The Evolution of Protest Repertoires in Hong Kong: Violent Tactics in the Anti-Extradition Bill Protests in 2019

Abstract The Anti-Extradition Bill protests in 2019 culminated in an unprecedented level of violence that departed from the established peaceful social struggles in Hong Kong. This paper examines the evolution of protest repertoires by analysing the interactions between protesters and state actors on a local and global scale. A dataset is presented to show the type, frequency and distribution of tactics. This paper reveals that structural and cultural changes as well as activists’ cognitive, affective and relational transformations at the micro- and meso-levels were pertinent to tactical radicalization. Cognitively, militant tactics were pragmatic responses to state-sponsored violence and police violence. They were also the affective outcomes of grief and anger. These processes were intertwined with the relational dynamics that advocated horizontal mobilization and that shaped, and were shaped by, the political-economic interactions between China and the West. The result was an extensive use of violent tactics alongside innovations in non-violent tactics.

In February 2019, the Hong Kong Special Administrative Region (HKSAR) government proposed to amend the Fugitive Offenders Ordinance, triggering one of the most spectacular political conflicts in local history. Ten months on, continuing protests against the law amendment, known as the Anti-Extradition Bill protests, had escalated and manifested a radicalization of protest tactics. From vandalizing government buildings and pro-China shops to hurling Molotov cocktails at the police, protesters adopted a range of violent tactics alongside peaceful actions like demonstrations, petitions and sit-ins.
What explains the widespread use of violent tactics during the Anti-Extradition Bill protests? And what explains the profound evolution of repertoires of contention in Hong Kong? Some studies provide self-reported data on the protesters' engagement in various actions and argue that radicalization was a result of a diversified participatory structure and interactions among different actors. 1 Others have examined the actions that targeted international audiences. 2 What remains to be understood, however, are the micro-dynamics that created and underpinned such a swift expansion of protest repertoires. Importantly, the ways in which local-global interactions steered the use of violent tactics have been left uncharted.
This paper seeks to explain the increasing use of violent tactics from June 2019 to January 2020 and explicate the evolution of repertoires of contention in Hong Kong. I first present data on the types, frequency and pattern of tactical choices during the Anti-Extradition Bill protests, along with a structural and cultural account of the broader change in repertoires of contention in the city. I then examine two key phases of the protests through an eventful analysis that highlights the interactions between actors at different geographical locales. I elaborate on how tactical innovations occurred and produced a political conflict filled with both violent and non-violent actions. The gist of my argument is that tactical radicalization and the expansion of protest repertoires were not simply driven by changes in political structure and political culture; they were also a result of cognitive, affective and relational transformations at the micro-and meso-levels, which allowed violent actions to become part of the local repertoires of contention.

Tactics and Repertoires of Contention in Protest Events
The dominant framework for studying movement tactics is Charles Tilly's repertoires of contention approach. Repertoires of contention refer to "the established ways in which pairs of actors make and receive claims bearing on each other's interests," meaning that actors tend to stick with a limited, recurrent and welldefined set of performances, even though other performances may be more effective. 3 These performances remain largely uniform for a significant period, but sometimes they change owing to macro transformations. 4 Regime change, for example, yields different sets of political opportunities and threats, leading to a tendency to use peaceful tactics in democratic states and violent ones in nondemocratic states. 5 As this study demonstrates, this is crucial to understanding how repertoires of contention are shaped in Hong Kong.
Repertoires of contention and tactics are moulded not only structurally but also interactively and relationally. As Doug McAdam maintains, ongoing interactions between powerholders and challengers engender a process of tactical adaptation and innovation. 6 Tactics evolve as actors' decisions affect one another, spawning a concatenation of outcomes. 7 One result is collective violence developed through the escalation of action repertoires within protest cycles. 8 In the process, political opportunities and state repression at the macrolevel, competition between movement groups at the meso-level and everyday experiences and radical ideologies at the micro-level all work together to prompt the use of violent tactics. 9 Despite the usefulness of the relational framework, existing studies tend to overlook the nuances of protest events upon which trajectories of collective violence are based. This might be owing to the fact that violence in the cases under study is often a product of decades of interaction in countries with a long tradition of radical movements. When explaining the lengthy phases of radicalization, the distinctiveness of events is inevitably flattened out. By comparison, the shift to violent protests followed a much shorter trajectory in Hong Kong, a process I will describe. It is thus possible to unpack the crucial protest events that motivated activists to use violence and discern the causal steps through which a dramatic evolution of protest repertoires occurred.
