「杀人有两种,一种是用枪杆子杀人,一种是用笔杆子杀人。」

“There are two ways to kill people: one is with the barrel of a gun, the other is with the pen.”

– Mao Zedong 毛泽东

Introduction

How state violence acquires legitimacy within a given political order is a question that occupied major philosophers and political thinkers of the twentieth century. Grappling with this question has yielded some of the key concepts in modern violence studies, including “state of exception” (Schmitt; Agamben), “monopoly on violence” (Weber), and “distribution of the sensible” (Rancière), which together lay the groundwork for understanding how states normalize the use of force and manufacture affective compliance. The present article integrates these theoretical insights with existing historical scholarship to probe the processes of legitimation of state violence in the People’s Republic of China (PRC). Sociologists and political scientists have identified full-scale war mobilization as the primary means through which the Chinese Communist Party grasped power and ultimately took over the country in 1949 (Perry; Shambaugh, “The Soldier and the State”; Walder; Xun). Literary scholars and art historians, while certainly recognizing the power of the gun, have lent more focus to formal and cultural mechanisms which harnessed political energy for the dual purpose of proletarian revolution and nation making (B. Wang, The Sublime Figure; Ji; Pang; Mittler; Lee, Revolution of the Heart). In this context, our study takes a novel approach which combines close reading and historical analysis with quantitative conceptuality and text mining. We focus on one of the most important official publications of the PRC, the People’s Liberation Army Daily or Jiefangjun Bao 解放军报 (hereafter PLA Daily), to explore three major questions: what was the journal’s role in legitimizing state violence, how was the Chinese soldier imagined and institutionalized in this important official publication, and how were violent sentiments distributed discursively across different topics in the aftermath of the communist takeover. A conventional reading of the PLA Daily over the course of four decades would be a massive and prohibitive undertaking; the growing computational resources available to humanistic scholarship allow us to examine the representations of violence at a macroscopic scale beyond the reach of traditional methodologies.

Before we proceed, it is important to point out that ours is not a historical study of violence per se but rather of its discursive-affective legitimation. While we do draw extensively upon existing definitions of violence (Ganesan and Kim's formulation (3), “the state’s utilization of its apparatus—the military, the police, and security agencies—in order to maintain the unilaterally defined order of the society in question,” is particularly relevant to our discussion), our focus is on the vocabulary distribution in a major PRC journal and how this distribution legitimated the use of physical force in the specific historical context of post-1949 China. The violent language used in the PLA Daily carries official approval, since everything that was published must have first passed reviews and censorship.[1] This makes the PLA Daily a nonpareil resource for a historically-informed exploration of how violence and the language of violence have been mutually implicated in a feedback loop, a loop that more than once spiraled out of control, with devastating consequences.

To Serve the People: Modeling the Chinese soldier

From personal memoirs and cooking recipes to revolutionary agitation and war coverage, the PLA Daily offers a wealth of insights into the sociopolitical realities of the People’s Republic of China. One of the few major periodicals that continued publication during the Cultural Revolution (1966-1976), the paper has received relatively scant attention from scholars, aside from political science studies devoted to the PRC propaganda system (Wang and Brady; Luo; Wu and Yu; LaRocca). The establishment of the journal was part and parcel of the aggressive post-1949 drive by the Chinese Communist Party to standardize communication channels. Organized in November 1954 by order of Luo Ronghuan, the Vice-Chairman of the China Military Commision (CMC), the creation of the PLA Daily was formally approved by Mao Zedong after request by Peng Dehuai, then China’s defense minister, in the middle of 1955. It was to be a single newspaper under the auspices of the CMC to ensure ideological consistency and loyalty to the Party’s core principles across the various branches of the PLA (Zhang). Commenting on the paper’s influence, Marshal Lin Biao stated that the journal “has not only been highly influential in revolutionizing and educating the military, but has also played a major role in revolutionizing the entire country” (quoted in Lu 2).

Lin Biao’s assessment speaks to the impact exerted by the PLA Daily on the Chinese public since its first official issue on January 1, 1956. How a newspaper originally intended for the armed forces could play such a crucial role can be explained by the special position of the PLA in both the political system and the social structure of the PRC. This position crystallized when the Red Army was rebuilding in the communist capital of Yan’an during the war against Japan and then developed to systematically involve the military throughout society (Gittings, The Role of the Chinese Army 79). The systematic involvement expanded the PLA’s image and its influence while extensive political work internal to the PLA burnished the military’s revolutionary credentials (Gittings, The Role of the Chinese Army 255). Commentaries in the PLA Daily, the official army journal, were central to several political movements, including the publication of the Quotations from Chairman Mao Zedong (the Little Red Book), the “Study the PLA” initiative, and the controversy surrounding the theater piece Hai Rui Dismissed from Office, which Mao interpreted as a thinly veiled critique of his leadership (Lu). Likewise, in their article “Media, Environment of Quotations, Cult of Personality,” Ting-Jun Wu and Pin-Lian Yu emphasize the journal’s significance in promoting Mao Zedong’s cult of personality and inciting the Red Guards’ violence. In particular, they refer to an article from August 29, 1966, which appealed to the Red Guards to learn from the PLA and to “study Chairman Mao’s books, listen to Chairman Mao’s words, to act in accordance with Chairman Mao’s instructions, and to be Chairman Mao’s good soldiers” (55, emphasis added).

But what exactly does it mean to be a good soldier in the PRC? We highlight the word “soldiers” in Wu and Yu's quote because the image of the soldier is one of the most ubiquitous in PRC culture. Yet, aside from the term’s circumstantial appearance in scholarly studies of revolutionary aesthetics, there has been little treatment of the PRC soldier qua soldier; that is, as an agent of violence.[2] “Soldiers” in the preceding quote is a translation of the Chinese term zhanshi (战士), the preeminent term in the PLA Daily referring to violent actors. There are several other terms carrying similar connotations to zhanshi, including junren (军人), bing (兵), shibing (士兵), and minbing (民兵), as well as terminological combinations, such as geming junren (革命军人) or renmin zhanshi (人民战士). Rather than lumping all of them together under one single translation, “soldier,” we acknowledge the semantic and political tensions between such official and semi-official designations.

