The Process of Translating

INTRODUCTION
My description of translating procedure is operational. It begins with choosing a method of approach. Secondly, when we are translating, we translate with four levels more or less consciously in mind: (1) the SL text level, the level of language, where we begin and which we continually (but not continuously) go back to; (2) the referential level, the level of objects and events, real or imaginary, which we progressively have to visualise and build up, and which is an essential part, first of the comprehension, then of the reproduction process; (3) the cohesive level, which is more general, and grammatical, which traces the train of thought, the feeling tone (positive or negative) and the various presuppositions of the SL text. This level encompasses both comprehension and reproduction: it presents an overall picture, to which we may have to adjust the language level; (4) the level of naturalness, of common language appropriate to the writer or the speaker in a certain situation. Again, this is a generalised level, which constitutes a band within which the translator works, unless he is translating an authoritative text, in which case he sees the level of naturalness as a point of reference to determine the deviation- if any - between the author's level he is pursuing and the natural level. This level of naturalness is concerned only with reproduction. Finally, there is the revision procedure, which may be concentrated or staggered according to the situation. This procedure constitutes at least half of the complete process.
THE RELATION OF TRANSLATING TO TRANSLATION THEORY
The purpose of this theory of translating is to be of service to the translator. It is designed to be a continuous link between translation theory and practice; it derives from a translation theory framework which proposes that when the main purpose of the text is to convey information and convince the reader, a method of translation must be 'natural'; if, on the other hand, the text is an expression of the peculiar innovative (or cliched) and authoritative style of an author (whether it be a lyric, a prime minister's speech or a legal document), the translator's own version has to reflect any deviation from a `natural' style. The nature of naturalness is discussed in detail in my exposition of the theory of translating below; `naturalness' is both grammatical and lexical, and is a touchstone at every level of a text, from paragraph to word, from title to punctuation.
The level of naturalness binds translation theory to translating theory, and translating theory to practice. The remainder of my translating theory is in essence psychological - the relationship between language and `reality' (though all we know of 'reality' is mental images and mental verbalising or thinking) - but it has practical applications.
If one accepts this theory of translating, there is no gap between translation theory and practice. The theory of translating is based, via the level of naturalness, on a theory of translation. Therefore one arrives at the scheme shown in Figure 2.

Three Language Function
Expressive Informative Vocative
Translation Theory
Semantic Communicative


Frame of Reference
Problem Contextual Factor Translation Procedures







Theory Of Traslating
Textual
Refential Levels
Cohesive
Natural






Translation Practice


THE APPROACH
A translation is something that has to be discussed. In too many schools and universities, it is still being imposed as an exercise in felicitous English style, where t lie warts of the original are ignored. The teacher more or less imposes a fair copy which is a `model' of his own English rather than proposing a version for discussion and criticism by students, some of whom will be brighter than he is.
Translation is for discussion. Both in its referential and its pragmatic aspect, it has an invariant factor, but this factor cannot be precisely defined since it depends on the requirements and constraints exercised by one original on one translation. All one can do is to produce an argument with translation examples to support it. Nothing is purely objective or subjective. There are no cast-iron rules. Everything is more or less. There is an assumption of `normally' or `usually' or `commonly' behind each well-established principle; as I have stated earlier, quali­fications such as `always', `never', `must' do not exist - there are no absolutes.
Given these caveats, I am nevertheless going to take you through my tenta­tive translating process.
There are two approaches to translating (and many compromises between them): (I) you start translating sentence by sentence, for say the first paragraph or chapter, to get the feel and the feeling tone of the text, and then you deliberately sit back, review the position, and read the rest of the SL text; (2) you read the whole text two or three times, and find the intention, register, tone, mark the difficult words and passages and start translating only when you have taken your bearings.
Which of the two methods you choose may depend on your temperament, or on whether you trust your intuition (for the first method) or your powers of analysis (for the second). Alternatively, you may think the first method more suitable for a literary and the second for a technical or an institutional text. The danger of the first method is that it may leave you with too much revision to do on the early part, and is therefore time-wasting. The second method (usually preferable) can be mechanical; a translational text analysis is useful as a point of reference, but it should not inhibit the free play of your intuition. Alternatively, you may prefer the first approach for a relatively easy text, the second for a harder one.
From the point of view of the translator, any scientific investigation, both statistical and diagrammatic (some linguists and translation theorists make a fetish of diagrams, schemas and models), of what goes on in the brain (mind? nerves? cells?) during the process of translating is remote and at present speculative. The contribution of psycholinguistics to translation is limited: the positive, neutral or negative pragmatic effect of a word (e.g. affecter, `affect', `brutal', befremden, drame, comedie, favoriser, denouement, extr¢ordinaire, `grandiose', grandioznyi, `potentate', pontif, `pretentious', 'arbitrary/arbitration', proposer, exploit, hauteur, `vaunt') e.g. Osgood's work on semantic differentials is helpful, since the difference between `positive' and `negative' (i.e. between the writer's approval and his disapproval) is always critical to the interpretation of a text. The heart of translation theory is translation problems (admitting that what is a problem to one translator may not be to another); translation theory broadly consists of, and can be defined as, a large number of generalisations of translation problems. A theoretical discussion of the philosoph; and the psychology of translation is remote from the translator's problems. Whether you produce a statistical survey through question­naires of what a hundred translators think they think when they translate, or whether you follow what one translator goes through, mental stage by mental stage, I do not see what use it is going to be to anyone else, except perhaps as a corrective of freak methods - or ideas such as relying entirely on bilingual dictionaries, ml,slituting encyclopaedia descriptions for dictionary definitions, using the best­vumding synonyms for literary translation, transferring all Graeco-Latin words, , oiil inuous paraphrasing, etc. But there is never any point in scientifically proving i he obvious.

