Introduction

Previous research on translated language has revealed its characteristics as a language variety that differs from both the source and target languages. Most studies have focused on translation universals between different language combinations, using linguistic indicators at the lexical, syntactic, semantic, and discursive levels (Baker, 1993; Chesterman, 2010; Feng, et al., 2018; Seracini, 2021). In contrast, interpreted language, or “interpretese” (Shlesinger, 2008), has received limited attention, which has mainly been paid to interpreted discourse in political settings. Interpreted discourse has been approached through attempts to verify the existence of universal features represented by linguistic indicators commonly chosen to study translated discourse. Recently, empirical research has opted for multivariate analysis to investigate clusters of linguistic features by drawing on the comparable corpora of interpreted and non-interpreted texts (Xiao, 2015; Xu, 2021; Zou and Wang, 2021). As is widely applied to register analysis of written and spoken discourses, multidimensional analysis (MDA) based on Biber’s (1988) model has recently attracted the attention of interpreting scholars. This is because MDA can help distinguish two types of discourses based on communicative dimensions rather than a pre-selected set of linguistic features that vary from one study to another. The present study aims to use MDA to distinguish between interpreted and non-interpreted discourse and to shed light on the factors that contribute to such differences in a political setting.

Literature review

From translation universals to interpreting universals

Studies on translation universals, greatly advanced by corpus-based translation studies (Baker, 1995), have revealed linguistic features either overused or underused in translated texts compared to non-translated ones. However, the features chosen to justify translation universals, namely simplification, explicitation, normalization, and leveling, are usually few and loosely connected across studies, such as the optional complementizer “that”, linking adverbials, mean word length (Baker, 2004; Hu, et al., 2019; Olohan and Baker, 2000). Nevertheless, such an approach is sustained by interpreting scholars to specify interpreting universals characterized by linguistic features whose uses consistently exceed expectations in interpreted texts as opposed to non-interpreted ones. Shlesinger and Ordan (2012) examined twenty-nine features (e.g., parts of speech, verb patterns), but found that interpreted texts derived from simultaneous interpreting share more similarities with original speeches of the target language than with translated texts of the same source language. The study suggests the invalidity of the putative laws of translation as applied to interpreted texts. Similarly, Kajzer-Wietrzny (2018) examined the optional “that” as a feature of explicitation in interpreted speeches of the European Parliament, and found that interpreters’ use of “that” was more similar to that of non-native speakers than to that of native speakers. The result concludes that the optional complementizer “that”, which is claimed to strongly distinguish interpretations, tends to be a feature shared by interpreted language and non-native language as constrained language. Moreover, some studies provide evidence against the simplification of interpreted language (e.g., Sandrelli and Bendazzoli, 2005) using features such as lexical density (Hu, 2016) and standard type-token ratio (Fu and Wang, 2021).

Recent studies that question the application of translation universals to interpreting universals tend to suggest that interpreting universals are tendencies rather than universal laws, just as translation universals may not exist (Pastor, et al., 2008). As more studies differ in terms of language pair, working mode, etc., results based on different combinations of linguistic features will, not surprisingly, contradict each other. This may suggest that interpreted language is not embodied by definite yet isolated features independent of the communicative situation but by communicative functions represented by co-occurrence patterns of multiple linguistic features exploited in a particular communicative situation. In light of this, some researchers are beginning to adopt Biber’s MDA to explore communicative functions of interpreted discourse, which are characterized by co-occurrence patterns of linguistic features used in a particular setting.

MDA and MDA in corpus-based interpreting studies

Proposed by Biber (1988), MDA conducts a factor analysis of a wide range of linguistic features to provide a comprehensive description and interpretation of the communicative functions of a register, which is defined as a functional variety of language associated with a given situation type (Halliday, 2004, p.27). Factor analysis is a technique that discovers latent relationships among a set of variables. The objective is to explain several observed variables by a set of factors, where the factor is much smaller in number than the variables (Ding, 2013, p239). The main idea of MDA is “based on the assumption that linguistic co-occurrence patterns reflect underlying communicative functions” (Biber, et al., 2006). According to Biber (1988), six important dimensions are proposed through factor analysis in MDA. Each dimension consists of a set of linguistic features that tend to occur frequently in texts because they serve a common communicative function. The features with positive and negative loadings tend to occur in a complementary distribution, i.e., if the positive features co-occur with a high frequency in a text, the negative features have a rather low frequency in the same text, and vice versa. From a functional perspective, the six dimensions are labeled as follows: involved vs. informational production; narrative vs. non-narrative concerns; explicit vs. situation-dependent reference; overt expression of persuasion; abstract vs. non-abstract information; and online information elaboration.

