern their status in a linguistic description; if the answer is positive the inevitable consequence is to weaken the position of the discreteness principle even further. In either case the difficulties of accounting for these elements remain the same, and the question of whether or not they belong to the level of semantics is ultimately of minor importance. Similarly one does not solve the problem of describing them by ascribing everything that does not belong to the domain of denotation a paralinguistic or nonlinguistic status. After these general statements that leave matters in limbo, it is necessary to take a closer look at the non-denotative elements themselves. It is no easy task to classify and order all the elements that contribute to forming a message, without denoting anything comparable to the signified. Often it is impossible to tell, without knowing the exact circumstances in which a given element manifests itself, whether we are dealing with attitudinal information or social data about the speaker. If the latter, one has to decide if the social element is relevant for the written text or oral messsage. It is also important to notice that some of the data provided in an oral message do not remain intact in the written transcription of the oral form. Finally one should keep in mind that some of the elements that we are about to discuss manifest themselves at different levels of a language and that it would lead to inconsistencies in an analysis if they were included at one level but ignored at another. Pronunciation illustrates most if not all of the problems touched upon in the preceding paragraph. In most cases the pronunciation of an individual speaker yields a good deal of information about his or her geographic and social background, as well as about sex and age. As a rule, however, a written verbatim account follows the ordinary spelling and does not even try to keep the information of the oral message. The different components of oral speech are subject to distinct treatments, when represented in writing. Pitch is traditionally disregarded, and if it is important for the reader to get some information about it, the writer has to resort to metalinguistic comments that are added to the oral original text. Intonation fares slightly better, yet graphic means are rather limited: punctuation, italics, and some less general devices. 3 If pitch and intonation are disregarded, what remains is what is usually called the 'pronunciation.' Some languages use a standard spelling that comes close to the phonological ideal of one grapheme or graphic sign

49 Discreteness and gradual transition for one phoneme; other languages have systems that, over the centuries, have become more and more distinct from the pronunciations they were supposed to represent. Usually it is the pronunciation that changes over the years, while the orthographic system lags behind. A special category is formed by languages that do not use alphabetic systems, but instead use ideograms or pictograms that have no connection with the pronunciation of the oral message they represent in writing. It is obvious that pictograms and ideograms are unfit for signalling deviant pronunciations. 4 This is easy to understand when one thinks of the current system of representing numbers. Even if we wished, we would not be able to indicate the deviant pronunciation tree for three in a direct quote if we used the number sign '3.' In order to report the pronunciation deviation, alphabetical spelling has to replace the number sign. This does not mean, however, that spelling always provides adequate means and, as we shall see shortly, many dialect features inevitably get lost unless special diacritic signs are introduced to complement the alphabet. Except in specialized works on phonetics these signs are not used, however, so we shall limit ourselves to graphic representations that are restricted to the graphic signs of the standard written language. First, the distinction between phonemic and phonetic deviations should be mentioned. Phonemes are distinctive units in the sense that if one phoneme is replaced by another in a signifier, the signified changes as well.5 Sometimes the resulting sound chain does not have a corresponding signified: / - /k.el/ (cat - cal?); sometimes there is a change of meaning: /k.etl - /k.esl, (cat - cash). If for some reason a person does not have all the phonemes of the standard language at his disposal, but uses one phoneme to represent two different phonemes of the model that prescriptive grammars offer, graphic representation may be difficult. If the unique phoneme coincides with one of the two phonemes, graphic representation is quite simple. In a substandard variant of Dutch, spoken in Amsterdam, Isl represents both Isl and lzl of the standard language, and so hij zegt ('he says') can be faithfully represented by hij segt as far as the simplification Isl for Isl and lz/ is concerned. (If the reverse happens and the deviant variant has an extra phoneme there is no such easy solution, of course.) If the one phoneme of the variant differs from both the phonemes of the standard language, finding itself somewhere in the middle, graphic representation often reflects the impression of standard speakers that the wrong phoneme is being used. So, the lrl phoneme of a Chinese or Japanese person speaking a language where lrl and Ill are two distinct units gives the impression of being an Ill in words like

50 Linguistics, literary analysis, and literary translation

right, arrow, and /r/ in light, flight, alley. Graphic representation, then, does not reflect the speaker's oral speech, but rather the hearer's impression of that speech. Phonetic deviations, where no change in phonemic value is involved, are much more frequent than phonemic discrepancies. In this purely phonetic category the phoneme inventories are the same, but some phonemes are pronounced in a noticeably different way. A few examples taken from English and French will illustrate some of the aspects of rendering phonetic deviations in writing. If an Englishman speaks French, one of the features of his foreign accent is likely to be the aspiration of initial stops before vowels: p, t, k pronounced [ph, t", kh]. Adding an h to indicate this aspiration works very well for t: thous les jours, unless standard spelling already has an h: le the, la these. For p the difficulty lies in the use of the digraph ph for the !fl phoneme in words of Greek origin, while /kl, when represented by c, runs into the problem of ch coinciding with/~/: chanter. Thus, car would get confused with char, if the aspiration [kh] were indicated by adding h after the c. The possibilities of indicating aspiration in French spoken by an Englishman or an American are thus limited, but when a Frenchman speaks English without the usual aspiration, no spelling device at all is available, as English spelling does not indicate the aspiration element and uses simply p, t, k, etc. Another example where spelling fails to provide phonetic information both in English and in French is the phoneme /r/. Both languages have only one phoneme /r/, but besides the standard pronunciation they both offer a variety of dialect deviations. Because there is only one grapheme that can be used, the reader who wants to know how a dialect text would sound when read aloud is reduced to guessing, as far as the r is concerned. That most inhabitants of Quebec City pronounce r more or less as in standard French, whereas the majority of Montrealers roll the r, using the tip of their tongue, is a fact one has to know in order to give a phonetically correct oral interpretation of a dialect text representing French spoken in Quebec City or in Montreal, because the spelling cannot reflect any particularities the respective r pronunciations may have. 6 The variants of standard speech, whether social or geographical, or both, are encountered in pronunciation, in the lexicon, and in morphology and syntax. So, if one decides to introduce a seme for social level or for rural origin when dealing with the lexicon, apart from the already mentioned difficulty at the level of the signified (see pages 47-8 above) the same kind of semes would have to be assigned to morphological, syntactical, and pronunciation variants (phonetic as well as

51 Discreteness and gradual transition phonemic). Unfortunately some pronunciations that are traditionally considered to be substandard, such as French quat [kat] for quatre [katr], coincide with neutral pronunciation in rapid speech. Moreover, levels of speech are not separated from one another in such a way that mixing of levels never occurs.7 Not only may a speaker use standard lexicon, morphology, and syntax while using substandard pronunciation, but it is also possible that speech samples of one and the same person will present features belonging to different speech varieties within the same category, be it pronunciation, the lexicon, or another category. The same observation can be made for archaic and innovative features that often have a social or geographic aspect as well. Even if each element, taken by itself, can be considered discrete, the picture gets blurred when we consider a larger sample of oral or written text where elements belonging to different varieties of the language that is used occur side by side. Instead of clear-cut divisions between the different sociolects and dialects, the transitions from one yariety into another are gradual. Moreover, the same feature may characterize social level, geographic origin, and generation; so it becomes very complicated to arrive at the correct interpretation of such a feature without the help of supplementary linguistic and even non-linguistic data. In summary, we may say that directly observable data at the level of the signifier, whether provided by discrete elements (lexicon, morphology, syntax, and phonology) or by phonic phenomena that cannot be structured into a range of discrete units but have to be considered as forming a continuum (phonetics), do not provide unequivocal information about the speaker. Moreover, some of the features of a phonetic or even phonemic character - an extra phoneme - cannot always be expressed when the usual alphabet is used. (None of these features are maintained in a pictographic or ideographic system.) Sometimes sociolect or dialect plays an important role in shaping a message, at other times the impact, if there is one, is negligible. In the latter case translators as a rule disregard the special markers, and their target text conforms to the rules of the target language. In order to be able to decide whether to look for social or dialect equivalents or instead to revert to standard language, the translator has to determine what the status of the elements in question is in the source text. Saying that they have to be maintained in translation when they are relevant in the original does not help very much if we do not have criteria for deciding when they are relevant. In the next chapter some such criteria will be discussed and their reliability and limitations examined.

9 Intentionality and relevance

When we analyse a text a question that often arises is: What was the intention of the author when choosing the forms we have in front of us? The answer is often influenced by the overall purpose of the text political pamphlets, advertising, didactic writings - under the assumption that the details are subordinate to the whole, and if a detail does not fit it may even be ascribed to an error of judgment of the author. In literary texts, however, aim and means are usually not so easily defined and detected, and unless the author himself explicitly declares what he wants to achieve with his work, the reader is reduced to guessing. Different readers may come up with different hypotheses and each hypothesis will colour the interpretation and evaluation of the smaller details of the text. Even when the author has made a statement about the aim of his work readers may wonder whether such a statement is trustworthy. So, even if the level of intent' of a work is crucial for its interpretation and translation, in many instances analysts and translators prefer not to rely too heavily on it and to use other criteria as well. Instead of the intentions of the author, which are hidden, the relevance or pertinence from the reader's point of view may provide some guidelines. However, as this relevance sometimes depends on the reader's understanding of the intentions2 of the author, the problem of being unable to verify hypotheses about these intentions remains. Intention and relevance are not, as the preceding remarks may have suggested, linked in such a way that one can be derived from the other. A writer may be totally unaware of some constant obsession that is manifested in his work and turns out to be one of its most relevant features. However, overt intentions - overt in the sense that the writer is aware of them and sometimes even makes them known to others -

53 Intentionality and relevance may not result in anything the reader considers pertinent. Of course, one could argue that the failure in itself is an important fact, but as it is not part of the text and depends exclusively on outside information that the reader may or may not receive, we prefer not to include this kind of failure in our discussion of intention and relevance. Before examining the intricacies of literary texts a few examples taken from oral-communication situations and from non-literary texts will be given by way of an introduction. As it is impossible to be exhaustive, the readers are asked to complement these examples with others taken from their own experience and to add categories that may have been overlooked or deemed not important enough to be included. The expression of emotions provides a good example of the complex relationships between the intention of the speaker/writer and the relevance of the signals for the hearer/reader. Inasmuch as intonation, and for some emotions voice volume as well, plays an important role, oral messages contain more indications than their written counterparts, and written messages are, as a rule, more controlled, so one has to keep in mind that only part of what is true for the oral level also applies to written texts. When someone is angry this mood may manifest itself in voice volume, intonation, and lexicon. If there is a high degree of emotion, the syntax may even be affected to the point of incoherence. All these indices are noticed by the hearer and interpreted as signs of anger. So far everything seems fairly straightforward. However, problems arise when the intention of the speaker is at issue. The following possibilities arise: a The speaker is angry and wants to make this clear. The raising of the voice, the intonation, and the lexical choice are all deliberate. The syntax remains unaffected. b The speaker is angry and does not want to show it. In spite of himself he gets worked up and carried away to the point where even syntactic rules break down. Are we allowed to attribute this to a subconscious desire that goes against his conscious efforts or not? It is obvious that psychology and psychoanalysis cannot be avoided when answering this question. c The speaker is not angry, but his idiosyncratic use of language is misinterpreted by the hearer. The features that normally signal anger belong to the neutral register of the speaker. Insiders who know this discard the usual interpretation, but outsiders are likely to err. d The speaker is not angry, but for some reason - say re-establishing quiet and order in a class-room, intimidating adversaries in negotia-

54 Linguistics, literary analysis, and literary translation tions - he pretends to be angry and uses some or all of the indices (except syntactic incoherence) to convey the impression. Most readers will remember class-room situations in which they wondered whether the teacher was really angry or was merely feigning anger. Even more complex situations can be imagined where the speaker thinks that he is putting on a show, but actually expresses his real feelings, 3 which are not recognized as such by the hearer, who believes the speaker is pretending. From this short enumeration it is sufficiently clear that no simple, generally valid scheme for the relation between expressing and perceiving an emotion can be given. In literary texts this uncertainty can be exploited to create tension and misunderstanding between characters. The Freudian slip or Fehlleistung was not unknown before Freud as a device to give away hidden emotions and preoccupations of fictional characters, although it obviously did not have that name. After Freud the Fehlleistung has been used more frequently as a device in literature, but instead of being proof of the psychological insight and intuition of the writer, it now often only testifies that the author has some knowledge of Freudian theories and applies them dutifully. In any case, the linguistic patterns of the Fehlleistung, where the replacing signifier often resembles the replaced while the signified is either the opposite of or totally different from the signified that was intended, have not changed. 4 When there is similarity between signifiers, translation may be difficult, but unlike puns and word-plays, slips can occur without such similarity at the level of expression. With socially substandard speech the picture is somewhat different from what we have just seen for emotions, although emotion may be a factor in the use of substandard language. The most important distinction is that contrary to the emotional aspect of language, the substandard aspect is in many circumstances irrelevant. Another characteristic linked with irrelevance is that substandard features are often used unintentionally: a The speaker uses substandard forms and/or pronunciations as they are normal for him; the hearer disregards the deviation from standard usage because 1 / the topic is neutral - a technical discussion, a lecture - or 2 I he knows the speaker talks that way. b The speaker uses substandard features unintentionally, but the hearer misconstrues the deviation from the prescribed norm as a lack of respect to him. c The speaker uses substandard features consciously or half consciously to adapt his level to that of his interlocutor.

55 Intentionality and relevance d The speaker uses substandard features to express his contempt for his interlocutor. e The speaker uses substandard features to express his negative feelings towards the subject he is talking about. f The speaker is aware of substandard habits and tries to suppress them, but is not entirely successful in his effort, thus giving away both his social background and his desire to hide that background. g The speaker is aware of the difference between substandard and standard, and deliberately uses substandard features in order to express his allegiance to a social group or cause. Most of the above examples also apply to situations where the deviation from the norm is geographical rather than social: a The speaker uses dialect because the dialect is his natural way of expressing himself; the hearer may have some trouble understanding, but is aware of the fact that the speaker does not have any special intention in using his dialect. b The speaker uses dialect forms to adapt himself to the speech of his interlocutor. This adaptation is often deliberate, but may be an automatic reaction when dialect and standard language have been equally mastered. c The speaker who is aware of the prestige of standard speech tries to suppress the dialect features that are felt to be negative; not being entirely successful in his endeavours, he shows his geographical origin as well as his hang-ups about this origin. d The speaker deliberately stresses the features that distinguish his dialect from standard speech in order to assert his geographical origin; in this case local resentment against a norm-setting capital city and regional political aspirations often play an important role. 5 If more than two speakers are involved, a dialect or substandard sociolect is sometimes used in order to exclude the speaker(s) who cannot express himself (themselves) in it (or in the extreme case cannot understand it). 6 The social codes prescribing certain variants of a language for certain occasions add to the complexity of the situation regarding sociolects and registers. This codification may be fairly loose as it is in North America, more rigid as in many European societies, or even extremely strict and precise, as in countries of the Far East such as Japan. It is not always possible to interpret the examples we have just given unambiguously; thus, intentionality and relevance, not having any precise formal criteria that allow them to be identified, often remain sources of disagreement and dispute.

56 Linguistics, literary analysis, and literary translation So far little has been said about literary texts, apart from a few remarks about the expression of emotions and the role of Fehlleistungen giving away hidden feelings and preoccupations of protagonists. It is obvious that when used by a writer in order to tell something about a character the indices are intentional.7 The same is true for features occurring in directly quoted speech that indicate the social or geographical background of the fictional person who is speaking. There are, however, certain discrepancies between intentionality and relevance that deserve some attention, and there is also a peculiar instance where the intention of the writer is picked up correctly by the reader, in spite of the fact that the means the writer uses are wrong from the point of view of linguistics. In many cases the linguistic characterization of a hero is not maintained consistently throughout a text. After a time, when the background of the character is sufficiently established to need no further linguistic reinforcement, direct quotations lose the specific features used earlier. Independent of the degree of persistency of the characterization, there is a scale of linguistic accuracy ranging from a rendering as faithful as the traditional supply of letter signs allows (see chapter 8, pages 4~50) to the use of one or two salient features. Writers may err in their representation, but unless the reader is an insider the errors will not be noticed and the writer's intention will be interpreted correctly. The extreme case of incorrect rendering occurs when the writer uses a phonetic representation of standard speech in order to convey the impression of deviant pronunciation habits, while using elsewhere non-phonetic customary spelling. Guy de Maupassant falls into this category when he writes: '11 est v'nu un m'sieu trois fois.' The omissions of mute e in venu and monsieur, as well as the dropping of the silent r in final position in monsieur, are in perfect agreement with standard pronunciation habits, yet Maupassant uses the phonetic, or rather partially phonetic, spelling to indicate the special, presumably substandard pronunciation of a maid. 8 In passages in which the author or the omniscient narrator relates events or describes settings, socially or geographically marked features have a different function and do not characterize speakers. Sometimes a special effect is created by the harmony between language and subjectmatter: when that harmony has a local rather than a social base the term c.ouleur locale applies even though it is generally used for foreign-language, rather than for dialect, elements. In some instances there is no standard equivalent for a lexeme (word) or for an idiomatic expression of the dialect, but often the dialect is chosen for other reasons. These reasons also underlie the dialect choice in works where dialect

57 Intentionality and relevance is exclusively used and does not alternate with standard language. However, the reader is supposed to make comparisons between the two and thus to grasp the special flavour of the dialect. It is a debatable question whether dialects are more expressive, more pictorial, than standard language, or rather provoke in the cultivated reader a sweet nostalgia not entirely free of condescension. Special mention should be made of writers who use dialect or substandard forms but want to make sure that the reader recognizes them as such while at the same time being aware of the fact that the writer does not use them in his ordinary everyday language. Italics and quotation marks are used for that purpose, thus offering an example of a very simple means in written texts which has no equally simple counterpart when the text is presented orally. The relevance is stressed and the intentionality is unambiguously indicated. When italics or quotation marks are used in an inconsistent way, the interpretation of a dialect element that has not been typographically marked may cause some problems, as will be seen in the discussion of some Quebec novels (see chapter 17, page 116). Until now, only the term literary text has been used without any further specification. There are some differences, however, between prose and poetry, and it is interesting to examine whether or not these differences give intentionality and relevance a special position in the analysis of poetry. In spite of different approaches and terminologies there is little disagreement about the following points among those who study poetry: 1 Form features are very important at various levels: a. rhythm and rhyme (unless we are dealing with free verse); b. regular patterns of verse length and division into stanzas; c. expressivity of sound, harmony between sound and meaning. 2 The denotative value of the signs is often subordinated to non-denotative elements. 3 From language to language and period to period there is a special poetic vocabulary, which varies in importance. Whereas ta and tb are intentional, 1c may not be deliberate. The same is true for all sorts of special features that, although heterogeneous, all fit into category 2. In regard to category 3, we are dealing with a special code the use of which is likely to be intentional. Categories 1c and 2 especially lend themselves to inference or Hineininterpretieren on the part of the reader and, even more, the literary analyst who tends to ascribe his own discoveries when exploring a text to a

58 Linguistics, literary analysis, and literary translation conscious choice or, using a recently highly overworked term, to a strategy of the author. Whether deliberate, subconscious, or fortuitous, features 1c and 2 may be perceived as pertinent. If the reader happens to be the translator of texts containing such features, he may find it extremely difficult if not impossible to preserve them - or come up with an adequate equivalent - in the target text inasmuch as many of these features are either language-bound or culture-bound, or both. (For a discussion of some of the problems, see chapter 16, page 110, where a poem by Heinrich Heine is examined from the point of view of the translator.) In conclusion, in many instances relevance and intentionality turn out to be dependent on individual qualities of the speaker and/ or hearer, instead of being unequivocally coded in the language system. One of the major sources of misunderstanding between linguists and literary analysts is their difference of perspective. Whereas linguists look for the general procedures a language has at its disposal to give certain elements of a sentence a prominent position that stresses their importance, literary analysts examine individual texts and are often more interested in the special devices a particular author uses for the foregrounding of the most important parts of his message. The means that are part of the langue are taken for granted by the latter while the parole of the text under study is examined in order to find out what gives it its special, unique character. This misunderstanding is reinforced by the fact that linguistics and literary analysis often use the same terminology without giving the same definitions. So, the similarity of the signifiers hides the dissimilarity of the signifieds. In the next chapter the short-lived love affair between linguistics and literary analysis will be introduced in order to give some insight into the terminological confusion, which goes back to the early sixties, at least.