In analysing protest events, Donatella della Porta proposes a tripartite framework that encapsulates three distinctive but overlapping processes experienced by activists. These processes affect tactical choices and shape the movements that carry them out. 10 First, cognitive transformation occurs when activists interact with state agents and countermobilization groups, giving rise to concrete experience and knowledge about what tactics are effective and legitimate. For example, state repression is often a crystallizer of tactical escalation as violence can be justified in the face of an oppressive opponent. 11 Second, affective transformation takes place as activists experience emotional upheavals during protest events. This process profoundly influences tactics, especially when activists face intense repression as violent tactics are often deployed to vent anger towards opponents. 12 The third process is relational transformation stemming from the formation of solidarity networks and/ or competition between movement groups. This is pertinent to tactical innovation as activists learn and evaluate tactics through these networks. 13 Relational dynamics, however, can go beyond the local scale. Through scale shift, a contentious issue can be brought to the global stage and transnational alliances can be fashioned, both of which ultimately feed back into local state-society dynamics. 14 Accordingly, this article explores the cognitive, affective and relational transformations of Hong Kong activists that prompted them to resort to increasingly violent tactics during the eventful Anti-Extradition Bill protests. I disentangle some remarkable protest events and dissect activists' interactions with local state actors during the period June 2019 to January 2020. I also study a range of foreign governmental and non-governmental actors whose actions might have had a bearing on local activists' tactical choices.

Methods and Data
This paper adopts protest event analysis to explore the interactive dynamics of the Anti-Extradition Bill protests. Protest event analysis is a method that performs content analysis on newspaper articles and generates datasets about a protest. 15 It is therefore well-suited for examining the evolvement of protest repertoires. 16 Hanspeter Kriesi and his team use "action" as the unit of analysis so that a continuous stream of interactions can be reconstructed to identify the rhythm, interactivity and critical moments of a contentious episode. They also suggest observing multiple actors when analysing interactions. 17 Echoing this, the unit of analysis for the current study is organized action undertaken by actors who act collectively or perform in the capacity of a collective. I further advance the method by documenting actors at different geographical locales. Apart from local activists, HKSAR officials and countermobilization groups, I also take expatriate Hongkongers, foreign governmental actors and intergovernmental organizations into account.
An action is recorded when it is initiated by actors who make a contentious claim. An action co-initiated by different actors in the same place and at the same time is counted as one event as all actors share the same performance. However, actions emerging out of other actions are coded as separate events, although they may have some overlap in terms of time and space. For example, a clash with the police stemming from a non-violent sit-in is coded as a distinct event. A total of 5,459 actions were recorded from February 2019 to February 2020. During an action, actors may use single or multiple tactics. I include both institutional (for example, organizing press conferences, publishing public letters, moving private members' bills in the legislature) and non-institutional tactics (for example, break-ins, physical attacks). The usage frequency of each tactic in the resulting dataset reflects not only different levels of accessibility of these tactics but also how a decentralized network of actors affects tactical usage.
Data are derived from both traditional and online news media. Five newspapers with different political stances were chosen, namely Apple Daily, Mingpao, Oriental Daily, Takungpao and Wenweipo. The first two embrace a liberal stance and the rest are pro-China, with the last two being Beijing's mouthpieces in Hong Kong. I also collected data from Stand News, the most popular pro-democracy online news platform. It reports on many small-scale protest actions with rich descriptions. Considering the relevance of information, I included only reportage written by Stand News's reporters and excluded commentaries published on the website. In total, more than 30,000 pieces of news about the Anti-Extradition Bill protests were collected. An advantage of using multiple data sources is that it can mitigate the selection bias and description bias commonly found in earlier works of protest event analysis. 18 The unlawful nature of many actions during the Anti-Extradition Bill protests and the highly decentralized activist networks made it difficult to get in touch with frontline activists and build a rapport with them. To overcome this, I conducted a textual analysis of articles that contain interviews with activists to unpack the micro-dynamics of tactical choices. These interviews came from Apple Daily, Mingpao and Stand News. Taking into account the reputation of these prodemocracy news outlets, these interviews provided fruitful accounts of activists' subjective and affective worlds. In addition, to understand the transnational dimension of the protests, I participated in and made ethnographic observations at two Anti-Extradition Bill protests in Manchester in June and September 2019.

