Introduction
Propagandists face a dilemma. They want to discredit political ideas that rival their own, but to do so they often must invoke those very ideas for their readers. This entails a risk: perhaps the reader/listener will become curious and wish to learn more. After all, consumers of authoritarian propaganda can sometimes “read between the lines” to glean information or meanings that the propagandists did not intend to disseminate (Weiss and Dafoe Reference Weiss and Dafoe2019, 964).
Including indicators of disapproval in the text is one technique to simultaneously invoke and discredit an idea. In China’s official discourse, one common marker is to place the phrase so-called, in English, or 所谓, in Chinese, before the idea to be discredited.Footnote 1 Usually, so-called is paired with ironic inverted commas around the relevant phrase to underscore the intended disapproval. For example, journalists Mary Hui and Dan Kopf noted the dramatic rise in the use of so-called in the official language of Hong Kong as the Chinese Communist Party (CCP or party) consolidated political control over the city in 2019 and 2020. The intent, they write, was for the authorities to “undermine information they find objectionable” (Hui and Kopf Reference Hui and Kopf2020; see also Colville Reference Colville2024).
China is not the only context in which the use of so-called or inverted commas can be found. In writing about how language was corrupted by the Nazi regime, Victor Klemperer noted that the ironic inverted comma “questions the truth of that which is quoted, declares that the reported remark is untrue” (Klemperer Reference Klemperer and Brady2013, 75–76). In the later years of the Soviet Union, Yurchak notes that when translating foreign texts for Soviet audiences “interpreters were advised to use special indexical markers—such as quotations marks, the term ‘so-called,’ and so forth—that would signal to Soviet readers that the foreign phrases were not used in the proper monosemic sense ‘accepted in our literature’” (Yurchak Reference Yurchak2005, 52).Footnote 2
In Xinhua, China’s main state news agency, from 2003 to 2022 we count 22,914 articles containing so-called in the agency’s English content and 22,524 articles containing 所谓 in Chinese. In this research note we apply quantitative text analysis techniques to these samples. We have three aims. First, we want to know how, if at all, usage changed over time between 2003 and 2022. Second, we aim to compare the Chinese and English text corpora to understand differences and similarities in what Xinhua discredits domestically and internationally. Third, we hope to better understand the specific ideas or entities that the PRC propaganda apparatus wishes to discredit.
More details will be provided below, but temporally we note a dramatic increase in the use of the term from 2018. Thematically our findings indicate that in both corpora, certain core ideas attract the use of the term so-called. These tend to relate to perceived challenges to China’s sovereignty, such as control over Taiwan or Hong Kong. A theme of rising importance in both corpora is systemic political rivalry with the United States. When it comes to differences between internal and external propaganda, we find broad similarities but that Chinese content focuses even more on discrediting the US and less on countering criticisms of China’s ethnic minority policies.
These findings are important because they provide a systematic view of the ideas and entities that the PRC discredits for different audiences. Although ours is a brief research note and does not build a theory to explain its findings, we believe our study contributes to key debates. This analysis adds to scholarship in comparative authoritarian propaganda by analyzing the discrediting of rival ideas and entities, PRC propaganda specifically by identifying temporal, thematic, and audience targeting patterns, and Chinese politics more generally by showing in new ways ideas that the CCP finds threatening.
Authoritarian propaganda and de-legitimation
Space constraints of the research note format preclude a comprehensive literature review, but we nonetheless wish to highlight the scholarship we build on. Propagandists must contend with ideas, evidence, and/or symbols that unsettle their preferred ideological universe. The most obvious option is to censor them. Authoritarian regimes, and the PRC specifically, censor a great deal. However, censorship brings its own risks insofar as awareness about censorship can motivate people to seek out alternative information (Roberts Reference Roberts2018, 137–145). There are ways to obscure evidence of censorship and censors learn and adapt over time (Guriev and Treisman Reference Guriev and Treisman2022, 86–113), but inevitably some heterodox ideas sneak through. In China, despite its extensive censorship system, researchers have shown how unofficial ideas still enter Chinese discourse (Lu et al. Reference Lu, Schaefer, Park, Joo and Pan2022). Furthermore, sometimes states encourage the dissemination of criticisms of their own policies if those criticisms are perceived as coming from a disreputable source, thus undermining the criticism itself (Gruffydd-Jones Reference Gruffydd-Jones2022). Thus despite strict censorship, some ideas will be reported in propaganda that need to be designated as unacceptable for the audience.