10 Terminological confusion

It is a fairly common phenomenon that when terms in use in one discipline are borrowed by another discipline they undergo a shift in meaning. It is equally common that within a particular discipline terms fluctuate in meaning, being used in different ways by different people. So, it does not come as a surprise that linguistic terminology when used in literary analysis should not be understood in its original meaning without checking. This checking is easier said than done because within linguistics the value of a given term may vary greatly, and literary critics usually do not indicate from which linguistic school the term they use has been borrowed. To make things even worse there is often some uncertainty about the precise meaning of a term as used by a specific linguistic school, and finally there are sometimes terminological inconsistencies in the writings of even a famous and much-quoted linguist. Because many of the terms that are widely used in linguistics as well as in literary analysis go back to Ferdinand de Saussure, it is appropriate to start with a second look at the saussurean linguistic sign, which was discussed above (chapter 5, page 30). In Saussure's work there is some wavering in the use of the terms signifier and sign. 1 In principle the signifier is linked to the observable element, whether in speech or in print, and internalized as an acoustic image. This acoustic image forms, together with the signified (or concept), the linguistic sign. Thus, both signifier and signified have to be seen as abstractions: the acoustic image allows one to recognize differently pronounced units as belonging to the same signifier, and the concept 'tree,' for example, incorporates all the different tree species as well as the different trees belonging to the same species. In both the signifier and the signified there is, thanks to the abstraction, an element of generalization without which linguistic

6o Linguistics, literary analysis, and literary translation

communication would be extremely cumbersome, if not impossible. If the sign tree occurs in an utterance it most likely refers to a specific tree, and this tree, belonging to the outside extra-linguistic world, is called the referent. One should not confuse the referent with the concept because they belong to totally different categories. This distinction is easy to see when we are dealing with concrete notions such as tree, but it becomes more difficult to separate the two in the case of abstractions such as 'immortality.' The confusion between signified and referent is increased by the fact that Saussure himself sometimes uses the term sign when speaking about the signifier. 2 In this way the original sign consisting of the two indissolubly linked parts - signifier and signified - is in danger of losing its identity, and the next step is to link the new sign (signifier) directly with the referent. This direct relationship bypasses the level of abstraction and generalization, thus undermining what was the comer-stone of Saussure's concept of language. For literary analysis it is very tempting to follow this erroneous interpretation of the saussurean sign, as it allows one to concentrate on unique relationships expressed in a particular text without paying attention to the question whether the relationship has general validity and belongs to the langue level, or lies outside the underlying code and therefore does not have a place in the langue stratum. As soon as individual messages are studied - literary texts being a special kind of message, the term message is appropriate here - without taking into account the aspects of abstraction and generalization that characterize the linguistic sign, accessory features of the signs may come to the fore. Moreover, context and extra-linguistic data may induce the analyst to postulate features (semes) that do not belong to the sign itself but are conditioned by the circumstances of its occurrence. In any case, readers' reactions and interpretations become much less conditioned by an underlying code, a fact that is reflected in some recent reader-oriented theories of text analysis (see chapter 11, page 68; chapter 13, pages 83-4; chapter 16, page 105). Unfortunately the terminology does not indicate whether we are dealing with an individual reaction, or with a linguistic analysis of general validity. The fact that many literary critics believe in the general validity of their personal interpretation does not help to clear up the confusion either. Another factor that gives rise to misunderstandings is the mixing of terms originating from different linguistic schools. The term symbol offers a good example of this mixing. Before looking at other interpretations let us examine the place that symbol occupies in saussurean linguistics.

61 Terminological confusion

Saussure and his school make a distinction between symbol and signifier. Sometimes a third term, symptom (French indice), is included in the discussion. Finally, there is a fourth term, namely signal, that shares some characteristics with symbol and some with symptom, thus blurring distinctions that are not always easy to make anyway. When motivation and conventionality are used as parameters, however, some differences transpire fairly clearly: a. a signifier is by definition conventional and not motivated; b. a symbol is conventional and motivated; c. a symptom is not conventional and not motivated; d. a signal is conventional and may or may not be motivated. Some examples will illustrate the differences between the four notions, and some explanatory remarks will complement the succinct statements given under a, b, c, and d. The conventionality and arbitrariness of the link between the signifier table and the concept 'table' has already been explained (see chapter 5, page 30; chapter 7, page 41). The acoustic image does not give away any of the characteristic features of the concept, so in order to know what table means one has to have been told and thus been made aware of the convention. In the case of a symbol - the maple leaf or the beaver standing for Canada - the link is motivated, as there are countless maple leaves in Canada and quite a few beavers. Yet the link is conventional: there are probably just as many pine needles as there are maple leaves in Canada, and many animals outnumber the beaver. A racoon would do just as well, although a dam-builder living in the wilderness is more respectable than an animal that has conquered the city and lives on garbage. It should be noticed that two different symbols may stand for the same symbolized notion and that one symbol can have more than one function: a dog symbolizes faithfulness but may also symbolize servility. A symptom falls into a category that is totally different from the one signifier and symbol belong to, because it lies beyond the realm of convention that is controlled by the human mind. Whether or not a symptom is noticed and correctly interpreted does not affect its status as a symptom. Whether or not a rash is correctly connected with the allergy that causes it, it is a symptom of that allergy. If one keeps in mind the special relationship between symptom and cause, it does not matter which label one gives to it, motivated or not. If one chooses to speak of motivation one should add that it is not the same kind of motivation as that pertaining to symbols. However, if one rejects the idea of motivation in the

62 Linguistics, literary analysis, and literary translation

case of symptoms it should be pointed out that the categories of motivation and arbitrariness do not apply to symptoms. Signals form a mixed category of which the only constant feature is conventionality. Sometimes a signal goes back to a symptom. Steam being a symptom that water has reached its boiling point is used to produce a sound (whistle), signalling that the water is boiling. Very often signals have the sort of motivation one can observe in pictograms: a skidding car represents the danger of a slippery road. Road signs and traffic signals are often a mixture of symbolic and arbitrary elements: the image of a bicycle means that there is a path for cyclists. This signal may be reinforced by a linguistic message. Some traffic signals such as traffic lights - red, amber, and green, and the advanced flashing green - are purely conventional and use neither pictorial representations nor linguistic reinforcement.3 So, even if signifier and linguistic sign sometimes get confused and signal and sign are both used in everyday language without precise definitions, leading to vagueness and overlap, it is possible to get an idea of what each term roughly represents in saussurean and post-saussurean structuralism and functionalism. Unfortunately in other linguistic schools 4 the term symbol is widely used with approximately the same meaning that signifier has in the saussurean linguistic-sign model. If the meaning that should be attached to the term when used in linguistically oriented literary criticism had been specified no harm would have been done, but definitions or specifications are rarely supplied. To add to the problem much of what is said about the expression level deals with sound motivation, which is also called 'sound symbolism.' This type of symbolism is completely different from and has very little in common with the symbolism of the maple leaf and beaver just discussed. Sound symbolism is not conventional: it doesn't have to be learned and it is not possible to indicate for a list of symbols (sounds) what exactly they symbolize. Some people are more receptive than others, and while, in the case of conventional symbols, it is possible to say, for example, when somebody thinks that the birch leaf is the symbol standing for Canada, that that person is mistaken, right and wrong do not apply to sound symbolism. As a phenomenon sound symbolism does not appear to be limited to a specific language, or language family. (See also chapter 5, page 30; chapter 7, pages 44-5; chapter 11, pages 69-70.) Another area where terminology varies from one school to another and may cause some difficulties is that of signification and valeur as they appear in the French text of the Cours . Saussure's example is very clear:

63 Terminological confusion in the sentence 'The sheep was grazing the grass,' sheep has the same 'signification' (meaning) as the French mouton in 'Le mouton broutait l'herbe.' Yet there is a difference in 'valeur' (value) because sheep can only indicate the live animal, while mouton can also refer to mutton, the meat of a sheep prepared for consumption. On the syntagmatic level, i.e., in the utterance in question, there is equivalence, but on the paradigmatic level, i.e., in the inventory, sheep and mouton are not identical because sheep has a more limited domain than mouton . So, for Saussure the value of an element is not context-bound and the signification is. When critics speak of the special value of an element in a given context, they obviously do not have Saussure's interpretation of the term valeur in mind. 5 Apart from this source of confusion, which is purely terminological, Saussure's example creates a more serious problem for semantic description and text analysis: does the value have any impact on the utterance level and, if so, of what does it consist and how do we include it in a text interpretation? Saussure's example seems to rely heavily on the referent when it is declared that there is identity on the syntagmatic level, and the context and extra-linguistic situation are crucial. 6 This way of conceiving of semantic analysis is even more explicitly defended by those linguists and philosophers who declare that there is no meaning at all without a context and a situation. Isolated terms do not have any meaning. At the extreme opposite one finds the opinion according to which terms do have a wide range of potential meanings. When they occur in a speech act, context and extra-linguistic situation eliminate most of the potential meanings. As a rule only one meaning is left. So, even if the two views are quite different, both ultimately rely on context and situation to arrive at the semantic interpretation of an element in an utterance. Yet the difference is very important, because the second view suggests that other meanings have been eliminated. If these meanings leave a trace in the mind, the saussurean value does have an impact on the syntagmatic meaning. In most instances such a trace would be an extra burden and is therefore not allowed to reach the conscious level, even if the short cut that comes close to or coincides with the first view is not taken. For texts that are denotative the value element does not play any role, and it does not matter very much whether the absence of this element is the result of a direct short cut or a subconscious elimination process, or can be explained by positing the formation of meaning in context and extra-linguistic situation. When literary texts are involved the denotative, directly given message shares the

64 Linguistics, literary analysis, and literary translation limelight with what is not immediately accessible, and value elements not favoured by the context may give hints to different interpretations. In this respect value elements do not differ from associative and connotative elements. As value depends on polysemy, however, it is more general and belongs to the language system, while connotation and association are less codified - especially association - and are more individual. When all the terminological inconsistencies, the fuzziness of some of the concepts, and the overlap are taken into account, it is hardly surprising that literary analysts do not always observe the necessary rigour in the use of linguistic terminology. The terms of different schools are mixed: sometimes the same term is used alternately in its technical meaning and in its everyday interpretation. It also happens that in a given passage a term used in the beginning is farther on equated with another term that is not at all equivalent when used by linguists. These shifts add to the lack of precision and create the impression that the terms have the effect of being an incantation rather than the means to express a coherent well-structured theory, or at least fitting into such a theory. In the next chapter some examples of runaway terminology and of literary analysis with a linguistic underpinning, and an attempt to incorporate all non-denotative elements into a semantic theory, will be discussed. Because of the general theoretical tenet of that attempt we shall begin the chapter by briefly examining it.


Short of denotation and beyond it

Not every text and not every oral utterance requires the same attention and the same interpretive effort from the receiver/decoder. Sometimes the only function of an utterance is to manifest the willingness of the speaker to communicate with the listener. This function, called the 'phatic function,' is characterized by the fact that the denotative semantic content of the utterance is of little importance. Greetings, routine inquiries about the interlocutor's well-being, and remarks about the weather belong to this category. Routine inquiries such as 'How are you?' do not solicit an answer, and that is the reason why the answer is invariably: 'Fine!,' no matter how miserable the answerer feels. Dutch people are more likely than their Anglo-Saxon counterparts to give an honest answer. Thus, in Dutch the denotative content of the question is not weakened to the same degree as in English. For translators this difference may cause some problems. Formally phatic utterances may be identical with utterances with full denotative function but sometimes they are characterized by indistinct pronunciation. French offers a curious example of special treatment of a phatic utterance: the t of comment is never pronounced in liaison with the vowel of the following word:1 Comment I a-t-il resolu ce probleme? [koma atil rezoly s~ problem]

but liaison takes place in the question Comment allez-vous? [komatalevu]

That comments on the weather have a phatic character is easy to un-

66 Linguistics, literary analysis, and literary translation derstand as they are usually made to people who are fully aware what the weather conditions are like and do not need to be reminded of them. The contact element is not so frequent in written texts. Although letters may have features that fall into the phatic category when they are written with the sole purpose of establishing a contact, reminding the recipient of the existence and the friendly disposition of the writer, it is rare to find a letter as reduced in denotative content as are oral phatic utterances. There is no equivalent in written text of the truncation and slurring that sometimes take place in the phonic expressions of phatic messages. Whether one classifies these kinds of oral and written texts together or - because of the differences that exist - relegates them to separate categories, both show what Uriel Weinreich called 'subdued semanticity.' This subdued semanticity is opposed to what could be labelled as 'fullfledged semanticity' and to 'enhanced semanticity.' For the latter one also encounters the term hypersemanticization. Full semanticity characterizes oral and written utterances where denotation is fully transmitted - the speaker/writer has the intention to transmit and the hearer/reader interprets the message in accordance with that intention. No hints are encoded or decoded, and elements that in other texts are interpreted as additional information are not noticed. This denotational purity may never be reached in actual communication because all texts (oral or written) either fall somewhere short of full denotation or go beyond pure denotation, or do both. As a theoretical middle stage between subdued and enhanced semanticity, the semanticity where signs are taken at face value is quite acceptable, however, and does not cause any difficulty. Enhanced semanticity is found in texts containing clues that either reinforce the denotative interpretation or modify it by adding new elements. Weinreich mentions two marks of hypersemanticization: 1 The phonic part of signs assumes an independent symbolic value (sound imitation, expressive sound, see, e.g., chapter 7, pages 43-5); signs with similar acoustic images (rhyme, etc.) may be correlated in a way that normal speech, based on the arbitrariness of the sign, does not know. 2 Over the scope of a given text (poem, etc.) meanings are imputed to some signs which are richer than, or otherwise deviant from, the meanings of the same signs outside the text. Weinreich stresses the difference between a message in standard usage that the receiver has only to decode, and a hypersemanticized text where

67 Short of denotation and beyond it the common code is modified and has to be guessed by the favourably inclined receiver. Without guessing the modification it is impossible to decode the message properly. 2 Without denying the importance of hypersemanticized texts and discussions about how to analyse them from a semantic point of view Weinreich, as a true linguist, warns against the danger inherent in the tendency 'to concentrate on the special effects of meaning [in these texts] without first accounting for the semantic workings of language in its more standard uses.' 3 This warning is not always heeded, as will become clear from some of the examples that follow. Jurij Lotman's ideas4 are in agreement with those expressed by Weinreich in so far as Lotman recognizes and emphasizes the special position of literary texts as compared with ordinary language messages. He also stresses the importance of sound symbolism, of special links created by rhyme and rhythm. Because there is in many texts a network of hints and hidden cross-references, Lotman arrives at the following conclusion: whereas in unmarked, normal language use the unit is the word,5 a literary text operates with units well below the word level, such as individual sounds, as well as with much larger units. Ultimately the text itself in its entirety becomes a semantic unit. Although one can understand that the unity of a text as manifested by what we shall call 'hidden' cross-references is an important presupposition for the unravelling of those cross-references, it seems that postulating a semantic unit the size of a text does not serve any purpose. In the first place, pointing out the coherence of the various text elements would be enough, and in the second place even the text may not be a sufficiently large unit to account for all the hidden hypersemanticized clues, as some clues can only be picked up by someone who has read other older texts that contain the needed explanations. This phenomenon of dependence on other texts is not a recent discovery in literary analysis: influences and filiations in literature have been a topic of investigation for a long time. What is new, however, is the emphasis on subconscious factors, where the author is not necessarily aware of the fact that texts he has read account for elements in his own work, whether these elements are a continuation of a tradition or a reaction against it. A good deal of scholarly research and of ingenuity has been invested in the study of what is called 'intertextuality.' 6 As literary influence and intertextuality are not limited to one language and one period, we are far removed from the synchronic linguistic approach

68 Linguistics, literary analysis, and literary translation where a sign is determined by and opposed to all the other signs of the same level of investigation, and close to cultural history and cultural tradition. Another problem connected with the hypersemanticized sign is caused by the lack of a general code. This difference is a crucial one because the linguistic code is based on a shared convention, while the hypersemanticized code is more like a riddle that has to be solved without any means of verifying that the solution is the correct one. Although there are a number of conventional literary devices, the non-conventional aspect of a text is what really matters and what gives it its own identity. It is in this area that there is no well-established code, and it may therefore be preferable not to speak of the correct solution, but of a correct solution, because there is more than one. So, one can expect many self-proclaimed authorities to offer a wide variety of views and interpretations, the acceptibility of which depends partly on the receptiveness of the reader. When Lotman discusses a poem by Aleksandr Puskin, he mentions the special effect of the vowels in Istolkovat' sebe ne smel: i-o-o-a-e-e-e-e (He did not dare to explain to himself) 7

where in the first part a transition between different vowels takes place, and in the latter half there is repetition of one vowel. Lotman's remark is sharply criticized by Karel van het Reve, a Dutch Slavicist and literary historian. His main objections are: first, that the order i-o-o-a, e-e-e-e does not say anything about the quality of the verse and good and bad verses may show identical vowel patterns.8 Second, that Lotman speaks about phonologically identical vowels, but the pronunciation of a phoneme in pre-stressed or post-stressed positions is quite different from its realization under stress. The e-e-e-e sequence, where the first and the third e are unstressed, sounds like i-e-i-e. Van het Reve could have added that, for the same reason, the first series does not sound i-o-o-a but rather i-a-a-a. 9 It is noteworthy that Lotman tries to point out euphonic effects without going into the relationship between signified and signifier here. In most cases where special phonic combinations are examined and admired, the meaning of the passage in question is taken into account. Yet meaning cannot be excluded as Raymond Queneau shows by using what he called isovocalism. Isovocalism consists of reproducing all the vowels of


