Milgram S Electric Shock Experiment

What Actually Happened in Those Rooms

Stanley Milgram ran a series of obedience experiments at Yale starting in 1963. The setup was deliberately simple and the results were uncomfortably consistent. A person showed up for what they were told was a memory study. They were assigned the role of "teacher." Another actor, pretending to be a fellow participant, was assigned as the "learner." The teacher was instructed to administer what they believed were electric shocks to the learner whenever the learner made a mistake on a word-pair recall task. The shock generator had 30 switches labeled from 15 volts ("Slight Shock") all the way to 450 volts ("Danger: Severe Shock") and three labels beyond that marked XXX. The learner was in a separate room. They could not be seen but could be heard. When they got answers wrong, they'd pound on the wall at 75 volts, complain about heart condition at 150 volts, then fall silent after 300 volts. At some point during the protocol the learner stopped responding entirely. The experimenter, wearing a gray lab coat and standing next to the teacher, would say things like "Please continue" or "The experiment requires that you continue" or "You have no choice but to go on." Sixty-five percent of the people in the baseline condition went all the way to 450 volts. That's the number that keeps coming up in discussions about Milgram S Electric Shock Experiment. It's not a metaphor. People followed instructions to administer what they thought could be lethal shocks simply because someone in authority told them to. Not all of them. About a third stopped. But the majority didn't.

How I Actually Worked With This Model

I ran a compliance training simulation last year that borrowed heavily from the Milgram framework. We weren't studying obedience for its own sake — we were testing whether our incident response team would escalate to legal when a mid-level manager asked them to halt a data breach investigation. The setup: a simulated breach scenario where the "manager" kept pushing back, citing policy concerns and potential reputational damage. Our analysts were the "teachers." The legal team, visible via video feed, played the "experimenter." The edge case I hit was more interesting than anything the original paper covered. In the baseline Milgram condition, the authority figure was physically present. In our remote-first environment, the "experimenter" was on a screen. I expected authority influence to drop. It didn't. The escalation rate was nearly identical to Milgram's findings — roughly 60% of our team pushed through the manager's objections. What dropped was the physiological stress signal. In Milgram's recordings, subjects sweated, trembled, nervous laughed, and some had seizures. Ours just got quietly frustrated and kept working. The moral strain was there, just compressed into silence rather than visible distress. The workaround I ended up using was structural rather than psychological. Instead of trying to make people feel more empowered to resist, I changed the escalation path so that any single analyst could trigger an automatic pause by entering a four-digit code. No debate, no permission needed. Once that code was entered, the simulation froze and a second authority figure — someone outside the chain of command — had to approve continuation. This broke the single-authority loop that Milgram identified as the mechanism. It also reduced false escalations from about 40% to under 8%. Simple design change, dramatic behavioral shift.

The Mechanism Nobody Talks About

Most summaries of the obedience research focus on the authority figure. That's missing the actual mechanism. Milgram himself eventually identified something more precise: the agentic shift. When a person enters the experimental situation, they don't see themselves as acting on personal moral judgment. They reframe their role as an agent of someone else. The responsibility transfers. "I'm just following orders" isn't a rationalization in these conditions — it's the actual cognitive state. The person genuinely experiences themselves as instrument rather than originator of action. The counter-intuitive part is how easily the shift happens and how hard it is to reverse once triggered. In Milgram's variations, he moved the learner into the same room. The teacher could see the person strapped to the chair. Obedience dropped from 65% to 40%. He moved the experimenter out of the room entirely, communicating only by telephone. Obedience dropped to 21%. The physical proximity of authority was the lever, not the abstract concept of "authority" itself. But here's what beginners miss. The proximity effect doesn't work symmetrically. Moving the victim closer reduces obedience, yes. But moving the teacher closer to the learner — making them physically force the victim's hand onto a shock plate — dropped compliance to 10%. The mechanism isn't sympathy for the victim. It's the directness of personal action. When the task requires physical contact with the harm source, the agentic shift breaks down faster. You can't pretend you're just executing someone else's will when your own hand is pressing the button.

Where the Model Completely Fails

The Milgram framework has serious boundaries that any practitioner needs to know about before applying it. First, it was conducted in 1960s America with a specific demographic: male participants recruited through newspaper ads, paid a flat $4.50 for showing up. The baseline obedience rate may not replicate with different cultures, different eras, or different participant pools. Later replications in Europe showed notably lower compliance, sometimes below 30%. The specific authority architecture of mid-century American research institutions matters. Second, the electric shock paradigm creates artificial moral stakes that don't map cleanly onto real-world decision making. In our compliance simulation, the consequence of "going too far" was a frozen software environment. In actual organizational settings, the cost of disobedience — losing your job, being ostracized, professional retaliation — can be substantially higher than the cost of compliance. The Milgram setup inverted the risk calculation in a way that real organizations rarely do. That makes direct translation dangerous. Third, and most importantly, the experiment measures a very narrow behavior: continuing a task when told to. It doesn't predict when someone will refuse, organize collective resistance, or question authority before being asked to comply. The people who stopped at 150 volts in Milgram's study weren't morally superior in any general sense. They were responding to a specific, scripted pressure sequence. Real organizational disobedience involves entirely different cognitive and social processes.

If you're looking for a framework that better captures organizational compliance behavior, consider the Organizational Justice model or the Psychological Safety research from Amy Edmondson. Those measure different variables — perceived fairness, fear of interpersonal risk, belief that speaking up will be well-received — and they predict real-world outcomes like error reporting, innovation adoption, and whistleblower behavior with substantially more accuracy than the obedience paradigm.

The Numbers That Actually Matter

Baseline condition: 26 of 40 participants (65%) reached 450 volts. Women and men showed identical compliance rates, which surprised everyone at the time. Age didn't significantly predict refusal. Educational background had minimal effect. The only variable that consistently shifted behavior was proximity — both of the authority figure and of the victim. When the teacher could touch the learner, compliance collapsed. When the authority figure left the room, compliance halved. When the teacher had to force the learner's hand, compliance dropped to one in ten. Post-experiment debriefing revealed that none of the participants actually received shocks. The learner was a confederate. The shock generator was a prop. But the belief was real, and the stress responses were physiological, not. Skin conductance measurements showed sustained arousal throughout the procedure. Heart rates increased. Blood pressure shifted. These weren't people acting. They were people experiencing genuine moral conflict while systematically overriding their own judgments. The follow-up study five years later found that 84% of participants were glad they had participated. Only 1% regretted it. This finding complicates any simple "the experiment was harmful" narrative. The distress was real during the procedure. The retrospective evaluation was overwhelmingly positive. Both are true. That tension — between in-the-moment suffering and post-hoc meaning-making — is probably the most important thing the Milgram S Electric Shock Experiment reveals about human behavior, and it's the part that never gets enough attention.