Authoritarian leaders delegitimize ideas because they threaten their own power or interests. In the Chinese case, we know that the authorities find alternative political ideas threatening because they admit as much. The leaked 2013 inner-party document “Communiqué on the Current State of the Ideological Sphere,” better known as Document 9, explicitly framed ideas like human rights, constitutionalism, free media, and questioning the party’s version of history as threats to party rule (ChinaFile 2013). A wide-ranging tightening of the public sphere and crackdown on groups associated with these ideas followed.
The propaganda apparatus is key for delegitimizing ideas threatening to the authorities’ political order. Studies of China’s propaganda system emphasize its importance to CCP rule and its penetration of the public sphere (Huang Reference Huang2015; Creemers Reference Creemers2017; Mattingly and Yao Reference Mattingly and Yao2022). Some argue that much Chinese propaganda is about signaling domination rather than persuading citizens (Huang Reference Huang2015; Carter and Carter Reference Carter and Carter2024). There is merit to this view, but the ideas themselves matter too, as evidenced by the fact that they often change over time.Footnote 3 While the study of authoritarian propaganda typically focuses on the ideas that the government wishes to popularize, less attention is paid to the ideas that the propaganda apparatus wishes to discredit (however, see Chester Reference Chester2024). Propagandists must decide what to do with ideas that they oppose but do not censor, either because they are unwilling or unable.
Authoritarians use communication strategically to target different audiences at different times (Baturo and Tolstrup Reference Baturo and Tolstrup2023). One basic distinction among many possible audience subdivisions is between internal (domestic audiences) and external (foreign audiences). External propaganda has a rich history in the CCP and has long been used strategically to shape foreign perceptions of the party’s goals and practices (Edney Reference Edney2014; Brady Reference Brady2015; Tsai Reference Tsai2017). The globalized expansion of the PRC’s external propaganda since around 2008, after pro-Tibet protests and the Beijing Summer Olympics, was ambitious and partially born of the party’s perception that it was unable to steer global conversations, advance its preferred concepts, and marginalize criticism of its policies (Kurlantzick Reference Kurlantzick2023; Gruffydd-Jones Reference Gruffydd-Jones2022, 54–59). Streams of internal and external CCP propaganda intersect (Edney Reference Edney2014; Weiss and Dafoe Reference Weiss and Dafoe2019), but exploring their differences and similarities remains worthwhile, if only because the party maintains this distinction. Operating “under the principle that ‘internal and external are different’ (内外有别),” for example, the Central Propaganda Department recommends different messages for internal and external audiences (Perry Reference Perry, Shue and Thornton2017, 38).
Data, classification, validation, and broad temporal trends
To determine how and when so-called is used, we searched for all articles containing the term so-called and 所谓 within the English and Chinese editions of Xinhua News Agency, respectively, over a 20-year period (2003 to 2022). Xinhua is chosen because, along with the People’s Daily, it forms the most authoritative party-state media in China and is geared toward a mass audience (Esarey Reference Esarey2021). Xinhua also functions as China’s main international newswire, thus allowing comparative analysis of internal, or Chinese language, and external, or English language, messaging.Footnote 4
The search yielded 22,914 English-language articles out of a total of 1,895,630 and 22,524 Chinese-language articles out of a total of 1,509,869.Footnote 5 We adopt a supervised classification approach, which entails training an algorithm on a sub-sample of the data by applying labels and then applying the algorithm to the remaining unlabeled data (Dai and Luqiu Reference Dai and Luqiu2022, 264). We use the Naïve Bayes classifier, which is regarded as a well-performing baseline model upon which more complicated models are based (Grimmer, Roberts, and Stewart Reference Grimmer, Roberts and Stewart2022).
To begin, labelled data must be created. Sentences with so-called were coded by the researchers into two categories, sarcastic or neutral, based on the context of the sentence. For example, a sarcastic use is: “A Chinese Foreign Ministry spokesperson on Friday refuted the so-called ‘genocide’ and ‘forced labor’ in Xinjiang that U.S. politicians, in collusion with some anti-China organizations and individuals, have been unscrupulously spreading and hyping up for their ulterior political purpose [text18145].” A neutral use occurs when so-called is used in place of a phrase like “what is dubbed as,” for example: “Indonesia … is laid on a vulnerable quake-hit zone, the so-called ‘the Pacific ring of Fire,’ where two continental plates, stretching from Western hemisphere to Japan, meet that causes frequent seismic and volcanic movements [text827].”