Short of denotation and beyond it

a famous line in the same order, but changing the lexical items that carry the most important meaning. Stephane Mallarme's use of vowels in Le vierge, le vivace et le bel aujourd'hui

is repeated by Le liege, le titane et le sel aujourd'hui. ' 0

Another experiment that could be done is to go to a poetry reading in an unknown language. Even when those who speak that language extol the musical qualities of the poems, saying that they are pure music, the foreigner may not perceive any musical qualities at all and prefer to go to a concert. The relation between signified and signifier in those cases where sound motivation is being perceived sometimes seems to take the signified as its starting-point, sometimes the signifier. In a signified-based relationship, it is, on closer examination, not the abstract signified but a specific image of the referents in general or of a particular referent that is linked with the sound impression. A typical example of such a link, based entirely on the specific image the referents call forth, can be found in the discussion of a line in Puskin's Evgenij Onegin by Vladimir Nabokov. 11 For the line in question Pod sen' ceremuch i akacij

Nabokov, after a lengthy discussion of the reasons for his choice, proposes: Beneath the racemosas and pea trees

(for Nabokov's translation theories see chapter 16, pages 105-6). Nabokov rejects translations for ceremuch other than his own as not being evocative and poetic enough to match the Russian word. The Padus racemosa is a kind of bird cherry with clusters (racemes) of white flowers in the spring. The fruits are dark cherries that attract many birds. Nabokov says about the tree and its name: 'The Russian word, with its fluffy and dreamy syllables, suits admirably this beautiful tree, distinguished by its long racemes of flowers, giving the whole of it, when in bloom, a gentle pendulous appearance. A common and popular woodland plant in Russia, it is equally at home among the riverside alders and on the pine barren; its creamy-white, musky, May-time bloom is associated in Russian hearts with the poetical emotions of youth' (Brower, page 104). Judging by the emotional description of the racemous bird cherry in the Russian spring, Nabokov has carried over his favourable association

70 Linguistics, literary analysis, and literary translation

from the tree to the name of the tree. For one who does not share Nabokov's Russian experience of the white blossoms proclaiming it is spring after a long harsh winter, ceremucha does not seem fluffy and dreamy at all. A link in the opposite direction where the sound influences the associations triggered by the signifier can be observed in Georg Trakl' s poetry. Heinz Wetzel made an in-depth study of sound and image in the work of this expressionist poet. 12 The difference between Wetzel's findings and Nabokov's remarks is partly the result of a difference in approach and a difference in scope: Nabokov gives one example of the phenomenon of sound symbolism in an article dealing with translation problems, whereas Wetzel makes that phenomenon the topic of a book. Puskin' s fluffy and dreamy syllables are not examined in the light of the poet's oeuvre and the reader of Nabokov's article does not know whether he is dealing with an example that is typical of Puskin' s poetry and idiosyncratic or with an instance of a more isolated character. Trakl's poetic work, however, is scrutinized for recurrent themes and characteristic sound combinations translating certain moods. The melancholy of a quiet autumn scene in nature with a pond and elms and elders is reinforced or created by the German for those trees: Ulme and Holunder. Both trees often recur in Trakl' s poetry and the mood is invariably pensive and melancholic. In contrast to Puskin's example it is not the sound symbolism but the referents that give rise to some questions. Elders are nice shrub-like trees that bloom in early summer. Especially when in bloom, they do not convey feelings of melancholy but rather of joy. Not unlike what the clusters of creamy white blossoms of the racemous bird cherry meant for Nabokov, the fragrant, creamy white, umbrella-shaped flower clusters are for me indelibly connected with the beginning of summer and the summer vacation on Texel, one of the Frisian Islands that form a string along the coast from the north of the Netherlands to Denmark, where my parents used to take us in the summer. Elderberries are dark, but so are the fruits of the bird cherry. Moreover, from elder-berries one can make wine or jelly; from bird cherries, nothing at all. So, even the darkness of the fruit in itself is not necessarily a reason for melancholy. As the shape of the elder does not offer any clue either - the weeping willow may serve as an example of a tree with a melancholy-inducing appearance - it is probably the sound combination Holunder that is the source of the symbolism of that tree in Trakl's poetry. 13 What emerges from this two-tree discussion is that it is impossible to treat sound symbolism uniformly. Another important point to remember

71 Short of denotation and beyond it

is that the abstract general character of the signified recedes to the background or disappears altogether because the sound symbolism is connected with only some of the qualities of the signified. 14 It should be noted that neither Nabokov nor Wetzel uses the saussurean terms, thus avoiding the risk of confusion between the sign in context and the sign as an abstract unit of the language inventory. It is unfortunate that many others do not show the same insight and instead of abiding by terms such as expression and content prefer signifier and signified without defining them so as to avoid misunderstanding about the level of abstraction - or concreteness - on which they place their analysis. It is undoubtedly because of the blurring of the levels of analysis that objects described in a text become signifiers of the social background of their owners although the objects are not part of the language and although the relationship between the object and the social background of the owner is not as exclusive as the one that unites signifier and signified. The problem here is that anthropology and sociology have borrowed the linguistic terminology, and that literary criticism borrows the same terms, sometimes from linguistics, sometimes from the other two disciplines. In the next chapter a special case will be discused where linguistics and sociology have conflicting views and literature is caught in the middle.

12 What's in a name?

One of the points about which linguists seem to agree without factional strife is the position of proper names. Unlike linguistic signs in the saussurean sense, proper names do not have a conceptual part linked in a bi-univocal way with the acoustic image. Names serve as labels or as means of identification, but they do not have any semantic content. This is easy to see when one compares different persons who are namesakes without being relatives: all they have in common is the name. Yet this simple truth seems to be contradicted by the fact that the mere name of a person one does not know raises all sorts of expectations about the characteristics of the bearer. These expectations are partly based on ethnic stereotypes: a Scandinavian name makes one expect a tall, blond, blue-eyed person; an Italian name, somebody with dark hair and brown eyes who is rather short. In the Netherlands there are first names and surnames that are associated with the northern provinces where people are supposed to be tall, taciturn, and reliable, and others that are connected with the south where the stereotyped inhabitant is more talkative and outgoing, has a higher appreciation of the good things of life, and is morally less strict than his northern counterpart. To complete the picture it should be added that the south is Roman Catholic and the north Protestant with Calvinistic overtones. Apart from these national and regional associations, there are social stratifications in proper names and surnames that are also taken for granted when people try to guess what kind of person is behind a name. So, the personal identification is accompanied by social data about the identified individual. Because there are fads in first names that last for some time before a new name becomes the favourite, replacing the old one that has lost its popularity, one can sometimes even guess the age of a person by the given name.

73 What's in a name? Of course, all these extra data that names seem to provide are based on statistical probabilities, and that means that they do not provide any data at all. If one compares table to Robin Smith, the difference between the two is striking. While table has to refer to a table, i.e., one object falling within the conceptual category of table, Robin Smith can label a variety of persons. They may be male or female, of English or AngloSaxon extraction, but also of any other extraction, as the name may have been adopted when the family in question immigrated to an Englishspeaking country. Moreover, one male English ancestor many generations ago can account for the name of an otherwise totally non-English family. If the Smith family chooses to call their cat Robin Smith there is nothing to prevent them. This means that Robin Smith does not even signal the element 'human' unequivocally. As first names are given by the parents in most cases, they sometimes indicate special religious or political preferences parents have at the moment of the birth of their offspring. This says, of course, something about the parents and very little about the bearer of the name (apart from the information about the kind of parents he probably had - 'probably,' because the name may have been given to an orphan by a government institution). So, a convinced republican born during the Second World War when the Germans occupied the Netherlands may be called after Queen Wilhelmina, and a pacifist, son of Nazis or parents sympathetic with Hitler's regime, may be saddled with the name Adolf, while a Russian woman may have to go through life as Elektrifikacia in honour of industrial development in the twenties and thirties. But, then again, if one's name is too much of a burden one can always change it officially, or adopt another name to be called by. So, it is very dangerous to draw any conclusions from names. In my own case the first name, Henry, has sometimes been interpreted as a sign of my desire to blend with Anglo-Canadians, many immigrants anglicizing their first names and giving their children, born in Canada, English names rather than following traditions of the country of origin. This interpretation of Henry is wrong, however, as I was named after my grandfather who happened to have this English name. Clearly, the label idea proposed by Saussure is still the only acceptable one, from a linguistic point of view. This observation applies to real-life situations, but it has to be qualified when we are dealing with proper names in literary texts. The only exception is the historical-biographical or autobiographical genre if true names are maintained, as in those writings real-life situations prevail.

74 Linguistics, literary analysis, and literary translation In all other works, though, names are created by the authors, and many heterogeneous factors may determine their choice. 1. There are subconscious links between persons who played a role in the life of the writer and some of his fictional heroes. If the same name is used for different heroes in successive books, positive or negative connotations with that name may become apparent. If each book is examined separately, though, the reader gets the impression of random choice. The connections between real-life encounters recede into the subconscious of the author and the names of his protagonists are not really interesting from a literary point of view but rather belong to the realm of psychology. 2a. The writer is fully aware of the reasons why a certain hero has been given a certain name. He may reveal them in private or in public, orally or in writing; or he may keep his reasons to himself. In the latter case there is, from the point of view of the reader, little difference between this situation and the one described under 1, above. Even if the information about the deliberate link is very interesting and sheds new light on the author it is doubtful whether it is of any importance from a literary point of view. The fact that the surname Levin in Lev Tolstoj's novel Anna Karenina indicates that the hero of that name often presents the views of the author and is very close to him is interesting, but does this knowledge affect the interpretation of the work, and does it alter the aesthetic appreciation of its literary qualities? When Lermontov calls the protagonist of his novel, A Hero of Our Time (Geroj nasego vremeni), Peforin, some readers may connect the name, which is derived from a river named the Pecora, with the name of Onegin, also going back to a river, the Onega, and then link Lermontov's hero with Puskin's, as was Lermontov's intention. The reader may also have been told about this feature of intertextuality in a literature seminar. Again the question can be asked to what extent this piece of information influences the reader. As was already discussed (chapter 9, pages 52-3) knowledge of the author's intentions sometimes helps in interpreting an otherwise ambiguous text. Names of the type of Levin and Peforin confirm what the text would have made clear, even without the clues provided by the names. The readers who are not aware of the special character of the names in question can easily detect the importance of Levin's opinions in Tolstoj's eyes, and provided they have read Evgenij Onegin, discover the affinities between Peforin and Onegin. In intertextually oriented studies the uncovering of name clues forms an important part that tends sometimes to become an aim for its own sake (see also chapter 14, page 90; chapter 16, page 106).

75 What's in a name? 2b. Some writers give names to their characters - or to some of their characters - that tell the reader something about their personalities. In Dostoevskij's The Idiot (Idiot), the protagonists are Prince Myskin (mys = 'mouse') and Rogozin (rog = 'horn'), the former being good, unrealistic, and defenceless, the latter personifying the harsh driving force of evil. In Crime and Punishment (Prestuplenie i nakazanie), the hero who does not recognize the laws of society is called Raskol'nikov, a name that is based on the religious term raskol or 'schism.' In the seventeenth century the Raskol'niki did not recognize the authority of the official church and, though severely persecuted, maintained their own beliefs. The character's friend, and later on brother-in-law, represents the common sense and wisdom without which society cannot function. His name, Razumichin, is derived from the noun razum, which means 'wisdom.' A similar device can be found in Pasternak's naming his hero Zivago (in the novel Doktor Zivago), based on zivoj meaning 'alive' or 'living,' as opposed to Antipov, who is his antagonist, as the element anti clearly indicates. Somewhat into the novel Antipov assumes the name Strel'nikov, containing the stem strel that is found in streljat' ('to shoot'), strela ('arrow'), etc. This assumed name is explicitly connected with the bearer's activity while the original Antipov is never explained. A whole range of names in Gogol's Dead Souls (Mertvye dusi) that, without ever being explicitly commented upon, indicate the main characteristics of their bearers fall in the same category as Antipov. All these names underline but do not create the impression the reader gets from the text about the heroes. If he does not notice the connection he does not miss very much, except the feeling of being an insider. Translators, not surprisingly, do not try to translate the revealing names and keep the original forms in their translations. Thus the allegorical element is lost. It is only in truly allegorical works that the names of all the characters are translated. In stories such as the Dutch Elckerlijc or the German Jedermann, or the French Roman de la rose' the characters have the name of the quality they personify, and translation does not create the problem that would result from the translation of only a few of the names among the many that occur in novels such as Crime and Punishment or Doctor Zhivago. 2c. Finally, a very common name-giving procedure consists of choosing a name that raises expectations about social and/or geographical background. The writer uses the fact that stereotypes are solidly entrenched in the minds of most readers. By choosing a name that fits the

76 Linguistics, literary analysis, and literary translation character according to readers' expectations he shows his insight. There are many Frenchmen with non-French-sounding names, but Frenchmen in English, German, or Russian novels have French names that sound really French. In some instances names are significant of political behaviour, even though the persons bearing them received the names at birth when nothing revealed their future political preferences. Aleksandr Solzenicyn' s novel The First Circle (V kruge pervom) offers a clear example of this kind of significance. Major General Petr Afanasievic Makarygin has three daughters. The oldest is called Dinera from ditja novoj ery ('child of the new era'), the middle one, Dotnara from doc' trudovogo naroda ('daughter of the working people'), while the meaning of the youngest daughter's name, Klara, is not known in the family. The two older daughters are opportunistic and similar to their father, a typical apparatschik, prosecutor of special (i.e., political) cases. The youngest daughter, Klara, however, has an open heart and an open mind that make her aware of all the pettiness, cruelty, and injustice around her. Solzenicyn thus follows two traditions: people have names that fit them, and as in fairy tales, the youngest of the three is good and noble, unlike her older sisters. If the middle daughter, Dotnara, had been good and Klara bad, two of the reader's expectations would not have been fulfilled. In some instances the writer deliberately creates a clash between the name and the hero, and this clash may even play a part in the story. In the Dutch novel by Harry Mulisch entitled De aanslag (The Assault), the first part takes place in the winter of 1945 when Holland was occupied by the Germans. The main character, a boy about twelve years old at that time, has the same first name, Anton, as the leader of the Dutch National Socialist party, Anton Mussert. When his older brother, Peter, is angry with him he calls him 'Anton Mussert.' One of those silly quarrels with name-calling takes place just before a notorious Dutch Nazi policeman is shot and killed near their house. This is the start of a series of tragic events, incomprehensible and brutal, that form a terrible contrast with the preceding brotherly quarrel. Mulisch lets his hero reflect on the names 'Anton' and 'Adolf' in retrospect, using the free indirect-speech form. However, apart from the obvious clash between the name Anton and the fact that Anton becomes a victim of Nazi reprisals, the name has little or no impact on the protagonist's behaviour and does not influence the further development of the story. The importance of a name that does not correspond to expectations is much greater in one of Dostoevskij's novels, The Adolescent (Podrostok). The adolescent, the natural son of a landowner, is officially the son of

77 What's in a name? his real father's gardener. Unfortunately, this gardener's name is Dolgorukij, also the name of a very-well-known family belonging to the Russian nobility. Being asked constantly if he is a member of this respected family adds to the humiliation and frustration of the sensitive hero and makes him want to assert himself. Humiliation and frustration, combined with ambition, form the ingredients of a typical Dostoevskij hero, and the fact that his surname, Dolgorukij, is all wrong probably merely reinforces features of character that would have been there anyway. This is of course only speculation, but it seems to me that the name factor should not be overrated when one tries to analyse and explain the hero's behaviour and thoughts. The picture that emerges from what has been discussed so far shows that names in literature are not always simply labels. The stereotypes that govern expectations in 'real life' are often used by writers in order to create an impression of reality. In other instances names stress the allegorical function of the heroes who bear them, while still another category of names links the work in which they occur with previous works, thus stressing cultural traditions and illustrating intertextuality. The same name may function differently in different texts, and no rule can be given on how to interpret a given name so that the interpretation will always be valid. The name can be used to frustrate expectations that are based on stereotypical representations or it can be a simple label. 2 The last two usages highlight the difference between a name and a linguistic sign, even when we limit ourselves to literary texts. In spite of all appearances to the contrary, the proper name is not a sign (unless it becomes a common noun as in a real Don Juan, un vrai Tartuffe, etc.), and it is therefore advisable not to use linguistic terminology when talking about the value of proper names in a literary work. Roland Barthes, when discussing proper names in Proust's A la recherche du temps perdu, does use linguistic terminology, and the result is rather confusing, because the terms are not defined nor do they correspond to the saussurean definitions. To make things worse there is a constant shifting in apparent meaning, and equations where the two terms are not at all synonymous in linguistics are fairly frequent. The following passage taken from the essay 'Proust et les noms' ('Proust and Names')3 is a good example not only of Barthes's handling of proper names in literary texts but also of his general use of saussurean terminology. 4 In Proust the [Proustian] narrator has to accomplish an ambiguous task (for it leads through a good many errors to the truth), which consists of desperately

78 Linguistics, literary analysis, and literary translation questioning the signs: signs emitted by the work of art, by the beloved person, by the milieu he frequents . The Proper Name is also a sign, and not, of course, a simple mark [French: indice] that, according to the current conception from Peirce to Russell, indicates without meaning. As a sign the Proper Name offers itself for exploration, for deciphering; it is at the same time a 'milieu' (in the biological sense of the term), in which one should plunge, bathing indefinitely in all the reveries it carries with it, and a precious, compressed, embalmed object that one should open as a flower. In other words, if the Name (that is how we will call the proper name from now on) is a sign, it is a voluminous sign, a sign that is always heavy with a dense thickness of sense, which is not reduced or flattened by any usage, contrary to the common noun, which never delivers more than one of its meanings [French: sens] per syntagm. The Proustian Name is by itself alone and in all cases the equivalent of an entire dictionary heading: the name Guermantes immediately covers all that memory, customs, culture can put into it: it knows no selective restrictions, the syntagm in which it is placed being indifferent to it: so it is, in a certain way, a semantic monstrosity, for, provided with all the characteristics of the common noun, it is nevertheless able to exist and function outside any projection rule. That is the price - or the ransom - of the phenomenon of hypersemanticity [see above, chapter 11, pages fxr7] of which it is the seat and which brings it, of course, very close to the poetic word. Although Barthes mentions Weinreich in a footnote, it is clear that his flowery description of the evocative value of proper names in Proust has very little in common with Weinreich's lucid analysis of hypersemanticized language. Barthes does not mention 'normal' usage, nor does he explain the difference between associations (of a personal or a more collective cultural character) with a specific referent indicated by a proper name and the semantic content of a common noun (offering the possibility of associations and connotation in addition to the cognitive, intellectual denotation). Barthes calls all associative values 'semes,' declaring: These semes are, of course, 'images,' but in the superior language of literature, they are none the less pure signifieds, offered like those of denotative language to an entire systematic of meaning. Some of these semic images are traditional, cultural: Parma does not indicate a town in Emilia, located on the Po, founded by the Etruscans, of the size of 138,000 inhabitants; the veritable signified of these two syllables is composed of two semes: stendhalian sweetness [French: douceur] and the reflection of violets. Others are individual, based on memories: Balbec has for semes two words spoken in the past to the narrator, one by Legrandin (Balbec is a place of storms, at land's end), the other by Swann (its

79 What's in a name? church is Norman gothic, half romanesque), so that the name always has two simultaneous senses: 'gothic architecture and storm at sea.'

