Eileen Chang's Translation of The Golden Cangue

1. Introduction

f all of Eileen Chang's fiction, Jinsuo Ji is doubtless the most important one. Ever since its first publication in 1943, the novella has aroused enduring interest among literary critics and researchers. Xun Yu, one of the earliest reviewers of Eileen Chang's fiction, despite his critical opinion towards most of Chang's works, remarks that Jinsuo Ji "should at least be ranked among the most beautiful fruits of our literary garden" (1994: 121). More credit is given to this work by C. T. Hsia, who regards it as "the greatest novelette in the history of Chinese literature" (1971: 398). After Hsia, more and more researchers have invested their efforts in this famous novella, covering such aspects as metaphor, theme, characterization, and narrative strategies (Chen 1983; Tang 1994; Fan and Ji 2003; Gao 2003; Lin 2003a, 2003b; etc.).

2. Method

2.1 The Transeme Model

Leuven-Zwart's transeme model is generally regarded as the most detailed and comprehensive approach for the purpose of translation analysis (Hermans 1999: 58; Munday 2001: 63). It consists of a "comparative model" and a "descriptive model." In the comparative model (Leuven-Zwart 1989), the ST and TT are first divided into segments of transeme. The invariant meaning shared by the ST and TT transemes is identified as the architranseme (ATR). The ST and TT transemes are then compared respectively with the ATR, which reveals different ST-TT relationships. A hyponymic relationship indicates a "modulation" shift; a relationship of contrast represents a "modification" shift; and no relationship means a "mutation" shift.

In modification, difference between the ST and TT may be semantic or stylistic, and with respect to the ATR, disjunction may occur in the ST transeme or the TT transeme. Four categories can thus be distinguished: semantic modulation/specification, semantic modulation/generalization, stylistic modulation/specification and stylistic modulation/generalization. The case of modification is more complex as disjunction may occur in the semantic, stylistic or syntactic aspect, resulting in three categories of shifts, i.e. semantic modification, stylistic modification and syntactic modification, of which the last one is further divided into syntactic-semantic modification, syntactic-stylistic modification and syntactic-pragmatic modification. For shifts concerning the stylistic aspect (i.e. stylistic modulation and stylistic modification), a finer classification is made according to the contributing element of a particular style such as register, profession, time, text-type, culture, syntagmatic feature, and paradigmatic feature. Under mutation, there are three sub-categories: addition, deletion and radical change of meaning. In all, the part of comparative model is made up of 8 categories and 37 sub-categories.

The descriptive model is designed to detect how smaller linguistic changes affect the metafunction of language that operates on the story level and discourse level. The main idea is that the interpersonal function is related to focalization and the identity of the narrator, the ideational function determines the image of fictional world and mind style, and the textual function corresponds to fictional order and syntactic ordering (Leuven-Zwart 1989: 173-79). Shifts in the macrostructure can thus be figured out through shifts in the microstructure.

With the use of a tertium comparationis, the elaborate classification of shifts and the extension of translation analysis to the domain of narratology, Leuven-Zwart's transeme model is a remarkable advance over the "usual impressionistic or intuitive statements about translations" (Hermans 1999: 62). On the other hand, disadvantages of model are also notable. For one thing, the practicality of the model is doubtful as it involves a rather complex description of shifts. For another, the objectivity of the model is challenged by the fact that the identification of ATR may vary from person to person (Gentzler 2001: 134; Munday 2001: 66; Hermans 1999: 62). In addition, the transeme as comparative unit may result in extra syntactic shifts. It is therefore necessary to revise the model before it is applied in the present study.

2.2 A Revised Model

Aiming at a feasible method, revisions are made to the transeme model on the comparative unit and classification of shifts. In the transeme model, the ST and TT are compared on the basis of transemes. A transeme is either a "state of affairs," i.e. a predicate plus its arguments, or a "satellite," i.e. an adverbial specification or amplification of the state of affairs (Leuven-Zwart 1989: 156). Distinction of the two types of transeme, which is meant to give a detailed analysis of shifts, however, cannot help giving rise to extra "shifts" that are not the result of the translator's decision, but of the comparative method itself. One example will suffice.

