Learning from criticisms of civil resistance

Critical Sociology, vol. 47, nos. 7-8, 2021, pp. 1191-1203, https://doi.org/10.1177/08969205211025819

pdf of published article

Monika Onken

Free University of Berlin, Germany

Dalilah Shemia-Goeke
Brian Martin

University of Wollongong, Australia


In recent decades, civil resistance, also known as nonviolent action, has become more widely used among social movements and recognised by researchers. Alexei Anisin has usefully offered a critique of civil resistance theory and practice. His ideas are used as a basis for reflection and deeper understanding of both strengths and weaknesses of this approach. Inspired by Anisin’s questioning of the dataset used to compare nonviolent and violent anti-regime campaigns, we point to multiple neglected factors. Anisin contrasts quantitative nonviolence research unfavourably with medical and scientific research. Although Anisin’s image of science is idealised, it does point to the value of recognising the role of values in civil resistance research and in critiques of it.


sociology, protest, revolution, nonviolence, social movements, civil resistance


There are long histories of two approaches to social and political change, armed struggle and what might be called conventional politics, which includes negotiation, lobbying, public debate, campaigning and elections. The writing and analyses in these areas are vast in extent, for example with detailed scrutiny of wars and political machinations. There is also a third approach to social and political change, which can be called civil resistance, nonviolent action, people power or satyagraha. It involves citizens who, in their roles as civilians, use methods such as strikes, boycotts, fasts, sit-ins and setting up alternative institutions. Although these methods have been used for hundreds of years, seeing them as a strategic approach to social change dates from the early 1900s, when Gandhi led campaigns in South Africa.

In the past century, this third approach has gradually gained recognition among social activists, and in many countries in recent decades has replaced armed struggle as the preferred approach in social movements. This is despite a lack of funding or official recognition by governments. Accompanying the greater use of civil resistance, there has been greater interest in this approach by scholars. Even so, civil resistance remains a minority interest in studies of social movements, where resource mobilisation, political opportunity structures, and framing continue to dominate the field.

Along with the rise in activist and scholarly interest in civil resistance, there have been a number of critiques (e.g., Churchill, 2007; Gelderloos, 2007). More recent critiques, such as Malm (2021) also show how relevant this debate is for practitioners, and why it may therefore be meaningful to engage in it on an academic level as well.  Critiques can be valuable for testing and developing ideas. Here, we focus on Alexei Anisin’s 2020 article “Debunking the myths behind nonviolent civil resistance”. It raises a number of valuable points worthy of further comment. While we disagree with some of Anisin’s assessments, we appreciate the effort he has put into his analysis. We share with him a search for effective ways for activists to challenge repression and oppression, and for this it is valuable to learn from different perspectives.

In the next section, we provide some background about the theory and practice of nonviolent action, as context for examining issues raised by Anisin. Then we look firstly at methodological and then at ideological issues involved with Anisin’s critique. Our examination points to some matters that need greater attention by researchers, such as the recognition of violence in otherwise nonviolent campaigns and new conceptualizations of campaigns. In summary, we highlight the value of critique for gaining greater knowledge about ways of challenging repression and oppression.

2. Background of nonviolence theory and practice

Methods of nonviolent action, such as strikes, boycotts and other forms of noncooperation, have been used for hundreds of years. It was only in the 1800s that awareness developed about a few major campaigns — notably in Hungary and Finland (Csapody and Weber, 2007; Huxley, 1990) — using these methods. Thinking of nonviolent action strategically is commonly attributed to Gandhi in campaigns that he led in South Africa and India (Sharp, 1979). Gandhi developed this approach over the following decades, seeing his efforts as “experiments with truth” (Gandhi, 1940). During his lifetime, Gandhi was lauded but also reviled, including by proponents of armed resistance in India. Gandhi was a prolific writer but did not try to organise his ideas into a standard framework. A few researchers tried to do this (Gregg, 1934/1966; Bondurant, 1958/1988). Gandhian nonviolence is based on an ethical imperative not to use violence against opponents, and is sometimes called “principled nonviolence”.

Gene Sharp (1973, 1980, 2005) developed a different approach to nonviolence. Initially following Gandhi’s approach, he then developed a framework grounded in the assumption that nonviolent action can be more effective than violence. Sharp’s approach is sometimes called “pragmatic.”

Gandhi’s campaigns and example inspired many activists around the world, for example Martin Luther King, Jr. Sharp’s work also inspired many activists, especially his book From Dictatorship to Democracy (1993/2010), which was translated into many languages and widely read. Despite his huge impact on activists, for decades Sharp had relatively little impact on social movement scholarship (Martin, 2013).

