ARTICLES

Dimensions of Fidelity in Translation with Special Reference to Chinese

Yuen Ren Chao 趙元任
University of California

There are translations and translations. Whenever a sentence of this type is heard by Bertrand Russell, he will almost invariably remark dryly: 'Then there must be at least four translations.' Now how would you translate the sentence 'There are translations and translations' into a language that has no distinction between the singular and plural forms of nouns, or for that matter into a language that has a dual number in addition to singular and plural forms? The answer is, you cannot. If you try to translate it word for word, or even if you smooth out the grammar, to infer from it that there are at least four translations would then be completely non sequitur. I cite this example in order to show that translation is such a multidimensional affair that for any given material there are not only four translations, but usually many more than four translations according to the relative importance to be assigned to various dimensions.

Fig.1 'Fluency'. (Photograph: GRB)

In the late 1800's the prolific Chinese writer Yen Fu 嚴復 (1853-1921), who translated into Chinese Thomas Huxley's Essays, John Stuart Mill's System of Logic, Herbert Spencer's Principles of Sociology, among other things, used to set up three requirements for translation: fidelity, lucidity, and beauty-well, to give up the sound effect (on which more later) for a closer translation of the content of his three criteria, let us say: fidelity, fluency, and elegance for what Yen Fu called shinn, dar, yea (信, 達, 雅). But the third requirement, namely, elegance, is not always valid. Suppose at a court trial, for example, a man is sued for having said in English : 'You are a damn fool,' and the Chinese court interpreter renders it as: Nii sh ig heen bu jyhhuey de ren (你是一个很不智慧的人), the translation has no doubt gained in elegance, but will certainly not be a faithful translation of the original and might even affect the outcome of the case. As for the second requirement, that of fluency, it is generally a desirable quality in a discourse, as for example when an interpreter translates for the doctor the inarticulate or incoherent speech of a sick or injured person. However, in the case of a novelist or dramatist who is portraying differences in personality by the differences in expressiveness in the speech of his characters, it will of course not do to translate all the dialogues with equal clarity and fluency.

Thus, we have to come back to the first factor, namely, fidelity, as the main desideratum in translation. But before I take up the various dimensions of fidelity I must first raise the question as to the nature and size of the unit to be translated. The material to be translated may be a book, a poem, a dialogue in a play, or a speech, and the medium in which it is to be translated may be either written or spoken. The size may vary anywhere from a word to a whole encyclopedia. One important aspect of the translational situation is that language, whether in the form of live speech or in the form of written text, is not apart from the rest of life, but forms a part of life. This truism would hardly need repeating if it were not for the fact that students, and sometimes even we linguists, often forget it and treat language as if it were something sui generis. But when you translate a text, it is always in a context, and when you translate something spoken, it is always spoken in a situation.

In this connection, attention should be called to the interesting borderline phenomena of language and non-linguistic symbolic behavior, such as voice quality, intonation,[1] gesture, and so forth. If the same desired effect is to be attained, sometimes a word or a sentence in one language may have to be ' translated,' so to speak, by a gesture. For example, in a certain situation an English sentence:

may better be translated into French by a shrug of the shoulders than by the words Je ne sais pas spoken in any French intonation. On one occasion, when I was giving a lecture in Chinese to a Japanese audience and punctuated the ends of paragraphs with pauses, my interpreter into Japanese translated my pauses into sḥ —, that is, a sort of s or sh, with the air drawn in, as he rose from a 90⁰ bow. Now is this language? If not, then we have a translation of language into non-language. Again, in the so-called simultaneous translation setup at the United Nations, the majority of the interpreters are quite good in the total fidelity of their translations. At one time, one interpreter for the Soviet Union, who was an American citizen, was so good in rendering the exact effect of the speeches that he constantly received letters of complaint, accusing him of unpatriotism or even of treason. It was of course simply his job, and if he did not do it somebody else could. It was not recorded, however, when a certain delegate from the Soviet Union emphasized his point by putting his shoe on the table, whether his interpreter on his part also put his shoe on the table.

To return to the question of the size of the unit to translate, there is no translation at the level of single phonemes. A distinction is usually made between translation and transliteration. For example, when Oxford appears as Nioujin (牛津) 'Ox-ford,' it is translation, while New York as Neou'iue (紐約) [2] is transliteration. But when Cambridge is rendered as Jiannchyau (劍橋), it is half transliterated [3] and half translated. Similarly, Longguoofu (龍果夫) for Dragunov is the same thing in reverse, namely, with the first part in translation and the second part in transliteration.