To facilitate the analysis, we have numbered each transeme of the ST and TT. A state of affairs transeme is marked by parentheses, and a satellite transeme, by slashes. Following the transeme model, S1 plus S2 is to be compared with T1. Now the first step of identifying corresponding ST and TT transemes has brought about a translation shift, which will be termed as "syntactic-stylistic modification/ implicitation," indicating that the TT uses less elements to convey information. However, if TT1 were to be changed to "Bit by bit it came out from the clouds," the TT would also have two transemes and the mentioned shift would not have occurred.

As a replacement for transeme, a flexible comparative unit in accord with the "natural" translation unit will be adopted in the revised model. Unlike the traditional type of translation unit which is fixed structurally or semantically, what is proposed here is a changeable, dynamic concept. This is because in actual translation practice, the translator does not always maintain a fixed, static translation unit throughout the translating process (Lörscher 1991, 1993; Séguinot 1996). It is perfectly possible, for instance, that he/she translates clause by clause on one occasion, and sentence by sentence on another. Therefore, we agree with Lörscher that the unit of translation is an attention unit--it is the portion of the ST on which the translator focuses attention so as to render it as a whole (1993: 209).

With the natural translation unit as the comparative unit, it is more likely to give an objective delineation of the translator's choices. To identify the natural translation unit, we can make use of semantic, syntactic and punctuation clues in the ST and TT. In the case that there are two possible translation units (e.g. phrase and clause), the larger one will be chosen for analysis for the sake of convenience. It should be noted that the unit of shift is different from the unit of comparison (or the unit of translation). A shift may occur on different levels and in different aspects of the language. It is not rare that more than one shift exists within one comparative unit, and a particular textual change serves as evidence of several shifts.

The classification of shifts is simplified and adapted to the characteristics of the Chinese language. The basic categories include semantic shifts, stylistic shifts and syntactic shifts. Semantic shifts are divided into four groups: specification, generalization, analogy and radical change. Specification has two subcategories, i.e. explicitation and supplementation, which represent different degrees of change: the former is based on the original context or world knowledge and the translator actually does not add new information to the original; whereas in the case of the latter, the translator expands the original text to a controlled extent by supplementing minor details. Generalization, the reverse instance of specification, also has two subcategories, termed here implicitation and condensation. Analogy refers to the case that the TT approximates the ST in primary language function, while they may differ in literal meaning. This type of shift often takes the form of paraphrasing. Radical change is gross deviation from the intended information conveyed by the ST. According to the form of change, it is subcategorized as addition, deletion and substitution.

Concerning stylistic shifts, we believe that a language characterized by generality or neutrality also represents a type of style, which is of equal status with those more "distinctive" types. Therefore, unlike the case of the transeme model, there is no "specification" or "generalization" of stylistic meaning in our classification. A stylistic shift always means a change from one type of variation to another. Revision is also made to the classification of factors that give rise to stylistic shifts. Of the seven types suggested by the transeme model, the syntagmatic element and paradigmatic element both concern the use of rhetorical device. They are thus combined as one element, i.e. the rhetorical element. For shifts in the rhetorical element, a convenient classification can be made between two groups. If the ST information is rendered with extra rhetorical device, the shift is taken as one of the subcategory "increase of expressiveness." In a reverse case, if the original rhetorical element is abandoned in the translation, the shift will fall into the range of "decrease of expressiveness." The attitudinal element is one added to the list of the transeme model. It reflects the language user's feeling or attitude, accounting for various attitudinal meaning, which can be objective, subjective, neutral, derogatory, commendatory, sarcastic, etc. The cultural element is at the root of two contrasting styles. One is characterized by expressions with exotic coloring, termed "exotic style," while the other, "natural style" is marked by transparent, idiomatic expressions. A shift from exotic style to natural style is called naturalization, and the other way round, exotization. The rest four elements, i.e. register, profession, time, and text-type, remain constant in the model to be adopted here.