THE TEXTUAL LEVEL
Working on the text level, you intuitively and automatically make certain `con­vcrsions'; you transpose the SL grammar (clauses and groups) into their `ready' TL ~-(lliivalents and you translate the lexical units into the sense that appears immedi­airly appropriate in the context of the sentence.
Your base level when you translate is the text. This is the level of the literal t r:nlslation of the source language into the target language, the level of the trans­I;i i ionese you have to eliminate, but it also acts as a corrective of paraphrase and the
parcr-down of synonyms. So a part of your mind may be on the text level whilst another is elsewhere. Translation is pre-eminently the occupation in which you have to be thinking of several things at the same time.

THE REFERENTIAL LEVEL
You should not read a sentence without seeing it on the referential level. Whether a text is technical or literary or institutional, you have to make up your mind, summarily and continuously, what it is about, what it is in aid of, what the writer's peculiar slant on it is: say, L'albumine et ses interactions medicamenteuses (It.: l.'albumine e le sue interazioni medicamentose) - it may be the action of drugs on blood, the need to detect toxic effects, the benefits of blood transfusion. Say, La pression quantitative - the large number of pupils in schools, the demand for better-quality education, the need for suitable education for all. Say, Recherches sur un facteur diuretique d'origine lymphatique - the attempt to find a substance in the body fluid that promotes urine production, the disorders that inhibit the formation of the substance, the attempts to isolate the substance. Always, you have to be able to summarise in crude lay terms, to simplify at the risk of over-simplification, to pierce the jargon, to penetrate the fog of words. You get an abstraction like Ce phenomene s'avere; ce phenomene, exact pour cellules et fibres - referring to a tumour becoming so large that it compresses the parenchyma next to it. Usually, a more specific reference is desirable in the translation: the tumour's swelling, deteriora­tion, etc. Thus your translation is some hint of a compromise between the text and the facts.
For each sentence, when it is not clear, when there is an ambiguity, when the writing is abstract or figurative, you have to ask yourself: What is actually happen­ing here? and why? For what reason, on what grounds, for what purpose? Can you see it in your mind? Can you visualise it? If you cannot, you have to `supplement' the linguistic level, the text level with the referential level, the factual level with the necessary additional information (no more) from this level of reality, the facts of the matter. In real life, what is the setting or scene, who are the actors or agents, what is the purpose? This may or may not take you away temporarily from the words in the text. And certainly it is all too easy to immerse yourself in language and to detach yourself from the reality, real or imaginary, that is being described. Far more acutely than writers wrestling with only one language, you become aware of the awful gap between words and objects, sentences and actions (or processes), grammar and moods (or attitudes). You have to gain perspective (distacco, recul), to stand back from the language and have an image of the reality behind the text, a reality for which you, and not the author (unless it is an expressive or an authorita­tive text), are responsible and liable.
The referential goes hand in hand with the textual level. All languages have polysemous words and structures which can be finally solved only on the referential level, beginning with a few multi-purpose, overloaded prepositions and conjunc­
tions, through dangling participles (`reading the paper, the dog barked loudly') to general words. The referential level, where you mentally sort out the text, is built up out of, based on, the clarification of all linguistic difficulties and, where appropriate, supplementary information from the 'encyclopaedia'- my symbol for any work of reference or textbook. (Thus in pour le passage de Flare, you find that Flore/Flora was an Italic goddess of flowers and gardens. As it is in Claudel you translate: `for the goddess Flora to pass' and leave the rest to the reader.) You build up the referential picture in your mind when you transform the SL into the TL text; and, being a professional, you are responsible for the truth of this picture.
Does this mean, as Seleskovitch claims, that `the (SL) words disappear' or that you `deverbalize the concepts' (Delisle)? Not at all, you are working continu­ously on two levels, the real and the linguistic, life and language, reference and sense, but you write, you `compose', on the linguistic level, where your job is to achieve the greatest possible correspondence, referentially and pragmatically, with the words and sentences of the SL text. However tempting it is to remain on that simpler, usually simplified layman's level of reality (the message and its function) you have to force yourself back, in as far as the readership can stand it, into the particularities of the source language meaning.