A further distinction is usually made between transliteration and transcription. Transliteration in the strict sense is the conversion from the elements of the writing system of one language into those of another, whether systematically or haphazardly ad hoc. For example, when words in Cyrillic or Greek letters are spelt in Roman letters, it is transliteration. But when English words are written in Japanese kana according to certain rules of writing the sounds (as opposed to the spelling) of English words, it is transcription. For writing systems in which the graphic unit is the syllabic morpheme, such as the Chinese, rules of both transliteration and transcription will be rather complicated to establish, whence the great divergence in writing foreign names in characters. Add to this the divergence in the dialects of the writers and the result is even more complicated. I used to know the name of the famous English natural philosopher as 奈端, which is Nayduan in Mandarin. It was not until many years later that I learned that in some variety of Cantonese the characters are pronounced Noaytoan, being a fair approximation to the German pronunciation of the name Newton. Efforts have been made, in which I have taken part, to establish equivalences between syllabic types in Western languages and Chinese characters for at least a one-way consistent system of transcription of sounds, if not a system of transliteration, but because of considerations of elegance and compatibility in length (see below on sound effects), no system has yet been adopted either officially or in practice. A somewhat unimportant form of transliteration is the conversion of one form of writing into similar-shaped elements of another system of writing, called by C. J. Catford 'graphological translation.' For example:

Original	СПУТНИК
Graphological Translation	CHYTHNK
Transliteration	SPUTNIK [4]

Another example is in the name of the honor society in oriental languages at the University of California. Because Chinese and Japanese are the major languages of the Orient, its name in Greek letters is 'Phi Theta,' that is, 中日. Trivial as such examples are, graphological translation may be of increasing importance in view of the possibility of graphical scanning in machine translation and other mechanical treatment of written text.

Translation proper begins when we deal with meaningful units from morphemes and words on. While everyone is more or less aware of the multiplicity of meanings for the same word, translators often forget that the levels of units between languages need not always correspond. For example, while Western translators usually render correctly each character in classical Chinese into one word or one morpheme, as in yii wei (以為) 'take (it) to be,' suoo yii (所以) 'wherewith,' swei ran (雖然) 'although (it is) so,' they often overtranslate when handling modern Chinese, in which many compounds should be translated as single words. Thus, the forms in the preceding examples would be yiiwei 'to think (mistakenly),' suooyii 'therefore,' sweiran 'although.' As to the multiplicity of meanings for the same word, it is usually a safe guide, as I. A. Richards has observed (in a conversation with the writer), to tell whether the same word occurring in different places is to be translated into the same word or different words by noting whether the meanings come under the same numbered definition in a monolingual dictionary. For instance, the word 'nice' under number one goes into German fein, under number two into German hübsch; or, again, the word 'state,' under number one is German Zustand, under number two, Staat. This is of course not to imply that one language is more ambiguous than the other, since it works both ways. Thus, we have:

in which tzuoh is ambiguously 'do' or 'make' and jiaw is ambiguously 'make' or 'call,' while 'make' is ambiguously tzuoh, shyy, or jiaw. Similarly, we have the following chain ambiguities:

The most specific kind of context in which a word or a sentence occurs is that of an actual instance of occurrence in a situation. This constitutes what is in the terminology of communication theory a token of the word or sentence as a type. Thus, when Mencius interviewed King Huey of Liang and the king said: 'Soou 叟 (Sir) !'[5] the word soou (which happened also to be a one-word sentence) was a token of the type soou. Because philologists are chiefly concerned with the analysis of actual texts in specific contexts, while linguists are primarily interested in typical forms in general, I often characterize the difference between the two disciplines by saying that philology is the study of tokens and that linguistics is the study of types. Translation of a historical text is then the translation of a token and should, after adequate research in the context, yield a definitive translation of the original. This is, however, only true in so far as the interpretation of the original is concerned. Since the user of the translating language and the hearer or reader may each vary as to his own background and as to the circumstances of hearing or reading, there may still be the necessity of differences in the translation even for the same specific text. Hence the controversies over the old versus the new versions of the Christian Bible, since to readers of the older generation the Authorized, or Douay Rheims, version will have very definite associations and overtones which they miss in the modern versions. On the other hand, the new generation may possibly get better approximations to the effect of the original from a modern version, so its defenders claim, than from an old version, since it never grew up with it in the first place.

So much for the problems of size in translation. Now, to examine more closely the various dimensions of fidelity, one important dimension is that on the scale of semantic versus functional fidelity. Is the translation to tell what the original means, or is the translation to do what the original does in the given situation of use? As an extreme case of purely functional translation, with zero degree of semantic fidelity, I shall cite the example of Dr. P. C. Chang's interpreting of the lectures by the famous female impersonator Mei Lan-fang. This was how it went at the beginning of one of Mei's lectures in 1930:

Mei: 'Sheaudih jehshie ryhtz cherng gehwey inchyn jauday, jensh gaanshieh de heen.' (小弟這些日子承各位殷勤招待, 真是感謝的很)

Chang: 'The fundamental principle of Chinese drama is simplicity itself. '

and so it went on for the rest of the hour.

But as examples of translation in a more serious sense, take the sentence: Ne vous dérangez pas, je vous en pris ! A semantic translation of it into English might be 'Do not disturb yourself, I pray you!' while a functional translation might be simply 'Please don't bother!' The second translation is functional because that is what one would say in English under the same circumstances. But if we look closer at the constituents being translated in this and in fact any other material for translation, we shall find that the difference between the semantic and the functional is a matter of degree. To be sure, there would be no point in equating dérangez to 'derange' since that would be giving the etymological cognate and not translating. But a close semantic translation could be 'disturb yourself.' On the other hand, 'l request you' for je vous en pris is closer semantically, while 'please' is functionally what one would more likely say in cases where one would say je vous en pris. But isn't the meaning of a word in a context or in fact isn't the meaning of any linguistic form that which one would normally say under those circumstances? If so, then the best semantic fit in a translation will have to be also functionally the most suitable to use. The idea of semantic translation, however, is not completely without meaning—no pun intended. By semantic translation one usually refers to the most commonly met with meaning of a word, and, other things being equal, to the etymologically earlier meaning. This is again a matter of degree, since all semantic meaning is in one sense functional.