As to syntactic shifts, due to the significant difference between Chinese and English in grammatical features, some categories of shift are actually obligatory shifts, and still some can be subsumed under the same heading. In the revised model, syntactic shifts are of two major manifestations: shifts in sentence type and shifts in complexity. The former is similar to syntactic-pragmatic shifts which deal with alterations in speech act. Shifts in complexity are changes toward either increase or decrease of complexity.

Focusing on the discourse level, the simplified model analyzes macrostructural shifts from the aspects of mind style, the image of the narrator and the syntactic order. Mind style is a cumulative effect, or a tendency, of focalization, addressing how the fictional world is presented. It can be approached from two perspectives. One is from the object it targets, namely the focalized. The other is from the subject who describes the fictional world, namely the focalizer. Obviously, mind style is closely associated with the ideational function, which determines the focalized, the information about the fictional world, and the interpersonal function, which underlies the focalizer's point of view on the fictional world. As the ideational function and interpersonal function are both realized through specific words, different mind styles may result from different word choices. To be specific, if the words used to describe the fictional world are merely general terms, the mind style will be general. In a contrary case, if the focalizer chooses physical and concrete words, the mind style will be physical and concrete. From the perspective of the focalizer, if the fictional world is described in a non-committal and distant manner, there will be an objective, neutral mind style. Whereas, if the focalization is tinged with personal feelings or opinions, the mind style will be subjective or emotionally charged. It should be noted that there are different types of mind style, representing varied ranges of influence. In this study, although we are mainly concerned with mind style based on the whole fiction, we will also examine mind styles related to some local stylistic effects, for instance, in the description of a character's speech or appearance.

The narrator is "the one who narrates, as inscribed in the text" (Prince 2003: 66). Basically, the image of a narrator is determined by four parameters. The first is the position of the narrator. If the position is outside the fictional world, that is, someone narrates the story from an "above" or superior level, the narrator is an external narrator; if the narrator's position is inside the fictional world, that is, one of the characters takes up the role of narrating, then the narrator is identified as an internal narrator. The second is the degree of perceptibility of the narrator. This is determined by the narrator's involvement in the story. With this criterion, a distinction is made between a covert narrator and an overt narrator. The former is a non-intrusive one, "presenting situations and events with a minimum amount of narratorial mediation" (Prince 2003: 17). The latter is intrusive, with various signs of showing his existence, such as report (indirect discourse) and commentary. The third parameter is the extent of knowledgeability of the narrator. If a narrator knows (almost) everything the situations and events recounted, he is an omniscient narrator (op. cit.: 68). Conversely, if the narrator only has a restricted knowledge about the situations and events, he is a non-omniscient or restricted narrator. The fourth contributing factor of the narrator image is distance, which refers to the metaphorical space between the narrator, characters and events narrated, and the reader (op. cit.: 22). It can be temporal, intellectual, cultural, moral, and so on. Since the narrator serves as the intermediary between the fictional world and reader, the distance we look at is practically that between the narrator and fictional world, and that between the narrator and reader.

Syntactic order is related with the textual function of language. Following the transeme model, the revised model will approach it from two respects. One is segmentation, namely, "the way in which the information is divided into sentences, clauses and phrases" (Leuven-Zwart 1989: 178). With the idea of Leech and Short, segmentation can be specified as three types: chronological segmentation, psychological segmentation and presentational segmentation (1981: 236). The first one means that sentences are organized in a way similar to the sequence of events taking place in the fictional world. The second one is based on the principle of "first is most important" (ibid.). It represents a natural order in which things or thoughts occur in one's mind (Leuven-Zwart 1989: 178). The last one follows the principle of climax, in which the most important information is retained until the last part of the sentence. The other respect of syntactic order concerns cohesion of the text, which is realized through such devices as reference, substitution, ellipse, conjunction and lexical cohesion (Bloor and Bloor 2004: 93-99). Thus, to find out shifts in cohesion, we need to pay attention to shifts concerning the sentence structure and word choices.

3. Describing Shifts in The Golden Cangue

In this part, we will look into both the micro- and macro-shifts manifested in Eileen Chang's translation The Golden Cangue. Before we proceed to detailed discussion, one point needs to be clarified. In our analysis, statistic results would be resorted to for identification of varying prominence of major types of shift. Inevitably, there will be some deviation between the measured value and the true value. However, we can be content with what we have obtained since we only aims at general tendencies of shifts, which does not require precise statistics.