THE COHESIVE LEVEL
Beyond the second factual level of translating, there is a third, generalised, level linking the first and the second level, which you have to bear in mind. This is the `cohesive' level; it follows both the structure and the moods of the text: the structure through the connective words (conjunctions, enumerations, reiterations, definite article, general words, referential synonyms, punctuation marks) linking the sentences, usually proceeding from known information (theme) to new information(rheme); proposition, opposition, continuation, reiteration, opposition, conslusion for instance - or thesis, antithesis, synthesis. Thus the structure I,dlwws the train of thought; determines, say, the `direction' of d'ailleurs (`besides', I i n t tier', `anyway') in a text; ensures that a colon has a sequel, that ulterieur has a Inmr reference; that there is a sequence of time, space and logic in the text.
The he second factor in the cohesive level is mood. Again, this can be shown as a -Imluctical factor moving between positive and negative, emotive and neutral. It wvms tracing the thread of a text through its value-laden and value-free passages rvlrrclr may be expressed by objects or nouns (Margaret Masterman (1982) has ,.I town how a text alternates between `help' and `disaster'), as well as adjectives or ,Irr;rlitics. You have to spot the difference between positive and neutral in, say, Olrlrr"c'ciate' and `evaluate'; `awesome' and `amazing'; `tidy' and `ordered'; sauber n m I rein; `passed away' (indicating the value of the person) and `died'. Similarly you have to spot differences between negative and neutral in say `potentate' and `ruler'. I Ircsc differences are often delicate, particularly near the centre, where most languages have words like `fair', `moderate', maft, passable, assez bon whose value ,vnnon always be determined in the context.
My third level, this attempt to follow the thought through the connectives mid the feeling tone, and the emotion through value-laden or value-free expres­-,iv ms, is, admittedly, only tentative, but it may determine the difference between a humdrum or misleading translation and a good one. This cohesive level is a rcaulator, it secures coherence, it adjusts emphasis. At this level, you reconsider t I rc lengths of paragraphs and sentences, the formulation of the title; the tone of the conclusion (e.g. the appropriateness of d tout prendre, en definitive (often tricky), en /nr de compte, enfin(!), d la fin, en somme, en tout etat de cause to summarise an :uyument at the beginning of a final sentence). This is where the findings of discourse analysis are pertinent.
THE LEVEL OF NATURALNESS
With all that, for all texts (except the ones you know are `odd' or badly written but authoritative, innovatory or `special', e.g., where a writer has a peculiar way of writing which has to be reproduced - so for philosophy, Heidegger, Sartre, 1 Iusserl; so for fiction any surrealist, baroque, and certain Romantic writers) - for t he vast majority of texts, you have to ensure: (a) that your translation makes sense; (b) that it reads naturally, that it is written in ordinary language, the common grammar, idioms and words that meet that kind of situation. Normally, you can only do this by temporarily disengaging yourself from the SL text, by reading your own translation as though no original existed. You get a piece like: Une doctrine nee dans une fraction du clerge de fAmerique latine qui foisonne sous diverses plumes et dons diverses chapelles'et qui connaft deja un debut d'application autoritaire sous la tutelle de 1'Ftat. (L'Express, July 1985.) The passage has various misleading cognates, and you can reduce it to sense by gradually eliminating all the primary senses (fraction, nee, plumes, chapelles, connait) to: `A doctrine originating amongst a fraction of the clergy of Latin America which proliferates among various writers and in various coteries and which already experiences the beginnings of an authoritarian applica­tion under the tutelage of the State'.
Now you still have to make that passage sound natural, which will usually depend on the degree of formality (see p. 14) you have decided on for the whole text. But you might consider: `A doctrine originating in a group of Latin American
clergy and proliferating among various writers and coteries, which is now just beginning to be put into practice in an authoritarian fashion under the auspices of the State' (note that deja often translates as `now').
A word on `naturalness'. A translation of serious innovative writing (maybe Rabelais, Shakespeare, Thomas Mann, maybe Hegel, Kant, maybe any authority) may not sound natural, may not be natural to you, though if it is good it is likely to become more so with repeated readings:
The funnel unravels an enormous mass of black smoke like a plait of horsehair being unwound.
La cheminee devide une enorme fumee noire, pareille d une tresse de crin qu'on detord.
(G. F. Ramuz, `Le retour du mort', from Nouvelles.)
A still new patient, a thin and quiet person, who had found a place with his equally thin and quiet fiancee at the good Russian Table, proved, just when the meal was in full swing, to be epileptic, as he suffered an extreme attack of that type, with a cry whose demonic and inhuman character has often been described, fell heavily on to the floor and struck around with his arms and legs next to his chair with the most ghastly contortions.
Ein noch neuer Patient, ein magerer und stillerMensch, der mit seiner ebenfalls mageren und stillen Braut am Guten Russentisch Platz gefunden hatte, erwies sich, da eben das Essen in vollem Gang war, als epileptisch indent er einen krassen Anfall dieser Art erlitt, mit jenem Schrei dessen damonischer und aussermenschlicher Charackter oft geschildert worden ist, zu Boden sturzte und neben seinem Stuhl unter den scheusslichsten Verrenkungen mitArmen und Beinen um sich schlug.
(Thomas Mann, Der Zauberberg. )
You may find both these sentences unnatural. Yet, in spite of numerous lexical inadequacies (we have no word for mager nor any as vivid as schildern, and few parallel sound effects) this is what Ramuz and Thomas Mann wrote, and we cannot change that.