Thus, the seme part of the signified is itself a signified and an image. One is far removed from the smallest meaningful element. The problem linguists face when introducing a unit that does not always have a recognizable corresponding element at the level of expression does not seem to bother Barthes, who does not mention the signifier that would have to accompany the seme-signified, if the rule no signified without signifier were adhered to. Barthes goes on in the same vein, using linguistic terminology to describe things quite different from the linguistic concepts: 'So, each name [no capital Nin the French text] has its own semic spectrum, variable in time, according to the chronology of its reader, who adds to or takes away from its elements, precisely as the language does in its diachrony.' Language in its evolution is a community phenomenon and has very little in common with the changing individual associations of a reader when confronted with a proper name. The Name, indeed, is catalysable; one can fill it, expand it, fill the interstices of its semic armature with an infinite number of additions. This semic widening of the proper name can be defined in another way: each name contains several 'scenes' that have first arisen in a discontinuous, erratic way but just beg to be combined and thus form a little story [French: recit], for telling is never anything but linking together by a metonymic process a reduced number of full units: so Ba/bee not only contains several scenes, but also the movement that can bring them together in one and the same narrative syntagm.

It is difficult to follow Barthes' s reasoning in this passage: there are

several scenes linked by a metonymic process - it is not explained what precisely a metonymic process is like - into a syntagm. But a syntagm is the result of the linkage of several formally distinct elements into a coherent chain. It is impossible to apply this to associative evocations triggered by the name Balbec. The formation of a syntagm is dependent on syntactic and semantic rules. In the case of Balbec there cannot be a syntagm, whether it is called narrative or not, in the linguistic sense of the term. When commenting on sound symbolism in Proust, Barthes introduces what he calls an intermediary concept between the signified and the signifier. His example is again Balbee: if it signifies by affinity a complex

8o Linguistics, literary analysis, and literary translation of high-crested waves, steep cliffs, and a rough, rising architecture, it can only be because we have a conceptual relay, that of 'rough' (French: rugueux), that is valid for touch, hearing, and sight. 'In other words, phonetic motivation requires an interior naming: the language surreptitiously returns to a relation that was - mythically - postulated as being immediate.' Metaphor and symbolic phonetism are thus closely linked (e.g., rough being applied to a sound). The reasoning is based on the sound effect of Balbec, and it is questionable whether the relay concept rough ('rugueux') is really connected with the signifier and not only an association with the signified. Barthes gives a typical example of what Weinreich warned his readers to avoid: he loses sight of normal language use and only focuses on literary language. He then applies the semantic terminology that Saussure created for describing language as a meaningful code shared by the members of a speech community to proper names, a category that Saussure explicitly excluded from his semantic description. That a label-name may evoke a complex of memories and associations in an individual hearer (e.g., Proust's narrator and via him Proust's reader) seems to depend on the knowledge the hearer has of the referent. But Barthes rejects this explanation emphatically when he concludes his essay on Proust: The name is nothing if it is unfortunately joined directly to its referent (what is, in reality, the duchess de Guermantes?), that is to say, if its nature of a sign is missed. It is in the signified that there is the place of the imaginary: that is, no doubt, Proust's new idea, for which he has shifted, historically, the old problem of realism that presented itself, before him, almost exclusively in terms of referents: the writer works, not on the connection between the thing and its form (that which was called in the classical period its 'painting,' and more recently its 'expression'), but on the connection between the signified and the signifier, that is to say on a sign.5

Barthes' s discussion of proper names in Proust shows an unfortunate mixture of intuitive literary insights that are very personal and a terminology that suggests general validity. Linguistics has not served Barthes and Barthes has done a disservice to linguistics. In the next chapters an answer will be sought to the question whether the relationship between linguistics and literary analysis is of necessity an unhappy one or whether there are more constructive elements in it

81 What's in a name? than would appear from the analysis of Barthes' s essay. First, some general linguistic insights will be examined from the point of view of their impact on literary analysis and then a few examples will be given of linguists trying their hand at discussing a literary text.

13 Linguistics and literary analysis: a happy alliance?

One of the most elusive problems involving linguistics and literature is the question what features make a text literary. Until now we have taken for granted that literary texts form a special category. It would be nice if linguistics could provide a definitive answer and give a list of necessary and sufficient criteria. Instead of such a clear-cut solution there are numerous suggestions of criteria that cover only part of what is intuitively considered to be literature and - a more serious drawback - tum out to be applicable to what is generally considered not to be literature. Formal constraints are part of most poetry, but rhyme and rhythm do not make a literary work if applied mechanically by would-be poets. However, Baudelaire called short prose texts Petits poemes en prose' and the Russian novel writer Turgenev published late in his life short nostalgic texts in prose that are called by the author himself Senilia or Stichotvorenija v proze (Poems in prose). 2 Saying that the semantic load of the constituent elements of the text- see above (chapter 11, pages 66-7) for the notion of hypersemanticity - is of decisive importance does not help solve the problem because this hypersemanticity is not necessarily formally different from normal semanticity. The criterion 'Recognized as a work of literature by generation after generation,' which is explicitly chosen by Jean Cohen when he discusses the structure of poetic language3 and tacitly used by many others, really begs the question because it does not explain why all those generations agreed in calling the works in question literary. Apart from the absence of directly observable linguistic criteria to determine the literary character of a text even if it has been considered a masterpiece through the ages, there are texts that in the eyes of some generations of readers qualify, while others reject them. Not only do people of the same generation disagree on the literary quality

83 Linguistics and literary analysis of texts that have recently been produced, but there is also a sort of collective shift in taste when we examine the fate of some writers over the years. These two kinds of differences in taste are difficult to keep apart, although one should not confuse them in spite of some apparent similarities. On the one hand, there are individual tastes and personal idiosyncrasies; on the other hand, there is a culturally and historically determined preference shared by the majority of the members of a certain community in a given period. This community may represent only a small part of the speech community to which it belongs or it may almost coincide with it, depending on the position of literature in the speech community as a whole. For the sake of simplicity we shall speak of the collective taste without mentioning every time that the collectivity is not to be confused with the speech community. As was said before, individual deviations from the collective taste occur; some mavericks may agree with a generation they do not belong to, and others fall outside any mainstream that is known for the past or the present. (For the future one cannot be sure, of course.) So, the taste that is representative for a period is not absolute, and the group of people who have any opinion on literature, whether following the trend or going against it, may be very small compared to the total number of people who speak the language in which the literature is written. Although especially those who go against the trend often have the impression of being independent in the opinion they arrived at by their own free choice, they are likely to have been influenced by their surroundings: a dissident can only dissent if he knows the general opinion and manifests his disagreement as a reaction against it. This is precisely the reason why it is impossible to re-create the conditions of a previous period in order to identify with the people who lived at that time in their appreciation of art or literature. No matter how careful a special setting is reconstructed with period furniture and candle-light and no matter how perfect the old instruments are on which baroque music is played for an audience that wants to enjoy the music as it was enjoyed by the first listeners, the whole enterprise is doomed to failure because the public cannot but judge the music from a twentieth-century perspective. In some respects a musical evening with music that has never been played before, staged in a very modern setting, comes closer to creating the conditions of first-time performance of baroque music than does a connoisseur's re-creation of all the authentic details that are known. In the same way an ancient manuscript or even a book that appeared

84 Linguistics, literary analysis, and literary translation a few decades ago cannot be read with the eyes of the first readers. It is impossible for the modern reader to block out of his mind all the linguistic and cultural-historical changes that have taken place since the text was created. The situation, then, is not without some similarities to that posited by Whorf and Sapir for interlinguistic communication: each speaker is a prisoner of his native tongue and the conceptual and grammatical inventory it offers him for formulating his thoughts (see chapter 6, pages 39-40). Although we can try to understand the literary preferences of bygone periods, there is inevitably a considerable amount of interference from our own era. Sometimes it is impossible to understand the success or the lack of success of a literary work at the time it appeared, but even when our admiration for a poem or a novel seems equal to that of the contemporaries of the writer, we may have completely different reasons for our admiration. So, the literary theory known under the name of hermeneutics4 and the linguistic views that stress the importance of an immanent analysis of a language at a given moment of its evolution do not contradict each other and could easily be combined. But even if one combines the linguistic theory with the postulates of hermeneutics according to which a reader's appreciation of a text is conditioned by the norms prevailing in that reader's lifetime, the question of what makes literature into literature remains unanswered. If literary specialists look in vain for solutions from the linguistic side, linguists, waiting for answers from the side of literary analysis, are equally frustrated in their expectations. Sometimes intentionality is mentioned, but the best of literary intentions and the most dutiful application of what are called 'literary devices' - in the case of poetry, rhyme, rhythm, and a special vocabulary; in the case of prose fiction, well-formed sentences and an interesting subject - do not guarantee that the result will be considered to be literature. However, some texts of a religious character such as 'The Song of Songs' are included in literature. In the case of private diaries the intentionality is often difficult to assess. The writer may have readers in mind and be aware of the literary quality of what he is writing (as is true of Andre Gide's diary), or he may honestly reserve his diary for the outpouring of his private reflections not meant to be read by others. Epistolary prose presents another example where literary intentions may or may not be an important factor. Even if we limit ourselves to narrative prose as it has been the subject of many theoretical studies in recent years, the difference between literary and non-literary texts does not become any clearer. Narratology

85 Linguistics and literary analysis examines the devices used in narrative prose and tries to discover the basic structures that are common to all stories. What started with the study by Vladimir Propp of the structure of the Russian folk-tale, 5 a genre in which tradition is very strong and stereotypes abound, has become a general examination of all sorts of narratives, from short folktales to experimental novels. Important distinctions are made between the chronological sequence of events - the fabula - and the way these events are ordered and presented in the story that is told in the text. Who tells the story is another question that is scrutinized. It is here that the linguistic category of personal pronouns enters the picture. Apart from this category one finds verbal voice and sometimes free indirect speech and its formal expression as elements of the discussion, but on the whole, linguistics plays a minor if not insignificant role in narratology.6 This fact may pass unnoticed, and the extensive use that is made of linguistic terms such as the text grammar, the morphology of the text, often creates the false impression of a close relationship between the two disciplines. This impression is reinforced by occasional equations of text structure and sentence structure: different actors in the story get labels such as 'dative' or 'accusative' assigned to them. This does not mean, of course, that it is wrong to use linguistic terms metaphorically to describe and analyse a literary text - the terminology may even turn out to be helpful. However, when one comes across such terms in a literary analysis or in a general literary theory, one should always try to determine whether or not they are used in their original meaning. It is not easy to draw the line between linguistics applied to literary analysis and linguistic metaphors. Even if the rule is applied that linguistic distinctions should have a formal basis, one occasionally runs into trouble, because not all linguistic theoreticians themselves abide by that rule without allowing for some exceptions. An additional difficulty lies in the fact that some distinctions are made by phonic means that are only partially reflected in the written text, if they are reflected at all. A few examples of ideas formulated by a linguist and taken over by literary analysts will shed some light on the positive as well as the negative aspects of this fairly recent alliance. As with narratology, no criteria are found that allow us to isolate literature as a separate category, and when we speak about literary texts the qualifying adjective 'literary' is based on collective and personal taste and used in the same loose way as it has been used throughout this book. 7 One of the linguists who is often mentioned in literary criticism is Emile Benveniste, whose work on general linguistics contains many

86 Linguistics, literary analysis, and literary translation ideas that are of direct interest to those who study literature. In his book Problemes de linguistique generale8 the question of the special character of the category of personal pronouns comes up several times. With a wealth of formal arguments taken from many different languages Benveniste argues in favour of the exclusion of the third person from the personal category. Unlike the first and the second person, the third person is the non-person. From a functional point of view, however, there are strong arguments for its inclusion, and as Claude Hagege shows convincingly in his discussion of the question,9 the third-person pronoun has very often a person as its referent so that the label 'non-person' is definitely misleading. Unfortunately it is precisely this label that is easily remembered while the morphological and typological arguments (pro and contra) brought forward by Benveniste and Hagege are quickly forgotten, if they have ever been known, by the literary analysts who are fascinated by the non-person status of the third person and give it an existential interpretation. Linked with the first-person-centred pronoun system is the division of French verbal tenses into a category belonging to discourse in which the speaker/writer is directly involved, and a category used for what Benveniste calls le recit historique. 10 In the latter category the facts are presented without personal involvement of speaker or hearer. Viewing the passe simple as an aorist, the author formulates the following trends: as the aorist is used for the relating of historical events without emotional interference, the occurrence of first- and second-person forms is rare: ils arriverent and il arriva are very frequent; nous arrivames and tu arrivas are not. When one keeps this distinction in mind, some shifts in tense use can easily be explained. In a passage of Voyage au bout de la nuit Celine seems to address the reader directly when he uses the passe compose to relate crucial events, but he returns to the neutral tone of the passe simple (even in the first-person singular) to continue his story when the events are less exciting or sudden. 11 The avoidance of the passe simple and its replacement by the passe compose in texts where the discourse character is not easily detected except for the fact that the passe compose is used creates the risk of circular reasoning in some instances, but on the whole the distinction is a helpful element in text analysis where the passe simple and the passe compose are concerned. Inasmuch as the imparfait occurs in both categories set up by Benveniste the thorny problems connected with the use of that tense 12 fall outside the realm of the opposition between discourse and recit historique. The opposition we just discussed was formally expressed and posited

87 Linguistics and literary analysis specifically for French, so unjustified extensions into other languages without formal oppositions comparable to that of the French verbal system are not being encouraged. One would expect this warning against unwarranted extrapolations also to apply to Benveniste's semantic analysis of the middle voice, 13 based mainly on Sanskrit, Greek, and to a lesser extent Latin, where the class of deponentia is, as it were, a remnant of the productive middle voice in the two other languages. The formal expression of the middle voice and its opposition to active and passive are the necessary conditions for its existence. After having examined cases where the same verb occurs in both active and middle voice - 'the priest sacrifices' (active, for somebody else), 'the man sacrifices' (middle voice - for himself) - Benveniste discusses verbs that only occur in the active voice (activa tantum) and verbs that are only encountered in the middle voice (media tantum). A careful analysis and comparison of the two groups from a semantic point of view yields the find that in the media tantum the subject is internal to the action whereas in the active verbs the process takes place externally. This rather vague global distinction is of a semantic character. Because there is no direct opposition between active and middle voice for each verb but only a categorial opposition between the two groups, the presence of an element of choice on which a linguistic opposition is based may be put into question. But even if one rejects the opposition for activa tantum and media tantum, there are verbs in the language discussed by Benveniste that do have both categories. It is taking a step farther away from the principle of formal expression when the middle voice is introduced in descriptions of languages where there are no morphological markers for it. The label middle voice, used in The French Fictional Journal by Valerie Raoul for the verb to write, is justified by a quotation taken from Benveniste's article on the middle voice and by a reference to Roland Barthes' s declaring that 'the middle voice corresponds exactly to the state of the verb to write.' 14 This intuitive semantic interpretation does not have any formal justification. Benveniste does not even mention the possibility of a middle voice for languages that do not have the formal means to distinguish it from the other voices. It should also be noted that in classical Greek the verb 'to write' (-ypact,Hv) had an active as well as a middle voice so that it is not unlikely that ecrire in French and to write in English cover both the active and the middle voices of the Greek verb. In any case Benveniste' s linguistic authority appears to be called upon to justify a literary interpretation of a form (lexeme) that he would never have included in his theoretical considerations.

88 Linguistics, literary analysis, and literary translation So, the alliance between linguistics and literary analysis is not an altogether happy one. The partners expect each other to provide an answer to the question what makes a text belong to literature and they are both disappointed in their expectations. When linguistics studies precise details of linguistic functioning in general or of the functioning of one particular language, the findings tend to be incorrectly extrapolated in literary studies so as to cover phenomena that fall outside the boundaries that had been carefully defined in the original study. The result is sometimes a dialogue of the deaf, as both sides, using the same terms, think they are speaking about the same things, and then do not understand what the other is saying because the concepts behind the terminology are quite different. Yet there is room for co-operation, and provided some effort is made to clarify terminological misunderstandings a fruitful dialogue is possible. On the one hand, linguists can help bring some precision to the often fuzzy concepts that are worked with in literary analysis, and on the other hand, literary texts and the questions literary analysts ask about them may provide linguists with a topic that is refreshingly new, after the many studies they have done about sentences and constructions taken from their own idiolects or made up by themselves as illustrations of what they think is not grammatical. In the next chapter a few studies by linguists of specific literary works will be reviewed, in order to show the positive side of the alliance between linguistics and literature.


Linguists and literary texts

Apart from formulating general theories about literary language- Weinreich's claim that such language is hypersemanticized is an example of such a general approach (see chapter 11, pages 66-7)- linguists also have studied specific texts over the years. The interest shown for this kind of study varies considerably from period to period and from one linguistic school to another. On the whole, North American linguistics, whether neo-bloomfieldian or transformational-generative, does not seem very preoccupied with the special features of literary texts, giving more attention to problems not connected with literary characteristics. So, in order to get some insight into what linguists derive from their study of literary texts one has mainly to turn to Europe or to linguists who have been exposed to European approaches. Before we examine a few examples of linguists working with literary texts, it will be useful to distinguish between several categories of linguists, according to the aim and the character of their studies. 1 . First, there are the linguists who use literary texts as material for their grammatical and syntactic studies. In general, such studies discuss a syntactic structure, or the special use of a grammatical form illustrated by a quotation from a recognized literary text. As a rule different authors are quoted, but in some rare instances the work of one author is the source of all quotations and the basis of the investigation. 1 This category is interesting in so far as it stresses the relation between norm and literary language. It shows on the one hand that literary examples can be used to illustrate standard usage. (One should keep in mind that standard or normative models do not necessarily reflect the most frequently used speech forms within a speech community.) On the other hand it shows that there is a possibility that some literary texts even influence standard

90 Linguistics, literary analysis, and literary translation usage. In any case, the movement is from the individual writer's usage to the general prescriptive model, and little or no attention is paid to idiosyncratic features as those features cannot be used in examples that should be followed by everybody. It is impossible to get an insight into the author's work based on the isolated sentences that are quoted, and the various texts that provide the examples are never studied as literature with special attention to features that belong to the realm of hypersemanticity, nor are effects taken into account that are produced by text segments larger than the sentence. 2 . Second, some linguists study literary texts, especially poetic texts, in order to discover devices that enhance the semanticity of the text. The Anagrammes by Ferdinand de Saussure published a long time after his death2 and many studies by Roman Jakobson on poetics in general and on individual poems3 have a prominent place in this category. Not only are sound-expressivity and the linking function of rhyme that brings together otherwise completely disconnected elements studied, but also sound patterns and their repetition or reversal and associations based on similarity of signifiers. In some instances this similarity applies only to part of the signifier of an element as it occurs in the text. This partial similarity opens up a whole area of hidden signs that painstaking probing can bring to the surface. Very often the results say a good deal more about the analyst than about the text itself, and the ingenious analyses become shows of originality rather than illuminations of the text. The psychoanalytical orientation of some of the studies undertaken in this way makes discussion very difficult. The author may have aimed deliberately at a given association, it may be a manifestation of his subconscious, or the association could have its origins in the reader's mind. If cueillir des violettes is associated with viol4 in a poem from around the tum of the century, there may even be an irritated, incredulous reaction by some readers who find the connection far-fetched or even detrimental to their enjoyment of reading it. In any case one can say that the second category of literary studies by linguists is in sharp contrast with the first; it stresses the particular and is moreover so strongly reader-oriented that the findings do not have general validity. They are the result of one reader's interpretation and each reader is more or less expected to follow his own inclinations. In spite of the peremptory tone of some of the studies in this category, it is unlikely that any author would deny the personal slant of his text analysis, and claim that everybody should view the text in the same way. This reader freedom inevitably causes great problems when a text has to be translated. Because the text can be read