Gandhi and Sharp are the two most prominent figures in the field of nonviolence, but there have been a great number of other contributions. There are studies of nonviolence in history (Bartkowski, 2013), new theoretical approaches (Vinthagen, 2015), and numerous case studies of nonviolent campaigns (Parkman, 1990; Schock, 2005; Zunes et al., 1999), and the possibility of civil resistance as an alternative to military defence (Boserup and Mack, 1974; Johansen and Martin, 2019). Despite the research and increasing attention to nonviolent struggles, though, it has been commonly assumed by most commentators — aside from those involved in nonviolent campaigns — that a sufficiently ruthless government can always defeat nonviolent challengers. This assumption was most obvious in comments that Gandhi’s methods might have worked against British colonial rule, with the British assumed to be gentle oppressors, but not against the Nazis.[1]

A number of scholars documented challenges to repressive regimes driven by nonviolent action — for example the Iranian revolution of 1978–79 — but the belief that nonviolence could be successful only against democratic or benevolent governments persisted (Chenoweth, 2021: 13). Statistical studies confronted this belief. First was a treatment by Karatnycky and Ackerman (2005). After this, and far more powerfully, was a comparison of armed and nonviolent campaigns for regime change, secession or against occupations carried out by Erica Chenoweth and Maria Stephan (Stephan and Chenoweth 2008; Chenoweth and Stephan, 2011). Their 2011 book in particular became widely recognised.

Chenoweth and Stephan drew several conclusions. One was that nonviolent anti-regime campaigns were twice as likely to be successful as armed campaigns. Another was that nonviolent campaigns were just as likely to be successful against repressive regimes as against less repressive ones. This latter finding was a frontal challenge to the hegemonic belief that violence, if sufficiently ruthless, will always triumph over nonviolent opposition.

Chenoweth and Stephan (2011) argue that nonviolence as a form of organized and strategic civil resistance activates multiple mechanisms, which force the state to either give in to demands or give up office completely. These mechanisms include mass mobilisation, backfire and loyalty shifts to undermine the state’s pillars of support. This understanding counters the common assumption of monolithic state power. Instead, Chenoweth and Stephan as well as predecessors, such as Sharp (1973), argue that the state — no matter how repressive — relies on the cooperation and obedience of its people. These dependencies, on for example elites, economic organizations and security forces, are targeted in an effort to encourage people to disobey orders and withdraw their support from the government. Fundamental to Chenoweth and Stephan’s study is that the chances of this happening are much higher with nonviolent campaigns than with violent ones. Among other reasons, this is due to more people seeing the resistance as legitimate, and state agents becoming more reluctant to use repression.

The attention to Chenoweth and Stephan’s work has shifted the field, to some extent, toward quantitative studies. Even so, a considerable amount of qualitative research continues. It is important to recognise that the nonviolence research field is much broader and diverse than quantitative studies of anti-regime campaigns. In particular, there continues to be research on how nonviolent campaigners can be more effective (e.g., Bramsen, 2018; Burrows and Stephan, 2015; Dudouet, 2015), in addition to studies of whether they are effective.

With this brief background, we now turn to issues raised by Alexei Anisin. Our aim is twofold: to examine his findings and to use his critique to point to areas where nonviolence research can usefully be expanded or directed.

3. Methodological issues

Anisin’s primary focus is on the quantitative study of campaigns, in particular Chenoweth and Stephan’s (2011) study of 323 anti-regime, secession and anti-occupation campaigns from 1900 to 2006. Chenoweth and Stephan selected campaigns with maximalist demands — overthrow of a government for an anti-regime campaign, independence for a separatist movement and ending of an occupation — and classified them as either violent or nonviolent. They then assessed whether they were successful, partially successful or unsuccessful in achieving their goals. Chenoweth and Stephan found that nonviolent campaigns were twice as likely to be successful as violent campaigns.

Anisin challenges this finding in several ways. He argues that Chenoweth and Stephan’s dataset is inaccurate in a number of respects, for example that relevant campaigns have been omitted or misclassified. Most importantly, he offers two modifications to the dataset.

First, he argues that rather than classifying campaigns as either violent or nonviolent, there should be four categories: nonviolent, unarmed violent, reactive violent and fully violent campaigns. Unarmed violence refers to rioting and, more generally, violence without arms, for example throwing stones or using Molotov cocktails. Reactive violence refers to violent methods of defending against state violence. Classifying campaigns into these four categories, Anisin finds that nonviolent campaigns are not as successful as claimed by Chenoweth and Stephan, while campaigns using unarmed or reactive violence are more effective.

Second, Anisin extends the dataset by including campaigns in the 1800s. He finds no nonviolent campaigns during this century but a considerable number of violent campaigns, a majority of which were successful. Adding these campaigns to the dataset increases the overall success rate for violent campaigns.

Anisin’s innovations raise a number of issues worth considering. One is the question of how to classify campaigns, most of which in this case are campaigns to overthrow a government or oust a dictator. Classifying a campaign can be difficult because few campaigns solely use one set of methods. Chenoweth and Stephan decided to use two categories, violent and nonviolent, making the decision based on an assessment of the dominant method of action. If protests, strikes and boycotts were the main methods, a campaign was classified as nonviolent even if there was some unarmed violence. This is a problem if the unarmed violence was instrumental in the success of the campaign.

We would go further than Anisin in pointing to the limitations of classifying campaigns. Many violent campaigns also use nonviolent methods. As is well known, guerrilla warfare is as much a political as a military mode of struggle. Guerrillas seldom can win in a frontal battle, and much of their success stems from obtaining support from the population, which can be achieved through means such as land distribution, stamping out corruption and setting up community justice procedures, all of which are part of the repertoire of nonviolent campaigners.