Correlated highly, though not identical, with the semantic-functional dimension, is that of literal versus idiomatic translation. The term literal is a misnomer, since it would seem to mean transliteration. In actual usage, of course, a literal translation means a word-for-word or morpheme-for-morpheme translation. The sign Tsyy Luh Bu Tong (此路不通) says literally 'This road doesn't go through,' but the idiomatic equivalent is 'Not a Through Street' (or 'No Thoroughfare' in England). Jau Tie Jih Sy (招貼卽撕) says literally 'Signs Pasted (will be) Immediately Torn,' but is idiomatically equivalent to 'Post No Bills.' From another point of view a literal translation may also be regarded as a fine-grained translation, but not necessarily of high overall fidelity if it is not idiomatic or functionally misleading. In this literal-idiomatic dimension there are also differences of degree on a sliding scale. Thus, between the French and the English forms of signs about smoking, there are the following possible steps to consider:

Prière de ne pas fumer	Original
'Prayer of not a step to smoke'	Literal translation
'Request of not smoking'	Grammatical but not idiomatic
'No smoking, please'	Idiomatically acceptable
'No smoking'	The usual sign

There is one usage by which a literal translation is applied to a smoother form than a word-for-word translation. To quote again from Catford,[6] adding a comparison with Chinese, we have:

English:	It's raining cats and dogs	Original
French:	II est pleuvant chats et chiens	Word-for-word tr.
	II pleut des chats et des chiens.	Literal tr.
	II pleut à verse.	Free (idiom.) tr.
Chinese:	Ta sh shiahj mhau her gooumen.	他是下著貓和狗們.
	Shiahj mhau goou ne.	下著貓狗吶.
	(Yeu dah de jeanjyr sh) shiah mhau shiah goou Ie.	(雨大的簡直是)下貓下狗了.
	Chingpern dah-yeu Ie.	傾盆大雨了.

Note that the French and the Chinese happen to agree literally, too.

A word of warning should be said here against the strong temptation to use an interesting literal translation at the cost of fidelity in the other dimensions. If a translation is both literal and idiomatic, well and good, as in the French and Chinese above. Again, in: Ta bu hwai hao-yih (他不懷好意) 'He doesn't harbor good intentions,' the equating of hwai with 'harbor' is very apt. When 'The style is the man' is translated as Wen ru chyi ren (文如其人), it is fairly close, though the Chinese is in wenyan, while the English is neutral in that respect. When, however, Shiawhuah! (笑話!) is equated to 'Ridiculous!' then there are problems. For while the Chinese can be used either as non-polite or as insulting language, the English can only be the latter if applied to the person being spoken to. Even more subtle are the shades of differences between donq₀syyle [7] (凍死了) and 'frozen to death.' Most of the time, both are used either in the literal sense or as a hyperbole and the Chinese form with or without neutral tone can be used either way too. But depending upon context, one may be idiomatic in one but not in the other language.

A very important dimension of fidelity which translators often neglect is comparability in frequency of occurrence, or the relative familiarity of the expressions in the original and the translation. Too great a discrepancy in this respect will affect fidelity, even though the translation is accurate in other respects. As is well known in information theory, the less often a thing is talked about, the more it means to talk about it. Sometimes the very things one talks about may be a familiar thing in one culture and strange and exotic in another. In such a case, if the thing is the main topic of the discourse, it cannot be helped. An account of a game in the World Series can very easily be translated into Japanese, but would make poor reading in Chinese, in which terms about soccer are heard every day, but not those of baseball. However, in cases where a familiar expression is used casually as a figure of speech, then sometimes a translation by a different figure of speech of the same import but with a comparable degree of familiarity will result in a higher degree of overall fidelity than an apparently faithful translation which is very unfamiliar. For example, to speak of reaching the third base might be rendered, in Chinese, as reaching the 'listening stage' in a game of mahjong, where the apparently 'free' translation has greater fidelity, because it is a better match in the frequency of occurrence. Technically, the third base is in Chinese dihsan leei (第三壘). But at the lecture on these problems of translation, at which there were probably thirty or forty Chinese-speaking members of the audience, I asked how many had heard the expression dihsan leei and not one of them raised his hand. My daughter, Rulan Pian, was in the audience, but did not raise her hand, because she had just learned the term that same afternoon, as I had myself.

