Does Torture Work?

Guest post by Christopher Sullivan

More precisely, does committing torture enable state forces to limit subsequent acts of violence? Over the course of the past few decades, political practitioners and scholars alike have debated the merits of using torture as a tool to combat the international and domestic threats of terrorism and insurgency. Concerns over the morality and potential effects of torture have emerged again recently in response to the revelation of new material on CIA torture practices as well as the increasingly widespread application of torture in the Syrian conflict. Yet despite the rigor of this discourse, torture remains one of the most widely practiced instruments of coercion. An average of 80% of states are alleged to have practiced torture in any given year between 1981 and 2010, including nearly all states engaged in counterinsurgency operations. This suggests an enduring belief torture should assist the state in its efforts to target its enemies and reduce their capacity to commit violence.

Where evidence has been brought to bear on the subject, it has been at best anecdotal and at worst cherry picked to support the position of the claims maker.[1] Torture opponents often identify the case of Abu Zubaydah, the first high profile al-Qaeda member captured by the US. According to the FBI agent in charge of interrogating Zubaydah, using “conventional interrogation methods,” such as deception and rapport building, he was able to convince Zubaydah to identify alleged 9/11 mastermind Khalid Sheik Mohammed. After Mohammed was named, however, the CIA began to torture Zubaydah, and, according to the FBI agent, this was the point at which he stopped talking.[2] Torture proponents, by contrast, tend to focus on Mohammed’s behavior once he was captured. The CIA claims that once they captured Mohammed he refused to provide them with any actionable intelligence. However, according to the Washington Post, after being “subjected to an escalating series of coercive methods, culminating in 7 ½ days of sleep deprivation, while diapered and shackled, and 183 instances of waterboarding,” Mohammed began cooperating with the authorities – providing them with details of both al-Qaeda’s organizational structure and plans for future attacks.[3]

The bias associated with selecting evidence to support particular arguments has severely limited our ability to draw general conclusions about the effects of torture from these cases. More evidence is likely to emerge in response to the recent US Senate vote to declassify a trove of relevant documents. However, at present there remains little systematic analysis on torture’s effects in either agency reviews of torture or in a nascent academic literature evaluating torture’s effects.[4]

Recognizing that the empirical evidence surrounding instances of torture presents inherent limitations and potential biases should not repudiate efforts to scientifically study the effects of this policy. Arguments negating our ability to systematically evaluate torture’s effects ignore recent advances in the measurement of state violence, and allow torture proponents to sidestep cause-effect questions with claims that, as Bagaric and Clark (2007) put it, such evaluations would lead to a “distracting and superficial numbers game.”

In a new article at the Journal of Peace Research, I bring to bear micro-level data from Guatemala to generate a systematic evaluation of how torture affects violence within the context of an organized insurgency. This is a case in which highly skilled military personnel tortured with near impunity. Among other tactics, agents of the Guatemalan military forced the victims to stand hooded for hours or days, forced them to eat excrement, forced them to stay awake for days at a time, refused to give them food or water, subjected them to electric shocks, stripped them naked, burned them with cigarettes, suspended them from chains, sexually abused them, submerged them in water, cut them and broke their fingers. Combining data from Guatemala’s Commission for Historical Clarification with a research strategy designed to overcome many of the hurdles associated with causal inference, the analysis identifies how local-level dynamics of violence change in the aftermath of torture. The study examines torture’s impacts on subsequent killings perpetrated by both insurgents and counter-insurgents.

Two trends emerge from the analysis:

• First, torture has no identifiable systematic association with decreases in insurgent perpetrated killings.
• Second, torture is shown to be robustly associated with increased killings perpetrated by counterinsurgents.

Such evidence should lend pause to those who would consider employing torture, at least within the context of an insurgency. Justifications for torture do not rest on the contention that engaging in torture will reveal information, but on arguments that engaging in torture will allow state agents to somehow stop challengers from engaging in violence. If torture cannot produce discernable effects on insurgent violence then any immediate effects by torture, including the revelation (or non-revelation) of information, are of little consequence.

In this case, not only did torture display no relation to decreases in killings perpetrated by insurgents, but it had a somewhat pathological quality of being strongly associated with increases in other forms of counter-insurgent violence. The evidence suggests that insurgents were able to outmaneuver the forces employing torture, for example by adapting their organizations and strategies in response to torture or by rallying popular support against the use of torture. Subsequent counter-insurgent strategies appear to have been far less variable. The use of torture was strongly associated with increases in killings committed by counterinsurgents in the locality where torture took place as well, in some cases, in surrounding areas.

The evidence shines light on the impacts of one particularly controversial form of political repression. But it can also help resolve a puzzle that has vexed scholars for nearly 40 years — understanding more generally how violations of human rights influence decisions to participate in conflict. Questions remain about how other forms of government coercion influence decisions to commit violence or otherwise engage in collective action. It is also relevant to question how the impacts of repressive tactics such as torture might be influenced by the broader context of civil war violence and the institutions it generates. By looking at the operation of specific tactics within clearly demarcated contexts, rather than the operation of aggregate measures of repression, it should be possible to add nuance and scope conditions to improve upon existing theory.

Christopher Sullivan is a PhD candidate in the Department of Political Science at the University of Michigan and a Fellow at Yale University’s Program on Order, Conflict, and Violence.

[1] As an example, former Vice President Dick Cheney has famously argued that the effectiveness of torture was “proven” by the fact that there were no “massive casualty” attacks in the US since 9/11.

[2] Specifically, as Hajjar (2009) notes, the alleged torture included being “waterboarded 83 times, confined in a coffin-like box with insects, and subjected to brutal and degrading treatment for the duration of his detention in CIA black sites.”

[3] Reports from other sources claim that after the stopped torturing Mohammed, the CIA returned to rapport based interrogation methods. These activities are not included in the Post’s article. See Shane, (2009).

[4] See CIA (1963), Lehner (2006), Rejali (2007), Walsh (2009), and Regan (2009).