Social Perceptions of Torture: Genuine Disagreement, Subtle Malleability, and In-Group Bias

Abstract

This article explores whether people agree on what acts do and do not constitute torture and, if not, what social and contextual factors affect such perceptions. Some have suggested that there is a shared, commonsense definition of torture. If so, people should agree on whether particular acts constitute torture, and judgments should be unaffected by extraneous factors. In Study 1, however, American undergraduates disagreed about whether many acts constitute torture. In Study 2, they judged fewer acts as torture when they were embedded among more aversive acts. In Study 3, conservatives considered fewer acts torture when they were performed by Americans upon Iraqis than when they were performed by Iraqis upon Americans. This article discusses the implications of differing definitions of torture and recent controversies over treatment of detainees, recognizing that peaceful negotiations may well depend on how negotiating entities view their treatment at the hands of the adversary.

ACKNOWLEDGMENTS

Portions of this research were presented at the annual meeting of the Association for Psychological Science, New York, May 2006; and the Judgment and Decision-Making Preconference at the Society for Personality and Social Psychology, Memphis, January 2007. We thank John Cacioppo, Erin Hardin, Bryan Hardin, Peter McGraw, Katharine Hayhoe, Barbara Mellers, and members of our lab for helpful comments; and Monica Warren and Rodrigo Neely-Recuero for research assistance.

Notes

¹Of the 100 scenarios, there were 61 ties (e.g., 5 scenarios were considered torture by 91.7% of participants). Ties must be broken randomly, but doing so can spuriously affect the coefficient of reproducibility (CR). Therefore, we conducted 30 iterations with ties sorted in different random orders. CRs ranged from .848 to .851. To examine whether the mere presence of ties spuriously reduced CR, we conducted 30 additional iterations, each with a subset of 39 items randomly selected to exclude ties. Resulting CRs ranging from .810 to .834. Errors were counted with the Goodenough-Edwards method throughout (McIver & Carmines, 1981).

²All reported effects were significant at the .05 level. Cohen's d (Cohen, 1977) is a measure of effect size; ds of 0.2, 0.5, and 0.8 are interpreted as small, medium, and large effects, respectively.

³The institutional review board was made aware that many of the scenarios we used came from public sources, and that participants were likely to be exposed to similar information in the news. Furthermore, the consent form made participants aware that they would be exposed to potentially disturbing information, and reiterated that their participation was voluntary and that they were free to withdraw at any time.

⁴Scenarios considered torture by all of Study 1's participants were excluded so that the extremely aversive contextual scenarios would more closely mirror the mildly aversive contextual scenarios.

⁵For Study 2, we were constrained by the dual necessities of maintaining a 2:1 context-to-target stimulus ratio and equal distances of mean percentage of endorsement among the three stimulus sets. For Study 3, we simply selected all the scenarios that fell an equal distance above and below 50% endorsement, many of which were also moderately aversive Study 2 scenarios. In both studies, three scenarios that fell within the relevant ranges were thought to potentially have differing implications depending on the religion or ideology of the actor or victim in question (either implied or specifically stated, as in Study 3), and were avoided: “The guard kicks the prisoner's religious text” (Studies 2 & 3), “The guard forces the prisoner to strip naked” (Study 3), and “The guard forces a vegetarian prisoner to eat a steak” (Study 3).