Currently, several MDA-informed corpus studies of interpreted discourse have involved Chinese-English interpreting practiced in consecutive and simultaneous modes and different communicative settings. For example, Li (2014) employed Biber’s MDA to compare a diplomatic interpreting discourse with various written and spoken registers in original English, such as official documents, academic articles, and interviews. The study found that interpreted discourse was characterized by non-narrative content and abstraction, making it more similar to written than spoken register, but the finding was only supported by a small corpus size. Xiao (2015) selected twenty-one features with the highest weight in each of Biber’s (1988) dimensions and conducted an exploratory factor analysis on a corpus of Chinese-English simultaneous interpretations of international conferences. The results attested that interpreted discourse was likely to be a spoken register based on five dimensions labeled as involvedness, constrainedness, context-boundedness, abstractness, and prosody. One acknowledged limitation was the small selection of features salient in Biber’s (1988) dimensions. Xu (2021) built a corpus of simultaneous interpretations for the sessions of the Legislative Council of Hong Kong and expanded the features specified in Biber’s MDA to seventy-nine. The study showed that the simultaneous interpretations into English were similar to the original English speeches on three of the seven generated dimensions, i.e., involvedness, abstractness, and narrative content. However, the results based on a less balanced and representative corpus cannot be generalized to interpreting in other settings. Zou and Wang (2021) used all the sixty-seven features proposed by Biber (1988) to conduct an exploratory factor analysis of four sub-corpora. It was found that three dimensions, namely oral vs. literate, interpreted vs. non-interpreted, and planned vs. unplanned, could demonstrate how Chinese-English translations and interpretations of political texts differed from each other as well as from the original texts. However, the labels of the proposed dimensions may be too generalized to be informative.

Despite differences in research design and procedures, these few studies mentioned above have confirmed that MDA plays an active role in differentiating the communicative functions assumed by interpreted and non-interpreted discourse through co-occurrence patterns of multiple linguistic features. However, the advantage of following Biber’s (1988) MDA has not been fully exploited to avoid inconsistencies in the selection of linguistic features. Thus, the comparability and replication of related studies need to be improved.

Government press conference interpretations in China

Interpreting practice for the annually convened and live-broadcast press conferences of the Chinese central government during the NPC and CPPCC sessions has been extensively researched in China for more than two decades, because the Chinese government is interested in promoting the understanding and acceptance of its political discourse among international audiences and such research is greatly facilitated by the efforts of building and using large-scale interpreting corpora (Hu, 2016; Wang, 2012). Studies based on these corpora have shed light on norms in conference interpreting, the role of interpreters employed by the government, and translation universals in interpreting products, among others. For example, Wang (2012) identified four norms of ST-TT relationships that constitute one of the most important forces shaping the online performance of Chinese interpreters. Pan and Wang (2021) observed a general trend of political/diplomatic interpreting in China becoming more target-oriented since the 2010s. Li and Zhang (2020) examined the interpreter’s filtering choices at the Chinese Foreign Minister’s press conferences to explore the institutional interpreter’s gatekeeping role. Hu and Tao (2013) and Hu, et al. (2015) attempted to uncover universal features of the interpreted language to test the applicability of translation universals in interpreted texts. The results of their studies either confirmed or disproved the claim that translation universals are suitable to explain certain patterns of linguistic features in interpreted texts. Hu, Pan and Li (2015) explained that interpreters’ decision-making choice of linguistic features was mainly influenced by similarities and differences between the source and target languages, the interpretation of a government press conference as a register, and the agency of the interpreter. As mentioned above, Li (2014) and Zou and Wang (2021) are among the first to take advantage of Biber’s MDA to investigate the communicative functions which distinguish interpreted discourse from non-interpreted discourse in political settings from a functional perspective.

