Shifts of agency in translation: a case study of the Chinese translation of Wild Swans

Li, Long

doi:https://doi.org/10.7202/1073641ar

1. Introduction

Translation studies (TS) in the 1990s saw the emergence of two approaches (Munday 2001/2016: 142, 198): the cultural turn (Bassnett and Lefevere 1990; Spivak 1993/2000) and discourse analysis. Criticizing the failure of previous linguistic turns to consider translation within its cultural environment (Bassnett and Lefevere 1990), the cultural turn has generated new interest towards pertinent cultural issues, such as ideology. However, text-based scholars have criticized purely cultural approaches for their lack of rigorous research methods (Baker 1996: 9; Boyden 2008: 151; Fawcett 1995: 182) and their low potential for generalization (House 2013: 50). This paper calls for a more constructive dialogue between the two approaches. Linguistic methodologies, especially those based on systemic functional linguistics (SFL), can be an empowering tool for translation scholars to form a bridge between discourse (or a text) and its situational and cultural context. Among the various SFL systems, one that is particularly relevant to the “translation of ideology” (Hatim and Mason 1997/2005: 119) is transitivity[1]: the whole of lexico-grammatical resources that “represent reality in language” (Eggins 2001/2005: 206). Transitivity has been usefully deployed by a number of translation scholars (Mason 2003/2004; Munday 2007; Huang 2013) and notably by Calzada-Peréz (2007). The seminal work of Calzada-Peréz (2007), which inspires the present study, usefully deploys transitivity in relation to causation and voice for a corpus of European Parliament speeches and translations. However, while transitivity continues to attract new interest in discourse analysis (Chaemsaithong and Kim 2018; Dreyfus 2017; Lukin 2013), it has not drawn much recent attention in translation studies, except for Yu and Wu’s studies (2016, 2018) of the English translation of Chinese Zen Buddhist texts and Chinese government reports.

This study calls for a return to transitivity analysis as an empowering tool for descriptive translation studies, especially those with an interest in uncovering how ideology and the attribution of responsibility can be represented linguistically in translation. A key component of ideology is a differentiation of ‘us’ versus ‘them’ (Thompson 1990), by representing an event from a particular perspective and by attributing responsibilities to different characters with particular consistencies. This function in this context needs to be achieved through patterns of meaning, which must be linguistically realised chiefly through the lexicogrammatical resources of transitivity. This study deploys the transitivity system to compare the characterisation between a source text (ST) and a target text (TT) and the semantic consequences on the overall narrative. It also strives for methodological innovation by combining the standard SFL transitivity system with other semantic and lexicogrammatical systems, including voice, theme, and modality. This combination of analyses enables a better understanding of the complexity of the interactions amongst different linguistic systems that take place in interlingual translation in order to differentiate obligatory and optional translation choices.

The data for this qualitative case study comes from two chapters in Jung Chang’s politically volatile autobiography Wild Swans[2] (and its Chinese translation[3] by Pu Zhang). Using the abovementioned methodology, this paper examines the shifts of responsibility attributed to the controversial Chinese leader Mao Zedong (Chairman Mao), with a hypothesis that transitivity and related linguistic systems may have been manipulated by the TT producer(s), whether consciously or subconsciously, which results in changes in meaning that alter characterization and narrative in the translation. Based on these analyses, this paper seeks to uncover the semantic and ideological consequences of the choices made from different systems. In addition, it sets out to propose a categorisation of the equivalences and shifts in the translation of agency.

2. Transitivity and other interacting linguistic systems in Systemic Functional Linguistics (SFL)

2.1. Transitivity within the ideational metafunction

Transitivity is “modelled as a configuration of a process, participants involved in it, and attendant circumstances” (Martin, Matthiessen, et al. 2010: 98). To put it simply, it involves how we use language to represent who does what to whom/what, where, when, how, and why (Hasan 1988: 63). A full transference of such happenings and goings-on can be considered the core of what is usually expected of translation accuracy. Thus, a translation that does not match the overall meaning of the ST created through transitivity may not be considered a translation at all; instead, it shall be considered, to borrow Bassnett and Lefevere’s (1990) term, a type of rewriting. On the other hand, minor shifts in transitivity may also have an accumulated effect which significantly alters the ideology of attributing responsibilities or blaming.

Transitivity is aligned with the representational function of language within an SFL framework. Halliday developed a tri-functional view on language based on the works of scholars such as Malinowski, Bühler, and Jakobson (Halliday and Hasan 1985: 15-17). This functional view argues that language is what it is because of the functions it must achieve in the physical and cultural environments: 1) to construe and represent human experience – the ideational metafunction, 2) to enact personal and social relationships with those around us – the interpersonal metafunction, and 3) to turn the first two metafunctions into coherent and cohesive discourse – the textual metafunction (Halliday and Hasan 1985). The three metafunctions serve as a link between language and the context outside language (Thibault 1987: 608). This functional alignment of language and its context is the reason why SFL is particularly useful in explaining translation choices within their context. Transitivity is the major lexicogrammatical system[4] within the experiential metafunction, which is one component of the ideational metafunction, alongside the logical metafunction. The basic analytical unit of lexicogrammar is the clause, which is the smallest unit onto which the three strands of meaning (ideational, interpersonal, and textual) are mapped.