Before continuing with the consideration of the other dimensions, let us consider for a moment one aspect of the translation which has to do with the dimension of size, literalness, and frequency, namely the phenomenon of calque, or translation borrowing. In ordinary borrowing from one language to another, a foreign word or expression is taken over and adapted to the phonemes of the borrowing language, as for example, English menu ['meniu] or ['meiniu] from French menu [məny], or English chopsuey from Cantonese dzaapsöy (雜碎). In such cases, whether there is a change of meaning or not—and usually there is—no translation is involved. In translation borrowing, on the other hand, one translates the constituent parts of foreign words and makes up new combinations, thus forming neologisms. For example, the German noun for telephone is Fernsprecher, tele- translated as fern- and -phone freely translated as -sprecher. On the other hand, in the verb telephonieren, there is direct borrowing from the Greek (except for the addition of the German verbal suffix). Another example is German Einfluss, from Latin in + fluens. Sometimes, especially in translation borrowings of phrases instead of compound words, the borrowings may be so naturalized that most users are hardly aware of their foreign origin. Examples are: 'That goes without saying' < Ça va sans dire, or the colloquial 'How goes it?' < Wie geht's? 'Long time no see,' however, is not a translation borrowing, since Hao jeou bu jiann Ie (好久不見了), if translation-borrowed, would come out as 'Good long not met.'

Much more tricky are what I call skewed translation borrowings. By a skewed translation borrowing I mean one in which you translate a foreign word with meanings A, B, C, D, etc. with a certain native word for meaning A and then, instead of choosing other suitable words for the other meanings B, C, D, etc., just go on using the same native word for A mechanically whenever you see the foreign word. The result amounts to an importation of foreign meanings which the native word never had before. Present-day Chinese is full of such skewed translation borrowings, such as:

	Old meaning	Added meaning
weimiaw (微妙)	'delicate (of things)'	'delicate (of situations)'
chyangdiaw (強調)	'stress (in pronunciations)'	'to emphasize'
chingsuann (清算)	'liquidate (accounts)'	'liquidate (persons)'
Iiisheangde (理想的)	'ideal (adj. of idea)'	'ideal (perfect)'

Such borrowings always take time before they are quite naturalized and are mostly limited to journalistic language or discourse in a journalistic style. Some of the new ones appearing in headlines, especially in overseas Chinese newspapers, are hardly intelligible without reading on in the text or retranslating them into the source language. For example, one headline says that the crime rate in San Francisco had a shihjiuhshinqde (戲劇性的) decrease last month.[8] It made no sense to me until I realized that shihjiuhshinqde did not mean 'theatrical,' but 'dramatic': there was a dramatic decrease in crime rate. Another news item, about a manifesto concerning the hydrogen bomb signed by Bertrand Russell, Albert Einstein, and others, said in Chinese: 'Ever since the tests at Bikini, lianghao de dangjyu (良好的當局)—excellent administrators—unanimously have pointed out the danger that a war of hydrogen bombs can destroy the whole of mankind.'[9] I had to read the column twice before I realized that what they called lianghao de dangjyu was a skewed translation borrowing of 'good authorities,' good authorities have pointed out etc. To be sure, this sort of lazy man's translation is constantly being committed by students in foreign language classrooms. But when a new meaning becomes established, even though through foreign influence, it becomes part of the language—shall I say lingo?—whether you like it or not. But I am sure that excellent administrators for 'on good authority' is still unintelligible at the present stage.

To continue with the consideration of dimensions of fidelity, another dimension in which a translator may fall into the trap of what may be called false fidelity is the presence of obligatory categories in languages. A noun in English has to be either singular or plural, a verb either present or past. A friend in German has to be either male or female. A cousin in Chinese has to be not only either male or female but also either on the father's side or on the mother's side, either older or younger than oneself. What a translator has to do is of course to omit the obligatory distinctions, whether lexical or grammatical, if they are not obligatory in the translating language and if they are not relevant in the context. For instance, Chinese beau (表) is an adjective for relatives of different surnames and mey (妹) is a female relative of the same generation younger than oneself.[10] But if the obligatory distinctions do not matter in a certain context, then the combination beaumey can very well be undertranslated simply as 'cousin,' otherwise one would have to say things like: 'Good morning, my femalecousin-on-mother's-or-paternal-aunt's-side-younger-than-myself!' Again, in Chinese, as in Russian, to be married is one word for men (cheu 娶, жeнaт) and a different word for women (jiah 嫁, эамужем). Once, when I was interpreting a lecture by Dora Black in Peking, she said something about unmarried men and unmarried women, and I got the words cheu and jiah mixed up and came out with meiyeou jiah de nanren gen meiyeou cheu de neuren (沒有嫁的男人跟沒有娶的女人). Of course the audience roared with laughter at this, and when the speaker was puzzled and asked me what they were laughing about, all I could do at the moment was to whisper to her, 'It'll take too long now, I'll have to explain it to you afterwards.'