91 Linguists and literary texts

in many ways - some go so far as to say that each reader constructs his own text - it becomes very difficult to decide about right and wrong or even about better and less good in translations (see also chapter 16, page 105) . .3. As is always true of classifications into categories there is a significant overlap between categories and one should not think that each and every linguist dealing with literary texts neatly fits the criteria given for one category to the exclusion of all features mentioned in connection with the other categories. Thus, the fact that the third group of linguistic studies of literary texts is characterized by what can be called the 'statistical approach' does not mean that those studies exclusively concern themselves with statistical data. The gathering of those data is a means to an interpretive end, and the interpretations often fit into another category, especially the second one, as phonetic frequencies are interpreted from the point of view of expressivity. Most statistical studies deal with sound, but computers have made it easy to count almost any element of a text. But frequencies in themselves do not tell very much, and one of the difficulties is that they have to be compared with standard frequencies yielded by the study of a representative cross-section of texts. This cross-section may comprise texts of the same type as that being studied - poems for a poem, novels for a novel, etc. - or may be the other works of an author examined in order to highlight the special character of a given work within that author's oeuvre. It is also possible to include literary as well as non-literary works in the cross-section. It is only after the comparison with whatever standard has been chosen is made that the interpretive work should start. A statistical study somewhat outside the mainstream carried out by a linguist-mathematician compares poetic texts with scientific prose in order to determine the frequency of words with multiple meanings relative to that of words with one meaning only in both types of texts. 5 Although the idea is interesting and the text material that is used perfectly suited to the purpose, there is one major flaw that seriously reduces the value of the results that are given in percentages calculated with rigorous precision. That flaw lies in the fact that no criteria are given for establishing the number of meanings of a word. The number of meanings listed under one entry by a good dictionary is taken as the basis for calculating certain relations between words with one meaning, with two meanings, and with three meanings, and the type of text in which they occur. Dictionaries, however, have to make a choice and do not list all the meanings of a word. It is often quite difficult to decide whether a certain meaning-

92 Linguistics, literary analysis, and literary translation

feature justifies an independent listing or not. Dictionaries have to simplify and do not take the problem of overlap into account. As there are also involuntary omissions, dictionaries are not a very reliable source for numerical data about meaning. The fact that no two dictionaries yield the same data should have warned the author not to base his count on one dictionary no matter how authoritative it is. This study shows the danger of assuming that findings expressed in mathematical formulas are always trustworthy: as the dictionary foundation is very shaky, the superstructure is shaky as well in spite of its apparent scientific precision. So, whenever one reads a statistical study of a literary text the following questions should be asked. How did the author obtain his data; which criteria did he use? If no control set of texts is used, how does one know whether the numerical data that are presented have any special significance? If there is a control set of texts, how was this set selected? Which criteria were used? 4. In the fourth category of linguistic studies there may be some statistics, but the emphasis lies elsewhere. Instead of trying to discover hidden or at least not very visible features of a te-xt, the authors direct their attention precisely to the most striking characteristics of a text. In some instances the study is aimed at checking the claims made by literary critics; in others there is no element of reacting against the writings of a particular critic, but rather linguistic analysis is made of elements or features that are normally approached more impressionistically. An article by Wiecher Zwanenburg on the language of Claude Simon's LA route des Flandres falls into the fourth category.6 Contrary to a claim made by Jean Ricardou according to which the disintegration that takes place in the story is reflected by the disintegration of standard language usage, Zwanenburg shows that in spite of the extraordinary length of many of the sentences and the apparent peculiarities in subordination, co-ordination, and punctuation Simon stays within the limits set by traditional French grammar as far as subordination and co-ordination are concerned, and only leaves out punctuation marks or capital letters signalling the beginning of a new sentence when these signals are redundant and their omission does not prejudice comprehension. So, the decomposition or deconstruction of traditional language usage is not nearly as drastic as Ricardou's commentary suggests. Zwanenburg does not reject everything Ricardou says, but his careful linguistic analysis shows to what extent Ricardou exaggerated. If he had been right in everything he claimed LA route des Flandres would probably have been unreadable because if a writer wanders too far away from the language

93 Linguists and literary texts code, his work ceases to belong to the literature of the speech community to which he belongs. A book-length study of works by Proust, Gide, and Le Clezio7 examines sentence structure and length for all three authors and investigates possible relations between semantic content and type of sentence in Proust. By carrying out an analysis of immediately observable data Conrad Bureau, the author of this very original study, creates a verifiable objective framework that may serve as a basis for further, less objective evaluations. Bureau limits his analysis of proustian prose to narrative sections, excluding all directly quoted dialogue. In this way he avoids the difficulty of having to decide how to deal with incomplete sentences in an exchange of questions and answers, or with statements followed by confirmations that are incomplete and leave out elements that occurred in the preceding statement. His criterion for the sentence is syntactic cohesion. Each sentence is examined in order to establish its length as expressed in units having a primary or secondary syntactic function, and its complexity as measured by what Bureau calls its 'levels of dependency.' This dependency is manifested in the relationship of subordination. 8 After having explained at great length what his criteria are and how to apply them, Bureau demonstrates his method by applying it himself to Proust and obtaining what he calls the 'signalement' of each sentence, i.e., its characteristics expressed in length and complexity. Not unexpectedly a comparison of two thousand sentences taken from Combray to five hundred sentences from Gide's La porte etroite and another five hundred from Le Clezio' s Laguerre shows the great complexity of Proust's sentences. A notation that follows the linear development of the sentences facilitates the discovery of the structural symmetry of proustian sentences. Further examination yields data about the connection between great complexity and length - not all sentences in Proust's narrative are equally long and complex - and certain topics: room descriptions, for instance, contain among the longest and most complex sentences of the investigated sample. Bureau - and Zwanenburg' s remarks on the subject go in the same direction - states categorically that aesthetic evaluation begins at the point where the linguist's work has come to an end . Whereas the linguistic analysis is based on objective, verifiable data and carried out in a way that is explained and accounted for, the aesthetic appreciation of a text and the associations and connotations it carries depend on the taste and the receptiveness of each individual reader. The four categories of linguistic studies of literary texts are thus very

94 Linguistics, literary analysis, and literary translation different from one another in spite of the elements they have in common. The first category does not have any great impact on literary analysis and is of marginal interest for those who are concerned with the relationship between linguistics and literary analysis. It should be mentioned, however, that the literary quotations used to illustrate grammatical rules stress the general aspect of literary language and not its specificity. The second category deals with phenomena of hypersemanticity that are characteristic of literary genres. There is no sharp dividing line between studies done by linguists and those carried out by people more interested in literature than in linguistics. There is a speculative element in many of the findings presented in this category and some of the authors tend to get carried away and become highly impressionistic. The third category is partly to be seen as a reaction against the impressionistic tendencies of the second category, as statistics counteract or corroborate vague first impressions and intuitions. However, frequencies do not mean very much if not compared to what are the standard frequencies, and interpretation of the features that are presented shifts inevitably towards the subjective side of literary analysis. There is nothing wrong with that and excellent interpretations may be given, even without any precise statistics (see, e.g., chapter 11, page 70). If statistical data are provided the reader should be aware of the shift from the objective verifiable to the subjective non-verifiable and not think that the general validity of the first part guarantees the general validity of the second part. Sometimes statistical counts serve no other purpose than to enhance the scientific prestige of a publication whose intrinsic value depends on entirely different factors. The fourth category steers clear of this danger of confusing objective linguistic data and aesthetic appreciation and is often a reaction against this confusion. While the third category uses procedures that can easily be followed by non-linguists, the fourth category requires a thorough knowledge of linguistic theories and, for that reason, it is much more clearly separated from studies by non-linguists than categories 2 and 3. This short discussion of the four types of involvement with literature by linguists shows that there are many areas where the distinctions between literary and linguistic approaches become blurred. One should also keep in mind that just as some linguists are interested in literature, there are literary analysts who include linguistic considerations in their discussions. Even those not interested in linguistics per se will take into account grammatical categories such as verbal tense, mood, and voice in their analyses when they are relevant from their point of view. So, all that is linguistics is not necessarily unknown to non-linguists.

95 Linguists and literary texts This fact becomes even more clearly noticeable when one examines what linguists have to say about translation, in particular literary translation, as we shall see in the next chapters, in which linguistic approaches to translation will be discussed.

15 Linguistics and translation

Before we examine some of the major problems facing the translator of literary texts, it should be pointed out that some of them are not restricted to literary translations whereas others have a specifically literary character. Moreover, it should be kept in mind that some of the problems are linked to language differences, while other difficulties arise from differences in extra-linguistic physical setting and in social and cultural customs. We begin with the category of physical setting. The source language may refer to objects or phenomena that are unknown in the speech community of the target language, which therefore lacks adequate means to render the source text's content. Extreme examples are snow and ice and everything connected with them when they have to be expressed in aboriginal languages of the Amazon delta, or descriptions of modem big-city life in an Inuit language of the Northwest Territories in Canada, or in a Polynesian language of a small isolated island in the Pacific. Cultural and social institutions differ greatly from one nation to another and can cause trouble for translators. When I want to talk about aspects of the intricate administrative system of the University of Toronto to Dutch academics it is very difficult to use Dutch because there are no Dutch terms that correspond to those used in Toronto, the Dutch setup not sharing the functions and divisions that characterize the Toronto system. The translator can either resort to explanatory circumlocutions or create a neologism. In the latter case he may borrow the term directly from the source language or coin a term that is transparent in the target language. No matter which solution he chooses, he has to make sure that the context, together with the term in question, gives the reader enough dues to interpret the neologism correctly.

97 Linguistics and translation Although those translation problems that go back to unknown referents manifest themselves in the lexicon, they are different from those where lexical structures do not match in the source and target languages. The referents are the same but there is no one-to-one relation between the linguistic signs of the two languages involved. A comparison of the semantic (lexical) fields (see chapter 6, page 35) enables the translator to determine where the trouble-spots are. Sometimes the number of lexical units in both languages is the same but their delimitation - whether there is a no-man's land or a zone of overlap does not matter - is not identical; in other cases the number of units differs. Louis Hjelmslev gives the example of Baum, Holz, Wald in German as compared to arbre, bois, foret in French, where Wald covers a larger area than foret, and he also provides the Danish equivalents trae and skov, which together cover the area occupied by three units in German and French. 1 Even a relatively simple difference such as between English city, town, village, and French ville, village can create serious problems. First, the distinctions between the units of either of the languages are not always clear. Sometimes size is used as a yardstick, sometimes administrative classifications going back in time as far as the Middle Ages. This fuzziness does not make things easier, when one has to decide whether in a given context town should be rendered by ville or by village. However, both ville and village have at least two possible translations in English: ville- city, town; village- village, town. In many cases it would help if the translator either knew the referent named in the text or could go and visit it, but often he has to go by whatever clues the context provides or resort to simple guesswork. A slightly different situation arises when one language makes a distinction that another does not make so that a two-to-one correspondence prevails. To Latin patruus ('uncle from the father's side') and avunculus ('uncle from the mothers's side') corresponds only one term in English (uncle) or French (oncle). If an article of Roman Law dealing with the responsibilities of patruus and avunculus is translated, the distinction has to be maintained as it is pertinent in that particular context. So, the translator adds from father's side and from mother's side. However, in many texts the compulsory distinction of the source language can simply be omitted in the target language. Maintaining it everywhere would result in many cases in overtranslation (see also the theoretical description of this case, chapter 5, page 29). It is more difficult, as Roman Jakobson 2 has pointed out, to go from the less differentiated to the more differentiated language. The non-differentiated source language may not pro-

98 Linguistics, literary analysis, and literary translation vide sufficient data to decide to which of the concepts of the differentiated target language the non-differentiated term of the source text corresponds. If contextual data are lacking and extra-linguistic verification is impossible (or too time-consuming and/or too costly) the translator is again reduced to guessing. So far only lexical problems and problems connected with referents and manifested at the lexical level have been discussed. Annoying as they may be for the translator, each problem has a low frequency of occurrence. Though the problem of how to translate ville and village into English may be pressing in sociological texts about demographic patterns and urbanization, there are many texts where this particular translation problem does not arise. Uncles probably present themselves even more rarely than cities, towns, and villages. When the source and target languages do not match in regard to grammatical categories and syntactic structures the frequency with which the translation problems arise is much higher, and the mere fact of this high frequency makes laborious circumlocutions less acceptable because they could become irritating if they are repeated too often in the same text. Not all conflicts are equally cumbersome to resolve. When a given language has a dual (category indicating two), while the other language involved in the translation has only a singular (one) and a plural (more than one) most of the time the problems are minor. Of course a plural in the one language means 'more than one' and in the other 'more than two,' so that some plural forms in a text in the first language may correspond to duals in the second language, while duals of the second language have to be translated by plurals, possibly with the addition of two. In any case, if someone talks about his two brothers (dual), it is not necessary to add the specification that there are two of them everywhere the dual of the source language occurs. It is sufficient to provide this information the first time the dual is found, and for the rest the target language can use the plural without any risk of misinterpretation. Here again it is more difficult to go from the less precise to the more precise system: when the language that has only singular and plural as its number categories is the source language, the translator sometimes does not know whether to use the dual or the plural in the target language. More serious difficulties arise when verbal systems do not match. The correspondence and lack of correspondence between the Russian system that relies heavily on aspect and the French system where temporal categories are very important has been studied extensively.3 The results indicate that often meaning elements of the source language are lost

99 Linguistics and translation unless unacceptably long explanatory adverbial phrases are added, and that the target language provides meaning elements that are not expressed in the source language, for the simple reason that those elements are compulsory and the target language thus cannot avoid them. There are no steadfast rules for rendering verbal forms of one of these languages into the other, but the comparative approach makes one at least more aware of the differences between and similarities of the two systems. Very interesting work is being done by a group of Leningrad linguists specializing in problems of aspect.4 They compare original works in Russian with their translations in other languages, mainly English, French, and German, and texts written in those languages with the corresponding Russian translations. These studies in comparative aspectology form one of the rare cases where linguists use translations systematically for the purpose of gaining a better insight into both source and target languages. The material they use is literary but their findings about procedures of substitution, compensation, simplification, and interpretation followed by the translators are of general validity and not limited to literature. Apart from specific difficulties when the source text is a poem featuring stress patterns that are not allowed by the target language, most literary translation problems are also encountered, though often much less frequently, outside literature. Syntactic structures vary greatly from one language to another. Though it is a moot point whether the Sapir-Whorf hypothesis according to which different structures entail different worldviews is correct, it is certainly true that certain stylistic effects that are quite common in one language are difficult to produce in another language because of a different syntax and morphology. In languages such as Russian and Latin morphological endings that form one whole with the lexeme to which they are linked indicate the grammatical category and the syntactic function of the element that is formed in this way. If a noun is accompanied by an adjective the correspondence between the two is reflected by the identity of categories (e.g., feminine, plural) and function (e.g., direct object expressed by a case-morpheme; here the case-morpheme of the accusative). Because functions and dependencies are morphologically expressed, word order is fairly free, and variations in word order can be used for subtle changes in emphasis and for mere stylistic effects. Inasmuch as lexical and syntactic elements are put together and cannot be separated so that the syntactic function is expressed by a separate independent element, such languages are called 'synthetic.' French and English, however, generally express syntactic func-


Linguistics, literary analysis, and literary translation

tion either by a fixed word order (e.g., subject, verbal predicate, direct object, but never direct object, verbal predicate, subject, an order that is possible in Russian) or by function words such as prepositions. 5 These types of languages, that seem to analyse the units of which synthetic languages are constructed and break them down into separate elements, are called 'analytic.' Analytic languages are considerably less free in changing word-order, and it is difficult to match the word order effects of a synthetic language in an analytic language. One can, of course, resort to other means such as special grammatical structures signalling emphasis (e.g., French: c'est X que Y aime for Y aime X) or lexical reinforcements (Y aime X specialement), but the effect is never the same as in the synthetic source language, the substitutes for synthetic word-order variations often being much heavier and more emphatic and/or having a different stylistic impact. Comparisons between different morphological and syntactic systems fall under language typology.6 For translators typological studies that show them what can and what cannot be done from a linguistic point of view are very important. Even those who study the different types of narrative structures may find interesting observations in general typological studies that show that in some languages certain types of narrative presentation are impossible, and that the assumptions about the general validity of some statements have to be re-examined. Typological studies do not concern themselves with stylistics and nondenotative aspects of language, but when comprehensive comparisons are made between two languages, cultural and aesthetic considerations enter the picture, so it is certainly not fortuitous that the titles of some major works dealing with two languages from a comparative point of view contain the term style or stylistic.? The stylistic and aesthetic elements grow gradually more important if one moves from scientific prose, to literary prose, to poetry. In scientific texts the signified always takes precedence over the signifier: as long as the denotative message remains intact, the form in which it is presented does not really matter very much. Of course a nice, elegant formulation is preferred when possible, but if clarity requires the use of heavy and cumbersome circumlocutions those are preferable to sacrificing clarity for elegance. The more important a form becomes, the more limited the options at the level of the signifier become. Rhyme and rhythm impose severe restrictions on the translator who wants to maintain these formal features in the target language. Sound-expressivity is another element that, although it is not as directly noticeable and cannot be established in the objective way in


Linguistics and translation

which rhyme and rhythm are given, plays a very prominent role in certain literary genres. Association and connotation, whether the result of the author's intention or not, also create serious problems for the translator of literary texts. An example of values attached to sounds without the author's conscious awareness can be found in psychoanalytical sound interpretations. 8 Thus, the sound [r] (it should be the trilled variety) is linked with virility and called 'erectile r.' If these values are to be maintained in a translation, the denotative content cannot be kept. However, the potential of the sounds is only realized in connection with meaning. The same problem has to be faced when one tries to translate texts without any apparent denotative meaning. Even though rhythmic and phonic effects are of overriding importance in such texts, it should be kept in mind that they are produced by language signs, or in the most extreme case by pseudo-signs. These pseudo-signs will have different evocative effects depending on which language serves as a point of reference. In other words, even those elements should be translated. Not surprisingly, translations of such texts are very hard to find (see also chapter 11, pages 68-9) . Finally, word-plays, puns, and overt metalinguistic comments about a person's pronunciation or about his idiosyncratic use of a word or a grammatical category are so language-specific that translating them often turns out to be impossible. The picture that emerges from this chapter is not entirely reassuring for those who might have high hopes that a linguistic approach will solve their translation problems. Linguistic comparative analysis helps to bring some of the difficulties more sharply into focus and enables one to distinguish more clearly between linguistic hurdles and non-linguistic obstacles. However, knowing what the problem is does not always help to find a satisfactory solution. It may be a consolation for a translator when a linguist gives him more insight into his difficulties by telling him why the text he has to translate (text, syntactic structure, expression, or lexical item, it does not matter what) cannot possibly be translated into the target language. Sometimes, however, the linguist's explanation causes annoyance, the translator himself realizing prefectly well why his task is next to impossible and only wanting practical advice or better still a general theory of translation that will enable him to solve his practical difficulties for himself. Asking for such a theory is unrealistic, and in spite of some misleading titles suggesting that there is a general theory of translation, most linguists who have written about translation


Linguistics, literary analysis, and literary translation

do not present their reflections on the subject under the flag of a general theory. In the next chapter it will become clear why there is no general theory and why very often the translator must decide for himself which practical road to follow.