Since transitivity concerns the linguistic representation of who does what to whom/what, where, when, how, and why, a clause with a full transitivity pattern can be divided into the who/whom (participant[s]), the what (process), and the where/when/how/why (circumstance), as shown in Table 1.

Table 1

**Functional labels in transitivity**

Transitivity can be analysed from two perspectives: an ergative model and a transitive model. On the one hand, the ergative model generalises across our experiences (Halliday and Matthiessen 2014: 334) so that only four participant roles are recognised: Agent, Medium, Beneficiary, and Range. The participant through which a process unfolds is called Medium. But in some clauses, another participant is recognised as the external cause of the process, or the Agent. A clause with a feature of agency is considered an “effective” clause in ergative terms, while one without an Agent is considered a “middle” clause. Such distinction between agentive and agentless clauses exists in both English and Chinese. Dreyfus (2017) exemplifies three types of effective clauses and one type of middle clause in English, as shown in Table 2.

Table 2

**Effective and middle clauses in the ergative model of English transitivity (adapted from Dreyfus 2017: 376-379)**

The ergative model is more useful for analysing the shifts of agency and responsibility in translation, since the person or organisation with agency is highly likely to be represented linguistically as Agent in ergativity; otherwise, the agency would be so obscure that readers may not be able to link the character’s agency and the goings-on.

On the other hand, the complementary transitive model has been the source of the traditional distinction between transitive and intransitive verbs in English (Halliday and Matthiessen 2014: 216). In this model, a process is either confined to one participant (“intransitive” clause), or it extends from one participant to another (“transitive” clause). This paper argues that it is useful to combine the ergative analysis of agency with the transitive analysis, because the transitive model particularises the linguistic representation of human experiences, as shown in Figure 1 for English, and thus provides a more delicate framework for different types of Agent. This further differentiation depends on the type of process for which the Agent acts as an external force.

Figure 1

**The grammar of experience: types of process in English (Halliday and Matthiessen 2014: 216)**

As shown in Figure 1, there are three main types of experiences: 1) the outer experience of doing and happening (mainly “material”), 2) the inner experience of feeling (mainly “mental”), and 3) the relationship of one experience to another (mainly “relational”). Mason (2003/2004) usefully argues that it is best to maintain the same process type to translate an experience accurately. For instance, it is generally best to translate an English material process into a Chinese material process to ensure the same experience is construed in translation, even though this is not always possible due to linguistic differences. However, a non-obligatory shift from a “material” process (the government stabilized the economy) into a “relational” process (the economy was stable again) would convey a significantly different experience by stripping the government’s agency in stabilising the economy.

The participants are labelled differently in each of the six processes depending on the grammatical relationship between the participant and the process, for example Actor in a material process and Sayer in a mental process. Likewise, Agents can be further labelled in six different ways depending on the transitive process type, as Table 3 shows below.

Table 3

**Participant types that are considered agentive in the ergative model (adapted from Halliday and Matthiessen 2014: 344)**

2.2. Investigating characterization and narrative through transitivity

Transitivity analysis allows the different participant types of a character to be identified and labelled in the ergative or the transitive model, which contributes to a better understanding of how a character has been represented and of the overall narrative. The popular term narrative, or story, has been defined by Ewick and Silbey (1995: 198) (in Baker 2006: 19) as “a sequence of statements connected by both a temporal and a moral ordering.” A core feature of narratives is causal emplotment, which “allows us to take the same set of events and weave them into very different moral stories” (Somers 1997: 82, cited by Baker 2006: 67). Baker (2006: 70) argues that “patterns of causal emplotment can thus be subtly changed in translation through the cumulative effective of relatively minor shifts that lend a different weighting to the elements of the original narrative.” Whereas she has usefully identified a number of manifestations of casual emplotment, the obvious contribution of transitivity analyses to this aspect of translation has not been explicitly recognised.

Character has been usefully defined by Fisher (1997: 316, cited by Baker 2006) as “an organized set of actional tendencies.” As Hasan (1988: 65) argues, “…part of the basis of our perception of what a person is like derives from knowing what sort of participant roles are ascribed to that person.” Therefore, consistent shifts in the ergative and transitive roles may result in an overall change in the characterization.