It is easy enough to take care of such striking and obvious cases of obligatory categories, but it is the less obvious cases that are more tricky and more easily mislead the translator. Take the innocent-looking or sounding sentence: 'He put on his hat and went on his way.' In nine cases out of ten, a French, a German, or a Chinese student of English would translate it 'faithfully' with the pronoun 'his' in both places, whereas if he were to start composing the message in his own language, say in Chinese, he would probably just say: Ta dayle mawtz tzooule (他戴了帽子走了).[11]

Of course if overtranslation of obligatory categories is written and gets read on a large scale, it can establish a new usage, at first as a neologism, then as an accepted new style. Thus, starting with an imperfect knowledge of the uses of tense in English, a Chinese translator adds mechanically the suffix Ie whenever he sees a verb in the past form, even though in his own talk and writing he does not use the suffix Ie in many instances of reference to the past. Again, he uses a preposition bey (被) for 'by' whenever he sees a passive voice in the English verb, unaware of the fact that Chinese verbs have no voice and the direction of action of a verb works either way, depending upon context, and also forgetting that the preposition bey for passive action is used only before verbs with unfavorable meanings. However, once this sort of translatese is written often enough, it gets to be written in originals, even when no translation is involved. When this happens, it constitutes what in linguistics is known as structural borrowing, that is, instead of borrowing specific words or phrases discussed above, one borrows functional ('empty') words or a whole type of structure. So nowadays, one suffers not only scolding and beating but also being praised or rewarded.

Besides the translation or omission of obligatory categories, there is also the natural tendency, unless one is on guard against it, to translate noun for noun, verb for verb, or in the case of phrases, nominal for nominal expressions, verbal for verbal expressions, etc. Other things being equal, this will of course be a contributing factor toward fidelity. But since other things are never equal, they must all be considered and given no more than proper weight. For example, quelle merveille! is a nominal expression, but to render it as 'what marvel!' would be too strong, nor is it comparable in the dimension of frequency of occurrence. Instead, 'how marvelous !' would have a higher degree of overall fidelity, even though it is an adjectival and not a nominal expression. Likewise, the adjectival phrase jen taoyann (真討厭) is better translated by the nominal phrase 'what a nuisance' than the adjectival phrase 'how annoying.' So is jen haowal (很好玩ㄦ), an adjectival phrase, better translated by the nominal phrase 'what fun,' whereas the corresponding adjectival phrase 'how funny' would be entirely wrong. In Luen daw nii le (輪到你了) 'It's your turn now,' luen is a verb and 'turn' a noun. In Nah sh shyunhwan de (那是循環的) 'It's a vicious circle,' shyunhwan de is an adjective and 'circle' a noun, with 'vicious' understood in the Chinese. A translator would be strongly tempted to translate keeren (可人), which has a nominal root, as 'personable,' which, however, is not as accurate as 'lovable,' with a verbal root.

Sometimes, especially in clichés and proverbs, the most faithful translation will be of an entirely different structure. In Woo terng (我疼), woo is subject, but in 'It hurts' the 'me (understood)' is object. Chii yeou tsyy lii (豈有此理) is a whole sentence in wenyan, but used in speech as an adjective and should be translated as 'ridiculous.' 'I wish' followed by a contrary-to-fact clause could be equated to Woo yuannyih…, as in Woo yuannyih nii bye nemmyanql long (我願意你別那麼樣ㄦ聾) 'I wish you were not quite so deaf,' but a closer translation is ... (nah) dwo hao ([那]多好), preceded optionally by woo yuannyih. Huu tour sher woei (虎頭蛇尾) lit. 'Tiger's head, snake's tail' is a phrase of two nominal expressions; its equivalent 'anticlimax' is one noun. Jiin-shanq tian hua (錦上添花) and 'carrying coals to Newcastle' are fairly close in structure, but its counterpart in Chinese sheue-lii sonq tann (雪裏送炭), a verbal phrase, has its best equivalent in 'A friend in need is a friend indeed,' which is a full sentence.

Sometimes, not only the form classes do not need to correspond, but even radically different categories of linguistic elements may turn out to be the best translational equivalent. There is a very common grammatical form in Chinese consisting of a predicate, which may be a verb or an adjective, followed by the verb 'to be' sh(yh) (是), then followed by a repetition of the same predicate, as in hao sh hao (好是好). One can analyze this as '(as for being) good, (it) is good.' But this is really explaining the Chinese to a student of the language and not actually translating it. How then would you translate sentences of this type? Well, you translate this Chinese formula of words into an intonation in English. The English intonation which fits this Chinese formula best is what Harold E. Palmer [12] calls 'the swan,' so-called because its time-pitch graph makes a double turn like the neck of a swan. The plain statement Hao means 'It's good': but in the form Hao sh hao it means 'It's good (but).' It is of course also possible to render this formula by such phrases as 'to be sure,' or the more colloquial 'all right,' as in '(It's good) all right',' (with a low rising intonation), but the swan intonation is about as faithful a translation of the Chinese formula as any translation by the use of words. In extreme cases, language is even translated by non-language, such as gesture, as mentioned above.

