One common trait of repressive governments or laws is the emergence of an organized resistance, often involving high-ranking officials and civil figures who aren’t keen on obeying their leaders.

Obedience has been studied for decades, as an insight into how humans respond to authority and factor in their own sense of morality. In the 1960s, social psychologist Stanley Milgram carried out a controversial experiment on obedience to authority. He asked participants to give a series of electric shocks of increasing voltage to a test subject (who was actually an actor) whenever he gave a wrong answer to a memory test. Participants obeyed and entered in what Milgram described as an “agentic” state, following the orders of the experimenter.

Recently, we took another look at obedience, drawing on Milgram’s tests, but safely within the realm of virtual reality. We recorded a real actor’s performance, using a cutting-edge Hollywood-type motion capture setup, and portrayed the actor and his movements as a virtual human character (an “avatar').

We then asked students to give shocks to the avatar we’d created. Unlike in Milgram’s test, our participants were not deceived regarding the actual harm to the subject when his answers to the memory test were wrong—since he wasn’t a real person. Using this setup, we could ethically test how humans respond to authority.

The participants in the experiment nevertheless had difficulty with shocking the avatar: Although 85 percent of them completed the task, they exhibited measurable signs of stress. However, when we looked more closely at the actual interactions that people had with the virtual victim, we found that participants were trying to cheat, even if unconsciously, by giving cues to the avatar, signaling the right answers with a louder voice tone.

These findings challenge the conventional accounts of obedience that have been in place since the Nuremberg trials after World War II and apparently supported by Milgram’s own experiments. In our case we had clear evidence of a kind of disobedience among our participants. They did not enter an “agentic” state, blindly and carefully carrying out the orders of the experimenter, as executioners of harmful behavior. Instead they fit more the profile of an “engaged follower,” someone who apparently engages but nevertheless tries to get around the specifics of the orders. Essentially, they were disobeying or quietly resisting while appearing to follow orders.

However, those who resist authority also pay a big psychological toll. Even in the act of apparently obeying, they will follow orders that go against their own moral principles. Our experiments were presented to participants as a way to contribute to scientific research, and two in three of the participants who identified less strongly with that goal were more likely to quit.

By contrast, those who declared themselves supporters of scientific advancement continued to carry out their duties. Interestingly however, the supporters of science were also the ones who tried harder to help the avatar. We believe that by the act of helping, they reduced their levels of stress and were able to continue their task longer, however harmful the task was to the victim.

This finding suggests that people inflict harm despite the fact that they might actually care about their victims. And anyone who watches the original films of Milgram’s experiment will see firsthand that the participants who continued that experiment suffered psychological trauma. In our experiment, we did not get that far, but measured the stress levels before and after the experiment. At the end, even those who had cheated showed an increased stress level.

If we look at our experiments as a proxy for resistance to authority, we can anticipate a psychological cost to the resisters. Even though their obedience isn’t genuine, those who persist endure additional stress compared to those who decide to quit. In the long term they will also be facing the moral dilemma of engaged followership, wondering whether they engaged too much and in essence enabled a leader they did not want to obey.

Read the full report on the experiments as published in PLOS One.