2.3. Agency, voice, and theme

Agency is closely related to the grammatical systems of voice (Calzada-Peréz 2007; Dreyfus 2017), because an Agent can only appear in clauses with certain voice choices and its unmarked position in a clause is also relatively fixed depending on the voice. Viewed from the ergative perspective, two voices are recognised in English. One is the effective voice, in which an Agent extends to a Medium; it includes operative (active) and receptive (passive) voices. A clause in the receptive voice may include or omit the Agent, but the voice in both cases is nevertheless effective due to an Agent being implied by the passive verb form. The other is the middle voice, in which there is no Agent grammatically present or implied. These have been exemplified in Table 2 above. The same distinction between agentive and agentless voices has been recognised in Chinese (Halliday and McDonald 2004: 376), although it is based on different grammatical markers.

Agency is also closely related to theme (Li 2007; Kim and Matthiessen 2015), because of the interrelation between Theme choice and voice choice. Theme is the main system in the textual metafunction, and Theme is defined by Halliday and Matthiessen (2014: 89) as “the element that serves as the point of departure of the message”; whatever is considered a Theme is given textual prominence as the starting point. Then the message progresses to the non-prominent rest, which is referred to as the Rheme. On a text level, meaningful Theme choices across clauses, or thematic progression, can usefully orient the readers to follow the flow of the message in a coherent way. Thematic progression is the “successive selection of themes from one clause to another according to one of a fairly small number of motivated patterns” (Kim and Matthiessen 2015: 337-338), which mainly includes a linear pattern (or thematic reiteration), in which the same element frequently occurs as Theme, and a zig-zag pattern, in which the Rheme, or parts of it, is regularly picked up in the Theme in the following clause (Eggins 2001/2005: 324-325).

Kim and Matthiessen (2015: 341) point out that the voice system may considerably differ, even between two closely related European languages. For instance, the passive voice is more frequent in English than some other languages, such as German. Therefore, when interpreting translation shifts in theme and voice, it is important to consider typological differences in voice between the two languages, which may impose restrictions on the Theme choice in translation. On the other hand, restrictions of Theme choices in translation may also result in a preferable or inevitable change of voice.

As shown above, English has three choices within its system of voice: active, passive, and middle. According to Halliday and McDonald (2004), Chinese also has three voices, though they are different from those in English: dispositive, passive, and neutral. Dispositive, which does not have an equivalent in English, is realised using the construction 把 (bǎ), as shown in Table 4. Passive is realised using the constructions 被 (bèi) or 给 (gěi), again exemplified in Table 4. Any other voice that does not have such formal grammatical markers is considered neutral in Chinese, which could be either transitive or intransitive. Some instances of neutral transitive voice show features of agency and may appear similar to an English active voice, and the neutral intransitive voice may appear similar to the English passive, but they are nevertheless considered neutral due to the lack of an explicit grammatical marker. A main reason why the Chinese voice system has to be recognised with such explicit markers by grammarians is that Chinese verb forms do not change according to voice as do English verbs (that is, be + verb -d in English passivity).

Table 4

**The Chinese voice system (tabulated from Halliday and McDonald 2004: 372)**

The voice systems in English and in Chinese differ not only in the structure (namely, an Agent generally precedes the process in the Chinese passive bèi or gěi structure, but follows the process in English passivity), but also in the interpersonal meaning: a passive structure in spoken Chinese is semantically loaded with an undesirable feature, although such a negative connotation may not apply in the formal written register (Halliday and McDonald 2004). Hence, it is likely that an English passive structure with positive connotation may be translated into an active voice in Chinese to avoid the negative connotation in the spoken mode.

The three systems of agency, voice, and theme are interrelated, because a change in voice (that is, from active to passive) typically results in a change in Theme and the position or even the presence of the Agent, as exemplified in Table 5.

Table 5

**Interrelation amongst agency, voice, and theme**

This interrelation raises one implication for TS: change in voice in translation may be motivated either by a stylistic need to ensure a smooth thematic progression (that is, to consistently select the pot as the point of departure) or by ideological reasons to vary the attribution of responsibility to an Actor (the one who broke the pot).

2.4. Cline of responsibility

Based on the combinatorial analysis of agency, voice, and theme, Dreyfus (2017) proposes a cline of responsibility in English, as shown in Figure 2.

Figure 2

**Cline of responsibility (Dreyfus 2017: 379)**

Using this cline, Dreyfus (2017) shows that even her three-year-old son was capable of manipulating transitivity resources to construe the same reality, but one which minimises any responsibility attributed to himself as the do-er by saying The pot broke. Although the active voice and the passive voice (with Agent) both explicitly assign responsibility to the do-er, Dreyfus (2017) further argues that the active voice assigns higher responsibility by placing the do-er I in the initial, thematic position in the clause, therefore foregrounding the agentive role with textual prominence. In comparison, the agency of the do-er can only be inferred in the agentless passive voice, as it is not explicitly represented. Further down the cline, the middle voice completely leaves out any responsibility as if the pot broke inevitably, without any external force. In the case of Dreyfus’s three-year-old son, she argues that he chose the middle voice for ideological reasons, that is, to minimise his role in the breaking of the pot. Correspondingly, Dreyfus shows that her son chose the active voice when he wanted to foreground his role in events, such as when he built a train (I made a train).