Similar to the problem of obligatory categories, there is the problem of translating the endless varieties in different cultures of the subcategories of things and qualities, units of measure, money and coinage, names of colors, and the very names of numbers themselves. English has no juotz (桌子) 'table' – 'desk'; no shia (蝦) 'shrimp' – 'prawn' – 'lobster'; no che (車), since 'vehicle' would be out of style in most contexts; no ta (他), though current Westernized writing differentiates 他: 她: 它; there is not even ren (人), and 'man' often has to serve as 'woman.' When you call a woman huay-ren (壞人) you can neither call her 'bad woman,' nor 'bad man,' and 'bad person' would again be out of style, and you may have to settle for 'bad girl.' There are four equally common auxiliary verbs in Chinese: neng, keen, keeyii, and huey (能, 肯, 可以, 會), with overlapping equivalences with English 'can' and 'may,' with keen equatable to the awkward and therefore less frequently used 'be willing to.' There is only one word 'hot' for both tanq (燙) for temperature and lah (辣) for the taste. Among Chinese dialects, the sentence in Mandarin: Jeh tang tay tyan, keesh bu gow shian (這湯太甜, 可是不狗鮮) 'This soup is too sweet, but not tasty enough' would be difficult to translate into Cantonese without some circumlocution, as both tyan and shian would be called dhim (甜) in Cantonese. In names of colors there is no 'brown' in Chinese and there is no ching (青) in English. Many languages have no word for a length comparable to a yard, and the conception of teen-age would not be translatable unless the language happens to have a common feature from thirteen to nineteen. It is easy enough to translate such items, even with a high degree of accuracy, if it is a matter of giving the mathematical, physical, or economic equivalents. But since such expressions are often used for other than their purely quantitative import, fidelity in the other dimensions such as function, idiom, frequency, etc. will have greater weight. For instance, for a language with no word for dozen, 'a couple of dozen' will appear better as 'a couple of tens' than as 'about twenty-four.' Incidentally, such linguistic and cultural differences sometimes even affect wholly non-linguistic matters. Thus, it is not only often difficult to translate 'quarter' into a language that has a dollar-like unit but divides it into five twenty-cent pieces, but the existence of the quarter (or 20-cent piece, as the case may be) actually affects the prices of things that can be conveniently sold over the counter-and in slot machines!—so that the dimension of frequency will be affected in the translation of such items. Nobody would have said pas un sou if there had not been such a coin as the sou. Nobody would have said meiyeou ig benqtz (蚌子) if there had not been such a coin as the square-holed 'cash.'

Style is another dimension in which too much discrepancy will obviously affect the fidelity of translation. One may jazz up serious literature into modern slang, but that would be parody and not translation. Today's style in one language can of course be best translated in today's style in another, especially if the subject is one which is being talked about today. If it is a text of a past age, the translation leads to problems. I have already mentioned the problems involved in translating the Bible, and whole treatises have been written about them. For example, that very readable book, Trials of the Translator by Ronald Knox (New York, 1949), is mainly concerned with such problems. As for the age of the languages involved, there is no necessity, or even special virtue, in matching period with period. Must one, for example, translate The Divine Comedy in the language of The Canterbury Tales? If such a translation already exists in its own right, well and good, but there is no special virtue as to fidelity in matching periods as such. Moreover, what if the text to be translated, say the Chinese classics, was written long before the age of the translating language, say before the formation of what might be called the English language? The wise course in such a case, and this is the course that has been commonly adopted by most translators of the older texts, is to write in as timeless a style as possible. This practice, to be sure, may involve a loss of color and life, but it will at least be free from suggesting the wrong color. It is true that in the long run what seems timeless to the translator of one age will eventually be dated and that is why there had to be retranslations of important works, as people have done with the Bible and as John Ciardi has been doing with his 'Englishment' of the Paradiso.[13] The important thing about handling the older texts is that one should at least avoid the use of local color and narrowly dated expressions. For nothing gets as easily off color as that which is full of local color and nothing so quickly out of date as that which is right up to date.

An extremely important but often neglected dimension of fidelity is what might be called the sound effects of the language. I refer to such elements as length, symmetry, and, in the case of verse, meter, rhyme, and other prosodic elements. Now, since the semantic range of words and the obligatory categories of two languages never coincide, if all that is in the original has to be accounted for, the translation will necessarily be longer; but in trying to include everything and not to lose anything in the original, the translator will unavoidably add extraneous elements because of overlapping categories in the translating language. In practice, therefore, a translator will have to make a compromise between the sins of omission and the sins of commission and try to take into account all the dimensions of fidelity, including that of aiming at comparability in length. Take the French expression for talking nonsense et patati et patata. If you translate it as 'gibberish,' it will sound rather weakish; 'yak yak' is better, since it has a more similar pattern, and 'yakety yakety' will be even closer to the sound effects of the French et patati et patata. Translating Woode shin putelputelde tiaw (我的心撲忒ㄦ撲忒ㄦ的跳) as 'My heart palpitates' seems to give a pretty close sound effect, but 'My heart goes thumpety thump' has the advantage of comparability in length and style. To quote from John Ciardi: [14]

Every word has a certain muscularity. That is to say, it involves certain speech muscles. Certainly any man who is word-sensitive is likely to linger over the difference between the long-drawn Italian carina and the common, though imprecise, American usage 'cute' when applied to an attractive child. The physical gestures the two words invite are at least as different as the Italian child's goodbye wave ('Fa ciao, carina') with the palm of the hand up, and the American child's ('Wave bye-bye') with the back of the hand up.