For the subsequent analysis, we have selected three of the most prominent soldier figures in the PLA Daily: zhanshi, junren, and minbing. As illustrated in Figure 1, which shows normalized frequencies of these terms in the articles published during the years 1956-1989, the three figures claimed the textual space of the journal to different degrees of success and stability. As there was only a limited number of pages and characters that could be printed in each issue, these varying trajectories reflect the evolving economy of violence in the early PRC.[3] In this context, Giorgio Agamben's theory of violence proves especially illuminating. One among many other mechanisms of what Agamben refers to as the “sovereign power,” the legitimation of violence upholds the state of exception which separates bare life, Greek zōe, from the politically meaningful way of life proper to men, bios (Agamben 191). Who can kill and who can be killed depends upon the soldier positionalities identified in the official state discourse. If we think of these positionalities as vectors in a politically defined space, then what counts is not only which vector is the largest but also how closely it is aligned with the officially recognized dimensions along which violence can be legitimately performed at any given time. The division between “the People” and “people,” or the agents and the subjects of the sovereign power, is not fixed but constantly re-established and re-negotiated on the discursive battlefield as different soldier figures vie for legitimacy.

Figure 1
Figure 1.Stack plot showing the distribution of soldier terms as a fraction of all words published each year in the PLA Daily, 1956-1989.

To explore the differences between the three soldier figures, we first identified words and expressions occurring in a statistically significant manner around them. We segmented the PLA Daily articles (218,494 texts) into individual words and sentences using the spaCy pipeline zh_core_web_lg for Python. We then used the Fisher’s exact test to identify statistically significant collocates of each soldier term (p < 0.01).[4] By considering different context sizes (5-word window, sentence, document), we aimed to cover the entire syntax-semantics spectrum. To generate the final lists of collocates, we filtered out words that co-occurred with the target terms less than 1,000 times and sorted the remaining words in decreasing order according to the ratio of the actual number of occurrences to the expected number of occurrences (r-exp), the latter calculated given the overall frequency of words in the corpus.

Although we will explore the results in more detail in the following sections, a few basic differences become evident already now from Table 1, which displays the top five terms for each category. Zhanshi conveys a romanticized and idealized image, often associated with Lei Feng 雷锋 (1940?-1962?), a model soldier revered in the PRC for his selflessness and dedication. With a deep connection to familial and rural roots, the zhanshi is deeply intertwined with communal ethos, drawing from the concepts of “Five Virtues” and “Five Togethers,” which emphasize moral excellence and solidarity, respectively. Junren leans more towards the structural and functional aspects of the military and is characterized by terms such as “demobilized,” “retired,” or “active duty,” which all carry the administrative and bureaucratic tones of the military life cycle. Minbing, finally, is grounded in the idea of a civilian militia and emphasizes collective responsibility and grassroots participation in defense matters. As illustrated in Figure 1, these three soldier positionalities are not granted equal importance. When contrasted with junren, which refers specifically to the individuals associated with the military institution that is the PLA, the much more dominant zhanshi is a floating figure burgeoning with revolutionary meaning and endowed with a higher degree of social mobility: while a junren can be a zhanshi, the opposite is not always the case. In other words, almost anyone with correct political characteristics and will to revolutionary action, from women workers in civilian militias (minbing) to teenage Red Guards and even literary figures, could be designated as a zhanshi. They thereby become implicated in the apparatus of state violence and acquire a mandate to engage in violence on behalf of the People. As such, the term zhanshi serves as a univocally positive marker which the journal utilizes to distribute the mandate for violence to its designated agents in the PLA and beyond.

Table 1.Top five most significant collocates of the three major soldier designations.
Term 5-word window nObs r-exp sentence nObs r-exp document nObs r-exp
战士
zhanshi
雷锋式 (Lei Feng-style)
五好 (Five Virtues)
超期 (extended service)
退伍 (demobilized)
警卫 (guard)
1398
11642
1144
7140
2332
24.964
19.403
16.113
13.864
13.48
五好 (Five Virtues)
退伍 (demobilized)
探家 (home visit)
家长 (parents)
入伍 (to enlist)
7784
5906
1490
2945
1004
9.779
7.992
7.641
6.913
6.877
爱兵 (love for soldiers)
五同 (Five Togethers)
尊干 (respect for officers)
同乐 (Joy Together)
农村籍 (rural hukou)
25405
5435
11396
2001
2201
6.291
6.169
6.15
6.027
5.933
军人
junren
残废 (disabled)
复员 (retired)
现役 (active duty)
退伍 (demobilized)
军属 (family dependents)
4042
7548
1642
6058
1468
336.833
260.276
149.273
123.633
112.923
残废 (disabled)
复员 (retired)
现役 (active duty)
军属 (family dependents)
退伍 (demobilized)
2697
4851
1237
1288
4244
128.42
107.8
65.105
61.333
55.117
复退 (retired)
残废 (disabled)
复员 (retired)
抚恤 (compensation)
优抚 (preferential treatment)
1729
27928
61112
2398
12370
82.333
57.822
53.326
52.13
51.116
民兵
minbing
基干 (core cadres)
三落实 (Three Implementations)
预备役 (reserve duty)
大办 (to organize big-scale)
武装部 (Armed Departments)
2906
2696
1348
1250
1338
193.733
179.733
61.273
33.784
23.474
三落实 (Three Implementations)
基干 (core cadres)
全民皆兵 (Everyone’s a Soldier)
武装部 (Armed Departments)
预备役 (reserve duty)
2737
2738
1132
11668
1409
97.75
70.205
43.538
38.508
31.311
三落实 (Three Implementations)
整组 (reorganization)
劳武 (labor and military)
民兵营 (militia battalion)
民兵团 (militia regiment)
57576
10133
20975
4927
12572
75.658
71.865
61.152
57.291
55.628

Between Junren and Zhanshi

Some of the earliest usages of junren can be traced back to the late nineteenth century and occur in the context of military westernization/modernization (Lo 5). The term is generally seen in similar contexts throughout the Republican Era, both in the nationalist Kuomintang (GMD) and Communist Party (CCP) military milieus. In the 1950s, junren appears in the context of the PLA’s attempts to professionalize along the lines of the Soviet Union’s armed forces (Li 114), which can be also seen in many PLA Daily articles published during this time:

“The more advanced technology and weaponry become, and the greater its power, the greater the demands on every military member’s understanding, and the more solidarity is needed; every soldier [junren],[5] no matter what job they do, needs to be more aware of the responsibilities they themselves have toward the entire enterprise and understand their role in the larger system” (“The Development of Military Technology and the Role of Humans in Modern Warfare,” PLA Daily, No. 8-29, 1958).[6]

Junren appears also in the context of the rational management of the military and current and former military service-members; most common are texts related to the systematic incorporation of the various formal, semi-formal, and informal groups of fighters from World War Two, the Civil War, and the Korean War, into the PLA or into civilian work units, a process in which these soldiers were known as fuyuan junren (复员军人).[7] Furthermore, many of the articles referencing junren from the 1950s and early 1960s deal with the dependents of servicemembers. For example, an article from January 14, 1956, focuses on the family life of military officers. While it is concerned predominantly with ensuring that the families of officers adhere to socialist family values, what is of particular importance here is the recognition of the “military officer [junguan 军官] as a long-term professional [changqi zhiye 长期职业] working in the armed forces” (“Effectively Supporting the Families of Military Officers,” PLA Daily, No. 1-14, 1956).[8] Mentions of junren in articles during the 1950s generally follow this trend of increasing professionalization and rational management of the PLA.