3.1 Shifts in the Micro Structure

Regarding semantic shifts, of the 190 samples collected, 93 fall into the range of specification, taking up nearly a half of the total. Second to specification in number is generalization, with 60 samples which amounts to 32% of the total. Of the rest, 29 samples go to the category of analogy, and 7 to the category of radical change. Among those sub-categories, the most conspicuous is explicitation, which incorporates 81 samples. That is to say, about 43% of semantic shifts take the form of explicitation. Less conspicuous but still notable is the subcategory of condensation, whose 49 samples constitutes one quarter of the whole. Supplementation and implicitation come very close in proportion, about 6 %. The least noticeable types of shift are the subcategories under radical change, each with a tiny proportion of no more than 2 %. On the whole, what the statistic results reveal is consistent with some general expectations about translation. The prominence of the type of specification proves again the "explicitation hypothesis" which has been gaining ground in recent translation research (Blum-Kulka 1986, Klaudy 1998, Olohan and Baker 2000, Heltal 2005, Pym 2005). In addition, although there are different degrees of change, translator's alterations to the original are primarily controlled and context-based, as can be seen from the dominancy of "minor shifts" (including explicitation, implicitation and analogy).

Looking into shifts of each sub-category, we notice several interesting points about the distribution of different types of shifts. First of all, explicitation, the most prominent type of shift, often takes place in translation of culture-specific expressions. To be sure, The Golden Cangue, being a story of typical Chinese characteristics, contains many culture-bound terms, which would make little sense to most of the English readers without additional explanation. More often than not, the translator has to make clear those terms by supplying specific detail, even in the case of colloquial conversations which are supposed to be kept simple and short. For instance, the underlined part in ST2 is about customs of old-fashioned Chinese wedding ceremony. The translator adds relevant explanation to the original so as to facilitate the target reader's comprehension.

The translator also tends to make explicit the implicature of the character's language. This kind of explicitation usually occurs in expressions implying criticism or sarcasm. And interestingly, it mainly concerns the young Ch'i-ch'iao's speech. For example, seeing Chiang Chi-tse is not tempted by her, Ch'i-ch'iao sneers at the dissolute hypocrite:

As shown in the example, buzaihu (not care) in ST3 can be used to describe the relationship between friends, family members or lovers. It is thus vague and ambiguous. But the corresponding translation in TT3 is quite straightforward. The supplemented word "having" has explicitated what is implied in the original.

Shifts of generalization are marked by deletion of minor details of the subject. In the translation of Jinsuo Ji, two types of detail are liable to be omitted. One is the tag clause of direct quotation in conversations, such as Ta xiao dao (she said smiling). As explained in the editor's note of the translation, this is because the writer Eileen Chang is "a dedicated student" of traditional Chinese fiction, where the tag clause is always added before reported speeches (Chang 1971: 139). But as a translator, Chang has to take into consideration the target reader's expectation, and probably for this reason she reduces the frequency of the tag clause. The other kind of detail often sacrificed occurs in description of an act or movement. The translator often omits adverbials indicating a stable frequency of repetition, such as yixiang (always), chuchu (everywhere) yilu (all the way), zhaochang (as usual) and zhengtian (all day long). It seems that the translator prefers to make up for the omitted meaning by means of tense, as illustrated by the example below.

With a total of 213 samples, stylistic shifts stand as the most pervasive type of shifts in the translation under discussion. According to the frequency of shift, the seven subcategories can be divided into three hierarchies. On the top with the highest frequency is the type of shift in the rhetorical element, which is about 35% of the total. The middle level is composed of three groups, i.e. shifts in the register element, attitudinal element and cultural element, the percentages of which range from 20 to 18. At the bottom level are shifts in the temporal element, professional element and text-specific element, with a total share of 10%. It is also noticeable that shifts in the rhetorical element and cultural element are both toward one direction: the former is represented by decrease of expressiveness, and the latter, by exotization.