Regime Features, Political Culture and Strong Repertoires in Hong Kong
Repertoires of contention in local protests have been closely related to regime features. In post-war Hong Kong, large-scale violent conflicts were rare, the exceptions being the riots in 1956, 1966 and 1967. 19 This can be attributed to the colonial regime's changing mode of governance. After 1967, a number of sociopolitical reforms were carried out, leading to the partial opening of district-level participatory channels and the growth of pressure groups. Political dispute was thus absorbed by the administrative machine, which led to the predominance of "polite politics"the use of non-confrontational forms of struggle by challengersin socio-political conflicts. 20 Repertoires of contention are determined as much by political culture as by structural features. 21 A specific political culture of post-war Hong Kong has underpinned the tendency to use non-radical tactics. Siu-kai Lau sees "utilitarianistic familism" as the cultural foundation of the conservative nature of the Hong Kong Chinese population. 22 Wai-man Lam also notes that a culture of depoliticization discouraged activists from staging radical actions. 23 Despite the increasing popularity of public assemblies, rallies, sit-ins and signature campaigns, tactics employed in socio-political conflicts were largely non-violent and nonconfrontational from the 1970s through the 1990s. 24 The non-violent and non-confrontational tactics developed in the colonial era remained potent in post-handover Hong Kong, epitomizing the features of what 18 Koopmans and Rucht 2002;Earl et al. 2004. 19 Lau 1982. 20 Chiu and Lui 2000Ho 2000;King 1975. 21 Traugott 2010. 22 Lau 1982. 23 Lam 2004. 24 Lui and Kung 1985Sing 2000. Charles Tilly calls "strong repertoires"political performances that are strongly preferred under different circumstances. 25 Peaceful rallies, for example, have been recurrently utilized and have even gained an aura following the protest on 1 July 2003, during which half a million people took to the streets and stopped the passage of Article 23, a law on national security. 26 The protest underscored the hegemonic "order imagery" in the city while reinforcing the preference for non-radical tactics. 27 Despite the sway of strong repertoires, minor and incremental innovations in tactics have occurred, partly driven by changes in the local regime. After 2003, the Beijing central government started to encroach on Hong Kong's autonomy, critically through proclaiming comprehensive jurisdiction over Hong Kong in its White Paper, "The practice of the 'one country, two systems' policy in the Hong Kong Special Administrative Region," which was promulgated in 2014. 28 The regime's decreasing level of openness has led to an increasing use of forbidden claim-making performances by challengers, a notable example being the Umbrella Movement in 2014 during which tens of thousands of people resorted to civil disobedience, occupying roads in Hong Kong for 79 days.
The innovation of repertoires of contention has also been spurred by changes in political culture. Following the mass rally in 2003, new political actors emerged. Some joined the newly founded radical political parties and learned the tactics of direct action in the 2005 Anti-World Trade Organization protest. These tactics were applied in urban social movements, leading to an expanded use of disruptive yet still non-violent tactics. 29 Furthermore, in the early 2010s, growing mainland-Hong Kong integration triggered the emergence of localist discourses and associated tactical calls for more confrontational actions. Political developments from the mid-2000s through the early 2010s gave rise to a discourse challenging the use of "peaceful, rational and non-violent" actions and an acrimonious debate on the use of violent tactics during the Umbrella Movement. 30 After that ultimately fruitless campaign, competition among movement organizations intensified and the call for radical actions resonated more loudly. 31 Subsequently, a violent clash between police and localist protesters in 2016 presented a clear attempt to renovate the local repertoires of contention.
Overall, repertoires of contention in Hong Kong have been broadening as a result of changes in political regime and political culture. Although peaceful and nonviolent tactics remain a strong preference in civil society, confrontational but nonviolent tactics have risen as a viable alternative since the late 2000s. In the mid-2010s, violent tactics began to gain a foothold, laying the groundwork for the dramatic expansion of protest repertoires in the Anti-Extradition Bill protests. 25

The Evolution of Protest Tactics in the Anti-Extradition Bill Protests
The continuing influence of peaceful and non-violent tactics was evident in the Anti-Extradition Bill protests. Table 1 shows the most frequently used tactics during the protests. In total, 72 different tactics were deployed in the 2,879 oppositional actions recorded. The top three tactics were public assemblies/sit-ins, releasing public statements, and displaying and/or creating physical artifact(s). Yet it is important to note that a tactic is recorded if it is reported in newspapers, so the number of public statements might be underestimated as statements released by less influential actors were usually underreported. Also, displaying and/or creating physical artifact(s) could be a distinct event of its own or be used during public assemblies. In the latter case, it might be underreported as well. That said, the actual frequency of these two tactics might be higher than the figures presented here.