As noted by Li, however, during the 1950s a “widening gap” emerged between “an increasingly radicalized society and a gradually regularized military” (114). While junren was primarily associated with the modernization and professionalization of the military during the 1950s, the appearance of the affectively charged zhanshi in the same period reveals the radicalization that Li refers to but also the mounting tensions within official discourse as to who can be considered the legitimate agent of state violence (Figure 1). While Shambaugh (“The Soldier and the State” 533) argued that “[p]rofessionalism and politicization are not necessarily mutually exclusive in any army, including the Chinese,” the divergence of the professional junren and the romanticized zhanshi becomes increasingly palpable during the most radical decades of the 1960s and 1970s. This radicalization was partly the result of the gradual takeover of the PLA by Mao Zedong’s allies, most prominently the ascension in 1959 of Lin Biao to Minister of Defense in place of Peng Dehuai, the latter of whom had consistently advocated for professionalization.[9]

Zhanshi (战士) is a much more venerable term than junren, appearing notably in the frontier poems of the Tang Dynasty, where it invoked a varied pathos of bravery, virtue, and sacrifice on behalf of the country.[10] In modern times, the term begins to appear in a similar romanticized and politicized context toward the end of the nineteenth century. News reports on the First Sino-Japanese War (1894-1895), for example, refer to soldiers fighting on behalf of the Chinese nation as zhanshi.[11] The term takes on revolutionary connotations during the nationalist uprisings in the first decade of the twentieth century and after the establishment of the Republic in 1912. While both junren and zhanshi exemplified an ideal citizenry, a distinctive feature of the zhanshi was its emphasis on romanticized death in the service of the revolution and the nation.

The intertwining of romanticized death with the zhanshi, and the veneration of martyrs in general, found resonance within Mao Zedong’s eschatological Communist ideology, exemplified most prominently in his eulogy to the soldier Zhang Side 张思德 (1915-1944). Titled “Serve the People,” Mao’s celebrated speech from September 8, 1944 “marked the inception of the cult of the red martyr in the history of the CCP” (Hung 280). In the PLA Daily, the association between the Chinese zhanshi and martyrdom on behalf of the nation and revolution is clear for the period of the late 1950s through the end of the 1970s. For example, an article from April 30, 1959 quotes Mao’s eulogy and commemorates the martyr An Yemin 安业民 (1938-1958), who had sacrificed himself to save an artillery piece during the Second Taiwan Strait Crisis in 1958, as “the figure of a truly sublime communist soldier [zhanshi].”[12] The author is quick to add, however, that An was also a farmer’s boy with an elementary school education; he was a “normal” person and an “everyday” zhanshi, and as such could be emulated by virtually anyone. This agrees with Louise Edwards's argument that “China’s contribution to the socialist realist theory of exemplars was to bring the exemplar directly into the present world of the consumers of the narrative… close to the people” (26). The celebration of An Yemin conveys the intention of the PLA Daily’s editors to create a tangible relationship between the zhanshi and the reader, whether that be a soldier in the PLA or a commoner.

As the distance between the martyr and the average person closed, model soldiers became only the most conspicuous and well-respected zhanshi. On the back of this highly mobile figure, the mandate for legitimate violence spread beyond members of the regular state military apparatus. For example, one article uses the term in reference to the writer Lu Xun (“Learning from Mr. Lu Xun’s Fighting Spirit,” PLA Daily, No. 10-20, 1956). Another relates the story of a mother and a sister who “not only lost their own family members in the long and arduous revolutionary struggle, but also were zhanshi in their own right,” since they served various civilian roles during the communist revolution (“Glorious Mothers,” PLA Daily, No. 11-13, 1956). PLA Daily also uses the term to refer to members of the people’s militias, as in an article which portrays the stalwart minbing zhanshi as “loyal to the motherland and loyal to the revolutionary enterprise” (“Exemplary Figures in Militia Battalions,” PLA Daily, No. 8-13, 1964).

Yet the increasingly prominent label was available not only to those coming from the ranks of the educated elite, those who had lost loved ones in military conflict, or the junren members of the PLA. Ultimately, anyone who sacrificed themselves on behalf of the nation could be described as and, perhaps most crucially, describe themselves as a zhanshi. A paean from September 10, 1958 glorifies an entire zhanshi nation:

“…Should an enemy invade the Motherland, the zhanshi of the Steel Army will emerge from beside the blast furnaces, from the cotton mills, from the fields, from the millstones, from schools, from shops… from every corner of the Motherland. They will unite into a powerful fighting force and, from all directions, pierce the hearts of the enemy with their deadly bullets!” (“Everyone’s a Soldier,” PLA Daily, No. 9-10, 1958)

Here, virtually everyone fighting on behalf of the Motherland is a soldier, military member or not. As the Cultural Revolution gets underway, the zhanshi’s mobility combines with the increasing concentration of command over the legitimate use of force within the person of Chairman Mao to give rise to many of the violent mass movements of the era, including that of the Red Guards (hongweibing 红卫兵), the zealous teenage adherents of Mao Zedong Thought who answered his call to violently attack class enemies allegedly still lurking among the People (Thurston, Enemies; Yang). In an article entitled “Always Following Chairman Mao to Wage Revolution” from September 2, 1966, Red Guards are quoted as stating that “[w]e must hold high the great banner of Mao Zedong Thought, study the PLA in great earnest, become Red Guard zhanshi who live up to the title bestowed by the great Commander-in-Chief Chairman Mao, and forever serve as faithful servants of the People.”[13]