The statistic results have shown clearly that most conspicuous in shifts of stylistic meaning is the type decrease of expressiveness, a change toward flatter or more general style of expressing. The ST has made good use of the repertoire of the Chinese language, abundant in idioms, sayings, four-word formulations, etc. The diversified style is most remarkable in description of character's language (language of discontent or grudge in particular), which is essential for creation of "solidly and frighteningly real" characters (Hsia, 1971: 397). But the TT, as the majority of shifts suggests, cannot always reproduce similar expressive effects of the original. Notably, stylistic shifts are inclined to take place when the four-word formulation is involved in translation. The excerpt below is a case in point.

ST5 features two four-word formulations, nanhun-nujia and tingtian-youming, each composed of two synonymous and symmetric phrases. For the former, the translator conveys the referential meaning only, as there is no corresponding pair of synonyms for hun (a man's marriage) and jia (a woman's marriage) in English; for the latter, the literal meaning is retained, yet the original conciseness, rhythmic effect and cliché-like stance can by no means be reproduced.

Also, in treating metaphors from sayings or conventional usages, the translator would usually retain the tenor, while give up the vehicle. This tendency, different from the case of the four-word formulation where shifts are largely due to linguistic constraints, seems to be caused by a consideration for the reading habit of the target pole, and perhaps the translator's linguistic incompetence as well. Consider the following example.

The metaphor in ST6 na damaozi ya ren is based on conventional usage, which is obviously our of the frame of reference of the target reader. A literal translation of the metaphor would only sound odd and confusing for the English reader. While an English analogue is elusive, it is certainly safer for the translator to render only the conceptual meaning. Hence comes the flatter composition of TT6.

It should be noted that accompanied with the two tendencies mentioned above are often shifts in the register element. In the first tendency, the literariness residing in most of the four-word formulation is reduced with the "deconstruction" of the original form in translation, such as the translation of xinling-bihuan (p. 232) as "declined his valuable offer" (p. 187) and yaque-wusheng (p. 213) as "became still" (p.142). Whereas, in the second tendency, the colloquial quality of some conversational clichés or conventional usages would be raised to a more standard level, such as bu shang taipan (p. 229) rendered as "not used to company" (p. 179) and shangtou-shanglian (p. 214) as "impertinent" (p. 145). As a result, the TT would be neither too literary, nor too colloquial in terms of formality. Apparently, shifts in the register element are toward the average, standard level.

For shifts in the attitudinal element, the translated fiction under discussion exhibits a variety of sub-groups, including shifts from derogatory or ironic to neutral, from uncertain to definite, from respectful to neutral, from intimate to distant, from neutral to negative, etc. But the main stream can be summarized as alteration from emotionally charged to neutral and objective. Compare, for instance, the following source and target excerpts:

In ST7, the narrator obviously holds a negative, critical attitude towards the character mentioned, as shown by the derogatory phrase zizuo-zhuzhang (to act on one's own without prior permission from relevant parties) and the adverb jing, which is used to indicate surprise or unexpectedness. Yet in TT7, the word choices sound quite objective, without any negative implication. As such, the narration is more like a matter-of-fact reporting than a statement of blame.

Finally, regarding the cultural element, the shifts are one-directional, as there is only the type of exotization observed. As we have mentioned, exotization is a change from a style characterized by naturalness or idiomaticity to one tinged by foreignness. Foreignness in our study is manifested in the use of foreign words or unusual, bizarre expressions. The former is the case of transliteration (usually with further explanation). For example, the Chinese dish Yipin guo (p. 233) is put as "i-p'in-kuo," followed by a note "the highest ranking pot" (p. 189). The latter, as one may expect, results from the strategy of literal translation. It seems that apart from catering to the target reader (as indicated by the tendency decrease of expressiveness), the translator also takes into consideration the idiosyncrasy of the source culture. Thus, some culture-specific expressions are rendered literally, especially when their meaning can be derived from the situation or context, as illustrated by the underlined part in the example below.

It will not be difficult for the reader to figure out the action of a playboy like the character Ch'ang-pai, especially when he is on bad terms with his wife. In addition, following the excerpt is the episode that the character's mother gives him a concubine in order to "hold him" (p. 177). The translator's literal rendition, therefore, is understandable though the metaphor involved is out of the target reader's frame of reference.