Given that armed struggles typically include nonviolent elements and that many campaigns called nonviolent include some degree of violence, what is the best way to proceed? Chenoweth and Stephan assessed campaigns according to what seemed to be the dominant mode of struggle, either arms or nonviolent action. As Anisin argues, this is questionable because many campaigns have other elements besides the mode considered dominant.

However, Anisin’s classifications can be subject to the same sort of questioning. He has not made clear whether his two new categories, unarmed and reactive violence, are based on an assessment that these are the primary modes of struggle. In each campaign that he so categorises, nonviolent as well as violent methods were used. What is the justification for categorising a campaign as unarmed violence? It seems that Anisin classifies campaigns in this category if unarmed violence played a role, perhaps even a seemingly secondary role. Consider the campaign in Madagascar from 1991 until 1993. Anisin classifies it as unarmed violence. It did indeed involve violence yet, according to newswire reports, only 2% of 147 campaign events included rock-throwing.[2] Is this enough to categorize the entire campaign as unarmed violence, and how does this differ from campaigns in which 30% or 60% of campaign events are violent?

It is further unclear why Anisin chose those two new classifications and decided against using already existing categories. Among others, there is research on the role of fringe terrorism in nonviolent campaigns, which is distinctively different from unarmed violence (Belgioioso et al., 2019). As armed actors are in fact sometimes part of otherwise nonviolent campaigns, as pointed out by Pinckney (2016), there seems to be insufficient justification to treat unarmed violence as a category of anti-regime campaigns. The only alternative mentioned by Anisin to his own classification is research on radical flanks, which he describes as a very broad label. Indeed, it can be argued that radical flanks, in this context better called violent flanks, encompass a variety of violent tactics (Chenoweth and Schock, 2015). Anisin`s category of unarmed violence raises the same sorts of issues. On the one hand, Anisin excludes the use of arms from the category, hence calling it unarmed. Yet, Molotov cocktails and homemade grenades (for example in the 1986 people power revolution in the Philippines) are included under the umbrella of unarmed violence. This gives the impression that neither Molotov cocktails nor homemade grenades are arms, which is open to question.

It is also unclear whether reactive violence is the immediate reaction to repression or a longer development. Reactive violence could take place in response to numerous state actions, including covert repression, arrests and intimidation tactics. Furthermore, violence is not always immediate but can also develop over the course of time, for example when participants become frustrated with the lack of achievements.

Anisin’s criticism of Chenoweth and Stephan’s dichotomy of violent versus nonviolent campaigns, and the questions about how he justifies his own set of four categories, point to a shortcoming inherent to most — if not all — research. We need to simplify the complexities of individual campaigns and, more generally, processes of social change in order to make claims and draw conclusions about their effectiveness. No matter how hard we try, every research design will have to rely on some form of simplification. Whether this simplification, for example into the dichotomy of violence and nonviolence, is useful depends on the research aim. As Chenoweth and Stephan set out to compare large primarily nonviolent and violent campaigns, the dichotomy works for their purpose. Introducing a new threshold for what constitutes nonviolence and violence also simplifies complex realities, in Anisin`s case for the purpose of drawing conclusions about the effectiveness of unarmed violence.

In terms of methodology and dichotomies, the question arises of why Anisin did not supplement the existing quantitative data with qualitative insights. One way of overcoming the simplification of quantitative analysis while harnessing its advantages could be a mixed method approach (Creswell and Clark, 2017). This could for example strengthen the conceptualization of violence in otherwise nonviolent campaigns and allow scrutinizing violent mechanisms by relying on case-specific insights and guaranteeing generalizability through large-n data. Following this, it needs to be pointed out that deciding what is nonviolence and violence, and whether to use either of them, is context-specific and regularly discussed by campaign members and leaders. In other words, the distinction into violence and nonviolence could be understood as a social construct.  As far as we know, neither Anisin nor Chenoweth and Stephan relied on campaign expertise for their classifications.

Without refining the research design to make more reliable claims and without offering more theoretical insights into how unarmed violence leads to success, it is possible that the enhanced likelihood of success found by Anisin is due to a reporting effect. On the one hand, large movements are more likely to be successful and to engage in diverse tactics, including unarmed violence. On the other hand, the media are more likely to cover large movements and report on the use of violent tactics. In other words, the media reports on those movements most likely to achieve change and that involve violent tactics. Yet, the common determinant for both could be size rather than violent tactics. Gamson (1989) in reflecting upon his own work from 1975 comes to a similar conclusion: just because violence is observed does not mean that it works. Instead, he says, “violence is more a symptom of success than a cause” (Gamson, 1989: 458). This highlights the problem of not being able to observe the counterfactual in a real-life setting. We cannot see what would have happened had a specific violent movement relied solely on nonviolent methods, and vice versa.

There is a different sort of shortcoming in Chenoweth and Stephan’s analysis, one not noted by Anisin. According to Dahl et al. (2021), a key factor in anti-regime campaign strategies is selection of methods according to their prospect of success. Campaigns with the potential for widespread popular participation may be more likely to choose nonviolent approaches: they have a mobilisation advantage. In contrast, movements with a more limited capacity to generate popular support do not have this advantage and hence may be more likely to choose armed struggle, which is less dependent on mobilising mass support. This can help to explain the higher rate of success of nonviolent anti-regime campaigns: they are the ones with a broader base in the population for bringing about change.