When you are faced with an innovatory expressive text, you have to try to gauge the degree of its deviation from naturalness, from ordinary language and reflect this degree in your translation. Thus in translating any type of text you have to sense `naturalness', usually for the purpose of reproducing, sometimes for the purpose of deviating from naturalness. In a serious expressive text, in the sentence: il promenait son regard bleu sur la petite pelouse, `son regard bleu' has to be translated as `his blue gaze', which is a deviation from the normal or natural les yeux bleus, `his blue eyes'. Again Si le regard du pasteur se promenait sur la pelouse, etait-ce pour jouir de la parfaite plenitude verte ou pour y rrouver des idees (Drieu la Rochelle) is i r;utslated as something like: `If the pastor's gaze ran over the lawn, was it to enjoy iv; perfect green fullness, or to find ideas', rather than `Whenever the pastor cast a glance over the lawn it was either to enjoy its perfect green richness, or to find ideas in it’.
Again, son visage etait mauve, `his face was mauve', sein Gesicht war mauve i nutlvenfarben) are virtually precise translation equivalents. `Mauve' is one of the Icw secondary colours without connotations (though in France it is the second ,ulnur of mourning, `his face was deathly mauve' would be merely comic), and w a mally, like `beige', associated with dress - compare a mauve woman, a violet mnnan (`shrinking violet'?), but a scarlet woman is different. In the `mauve' cx;unple, a retreat from the unnatural `mauve' to the natural `blue' would only be tost ified if the SL text was both `anonymous' and poorly written.
You have to bear in mind that the level of naturalness of natural usage is p,runmatical as well as lexical (i.e., the most frequent syntactic structures, idioms :uul words that are likely to be appropriately found in that kind of stylistic context), ;in(], through appropriate sentence connectives, may extend to the entire text.
In all `communicative translation', whether you are translating an informa­avc text, a notice or an advert, `naturalness' is essential. That is why you cannot t r;utslate properly if the TL is not your language of habitual usage. That is why you ,w often have to detach yourself mentally from the SL text; why, if there is time, vvu should come back to your version after an interval. You have to ask yourself (or on hers): Would you see this, would you ever see this, in The Times, The Economist (watch that Time-Life-Spiegel style), the British Medical,Journal, as a notice, on the hack of a board game, on an appliance, in a textbook, in a children's book? Is it usage, is it common usage in that kind of writing? How frequent is it? Do not ask vwurself: is it English? There is more English than the patriots and the purists and t lie chauvinists are aware of.
Naturalness is easily defined, not so easy to be concrete about. Natural usage n,mprises a variety of idioms or styles or registers determined primarily by the 'sctting' of the text, i.e. where it is typically published or found, secondarily by the author, topic and readership, all of whom are usually dependent on the setting. It may even appear to be quite `unnatural', e.g. take any article in Foreign Trade (Moscow): `To put it figuratively, foreign trade has become an important artery in the blood circulation of the Soviet Union's economic organism', or any other axample of Soviet bureaucratic jargon; on the whole this might occasionally be tactfully clarified but it should be translated `straight' as the natural language of participants in that setting.
Natural usage, then, must be distinguished from `ordinary language',, the plain non-technical idiom used by Oxford philosophers for (philosophical) uxplanation, and `basic' language, which is somewhere between formal and informal, is easily understood, and is constructed from a language's most fre­, It ently used syntactic structures and words - basic language is the nucleus of a language produced naturally. All three varieties-natural, ordinary and basic-are formed exclusively from modern language. However, unnatural translation is marked by interference, primarily from the SL text, possibly from a third language known to the translator including his own, if it is not the target language. `Natural' translation can be contrasted with `casual' language (Voegelin), where word order, syntactic structures, collocations and words are predictable. You have to pay special attention to:
Word order. In all languages, adverbs and adverbials are the most mobile components of a sentence, and their placing often indicates the degree of emphasis on what is the new information (theme) as well as naturalness. They are the most delicate indicator of naturalness:
He regularly sees me on Tuesdays. (Stress on `regularly'.)
He sees me regularly on Tuesdays. (No stress.)
On Tuesdays he sees me regularly. (Stress on `Tuesdays'.)
Common structures can be made unnatural by silly one-to-one translation from any language, e.g.:
Athanogore put his arm under that of (sous celui de) the young man: (`under the young man's').
After having given his meter a satisfied glance (apres avoir lance): (`after giving').
Both these translations are by English students.
The packaging having (etant muni de) a sufficiently clear label, the cider vinegar consumer could not confuse it with . . . : (`as the packaging had. . .').
Cognate words. Both in West and East, thousands of words are drawing nearer to each other in meaning. Many sound natural when you transfer them, and may still have the wrong meaning: `The book is actually in print' (Le livre est actuellement sous presse). Many more sound odd when you transfer them, and are wrong - avec, sans supplement, le tome VII, `with, without a supplement, Vol.7' (`without extra charge'). Thousands sound natural, have the same meaning, are right.
The appropriateness of gerunds, infinitives, verb-nouns (cf. `the establishment of', `establishing', `the establishing of', `to establish').
Lexically, perhaps the most common symptom of unnaturalness is slightly old-fashioned, now rather `refined', or `elevated' usage of words and idioms possibly originating in bilingual dictionaries, e.g.
it ses necessites: `He relieved nature.'
Je m'en separe avec beaucoup de peine: `I'm sorry to part with it.'
Er strdubte sich mitHanden und Fussen: `He defended himself tooth and nail.'

