Coercion Changes the Sense of Agency in the Human Brain

Abstract

People may deny responsibility for negative consequences of their actions by claiming that they were "only obeying orders." The "Nuremberg defense" offers one extreme example, though it is often dismissed as merely an attempt to avoid responsibility. Milgram's classic laboratory studies reported widespread obedience to an instruction to harm, suggesting that social coercion may alter mechanisms of voluntary agency, and hence abolish the normal experience of being in control of one's own actions. However, Milgram's and other studies relied on dissembling and on explicit measures of agency, which are known to be biased by social norms. Here, we combined coercive instructions to administer harm to a co-participant, with implicit measures of sense of agency, based on perceived compression of time intervals between voluntary actions and their outcomes, and with electrophysiological recordings. In two experiments, an experimenter ordered a volunteer to make a key-press action that caused either financial penalty or demonstrably painful electric shock to their co-participant, thereby increasing their own financial gain. Coercion increased the perceived interval between action and outcome, relative to a situation where participants freely chose to inflict the same harms. Interestingly, coercion also reduced the neural processing of the outcomes of one's own action. Thus, people who obey orders may subjectively experience their actions as closer to passive movements than fully voluntary actions. Our results highlight the complex relation between the brain mechanisms that generate the subjective experience of voluntary actions and social constructs, such as responsibility.

Figure 1

Experimental Setup Schematic representation of the coercive condition (top) and the free-choice condition (bottom). In this condition, the experimenter looked elsewhere. In the coercive condition, the experimenter ordered the agent at each trial either to take money from her co-participant (financial harm group) or to deliver a shock (physical pain group). The experimenter stood next to the agent and looked at her throughout the whole condition.

Figure 2

Schematic Representation of the Apparatus during the Experiment The agent saw trial-by-trial feedback on the computer screen, whereas the “victim” did not. The agent pressed “F” on a keyboard to inflict harm and earn money or “H” not to inflict harm/earn money. Both the agent and the “victim” gave independent written estimates of action-tone intervals on an answer sheet. Electrodes connected to the stimulator were placed on the “victim’s” left hand, which was clearly visible to the agent.

Figure 3

Interval Estimation Results (A) Effects of coercion on interval estimates in experiment 1 in each group. The data for different action-tone intervals are shown to demonstrate interval estimation performance, but this factor was not central to our predictions. Error bars show SEs. Coercion consistently prolonged interval estimates. ^∗∗∗ indicates a p value <0.001. (B) Effects of coercion on interval estimates in experiment 2. ^∗∗∗ indicates a p value <0.001.

Figure 4

Judgments of Responsibility in Each Condition in Experiment 2 ^∗∗∗ indicates a p value ≤0.001. ^∗∗ indicates a p value between 0.001 and 0.01. Error bars show SEs.

Figure 5

Neural Response to Outcome Tones: Experimental Conditions Graphical representation of the auditory N1 amplitude in the free-choice (light blue) and the coercive (dark blue) conditions when (A) an electrical shock was delivered at the same time as the tone or (B) no electrical shock occurred. Topographical representations display the activity along the whole scalp.

Figure 6

Neural Response to Outcome Tones: Control Conditions (A) Auditory N1 amplitude in active (light green) and the passive (dark green) conditions. (B) Mean amplitude of the auditory N1 in all conditions. ^∗∗ indicates a significant difference (two-tailed, p ≤ 0.01). Error bars show SEs. (C) Topographical maps in active and passive conditions.