It is hereby argued that both the dispositive voice and the neutral transitive voice in their unmarked forms[5] are agentive and equivalent to the English active voice in maximising responsibility, and the Chinese intransitive neutral voice is the equivalent of the English middle voice in minimising responsibility. Taking into consideration the typological differences, it is hereby proposed that Dreyfus’s (2017: 379) cline of responsibility can be adapted as shown in Figure 3 for a cross-linguistic comparison of the cline of responsibility between English and Chinese based on Theme choice and voice.

Figure 3

**The application of Dreyfus’s cline of responsibility to Chinese**

2.5. Agency and tense/modality

While agency can be enhanced or weakened through shifts in voice, it should also be seen as being diminished when an agentive process shows modality markers or non-finite predicators, both of which add an irrealis sense that the agency is not actually exercised (Butt 2008: 76-77). In Figure 4, the agency of he in destroying people must be seen as being the most realis in the clause with the primary past tense destroyed, showing it was an actual past event, and the second agency, marked by modality would, is lower as it is only an expression of his inclination to exercise agency. The clause with a non-finite predicator shows the lowest level of agency, as it only conveys a hypothetical condition.

Figure 4

**A cline of finiteness in representing agency**

Again, linguistic differences between a source language and a target language may have implications on examining the finiteness of agency in translation. For instance, since Chinese does not have a grammaticalized system of tense, only a partial comparison between English and Chinese, based on modality markers, is possible when it comes to agency in relation to finiteness. The differentiation between actualised and unactualised in Chinese will be based on whether a clause contains a modality marker such as would.

3. Data and methodology

3.1. Jung Chang’s Wild Swans

This paper sets out to compare the representations of Mao Zedong at the beginning of the Cultural Revolution in 1965-1966 in Jung Chang’s politically charged English-language autobiography Wild Swans and its Chinese translation published in Taiwan. The ST represents, to date, the summit of success ever achieved by a contemporary Chinese writer in the English language. Initially published in 1991, it went on to sell over 15 million copies in 37 languages as of 2018.[6] The ST was initially published in 1991, two years after the fall of the Berlin Wall and the Tiananmen Square Incident, both in 1989. At the time, the dissolution of the Soviet Union marked the official victory of U.S.-led NATO in the Cold War. This publication was timely so as to capture the keen Western interest in life under China’s communist regime. Despite its commercial success, Wild Swans has attracted a number of criticisms for its interpretation of Chinese history and the overwhelmingly negative portrayal of Mao (Chun 1992; Gao 1993; Gao 2002; Goodman 2006; McDougall 2014). For example, McDougall (2014: 57) comments that the success of Wild Swans “reinforces the impression that English-language readers welcome or at least don’t object to books that are critical of contemporary China.” Carrying this monstrous representation of Mao in Wild Swans further, Jung Chang and her Irish historian husband Jon Halliday published a biography of Mao in 2005, Mao: The Unknown Story, which has also attracted criticism from academics (Benton and Chun 2013; Goodman 2006) for its unbalanced representation of Mao.

Wild Swans was translated into Chinese by Chang’s brother, Pu Zhang[7], who is not a professional translator but rather a London-based Chinese language writer. As the ST author’s brother, he is a represented character in the autobiography. It is unclear why Chang did not translate the book herself, since she has subsequently demonstrated her translation ability by translating two of her later biographies, on Mao and on Empress Dowager Cixi, into Chinese. Nevertheless, Chang has personally contributed to the decision-making in translating Wild Swans, suggesting that she approves of the Chinese language edition. Also, the translator’s proximity to the ST author and his role as a represented character may have given him unusual liberty in translating the book, something ordinary translators could not afford to do.

The Chinese translation was published in Taiwan, which is claimed to harbour one of the freest media environments in Asia.[8] The time gap between the ST (1991) and TT (1992) is minimal, although the TT context is marked by the thawing of Cross-Strait relations since 1979. The years 1991 and 1992 saw the adoption of the 國家統一綱領 (Guójiā tǒngyī gānglǐng) [Guidelines for National Unification][9] in Taiwan and the so-called 九二共識 (Jiǔ’èr gòngshì) [1992 Consensus], both of which supported the concept of “One China” and were welcome by Beijing.

This qualitative study mainly focuses on chapters 15 and 16 of Wild Swans, which portrays the beginning of the Cultural Revolution (1965-1966); three examples from other chapters are included (in Table 3 and in Section 4.3.1) when a suitable example is unavailable in the two selected chapters. Prior to the Cultural Revolution, the 大躍進 (Dà yuèjìn) [Great Leap Forward] (1958-1961) had resulted in a great, 3-year famine. China had also suffered through internal fights against class enemies and through numerous other political campaigns in the name of preparing for wars against the U.S., Taiwan, and the USSR. This desperate desire to catch up with developed countries and to prepare for war was influenced by volatile global politics. The Cold War peaked in 1962 with the Cuban Missile Crisis. With the Sino-Soviet Split as of 1960, there were conceivable threats to China from both America and the USSR. On the one hand, Sino-Soviet relations remained sour after Mao criticized Khrushchev’s call for de-Stalinisation in 1958. On the other hand, the arrival of U.S. troops in Vietnam in 1965 created an immediate threat.