As noted earlier, Anisin supplements Chenoweth and Stephan’s dataset with campaigns in the 1800s. He says there were no nonviolent campaigns during this century. It is not clear why he has omitted the Hungarian struggle for autonomy from the Austrian empire, which was an inspiration for Gandhi and other supporters of nonviolence (Csapody and Weber, 2007). That aside, to include campaigns in the 1800s raises an issue addressed neither by Chenoweth and Stephan, nor by Anisin: the roles of technology and training.

The availability of deadly weapons can make a big difference to a battle. For most of the 1800s, weapons might be said to have been relatively “democratic” in the sense that they were available to both governments and challengers. In hand-to-hand combat, strength and skills are vital, and governments have no great advantage. For remote combat, weapons make a much greater difference. Rifles were crucial in the 1800s, and these could be acquired and easily used by challengers. The development of the machine gun marked a turning point: it provided an overwhelming advantage against opponents without their own machine guns, which is an explanation for the one-sidedness of many colonial wars (Ellis, 1975). By the 1900s, governments were consolidating their weapons advantages, a process that has continued over the decades with air power, tanks and guided missiles (Coker, 2013; Dunnigan, 1983; Scharr, 2018).

The implication is that armed struggle against a well-armed regime has become ever more difficult in the strictly military domain. When looking at contemporary campaigns, challengers seldom have weaponry comparable to governments, so success rates are likely to be lower. Including campaigns from the 1800s skews the success rate of armed struggles, giving a misleading impression for contemporary conditions. To put it this way, as governments acquire ever more sophisticated weaponry, nonviolent campaigns are likely to have an increased rate of success compared to armed campaigns, at least when considering the fighting dimension of the struggles.

There is another factor not taken into account by either Chenoweth and Stephan or by Anisin: training techniques. As argued by Grossman (1995), most men have a great reluctance to kill others, even preferring to die rather than kill. Alarmed by a study showing that only one quarter of US soldiers fighting Germans on the front line during World War II fired their rifles, the US military developed training techniques, based on psychological research, that improved firing rates to 90% by the time of the Vietnam War. It is safe to say that few fighters in liberation movements or insurrectionary campaigns have similar training. The implication is that wherever militaries provide training to overcome instinctive reluctance to kill, they will have a huge advantage over challengers without such training. Note that this applies mainly to killing at relatively close range and not to dropping bombs, firing missiles or setting up improvised explosive devices.

If the impact of training were included in studies of the success rates of anti-regime struggles, this might help explain a declining success rate for armed struggle in the past half century. On the other hand, most of the successes of armed struggle during this period have been in parts of the world where this sort of training may not have been standard. In any case, the implication is that statistical studies need to be supplemented with analysis of the circumstances of individual campaigns. The weapons used and the training in using them, including training to overcome a reluctance to kill, should be considered.

In statistical studies of anti-regime campaigns, success is judged according to whether the stated aims of the campaigns have been achieved. This seems like a logical criterion, but there are other possible considerations that can be taken into account. One of them is the level of freedom years after the success of a campaign. Karatnycky and Ackerman (2005: 19) find that nonviolent campaigns are far more likely to lead to free societies later on.

There is another important factor worth including: death and destruction due to a campaign. On average, armed struggles lead to far more death and suffering, including trauma and its long-term effects on the individual and societal level, than nonviolent struggles. An example is the Algerian war for independence from France, 1952–1962, listed by Chenoweth and Stephan as a violent campaign that was successful. This war had a horrific toll in torture, injury and death, with sources reporting hundreds of thousands of deaths, perhaps even a million or more, in a population of only nine million. If achieving success causes massive human suffering, then this needs to be taken into account. However, there is no simple way to weigh the benefits of a successful liberation campaign against the human cost of the struggle. It is possible though to take note of the human cost and to compare it in different sorts of campaigns.

Anisin questions the causality of nonviolent action for successful political change. What he fails to consider though is that research on social movements and revolutions tends to emphasize structural conditions that lead to change and to neglect the role of agency in such socio-political transformations. Nonviolent action research contributes to this academic field, by highlighting the impact of strategic choices of protestors and resisters on societal change. Nowadays the fact that a chosen campaign strategy does make a difference with regards to the outcome of a movement is reflected in public and academic debate. Until very recently though most research accounts of revolutions were structural analyses that left out the agency of people. Chenoweth and Stephan in their book “make the case that voluntaristic features of campaigns, notably those related to the mechanisms put into effect by the resistors, are better predictors of success than structural determinants” (2011: 66). Their quantitative findings seem to have encouraged some social movement researchers to give more attention to movement dynamics, in particular to campaigners’ strategies and tactics. This contribution of nonviolent action research should be acknowledged despite all legitimate criticism of how to measure and operationalise the complex underlying dynamics. It is a positive development that the number of academics who take on a diligent empirical examination of dynamics of nonviolent action is growing and we welcome this trajectory.

In summary, Anisin’s critique of Chenoweth and Stephan’s comparison of nonviolent and violent campaigns points to several areas worthy of further exploration and development. The impact of weapons technology and the training of soldiers on the success rate of armed campaigns needs to be taken into account, and in assessing success of campaigns, more needs to be taken into account than simply whether the campaign’s goals are achieved. In particular, the human cost of the campaign — death, suffering and lasting impacts — needs to be considered. However, it is very difficult to include such factors in a statistical analysis, which points to the value of qualitative analysis.