During the period that the zhanshi was in the ascendant, from the increasing politicization of the PLA in the late 1950s to the end of the Mao era in the late 1970s, there are also articles dealing with the junren, albeit they are much less frequent. Different from the focus in the 1950s on the management of retired veterans and/or military families, or the reincorporation of veterans into new military units, the emphasis on junren in the 1960s and 1970s is on reinforcing the glorious image of veterans, their families, and, by association, the PLA and its revolutionary accomplishments of the past. This was done through mass campaigns, referred to variously as “embrace the military and privilege military dependents” (yongjun youshu 拥军优属) and “privilege and care work” (youfu gongzuo 优抚工作), among other terms (Table 1). At the same time that the PLA Daily led these mass campaigns to burnish the revolutionary image of junren, the military dependents, and the PLA in general, it viewed junren and other elements of the PLA bureaucracy as potential sources of revolutionary force in and of themselves.[14] Yet even in the midst of heated revolution and attempts to mobilize the junren and its associated groups, the term appears to maintain its association with military organization and lacks strong affective connotation; in other words, during this era junren remains for the most part professional as opposed to political.

As we move to the late 1970s and 1980s, there are still a large number of articles relating the term junren to the topics commonly associated with it in previous eras, such as military dependents and veterans. The junren is also still associated with military professionalization and modernization. Noteworthy during this period, however, is the junren’s increasing politicization, as it begins to take on some of the romantic characteristics of zhanshi from the Mao era. In what is a solid confirmation of Shambaugh’s statement that professionalization and politicization are not mutually exclusive in the PLA (Shambaugh, “China’s Military”), there are a large number of articles from this period focused on the spirit and character of the junren and the role that they are to play in and for society. For example, an article from September 1980 calls on soldiers and cadres to avoid using foul language (“Be Civilized, Polite Soldiers,” PLA Daily, No. 9-1, 1980). No longer is the figure of the junren simply an individual in a military bureaucracy. Here, at the beginning of “Reform and Opening Up,” junren is also taking on a new moral function in society. This moralized figure of the junren is apparent in many other articles as well, such as one which references the ongoing military conflict between the PRC and Vietnam: “Sacrificing Oneself in the Interest of the People is a Soldier’s Most Sublime Moral Precept” (No 12-10, 1985). Quoting Sunzi’s The Art of War, the text lays out the essence of what the “soldier of the new era” (xin shiqi junren 新时期军人) is supposed to be:

“The profession of the soldier [junren] is warfighting. War is an act of violence. It is ‘the realm of life or death, the path of being or nonbeing.’ War has always been closely linked to valor. As such, from the first day He joins the military, a soldier must be mentally prepared to bravely sacrifice himself to protect the Motherland.”

And yet, despite this new ethical content of the junren in the 1980s and its thrust toward nationalized sacrifice, its affective potential still pales in comparison with the zhanshi. This is particularly striking in the articles dealing with the Tiananmen Square massacre of 1989. When mentioning members of the PLA who fought against the demonstrators and/or who were injured or killed during the conflict, the journal commonly refers to those soldiers not as junren, a figure close at hand and imbued with the necessary politicized connotations, but as zhanshi:

“…a small group of heartless and fanatical criminals, along with rioters… brazenly intercepted and destroyed over a hundred military vehicles, setting fire to and burning thirty-one of them. Several young zhanshi were burned alive” (“Martial Law Troops Suffer Savage Attacks from Malicious Assailants; Forced to Take Decisive Measures after Bearing the Unbearable,” PLA Daily, No. 6-4, 1989).

Considering the dynamic landscape of violent agency from the founding of the PLA Daily in 1956 to the Tiananmen Square massacre of June 4th, 1989, it becomes evident that the legitimation for this violent agency is closely tied to the shifting positionalities of the figures of the junren and the zhanshi within the PLA Daily. In this context, the journal emerges as a key player in legitimizing a spectrum of violent actors, all fervently dedicated to the cause of the People and their struggle against perceived adversaries. One can see how the PLA Daily’s legitimation of the zhanshi, with its social mobility and its more radical violent potential, stimulated the excessive civil violence of the Cultural Revolution years: anyone could take on the mantle of the People’s zhanshi and mete out fury on those people deemed enemies. It is here, in an almost total state of exception, that Agamben's civil war between “people” and “the People” reaches its apogee. No wonder, then, that the PLA Daily attempts in the 1980s to walk back its legitimation of the zhanshi in favor of the junren as the bearer of the PRC state’s mandate for violence. With the ascent of the junren, violence on behalf of the People is safely sequestered within the ranks of the PLA, a bureaucratic military organization under the Party’s control. The zhanshi, that infinitely more volatile figure of the soldier, is thereafter kept on reserve, to be unleashed whenever the simple patriotism and professional ethic of the junren prove insufficient to spur soldiers and civilians into violent action.

The Rise and Fall of the People’s Militia

Having outlined a genealogy of the two major soldier designations, junren and zhanshi, we now move on to the enigmatic figure of the People’s Militia (minbing). Here, we focus only on the militia’s accession to the status of legitimate agents of violence during the late Cultural Revolution as well as their subsequent fall from grace in the immediate post-Mao period, as seen in Figure 1; the militia’s role in the PRC’s revolutionary history calls for an independent, full-length examination.

June Dreyer's remains one of the few scholarly treatments of the subject available to date. As she puts it, although in theory “the militia member is both civilian and soldier – a productive member of society in peacetime and a valiant warrior in time of hostilities,” in fact the militia’s role has been much more complex in China (71). Located between people and the People, the liminal status of the militia was epitomized in party slogans describing militia members as striding forward “with a shovel on one shoulder and a rifle on the other” (quoted in Dreyer 71). Despite his initial misgivings about the local militias as remnants of the Nationalist regime, Mao Zedong’s concept of the People’s War (renmin zhanzheng 人民战争), the “everyone’s a soldier” campaign (quanmin jiebing 全民皆兵), and the “three implementations” directive (san luoshi 三落实) emphasized the militia’s role in increasing domestic production and military matters (J. C. F. Wang 542). According to some estimates, approximately one-third of China’s entire population enrolled in militia organizations (McCord 125). The “sea of militiamen” was supposed to protect the country against nuclear attacks from the West (Powell; Gittings, “China’s Militia” 104–06). These developments enhanced the role of the militia, albeit not in a way that emphasized professional training, thus placing them in tension with the regular armed forces (Gittings, “China’s Militia” 107–09). Already in 1960, Ralph Powell noticed that “since the militia is an ‘organization of the masses’ rather than a professional army, the use of military ranks, uniforms, and salutes is frowned upon” (104). Similarly, Dreyer argues that “the greater the defense role given to an armed citizenry, the less the prestige of the regular army – in particular, the less the prestige of a professionally oriented, highly specialized army equipped with modern weapons” (74).