Overall, the focus on interpreting practice at such high-profile political events has shifted from interpreters’ expertize to their active role as mediators in the reconstruction and reframing of Chinese political discourse, and to global styles of communicative functions of interpreting for government press conferences. However, previous studies that applied Biber’s (1988) MDA to investigate communicative functions either compared interpreted political discourse with non-interpreted non-political discourse or did not strictly follow Biber’s (1988) MDA. Thus, the present study attempts to overcome these shortcomings by adhering to Biber’s (1988) model and limiting the communicative setting of both interpreted and non-interpreted discourse to government press conferences. Therefore, the research questions that the present study attempts to address are as follows:

  1. (1)

    Is the interpreted speech at the government press conference different from the non-interpreted one from a communicative perspective?

  2. (2)

    What linguistic features serve to distinguish the interpreted speech from the non-interpreted one at such an event in terms of communicative dimensions?

  3. (3)

    What reasons account for these differences?

Methods

The corpus

The multi-dimensional analysis devised by Biber (1988) is in nature a corpus-driven approach integrated with inferential statistical methods. The comparable corpus under study consists of two subcorpora. One is a collection of the transcriptions of the Chinese English consecutive interpretations at the Chinese Premier’s press conferences convened during the NPC and CPPCC sessions from 1998 to 2017 (hereinafter referred to as the IE subcorpus). The other contains the original English transcripts of the American government’s press conferences (hereinafter referred to as the OE subcorpus) during the same period (obtained from the website https://www.presidency.ucsb.edu). The reason for selecting the latter texts for comparative analysis is their proximity to the former in terms of the communicative setting, the speaker’s identity and delivery mode of spontaneous speeches as well as the time span of the communicative event (see Table 1).

Table 1 The corpus.
Full size table

Data analysis

In the present study, Biber’s (1988, p.89) model involving six dimensions and sixty-seven linguistic features is rigorously adopted. The software program for MDA is Multidimensional Analysis Tagger (MAT) (Nini, 2019), which replicates Biber’s analysis as closely as possible. The program not only generates a grammatically annotated version of the corpus for statistical calculations but also provides visualizations that display the position of the corpus on Biber’s (1988) dimensions and reports the closest text type of the corpus. MAT calculates dimension scores based on z-scores of frequency data of the sixty-seven linguistic features proposed by Biber (1988). The dimension scores and z-scores of the two subcorpora are then imported into SPSS 20.0 for the Student’s t test. The advantage of using the tagger is that it facilitates the comparison of the results of the present study both within the corpus and with those of similar studies.

Results and Discussion

Difference in text type

According to the MAT reports, the corpus scores on the six dimensions decide that the closest text type of the OE subcorpus is involved persuasion, while that of the IE subcorpus is general narrative exposition. This summary finding highlights the fundamental register difference between the two types of discourses. The finding about the interpreted discourse is somewhat contrary to expectations since the source speech that was interpreted at the government press conference was persuasive in nature as political discourse and its interpretation was assumed to be endowed with the same nature. However, it is consistent with Zou and Wang (2021)’s findings, also generated by MAT, about the text types of the American government’s daily press briefings and State of the Union addresses, and the interpreted discourse of the Chinese Premier’s press conferences. It seems that American presidents and spokespersons do not use high-profile press conferences as a venue for information dissemination and diffusion. Their political speeches serve a function of argumentation and persuasion, similar to impromptu speeches, interviews, etc., illustrating the American government’s attempts to persuade its own and foreign public to engage in certain activities or accept certain viewpoints proposed by the government. Convincing or not, the speeches seek to exercise and maintain the discourse power of the government, a power that is quite influential given that it is backed by the national strength of the country.

Against expectation, the interpreted discourse deviates from the persuasive function inherent in political discourse. As a text type of general narrative exposition, it assumes the purpose of conveying accurate and objective information of general interest, which is shared by press reports, editorials, etc. This can be explained by the Chinese government’s intention to provide interpreting services at such high-profile political and media events, i.e., to facilitate reliable and effective dissemination of information about the government’s policies and positions taken in the past year. It should be noted that by providing a variety of information and facts, the government seems to give priority to winning trust before winning hearts. Therefore, the government’s unwillingness to persuade may help promote its image before the foreign press and media, which often see the country as power-hungry and high-profile in the global arena.