In the case of syntactic shifts, there are altogether 90 samples collected. A considerable difference in distribution can be observed between the two subcategories. Shifts in sentence type are in the minority, which constitute only 21% of the whole. Shifts in complexity take away a dominant percentage of 79, of which 62% suggests an increase of complexity, and 17% a decrease of complexity.

The statistics tells clear that increase of complexity is the governing type of change in syntactic shifts. With a close look, we notice that long sentences tend to occur in descriptions of environments (both indoor and outdoor) and descriptions of the external appearance of a character. In the following example, ST9 depicts emphatically and economically the character's features with four nominal groups. But the conciseness is totally lost in the translation, as TT9 adopts the form of complete sentence and rearranges the way of presentation.

Another trend concerns description of actions. Usually, the ST uses several coordinate minor sentences to indicate successive movements. The TT, rather than following the original "chronic style," would render some of the verbs that function as predicate in the ST as adjectives or adverbs. In this way, sentence components of the TT would have more varieties, and thus the TT sentence would show greater complexity. This kind of shift can be illustrated by this example:

3.2 Shifts in the Macro Structure

As we mentioned in the explanation of the revised model, mind style can be specified from different facets of the fictional world. To put it simple, the fictional world is about "when," "where," "who," and "what happens." The first is the time, the historical setting of the story; the second is the place, the social and geographical environments of the story; the third is the character; the last one is the event. To find out general tendencies about these aspects, we shall look into semantic shifts and stylistic shifts, which reflect alteration in linguistic meaning and stylistic meaning respectively.

In semantic shifts, there are both the tendency of specification and generalization, but as the former overwhelms the latter in terms of the quantity of shifts, the image of the fictional world presented in the translation is more concrete than that in the original. That shifts of explicitation occur in speeches of the protagonist in her prime seems to indicate a change in characterization. We notice that when Ch'i-ch'iao is a daughter-in-law, she usually uses implicit language that is not so tart and mean; with her position rising, her language becomes more and more bitter and cutting. The contrast in language implies the gradual degeneration of the character, a significant aspect of the theme of the fiction. In this point of view, explicitation of the character's implicit language in effect obscures the two stages in the development of character's personality, which inevitably impairs characterization of the fiction. Thus the image of the character is not as natural as that in the original.

On the part of stylistic shifts, the translation employs both free and literal translation strategies. In the case of the former, the TT sacrifices the original vitality, conciseness and fluent rhythm that underlie those four-word formulations and conventional usages of the ST on the one hand; on the other hand, because of language constraints, it can by no means reach the transparency of the original despite its accommodating policy. Accordingly, the image of the fictional world presented to the reader by the translator carries less distinctive stylistic features, hence less zest and savor. The character's language and the narrator's languages in particular loses the varieties that indicate different emotions and pertain to different personalities and the particular temporal and spatial settings of the story. The fictional world thus appears to be less colorful and less emotionally charged. In the case of literal translation, exotization is the evident product, which, while adding some "local" color to the translated fiction, demands more efforts in capturing the meaning. For the ST reader, the fiction, with all its naturalness and idiomaticy, is an enjoyable reading. Whereas, in the eye of the TT reader, it is more or less a strange or even mysterious picture of old China with its typical myths such as arranged marriage, opium-smoking, complex personal relationships, bounding feet, concubines and various bizarre expressions. The image of the fictional world is then one featured with special alien favor, which adds to the appeal of the story and yet also imposes extra difficulty in comprehension.

Concerning the image of the narrator, the narrator of the story of Jinsuo Ji is an omniscient narrator, who looks from outside at the fictional world and tells the story. The interpersonal function of narrative fiction thus concerns the narrator-reader relation and narrator-character relation. According to theories of narratology, the more the narrator reveals himself in the fictional world, the closer he is to the character while the farther to the reader (Leuven-Zwart 1989: 176). In the case of The Golden Cangue, the narrator's involvement into the story is less than that of the original. As a result of the neutralizing tendency in shifts of attitudinal meaning, the TT brings forth a less "enthusiastic," and sometimes non-committing, narrator who narrates with less subjective language and thus has more proximity to the reader's world. In terms of perceptivity, the TT narrator is not as overt as the ST narrator whose narration is under heavy influence of personal judgments. And as such, the narrator also seems to be less knowledgeable that that of the ST.