In the local history of contention, all three tactics were "strong repertoires" which have been used for decades. Organizing public assemblies, for example, was a "radical" and ground-breaking tactic founded during the urban conflicts of the 1970s. 32 It has since been conventionalized and drawn upon repeatedly in different settings. The adoption of strong repertoires in the Anti-Extradition Bill protests could also be attributed to the specific context of the late 2010s. When the law amendment was proposed in 2019, the pro-democracy movement was experiencing a period of abeyance. Major political parties and movement organizations who spearheaded the campaign to mobilize against the amendment had to resort to familiar tactics to gradually gain impetus. As Sidney Tarrow contends, conventionalized, modular tactics can be used with minimal organizational effort even during quieter times of protest cycles. 33 Thus, modular tactics, which include demonstrations, signature campaigns, promotional activities and holding press conferences, played a major part throughout the Anti-Extradition Bill protests, especially in the early stages. Table 1 also shows the remarkable role played by confrontational and violent tactics. Three violent tactics, namely arson, damaging property and physical attacks, were all recorded more than 100 times. Meanwhile, confrontational yet non-violent tactics, such as blockades, disruptions, spatial occupations and non-physical attacks (for instance, using laser beams to intimidate the police), also featured heavily in the protests.
I constructed a 15-point scale to categorize different tactics. The four basic types of tactics are non-confrontational (non-opponent specific), nonconfrontational (opponent specific), confrontational (non-violent), and confrontational (violent). Non-confrontational (non-opponent specific) actions refer to tactics that aim to interact with the broader population. For example, while collective singing and hunger strikes do convey protest claims, their audiences are not confined to specific opposition groups or individuals. By comparison, opponent-specific non-confrontational actions, such as petitions and demonstrations, refer to tactics that directly communicate with an opposing group or person. These actions contain a more obvious target in claim-making as protesters attempt to interact with, and make demands on, their opponents directly. Non-violent confrontational actions include tactics like blockades and break-ins. Although they often involve bodily clashes with state agents, they are not aimed at attacking anybody or sabotaging any property. The last type of tactic is violent confrontational action, which refers to actions involving intentional assaults on human beings and/or damage to physical property.
From non-opponent specific non-confrontational actions to violent confrontational actions, there is a rise in the specificity of actions when activists interact with opponent(s). It is assumed that activists will first exhaust the relatively indirect non-confrontational means and then turn to more confrontational actions when they try to influence or attack their opponents more directly. Yet multiple tactics can be used in combination to create various levels of radicalness. An action is considered more radical if it deploys more direct tactics and has a more extensive range of combinations of tactics, hence the most radical action would be the one applying all four types of tactics. The progression of radicalness is described in Table 2. Figure 1 depicts the temporal distribution and evolvement of tactics. Non-confrontational actions (scale 1 to 3), be they non-opponent specific or opponent specific, were predominant throughout the Anti-Extradition Bill protests, signalling the nature of strong repertoires. From March to June, several non-violent confrontational actions (scale 4 to 7) were attempted, sometimes in combination with non-confrontational actions, by some radical political parties and non-party affiliated activists/politicians. As Table 3 further shows, although it appears that a radical wing already existed at the early stage, confrontational actions were not yet prevalent. The first tactical breakthrough took place in June. By that time, some nonorganizationally affiliated protesters had issued mobilizing calls through the LIHKG online discussion platform and Telegram, which resulted in fierce  clashes with the police. Although those events marked the start of the use of violence, violent confrontational actions (scale 8 to 15) were not widely employed until the second tactical breakthrough in mid-July. As Table 3 reveals, the radical wing had expanded since mid-July and conducted more violent confrontational actions. These actions co-existed with the non-violent actions staged by the moderate wing; both types were intensively applied until January 2020.
Next, I will unpack some crucial events occurring before and after mid-July. I will illustrate how a chain of events created transformative impacts on protesters in cognitive, affective and relational terms and led to violent tactics becoming part of the local repertoires of contention.

Innovations in Repertoires during the Anti-Extradition Bill Protests
Cognitive, affective and relational transformations before 21 July A protest is a transformative event where protesters gain knowledge about the empirical world and tactics, experience emotional upheavals and form networks and alliances. 34 During the Anti-Extradition Bill protests, events in mid-June informed the protesters about the brutal reality of state authoritarianism and police violence, prompting them to respond with progressively radical tactics.