Various options

Both linguistics and literary analysis raise a number of questions that are of the utmost importance to the translator. To a large extent his translation practice will depend on the answers to these questions .. Because most questions can be answered in more than one way, he is faced with a series of choices. Which answers he ultimately chooses depends partly on the type of text he is dealing with, and partly on personal taste. In other words, the same translator may treat different texts in different ways, and the same text may be treated differently by different translators. 1 As we have seen in the previous chapter, typological and comparative studies provide an inventory of what might be called the 'linguistic trouble-spots' in the translator's task. When only denotation is taken into account - on the assumption that such a restrictive approach is possible - circumlocutions and neologisms may provide some solutions, but when associations, connotations, and expressivity enter the picture the situation becomes much more complicated. The linguistic elements are inextricably mixed with cultural and aesthetic components. This amalgamation makes it impossible to formulate a theory that covers all possible translation situations, and is at the same time specific enough to provide some guidance to the translator when he is faced with a certain linguistic problem. General linguistic studies on translation, such as the works by Georges Mounin, Jean-Rene Ladmiral, and John C. Catford2 give the reasons why translating is difficult rather than indicating how to solve the difficulties. Comparative studies such as the one by Jean-Paul Vinay and Jean Darbelnet on English and French3 do give some practical advice based on linguistic analysis and on relative frequency when a construction is known in both target and source Ian-

104 Linguistics, literary analysis, and literary translation

guages (e.g., though French has a passive-voice construction it is used much less frequently than in English and maintaining all the occurrences of the passive voice of an English source text in the French target text would give that text a definitely un-French flavour; see, however, below, page 107, for a different opinion). However, Vinay and Darbelnet do not pay special attention to literary texts and the particular choices their translators have to make. The most comprehensive translation theory where all the different aspects are linked together and very explicit practical advice concludes each theoretical discussion is offered by Eugene Nida and Charles Taber.4 However, the authors are only able to offer a coherent and comprehensive analysis combined with guide-lines for translators by limiting themselves to one specific text, the Bible, seen not as a work of art, but as a text written to educate and instruct, to convince and to convert. Therefore, some of the choices that the translator has to make and that may be far from obvious in other circumstances do not cause any problems for Nida and Taber. In view of the primary function of the Bible, which is to spread the message of the Christian faith, the clarity of the message is given priority over all other considerations: level of speech and accessibility should be adapted to the audience for which the message is intended regardless of the level of the original passage. If stylistic features can be preserved, so much the better, but they are not vital. The same holds for metaphors, idiomatic expressions, and proverbs: maintain them in the target language if possible, replace them by equivalent but not identical elements if they cannot be kept unchanged, and resort to circumlocutions if those equivalents are not available. The choices made by Nida and Taber favour denotative content over connotative and associative elements, favour content over form, and are decoderoriented. They even stress the importance of accessibility at the oral level because the texts of the Bible are often used in sermons that are orally delivered before an audience. It is, of course, an exception in literature that the purpose can be as clearly defined as it is for the Bible, and even when the author's intentions are unequivocally expressed, few translators are likely to simplify the text stylistically in order to make it more easily accessible for a larger circle of readers. Yet adaptations to readers' expectations are rather common, as will become evident from the discussion of some of the major dilemmas that face the translator of literature. 1. The most general problem is undoubtedly that of the identity of the text. Structuralist linguistics stresses the interdependence of Ian-

105 Various options

guage elements belonging to the same level. A phoneme is defined with respect to all other phonemes of the system and could not be defined without them. For other units belonging to limited and closed inventories (see chapter 6, page 36) the defining process is identical; for the lexicon it is similar, the only difference being that the lexicon is not treated as one whole, but is first split up into smaller coherent lexical fields. As the lexicon forms an open, unlimited inventory, fluctuations from one idiolect to another are not easy to document, and many persons have a passive lexicon at their disposal (lexical terms they are able to interpret correctly) that is much richer than the active lexicon they possess (lexical terms they can correctly use themselves when formulating a message). In any case, when a reader decodes a text, he falls back on his own linguistic inventories - limited and unlimited, active and passive - and on his own extra-linguistic experience. It is undeniable that different readers do not interpret a given text exactly identically. These individual differences in readings have led, especially in modern text analyses, to the formula that each reader creates his own text. One step farther and this creative act of the reader is a necessary condition for a text to exist: without a reader there is no text at all. This paradoxical statement has far-reaching implications for translating. 5 a After the translator has created his own text in the source language he can translate it into the target language. Without a reader there is no target text, but readers will create their own text based on the results of interpretation, creation, and translation activities of the translator. b As the emphasis is on the interpretive-creative aspect of the translator's task, the identity of what used to be called the 'original text' (written by a novelist, a poet, a playwright) recedes into the background. c From a practical point of view discussions about the quality of different translations become meaningless; it is only possible to compare interpretations. One is inclined to like interpretations that come close to one's own better than others, but such preferences do not imply that translations based on other interpretations are not equally good. Not every translator will be happy with the responsibilities of creative freedom, because they might feel that the author and his text deserve a more prominent position and that there are limits to the freedom of creating one's own text and one's own translation. 6 2. The extreme opposite position, denying the translator any creative freedom, is taken by Vladimir Nabokov. The title of his contribution to a colloquium on translation is very eloquent: 'The Servile Path.' 7 The

106 Linguistics, literary analysis, and literary translation

servile path that the translator must follow, according to Nabokov, is in some respects easier, in other respects much more arduous, than the road of individual creativity. The translator does not have to emulate the author; he should not even try to maintain formal features or find their equivalent in the target language. A simple prose translation of a poem is much more honest than, and to be preferred to, a rendering in verse that suffers from formal constraints that lead to artificiality and clumsiness. However, minute details pertaining to the author's background should not escape the translator's attention. In a translation of Evgenij Onegin the fact that Puskin used bad French translations instead of reading the original texts by Byron must in some way be expressed, Puskin' s Russian quotations and allusions going back to the French rather than to the English. Especially when the target language is either English or French this requirement makes the translator run the risk of being accused of not knowing Byron unless he explains the reasons for his translation in a footnote. In other target languages the difference caused by referring to the French translation of dubious quality rather than directly to Byron's text is likely to remain unnoticed, as it is unnoticed by the overwhelming majority of the Russian readers of Puskin's original text. This raises two questions, the first of which is directly related to Nabokov's views on translation, while the second has a much wider scope: First, is the servile path really serving the text, or does it lead to demonstrations of scholarly erudition and an amazingly broad cultural background? In the latter case translations made according to Nabokov's requirements would address themselves to the scholarly reader rather than to the readers who want to read a work in translation because they cannot read the original, but who do not want to be distracted by a great amount of background information, no matter how interesting it may be, given in footnotes or in explanatory additions to the text. Second, how does the translator deal not only with rare cases such as Byron in Pu:skin but also in general with flaws and stylistic peculiarities in the source text? 3. When I was a student I once wrote an essay on Les Faux-Monnayeurs by Andre Gide. A helpful assistant looked over the first draft before I handed in the final version to the professor, who was known to pay special attention to our command of French. When I was handed back the rough copy there were quite a few suggestions for improvement in a long passage that was taken from the Journal des Faux-Monnayeurs and that was inserted in my own reflections marked only by quotation marks that were not very clear and had not been noticed. The lesson to be


Various options

drawn from this anecdote is clear: a widely respected author can permit himself liberties with his mother tongue that are considered mistakes or infelicities when used by a foreigner. The translator is usually in a different position vis-a-vis the foreign language, because he expresses himself in the target language, which as a rule is his mother tongue, translators who translate literary works from their mother tongue into a foreign language being very rare. In any case, a stylistic peculiarity in the translation that is a faithful rendition of the style of the source text runs the risk of being ascribed to the incompetence of the translator by monolingual readers who are unable to compare source and target texts. Moreover, there is a general danger of staying too close to the constructions of the source language. On the whole, translation specialists voice strong warnings against using an identical construction when that construction, though possible, is very rare. Yet Antoine Berman8 is of a different opinion. Readers, he claims, easily accept and even appreciate what he calls l'etrangete du texte ('oddness of the text'). Most translators try to avoid this oddness, though, and many stylistic peculiarities of particular writers are smoothed out in the process. If the translator, in spite of the risk we just mentioned, chooses to maintain certain idiosyncratic stylistic features, he can only do so if the target system allows it. If not, he may look for a substitute which he hopes will have an effect similar to that of the construction of the original text. Compromises are not uncommon. The extraordinary length and complexity of the Thomas Mann sentence is partly based on synthetic features of German (case morphemes and gender distinctions) that are unknown in English. The length and complexity of sentences in an English translation of Der Zauberberg do not equal those of the German original, many of Mann's sentences having been split into two or three independent sentences in English. Yet the impression is maintained because, relative to English standards, the average sentence of the translation is as long and complex as Mann's sentence is in relation to the norm in German literary prose. 9 The situation is different when either the target language or the source language has a strict codified system of stylistic levels whereas the other language does not. If a certain type of poetry is associated with a specific stylistic level, the translator will choose that level if the target language is differentiated. In the other direction, when the source language is stratified and the target language is not, translation does not involve any choice. (See for similar situations at the lexical and the grammatical level, chapter 5, page 29; chapter 6, page 36; chapter 15, pages 97-9.) One could say that the choice of the usual stylistic level is in ac-

108 Linguistics, literary analysis, and literary translation

cordance with the target code and therefore legitimate. It also caters to readers' expectations, and the result may be so different from the original that it hardly qualifies as a translation. Voltaire's 'improvement' on Shakespeare is a good example of stylistic adaptation to readers' expectations. Hamlet's monologue turns into a passage of a French tragedy, written in inevitable alexandrines with characteristic lofty vocabulary.10 4. Voltaire's adaptation of Hamlet's monologue to the laws of French tragedy is in many respects the result of a clash between the formal requirements of English and French. Yet it is not a typical example of a linguistically conditioned conflict (and in this case, its not very satisfactory solution) because the French language allows - and allowed in Voltaire's days - translations that come much closer to the original than does Voltaire's text. Genuinely linguistic conflicts arise when formal features of the source language cannot be reproduced in the target language, or can only be reproduced by distorting the natural form of the target language. In poetry, formal conflicts are more likely to manifest themselves than in prose, as formal constraints are much more stringent. Fixed rhythmic patterns are very common in poetry. Rhythm is not an independent entity, but is closely linked to the normal accentuation patterns of a language and, if they are present, of quantity oppositions. Classical poetry is completely changed when translations replace quantity patterns by stress patterns, but this is not very noticeable, because classical poetry is read by modern readers with stress replacing quantity. The syllabic structure of poetry in languages with a fixed accent (French, Polish) cannot be maintained in languages with a free accentuation (Russian, English). In the seventeenth century, Russian translations and imitations of Polish poetry maintained the structure of the original. From the Polish wierszy (verses), the Russians coined virsi to indicate syllabic verses. Later on, when it was realized that Russian did not lend itself to Polish versification patterns, the term survived with the meaning 'doggerel rhymes.' Certain rhythmic patterns are very suitable for the expression of various emotions or for the creation of an impression of anxiety, languor, or vitality. In so far as the rhythm reinforces the denotative message of the text, a change in rhythm in the target text may considerably weaken the impact of the text. The formal conflict between source and target languages thus affects the message, and an element of meaning, though its precise contents are difficult to define, enters the picture. There are many different types of formal conflicts directly involving meaning: in all cases the target language is unable to match the source

109 Various options

language in such a way that the special relationship between form and meaning characterizing the source text is preserved. Puns, word-plays, and metalinguistic comments form a category where the problem is very easily detected and rarely solved in a satisfactory way. Even a silly wordplay such as C'est ma belle mere. Tu te trompes, elle est laide ('She is my mother-in-law. You are wrong, she is ugly') creates a problem. Metalinguistic commentaries can sometimes be replaced. When Clamence, the hero of Albert Camus's novel La chute, says about himself that he likes the imperfect subjunctive, he thus characterizes himself as an oldfashioned conservative who likes elegant language. The translator of the English text takes the English subjunctive as an equivalent, puts an extra subjunctive in just before the metalinguistic comment, and solves the difficulty created by the absence of the imperfect subjunctive rather successfully. 11 In other instances, when substitution by a more or less equivalent category is not possible solutions are much harder to find for instance, when the source language has a well-develop~d morphological set of gender markers and grammatical gender is unknown (or differently structured) in the target language. When one of the secondary characters in Dostoevskij' s novel The Idiot states that he always says Rimskaja papa, he knows it should be Rimskij papa ('Roman pope'), papa, in spite of the final a that is normally associated with the feminine gender, being masculine and therefore requiring the masculine form of the adjective (rimskij). 12 The switch of gender that is probably intended to have a derogatory effect cannot be rendered in English. French offers romaine vs. romain, but the translator has to decide whether the definite article should join the adjective (la pape romaine) or remain masculine, and German der romische Pabst or die romische Pabst does not distinguish between masculine and feminine after the definite article in the nominative singular so that the translator has to rely on the definite article in order to express the gender switch. Dutch has a gender distinction opposing neuter on the one hand to masculine/feminine on the other, but it is not morphologically equipped to render the opposition of the Russian text between masculine and feminine. 13 Whatever the translator eventually decides to do, he knows that the troublesome passage is part of the intentional message of the original. Other aspects of hypersemanticized texts (see chapter 11, pages 66--9 and also chapter 9, pages 57-8) show varying degrees of intentionality and may be so inconspicuous that their omission in the translation may be explained by the fact that the translator did not notice them. It is, of course, also possible that he did perceive them but was not able to match


Linguistics, literary analysis, and literary translation

them because of a formal conflict. Sound expressivity is difficult to maintain if the denotative message has to remain unchanged. Trak.l's Holunder (see chapter 11, page 70) becomes a vlier [vlir] in Dutch, an elder in English, and a sureau [siiro] in French. The special link mentioned by, among others, Lotman and Weinreich, created by rhyme, between elements that would normally not be associated with one another is, when the translation is also rhyming, usually replaced by another link because other signs than those that translate the rhyming signs (to be more precise the rhyming signifiers of the signs) of the original are in the rhyming positions of the translation. This new link, forced upon the translator by formal constraints and content requirements, is rarely motivated in the way the original link was. Another example of hypersemanticized elements is when grammatical gender suggests that a personified inanimate noun has a sexual identity. In an article on Southern illusion and Northern reality in the early lyrical poetry of Heinrich Heine, Heinz Wetzel1 4 analyses, among others, a poem about a fir tree and a palm tree, separated from one another, the fir tree standing in snow and ice, the palm tree, as lonely as the fir tree, being scorched by the sun near a wall of rocks. In German the fir tree (der Fichtenbaum) is masculine and the palm tree (die Palme) is feminine, stressing the opposition between the North and South, 15 cold and warm, male and female, and hinting at the impossibility of a happy union of man and woman who remain lonely and sad while longing for each other. Roman Jakobson compared the same Heine poem with a translation into Russian by Lermontov in a lecture he gave in Toronto in the late sixties. 16 In spite of the high poetic qualities of Lermontov' s translation, a poem well known in its own right throughout Russia, Jakobson confessed to feeling uneasy because both trees have the feminine gender in Russian. In such instances target languages without grammatical gender that only leave out the hypersemantic gender element are preferable to languages that do indicate gender but cannot match the distribution of the source text for lexical reasons. Although one cannot prove that the gender distribution in Heine's poem is intentional, it is highly unlikely that the masculine/ feminine reinforcement of the main oppositional theme of the poem is purely fortuitous. 5. ·Intention is not at issue in the category of old literary texts where archaic language features are often linked with cultural heritage. The authors of those texts used the language of their time, and their contemporaries read their works and watched their plays as modern, new creations. Thus, there is some justification for the choice to translate


Various options

such texts into the target language without paying attention to their archaic features. The claim to present a text that modem readers will appreciate exactly in the same way as the original text was appreciated long ago is based on false assumptions, however, for it is impossible to change the readers into people who are contemporaries of the author of the original (see also chapter 13, pages 83-4). The other extreme would be to translate Fran~ois Villon into fifteenth-century English, but no modem translator would be able to do that. A special additional problem would be to select the period of the target language that corresponds to that of the source language. Nineteenth-century Russian may well be more adequate a tool for rendering seventeenth-century French literature than seventeenth-century Russian would be. Many translators opt for a compromise, using the language of their own time but introducing some archaic expressions into their text in order to give it a flavour of dignity that conforms to readers' expectations about an ancient text. This sort of compromise is not without some similarities to the solution translators find when faced with a text containing passages written in a dialect. Whether as the main language used or in contrast to another variant (in most cases the standard language) dialects are not easy to deal with in translations. Foreign languages used in contrast to the main language of the text also raise some questions as will be seen in the next chapter, in which dialects and foreign languages will be examined from the point of view of the translator.

17 Foreign languages and dialects

In chapters 8 (page 50) and 9 (page 55) some of the intricacies of dialect representation in writing and of the interpretation of dialect forms when used in alternation with standard language, or even as the main vehicle of expression, were discussed. Depending on the function of the dialect form it will be more or less important to look for an equivalent in the target language. A survey of a few literary texts in which Quebecor Acadian-French dialect forms occur and of their English translations will show whether there is any correlation between the function of the dialect in the text and the way the translators render it in their target texts. The four works selected as a sample also provide the opportunity to examine a special aspect of the translation of single words or sentences, and even entire passages, belonging to a language other than the main language in which the text is written. Before discussing the special situation where the embedded foreign-language elements of the source text belong to the target language, we should have a brief look at the phenomenon of foreign-language use in literature in general. It is not easy to define what constitutes a foreign element in a text. On the one hand, there are words spoken by a foreigner in his own language or terms referring to a specific referent in a foreign country. On the other hand, one finds elements that are more or less integrated, the integration often manifesting itself by orthographic and/or morphological adaptation to the systems of the receiving language.' When, in an English text, words like milieu and gourmet are found their foreignness is reduced by the fact that they are commonly used. Their foreign origin is recognizable but they have ceased to be foreign enough to function as special markers. Special marking functions of foreign elements are 1 / to signal the

113 Foreign languages and dialects

foreign origin or nationality of a character; 2 / to describe a foreign setting - couleur locale - even if the main language is adequately equipped for the description; 3 / to create an impression of cosmopolitan elegance. In an in-depth study of the effects that can be obtained by inserting foreignlanguage elements in a literary text Petra Braselmann2 examines the techniques used by a French novelist, Maurice Dekobra, a writer of pulp novels that were very popular between the wars. Her conclusion is that the foreign element never carries vital information without which it is impossible to understand what is going on. In some fairly rare cases the meaning does not become directly clear, but the reader who cannot miss the main function the element has (i.e., marking foreignness) also understands that the denotative function is secondary if not entirely dispensable. More often the meaning can be deduced from the context in which the foreign element is placed; the unknown foreign term can be interpreted either directly or after it has been used in a few different settings. Finally, a very common technique consists in repeating the foreign element immediately in the next sentence or in the same paragraph, using its equivalent in the main language, or in using the foreign element as a repeat of something that has been expressed just before in the main language. No matter which technique is used the position of the foreign word or sentence is marginal in the development of the story. When its function consists of announcing the ethnic origin of the speaker, or of reminding the reader of this origin, very often the reader can expect to be confronted with ethnic stereotypes. Dekobra's rather superficial use of foreign languages does not exhaust all the possibilities they offer; far from it. However, the more important the function of the foreign language becomes, the more difficult it gets to make a monolingual reader understand what is said in it without resorting to translations offered in footnotes. In nineteenth-century Russian novels French is very frequently used in directly reported speech. In War and Peace (Vojna i Mir) one finds pages where the only Russian Tolstoj uses is 'said the count,' 'answered the princess,' little sentences that are almost lost in the sea of French, the language in which all the conversations of members of the Russian gentry take place. 3 Because French was familiar to the public for whom Tolstoj wrote, the question of how to make sure that monolingual Russian readers understood the French passages did not arise. However, when French was no longer the language of communication of an important part of the Russian reading public, a change caused by the Revolution of 1917, it became necessary to provide translations.