Note (a) the fact that the SL expression is now old-fashioned or refined is irrelevant, since you translate into the modern target language; (b) however, if such expressions appear in dialogue, and are spoken (typically or say) by middle-aged or elderly characters, then a correspondingly `refined' translation Is appropiate; (c) naturalness has a solid core of agreement, but the peripheryis a taste area, and the subject of violent, futile dispute among informants, who ,,ill, Ini m that it is a subjective matter, pure intuition; but it is not so. If you are a translator, check with three informants if you can. If you are a translation teacher, welcome an SL informant to help you decide on the naturalness or currency (there is no difference), therefore degree of frequency of an SL expression.
other `obvious' areas of interference, and therefore unnaturalness, are in the vr,v wl the articles; progressive tenses; noun-compounding; collocations; the , m rrncy of idioms and metaphors; aspectual features of verbs; infinitives.

how do you get a feel for naturalness, both as a foreigner and as a native i.vl w ' The too obvious answer is to read representative texts and talk with w p v .wnative TL speakers (failing which, representative TV and radio) - and to mmrsclf fearlessly corrected. Beware of books of idioms - they rarely dis­tinguish between what is current (e.g. `keep my head above water') and what is h id i r.t;. `dead as a door nail').
There is a natural tendency to merge three of the senses of the word `idiom': (a)a group of words whose meaning cannot be predicted from the meanings of their .v r.i imcnt words (e.g. dog in the manger; Spielverderber; l'empecheur de tourner en rond) the linguistic usage that is natural to native speakers of a language; (c) the characteristic vocabulary or usage of a people. (Elle avait frappe d la bonne porte. Ca,c’est du francais!) when the original was merely Elle avait trouve la solution (`She had found the solution'), which is also perfectly good French.) The danger of this p i)ccdure is that it tends to devalue literal language at the expense of `idiomatic' Innguage, as though it were unnatural. If anything, the reverse is the case. i .cnainly, idiomatic language can, being metaphor, be more pithy and vivid than Incral language, but it can also be more conventional, fluctuate with fashion, and become archaic and refined (`he was like a cat on a hot tin roof) (sur des charbons wvlunts; wie auf gluhenden Kohlen sitzen), and, above all, it can be a way of avoiding i he (literal) truth. In translating idiomatic into idiomatic language, it is particularly difficult to match equivalence of meaning with equivalence of frequency.
Check and cross-check words and expressions in an up-to-date dictionary (I,ungmans, Collins, COD). Note any word you are suspicious of. Remember, your mind is furnished with thousands of words and proper names that you half lake for granted, that you seem to have known all your life, and that you do not properly know the meaning of. You have to start checking them. Look up proper names as frequently as words: say you get Dax, cite de petites H.L.M. - `Dax, a small council flat estate' may sound natural, but looking up Dax will show you it is incorrect, it must be `Dax, a town of small council flats' - always assuming that `council flat' is good enough for the reader.
Naturalness is not something you wait to acquire by instinct. You work towards it by small progressive stages, working from the most common to the less common features, like anything else rationally, even if you never quite attain it. There is no universal naturalness. Naturalness depends on the relationship between the writer and the readership and the topic or situation. What is natural in one situation may be unnatural in another, but everyone has a natural, `neutral' language where spoken and informal written language more or less coincide. It is rather easy to confuse naturalness with: (a) a colloquial style; (b) a succession of cliched idioms, which some, particularly expatriate teachers, think is the heart of the language; (c) jargon; (d) formal language. I can only give indications:
(avant tout) (F)
(a) first of all
(b) before you can say Jack Robinson
(c) in the first instance
(d) primarily
plus ou moins (F)
(a) more or less
(b) give or take
(c) within the parameter of an approximation
(d) approximately


COMBINING THE FOUR LEVELS
Kunststuck, tour de force, `feat of skill', dimostrazione di virtuosismo: summarising the process of translating, I am suggesting that you keep in parallel the four levels-the textual, the referential, the cohesive, the natural: they are distinct from bu1 frequently impinge on and may be in conflict with each other. Your first and lasl level is the text; then you have to continually bear in mind the level of reality (which may be simulated, i.e. imagined, as well as real), but you let it filter into the text only when this is necessary to complete or secure the readership's understanding oi the text, and then normally only within informative and vocative texts. As regard, the level of naturalness, you translate informative and vocative texts on this leve: irrespective of the naturalness of the original, bearing in mind that naturalness in. say, formal texts is quite different from naturalness in colloquial texts. For expres. sive and authoritative texts, however, you keep to a natural level only if the origina: is written in ordinary language; if the original is linguistically or stylisticall} innovative, you should aim at a corresponding degree of innovation, representin€ the degree of deviation from naturalness, in your translation - ironically, ever when translating these innovative texts, their natural level remains as a point o1 reference. For sincerite explosive, `impassioned, enthusiastic, intense or violent sincerity' may be natural, but sincerite explosive is what the text, a serious novel says, so `explosive sincerity' is what you have to write, whether you like it or nol (you will get accustomed to it, on sy fait d tout)-unless, of course, you maintain (1 disagree) that the figurative sense of explosif (temperament explosifl has a widel currency than the figurative sense of `explosive' (`an explosive temperament') when you are justified in translating explosif by another word you claim come within its semantic range (`fiery sincerity'?).
Paradoxically, it is at the `naturalness' rather than the `reality' stage of translating that accuracy becomes most important - therefore at the final stage.when you (reluctantly!) realise that a literal translation will not do, that it is either unnatural or out of place, there is a great temptation to produce an elegant variation simply because it sounds right or nice; say, for Si mince, si depouruu de chair, qu'on . es bien oblige de comprendre les petits copains feroces de la communale, qui l'ont wmnnmu Baton. (Bazin, L'Eglise verte.) You translate: `So thin, so deprived of flesh that you really can't blame his spiteful little friends at the local primary school who have nicknamed him "Stick".' Here the main trouble is `spiteful' for feroces: p(11U' simply isn't in the word feroce, it will not stretch that far and it is wmw cssary. The pragmatic (not the referential) component of copain is missed I~m 'pals' or `mates' won't fit). On est oblige is stretched a little too far, whilst l~lvnwnu de is deceptive, it is such a common construction that even `lacking in' is a In r la 'refined' or elevated. I would suggest: `So thin, so fleshless that you have to I n nv uuderstanding for his fierce (alt. `ferocious') little friends at the local primary school, who have nicknamed him "Stick".'
This is a stab at accuracy as well as naturalness, and in the case of the on est oblige de comprendre, it is not at the colloquial level of the first translation, but one ~ -u r Id maintain that the French is not racy or colloquial either. Admittedly, except for technical terms and for well-used words for culturally overlapping familiar object and actions, accuracy in translation lies normally within certain narrow of words and structures, certain linguistic limits. It is not so precise as pmisc, it is not `this word and no other'. It is not an absolute (there are no absolutes in translation). It represents the maximum degree of correspondence, refentially and pragmatically, between, on the one hand, the text as a whole and its various units of translation (ranging usually from word to sentence) and, on the other, the extralinguistic `reality', which may be the world of reality or of the mind. Admittedly it is harder to say what is accurate than what is inaccurate-translation r. like love; I do not know what it is but I think I know what it is not - but there is ;ilw;Iys the rappel d fordre, usually to bring you back to a close translation, and at least to show you there is a point beyond which you can't go.
THE UNIT OF TRANSLATING
Normally you translate sentence by sentence (not breath-group by breath-group), nnming the risk of not paying enough attention to the sentence-joins. If the r r;lnslation of a sentence has no problems, it is based firmly on literal translation (t tic literal translation of comprehensif is `understanding' and of versatile, `fickle'), plus virtually automatic and spontaneous transpositions and shifts, changes in word order etc. Thus:
MB, arrete d Perigueux le 13 fevrier, observe actuellement une greve de la faim,
MB, who was arrested in Perigueux on 13th February, is at present observing a hunger strike.