Applying a trained classifier on the full corpora, we find that 23.4 percent of the English-language corpus (5,367 articles) was predicted to be sarcastic, compared to 58.5 percent for the Chinese-language corpus (13,193 articles). These proportions correspond well to those in the human-coded samples. While it is difficult to say with certainty, differences between the corpora may result from linguistic differences, audience targeting strategies, or some combination of the two (see Alvaro Reference Alvaro2013; Reference Alvaro2015).
We further validated by manual inspection of a random sample of 10% of the machine-predicted sample (535 articles in English; 1,238 in Chinese), finding that 83.9% and 84.8% of the English and Chinese samples were correctly classified, respectively. In English 69.3% of the mentions of so-called were followed by ironic inverted commas and in Chinese 77.5%. Table A1 in the Appendix presents examples.
Since we are only interested in the sarcastic usage of the term, we focus on those in the analysis, but the fact that neutral usages exist allows us to establish time trends. Figures 1 and 2 show the distribution of sarcastic articles in proportion to all articles containing the term so-called. The figures reveal a dramatic increase in the proportional usage of so-called as a marker of disapproval from 2018 in both languages. We are not able to determine with certainty what drives the changes, but we suspect they stem from real-world events, in particular the combination of increased hostility in US–China relations and Xi Jinping’s efforts to legitimate his personalistic rule. The US is seen by the party as a major ideological threat (Garver Reference Garver2016) and Xi himself has called for a more active ideological battle with the West (Shirk Reference Shirk2023, 184, 263) which would help explain an increase in targets for de-legitimation in a time of high tension.
Figure 1. Distribution of sarcastic articles, in proportion to the entire corpus of 22,914 English-language articles containing the term ‘so-called.’
Figure 2. Distribution of sarcastic articles, in proportion to the entire corpus of 22,524 Chinese-language articles containing the term 所谓.
Evaluating themes and variation between language editions
With these basics established, we evaluate whether topics and words associated with sarcastic so-called markers vary between the Xinhua English and Chinese editions. We do this by employing a machine-learning technique known as topic modelling. First, we used the spectral initialization function within the Structural Topic Models (STM) package (Roberts, Stewart, and Tingley Reference Roberts, Stewart and Tingley2019) to identify a baseline number of topics (Lee and Mimno Reference Lee and Mimno2014).Footnote 6 Next, we used Keyword-Assisted Topic Models (keyATM) (Eshima, Imai, and Sasaki Reference Eshima, Imai and Sasaki2023). Unlike popular unsupervised models such as the LDA and the STM, keyATM is a semi-supervised approach which “allow(s) researchers to label topics via the specification of keywords prior to model fitting” (Eshima, Imai, and Sasaki Reference Eshima, Imai and Sasaki2023, 2).
Based on the literature, preliminary analyses of word co-occurrences, and the STM analysis, we identified seven theoretically important topics and generated a list of keywords for each topic.Footnote 7 The seven topics are: territorial disputes, “One-China,” foreign affairs, cooperation, rival systems, economy, and ethnic minorities. We then ran the keyATM models based on the seven topics.
Figures 3 and 4 show the distributions of those topics in the two language editions. In both editions, “Rival Systems,” “Cooperation,” and “One China” appear within the top five topics.Footnote 8 However, domestically the CCP propaganda apparatus appears more preoccupied with delegitimizing the United States and its role in the world. In the Chinese corpus, it appears especially evident that the top two topics, “Rival Systems” and “Foreign Affairs,” focus on criticism of US foreign policy.
Figure 3. Distriution of top 10 topics and corresponding keywords across English-language articles (keyATM).
Figure 4. Distribution of top 10 topics and corresponding keywords across Chinese-language articles (keyATM).
In line with existing research on the international editions of Chinese state media (Zhou Reference Zhou2022), “Cooperation” is more highly emphasized in the English edition, ranking higher than all other topics besides “Rival Systems.” This suggests there is more positive messaging towards international audiences and less wholesale repurposing of domestic stories of chest-thumping nationalism that may be unpalatable to international audiences. These articles tend to extol China’s cooperation in contrast to negative framings of its actions that are prefaced with “so-called.” This slight shift in emphasis may be a strategy for Xinhua to deal with the dilemma in Chinese external propaganda (Edney Reference Edney2014), which is torn between the need to be consistent with the party line domestically and the need to maintain international credibility.
The “Ethnic Minorities” topic appears within the top five topics of the English corpus but is relatively de-emphasized in the Chinese corpus, ranking seventh. To dig deeper, we filter for top articles that mention “Xinjiang” / “新疆” across a sample of 2,000 articles within the “Ethnic Minorities” topic in 2021. We get 50 English articles and 38 Chinese articles and apply qualitative content analysis to the sample.