The rapid increase of the militia’s presence in the PLA Daily (Figure 1) in the 1970s speaks to the editors’ recognition of the militia’s importance, but also the militia’s own growing claims to legitimacy as agents of violence blessed by Mao Zedong’s official endorsements (da ban minbing 大办民兵), which could be traced back to his celebrated report on Hunan peasantry praising the outbursts of revolutionary violence among the rural poor as “necessary and liberating” (Walder 43). The militia’s growth posed a challenge to the PLA, however. Close reading of the minbing-related articles published between 1970 and 1976 reveals that the PLA Daily’s editors were at pains to distinguish between the militia and the regular army, as in the following Tibet-focused excerpt:

“The militiaman Tsering Dorje found an uninhabited mountain pass [in the Kailash Mountain Range], over 60 kilometers away from his home. He encouraged seven families of freed serfs to settle there. One early morning, his wife was alerted by a dog’s bark. With her two sons, she followed the sound and discovered a deep footprint near a mountain spring. Soon, all mountain passes and critical routes were secured by border troops [zhanshi] and militia [minbing]. After a relentless 24-hour search, they captured a counter-revolutionary hiding in a rock crevice.” (“Promote the Militia, Every Citizen a Soldier,” PLA Daily, No. 9-29, 1973)

In such quasi-novelistic passages, the official zhanshi designation is clearly separated from the minbing, despite the overall emphasis placed on the collaboration between the two groups, united against a common enemy, and despite the occasional labeling of the militia as zhanshi, which we explored in the previous section.

In addition to the patronizing distinctions, which implicitly emphasized the higher status of the regular armed forces, the PLA also employed more direct means to bring the violent minbing under control. Already in November 1964, a PLA Daily editorial warned that “if political implementation is not fulfilled […] we will not be able to ensure that the guns are entirely in the hands of reliable workers and poor and middle peasants” (“Three Implementations for Enhancing Militia Work During the Socialist Education Movement,” PLA Daily, No. 11-17, 1964). Similar announcements had next to no effect in the midst of revolutionary turmoil, however. The rebel leaders in major cities declared that the local militias were ready to seize military control from the army (Walder 297). The tug of war continued for another decade (J. C. F. Wang 544–47); an editorial published jointly on September 29, 1973 in the People’s Daily and the PLA Daily called again for local party committees to reassert direct control over the armed civilians: “to properly establish the militia, the key is to get the policy right” (“Organizing the Militia Well,” PLA Daily, No. 9-29, 1973), but soon afterwards Wang Hongwen, Zhang Chunqiao, and other party radicals’ support for the People’s Militia regained momentum; based on various trade union federations and municipal party committees in which they were personally involved, “a command structure somewhat independent from the PLA” (J. C. F. Wang 550) was established and a change in the PRC constitution elevated the militia to parity with the PLA. Specifically, Article 15 of the PRC Constitution of 1975 reads: “The Chinese PLA and the militia [minbing] are the workers’ and peasants’ own armed forces led by the Communist Party of China.” This officially acknowledged militia was instrumental in suppressing the Tiananmen demonstrations in April 1976 sparked by the death of Premier Zhou Enlai 周恩来 (Dittmer 53; J. C. F. Wang 553).

The trajectory of the militia’s prominence in the PRC’s socio-political sphere, particularly as narrated in the PLA Daily, underscores the tension and the balance of power between grassroots mobilization and centralized authority in the Chinese Communist narrative. The frequent distinctions between the militia and the PLA in official publications reflect a cautious endorsement of the former (supported, after all, by none other but Mao Zedong), while simultaneously keeping them distinct from the regular army. The sheer fact that the militia could be considered legitimate agents of violence did not diminish but rather increased the potential for tensions between the armed civilians and the professionally trained soldiers. With the growing professionalization of the PLA’s armed forces and the ascent of the junren as the recognized agent of violence in the 1980s, Mao Zedong’s passing also marked the decline of the minbing.

Distribution of violence: topic modeling the PLA Daily

If the foregoing discussion focused on different soldier figures as agents of violence competing for legitimacy, this final section proceeds to analyze the larger discursive-affective contexts in which those figures necessarily participated. Here again, the concept of distribution remains crucial. Jacques Rancière has coined the term “distribution of the sensible” (partage du sensible) to emphasize the aesthetic dimensions of political power; the organization of sensory experiences determines what is visible, audible, and thinkable within the social sphere and thereby shapes the possibilities for action and participation (Rancière 12). The term refers at once to the “conditions for sharing that establish the contours of a collectivity (i.e., ‘partager’ as sharing) and to the sources of disruption or dissensus of that same order (i.e. ‘partager’ as separating)” (Panagia 103). This echoes Agamben's notion of “civil war” between the included “People” and the excluded “people,” which both divides society and yet “keeps it united and constitutes it more securely than any identity” (Agamben 191). The following analysis operationalizes this dual perspective, identifying themes in the PLA Daily which designated the legitimate objects of violent actions and thereby yielded themselves to violent participation.

Our first step was to collect a list of terms indicative of whether a given text could be considered violent or not. We were interested above all in state power and military-related violence, thus in vocabulary related to skirmishes, battles, and full-fledged army operations, such as 屠殺 (“to slaughter”), 攻打 (“to assault”), 灾难 (“disaster”), 烧毁 (“to burn down”), 毒死 (“to poison to death”).[15] In the next step, we identified articles in which such vocabulary was particularly prominent (at least one violent word per ten words on average) and labeled them as “violent.” One representative passage reads as follows:

Under the command of the right-wing military authorities in Indonesia, fully armed army troops, together with thousands of rioters armed with knives, axes, clubs, and other weapons, brutally beat, tortured, and bloodily massacred innocent and unarmed Chinese expatriates. Not even the elderly, weak, women, and children were spared. The atrocities of right-wing soldiers and rioters raping Chinese women continue to occur everywhere. In October last year, right-wing soldiers and rioters in Qibili [齐必力, Kabil?] forced all overseas Chinese who had once hung the flag of the People’s Republic of China to gather together. They beat them, stole all their possessions, and cruelly murdered local Chinese community leaders (“Indonesian Right-Wing Forces Owe the Chinese People a Heavy Blood Debt,” PLA Daily, No. 5-19, 1966).