Even street names have to be translated with due regard to comparability in length. One writer, in making fun of the street name 'Avenue of the Americas,' says: 'Yes, this is the Street-of-the-Great-Leap Forward -of-our -glorious -People's -Commune -System -over -the -Capitalist-Butchers-of-the-West, but everybody here still calls it Sixth Avenue.' To keep on the same theme, it is reported that the street in Peking on which the Russian embassy is situated has been renamed 'Anti-Revisionist Avenue': nine syllables. But in an actual photograph of the new street sign that I saw in The New York Times, it says: 反修路—only three syllables. Proverbs and common sayings are often equatable between one language and another, preferably with similar rhythmic effects. For example, 'As ye sow, so shall ye reap' goes quite well into Chinese as Jonq gua der gua, jonq dow der dow (種瓜得瓜, 種豆得豆), which says something like 'Plant melons (and you) get melons; plant beans (and you) get beans.'

In translating songs to be sung to the same melody, the requirement of sound effects is of course even more strict. Take, for example, the first two lines of Schubert's Erlkönig:

Here the words marked + and - are those which have been added or omitted, respectively, for reasons of rhyme and rhythm. (Note also the bad stress pattern in 'his young.') The preceding is still a fairly close translation. In the Haiden-Röslein, however, the demands of rhyme and rhythm are so strong that there is even no point in counting the pluses and minuses, as can be seen in the opening lines:

At the other extreme, as examples of sacrifice of sound for the sense is the usual type of translation of classical Chinese verse, such as that of the Book of Odes by James Legge or T'ang poems by Arthur Waley, in which the number of syllables is three or four times that of the original. While the message and imagery is usually very well conveyed in such translations, they give the feeling, to us who were raised in the concise and rhythmic swing of the shorter lines, of big mouthfuls of dough, if indeed not quite wey ru jyau lah (未如嚼蠟).

The more rhythmical is, however, not necessarily the more concise. Take the no spitting notice on trains of the Shanghai-Nanking Railway. The Chinese says:

Swei chuh tuu-tarn,	隨處吐痰	'Everywhere spit,
Tzuey wei eh-shyi.	最為惡習	Most bad habit.
Jih ree ren yann,	既惹人厭	It is loathsome
Yow ay weysheng.	又礙衛生	And bad for health.
Chejann yuehtair,	車站月台	Stations, platforms,
You shiu chingjye.	尤須清潔	Must keep clean, neat.
Taang yeou weiJann,	倘有違反	If you violate,
Miann chyh moh guay.	面斥莫怪	We will rebuke.'

The translation above is more rhythmic than literal. But the actual sign in English says in one sentence:

IN THE INTEREST OF CLEANLINESS AND PUBLIC HEALTH PASSENGERS ARE REQUESTED TO REFRAIN FROM SPITTING IN THE TRAINS OR WITHIN THE STATION'S PREMISES

To be sure, in the early days of that railroad, which was run by foreigners, there were in the Chinese notice overtones of the civilized management instructing those uncouth country people how to behave, while the English version was in a language of equals talking to equals. But the use of rhythmic forms in notices is very common in Chinese in any case.

When, however, it is a matter of translation between English and modern spoken Chinese, as I did for the Lewis Carroll books, I did not have the handicap of having to work with such disparate states of languages, and the rendering of sound effects was easier without sacrificing as much fidelity in the other dimensions. In Through the Looking-Glass [16] especially, I was able not only to make point for point in the play on words but also keep practically the same meter and rhyming patterns in all the verses. Take, for instance, the first stanza in Jabberwocky:

'Twas brillig, and the slithy toves

Did gyre and gimble in the wabe.

All mimsy were the borogoves

And the mome raths outgrabe.

In Mandarin it is:

While this sounds almost like the English—if Jabberwocky can be called English—all the sounds in it are nevertheless within the phonemic inventory of the initials, finals, and tones of Mandarin. When spelt in the National Romanization, it even looks like the English in places:

Yeou 'tian beirlii, nehshie hwojihjide toutz

Tzay weybial jiinj gorng jiinj berl.

Hao nansell a, nehshie borogoutz,

Hair yeou miade rhatz owdegerl.

And later on, when Humpty Dumpty explains the etymology of the difficult words, it will of course have to come out right in the translation. For example, 'in the wabe' is translated as tzay weybial, since just as 'wabe' comes from 'way before,' 'way behind,' and 'way beyond,' so does weybial come fromjeybial, neybial, and waybial, that is, 'this side,' 'that side,' and 'outside.'

In connection with the liberty taken with the original text for reasons of rhythm, length, etc. is it legitimate to add what was not in the original beyond just some necessary fillings? For example, to quote from Through the Looking-Glass again, when the Lion asks whether Alice is animal or vegetable or mineral: donqwuh, jyrwuh, kuanqwuh (動物, 植物, 礦物) and the Unicorn says she is a monster, the only natural translation for the word is guaywuh (怪物), which, though quite literal, is an overtranslation. Again, when the penultimate stanza in the epilogue:

In a Wonderland they lie,
Dreaming as the days go by,
Dreaming as the summers die:

is translated as:

Beenlai dou sh menqlii you,	本來都是夢裏遊,
Menqlii kaishin menqlii chour,	夢裏開心夢裏愁,
Menqlii sueyyueh menqlii liou:	夢裏歲月夢裏流:

lines 2 and 3 in the English say the same thing, while line 2 in the translation which, though it is in the mood of the poem, has been added rather gratuitously. Perhaps this overtranslation could compensate a little for the sin of omission in failing to translate the initial letters in each line of the poem to spell out the name ALICE LIDDELL.