The two chosen chapters describe Mao launching the Cultural Revolution against capitalist roaders (people who sought to push China towards a capitalist road) and “bourgeois intellectuals” such as scholars and teachers. He did so with the assistance of officials such as the Gang of Four, but also by encouraging the Red Guards and young students to rebel. The ST author, Chang, was around 13 or 14 at the time, a privileged student in Chengdu whose father was the director of Public Affairs of the province of Sichuan (Chang 1991/2003). The beginning of the Cultural Revolution impacted China as a whole, including the author’s family.

The two chapters are considered suitable and unique materials for a study of agency in translation. In these chapters, the author explores the reasons why Mao launched the Revolution and how the youth were mobilised to wreak havoc. Thus, frequent linguistic attributions of responsibility are to be expected in both the ST and the TT. The size of the parallel corpus is shown in Table 6.

Sources and sizes of the ST and TT in the present study10 — Table 6

**Sources and sizes of the ST and TT in the present study[10]**

3.2 Data analysis

The representation of Mao is firstly explored impressionistically throughout the whole book and subsequently analysed through detailed transitivity analysis in the two selected chapters. All clauses in which Mao and other characters are represented as a participant in the chosen chapters were exported to an Excel sheet for a close analysis of the process type, transitive participant role, ergative role, modality marker, and voice. Instances were then compared between the ST and the TT to identify and categorise any equivalences and shifts in terms of the linguistic representation of agency. See for instance Example 1, whose analysis is carried out in Table 7.

Table 7

**An example of variables in data analysis**

All analyses were revisited several times to keep errors to a minimum. Instances that were ambiguous in process or participant types were discussed with colleagues in SFL.

These variables also allow for further quantitative analyses (Li and Li 2020) and qualitative analyses to compare the shifts of responsibilities amongst different characters in the translation. Within the scope of this study, the focus will be the translation of Mao’s agency through qualitative discussions of some typical shifts and equivalences in the translation.

4. The translation of agency: equivalences and shifts

4.1. The overall shifts of the representation of Mao in translation

An impressionistic reading of the full English ST shows that Mao Zedong is characterized as:

an almost exclusively negative character;
an emperor-like dictator exercising absolute authority;
one who could not take the slightest criticism;
a firm believer in violence and human struggle as being necessary for social development.

The representation of Mao in Wild Swans is almost completely negative. In addition, when describing negative consequences of policies and campaigns related to Mao, his agency is usually made explicit: for example, Mao let this happenin order to… (Chang 1991/2003: 363). To hold Mao more accountable, Mao’s mental activities are extensively explored. The mental activities are then represented as having turned into his action/speech, which subsequently led to the drastic and disastrous changes that swept across China. Those changes in turn had a direct impact on the Chang household. A narrative pattern has been frequently found to trace tragedies in the author’s family directly back to Mao’s mind: Mao “felt threatened” (Chang 1991/2003: 348)… therefore he started a witch-hunt (Chang 1991/2003: 348-349); “Mao’s vague battle calls threw the population and the majority of Party officials into profound confusion” (Chang 1991/2003: 350); Chang’s parents (party officials) were overwhelmed by apprehension (Chang 1991/2003: 350); Red Guards group sprang up around the country (Chang 1991/2003: 359); Chang’s father fell victim to the rage of the Red Guards (Chang 1991/2003: 382-384); Chang became “fearful and confused” (Chang 1991/2003: 385).

The Chinese TT shows significant quantitative shifts in the representation of Mao. A dramatic 77% reduction has been found (Li and Li 2020) in the number of instances of Mao being represented as a participant in the Chinese TT, from 153 to 77 in the same two chapters. Mao’s agentive roles have also been reduced from 37 instances to merely 20. This decrease in Mao’s overall representation and agency is foregrounded against an increasing trend of the overall representation and agency of the author’s peers – the youth of the day who committed the bulk of the actual violence during the Cultural Revolution. The following subsections will discuss the equivalences and shifts in translating Mao’s agentive roles in the Chinese TT.