We randomly selected 100 violent articles from each year and transformed them into vector representations (embeddings) with the help of a large encoder model. These representations were then averaged into “vectors of violence” through k-means clustering.[16] In the next step, we computed cosine similarities between the vectors of violence and the individual embeddings of all articles published between 1956 and 1989.[17]

Figure 2
Figure 2.Monthly average of cosine similarities between the vector representations of PLA Daily’s articles and the vectors of violence.

As shown in Figure 2, the 34 years in question are characterized by cycles of “repression and relaxation” (He), the graph bearing witness to PRC society’s painful journey through a series of movements, mobilizations, and purges. We see a steady growth of violent language from 1956 all the way to the Cultural Revolution beginning in 1966, and then a series of peaks in the years that follow. The intra-party conflicts surrounding Deng Xiaoping’s “three directives” (san xiang zhishi 三项指示) in the early months of 1976,[18] the protests in Tiananmen Square in April 1976, the Sino-Vietnamese conflict (1979), and the June Fourth demonstrations in 1989 are also strongly marked by violent discourse. On average, the 1980s are charged with militaristic language much less than the preceding decades. That the model can identify the relevant periods in conformity with existing historiography is reassuring; on the other hand, the fact that both the years of 1976 and 1979 are marked with violent sentiments on a par with, and at times exceeding, the intensity of the Cultural Revolution has been rarely acknowledged in literature and calls for further research.

We decided to probe further behind this familiar narrative. Although two texts produced at different times might be equally violent according to their placement in the vector space, the violent charge might be in fact distributed unevenly among different themes. In the next step, we took inspiration from topic modeling, a text mining technique which considers a text or a set of texts as an assemblage of discourses, each of which is essentially a group of words that tend to occur together. As such, topic modeling is a way of partitioning the textual space and aligns with the distributional perspectives of Rancière and Agamben. We used the MALLET implementation of LDA (Latent Dirichlet Allocation) to train a topic model on the selected pre-processed PLA Daily articles (150,988 documents containing 60,029,332 words in total), with 200 topics, 1,500 iterations, and default hyperparameters.[19] After training, we calculated the contribution of each topic to the total amount of violence illustrated in the previous figure. The color intensity in each cell of the resulting heatmap (Figure 3) signifies the weighted average violence score of a specific topic for a particular month.[20]

Figure 3
Figure 3.Temporal analysis of violence-related language by topic (1956-1989). The color intensity in each cell indicates the weighted average violence score of a specific topic for a particular month, normalized against the total presence of the given topic in the given month. Only selected topics are displayed.

Our visualization demonstrates that violence, represented as a single line in Figure 2, is in fact a distribution across different topics and different time scales. Most prominent are themes with long-standing connection to violence, which can be recognized in the graph as long, horizontal stripes of red and black. Such themes converge around general terms related to war making (topic 15), strategic and combative activities (197), military equipment and exercises (15, 103, 191), as well as ideological and political conflicts (194). The space of violence is also claimed by more ephemeral, semi-violent topics, such as the Vietnam War (61) and the Arab-Israeli (137) conflicts. During the mid-phase of the Cultural Revolution (1968-1973), one observes a surge in violent discourse revolving around armed confrontations in Laos, Indonesia, Cambodia, the Korean peninsula, India, Latin America, Western Europe, and the Eastern Bloc. Violence is redistributed by attaching militaristic descriptions to the international and global themes and removing such vocabulary from the more local ones (12, 22, 57, 155, 171, 172, 177). Such a pattern of distribution might indicate a strategic redirection of narrative focus: by highlighting international conflicts and lessening the emphasis on internal strife, the PLA Daily portrays the internal revolutionary discourse as being under the protective and guiding hand of the military leadership. The consistent violent associations with terms like “the People” and “the enemy,” by contrast, reinforce the notion of external threats and further validate the need for a unified internal front (Table 2).

Table 2.Examples of high-violence, mid-violence, and low-violence topics with the top 10 most representative terms.
High-Violence Topics
103 敌人 (“the enemy”), 战斗 (“combat”), 阵地 (“position”), 战士 (“warrior”), 高地 (“high ground”), 手榴弹 (“grenade”), 炮弹 (“shell”), 炮火 (“artillery fire”), 消灭 (“eliminate”), 同志 (“comrade”)
132 敌人 (“the enemy”), 武装 (“armed”), 人民 (“people”), 力量 (“power”), 解放 (“liberation”), 地区 (“area”), 军民 (“military and civilians”), 敌军 (“enemy army”), 伪军 (“puppet army”), 进攻 (“attack”)
Mid-Violence Topics
87 印度 (“India”), 中国 (“China”), 政府 (“government”), 边界 (“border”), 巴基斯坦 (“Pakistan”), 领土 (“territory”), 尼赫鲁 (“Nehru”), 地区 (“area”), 进行 (“conduct”), 照会 (“inquiry”)
122 老挝 (“Laos”), 爱国 (“patriotic”), 泰国 (“Thailand"), 政府 (“government"), 日内瓦 (“Geneva”), 万象 (“Vientiane”), 战线党 (“Pathet Lao”), 民族 (“national”), 中立 (“neutral”), 集团 (“group”)
Low-Violence Topics
12 技术 (“technology”), 革新 (“innovation”), 成功 (“success”), 试验 (“test”), 设计 (“design”), 设备 (“equipment”), 提高 (“improve”), 创造 (“create”), 试制 (“trial production”), 制造 (“manufacture”)
155 全国 (“nationwide”), 中央 (“central”), 副主席 (“vice chairman”), 中共 (“Communist Party of China”), 委员 (“member”), 政协 (“Political Consultative Conference”), 代表 (“representative”), 今天 (“today”), 人民 (“people”), 委员会 (“committee”)

Another important insight from Figure 3 is the totalization of violence observable during the Cultural Revolution, as nearly all topics detected by the model become violent and contribute to the high overall score of that period. This can be seen in the form of red vertical columns traversing the entire graph. In such violent circumstances, virtually anything that can be talked or even thought about acquires violent connotations and becomes surrounded with violent vocabulary. While apparently in agreement with the Maoist paradigm of “continuous revolution,” this all-encompassing upsurge in violence paradoxically reflected the CCP leadership’s decreasing ability to control how violence was distributed (Ji 139–50).