4. “Ideological” issues

Anisin raises a number of issues concerning the validity of analysis of campaigns that point to the role of values, goals and what he calls “ideological bias.” We agree that these matters are important and need more attention. We look first at the role of values in scholarly inquiry and then at their roles in classification systems, research systems, goals, and the purpose of research.

In his critique of research by Chenoweth and Stephan, Anisin makes a comparison with scientific research, describing how caffeine researchers set up experimental and control groups. He gives the example of the sugar industry to indicate how research findings can be distorted by the influence of vested interests, suggesting that the same sort of influence affects Chenoweth and Stephan’s study. There are several points worth making about this.

Anisin’s portrait of the scientific research enterprise is an idealised one, built on the assumption that scientists are, or can be, neutral, objective searchers for the truth. This picture has been criticised by generations of scholars in the social studies of science. In practice, according to researchers in this field, scientists are always influenced by current ideas in society, by struggles for recognition and by various biases (e.g., Barnes, 1977; Mitroff, 1974). Anisin seems to subscribe to what has been called the myth of the scientific method, namely that there is a single method used by scientists that, if followed, leads to unbiased scientific knowledge (Bauer, 1992; Feyerabend, 1978).

The implication is that all researchers are influenced by their social roles, their aspirations, their peer groups, their sources of funding and prevailing assumptions in society. This is not something that can be overcome by a “pure” scientific method. Rather, it needs to be acknowledged. We agree that Chenoweth and Stephan are influenced in various ways, but so is Anisin and so are we!

With this context, it is useful to look again at the classifications of campaigns by Chenoweth and Stephan and by Anisin. The existence of a category is not neutral because it suggests a particular way of understanding the world and potentially obscures other possible ways. In other words, we find patterns in data according to our interest and expertise.

We need to ask whether certain methods of action are the driving forces behind campaigns. This is easier with nonviolent and violent campaigns. It is not so obvious in the case of unarmed violence campaigns. Are there instances in which campaigns were initiated by unarmed violence and this remained the dominant method throughout? How frequently did unarmed violence occur? Did it coincide with nonviolent methods, such as demonstrations, or did it take place in isolation?

Given that researchers are not neutral, it is likely that they will be influenced in various ways when classifying campaigns. In other words, it is worth looking at potential biases in assigning campaigns to particular categories. Chenoweth and Stephan addressed this by having subject and area experts examine documentation about campaigns and make judgements about their classification, including the primary resistance type, conflict type, outcome and peak participation (NAVCO, n.d.). By drawing on judgements by experts outside their project, they reduced the likelihood that their biases would influence their dataset.

Anisin, in contrast, drawing on a perspective different from Chenoweth and Stephan’s, uses his own judgement to assign campaigns to classifications. An approach less vulnerable to bias would be triangulation by relying on individuals with no stake in the conclusions to assign campaigns to his categories.

Gene Sharp (1973) presented a set of stages or facets of nonviolent campaigns in what he called the “dynamics of nonviolence.” Among these facets are laying the groundwork, challenge brings repression, maintaining nonviolent discipline and political jiu-jitsu. Political jiu-jitsu occurs when violence used against nonviolent campaigners generates greater support for the movement by mobilising members of the grievance group, third parties and sometimes opponents. Maintaining nonviolent discipline, which means avoiding using violence even in retaliation, as with reactive violence, is crucial to the likelihood of political jiu-jitsu.

A similar process can occur even when protesters use violence. The process of political jiu-jitsu can be generalised to other interactions, in what has been called backfire dynamics (Martin, 2007). For example, in a war, excessive violence by one side can be counterproductive, as in the case of the My Lai massacre during the Vietnam War (Gray and Martin, 2008). From this perspective, it may be more useful to look at those campaign tactics that enable backfire. Unarmed violence may not matter when the regime’s violence is much greater.

Underlying political jiu-jitsu and backfire dynamics is the question of legitimacy. It is crucial to be seen as having a legitimate cause and in being justified in actions taken (Feinberg et al., 2017; Sharp, 1959). Recall that nonviolence relies on a perception of moral superiority, from which it draws its legitimacy to make claims. This allows mobilizing participants and encourages loyalty shifts. When faced with repression, legitimacy plays an even bigger role, as the backfire effect is only activated and participation increased if repression can be framed as disproportionate and/or illegitimate. The chances of this happening decrease when nonviolent discipline is broken and campaign violence occurs (Abbs and Gleditsch, 2021). It is less likely for potential participants and supporters to identify with a campaign that uses violence, as it is more easily portrayed as unjust and illegitimate (Benford and Snow, 2000). After all, violence is often seen as morally wrong (Gandhi, 1940; Gregg, 1934/1966).

There are contexts in which an unarmed but still violent reaction by protestors may be viewed as legitimate by potential supporters, such as had been the case in some instances of the labour movement of the late 19th and early 20th century, when workers defended occupied factories against police and soldiers (Piven and Cloward, 1979: 69; 140). However, the question is always how easily it is for actions to be discredited in the eyes of the public. In general, mass support will more likely be generated for those that seem to have the moral high ground, and less likely for those who engage in human rights abuses (Feinberg et al., 2014). In other words, the outrage over repression has to outweigh the outrage over the campaign’s use of violence in order for backfire to still occur (DeNardo, 1985). Yet, because campaigns are often held to higher moral standards than state forces, even unarmed violence can have negative consequences for campaigns.