The first sign of a translation problem is where these automatic procedures from language to language, apparently without intercession of thought (scornfully referred to as transcodage by the ESIT School of Paris), are not adequate. Then comes the struggle between the words in the SL-it may be one word like `sleazy', it may be a collocation like `a dark horse', it may be a structure like `the country's government' (who governs what?), it may be a referential, cultural or idiolectal problem - in any event, the mental struggle between the SL words and the TL thought then begins. How do you conduct this struggle? Maybe if you are an interpreter, a natural communicator (I write half-heartedly), you try to forget the SL words, you deverbalise, you produce independent thought, you take the message first, and then perhaps bring the SL words in. If you are like me, you never forget the SL words, they are always the point of departure; you create, you interpret on the basis of these words.
You abandon the SL text - literal translation if you like (which, for the purpose of this argument, I couple with mandatory or virtually mandatory shifts and word-order changes) only when its use makes the translation referentially and pragmatically inaccurate, when it is unnatural, when it will not work. By rule of thumb you know literal translation is likely to work best and most with written, prosy, semi-formal, non-literary language, and also with innovative language; worst and least with ordinary spoken idiomatic language. Further, it is more often effectively used than most writers on translation, from Cicero to Nida and Neubert, (but not Wilss) lead you to believe.
Since the sentence is the basic unit of thought, presenting an object and what it does, is, or is affected by, so the sentence is, in the first instance, your unit of translation, even though you may later find many SL and TL correspondences within that sentence. Primarily, you translate by the sentence, and in each sen­tence, it is the object and what happens to it that you sort out first. Further, if the object has been previously mentioned, or it is the main theme, you put it in the early part of the sentence, whilst you put the new information at the end, where it normally gets most stress:
Die Vignette hatte Thorzualdsen 1805 in Rom entworfen.
The vignette was designed by Thorwaldsen in 1805 in Rome
Your problem is normally how to make sense of a difficult sentence. Usually you only have trouble with grammar in a long complicated sentence, often weighed down by a series of word-groups depending on verb-nouns. Grammar being more versatile than lexis, you can render a sentence like the following in many versions:
rum de particularismes dans la legislation, la creation d'une fonction publique territoriale, 1'anLrlnation des regles anterieures de deconcentration aux nouveaux rapports entre Etat et -llrotivites locales ant cree une effervescence institutionnelle comme notre administration h w aln n'en avait pas connue depuis un siede.
(M. Duverger, Les Institutions franqaises. )

You can either plough through this sentence, keeping roughly to the French w:uwncw and keeping the reader guessing, or you can make compromises, or, at the other end of the spectrum, in order to clarify the sentence as far as possible, you , can try:
The following measures have profoundly shaken French institutions in a way that has i rot been known in local government for a century: what has remained of government %yocrvision has been abolished; control of procedural legality has been reorganised and regional audit offices established; executive power has been transferred to the ~Iraitmen of deliberative assemblies; regions with full powers have been created; powers of economic intervention have been extended to regional and local authorities; powers previously exercised by the State have been transferred in complete stages to r hc various types of authorities; corresponding State resources have been transferred to these authorities; specific local characteristics have been introduced into legislation; u territorial civil service has been created and previous devolution regulations have been adapted to the new relations between the State and the local authorities.