Finally, a dimension of fidelity of practical import which has already been touched upon briefly is the situation of use of the original language and that of the translating language, and this often involves the interchange of language and non-language. In translating plays from English into Chinese, I have often met with cases where dialogue has to be translated as stage direction and vice versa. There is a Chinese character 唉! which in certain contexts every reader will pronounce as [ɦai]. Now this involves the use of the 'voiced h,' a non-existing sound in the normal list of Mandarin phonemes and is therefore on the borderline of language and non-language. To put it in English, the usual practice is of course simply to write the word sigh, which is then translating quasi-language and not ordinary language. One would then be giving a stage direction in place of giving a translation of the dialogue. Sometimes, during the act of translating live speech, the situation itself changes before the translation is finished. Then what should the translator do? If he finishes the translation, he will be translating a true sentence into a false sentence. If not, what? Here is what a resourceful airline pilot did in announcing an emergency landing, presumably on a transatlantic flight. He starts with French:

Attention, mesdames et messieurs. C'est votre commandant. Attachez vos ceintures de sécurité et préparez-vous pour un atterrissage d'urgence.

Achtung, meine Damen und Herren, hier spricht ihr Flugzeugführer. Bitte, befestigen Sie ihren Sicherheitsgürtel und bereiten Sie sich auf einer Notlandung vor.


Ladies and gentlemen, forget it. Everything is A-OK.[17]

Now is this a translation? And if so, what is the degree of fidelity?

In all the preceding discussions about dimensions of fidelity, treating them as if they were measurable, independent variables, it must be admitted that they are really neither measurable nor completely independent. We are far from reaching a workable quantitative definition of any of the dimensions, not to speak of formulating a mathematical function with a view to maximize its value.[18] The present state of affairs is still what in some of the formal disciplines is known as the pre-systematic stage, which is just another way of saying that the ideas are still half-baked. We are still not much beyond the stage, as stated by J. P. Postgate more than fifty years ago: 'By general consent, though not by universal practice, the prime merit of a translation proper is Faithfulness, and he is the best translator whose work is nearest to his original.'[19] But since nearness is a matter of degree, we are back to the problem of measurement of fidelity-back where we started. One useful test is to retranslate the translation into the original language and see if one can find a better fitting equivalent in the original language. If one can, then the translation is not faithful enough, as Mark Twain has well demonstrated. This is to be sure only a testing procedure and the problem of multidimensionality is still with us. But so far as that is concerned, in what field of inquiry is one not troubled with the problem of multidimensionality?

Related material from China Heritage Quarterly:

• The Wujing Project—towards a new translation of the Five Classics into the world's major languages
• Lionel Giles: Sinology, Old and New, by John Minford
• Contemplating Reading: Sixteen Views, by Chen Jiru 陳繼儒

Notes:

[1] That is, in so far as pitch characteristics are not a part of the phonemic system of the language being used.

[2] Naoyeuk in standard Cantonese, but pronounced Niouyoak in another southern dialect, presumably spoken by the original transliterator of this name.

[3] 劍 in Cantonese is kimm.

[4] C. J. Catford, A Linguistic Theory of Translation (Oxford, 1965), p. 66.

[5] Probably not as blunt as 'Old man!' or as deferential as Legge's 'Venerable sir!'

[6] Op. cit., pp. 25-26.

[7] A subscribed circle indicates optional neutral tone, that is, either donq-' syyle or donq.syyle.

[8] The Chinese World, San Francisco, February 14, 1968.

[9] The Chinese World, San Francisco, July 11, 1955.

[10] For details on these terms, see Y. R. Chao, 'Chinese Terms of Address,' Language 32 (1956).1.217-241.

[11] In my translation of Through the Looking-Glass, in which the Red Queen objected to Alice's saying that she had lost her way because all the ways belonged to the queen, I had of course to render 'her' literally in order to make the point.

[12] For further details, see CYYY (Ts'ai Yüan-p'ei Commemorative Volume) 1933, p.148. An example of Yiddish intonation as a grammatical form is found in Catford, p. 54.

[13] To be published, according to a letter from Mr. Ciardi, in 1968 or 1969.

[14] Saturday Review, October 7, 1961.

[15] Schirmer's Library ed., Vol. 343, Eng. tr. Th. Baker, 1895, 1923, pp. 214, 228.

[16] Under the title of Tzoou Daw Jinqtz Lii (走到鏡子裏), it will form Volume II of Readings in Sayable Chinese (Asian Language Publications, San Francisco, 1968), where a better version of the following lines can be found on p. 32, lines 1-4 (first stanza) .

[17] From a cartoon in Punch, October 19, 1966, p. 577.

[18] A beginning in the quantitative study of quality is found in John B. Carroll's 'An Experiment in Evaluating the Quality of Translations,' Mechanical Translation and Computational Linguistics 9.3; 4.55-66 (1966).

[19] J. P. Postgate, Translation and Translations (London, 1922), p. 3.