Such difference reveals the different functions underlying each type of political discourse produced in very similar settings. The purpose of the political discourse is the main determinant of its function, which to some extent reflects how a government seeks to present itself and be presented by others in international political communication.

Differences in communicative dimensions

The Student’s t tests on the six pairs of dimension scores of the two subcorpora show that the two types of discourses are significantly different on five of the six dimensions, except for Dimension 4 (see Table 2). The results of the difference in the frequency data of the linguistic features reveal the cluster of linguistic features and their co-occurrence patterns that contribute to the differences on each dimension of the two types of discourses.

Table 2 Student’s t test results for dimension scores and linguistic features with significant differences in frequency in the two subcorpora.
Full size table

Dimension 1: Informational vs. involved production

Dimension 1 distinguishes between the discourse with a personal/involved focus and the discourse with an informational focus (Biber, 2014). This dimension can capture the most salient characteristics of the two subcorpora because it is identified by factor analysis as the most important factor, accounting for the co-occurrence patterns of the largest number of linguistic features involved in MDA. Positive features on this dimension characterize informal, interactive spoken language, and features with negative loadings characterize information-dense, typically written language. Dimension 1 is therefore particularly relevant to the tension between oral and literate discourse (Kruger, et al., 2019). As shown in Table 2, the two subcorpora are remarkably different on Dimension 1. The OE subcorpus is characterized as highly involved with a high positive score (10.35), while the IE subcorpus is aimed at providing informational content with a high negative dimension score (−7.37). In comparison, the original political speeches show typical characteristics of oral discourse, while the interpreted discourse is positioned closer to literate discourse, although it is delivered in the oral mode. The findings about the consecutive interpretations of government press conferences are in agreement with the studies of Li (2014), Xiao (2015), and Fu and Wang (2021) on either consecutive or simultaneous interpretations.

According to Table 2, the high positive score of the OE subcorpus is attributed to the higher frequencies of a majority number of positive features loaded on this dimension, whereas the low negative score of the IE subcorpus is due to the much higher frequencies of four out of the five negative features involved. Considering example (1) of the OE subcorpus, it is found that the U.S. speaker frequently uses the personal pronoun “we” to refer to themselves or the government and the audience inclusively, and meanwhile uses the private verb “know” to openly express their thoughts and feelings. In addition, the speaker uses demonstrative pronouns, contractions, possibility modals “could” and non-phrasal coordination to connect his utterances, resulting in relatively general and fragmented information. The frequent co-occurrences of the positive features both contribute to the active involvement of the audience and serve to indicate that the speaker is not committed to precise content. In contrast, in example (2) of the IE subcorpus, the Chinese speaker uses nouns and words of greater length more frequently to elaborate on the government’s positions on domestic and international affairs and employs prepositions and predicative adjectives (e.g., “democratic, scientific, and responsible”) to compress intensive information into condensed sentence structures. The high frequency of co-occurrence of these negative features in the IE subcorpus illustrates the similarity of interpreted discourse to the written register in terms of information density.

(1) OE: I also know that there are businesses around the country that could be growing even faster, that our deficits could be coming down faster, that we would have more customers in our shops, if we get this thing resolved. We know what the right thing to do is. It’s a matter of political will. It’s not any longer a matter of policy. And I’m going to continue to encourage them to get this done.

(2) IE: Only through reform, opening-up and building socialism with Chinese characteristics under the leadership of the Chinese Communist Party can we succeed in bringing about a strong and prosperous China. And only by adhering to a democratic, scientific, and responsible spirit can we achieve our objectives.