Internationalization of violent discourse during the Cultural Revolution

The expediency of the attention shift away from the local developments towards the international stage has not been sufficiently emphasized by scholars. It has been often assumed that from its very inception the CCP was “internationalist in premise and in promise” (Kirby 874), in accordance with the Marxist-Maoist ideology and the CCP’s push for “Eastern revolution” meant to redefine the rules of international relations (Dirlik; Levenson 1–18; Volland 9–15; B. Wang, “Art”). As noted by Jian Chen (11), however, “revolutionary foreign policy helped to make Mao’s various state and societal transformation programs powerful unifying and national themes supplanting many local, regional, or factional concerns.” The idealized accounts of socialist internationalism as a “global struggle against capital” (Tyerman 7, 36) and an attempt to build “a socialist world economy stretching from Berlin to Canton” (Kirby 870) frequently overlook the pragmatic aspects of internationalist imagination. We suggest that in addition to its well-documented role in furthering domestic mobilization, internationalism served also as a rhetorical-distributional move that pushed the images of destabilizing violence towards remote lands located far from the center, not unlike the more recent Wolf Warrior 2 imagery contrasting domestic safety with overseas violence. This strategy was all the more important given that China’s armed forces were increasingly beset by factionalism in the late 1960s, the main line PLA forces and the local militia-driven PADs (People’s Armed Departments, or renmin wuzhuang bu 人民武装部) finding themselves at odds with each other (Guoqiang and Walder 5–6, 83; Dreyer 67–68).

For example, in mid-1967 an “unprincipled civil war” (Bridgham 18) engulfed the PRC and the antagonism and hostility increased between the PLA and the revolutionary left. Instead of supporting army commanders, revolutionary mass organizations in the provinces and major cities violently attacked local military control committees. On May 7, Mao launched an intensive rectification campaign within the army; clashes between rival factions disrupted several major rail lines, and the Wuhan Incident unfolded in mid-July, where the PLA-backed “Million Heroes” (bai wan xiongshi 百万雄师) and the Red Guards-backed “Wuhan Workers” (gongren zongbu 工人总部), both counting hundreds of thousands of members, engaged in an intense propaganda war and military confrontations; Red Flag editorial on July 31 fueled military sentiments, appealing to “bring down the capitalist roaders” in the PLA. Large-scale conflicts occurred in Beijing, Chongqing, Hangzhou, and other big cities (Thurston, “Urban Violence”). Such violent circumstances at home seemed of little importance to the PLA Daily’s editors, however. According to our model, some of the most violent articles published between May and July 1967 focus on events happening far from major PRC agglomerations, denouncing US imperialism and the Western incursions into PRC territory. In the following years, there are also hundreds of articles related to significant victories and developments in the military engagements in Southeast Asia and Palestine:

“Following the teachings of our great leader Chairman Mao, we, the young Red Guards, vow to be the strong backing for the people of the three countries of Indochina, firmly supporting the people of Cambodia and other Indochina countries in their war of resistance against America to save their nations, firmly supporting the American people in their revolutionary struggle against the Nixon administration, and firmly supporting the righteous struggle of the people all over the world against American imperialism!” (“Chairman Mao’s Solemn Declaration is the Highest Order to the Red Guards,” PLA Daily, No. 5-22, 1970).

Such internationalist coverage was disconnected from the developments within the PRC, however, where local, unofficial outlets brimmed with uncompromising denunciations of opposing factions. Writing about the revolution in Chongqing, Guobin Yang in his Red Guard Generation notes that “factional battles were intense and deadly in many places, […] involving the use of rifles, heavy artillery, and even tanks and military battleships” (41). The rebels would seize machinery factories to produce makeshift weapons and vehicles and then deploy them against their enemies. Across the country, different military groupings engaged in public accusations of their opponents and commemorated their own dead through public memorial services and newspaper pages devoted to the martyrs (Ji 180–218). Condemning the cruelty of their enemies, Red Guards would intersperse violent accounts of battles with Mao’s quotations, vying to prove their revolutionary credentials. The violence typical of such texts, however, comes from a completely different distribution than the one that characterizes official publications such as the PLA Daily.

Thus, while propagating an anti-colonial ethos and championing global justice, the official narratives published in the PLA Daily simultaneously served an important rhetorical purpose at odds with the happenings inside the country. A computational analysis which allows for the observation of shifts in the topical distribution of violence over time makes such narrative strategies more easily identifiable when combined with traditional historiographical approaches.

Conclusion

In this article, we use quantification as an exploratory and conceptual tool; not to arrive at a final conclusion as to which day and which hour in modern China’s history was most violent, but rather to suggest a distributional way of thinking about state violence and its legitimation. Quantitative approaches to texts such as the PLA Daily offer unique perspectives on the distribution of violence as a way of legitimizing distinct forms, agents, and targets of state power. We show how the three different soldier vectors: junren, zhanshi, and minbing acquired political legitimacy to engage in violence on behalf of the People: whereas the junren in the 1950s conveyed a focus on professional violence, it was relegated to a secondary role by the romanticized zhanshi during the Cultural Revolution, only to regain prominence after Mao Zedong’s death and the rise of new nationalism in China. Minbing, on the other hand, vacillated precariously between the two in an uneasy relationship with the regular armed forces. In addition to tracing such changes genealogically, we also show that violence in the PLA Daily has been distributed hierarchically: the “people vs. the People” narrative concentrated military sentiment throughout the analyzed period, whereas international topics manifested only intermittently. The selective reportage in the PLA Daily during the Cultural Revolution was not merely a reflection of global events but a distributional maneuver to shape domestic perceptions. Further big data research is necessary to examine similar narrative strategies in other sources.

Peer reviewers: Marilyn Levin, Henrike Rudolph
Dataverse repository: https://doi.org/10.7910/DVN/DLKIAC


  1. Likewise, the choice of the journal as our primary source delimits the scope of violent actors considered in this article. We are not dealing with thugs for hire, for example, as they are not part of the state apparatus; we are not dealing with police either, as the PLA Daily is about the military, not police.