There are a number of studies showing a negative relation between unarmed violence and campaign success. Huet-Vaughn (2013) finds that riots, and in particular property destruction, lower the chances of the campaign to achieve success. Steinert-Threlkeld et al. (2021) argue that protest violence reduces participation. Abbs and Gleditsch (2021) conclude that riots have a damaging effect on nonviolent activism.

Anisin neglects to establish any explanatory correlation between unarmed violence and success rates based on violence-specific mechanisms. The question thus remains whether and how unarmed violence could be responsible for the success of the campaigns — and not just occurring despite it.

Anisin says that nonviolence research is biased towards elite interests. This may be true in some instances, but there are also many nonviolent campaigns that have challenged Western powers, for example against the neoliberal dictatorship of Pinochet in Chile or movements in Indonesia, Iran, Philippines, South Africa and West Papua. Indeed, within the nonviolence research community, and among nonviolent movements, there is considerable support for challenging US imperialism (Johansen et al., 2012). According to activist and writer George Lakey, as Gene Sharp’s concepts of the logic of change of nonviolent action played a role within movements that accelerated the collapse of the Soviet Union, people may be

confused about the nature of his project, which was to amplify the power of nonviolent struggle for whoever chooses to try it instead of using violence. Some groups did choose civil resistance instead of armed struggle to elude the grasp of Russia. Others chose it to abolish apartheid in U.S.-allied South Africa. It is available for all. (Lakey, 2019)

Nonviolent action can be viewed as a tool and, like any tool, it can be used for a variety of purposes, but — importantly — some more easily than others. A toothbrush is designed to clean teeth and can be used for many other purposes, some more easily than others. Similarly, nonviolent action has been widely used to combat elite interest, for instance against racial segregation in the US, the achievement of universal suffrage, and independence from colonial rule. Yet on the other hand, elites may have come to understand that nonviolent action may be at times more effective than more traditional approaches of secret intelligence, such as covert killings of leaders, supporting guerrillas or coup d’états that blow back against those who instigated them. It is thus plausible that the CIA or other intelligence institutions have tried to initiate or support nonviolent movements to topple a government that does not serve US elite interests. (On the role of foreign support for nonviolent movements, see Chenoweth and Stephan (2021).) Moreover, there is evidence that right wing activists have also started appropriating the toolkit of nonviolence with the purpose of spreading their explicitly Islamophobic ideology, as they have been convinced by historical and scientific evidence of the effectiveness of nonviolent methods of change (Volk, 2021).

Even if there are exceptional situations in which violence may not be detrimental to campaign success, it always causes suffering. A further question is, what type of society may result from this? There can be advantages of using violence in campaigns but also significant disadvantages, including legitimating increased repression, making violent escalation more likely and modelling a world built upon violence and human rights abuses.


Alexei Anisin (2020), in presenting a critique of statistical comparisons of nonviolent and violent anti-regime campaigns, raises a number of points worthy of further examination. Most importantly, statistical comparisons reduce complex social interactions into simplified sets of numbers that cannot capture the nuances of actual cases. The question then becomes, which simplifications are most useful for gaining greater insight about general patterns and for guiding campaigns against repression and oppression? Furthermore, how can qualitative and statistical studies be used together to inform each other?

Chenoweth and Stephan (2011) have done pioneering research in compiling a dataset of anti-regime, secession and anti-occupation campaigns during the period 1900–2006, classifying each campaign as either nonviolent or violent. Anisin points to a shortcoming in the dichotomy between nonviolent and violent, and proposes two additional categories, unarmed violent and reactive violent. However, given that nearly every campaign contains a mixture of methods — for example, armed struggles nearly always involve methods of nonviolent action — it becomes important to justify classification systems. One way to do this — the way used by Chenoweth and Stephan — is by judging the predominant mode of struggle. The mere presence of unarmed violence, for example at the end of a campaign, may not be enough to justify putting the entire campaign in the category of unarmed violence. Anisin does not discuss this vital methodological point.

It is worth noting that talking of a predominant mode of struggle is as much normative as scientific — as is assessing this mode from a scholarly as well as an activist point of view. Part of what social movements do is to influence understandings of what is considered violent and nonviolent action, by shifting public opinion and what is accepted and perceived as common sense by raising awareness through their actions, internal discussions, public statements and how they frame their activities. One result is the interplay between movement and scholarly understandings and hence the making sense of tactics, dichotomies and classifications.

Anisin, in presenting his revised dataset, includes campaigns in the 1800s. This raises an issue addressed neither by Anisin nor by Chenoweth and Stephan: the roles of technology and training. We have noted that weapons technology after the 1800s, including the machine gun and aerial bombing, provided a major advantage to regimes with these capacities. Further, training techniques developed after World War II make soldiers far more willing to kill, giving a great advantage to forces so trained. Developments in both weapons and training mean that armed anti-regime struggles are less likely to succeed. It would be worth taking these developments into account in future studies.