114 Linguistics, literary analysis, and literary translation

Soviet editions of the Russian classics use the system of footnotes, which is adequate and satisfactory in the overwhelming majority of cases. In some instances French is used, not to render the conversations that take place in high society, but to characterize someone who would like to belong to that society. When the sister-in-law in Cechov's play The Three Sisters (Tri sestry) says something in French, she shows that she has not mastered that language, thus giving away at the same time both her lack of education and her ambitions. Ideally, the Russian translation should throw the same light on the character as the French does, but even if the Russian translation had some grammar mistakes or an air of clumsiness similar to that of the French, one wonders whether the monolingual reader would draw the same conclusions from the Russian text as the bilingual reader did from the French original. An explanatory note may be needed to tell the reader what the impact of the clumsy or incorrect French is meant to be. 4 When bilingual or multilingual texts are translated, as a rule only the main language of the text is replaced, the foreign elements remaining unchanged. However, when the main language and the foreign languages are mutually understandable, be it imperfectly, the translator will usually treat the secondary language of his text as if it were a dialect, or ignore the difference altogether. French, English, German, and Dutch translations of Solochov's novel And Quiet Flows the Don (Tichij Don) do translate the Ukrainian while translating the Russian.As Polish and Ukrainian are approximately as far apart as Russian and Ukrainian, a translation into Polish could keep the Ukrainian passages of the original. Even when the foreign languages are kept untouched, as they are in most translations, there may be a shift in connotations because ethnic stereotypes and prejudices vary from one people to another. There being no solution to this problem, one has to be resigned to accept this state of affairs. A special complication that cannot always be resolved arises when the target language coincides with a functionally important second language of the source text. In French translations of the Russian nineteenth-century classics French passages in the original are frequently singled out by a footnote. As long as the register of the French is comparable to that of the Russian spoken by the same character there is no real problem. Tolstoj's gentry is equally urbane and polished in both languages and the French that results from the translation of what is said in Russian is in harmony with the French that is kept unchanged from Tolstoj's original work.5 For Cechov's character, the sister-in-law

115 Foreign languages and dialects

of The Three Sisters, there is a wide gap between the native Russian and the foreign French, a language that she is not sufficiently fluent in to express herself with any naturalness. The French translation ends up with a mixture of fluent French corresponding to the Russian and bad French used in the original. This bad French that sheds a light on some aspects of the background and aspirations of the sister-in-law clashes with the rest of what she says. A footnote in the printed text may go some way in explaining the difference but a performance of the play in French cannot rely on footnotes. So, the translation either accepts the inconsistencies of two levels of French used in the same circumstances by the same character without any social or psychological explanations, or it simply brings the deficient French up to the normal standard. In either case an important clue is lost. The same clash between two levels of mastery of the same language can be observed in African novels written in French. Although these novels are not translations in the strict sense of the term and therefore are beyond the scope of this chapter, a few words will be said about one novel, Une vie de boy by Ferdinand Oyono, 6 in order to give some insight into the language problems African writers who have chosen either French or English as their means of expression have to cope with. Apart from the problem of using a language that is not always adequately equipped to express the extra-linguistic reality and that presents, by the structuring of its notional fields and grammatical categories, a worldview different from that of his first language, Oyono has to solve the problem of presenting a faithful picture of communication in a bilingual colonial society while using only French. Conversations among blacks are presented in French with a touch of what Berman has called etrangete ('oddness') in a slightly different translation situation (see chapter 16, page 107). These same people speak what the French call petit negre in their contacts with the white colonial masters. The protagonist who speaks French fluently, having learned that language from a white priest who was kinder than the other whites, forms an exception. Oyono handles this difficult task on the whole quite well. Because the level of French of the same person is conditioned by the social setting in which his conversation takes place, switches are always understandable for the reader, unlike what happens in the techov play discussed above (page 114). Yet there is some loss when a bilingual situation is presented in only one language. No matter what kind of solution is chosen for the rendition of regionalisms and English elements, English translations of French-


Linguistics, literary analysis, and literary translation

Canadian works in which English plays some role will have great trouble maintaining the effects of the bilingual situation prevailing in the original texts. As for dialect features (see for a general discussion chapter 9, pages 55-7), they are always difficult to match in the target language, especially when the dialect in the source language is geographically marked. Even if the Scottish Highlands form, in many respects, a good counterpart for the Massif Central, translating dialect features of the speech of a peasant from the Massif Central by peculiarities of English spoken by Highlanders is risky, because the English reader tends to have precise associations with that dialect that would be disturbing in a French setting. Sociolects that are less determined geographically are less difficult to translate although there may be problems if the sociolinguistic stratification of one of the two languages is much more elaborate and rigorous than that of the other. The dialect situation in French Canada is far from homogeneous. Some lexical items are known throughout French-speaking communities, often referring to a Canadian reality that is not commonly known in France. Even purists who advocate the elimination of almost everything that distinguishes Canadian French from the French spoken in France accept these items, sometimes called canadianismes de bon aloi ('Canadianisms that are respectable'). Other Canadianisms are not so widely used, being either socially stigmatized or restricted to a small area. As was stated before, the two types of deviation from standard language, social and geographical, often fuse so that it is impossible to separate them from each other. The problems of different levels of Canadianisms and of deliberate and unintentional Anglicisms is examined by Jules Tessier in his study 'Les particularites de vocabulaire dans l'oeuvre de Felix-Antoine Savard.' 7 Tessier compares the five successive editions of Menaud, maftredraveur, and arrives at the conclusion that it is not always possible to know why Savard replaced standard by non-standard terms or vice versa, from one edition to the next. For the translator it is thus, at times, very difficult to determine whether an Anglicism or dialectal feature should be ignored, because it was not intentional, or should be maintained as a deliberate element of the text. In the four French-Canadian prose works of fiction chosen as a sample, Canadianisms and English elements of different types occur. 8 The Canadian aspect manifests itself at the level of the pronunciation (a deviant orthography representing a deviant pronunciation) of the lexicon and less frequently at the level of the syntax and the morphology. English is found in the form of direct borrowings with the original English or-

117 Foreign languages and dialects

thography - these borrowings being single words, or entire sentences and idiomatic expressions - or as more or less recognizable Anglicisms that have undergone a process of phonetidphonemic and morphological integration into the French systems. Where that integration occurs, dialect element and foreign element coincide. English in French-Canadian literature is not treated in the manner described by Petra Braselmann (see above, page 113). The reader is supposed to know English sufficiently to understand the English elements without contextual clues. In spite of the fact that English and Canadianisms play a role in Maria Chapdelaine by Louis Hemon, Bonheur d'occasion by Gabrielle Roy, La guerre, yes sir! by Roch Carrier, as well as La Sagouine by Antonine Maillet it is better not to make any general statements, there being significant differences from one book to another. 9 In Maria Chapdelaine English and regional French-Canadian features are used to reinforce the couleur locale of the harsh and yet idyllic pioneer existence in the Lac-Saint-Jean region. There is no attempt to give any special psychological motivation for the use of English terms such as rough, toffe ('tough') or to differentiate the protagonists linguistically by having them use more or fewer Canadianisms. As the author Louis Hemon came from France and spent only a little over a year in the Lac-Saint-Jean region, the marginal function of foreign and non-standard elements is not surprising. The English translation does not try to use dialect features to match the Canadianisms, nor does it prevent the English elements of the original text from fusing completely with the English of the translation. Even when the original has a deviant spelling (toffe, bddrant or badrer 'to bother,' which are more or less integrated forms), the translation does not show any special treatment, as the normal English orthography is reinstated. Inasmuch as the translator does not justify his omissions, one does not know whether he opted for an undifferentiated English text deliberately, because he judged the deviations in the French text superfluous, or left them out, not being able to keep them. Gabrielle Roy depicts in Bonheur d'occasion (The Tin Flute) a workingclass Montreal family. The socially stratified big-city milieu gives her the opportunity to make a subtle use of dialect and English not only to differentiate between the main characters' social backgrounds but also to indicate moods, English being used when emotions run so high that without the distance created by the foreign language they might be impossible to control, or to sketch a social setting, the same character using standard French or Quebec French expressions according to the situation. Unlike the discourse of Hemon's characters, which is unnat-

118 Linguistics, literary analysis, and literary translation

urally elegant with complex subordinated clauses that are unexpected, to say the least, in the mouth of a timid taciturn pioneer who leads a solitary life in the woods, the discourse of Gabrielle Roy's characters has an authentic ring and English and Quebec French have a more important function than reminding the reader that the action takes place in Montreal. The English translation does not keep the dialect features, Quebec peculiarities being treated in the same way as standard French, and the English of the original fusing with the English of the translation. One example is enough to show what is lost. Mais le petit gars qui marchait a cote de moi c'ui de la bargain et du din d'oeil, ii etait pas lonely pantoute, ce petit gars-la

becomes But the little chap beside me, the one who winked and made the bargain, he was not the least bit lonely.

Although no justification for this kind of simplification is given, at least the reader is warned in the introduction to the English edition that some elements are lost: 'Miss Roy's second gift ... is her flair for capturing with an uncanny fidelity the accent and idiom of French Canada (whether rural patois or Montreal argot). Unhappily the reader of the English translation must take this gift largely on trust.' In contrast to the translators of Hemon and Gabrielle Roy, the translator of Roch Carrier's Laguerre, yes sir! provides some information about the choices that were made, in a 'Note de la traductrice' preceding the actual translation. In this novel the functions of Quebec French and English are completely different from their role in Bonheur d'occasion. English is often motivated by the presence of an Anglophone, the whole novel dealing with a confrontation between French-Canadian villagers and some English-speaking outsiders belonging to the army. There are few dialect features in the speech of the villagers, and as their community is very homogeneous from a social point of view, those features do not function as social differentiators. However, one characteristic of Quebec speech habits absent in Maria Chapdelaine and sparsely used in Bonheur d'occasion is abundantly used by Carrier, whose characters all swear profusely. The translator correctly states that there is no equivalent in English for the typical French-Canadian profanities. An added difficulty stems from the fact that the villagers misinterpret prayers, saying pleine et grasse ('round and fat') instead of pleine de grace ('full of grace'). However, her solution to keep garbled religious expressions and profanations in French in her English text seems too easy a way out and leaves the

119 Foreign languages and dialects

monolingual English reader baffled. Nothing is said in the introductory note about how to translate the English of the source text or how to deal with the garbled French of the 'Anglos,' except that the translation of the title 'The war, oui monsieur!,' is explicitly rejected in favour of the original title. Yet some effort has been put into maintaining the bilingual situation; especially when the English speak incorrect French, the incorrect English that matches it is accompanied by a metalinguistic comment 'replied in French' or 'still speaking French.' Except for one case where the English translation has Atten ... shun! for the original Attention! also English as the context clearly shows, an orthography that may be interpreted as an attempt to indicate that the source text was in English here, the English of the original loses its contrastive position in the English target text. The same is true for the English elements in La Sagouine, a monologue in the Acadian French of a fishing village. The special role of the language is emphasized in the preface: 'C'est par elle [i.e., the language] que la Sagouine est le plus fortement ce qu'elle est. [It is through her language that la Sagouine most strongly is what she is.]' What can a translator do in such a case? There is only one character who speaks a strictly localized dialect. No contrastive text, no differentiated levels of speech of various characters, break the unity of the text. A comparable, equally strictly localized English dialect has the already-mentioned drawback of calling forth special associations that clash with the setting of the original. Moreover, translators who have a sufficient active knowledge of such a dialect to use it for their translation of La Sagouine are extremely hard to find . The solution the translator has chosen can be called compensatory: dialect is replaced by very colloquial and substandard English. The result is not completely satisfactory, because La Sagouine's language, which is such an important part of her personality according to the just-quoted preface, comes very close to that of Holden Caulfield in Salinger's The Catcher in the Rye. The first impression one gets from the brief survey of English translations of French-Canadian prose fiction is that translation theory and translation practice are worlds apart. The solutions translators choose do not depend on theoretical considerations, but on practical abilities. The second impression is that the literary importance of an element is far from being a guarantee that the translation will maintain it. In the next chapter we shall discuss the question these impressions inevitably raise: What can linguists, literary analysts, and translators learn from one another?


Answerable and unanswerable questions

When turning to another discipline for help or clarification of a moot point in one's own field, one should not have unreasonable expectations and blame a colleague for being incompetent if he cannot provide answers that simply are not there. When three disciplines are involved, each of which may find itself in the position of provider of help to, or in that of receiver of aid from, another discipline there are six different situations to be distinguished. Without being exhaustive the following enumeration may give some idea of what kind of benefits can be gained from interdisciplinary contacts in each of these six situations. 1. LINGUISTICS TO LITERARY ANALYSIS

Apart from providing literary analysis with an example of a discipline dealing with language data, using clearly defined terms and making statements that are verifiable, linguistics stresses the basic principle of communication according to which speaker and hearer, writer and reader, must start from the same code in order to have felicitous communication. When too much emphasis is placed on reading as a creative act, literary analysis runs the risk of eliminating the identity of the text and of rejecting the very basis of literature. Semantic field studies and more recently socio-linguistic surveys can be used in literary analysis in order to gain a deeper insight into the possible motives an author may have had for choosing certain forms. Linguistics is unable, though, to give unequivocal answers to questions about the associations and connotations that a text may trigger. Only the principle of hypersemanticization can be formulated, but no precise statements about a given text can be made in such a way that they are objectively verifiable. In the area of expressivity the same ob-


Answerable and unanswerable questions

servation can be made. Although certain trends concerning the link between sound (signifier) and meaning (signified) can be shown to be a general phenomenon not limited to specific languages, and although the likelihood that expressive features occur more frequently in certain types of literature than in others can also be demonstrated, linguistics cannot give a definitive answer, yes or no, when a literary analyst is in doubt whether or not, in a specific case, he is dealing with expressivity. 2. LINGUISTICS TO TRANSLATION

i.inguistics provides the translator with typological descriptions that enable him to compare how various languages express certain basic relationships and actions. Contrastive analyses of two languages including not only a typological description of each of them but also comparisons between semantic fields help in many instances in finding the best equivalent in the target language for an item in the source language. Linguistics can point out that there is a problem when there is a one-to-two correspondence, and that the problem is more difficult when the source language has one item against the target language's two than when it is the other way round, but linguistics does not come up with a generally valid solution. The translator who did not know that his difficulty was caused by intrinsic incompatibilities of the two languages may find some consolation in the knowledge that nobody can provide a solution, but the consolation is meagre. If there is an onomatopoeic or expressive link in the source language that cannot be preserved in the target language linguistics can again analyse the translation problem in terms of motivated versus unmotivated signifiers but does not go beyond the analytical stage and does not suggest any answer to the problem. Dialect and foreign-language items also fall into the category of trouble-causing elements that can be neatly described in linguistic terms so that the translator gets a clear picture of his possible problem, without receiving further guide-lines. This absence of answers that have a general validity for the translator can be explained by the fact that different texts require different translation approaches. What has to be maintained in one type of texts can be omitted without any noticeable harm in another type of texts. Moreover, different translators may disagree about the relative importance of the various levels on which a text operates: phonic, denotative, associative, connotative, etc. - so that ultimately personal preferences play an important role. Summarizing, one can say that linguistics offers deeper insights into


Linguistics, literary analysis, and literary translation

various translation problems and can in some cases even provide solutions, but that there is no general linguistic theory of translation.



While literary texts of bygone periods provide linguistics with valuable information about earlier stages of various languages and of languages that, but for this documentation, would have disappeared completely, literary analysis does not make important positive contributions to linguistics. However, the negative contributions - if such a strange term may be used - are very valuable. Literary analysis, by asking linguistics a series of unanswerable questions about the texts it analyses, has helped to define the area of linguistics proper, and to show the importance in literature of elements that fall outside that area. Intertextuality, association, and connotation cannot be expressed in linguistic formulas. These elements, as well as aesthetic features, reduce the importance of what is considered the core of linguistic preoccupations: the denotative message. Linguistics can introduce the term hypersemanticization, but when literary analysis needs criteria to determine whether a text is hypersemanticized, linguistics is unable to provide them unequivocally. 4.


Literary analysis presents the translator with an interpretation of the source text based on the contributions that each level of observation makes to the overall message. As long as the interpretation is not used as a means of demonstrating one's ingenuity or erudition, the translator will try to keep as many of the elements as possible that are singled out as important in the target text. The knowledge that form features play a crucial role is not enough, however, and may even exacerbate the predicament in which the translator finds himself, if the target language does not offer the possibility of matching them. When the source text can be interpreted in two or more equally acceptable ways, each requiring a formally different translation, the translator may ask himself or a literary analyst which interpretation corresponds to the author's intentions. Unless he has supplementary data that point to one interpretation - the author may have written somewhere about his intentions, or spoken about them in an interview- the literary analyst can only say what he thinks is the most likely intention. If unintentional,


Answerable and unanswerable questions

subconscious factors are taken into account, literary analysis may in some instances complicate even further the task of the translator who wants to follow 'the servile path' (see chapter 16, pages 105-6).



Although literary translations are rarely used systematically in linguistic studies (for an exception see chapter 15, page 99), they demonstrate that different languages use different means to say approximately the same thing. Their mere existence tends to put into question the validity of the Sapir-Whorf hypothesis. Yet the inevitable imperfections of translations focus the attention on the limits of interlinguistic communication. The comparison between a source text where sound-expressivity is very important and the corresponding target text that only reflects the denotative meaning of the source text may lead to a reassessment of the arbitrariness of the link between signifier and signified, because it shows, on the one hand, that sound motivation can be very important, and on the other, that the link between an expression and its content is arbitrary: in the one language there is sound-expressivity; in the other there is not (see chapter 7, passim, and chapter 11, pages 69-71).