From Table 1, we see that Xinjiang was actively reported on in the English edition in the first half of 2021, compared to the Chinese edition, which only increased coverage of the topic in the second half. Based on manual inspection of the articles, we find that the English content typically responded to international condemnation of state repression in Xinjiang. On the other hand, the Chinese content reflected reporting of the state’s public-relations efforts, such as press conferences, white papers, and state-organized trips to Xinjiang for international guests. We take this as evidence for internal/external audience targeting. While the translated English article ‘twin’ of a Chinese article sometimes exists, these articles are not identical and often differ in framing and narrative style.Footnote 9 More importantly, not everything that is published by the Chinese edition appears in the English edition, and vice versa.
Table 1. Distribution of the sample of sarcastic articles mentioning “Xinjiang” in 2021
Without wishing to over-extrapolate from differences in these samples, overall, the results suggest that for both audiences discrediting ideas associated with rival political systems is important but that the US is further discredited for a domestic audience in the context of foreign affairs while cooperation is emphasized more in external propaganda.
Evaluating specific variation over time
Next, we examine the temporal trends of each topic using the dynamic keyATM function. Figures 5 and 6 display results. We observe that the topic proportions change in response to real-world events and the news cycle, and this allows us to validate the performance of the keyATM model. For example, the spikes in the topic prevalence of territorial disputes reflect the 2012 Senkaku/Diaoyu islands dispute between China and Japan, and the Permanent Court of Arbitration’s ruling in favor of the Philippines in the South China Sea dispute in 2016. The spike in the topic prevalence of ethnic minorities in 2008 is likely to reflect the 2008 Tibetan uprising and repression and the global attention it garnered in the lead-up to the Beijing Olympics. The intensification of the US–China trade war is likely to be captured by the spike in the topic prevalence of the economy in 2018. Meanwhile, the spikes in the topic prevalence about ‘one China’ corresponds to periods (2003–2008 and 2016–present) where the pro-sovereignty Democratic Progressive Party (DPP) held power in Taiwan.
Figure 5. Topic prevalence over time in the English-language corpus (keyATM).
Figure 6. Topic prevalence over time in the Chinese-language corpus (keyATM). The order of the plots corresponds to the order in Figure 5.
Figure 7. Top country mentions across both corpora.
While most of the topics experience variation over time as events arise, the “Rival Systems” topic is the only one which appears to show a clear continuous upward trend over time in both samples, indicating an underlying strategy to bolster the Chinese model and undermine the US. We further validated the performance of the keyATM model with manual inspection of the top articles per topic, finding that 76.2 percent and 85.7 percent of the English and Chinese samples fit well to their assigned topics.Footnote 10
Aside from the US focus, sarcastic usage of so-called appears to depend on political realities and external events that drive mentions of the concepts. For example, in filtering for top articles that mention “Taiwanese independence” / “台独” across a sample of 2,000 articles within the “One China” topic, we end up with 65 English articles and 177 Chinese articles (Table 2). Remarkably, none of the articles were published when the China-friendly Kuomintang (KMT) was in power. All such articles were published when the DPP, regarded by Beijing as separatists, held power.
Table 2. Distribution of the sample of sarcastic articles mentioning “Taiwanese independence” from 2003–2022
Exploring ideas targeted for discrediting
Overall, sarcastic uses of so-called tend to precede ideas, claims, or criticism that the Chinese state finds objectionable. In some cases so-called is used to convey that the intended meaning is the opposite of the literal meaning in the text, for example, so-called respect for human rights in the US, or so-called genocide in Xinjiang. In other cases, so-called is used to signal that the concept/entity in question is not recognized or is not legitimate in the eyes of the Chinese state, for example, Taiwan’s sovereignty or Hong Kong democratic primaries.
To better understand the specific ideas that are being discredited with the so-called rhetorical device, we extracted the top ten articles for “Territorial Disputes” and “Ethnic Minorities,” topics with similar temporal trends across both English and Chinese corpora.Footnote 11 Within the sample of articles that are highly associated with the “Territorial Disputes” topic, 88 percent of the time so-called was used to discredit Japanese and Filipino maritime territorial claims and their corresponding actions. For example, so-called was used to call into question Japan’s “nationalization” of the islands, which involved its “purchase” from private Japanese citizens in 2012. The device was also used to undermine the interpretation of the United Nations Convention on the Law of the Sea, which forms the basis of the Filippino arbitration case against China. In a small number of cases in the English corpus, so-called was directed at US actions deemed illegitimate, such as “transferr[ing] the so-called ‘administrative authority’ [of the Diaoyu islands] to Japan in the 1970s,” and the staging of “so-called ‘freedom of navigation operations’ close to Chinese waters, ahead of a July 12 ruling by the Permanent Court of Arbitration (PCA) in The Hague.”