At a more ideological level, Anisin invokes the methods of natural science to imply the existence of bias in studies by civil resistance scholars. Although natural science is not value-free, Anisin’s general point is worthwhile: it is valuable for scholars to take note of influences on their work from sources of funding and, as well, presuppositions deriving from social position, disciplinary focus, and possible applications. Anisin points to the influence of US government interests on civil resistance research; we noted that a considerable portion of research (and practice) in the field is counter to the interests of US imperialism.

Military research and development is a vast enterprise employing thousands of scientists and costing billions of dollars annually. Arms manufacturers sell their wares to governments worldwide, and technology corporations sell surveillance equipment to repressive regimes. To challenge the massive military and policing systems is a daunting enterprise. Supporters of civil resistance, though, have comparatively meagre human and material resources, yet many campaigns have succeeded despite the massive imbalance in weapons and training.

Nonviolent protest does not involve human rights abuses by the protesters, whereas guerillas may perpetuate spirals of violence and trauma; nonviolence also has strategic advantages in contexts in which the use of violent means may reduce legitimacy and support in the public domestically and internationally. This is also known to governments who may resort to forms of soft repression when overt physical violence might backfire against them and thus can be detrimental to the goal of silencing protest.

The hope of many civil resistance advocates is that more research and experience will improve the effectiveness of campaigns. A valuable part of improving nonviolence theory and practice is critical analysis, and for this we thank Alexei Anisin for his thought-provoking critique.


We thank Chris Brown, Véronique Dudouet, Janjira Sombutpoonsiri and Stellan Vinthagen for helpful discussions, and Kristian Skrede Gleditsch, Benjamin Isaak-Krauß, Tom Weber and an associate editor for valuable comments on drafts.


1.         For example, Burrowes (1996: 239), in contrasting British colonial practices in India and Kenya, refers to “the frequent and naive claim that it was the British moral code that made Gandhian nonviolence possible in India,” citing three authors who made this claim.

2.          The 2% figure is drawn from a database being prepared by L Mitchell and M Onken titled Measuring the Intensity of Nonviolent Discipline, intended for publication when complete.


Abbs L and Gleditsch KS (2021) Ticked off, but scared off? Riots and the fate of nonviolent campaigns. Mobilization 26(1): 21–39.

Anisin A (2020) Debunking the myths behind nonviolent civil resistance. Critical Sociology 46(7/8): 1121–1139.

Barnes B (1977) Interests and the Growth of Knowledge. London: Routledge and Kegan Paul.

Bartkowski MJ, ed. (2013) Recovering Nonviolent History: Civil Resistance in Liberation Struggles, Boulder, CO: Lynne Rienner Publishers.

Bauer HH (1992) Scientific Literacy and the Myth of the Scientific Method. Urbana: University of Illinois Press.

Belgioioso M, Costalli S and Gleditsch KS (2019). Better the devil you know? How fringe terrorism can induce an advantage for moderate nonviolent campaigns. Terrorism and Political Violence 33(3): 596–615.

Benford RD and Snow DA (2000). Framing processes and social movements: An overview and assessment. Annual Review of Sociology 26(1), 611–639.

Bondurant JV (1958/1988) Conquest of Violence: The Gandhian Philosophy of Conflict. Princeton: Princeton University Press.

Boserup A and Mack A (1974) War Without Weapons: Non-violence in National Defence. London: Frances Pinter.

Bramsen I (2018) How civil resistance succeeds (or not): micro-dynamics of unity, timing, and escalatory actions. Peace & Change 43(1): 61–89.

Burrowes RJ (1996) The Strategy of Nonviolent Defense: A Gandhian Approach. Albany: State University of New York Press.

Burrows M and Stephan MJ, eds. (2015) Is Authoritarianism Staging a Comeback? Washington, DC: The Atlantic Council.

Chenoweth E (2021) Civil Resistance: What Everyone Needs to Know. New York: Oxford University Press.

Chenoweth E and Stephan MJ (2011) Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict. New York: Columbia University Press.

Chenoweth E and Stephan MJ (2021) The Role of External Support in Nonviolent Campaigns. Washington DC: ICNC Press.

Chenoweth E and Schock K (2015) Do contemporaneous armed challenges affect the outcomes of mass nonviolent campaigns? Mobilization: An International Quarterly 20(4): 427–451.

Churchill W (2007) Pacifism as Pathology: Reflections on the Role of Armed Struggle in North America. Edinburgh: AK Press.

Coker C (2013) Warrior Geeks: How 21st-Century Technology is Changing the Way We Fight and Think about War. London: Hurst & Company.

Csapody T and Weber T (2007) Hungarian nonviolent resistance against Austria and its place in the history of nonviolence. Peace & Change 32(4): 499–519.

Creswell JW and Clark VLP (2017) Designing and Conducting Mixed Methods Research, 3rd ed. Los Angeles: Sage.

Dahl M, Gates S, Gleditsch KS and González B (2021) Accounting for numbers: group characteristics and the choice of violent and nonviolent tactics. Economics of Peace and Security Journal 16(1): 5–25.

DeNardo J (1985). Power in Numbers: The Political Strategy of Protest and Rebellion. Princeton, NJ: Princeton University Press.

Dudouet V, ed. (2015) Civil Resistance and Conflict Transformation: Transitions from Armed to Nonviolent Struggle. London: Routledge.