On 9 June, the Civil Human Rights Front (CHRF) mobilized a demonstration that attracted more than a million participants. Despite such a massive, peaceful rally, the government refused to withdraw the law amendment bill. Infuriated by the government's authoritarian attitude, some radical protesters resorted to tactics of physical attacks and arson that night. A confrontation with the police ensued, but the second reading of the law amendment bill was still going to take place on 12 June. On that day, the moderate and radical protesters mobilized an assembly and organized a road blockade around the Legislative Council complex. Some radical protesters again attacked the police with bricks and sticks. This triggered one of the harshest episodes of police repression of civilians in local history as the police responded with over 150 canisters of tear gas, 20 bean bag rounds, several rounds of rubber bullets and countless attacks with pepper-spray, pepper-spray pellets and batons. Aiming to denounce the police violence and call for the withdrawal of the bill, the CHRF mounted another peaceful demonstration four days later, which the organizer claimed was attended by two million people. Meanwhile, protesters formulated the so-called "Five main demands," which demanded that the government withdraw the bill, set up an independent commission of inquiry to investigate police violence, retract the characterization of the "6.12" protest as a riot, drop the charges against all arrestees and implement universal suffrage. As the HKSAR government promised only to postpone the law amendment, the protesters gradually went through a cognitive transformation. The uncompromising stance of the 34 Della Porta 2008b. government informed them, particularly the radical wing, about the need for a tactical escalation.
Cognitive transformation went hand-in-hand with affective changes. In political conflicts, anger is a crucial emotion for mobilization, but it is also carefully contained to avoid crackdowns. 35 Anger over the police violence had been mounting since the "6.12" clash, but the radical protesters did not immediately respond with violent tactics. Instead, they mobilized two largely non-violent and restrained actions. On 21 and 26 June, they laid siege to the police headquarters. Besides blocking the entrance and nearby roads, they condemned police violence mostly in symbolic terms through, for example, the use of foul language, throwing eggs at the walls and defacing the police emblem with graffiti. Anger towards the police was evident but it was also restrained. Protester M recounted his confrontation with the police in Shatin on 14 July. He claimed that the police had initially allowed the protesters to leave via a shopping mall, but then a police squad entered the mall and beat them. M was beaten by a policeman with a baton until around ten protesters came to help and hit back at the policeman with punches and kicks. "Everyone was so angry. It is unreasonable for the police to use such a level of violence against us and that we had to suffer that." Yet he emphasized that their anger was not unrestrained: "No protester lost control; that policeman would have been dead otherwise." 36 Although the protesters were repeatedly enraged by police violence and responded with increasingly radical actions, they kept control of their anger and their use of violence to avoid damaging the legitimacy of the campaign.
Events in June provoked not only anger but also grief. These emotions helped to boost mobilization and solidarity, which was germane to the protesters' relational transformation. 37 On 15 June, Ling-kit Leung, in an extreme act of protest, committed suicide, evoking a huge wave of grief in the city. Anger intermingling with grief gave way to a huge mobilizing force, evident in the scale of participation in the demonstration the next day. On 16 June, many protesters paid tribute  Jasper 1998. to Leung at the site where he died. The demonstration became a public mourning event, the shared grief forging a sense of solidarity among the protesters. As Judith Butler argues, the body implies mortality, vulnerability and agency. Deaths and mourning are powerful events that connect political subjects. 38 At the same time, emotions helped justify alternative tactical choices. One protester, now in exile, articulated her feelings in the following way: I'm sorry that I didn't take action to avenge Martyr Leung on 16 June but just chanted a few more slogans and wrote a few more posts on social media; I regret it that I was one of the people who let the paramedics into the police station on 21 June to bow down to the good-for-nothing cops; I cannot forgive myself for just writing down "An eye for an eye" on the walls in the [Legislative Council complex] and didn't carry it out on the day. 39 This quote expresses an urge for revenge and a preference for more radical actions. Remarkably, activists' solidarity and the radical wing's position were strengthened by another emotional event: the storming of the Legislative Council complex on 1 July. Despite initial doubts about police infiltration, the protesters' dramatic retreat from the meeting hall exhibited a strong sense of brotherhood and sisterhood, which became an affective image that united different factions. Unlike the Umbrella Movement, debates over radicalization were kept at a minimal level through using slogans such as "No severing of ties" and "Brothers climb a mountain together, each makes his own effort." These slogans overcame the tactical divisions between the radical and moderate wings, allowing radical actions to become a viable option.
The three transformations worked not just locally but also globally. The extradition bill and events in June activated the diasporic communities of Hongkongers and galvanized a global network of Hong Kong dissidents to mobilize around the world. In total, 147 oppositional actions were staged by expatriate Hongkongers and 178 actions were organized by expatriate Hongkongers and foreign supporters together. Responding to the violent "6.12" clash, for example, the diasporic communities mobilized public assemblies in their cities on 16 June to support protesters in Hong Kong. I joined the protest in Manchester. Although the assembly was loosely organized through a Facebook group, participants all displayed a grievance towards the Hong Kong police. The improvised slogans we chanted condemned both the HKSAR officials and the police chief. We also held a minute's silence to mourn the death of Ling-kit Leung. A participant shared his thoughts, calling for solidarity between different factions in the face of a common enemy. At the end of the gathering, we sent our group photos to Apple Daily, which reported instant coverage of the protest actions around the globe in a symbolic act of solidarity with the protesters.