  2. In addition to the numerous studies of model soldiers such as Lei Feng and Ouyang Hai, treatments focusing specifically on the image of the soldier or the PLA in PRC culture are rare. Some recent examples include Haiyan Lee’s studies “The Charisma of Power” and “The Soft Power of the Constant Soldier”; Yizhong Gu's dissertation “The Myth of Voluntary Death,” which deals with sacrifice and martyrdom in Maoist films; Xiaomei Chen’s chapter “Worker-Peasant-Soldier Literature,” which surveys several Maoist literary works written by and/or about communist soldiers; and Ban Wang’s chapter “Art, Politics, and Internationalism in Korean War Films.” While these latter studies address the image of the soldier or the PLA, none of them focus on the relationship between this image and the management of violence. While Yan Xu's work The Soldier Image and State-Building in Modern China, 1924-1945 addresses the relationship between state ideology and the soldier, Xu's focus is less on the question of violence than the model for conceptualizing the sociopolitical structure of modern China embodied in the military. Other recent works, such as Diana Lary's China’s Civil War, Xiaobing Li's A History of the Modern Chinese Army, and Neil Diamant's Embattled Glory also address the relationship between the PLA soldier and broader PRC culture and society, yet they do not directly concern the relationship between the Chinese soldier and the management of violence.

  3. In this sense, our argument is comparable to Alex Woloch's analysis of how narratives distribute limited textual space attention among literary characters (Woloch 2004).

  4. The test determines the statistical significance of an event: the less likely the event (the lower the p value) the more significant its occurrence. For instance, if an otherwise highly frequent term appears together only with the junren, this will count as a statistically unlikely, and hence significant, event, since the null hypothesis assumes that all words are distributed randomly across the corpus. This technique has been successfully employed in DH projects (Algee-Hewitt et al. 2020). To reduce the amount of memory required to generate the co-occurrence matrices, we have considered only the words that appear at least 10 times in the whole corpus.

  5. The Chinese for “military member” here is jundui quanti renyuan 军队全体人员; this is terminologically not the same as junren, but the equivalence is implied in the context of the subsequent lines.

  6. Here and in the following quotations from PLA Daily, we provide the article’s title together with the date of publication for easier identification, as we do not have the exact page numbers. The texts have been acquired from the digitized version of the journal available through the East View library (https://dlib.eastview.com).

  7. In his study Embattled Glory, Neil Diamant translates this latter term as “retired veterans” (Diamant; Vogel 54–59). For an example of this process in the PLA Daily, consider the title of an article from April 6, 1957: “All Provinces and Municipalities Actively Prepare Reintegration Work - Welcome Retired Veteran [fuyuan junren] Participation in Rural Construction.”

  8. While the terms used here, junguan, is not junren, the military officer (junguan) is the managerial stratum of the junren-based military apparatus.

  9. For a more detailed discussion of the internal political mechanisms by which the PLA was slowly radicalized at the end of the 1950s and during the early 1960s, see (Shambaugh, “The Soldier and the State” 538–39) and (Walder 196–97).

  10. For an overview of Tang Dynasty frontier poetry in English, see the discussion in Owen 1996, 459–77.

  11. A March 16th, 1895 article from the influential Shenbao titled “Remembering Aid for the Troops [zhanshi]” points out that the “valiant” soldiers wounded on the battlefield are “loyal and righteous members of the Chinese nation” and that monetary contributions to provide medical services for the wounded guarantee “that the vast expanse of China… can be preserved” 戰士銘恩," "…我兵奮勇直前多有受傷之人… 查此受傷之人皆我華忠義之士… 我中國幅之廣人物之多茍能似此存心何患不能挽回…

  12. “…Comrades, from here we witness the figure of a truly sublime communist soldier [zhanshi]!… Placing the interests of the Party and the People above one’s own, caring for the revolution, the collective, and comrades more than for oneself, these are the fine virtues instilled in the People’s soldiers [zhanshi] by the Party’s teachings; valiant and unyielding, with an indomitable spirit, aiming to overpower all enemies and never submitting to any enemy, this is the glorious tradition of the People’s army under the leadership of the Party” (“The Figure of a Sublime Communist Soldier,” PLA Daily, No. 4-30, 1959).

  13. “Always Following Chairman Mao to Wage Revolution; Implementing the Sixteen Points Unswervingly and Fully,” PLA Daily, No. 9-2, 1966.

  14. “Calling on the Families of Martyrs and Demobilized Soldiers to Eagerly Study Chairman Mao’s Works,” PLA Daily, No. 7-31, 1966.

  15. We have further augmented this list with semantically-proximate words identified through static word embeddings (word2vec) trained on our corpus. The final list includes 815 unique terms.

  16. The semantic content of each article labeled “violent” was transformed into a 768-dimensional vector representation by accessing the [CLS] special token in the top layer of the pre-trained Chinese BERT model (bert-base-chinese, 110M parameters). Prior to generating the embeddings, the model has been fine-tuned for one epoch on a binary classification task (violent/non-violent). By taking an average of all such representations (embeddings), we arrived at a single vector of violence, which, in theory, could be used to quantify how violent a given text is: the higher the similarity between this violence prototype and the vector representation of some other text, the more violent the text. Since violence appears in many forms and shapes, however, such an approach would necessarily lose nuance. We adopted the multiple prototypes approach, clustering similar embeddings with the help of elbow method and using the centroids of these clusters as representative vectors (prototypes) for each cluster.

  17. For each article, we calculated the cosine similarities between its vector representation and the violence prototypes, selecting only the highest score, which then contributed to the average violence for any given month. We averaged the vector representations for each article longer than 510 characters (the maximum input for bert-base-chinese, with two special characters added by the tokenizer).

  18. While couched in Marxist-Leninist terms, the so-called “three directives” was a set of policy guidelines launched by Deng Xiaoping in May 1975 to steer the country away from the ideology of class struggle in the interest of economic modernization.

  19. The pre-processing involved removing Chinese stopwords and monosyllabic words from the corpus and selecting only articles containing at least 100 words. For topic modeling, we did not split the articles into smaller chunks as we considered each article as a thematically-coherent entity.

  20. The violence score was derived by multiplying the topic’s presence in each article with the corresponding cosine similarity calculated beforehand (see notes 16 and 17), leading to the aggregate weighted scores for each topic and month. The scores were then normalized using the total topic presence in each month. The normalization assures that the high presence of a topic in many non-violent articles does not by itself lead to a high weighted average. In other words, the topic must be present in highly violent texts to contribute significantly. Only the top three terms for each topic are shown.