Short answer
The answer in plain English
Ordinary people often obey bad instructions because the request arrives through a trusted role, begins with a plausible small step, and keeps moving before they reassess it. Authority and familiar social scripts reduce uncertainty; gradual commitment makes stopping awkward; and a chain of command can make responsibility feel as if it belongs elsewhere. None of these forces removes choice, but together they can make compliance feel like the normal option.
Why it matters
What to understand
Obedience is a useful coordination shortcut, not evidence of a defective mind. It becomes dangerous when legitimacy is imitated, requests escalate in small increments, and nobody is clearly accountable for the outcome. Milgram's experiments showed how strongly a situation can pull behavior, but they did not establish a fixed obedience rate or prove that participants stopped thinking. Practical resistance works by breaking momentum: pause, verify through an independent channel, name the concrete consequence, and use a pre-agreed procedure for refusing unsafe requests.
Visual guide
How the pieces fit together
Obedience usually begins as sensible cooperation
A caller claims to be from your bank. They know your name, describe a plausible security problem, and ask you to confirm one harmless-looking detail. The dangerous request comes later, after the interaction already feels official.
That sequence matters. Most people do not accept an obviously destructive command in one clean decision. They cooperate with a role that appears legitimate, take a small first step, and then continue through a series of requests whose meaning changes gradually.
Obedience is not inherently foolish. Hospitals, aircraft, emergency crews, and ordinary workplaces depend on people accepting expertise and coordinating quickly. The same shortcut can fail when the signs of expertise—the title, uniform, confident voice, official vocabulary, or polished dashboard—are easier to copy than the expertise itself.
The useful question is therefore not “Why are some people gullible?” It is “What features of a situation make continued compliance feel normal?”
Social scripts make refusal feel like a new problem
Familiar interactions come with scripts. A receptionist requests identifying information. A manager assigns work. A technician asks you to follow a sequence of steps. The script saves time because neither side must renegotiate the relationship with every sentence.
A bad instruction can borrow that structure. Once the interaction is underway, refusing may require you to accuse someone of dishonesty, challenge a superior, or admit that you do not understand a process everyone else appears to accept. The social cost is immediate and visible; the possible harm may still feel abstract.
This changes the internal question. Instead of asking whether the action is safe or justified, a person asks whether they are expected to perform it. The role seems to have answered before the evidence has been examined.
Imposter scams use exactly this combination of authority and momentum. The FTC reported $3.5 billion in losses to imposter scams in 2025. A demand for an entire savings balance would trigger suspicion. Confirming a name, reading a code, or staying on the line can feel like routine account security. Each completed step makes the next one look less like a fresh decision.
Small commitments can conceal a large boundary crossing
Psychologists use “foot in the door” for situations in which agreement to a small request can increase compliance with a larger related request. It is not a spell and does not work reliably on everyone. Its importance here is simpler: gradual escalation makes the boundary hard to see.
Moving from step four to step five may feel trivial even when step ten is something you would have rejected immediately. Continuing also protects a comfortable self-story. If you have cooperated for several minutes, stopping means considering that the earlier cooperation may have been a mistake.
Good procedures counter this by defining boundaries in advance. A payment-account change might always require a second person. Passwords and one-time codes might never be shared during an incoming call. A worker might have explicit authority to stop equipment without first proving that a hazard exists. These rules replace improvisation under pressure with a safer script.
Responsibility can feel diluted without disappearing
Harmful actions are often split across a chain. A manager approves a plan, software recommends a choice, a policy supplies the language, and an employee presses the button. Each participant handles only part of the process, so the consequence can feel as if it belongs somewhere else.
Laboratory work on coercion found that orders can reduce both reported responsibility and an experimental measure of agency. That does not erase moral or legal accountability. It helps explain why someone may experience an action as execution of another person’s decision rather than as their own choice.
The remedy is to reconnect action and consequence. Ask who bears the loss if the instruction is wrong. Name the person affected. Translate “complete the security procedure” into “transfer money to an account controlled by someone you have not independently verified.” Bureaucratic language becomes less protective when the outcome is stated plainly.
What Milgram showed—and what the slogan misses
Stanley Milgram’s early-1960s studies placed participants in the role of a “teacher.” An experimenter instructed them to administer what appeared to be increasingly severe electric shocks to another person for wrong answers. The shocks were simulated, but participants believed the setup was real enough to produce visible distress.
In the best-known condition, 26 of 40 participants reached the final 450-volt switch. That number is memorable, but “most people blindly obey” is a poor summary. Milgram ran multiple variations, and compliance changed with the experimenter’s proximity, the learner’s proximity, the institutional setting, and the behavior of peers. Some participants refused; others argued while continuing.
Interpretation remains contested too. Archival research suggests that some participants understood themselves as supporting a worthwhile scientific project rather than merely surrendering thought. If so, the danger is not only submission to orders. A legitimate-seeming mission can make harmful actions appear necessary and helpful.
Jerry Burger’s later partial replication used stricter screening and stopped at 150 volts, the point at which the learner strongly objected. A majority were prepared to continue beyond that point, but the design cannot tell us how many would have gone to the end. Neither study provides a universal percentage for human obedience.
The defensible conclusion is narrower: ordinary people can be pulled surprisingly far when authority appears legitimate, commitment grows gradually, responsibility feels remote, and resistance requires breaking the social script.
Build an interruption before you need courage
Bad compliance thrives on speed and continuous contact. A pause turns the next action back into a decision. Hang up and call the institution using a number you found independently. Leave the room. Ask someone outside the chain. Refuse secrecy around a supposedly urgent procedure.
Workplaces can make resistance procedural rather than heroic. Useful phrases include: “I need written confirmation,” “Which policy authorizes this?”, “Who owns the risk if this fails?”, and “Let’s verify this with someone outside the approval chain.” They create a legitimate counter-script instead of requiring one person to win a confrontation.
The uncomfortable lesson is not that cooperation is a hidden character flaw. Cooperation is a social tool. It becomes dangerous when a situation directs attention toward the next instruction and away from authorship, evidence, and consequence. The earliest warning is often not a dramatic command. It is the moment when everybody is following the process and nobody can clearly explain who decided it was safe.