Dunnigan JF (1983) How to Make War: A Comprehensive Guide to Modern Warfare. New York: Quill.

Ellis J (1975) The Social History of the Machine Gun. London: Croom Helm.

Feinberg M, Willer R and Kovacheff C (2017) The activist’s dilemma: Extreme protest actions reduce popular support for social movements. Journal of Personality and Social Psychology 119(5): 1086–1111.

Feyerabend P (1978) Science in a Free Society. London: NLB.

Gamson WA (1975) The Strategy of Social Protest. Belmont, CA: Wadsworth Publishing.

Gamson WA (1989) Reflections on The Strategy of Social Protest. Sociological Forum 4(3): 445–467.

Gandhi MK (1940) An Autobiography or the Story of My Experiments with Truth, 2nd ed. Ahmedabad: Navajivan.

Gelderloos P (2007) How Nonviolence Protects the State. Cambridge, MA: South End Press.

Gray T and Martin B (2008) The American war in Indochina: injustice and outrage. Revista de Paz y Conflictos 1: 6–28.

Gregg RB (1934/1966) The Power of Nonviolence, 2nd ed. New York: Schocken Books.

Grossman D (1995) On Killing: The Psychological Cost of Learning to Kill in War and Society. Boston: Little, Brown.

Huet-Vaughn E (2013) Quiet Riot: The Causal Effect of Protest Violence. PhD Thesis, University of California, Berkeley, USA.

Huxley SD (1990) Constitutionalist Insurgency in Finland: Finnish “Passive Resistance” against Russification as a Case of Nonmilitary Struggle in the European Resistance Tradition. Helsinki: Finnish Historical Society.

Johansen J and Martin B (2019) Social Defence. Sparsnäs, Sweden: Irene Publishing.

Johansen J, Martin B and Meyer M (2012) Nonviolence versus US imperialism. Economic & Political Weekly 47(38, 22 September): 82–89.

Karatnycky A and Ackerman P (2005) How Freedom is Won: From Civic Resistance to Durable Democracy. New York: Freedom House.

Lakey G (2019, July 16) Will the real Gene Sharp please step forward? Waging Nonviolence. https://wagingnonviolence.org/2019/07/gene-sharp-cold-war-intellectual-marcie-smith/

Malm A (2021) How to Blow Up a Pipeline. London: Verso.

Martin B (2007) Justice Ignited: The Dynamics of Backfire. Lanham, MD: Rowman & Littlefield.

Martin B (2013) The politics of Gene Sharp. Gandhi Marg 35(2): 201–230.

NAVCO (Mapping Nonviolent and Violent Campaigns and Outcomes) (n.d.) Variable descriptions. (Cambridge, MA: Harvard University, Weatherhead Center for International Affairs) https://navcomap.wcfia.harvard.edu/variable-descriptions

Parkman P (1990) Insurrectionary Civic Strikes in Latin America 1931–1961. Cambridge, MA: Albert Einstein Institution.

Pinckney J (2016) Making or Breaking Nonviolent Discipline. Washington, DC: ICNC Press.

Piven FF and Cloward R (1979) Poor People’s Movements. New York: Vintage.

Scharre P (2018) Army of None: Autonomous Weapons and the Future of War. New York: Norton.

Schock K (2005) Unarmed Insurrections: People Power Movements in Nondemocracies. Minneapolis, MN: University of Minnesota Press.

Sharp G (1959) The meanings of non-violence: a typology (revised). Journal of Conflict Resolution 3(1): 41–64.

Sharp G (1973) The Politics of Nonviolent Action. Boston: Porter Sargent.

Sharp G (1979) Gandhi as a Political Strategist. Boston: Porter Sargent.

Sharp G (1980) Social Power and Political Freedom. Boston: Porter Sargent.

Sharp G (1993/2010) From Dictatorship to Democracy: A Conceptual Framework for Liberation, 4th ed. East Boston, MA: Albert Einstein Institution.

Sharp G with Paulson J, Miller CA and Merriman H (2005) Waging Nonviolent Struggle: 20th Century Practice and 21st Century Potential. Boston: Porter Sargent.

Stephan MJ and Chenoweth E (2008) Why civil resistance works: the strategic logic of nonviolent political conflict. International Security 33(1): 7–44.

Steinert-Threlkeld Z, Chan A and Joo J (2021) How state and protester violence affect protest dynamics. APSA Preprints. doi: 10.33774/apsa-2019-bv6zd-v3.

Vinthagen S (2015) A Theory of Nonviolent Action: How Civil Resistance Works. London: Zed Books.

Volk S (2021, March 25) How the German far Right appropriates ideals of non-violent resistance. Open Democracy. https://www.opendemocracy.net/en/countering-radical-right/how-german-far-right-appropriates-ideals-non-violent-resistance/

Zunes S, Kurtz LR and Asher SB, eds. (1999) Nonviolent Social Movements: A Geographical Perspective. Oxford: Blackwell.

For correspondence: Brian Martin, Humanities and Social Inquiry, University of Wollongong, NSW 2522, Australia. Email: bmartin@uow.edu.au

Go to

Responses to critiques of nonviolence

Brian Martin's publications on nonviolence

Brian Martin's publications

Brian Martin's website