The three transformations also featured in the interactions between Hong Kong activists, the Chinese government and the West (English-speaking countries and EU member states). Against the backdrop of the arrest of Huawei's vice-chairwoman in Canada and the US-China trade war, local protesters 38 Butler 2004. 39 "Letter from exile. 2." Stand News, 25 September 2019. cognitively perceived and exploited the opportunity to persuade political actors around the world to put pressure on China. The worsening relations between China and the West also fostered hope that Western governments would offer assistance to protesters in Hong Kong. In total, 385 oppositional actions set nonlocal actors (including intergovernmental organizations such as the United Nations, foreign governments and foreign politicians) as targets of communication. In June, the G20 summit in Japan provided a timely opportunity for activists to rally international support. While local activists petitioned the consulates of the G20 member states to press China to withdraw the extradition law amendment, pro-democracy supporters set up a crowdfunding campaign to pay for advertisements in newspapers around the world in an attempt to attract global attention. Figure 2 shows the pattern of reactions from Western countries and intergovernmental organizations (0 = supportive stance towards the HKSAR/Chinese governments, 1 = antagonistic stance against the HKSAR/ Chinese governments, 2 = ambivalent). Prior to June, instances of antagonistic reactions were fewer than reactions with an ambivalent stance. After June, both types of action increased, proving the potency of activists' transnational mobilization and the impact of local conflicts.
The significance of events in June is shown in Table 4, which documents the top 30 events that sparked the most follow-up or preparatory actions. In coding the newspaper articles, I assume that each action was an attempt to respond to past events or prepare for upcoming events. For example, police violence in a protest may trigger another protest to condemn the violence, what I term a follow-up action. Also, protesters sometimes envision capitalizing on certain occasions for claim-making. This constitutes a preparatory action. Then, I identify the three main events that a current action was responding to or preparing for. If an event triggered a large number of follow-up or preparatory actions, it can be interpreted as an event highly influential to protesters' strategic choice. As shown in the table, the "6.9" demonstration, the "6.12" clash, the two suicide protests on 15 and 29 June, and the G20 Summit yielded 80, 272, 61, 50 and 31 follow-up and/or preparatory actions respectively, implying their remarkable cognitive, affective and relational impacts on activists.

Cognitive, affective and relational transformations after 21 July
While the events in June laid the foundation for radicalization, the turning point towards violence occurred on 21 July. On that day, the CHRF staged another demonstration. Some radical protesters departed from the main protest line and headed to the Liaison Office of the Central People's Government (LOCPG). They smeared the national emblem with black ink in a symbolic condemnation of Beijing's political intervention, provoking another heavy-handed response and confrontation with the police. As some protesters returned home late in the evening, a group of pro-China thugs, wearing white T-shirts and armed with sticks and canes, launched an attack on protesters, journalists and even ordinary citizens at the Yuen Long train station. Some online videos revealed that the thugs had the support of a pro-China legislator. Furthermore, the police initially ignored emergency calls from citizens, only arriving at the scene 39 minutes after the first call for help. This gave many Hongkongers the impression that the organized attack had the tacit approval of the police.
The alleged collusion between the police and the thugs came as a "moral shock" to many Hongkongers, changing their cognitive perception of reality and fuelling their anger. 40 Although the countermobilization tactics of recruiting mobs and thugs also appeared in the 2014 Umbrella Movement, this was the first time Hongkongers were made aware of the scale and level of violence of countermobilization. The radical wing of pro-democracy protesters thus used the evident moral decay of the police force and state-sponsored violence to legitimize their use of radical and even violent tactics. As Table 3 shows, instances of violent actions (scale 8-15) increased five-fold from July to August. Table 5 shows the result of a multiple regression analysis on the use of various tactics, indicating that attitudinal transformation which endorsed the co-use of peaceful and radical actions was a strong predictor of the increasing use of violent tactics. It also confirms the significant impact of more violent countermobilization tactics on both peaceful and radical actions on the prodemocracy protesters' side.