Translations of the same text by different translators can be very dissimilar, thus drawing attention to the fact that many literary texts allow for more than one interpretation. Translations bring into focus the unique position of the source text. Whereas good translations rarely last (Baudelaire's translation of Edgar Allan Poe being one of the few exceptions to that rule) and are replaced by newer translations that are more accessible for the modern reader, 1 the original work remains untouched and only receives explanatory footnotes if necessary. The patina of age is not transferable. Another quality that is not easily transferable can be found in texts written in an idiosyncratic style and/or with an idiosyncratic use of words. What is accepted as a positive quality in the original may be rejected as incorrect language use in the translation. In such a situation the literary analyst should ask himself whether the positive evaluation of the specific traits of the original is an honest aesthetic judgment or an uncritical admiration based on the reputation of the author. Because translators have to make conscious what may remain sub-

124 Linguistics, literary analysis, and literary translation

conscious in ordinary reading it is sometimes claimed that translations can be an enrichment of the original text. This is perhaps true in some instances, for the bilingual reader who is fluent in both source and target languages. It is certainly not true for the monolingual reader who can only read the target text: if there are any gains, they are very rare, while losses are very frequent and sometimes considerable. So, both the translation activity itself and the comparison between an original literary text and its translations can make one aware of important features of the source text that otherwise might have remained unnoticed. The conclusions to be drawn from this short survey of possible points of interaction are threefold: a Even if not all questions can be answered it is worth while to look for solutions and for new approaches outside one's own field. b A dialogue with the other disciplines makes one realize that many of the problems are shared. c By realizing that some questions the other disciplines ask fall outside the area of one's competence, one becomes aware of the possibilities and the limitations of one's own discipline. For these reasons I hope that contacts between linguists, literary analysts, and translators will continue to grow, with not only a better understanding of the disciplines that others work in, but also a deeper insight into one's own field of activity, as the ultimate result.



1 The use of 'he' and 'his' does not reflect a bias but rather a desire not to

burden the text with cumbersome repetitions of formulae such as 'he or she' or 'his and her.' CHAPTER 1 1


Linguistics as a general term does not specify which school of thought is followed . It is a well-known fact that on the whole linguists are not very tolerant of theories that differ from their own approach . It should be stressed, however, that the different schools (structuralist, functionalist, transformational-generative, neo-bloomfieldian, glossematic, etc.) have many ideas in common, and that in many areas they are complementary, rather than contradictory of one another. For a tolerant introduction, see Mahmoudian (1982). CHAPTER 2

System, norm, usage

1 Saussure (1967; 1972); Godel (1957); Mounin (1968); Engler (1968). 2 Chomsky (1957; 1965). 3 Palmer (1981 : 65-6): 'In practice linguists often attempt to rule out context as far as possible - to deal with "marginally decontextualised sentences" ' (Lyons, 1977: 590). These are the objects of study of most grammars. On methodological grounds this approach is essential because of the enormous variation in language, but the dangers are obvious, and it is difficult to accept, without severe reservations, Chomsky's (1965: 3) view that

126 Notes to pages 7-8



6 7


'Linguistic theory is concerned primarily with an ideal speaker-hearer, in a completely homogeneous speech community, who knows its language perfectly.' See also the attempt to create a special T-G grammar for literary texts (Bierwisch, 1971). Neogrammarians Uunggrammatiker): a group of linguists at the University of Leipzig, who in the last third of the nineteenth century established a theory according to which the evolution of language is governed by strict laws. These laws are most evident in the domain of sound evolution. Jede Sprache besitzt zur Zeit ihres Entstehens als Einzelsprache eine gewisse Anzahl von Casusformen, heriibergebracht aus der Periode ihres Zusammenlebens mit einer oder mehreren verwandten Sprachen. Diese Formen nehmen ausnahmslos die Gestalt an, wekhe die Wirkung der Lautgesetze ... hervorbringen muss. (Leskien, 1963: 2) [translation] At the time of its birth as an independent language, each language has a certain number of case-forms, carried over from the period when it lived together with one or more closely related languages. These forms take, without any exception, the shape that the action of the phonetic laws ... must produce. The most prominent Neogrammarians were August Leskien (184-1916), Hermann Paul (1846--1921), Karl Brugmann (1849-1919), Hermann Osthoff (1847-1909), and Berthold Delbruck (1842-1922). In France the most famous Neogrammarian was Gaston Paris (1839-1903). Andre Martinet is the founder of the functionalist-structuralist school of linguistics with its centre in Paris. An important general theoretical work is Martinet (198o). The principles of diachronic phonology and phonetics are discussed in Martinet (1955). For an exhaustive bibliography of the works of Andre Martinet to 1979 see Martinet (1979: vii-xxvii). A fairly early work in structural phonology is Haudricourt and Juilland (1949). William Labov has published many important works on the links between social stratification and language varieties. See, for example, Labov (1972b). For the recent history of some dialect markers on the island of Martha's Vineyard see Labov (1972a). Labov's findings put into question 1 / the validity of evolution description exclusively explained by internal systemic pressures; 2 / the usefulness of the idealized homogeneous competence (Chomsky) or langue level for describing linguistic phenomena. For French see Martinet and Walter (1973) and Walter (1977). The phonetic changes are systematically described in Bourciez (1958) and Fouche (1952; 1958; 1961). For two special approaches see Schogt (1960), where the blending of two dialect groups (East and West) in the central

127 Notes to pages 8-16

9 10


12 13



dialect of Paris is discussed; and Schogt (1965), where structural methods are applied similar to those used by Andre Haudricourt and Alphonse Juilland (1949) . Ettinger (1976) follows the model presented by Coseriu (1952). Grevisse (1975: 636 #647) writes about the imparfait du subjonctif: 'L'imparfait du subjonctif a disparu, ou peu s'en faut, de la langue parlee. Dans la langue ecrite, I'emploi en est restreint. Des formes comme naquissions, contentassiez, mena,assions, debarassasse, etc. etonneraient I'oreille, quoiqu'elles n'aient en soi rien d'offensant; c'est a cause de leur rarete que Jes flexions de l'imparfait du subjonctif paraissent choquantes ou ridicules - La defaillance de l'imparfait du subjonctif a entraine celle du plus-queparfait.' It is impossible to mention all the Romanian linguists involved in the creation of a unified standard norm. An important book, mentioned by Ettinger, is Iordan et al. (1967) . Walter (1981). Gustave Guillaume (1965) discusses all possible surcompose forms, but does not mention the problem of the surcompose forms of verbs requiring etre instead of avoir as the auxiliary of compound tenses, nor does he indicate relative frequency of those forms that in his terminology belong to the aspect-biextensif (i.e., all surcompose forms). A possible reason for the confusion of native speakers when asked what the surcompose form of a pronominal verb would be like lies in the fact that est - a ete results in the direct encounter of the reflexive pronoun and a form of the verb avoir: s'a ete, a combination they are conditioned to reject immediately. See Jolivet et al. (198o: 133-211).


Appendix Probi. For data about this glossary containing 227 words of common usage in their proscribed form and in the correct Latin form, see Tagliavini (1964: 169--70). The name Appendix Probi was given to the list because it was found as an addition to a manuscript of the grammarian Valerius Probus. For a sample of words found in the Appendix see below (page 128).


3 Literature, norm, usage, and evolution

The story of [we/ wa / c], all going back to stressed closed e in open syllables (in free position) and the survival of two of these three in Parisian French, is the subject of Schogt (1¢o). For the wavering of prescriptive grammarians

128 Notes to pages 16-19





between [we] and [wa] see chapter 5, page gJ. See also Thurot (1881/1: 356, 35()-63) and Beaulieux (1927: 296). Grevisse (1975: 1037 sous 5, 939) states: 'La langue populaire emploie le conditionnel apres si marquant le potentiel et l'irreel: Si tu voudrais, on travaillerait ensemble (Fr. Carco, L'Equipe p. 48 cit Sandfeld, t. 11 p. 343). Pour peu qu'on ait souci de bien parler ou de bien ecrire, on se gardera de cet emploi.' Frei (1929: 200): 'L'imparfait apres si est un procede de confonnisme inutile a )'intelligence de la phrase, et qui entrave l'interchangeabilite entre l'independante et la subordonnee. En meme temps, il empeche )'expression du mode quand ce dernier demande a etre exprime; dans ce cas, le Iangage populaire se sert du conditionnel d'eventualite, absolument comme dans la phrase independante: Je pourrais peut-etre le voir > Si je pourrais peut-etre le voir (Z Je pourrais peut-etre le voir / Si je pouvais ...)' > leads to Z instead of Some examples from the Appendix Probi are: pauper non paupera (change of declension) senatus non sinatus (confusion of pretonic closed front vowels) olim non oli (disappearance of final -m) stabulum non stablum (syncope) equs non ecus (correct form should have been equus) In Dutch the diminutive ending -je is replaced by ie [i) in popular speech: thus kopje ('little cup') > koppie. When an effort was made to avoid this substandard mark, kopje koffie, which normally became koppie koffie, was replaced by kopje kofje. This hypercorrect form was on the way out when I was a child, and is no longer used. Wartburg (1946: 89ff) describes the rise of the dialect of the Ile-de-France as the national language. The following two quotations clearly illustrate the prestige the language of the Ile-de-France enjoyed already in the twelfth century. Mis langages est boens, car en France fui nez. (Guernes de Pont-Sainte-Maxence (1936: vers 6165)) La Roine n'a pas fait ke cortoise, Ki me reprist, ele et ses fieus Ii Rois Encoir ne soit ma parole franchoise, Si la puet on bien entendre en franchois; Ne chil ne sunt bien apris ne courtois, S'il m' ont repris se j' ai dit mos d' Artois,

129 Notes to pages 19-24 Car je ne fui pas norris a Pontoise. [written around 1180) (Conon de Bethune (19211) 7 The pronunciation of the stressed vowel of water is not uniform either, and fluctuates within the British as well as the North American speech community. 8 The Polish Romance scholar Halina Lewicka (1959) pointed out that in the Picard literature of the Middle Ages the frequency of occurrence of regional features in a text depended more on the type of the text (literary genre) than on its author. 9 Sagan (1957: 122-3): 'Maligrasse se passa la main dans Jes cheveux. Comrne ii en avait peu, cela fit un geste pauvre. Jolyet cherchait vaguement un tenebreux moyen de livrer Beatrice a ce cher vieux Maligrasse, apres qu'il l'eut possedee lui-meme, bien entendu.' As eut replaces aurait of free indirect discourse, it is difficult to decide whether the replacement is an unusual extension of 'eut for aurait after conditional si' or a subjunctive triggered by the analogy with avant que followed by that mood. For a short description of the substitution of the imperfect subjunctive for the conditional consult Chevalier et al. (1964: 364) in which the different instances are discussed under the general heading: '11 [i.e., the imperfect subjunctive] equivaut a un conditionnel passe dans un systeme hypothetique avec si.' 10 Barbusse (n.d.: 86): 'et bien que je ne le vis pas, je sus le moment oil sa chair etait entree dans celle de la femme.' CHAPTER

4 Cost and yield

A clear example of reducing the speaker's effort can be found in the evolution of feminine singular past participles from Latin to Modern French: don[ata > ada > eda >ea> e]. Compare also the North American voicing: Peter> Peder. 2 Native speakers are able to fill in the missing elements that they did not catch when listening, or that are illegible in a written text, as long as the subject-matter is familiar. They rely on their expectations and probabilities. 3 In English the generic term to go to is less explicit than to fly to, to ride to, to drive to, to walk to, etc., that provide information about the special character of the going and are each much more limited in their applications. 4 Harlequin Romances is a highly successful series that started in Toronto, but has acquired an international distribution. The recipe of love and rom1


Notes to page 24

ance is quite simple and the instruction sheet for those who want to write for the series is very precise about the level of difficulty, etc. Harlequin Enterprises provides guide-lines to prospective authors for their series. The following passage taken from guide-lines that were sent out a few years ago gives a good idea of what the series aims at: Harlequins are well-plotted, strong romances with a happy ending. They are told from the heroine's point of view and in the third person. There may be elements of mystery or adventure, but these must be subordinate to the romance. The books are contemporary and settings can be anywhere in the world as long as they are authentic. Remember, our readers enjoy 'visiting' new and unknown places (to them) and they also enjoy learning about local food, dress and customs. Characters should be interesting, well-developed and convincing in their roles. Heroines are of all types. They have all the interests and occupations of today's women and a desire for a satisfying man/woman relationship based on love and marriage. Though love scenes may be sensuous, sex must not be explicit. Authors are warned: 1) Your manuscript will be rejected if it does not come up to our standards for a number of reasons - if the characters are not sympathetic or not 'Harlequin people' - if the plot is weak, or alternatively if the book is too heavy with plot so that the romance is 'lost' - if the locale is uninteresting - if the book lacks warmth - and, finally, if there is just not a high enough proficiency of writing. 2) Your manuscript will also be rejected if it is not a romance. We do not publish non-fiction, war novels, family chronicles or the like. 5 This often quoted line by Paul Eluard (1968: 232) is the beginning of the following poem 'L'amour la poesie v11'): La terre est bleue comme une orange Jamais une erreur les mots ne mentent pas Us ne vous donnent plus a chanter Au tour des baisers de s' entendre Les fous et les amours Elle sa bouche d' alliance Tous les secrets tous les sourires Et quels vetements d'indulgence A la croire toute nue. Les guepes fleurissent vert L'aube se passe autour du cou

131 Notes to pages 24-32 Un collier de fenetres Des ailes couvrent les feuilles Tu as toutes les joies solaires Tout le soleil sur la terre Surles chemins de ta beaute. 6 The whole question of redundancy, probability, and predictability is rather complicated. Whereas probability can be calculated by using objective inventory data, as can redundancy as well, predictability is a dynamic characteristic that depends on the growing familiarity the reader obtains in the process of reading a given author. From the reader's point of view there is no absolute measure of predictability, each reader reacting differently to a text when starting, and there is no absolute generally valid growth rate of predictability, one reader being quicker than another in discovering which collocations are favoured by a given author in a given text. The homogeneity and coherence of a text enable the reader to form expectations. Greimas (1966: 69--101) coined the term isotopie to describe these features of inner consistency. 7 This statement does not take into account the construction nous, on danse that is mentioned at the end of this chapter in connection with nous

sommes - nous on est.

8 For entropy we are following the definition given in Dubois (1974: 193): 'L'entropie represente le degre d'incertitude ou l'on est de !'apparition de chaque signal. Ainsi, pour un nombre donne de reponses possibles, l'entropie est maximale lorsque toutes les reponses ont la meme frequence.' 9 Jakobson (1963: 89).


English and Dutch do not have the same kinship terminology: English cousin Dutch neef nephew English cousin Dutch neef nicht French


5 Formal identity and formal opposition: the linguistic sign

ville village

English city town village French does not have a generic equivalent of the English nut.

132 Notes to pages 32-5 3 There are a number of definitions of the seme. Apart from a few linguists like Eric Buyssens and Luis Prieto, who use the term for indicating a semantic unit consisting of a signified and a signifier, most linguists consider the seme to be the smallest meaningful unit, a feature of content. There is a great deal of controversy surrounding the concept, and therefore one should always try to find out in what sense the term seme is used by a given author. For a more detailed discussion of the various interpretations of seme see the article 'Seme' by Schogt in Sebeok (1986). 4 Andre Martinet and Henriette Walter consistently incorporate dynamics in their synchronic descriptions of contemporary French. See, for example, Martinet (1968) and Walter (1976). 5 There are exceptions to the general rule of complementary distribution of plural forms, as a few examples will show. In Dutch aardappel (potato') has two plural forms, aardappels and aardappelen, that are semantically interchangeable. A somewhat more common exception is found when one singular has two plural forms linked with it that are semantically.and/or stylistically distinct. Russian syn ('son') has a semantically unmarked plural, synov'ja, and a plural, syny, that is used in special solemn contexts such as syny otecestva ('the sons of the fatherland'). Italian and Romanian both offer a considerable number of double plurals with distinct meanings. ii braccio 'the arm' le braccia 'the arms' Italian le bracci 'the arms' (figurative) i muri ii muro 'the wall' 'the walls' (of a house) 'the walls' (of a city) le mura Romanian porn 'tree' 'trees' pomi pomuri 'Christmas trees' 'time' timp timpuri 'times' 'beats in music' timpi 6 See Jakobson (1968: 350--77). 7 See Benveniste (1967: 16~75).


6 Field models, universals, and language-bound world-view

1 The history of field theories can be found in Gordon (1972). Germain (1981) incorporates a comprehensive description of semantic field theories and their implications in a general functional semantic theory. When semantic field theories are discussed the first scholar who comes to mind is Jost

133 Notes to pages 35-9






Trier (1894-1970) . His work on the subject spans several decades. See Trier (1973), where the editors give an introductory survey of the history of field theories. Saussure (1972: 160-1). Strictly speaking there is a slight difference between the evolution of craindre, redouter, and avoir peur on the one hand, and singular, dual, plural on the other. In the hypothetical case of the disappearance of one of the verbs Saussure assumes that the remaining two would share the space that would have become vacant. In the real case of the disappearance of the dual, the singular remained unchanged whereas the plural incorporated all former duals. A comprehensive approach based on the division Man, the Universe, and Man and the Universe is used in Wartburg and Hallig (1963). The criteria for structuring the vocabulary are non-linguistic, often arbitrary, and frequently ethnocentric. The best-known diachronic study written by Trier (1931) deals with the terms Kunst, List, Wislteit, and Wissen in Old High German. In 1897 Brea! (1921: 101) already wrote: 'L'allemand List, "ruse," a commence par etre un synonyme de Kunst "savoir, habilete." On disait Gottes List "la sagesse de Dieu." ' Trier does not mention Brea! in his book; thus there may or may not have been a direct influence from Brea! to Trier. For kinship see, for example, Leech (1974). For colour terms see also Berlin and Kay (1969). A discussion of kitchen terms can be found in Lehrer (1974) . Pottier (1963) examines the semantic field of seats, while Mounin (1965) deals with domestic animals. See also Pottier (1974: 99). An exception to the rule that semantic fields deal with notions the structuring of which is based on verifiable extra-linguistic data is Duchacek (196o).

7 For the important place kinship and colour terminology have in field studies see Gordon (1980). 8 Matore (1950). 9 Guiraud (1967). 10 See, for example, Bondarko (1974). 11 Absolute and relative universals are discussed from a variety of points of view in a volume of articles edited by Greenberg (1963). For the translation aspect of universals, see Nida (1966). 12 The principle of double articulation is explained in many of Andre Martinet's works. See, for example, Martinet (1957; 1965: chapter 1). 13 Notice that in Canadian English the raising of [ay] to [Ay) before voiceless

134 Notes to pages 39-45 consonants ('eyes' [ayz] but 'ice' [Ays)) creates a difference in pronunciation that is not observed in most other varieties of English. 14 Sapir (1921); Whorf (1956); see also Humboldt (1836). CHAPTER

7 Arbitrariness, convention, and motivation

1 Ullmann (1¢o: 104-15) gives, in the section dealing with sound motivation,

onomatopoeia, and expressive words, a list of names for the cuckoo in different languages, all in the category of onomatopoeia. 2 The technical term for distribution rules is 'phonotactic rules.' Examples of differences in phonotactic rules for individual languages abound. To give only two: 1 / English, French, Russian, German, and Dutch all have the opposition voiced/voiceless. However, whereas voiced consonants occur in English and French in final position, Russian, German, and Dutch allow only voiceless consonants there, thus offering an example of neutralization. 2 I In English the consonant I does not occur after the initial cluster st; in Russian st+l is not excluded by any phonotactic rule. 3 The phenomenon of sound-expressivity has been the subject of controversy and speculation for many centuries. Plato and his contemporaries were divided on the question whether words had a natural, inherent meaning (