Dimension 2: Narrative vs. non-narrative concerns

Dimension 2 reflects narrative vs. non-narrative concerns, or distinguishes between “an active, event-oriented discourse and a more static, descriptive or expository type of discourse” (Biber, 1988). On this dimension, both subcorpora show a focus on non-narrative content, given their negative dimension scores (see Table 2). The result about the IE subcorpus is supported by Li (2014). Nevertheless, the interpreted discourse contains more non-narrative content than the non-interpreted discourse, as the score of the former is significantly lower than that of the latter (p < 0.001). The more non-narrative content in the IE subcorpus is produced by the higher frequency of predicative adjectives than that of present tense verbs in the OE subcorpus. Due to the higher use of predicative adjectives loaded as a negative feature on both Dimensions 1 and 2, the interpreted discourse not only signals information density but also represents a static type of discourse.

Besides, the two subcorpora present narrative content with different preferences for verb tense (see Table 2). The OE subcorpus uses more past tense verbs to refer to past events or statements made by a third party, such as “they” in example (3). In example (4), the IE subcorpus prefers the present perfect tense to emphasize things that happened before and have an impact on the current situation, especially when it predicates the positive effect or progress of the Chinese government’s work and policies (Gu, 2018). Considering both narrative and non-narrative content, it can be inferred that the interpreted and non-interpreted discourse use different patterns of linguistic features for similar communicative functions.

(3) OE: They still can’t bring themselves to admit that the Affordable Care Act is working. They said nobody would sign up; they were wrong about that. They said it would be unaffordable for the country; they were wrong about that.

(4) IE: In this government, we have issued a total of 510 billion yuan of treasury bonds…we have managed to mobilize capital from the banks…over 2 trillion RMB yuan have been spent to undertake various kinds of projects…we have also built 5000 km of trunk railways…

Dimension 3: Explicit vs. situation-dependent reference

Dimension 3 captures the distinction between text-dependent reference and situation-dependent reference that rely on the shared context of the addresser and addressee for interpretation (Biber, 1988). According to Table 2, the IE subcorpus is more context-independent than the OE subcorpus, which is corroborated by the findings on Dimension 1 indicating the proximity of interpreted discourse to the written register, since the written register is characterized by context-independent reference. The result is in line with studies on interpretations by Li (2014) and Xiao (2015) and is also supported by research on translations into English (Kruger and van Rooy, 2016).

With a relatively low positive score, the OE subcorpus shows a slight tendency toward situational reference. Example (5) shows that the U.S. speaker frequently uses such positive features as pied-piping relative clauses and WH relative clauses on the subject position to clarify references. In example (6) of the IE subcorpus, the use of nominalizations is more often associated with references to the Chinese government and abstract concepts describing social activities and purposes. Also, the heavier use of phrasal coordination helps to specify two parties involved or to connect complex concepts that involve two independent ideas. These positive features characterize the integration and density of information while clarifying references in the interpreted discourse, easing the burden of shared context and prior knowledge on the part of the non-Chinese audience.

(5) OE: But I’m clear also that there are many issues on which the United Kingdom and the United States stand alongside one another, many issues on which we agree.

(6) IE: I wish to state, in clear-cut terms, that the central government will as always continue to give support to Hong Kong to ensure its stability and development. As a matter of fact, we are losing no time in bringing about the arrangement for closer economic ties and trade between the mainland and Hong Kong.

Dimension 4: Overt expression of persuasion

Dimension 4 is characterized by “overt expression of persuasion” (Biber, 1988, p.111). The result shows no significant difference between the two subcorpora on this dimension (p = 0.552), indicating that both discourses of press conferences have explicit marking of persuasion. It contradicts the claim by Xu (2021, p.136) that simultaneous interpretations of the Hong Kong Legislative Council were not persuasive, which was explained by the interpreter’s attempt to remain neutral and risk-averse due to a lack of shared context with the speaker.

However, the two discourses fulfill the purpose of persuasion by means of different co-occurrence patterns of positive features. Example (7) from the OE subcorpus shows that the American speaker tends to achieve persuasion indirectly by using conditional adverbial subordinates and the infinitive as adjective and verb complements. In comparison, example (8) from the IE subcorpus illustrates that the Chinese speaker tends to use necessity modals “should” and split auxiliaries (i.e., auxiliaries followed by one or two adverbs and a verbal base form) to show directness in giving instructions or making an appeal to the audience.

(7) OE: If we want to continue to enjoy success, we must continue our commitment to dynamic action. There is important work to be done in America this year, …

(8) IE: In security, we should never forget about the dangers and in times of peace, we should always be alert to the potentials for chaos.