Media interviews with radical protesters disclose the details of their cognitive and affective transformations in the aftermath of the twin events on 21 July. Protester A, who maintained a moderate stance in the Umbrella Movement, was on the frontlines battling the police in 2019. Now, he asked: Why should we be so lenient to the enemy and so restrictive to ourselves? … We had not caused any fires until the confrontation in Sheung Wan on 21 July. On that day, the police kept shooting us with rubber bullets. How could we escape if we did not light a fire as a barricade? 41 His words indicate that violent tactics were indeed pragmatic responses to repression. Similarly, protester M recalled that following the "7.21" thug attack, he and his fellow activists began preparing additional protective gear. Their gear proved useful during the subsequent protest on 27 July, which was intended to exact revenge on the pro-China thugs, as the police quashed the protest to "protect" the thugs. He said, "my friends asked me how a moderate protester gets radicalized. I replied that there is nothing special. If she or he goes to the frontline and witnesses incidents like this, she or he will automatically become a radical protester." 42 Again, episodes of police violence aided the justification of violent tactics, as activists' cognitive and affective experiences of police violence motivated them to become more militant.
The twin events on 21 July also provoked hostile state responses that changed state-society dynamics. Eight days after the protest at the LOCPG, the Hong Kong and Macao Affairs Office (HKMAO) of the State Council held a press conference to praise the efforts of the HKSAR government and the police force and to call for a cessation of the "violence and chaos." In subsequent press conferences, the HKMAO even warned against emerging signs of terrorism in Hong Kong. The high-profile support from the central government amounted to an endorsement of police repression, which generated a spiral of violence in local society.
As repression intensified, the protesters underwent further cognitive and affective transformations that led to an expansion of the radical wing. As Table 5 indicates, the intention to blame the local officials and/or the police (as measured by protest claims presented in actions) was a strong predictor of both violent and non-violent confrontational tactics. The spiral of violence was evident from August to November. On 11 August, several incidents occurred: a female protester's eye was badly injured by the police, the police fired a canister of tear gas into the lobby of a train station and some pro-China gangs launched another attack on pro-democracy protesters. These events led to 37, 29 and 25 follow-up actions respectively (Table 4). In the same month, there were also rumours of deaths caused by police violence at the Prince Edward train station and of sexual abuse occurring in the San Uk Ling detention centre. Martyrs and stories of deaths formed the sources of anger and hatred, compelling activists to fight against the "evil" police. One protester declared: "I really just want to avenge our fallen heroes with everyone, and win this fight together … I hope that everyone can bear the martyrs from San Uk Ling and Prince Edward, and those who were murdered for knowing the truth, in mind in the coming days. self-restraint was now loosening, with "Hongkongers, revenge!" gradually replacing "Hongkongers, resist!" as a protest slogan from September. While the radical protesters launched selective attacks against policemen and government property with ever-increasing frequency, they also targeted the property of pro-China gangs and businesses. Violence further escalated in October owing to the implementation of the Anti-Mask Law. It reached a peak in November with 109 instances of violent actions (Table 3). In early November, a university student, Tsz-lok Chow, was badly injured during a street battle and died a few days later. Chow's death sparked 137 follow-up actions as the protesters accused the police of killing Chow and mounted a city-wide strike. The strike culminated in fierce confrontations on two university campuses and nearby streets, producing 59 actions that applied both violent and non-violent confrontational tactics (scale above 11) against the police that led to thousands of arrests and hundreds of injuries between 11 and 29 November. Cognitive and affective transformations intertwined with some evolving relational dynamics to allow both violent and non-violent tactics to develop further after 21 July. Innovations occurred and added new impetus to the existing strong repertoires of peaceful protests. Notable examples were the tactics of buycott and boycott, which originated in the 2014 Umbrella Movement but only really caught on in 2019. 44 Another example was the collective singing of "Glory to Hong Kong" at a city-wide level. Had this non-violent, symbolic tactic be adopted during the Umbrella Movement, participants would have been accused of being defeatist. Yet, in 2019, protesters were connected through horizontal working cells. In principle, cells or individuals can propose their own tactics. According  (2020); survey results of 10, 11 and 13 August were combined by taking an average to construct a meaningful timeframe to observe tactical change; results of 16 and 18 August were combined; the two surveys on 4 August were combined; figures for November were estimated through taking the mean of the 20 October and 8 December surveys. b Tactic scores were computed through the same 15-point scale.
to one protester: "even if the propaganda cell and the tactics cell disagree with each other, they respect each other's decisions. Then both groups would send out a message to ask people to decide which path they want to follow. If one is determined enough, she or he can go on." 45 Horizontal mobilization thus allowed diverse tacticsboth violent and non-violentto flourish. Violent actions could also be justified in the name of solidarity, although such justifications sometimes marginalized criticism of violent tactics. 46 Relational transformations also occurred at the global-local nexus. After 21 July, violent and non-violent tactics were deployed simultaneously to connect local protesters with the outside world. The protesters invented new non-violent tactics not seen in past protests, such as displaying symbolic artifacts (for example, national flags and masks of foreign political figures). They also organized assemblies at Hong Kong International Airport to communicate their protest message to foreigners. Later, inspired by the action that had originated in communist Estonia, Latvia and Lithuania, they formed human chains to appeal to global audiences. On 29 September, local protesters and expatriate Hongkongers staged anti-totalitarianism protests, calling for global solidarity against the Chinese communist regime. I joined the assembly in Manchester. We gathered outside the venue of the Conservative Party Conference, hoping that our message could be added to the Tories' agenda. Some protesters were dressed in black, wore respirator masks and waved flags displaying the slogan, "Liberate Hong Kong, revolution of our times," in solidarity with frontline activists in Hong Kong who shared a similar outlook. We chanted slogans such as "Dirty cops, an eye for an eye" and "Cops did nothing on 7.21, and beat protesters to death on 8.31," expressing our shared anger over the police violence.