The above translation has converted a dozen verb-nouns into verbs, which gv ws against the noun-forming tendency of most languages but perhaps clarifies the woucnce.
Below the sentence, you go to clauses, both finite and non-finite, which, if you are experienced, you tend to recast intuitively (see Chapter 8 on shifts or i t:mspositions) as in the previous long sentence, unless you are faced with an obscure or ambiguous sentence. Within the clause, you may take next the two obviously cohesive types of collocations, adjective-plus-noun or verb-plus-object, w the various groups that are less context-bound. (I think Masterman's breath­group units may be more applicable to interpreters than to translators.)
Other difficulties with grammar are usually due to the use of archaic, little ii scd, ambiguously placed or faulty structures. You should bear in mind, however, t Ilat if long sentences and complicated structures are an essential part of the text, and are characteristic of the author rather than of the norms of the source language, you should reproduce a corresponding deviation from the target language norms in your own version (as in Proust).
THE TRANSLATION OF LEXIS
However, the chief difficulties in translating are lexical, not grammatical - i.e. words, collocations and fixed phrases or idioms; these include neologisms and `unfindable' words, which I deal with separately.
Difficulties with words are of two kinds: (a) you do not understand them; (b) you find them hard to translate.
If you cannot understand a word, it may be because all its possible meanings are not known to you, or because its meaning is determined by its unusual collocation or a reference elsewhere in the text.
We have to bear in mind that many common nouns have four types of meaning: (a)physical or material, (b) figurative, (c) technical, (d) colloquial;
The first thing to say about this diagram is that it is schematic, and that the colloquial meanings are tied to collocations or fixed phrases. Secondly, the tech­nical meanings are often the worst translation traps (take enjoliveur, not `pretti­fying' but `hub cap') since you expect technical terms to be monosemous, i.e. have one meaning only -a widespread illusion. (Admittedly, some of the technical term, mentioned are `familiar alternatives', and others are often compounded with theit classifiers, e.g. Orgelzug, Orgelpfeife.)
My next point is that most nouns, verbs or adjectives can be used figuratively and therefore can have figurative meanings-the more common the word, the more contagious and accessible the figurative meanings. If we are desperate, we have tc test any sentence for a figurative meaning e.g., `The man loved his garden'. The garden may symbolise privacy, beauty, fertility, simple hard work, sexual bliss. etc.
Other possible solutions to the `word problem' are that the word may have are Archaic or a regional sense (consult appropriate dictionaries), may be used ironic­ly, or in a sense peculiar or private to the writer (idiolect), or it maybe misprinted.
But be assured of one thing: the writer must have known what he wanted to say he would never have written a drop of nonsense in the middle of a sea of sense, and some how you have to find that sense, by any kind of lateral thinking: misprint, miscopying (anatomie for autonomie), author's linguistic or technical ignorance, freudian slip (prostate craniale; craniale doesn't exist, craniennie; fine, but what has a prostate to do with a skull? Skull, head, top? Upper prostate?). You have to force word (usually it is a word) into sense, you have to force your word at they appear in compounded nouns (maison centrale, prison; maison close, I u -n I ml; rnaison de culture, arts centre; maison de rapport, apartment block; maison de, convalescent home; maison de maitre, family mansion, etc.), in idioms or as an it- to in a lexical set (e.g., root, racine, Stamm in a text on linguistics). Very rarely, i l try van only be clarified by a reference to the adjoining paragraphs or beyond: any object qualified by `the' may take you far outside your sentence Another general point about translating is that, in principle, since corre­,p,nul ing SL and TL words do not usually have precisely the same semantic range f i I n a yh many do in cognate languages), you are over- or under-translating most , usually the latter. In fact, since in the majority of texts you are more ..nwcrncd with the message (function) than with the richness of description, and : n u c t he meanings of all but technical words are narrowed down in their context, n,wsl:uion correspondence is usually close. However, we must remember that a F~m :u number of words in one language include and overlap in varying degrees of nwnning the words they appear most obviously to translate into another language. I Inm French words like silhouette, discontinuite, assurer, descendre, phenomene, rvnlulion, egalement are much more common and have a wider semantic range than their cognates in English, and therefore more often than not they are trans­Iawd by several different more specific words. This illustrates one of the main p„hlems in translation, the enforced shift from generic to specific units or vice Vvrsa, sometimes due to overlapping or included meanings, sometimes to notorious Ivxical gaps in one of the languages, which may be lacking in a generic word for ,J,jerts or processes (amenagement) or in common specific terms for common parts wl the body (nuque, reins, `shin', `knuckle', `freckle'). Notoriously, there are .orpising lexical gaps and virtual duplications (visage, figure, Meer, See) in every Ivul;uage, and languages group objects differently (un fauteuil is not une chaise)-it i needs a translator to expose the apparent bits of linguistic chaos in another language i rnluoe is a grandson, a granddaughter, grandchild, nephew, niece). English, apparently the richest language in the world, cannot do better than `bank', `funny', `plane', etc. for denoting very different referents. (Its numerous monosyllables make it the most pun-prone language in Europe.) However, as long as you are sensitised to these lexical facts, you will not find them a problem unless they are used metalingually.
One little item - say, the precise meaning of a Hohenvergleichtafel: what is a `panorama'? Is it the same in German? Can it be a Kupferstich? What is the difference between an etching and an engraving? Between gravieren and einschnit­zen? All this, if you have no informant accessible, can take you longer than the 10-15 pages of the text which follow, and you have to be prepared to give all that time to it (but not in an exam). In real life, you have to be ready to take more time over checking one figure, chasing one acronym, or tracing one `unfindable' word than over translating the whole of the relatively easy and boring piece you find it in.
THE TRANSLATION OF PROPER NAMES (see also p. 214)