4.2. Equivalence in the translation of agency

An instance of the translation of agency is considered an equivalence in this paper at the lexicogrammatical stratum when the character has been translated into the same participant at an equivalent level on the cline of responsibility in Figure 5. This almost inevitably requires the same process to be maintained. There are a number of instances in the data where the equivalence of Mao’s agency has been achieved on a clause level. For instance, in the example in Table 7 above, Mao’s agency as an Actor in “using” animate goals – the young students of the day – to incite violence has been translated as 是利用 (shì lìyòng) [be using], which represents an equivalent level of agency of acting on animate goals. As shown in Figure 5, the Chinese neutral voice (transitive) is considered one of the two equivalences of the English active voice. The elliptical 他 (tā) [he] in the Chinese clause does not affect Mao’s agency since it is recoverable from the one in the immediately preceding clause, 他也不特別喜歡或信任他的小紅衛兵 (Tā yě bú tèbié xǐhuān huò xìnrèn tā de xiǎo hóngwèibīng) [He also did not particularly like or trust his young red guards]. A distinction should be made between an agentless passive, in which an Agent is implied but not grammatically recoverable, and the operative voice, in which the Agent is recoverable grammatically. Elliptical reference is common in Chinese (Halliday and McDonald 2004; Hu 1994).

This equivalence in translating Mao’s agency may appear as common-sense to translators. However, there are potential alternative translation techniques that would reduce Mao’s agency, as shown in Table 8. This shift of Mao from an agentive Actor to an agentless Senser would lead to an omission of Mao’s role as an Actor.

Table 8

**An alternative translation technique**

Similar lexicogrammatical equivalences in translating Mao’s other agentive roles, such as Initiator (+ Animate Goal) and Sayer (targeting), have also been found, as listed in Table 9, based on Example 2.

Table 9

**An example of an equivalence in translating Mao’s agentive Initiator role**

In Table 9, Mao’s highly agentive role in initiating the population to act is translated as an equivalent Initiator role in Chinese. The translation also shows explicitation of this initiation as a verbal process in the additional embedded clause 按他說的 (àn tā shuō de) [as he said], which may have been influenced by a need for fluency in Chinese in translating the initiating process get…to act.

Example 3, analysed in Table 10, also shows how Mao’s agency as a Sayer is maintained in the translation.

Table 10

**An example of an equivalence in translating Mao’s agentive Sayer role**

Sayer (targeting) is the only instance of Sayer being recognised as an Agent lexicogrammatically (Halliday and Matthiessen 2014: 207), as if the Sayer is verbally acting on the listener by means of criticism or praise. In Table 10, Mao’s agentive Sayer role has been translated as an equivalent agentive Sayer in the neutral (transitive) voice in Chinese.

4.3. Shifts in the translation of agency

While many lexicogrammatical equivalences of Mao’s agency have been maintained in the chosen translation, more shifts have been identified. In the Chinese translation of Wild Swans, the overall quantitative trend is to reduce, and at times, omit Mao’s agentive roles. Therefore, this section will focus on examples where Mao’s agency has either been diminished or omitted lexicogrammatically. While there may also be isolated examples of translation shifts where Mao’s agentive role has been enhanced or even added in the Chinese translation, they will not be discussed in this paper, as they have not been identified in the chosen chapters nor do they represent the overall quantitative shifts.

4.3.1. Reduction of agency through the change of voice and Theme

Table 11 below shows an example of a reduction of Mao’s agency by changing both the voice and Theme choice.

Table 11

**An example of reduced agency through shifts in voice and Theme choice**

Although the TT represents the same experience and maintains Mao’s agentive role as an Actor that has created changes to the abstract Goal of 解放 (jiěfàng)[rehabilitation] readers may notice that the TT places Mao not in a thematic position as does the ST, but in the middle of the clause, after the Theme 大規模「解放」 (dà guīmó ‘jiěfàng’) [the mass-scale rehabilitation] and before 批准 (pīzhǔn) [approve]. This reflects a grammatical shift in voice. While the voice is active in English, the Chinese clause is a textually motivated variant of the neutral voice according to Halliday and McDonald (2014), due to its lack of a formal marker of the dispositive or passive voice and due to the Goal being placed in the thematic position. This variation attributes less responsibility to Mao by not foregrounding his agency at the thematic position. However, it is unclear whether the change of voice is the driving force or a consequence of the Theme change. Therefore, it becomes necessary to compare the thematic progression in the ST and the TT. The three preceding clauses and one following clause have been extracted to compare the thematic progression in the ST, as shown in Figure 5, and in the TT, as shown in Figure 6.

Figure 5

**Thematic progression in the ST (Chang 1991/2003: 583)**

Amongst the four English clauses, two lines of smooth thematic progression can be identified. One line of a zig-zag progression can be drawn from the Rheme of Clause 1 to the Topical Theme of Clauses 2. One line of a linear pattern can be drawn between the topical Themes of Clauses 3 and 4. However, the Theme choice of Mao in Clause 3 is foregrounded as it cannot be traced back to either the Theme or Rheme of any previous clauses here, creating a gap in thematic progression, which may draw the reader’s attention to this abrupt Theme choice. At the same time, the ST uses the active voice, which attributes maximum responsibility to Mao according to Dreyfus’s cline.

The thematic progression in the TT has been visualized in Figure 6 below.