Dimension 5: Abstract vs. non-abstract information

Dimension 5 distinguishes discourses that are “abstract, technical and formal from other types of discourses” (Biber, 1988). This dimension, similar to dimension 1, captures a contrast between informationally dense and colloquial discourses. The difference between the two types of discourses is significant (p < 0.001) on Dimension 5, though not as big as the differences on Dimensions 1 and 3. The OE subcorpus is slightly concrete, given a negative score close to zero, while the IE subcorpus contains more abstract information, but its abstractness remains at a low level. However, Li (2014) argued that interpretations for Chinese prime ministers contained more concrete information, and Xiao (2015) found that simultaneous interpretations for panel discussions were less abstract than such discussions in original English. The different result of the present study may indicate that the more abstract interpretations are subject to a higher degree of formality and magnitude of the high-stakes political event.

Example (9) shows the more frequent use of past participle WHIZ deletions and by-passives in the IE subcorpus. The patient is emphasized when the agent is by fault the Chinese government, or in other cases a particular government agency is specified by the by-passive. Also, the co-occurrence of conjuncts “for example” with the past participle “listed” indicates complex logic between the components of a long sentence. The frequent co-occurrence patterns of these positive features make the interpreted discourse more abstract than the non-interpreted discourse.

(9) IE: Of course, there are also new projects listed in the plan, for example, the government-subsidized housing project, in a space of 3 years we’re going to resolve the housing difficulties for 7.5 million households …

Dimension 6: On-line informational elaboration

Dimension 6 contains features that mark information elaboration in unplanned or well-planned discourses. According to Table 2, both American speakers and Chinese interpreters can convey elaborate information, with the former slightly more able to do so than the latter. The result is supported by Li’s (2014) and Xu’s (2021) investigations of either consecutive or simultaneous interpretations.

Example (10) of the OE subcorpus shows that various “that” clauses and demonstratives provide detailed explanations of government policies or actions. However, the non-interpreted discourse has low information density due to tacking on additional dependent clauses and less careful planning of sentence structures. Conversely, the IE subcorpus shows low frequencies of positive features but prefers a negative feature, i.e., phrasal coordination, which results in low information elaboration but well-planned information integration. The result of Dimension 6 is also aligned with those of Dimension 1.

(10) OE: And in exchange for a temporary extension of the high-income tax breaks … a policy that I opposed but that Republicans are unwilling to budge on, this agreement preserves additional tax cuts for the middle class that I fought for and that Republicans opposed two years ago.

Reasons for differences

Based on the above-mentioned findings of the comparable corpus, the reasons for the differences in text type and communicative functions between the interpreted and non-interpreted discourse of government press conferences can be summarized in the following three perspectives.

First, government press conferences are venues for verbal interactions between government spokespersons and media reporters. Compared to the non-interpreted discourse, which is represented by the interactive and non-abstract nature on Dimensions 1 and 5, the interpreted discourse reflects an informative and abstract nature that characterizes the written register. The style of interpreted discourse seems to be strongly influenced by the communicative context and cultural background in which it is situated (Halliday, 1988; Hu, 2011), in addition to the equilibrium effect of interpretation (Shlesinger, 1989). Similar as they are in terms of discourse field and discourse mode, the two types of discourses are different in terms of discourse tenor (Li, 2018). As a formal, high-stakes institutional discourse in Chinese culture, the political discourse of government press conferences is characterized by a relatively distant power relationship and a low degree of contact between the government speaker and the audience. Moreover, the interpreted discourse is primarily aimed at informing rather than persuading the domestic and foreign public about the government’s policies and positions. Therefore, the communication purpose and the role relationship between the communication actors combine to ordain the interpreting style to be formal and abstract rather than colloquial, and more concerned with informativeness rather than interactivity. This, in turn, may increase the cultural distance between the speaker and the foreign audience through the interpreted discourse (Du and Wang, 2021), reflecting the purposive and power-invested nature of interpretations of government press conferences as an institutional discourse.