You have to look up all proper names you do not know. First, geographical terms. In a modern text, Beijing is no longer Peking; nor is Karl Marx Stadt now Chemnitz; nor is Mutate (Zimbabwe) any longer Umtali; and in 1997 Hong Kong will be Xianggang. Im Saaletal is `in the Saale valley' not `in Saaletal'. Do not normally call Polish or Czechoslovak towns by their German names: Posen/ Poznan, Breslau/Wroclaw, Karlsbad/Karlovy Vary, Teschen/Decin. (The Polish Minister of Information rightly protested to the West Germans about this habit recently,) Only the English refer to the Channel as theirs. Consider giving clas­sifiers to any town, mountain or river likely to be unknown to the readership. Check the existence of any place name used in a work of fiction: Tonio Kroger's Aarlsgaard does exist, but not in my atlas. Bear in mind and encourage the tendency of place-names to revert to their non-naturalised names (Braunschweig, Hessen, Hannover), but do not overdo it-let Munich remain Munich. Do not take sides on any political disputes about place-names.
Be particularly careful of proper names in medical texts: a drug in one country will be marketed under another brand name in another, or it may merely be a chemical formula such as `aspirin'. Tests, symptoms, diseases, syndromes, parts of the body are named after one `scientist' in one language community and a different one, or are given a more general term, in another. Check the spelling of all proper names - this is where misprints are most common. Remember that while English keeps the first names of foreign persons unchanged, French and Italian sometimes arbitrarily translate them, even if they are the names of living people.
In the period between translating and revision, you should not lose sight of the linguistic problems of the text. (All translation problems are finally problems of the target language.) Do not always be searching for synonyms. A change in word order may be the answer (. . , de nouveaux types d'e(ectrodes indicatrices - . . . `new indicative types of electrodes'- i.e. types indicative of future ranges). If it is a fact, not a word, you are searching for-How many casualties at Cassino?-let your mind play over the various types of reference books - or your own memories. I am not doiiving neurolinguistic, psychological processes in translation, far from it. I am merely saying you cannot analyse or schematise them; they are unconscious, part of the imagination. If you are lucky, when you brood, you find a solution suddenly surfacing.

Revision
during the final revision stage of translating, you constantly try to pare down your wvrsiun in the interest of elegance and force, at the same time allowing some v w lundancy to facilitate reading and ensuring that no substantial sense component is lost. (Another tension - the translator of a demanding text is always on some or other, like Nietzsche's Ubermensch.) This means translating le pour­.rnntge de grossesses menees d terme not as `the percentage of pregnancies brought to a ,successful conclusion', far less `pregnancies taken up to term' but perhaps as successful pregnancies'; faire fonctionner as `operating' not `putting into operation'. 1'mu are trying to get rid of paraphrase without impairing your text, the reality k-hind the text or the manner of writing you have preferred (natural, innovative or ,mlc). The virtue of concision is its packed meaning - and the punch it carries. 1' on r text is dependent on another text but, paradoxically again, in communicative i r:ulslation you have to use a language that comes naturally to you, whilst in ,w n:mtic translation, you have to empathise with the author (the more you feel with f lic author, the better you are likely to translate-if you dislike a literary text, better not translate it at all) - and in your empathy you should discover a way of writing which, whilst normally.not natural to you, expresses a certain side of you `naturally' :ant sincerely. A great translation is also a work of art in its own right, but a good i rmslation, even of a great work, need not be so.
But my last word is this: be accurate. You have no licence to change words fli:u have plain one-to-one translations just because you think they sound better iluut the original, though there is nothing wrong with it; or because you like •:vnonyms, because you think you ought to change them to show how resourceful you are. Mind particularly your descriptive words: adjectives, adverbs, nouns and verbs of quality. The fact that you are subjected as a translator to so many forces uud tensions is no excuse for plain inaccuracy.
`But that's what the author wrote.' Why do you want to change it? You rmlldn't have a clearer indication that this is what the author would write in the Ivrcign language, if he could. Why do you think he wrote cigogne when you
f r:mslate it as `migrating bird'? Why did he not write oiseau migratoire? Is it because vow're into text-linguistics, because your overall text strategies, your proto-typical n ructures, the global superstructures, the exciting new developments in the broad interdisciplinary field of the science of cognition demand this change? Surely not.
Many translators say you should never translate words, you translate sen­icnces or ideas or messages. I think they are fooling themselves. The SL texts consist of words, that is all that is there, on the page. Finally all you have is words to translate, and you have to account for each of them somewhere in your TL text, sometimes by deliberately not translating them (e.g., words like schon and dejd), or by compensating for them, because if translated cold you inevitably over-translate them.
In another chapter (Chapter 19) I detail the various points you have to look out for when you revise. Revision is also a technique that you acquire. I suggest you spend on revising 50-70% of the time you took on translating, depending on the difficulty of the text. If you have the time, do a second revision a day or so later. It is difficult to resist making continual `improvements' in the taste area, and this is harmless provided you make sure that each revised detail does not impair the sentence or the cohesion of the text. If appropriate, the final test should be for naturalness: read the translation aloud to yourself.

Conclusion
I have tended to assume a demanding and challenging SL text. One can admittedly find, somewhat artificially, translation problems in any text, any metaphor. Unfortunately, there are a great many run-of-the-mill texts that have to be translated which present few challenges once you have mastered their terminology, which carries you through into a series of frankly boring and monotonous successors. They become remotely challenging only if they are poorly written, or you have to skew the readership, i.e. translate for users at a different, usually lower, level of language and/or knowledge of the topic. Many staff trans­lators complain of the wearisome monotony of texts written in a humdrum neutral to informal style, full of facts, low on descriptions, teetering on the edge of cliche: certainly my account of the translating process will appear largely irrelevant tc them. Enterprising translators have to appeal to the research departments of their companies for more interesting papers, or themselves recommend important original foreign publications in their field for translation. Others transfer from, say; general administration to the human rights department of their international organisation to find something worthwhile to do.
It is one of the numerous paradoxes of translation that a vast number of texts, far from being `impossible', as many linguists and men of letters (not usually ir: agreement) still believe, are in fact easy and tedious and suitable for MAT (machine-aided translation) and even MT (machine translation) but still essentia: and vital, whilst other texts may be considered as material for a scholar, a researcher and an artist.