Figure 6

**Thematic progression in the TT (Chang 1992/2006, translated by Pu Zhang: 399)**

In comparison, the Chinese TT shows smoother thematic progression, with three lines of smooth progression identified amongst the four clauses. The main difference is that the Theme in Clause 3, 大規模「解放」 (dà guīmó ‘jiěfàng’) [the mass-scale “rehabilitation”] is linked to the Rheme 像大多數走資派那樣平反「解放」 (xiàng dà duōshù zǒuzīpài nàyàng píngfǎn ‘jiěfàng’) [rehabilitated like the vast majority of the capitalist-roaders] in Clause 2, which bridges the gap created in the Theme progression in the ST by changing the active voice into a textually motivated neutral voice. Thus, one may reasonably argue that this shift in voice may have been motivated by the translator’s intention to create a smooth thematic development, although it is important to note that this shift is by no means mandatory, and that a similar progression with a marked Theme 毛 (máo) [Mao] would be possible in the Chinese translation. However, a semantic consequence of this shift is the loss of the foregrounding effect of an unexpected thematic position of Mao’s agentive role and subsequently diminished responsibility attributed to him. This consequence remains valid regardless of whether the translator has failed to grasp the subtlety of the foregrounding effect, or whether he has made a deliberate choice to ‘back-stage’ Mao’s agency. The second scenario appears to be more in line with the dramatic overall quantitative trend to represent Mao less and as being less agentive in the Chinese translation.

This example echoes Baker’s (1992/2011) view that Theme choices are highly meaningful in translation, and that of Kim and Matthiessen (2015) that Theme choices may interact with voice. However, it demonstrates how optional shifts in Theme, induced by the change of voice, can lead to diminished agency of a represented participant.

4.3.2. Omission of agency in translation

The next category of translation shifts is the omission of agency. This could take the form of minor omission through representing the same experience but using either an agentive passive or middle voice, which does not represent external agency, or major omission of an entire experience in which agency is featured.

4.3.2.1 Omission of agency through agentive passive and middle voice (implicitation of agency)

Dreyfus (2017: 378) argues that agentive passives attribute little responsibility to an Agent, because it is not mentioned even though it is implied, and that middle voices attribute the “least or no responsibility.” Whereas there is certainly grammatical potential in Chinese for an agentive passive, no examples have been found in the chosen data. However, translation shifts from the English active voice to the intransitive neutral voice in Chinese have been found, as shown in Examples 5 and 6, analysed in Tables 12 and 13, respectively.

Table 12

**An example of a shift in voice in translation**

The Chinese translation here shows an intransitive neutral voice, as if the Cultural Revolution Authority created itself without any external force. Although one may argue that Mao’s role in setting up this authority can be inferred in the co-text, his agency is omitted lexicogrammatically in this clause. It is argued that this is an example of an implicitation of Mao’s agency in translation.

The following example, analysed in Table 13, shows a similar trend of implicitation of Mao’s agency in Chinese:

6

On 10 November 1965, having repeatedly failed to have the article condemning Wu Han’s play published in Peking, Mao was at last able to get it printed in Shanghai, where his followers were in charge.
Chang 1991/2003: 348

最後，在毛安排下，毛夫人江青找上海慣打棍子的評論家姚文元寫了篇批判文章，十一月十日在上海發表。

(zuìhòu, zài máo ānpái xià, máo fūrén jiāngqīng zhǎo shànghǎi guàn dǎ gùnzi de pínglùn jiā yáo wényuán xiě le piān pīpàn wénzhāng, shíyī yuè shí rì zài shànghǎi fābiǎo.)

[In the end, under Mao’s arrangement, Madame Mao Jiang Qing found the commentator Yao Wenyuan who made frequent criticism to write a critical article, on 10 November in Shanghai (it) published.]
Chang 1992/2006: 237, translated by Zhang

Table 13

**An example of an implicitation of Mao’s agency**

The translation shows added experience that does not derive from the ST, such as the printing date, Madame Mao’s role, and additional detail about the writer of the article, reaffirming the nature of the Chinese translation as a form of rewriting. However, the shift under interest is the change of voice from the English active voice to the intransitive neutral voice in Chinese, removing Mao’s agentive role in initiating the printing in Shanghai. Again, Mao’s agency can be semantically inferred from the co-text. Since he arranged for his wife to find a writer, he should have a vested interest in getting the article printed. However, his agency is not represented in the clause. Hence, this shift also exemplifies the implicitation of Mao’s agency in the Chinese TT.

4.3.2.2. Complete omission of an Agent-induced experience

At the other end of the cline of the reduction of Mao’s agency in translation, one finds complete omissions of ST clauses or even clause complexes in which Mao is represented lexicogrammatically as an Agent. Although such forms of omission can be considered a major violation of translation accuracy, which ordinary translators dare not deliberately commit, it has been frequently found in the present translation, especially in relation to Mao’s role.

For instance, two clauses, which follow If he was to get the population to act, within the same clause complex, have been omitted in their entirety in the Chinese TT, as shown in Example 7, analysed in Table 14.