Within the sample of articles that are highly associated with the “Ethnic Minorities” topic, 75 percent of the time, so-called was used to discredit the idea that China represses its Uyghur minority in Xinjiang, including research that has found evidence of forced labor, forced sterilizations, and cultural genocide. In the remaining instances, so-called was directed at US actions, including the Uyghur Human Rights Policy Act, the Hong Kong Autonomy Act, and a “conference on so-called promoting religious freedom,” which refers to the 2019 Ministerial to Advance Religious Freedom hosted by then Secretary of State Mike Pompeo.
In both samples, US actions that opposed the CCP’s interest or supported its critics, such as freedom of navigation patrols or legislation relevant to Xinjiang or Hong Kong, were countered more prominently in English than in Chinese. While US actions perceived as hostile to China need to be countered by Xinhua’s English edition as part of the push to win international discourse power, these actions appear to be less widely publicized in Xinhua’s Chinese edition. This suggests the existence of a domestic strategy to selectively reduce the issue salience of inconvenient or sensitive foreign affairs topics, thereby contributing to research on this information management strategy in autocracies (see Carter and Carter Reference Carter and Carter2024, 185–190; Widmer Reference Widmer2024).
Detecting country mentions using word-frequency analysis
Finally, given the prevalence of the United States in these results, we wanted to dig further to compare it with other countries whose ideas or actions are discredited with the so-called device. We conducted a word-frequency analysis using multi-language geographical dictionaries in Newsmap (Watanabe Reference Watanabe2018). Figure 7 illustrates the results. Excluding China, Hong Kong, and Taiwan, across both corpora, the US was by far the most frequently mentioned country, followed by Japan.Footnote 12 In the English dataset, the US was mentioned 12,782 times, accounting for a third of all country mentions. In the Chinese dataset, the US accounted for 40 percent of all country mentions. It had 66,491 mentions, nearly four times that of second-placed Japan. It is worth noting that there are countries—Iraq, for example—that rank within the top ten, where the so-called device is often being used to criticize US foreign policy actions. These results lend further support to the conclusion that ideas and actions supported by or associated with the United States are target number 1 for the use of so-called as a propaganda device in China.
Conclusion
This research note has investigated the dynamics of the linguistic marker so-called / 所谓, usually along with inverted commas, in PRC propaganda. In sum, Xinhua not only continues to use so-called as a propaganda device, but also its frequency has increased since around 2018. It can be inferred that the propaganda apparatus sees a renewed need to de-legitimize words or actions it perceives as threatening to the party’s interests. In addition to criticisms of its core interests, the US emerges as a particular focus in Xinhua’s de-legitimating rhetoric in both samples, but especially for its Chinese-language audience. This is consistent with Garver (Reference Garver2016: 812–814), who argues that the CCP leadership sees the US as leading an ideological campaign against it and that the party aims to inoculate its population in response. In both samples the focus on the US has dramatically increased in recent years, which is consistent with Xi Jinping’s emphasis on combatting “Western” influence (Buckley Reference Buckley2013).
Given the space constraints of the research note format our study has many limitations in terms of timespan, source selection, and audience reception and is by no means the final word on this topic. Nevertheless, two brief conclusions can be drawn. First, this research note demonstrates the utility of focusing on the ideas that propagandists wish to discredit. Doing so allows a window into the political ideas that authorities find threatening to their power.
Second, in the PRC context, clearly the United States is perceived as a threatening rival that the authorities wish to delegitimize. The proportion of articles that focus on discrediting the US out of all possible topics is remarkable and appears to be increasing with time, likely partly in response to more hawkish US policy towards China in recent years and partly accelerated by Xi Jinping’s increased domestic control and repression. This research is thus consistent with arguments that China perceives the US as its main rival (Doshi Reference Doshi2021) and that Xi Jinping’s personalist power is entwined with a more nationalistic foreign policy and anti-West ethos (Shirk Reference Shirk2023). The US–China rivalry will continue to be important not only for China’s foreign relations but also its domestic politics as its population continues to be presented with material that discredits ideas associated with the US.
Data availability
The data and R scripts required to verify the reproducibility of the results in this article are available on CodeOcean at https://codeocean.com/capsule/8377953/tree.
Funding statement
Lim is supported by the Irish Research Council’s Government of Ireland Postgraduate Scholarship Programme.
Competing interests
The authors declare none.
Open access