In semantic generalization about the character's speech and thoughts, the translator often omits the tag clause that indicates a quotation, which may result in a shift of focalizer--from character focalizer to narrator focalizer. Words originally from a character are changed into the narrator's narration, which inevitably affects the reliability of the words and hence reader's judgment. In other words, the shifts have adjusted the narrative slightly away from the traditional story-telling. The narrator in this context can be called the "showing" narrator, which contrasts the "telling" type.

From the aspect of distance, since there are a considerable number of shifts related to the cultural element, we can infer that while the narrator in the ST shares the same culture with the reader and the fictional world, the narrator in the TT keeps a remarkable cultural distance with his reader.

Shifts in the syntactic order can be analyzed from segmentation and cohesion. For segmentation, syntactic-semantic shifts have shown that in describing successive actions or movements, the translator tends to render some verbs that function as predicates in the ST in the form of adjective or adverb. This tendency has, in fact, altered one aspect of the original segmentation. In the ST, a continuum of movements being described represents a chronological segmentation, that is, following the natural chronological order in narration. But the way of rearranging sentence components in the TT often means another type of segmentation--psychological segmentation, which follows the principle of "first is most important" (Leech and Short, 1981: 236). Therefore, in syntactic order, part of the TT undergoes the change from chronological segmentation to psychological segmentation.

As for the aspect of cohesion, all of the three basic types of shifts may exert some influence. In semantic shifts, as specification is more prominent that generalization and more function words are used in the translation, it seems that the TT ought to be more cohesive that the ST. Nevertheless, because of shifts in stylistic and syntactic meaning, the TT expressions are tinged with strong foreign flavor that demands greater efforts in comprehension, and its sentences are generally longer and more complex than those in the ST. These reduce considerably the textual cohesion of the TT. On the part of the ST, with idiomatic expressions familiar to the reader and concise, neatly organized sentence patterns, the language of the text will by no means become an obstacle of reading. The contrast between the TT and TT has shown clearly that the TT is doubtless less cohesive that the ST.

3.3 Summary

In summary, it is shown that on lower linguistic level, shifts of explicitation and decrease of expressiveness are most conspicuous; second to them is the increase of complexity of the sentence structure. Of various elements contributing to stylistic shifts, in addition to the rhetoric element, the register element, cultural element and attitudinal element play a significant role in the makeup of the translated text.

For shifts in the macrostructure, the analytical results suggest that the fictional world is presented in the translation with greater detail, and the mind style of the TT on the whole is more concrete and less subjective; the narrator in the translation is an external narrator, less involved and less knowledgeable than the one in the ST. Also, there is considerable cultural distance between the narrator, reader and fictional world in the translated fiction. In syntactic order, there are shifts from chronological segmentation to presentational segmentation, and the translation, with more information provided, more complex sentence patterns and strong foreign coloring, is found less cohesive than the original text in textual structure.

4. Conclusion

This paper has made a detailed comparative study on Eileen Chang's renowned novella Jinsuo Ji and its English translation The Golden Cangue. Focusing on textual relationships between the ST and TT, it remolds Leuven-Zwart's transeme model and discusses in detail translation shifts in both the micro and macro structures.

Viewed as a whole, Eileen Chang's treatment of her own work is controlled by a reader-oriented principle. To facilitate the reader's understanding, Chang adopts a common, standard language for expressing, supplements relevant background information for culture-bound words, simplifies or removes some metaphoric expressions exotic to the English reader, and rearranges the syntactic order of presentation. These account for the explicitating and flattening shift tendencies. On the other hand, it should be noted that despite the translator's wish for an easy reading, the comprehensibility of the translation is restricted by the translator's proficiency in English and the content of the story, as suggested by the rise of increase of complexity in sentence structure and the overall foreign coloring of the translation.