Table 14

**An example of an omission of Mao’s irrealis agency**

The second and the third clause, in which Mao was represented as an agent, have both been omitted, whereas the first clause forms a new clause complex with a different clause in the TT. Likewise, the clause complex in Table 15 below, which contains a clause with Mao’s agency, has been omitted in the TT it its entirety.

Table 15

**Another example of an omission of Mao’s irrealis agency**

Mao’s agency in both examples above is made irrealis by being marked either by modality would have to or by non-finiteness to. This means Mao’s represented agency here is not actual but hypothetical, in the sense that the agency was what Mao was obliged, or desired, to exercise. Since the agency is irrealis, it could not be objectively observed. Thus, it is argued that this is the ST author’s subjective interpretation and speculation of Mao’s mental activities and Chinese history. The omission of many more such examples of Mao’s irrealis agency is coupled with the finding of Li and Li (2020) that Mao’s role as a Senser is significantly reduced in the Chinese TT. Together they reflect a general tendency in the Chinese translation to not provide interpretation of why Mao exercised agency, but to focus more on his agency that can be reported on and verified more objectively.

4.4. Summary of findings

The discussion above categorises linguistic equivalences and shifts in the translation of Mao’s agency, as shown in Figure 7.

Figure 7

**Categorisation of the translation of agency in the present translation**

Equivalences can be achieved through maintaining equivalences in the character’s participant type, the process type, voice, and Theme in the same clause. While there is a number of such equivalences, more shifts have been identified in the selected corpora, including the change of the active voice to a transitive or intransitive neutral voice in Chinese, and even to complete omissions of clauses, all of which result in an overall reduction, implicitation, and even omission of Mao’s agency in the Chinese TT. Examples of complete omissions are found to occur particularly where Mao’s mental activities and irrealis agency were being speculated in the ST. This shows a clear tendency in the Chinese translation to offer a non-compliant reading of the English source text, reducing the overall representation of Mao and his agency but also placing him in a less central and foregrounded position in the narrative of the Chinese translation. Mao is a major character in the ST who is frequently and explicitly held responsible for the tragedies suffered in China and in the author’s household. In stark contrast, he is a minor character in the Chinese translation who is only at times, and often implicitly, held accountable for the unfolding events in the narrative. This shift in characterisation, especially the implicitation of Mao’s responsibility in the Chinese translation, is highly foregrounded considering the hypothesis that translation generally shows a trend of explicitation. In fact, the agency of some other characters, such as the Red Guards, have been explicitated. The two different narratives reflect a deviation in the Chinese translation from the keener interests towards and higher tolerance for criticisms of Mao in the English-speaking context when targeting the Taiwan context.

Another main finding is the intricate interrelation amongst the systems of agency, voice, and theme. This paper echoes the relevant literature in Theme choices in translation that a shift in voice, which may enhance or diminish agency, may be motivated primarily by the need for smooth thematic development. However, it also shows an interesting example of the removal of a textually foregrounding Theme choice in the ST, which results in a plainer translation that draws less attention to Mao’s responsibility. Patterns of such translation choices, whether consciously or unconsciously made, can have semantic and ideological consequences in altering the attribution of responsibility.

5. Conclusion

This paper has built on existing literature in transitivity and theme in translation and discourse analysis. In particular, it adopts Dreyfus’s (2017) cline of responsibility to propose a category of equivalences and shifts in the linguistic realisation of agency in English-Chinese translation, taking into account typological differences between English and Chinese. It also takes into explicit consideration the linguistic systems that may interact with transitivity, including theme, voice, and modality, to differentiate translation shifts that can be considered motivated from those that may be considered mandatory due to the constraints of other linguistic systems. The examples under discussion have usefully demonstrated how localised shifts in transitivity can build up significant overall shifts in the representation of the agency of a character, which alters the characterization and the overall narrative. The proposed category in Figure 7 may be extended to investigate the translation of agency in other text types, language pairs, and directions.

A limitation of this study is a lack of interviews with the producers of the ST and TT. Although the translation process may be inferred based on text analysis, this paper does not claim concrete insights into the process, as no interview has been accorded so far. Rather, it has categorised and discussed equivalences and shifts in agency based on textual evidence and it has discussed the repercussions of transitivity choices on the function, ideology, and narrative of the translation. Regardless of the process and motivation, patterns of meaning created through localised lexicogrammatical choices in translation have semantic and even ideological consequences.

It is hoped that this study will spark more interest in discourse analysis approaches to translation studies, especially that it will foster the consideration of transitivity analysis as a useful tool for the study of agency, characterization, and narratives in different text types and language-pairs. Another aim of this study is to reiterate the importance of typological differences in any text-based study of interlingual translation.

Shifts of agency in translation: a case study of the Chinese translation of Wild Swans

Abstract

Résumé

Resumen

1. Introduction

2. Transitivity and other interacting linguistic systems in Systemic Functional Linguistics (SFL)

2.1. Transitivity within the ideational metafunction