On the radical front, the protesters continued to use violence in local street battles to expose the evil side of the regime and rally for international support. In the words of one radical protester: "[non-violent actions] were mostly unspectacular, but scenes of violent clashes were eye-catching. The latter could show to the outside world that we were still fighting against this regime." 47 Although the actions of Western governments are not a significant predictor of local tactics as shown in Table 5, some local-global dynamics are worth mentioning. Since the Umbrella Movement, pro-democracy activists, spearheaded by Demosistō, a political party, have continued to lobby US politicians to assist in the struggle in Hong Kong. In 2019, they urged members of the US Congress and Senate to pass the Hong Kong Human Rights and Democracy Act (HKHRDA), which, if implemented, would amount to an economic sanction on China. Fierce clashes in October and November apparently accelerated the legislation of the HKHRDA and its final endorsement by the-then US president, Donald 45 "Kangzheng xunhaota, suishishi jinhua, niquan wenxuan, wuming yingxiong" (Evolving protest information centre, propaganda for rights struggles, nameless heroes). Apple Daily, 15 July 2019, A10-11. 46 Lee 2020. 47 "Yongyunü: chongji ling zhengfu you fanying" (Radical female protester: conflicts induced governmental responses). Mingpao, 5 November 2019, A06.
Trump. Besides the US, other Western countries also expressed support for local protests. As Figure 2 demonstrates, since mid-October instances of antagonistic reactions against the HKSAR/Chinese governments (code 1) outnumbered ambivalent ones (code 2). In sum, Western reactions and local violent actions appear to have a circular causal relationship, yet more research is needed to confirm the actual relationship.

Conclusion
By examining a broad set of actors whose actions shaped political interactions through transnational connections, this study has shown how the evolution of the repertoires of contention in Hong Kong was engendered through cognitive, affective and relational mechanisms during the Anti-Extradition Bill protests. Cognitive and affective transformations occurred when the protesters encountered, and were enraged by, police violence and the government's authoritarian attitude, events which were entwined with a solidaristic bond facilitated through grief. These transformations also took place on a global scale through activists' efforts to communicate with foreign political actors and the mobilization of the Hong Kong diasporic communities. The three transformations took a turn on 21 July. State-society conflicts escalated as the Beijing government became more determined to clamp down on the protests. The protesters also justified violent tactics as a pragmatic response to, and an expression of anger towards, police violence and state-sponsored violence. Violent tactics therefore began to be incorporated into the local repertoires of contention alongside some innovations in non-violent tactics.
Examining the evolvement of protest repertoires is crucial to advancing our knowledge of movement dynamics. Although the social movement scholarship emphasizes the need to appreciate relations and interactions, few studies have looked beyond local-and national-level politics. This paper contributes to the literature by showing how tactics evolve through interactions between local, national and global actors. Analysing movement dynamics through a wider spatial lens is also indispensable to our understanding of emotions. Despite the emotional turn in the 2000s, the study of emotions in social movements has stayed at the bodily or behavioural levels. 48 This paper nevertheless argues that emotions can be influenced by global actors' actions and can inspire transnational actions.
The radicalization of Hong Kong's political conflicts prompted Beijing to promulgate the National Security Law in 2020. While the law further reduces the local regime's level of openness, the space for claim-making activities has become increasingly limited. As Charles Tilly contends, in high-capacity nondemocratic states, protests can only emerge in the tiny space that escapes governmental control and these actions often invoke a high degree of violence. 49 This explains why many claim-making activities in Hong Kong that were previously toleratedincluding the June 4th commemorationare now forbidden despite activists' attempts to mobilize through the strong repertoire of lawabiding, peaceful actions. Some established movement organizations, such as the CHRF, have also been forced to disband. Under these circumstances, less organized, wildcat actions that involve intentional physical assaults arise. It remains to be seen whether local protests will head towards a mode of "mobilization without the mass" that is commonly found in mainland China. 50