Second, the binding effects of norms on government-employed interpreters cannot be ignored, since they are value judgments shared by the groups of interpreting professionals and their clients. On the one hand, on Dimension 4 concerning persuasiveness, both interpreted and non-interpreted discourse function to persuade, but the former tends to overuse necessity modals and split auxiliaries instead of conditional adverbial subordinates preferred by the latter. Additionally, the interpreted discourse is less detailed with respect to Dimension 6 concerning information elaboration. These findings suggest that government-employed interpreters as interpreting professionals adhere to the norm of “adequacy” (Toury, 1995) so that they are primarily responsible for language rather than content (Ren, 2017). On the other hand, on Dimension 5 regarding abstraction, the interpreted discourse shows a preference for conjuncts to present complex logical relationships between sentences, rendering explicitness in grammatical cohesion (Hu et al., 2015). These choices demonstrate the interpreters’ conformity to the norm of “acceptability” (Toury, 1995), resulting in interpretations that become more hypotactic or target-oriented (Pan and Wang, 2021) in sentence construction. After all, naturalness, as evidenced by maximum proximity to the target language, is commonly regarded as a sign of professionalism and expertize by interpreters employed by the government, who are believed to reflect the highest level of interpreting expertize in China.

Finally, the style of interpreting for government press conferences is influenced by the habitus of interpreters employed by the government. The results of Dimensions 2 and 3 show that the government interpreters tend to use the present perfect aspect of verbs to express narrative content and phrasal coordination and nominalizations for denotative clarity, which are not prevalent in the non-interpreted discourse. Due to the gradual routinization of the live broadcast mode of Chinese government press conferences and the fact that the events aim to involve a wider range of “absent” global audiences, the interpreters become more aware of their gatekeeping role (Li and Zhang, 2020) and inclined to risk avoidance (Kajzer-Wietrzny, 2018; Pym, 2004). They are more vigilant about the clarity and cohesion of the interpreted discourse broadcast live. Furthermore, the working mode of consecutive interpreting requires the interpreters to improve efficiency through near-automatic selection of expressions and high compression of information. This requirement not only reflects the expertize of interpreters affiliated with the government but also leads to the convergence of their individual interpreting styles. Overall, the interpreters’ habitus, exemplified to some extent by their choices for narrative content and referential clarity, illustrates that the interpreted discourse is influenced by non-linguistic factors such as the working mode of interpreting, the broadcast mode of government press conferences, and the interpreters’ perceptions of their role, thus subjecting the shaping of interpreting style to tightly interwoven socio-cultural factors.

Conclusion

Through a comparative MDA of the interpreted and non-interpreted discourse of government press conferences, several conclusions are drawn based on the quantitative and qualitative analysis of the comparable subcorpora.

First, the present study suggests that communicative functions, rather than a predefined set of linguistic features, determine the universal characteristic of interpreted discourse and distinguish it from non-interpreted discourse.

Second, the findings present the unique communicative style and structural system of the interpreted discourse of government press conferences, as it differs from the non-interpreted discourse in text type and the co-occurrence patterns of multiple features on different communicative dimensions. Even under the constraint of the working mode of consecutive interpreting, the interpreted discourse is higher in information density and integration, clearer in reference, and more formal and rigorous in style. As an institutional discourse associated with the power of the central government, the interpreted discourse is characterized by more attention to non-narrative content, focus on explaining government work and principles, a lower level of abstraction, and a persuasive function as effective as the non-interpreted political discourse. Overall, the factors influencing the formation of interpreting style are derived from the nature of the interpreted discourse as an institutional discourse, interpreting norms, interpreters’ habitus, and other socio-cultural facets such as the working mode and broadcasting mode of interpreting for government press conferences.

Last but not least, the MDA used in this study demonstrates its effectiveness in identifying the differences in stylistics and discourse functions between interpreted and non-interpreted discourse in similar communicative settings, which are attributed to varied co-occurrence patterns of feature combinations that cannot be observed in conventional frequency-based analyses. It aids in improving the comparability of linked research and in giving a general overview of the characteristics of interpreted discourse. The effects of chronological change in political discourse, the interpreter’s idiosyncrasy, and a comparative analysis of interpreted English discourse from Chinese and other languages may be subjects for future research.