The Anatomy of Power: Why Do People Obey Authority?

People obey authority for several reasons: they may trust its expertise, accept its legitimacy, identify with its goals, fear penalties or rely on others when a situation is uncertain. Obedience is a relationship shaped by context. Research does not show that humans inevitably surrender judgment when someone gives an order. It shows how institutions can encourage compliance, and how people also resist.

A command rarely arrives as a naked demand. It arrives with a title, a uniform, a familiar procedure or a story about a shared purpose. Often following it is sensible. Air traffic controllers and emergency teams coordinate complex work through accepted authority. The difficult question is how the same capacity for cooperation can help people continue when an instruction begins to conflict with another person’s welfare.

What counts as obedience?

Obedience means responding to a directive from an authority. Conformity means adjusting behavior or judgment toward a group norm. Compliance is the broader act of agreeing to a request, which may come from an equal. A person can conform without receiving an order, and can obey while privately disagreeing. Separating these processes helps explain why a silent room and a forceful supervisor can exert different kinds of pressure. Asch: independence and conformity

The distinction also prevents moral shortcuts. Cooperation is not automatically submission, and disagreement is not automatically courage. We need to examine the instruction, the evidence supporting it, the opportunities to question it and the likely effects on other people. Authority becomes dangerous when its status is treated as a substitute for justification, especially when it removes meaningful ways to challenge mistakes.

What happened in Milgram’s experiment?

In Stanley Milgram’s 1963 published study, 40 men were placed in a supposed investigation of learning. Assigned the role of teacher, they were instructed to deliver increasing electric shocks for a learner’s errors. The learner was a confederate and the shocks were simulated. Participants confronted an apparently distressed person and an experimenter urging continuation. Twenty-six reached the maximum labeled level of 450 volts, the origin of the famous 65% figure. Milgram: Behavioral Study of Obedience

The statistic describes one condition and a particular recruited sample. It is not the percentage of humanity willing to kill, and the experiment did not measure homicide. Participants’ visible distress also matters: continuing did not necessarily mean indifference. The study exposed a conflict between the apparent demands of a legitimate scientific activity and the signs that another person wanted it to stop. A percentage alone cannot capture that conflict.

Why the setting mattered

Milgram conducted many variations. Changes in the proximity of the learner, the experimenter’s presence and the behavior of other participants changed continuation rates. These variations undermine the idea of a fixed obedience reflex. The authority relationship was created through a setting, a sequence of instructions and an account of why the task mattered. Different arrangements could weaken it. Haslam and Reicher: fifty years of obedience research

One influential interpretation emphasizes identification with the scientific project. People could regard themselves as helping a worthwhile enterprise and interpret the experimenter’s instructions within that commitment. This “engaged followership” account challenges the familiar image of a mindless person mechanically executing commands. It is a theoretical interpretation supported by research, not a final explanation of every participant’s motives. Reicher, Haslam and Smith: identification-based followership

Why the research remains contested

Milgram’s studies involved deception and substantial distress. Later scrutiny has questioned aspects of procedures, debriefing, participants’ beliefs about the shocks and the way findings entered textbooks. These questions affect how confidently the results can be generalized. They do not make the recorded behavior disappear, but they require us to ask what participants believed they were doing and how much confidence they placed in the laboratory’s assurances. Haslam and Reicher

The ethical difficulty is inseparable from the scientific one. If participants reasonably trusted that a university would prevent serious harm, continuation may not mean the same thing as knowingly inflicting severe injury. If they believed the danger was real, the experience raises profound concerns about their treatment. Either way, the study cannot be responsibly reduced to a reassuring explanation that ordinary people have no responsibility because authority is irresistible.

Would people still obey today?

Jerry Burger’s 2009 partial replication studied 70 adults with additional screening and safeguards. The procedure stopped around the 150-volt decision point rather than proceeding through the full original sequence. In the base condition, 70% were willing to continue beyond that point. This provides evidence that the situation could still generate substantial continuation. It does not establish that 70% would have reached 450 volts. Burger: Replicating Milgram

The distinction is easy to lose in a headline. A partial replication is deliberately partial, and its ethical protections change the research design. It can test a narrower behavioral comparison without re-creating every feature of the earlier work. Differences in recruitment, exclusion criteria, procedures and historical context also limit sweeping claims that human obedience has remained completely unchanged across generations. American Psychological Association: the replication

How does a group change what we say?

Solomon Asch’s line-judgment experiments placed a participant among people who gave an obviously incorrect answer on selected trials. Some participants publicly agreed with the majority even when their own perception pointed elsewhere. The design demonstrated that social agreement can influence expressed judgment. It did not show that everyone conforms, or that private perception always changes whenever a public answer does. Asch: the original study

Researchers distinguish normative influence, such as avoiding rejection, from informational influence, such as believing others know better. Real situations can combine them. In an unfamiliar meeting, people may stay quiet both because they fear embarrassment and because they assume colleagues possess relevant information. Bond and Smith’s meta-analysis of Asch-type studies also found variation across settings and cultures. A laboratory result should not become a ranking of the supposed character of entire populations. Bond and Smith: culture and conformity

Manipulation often changes the question

Manipulation can work by making a troubling request appear to be evidence of loyalty, competence or membership. The immediate question becomes whether someone is a good colleague rather than whether the action is justified. That framing helps explain why sincere commitment can support harmful conduct. People may defend a leader because the leader seems to represent a valued group, even when the instruction conflicts with principles the group claims to uphold. Reicher, Haslam and Smith

Consider a hypothetical workplace request to hide a failed safety check. The pressure may come through phrases about protecting the team, avoiding delays or trusting experienced managers. This example is an application of the research, not a measured outcome from Milgram’s laboratory. Useful questions restore what the framing obscures: what does the check show, who bears the risk, and who can independently review the decision?

Authoritarianism is larger than a personality label

Psychological authoritarianism concerns patterns such as submission to accepted authorities, aggression sanctioned by those authorities and adherence to established conventions. Political authoritarianism describes institutional concentrations of power. The concepts overlap but are not interchangeable diagnoses. A person’s response on a questionnaire does not, by itself, explain an entire political system or predict every decision that person will make. Authoritarianism and threat in 59 nations

Research links perceived threats and authoritarian attitudes, but causal direction and measurement remain important questions. Leaders can respond to fear, amplify it or define who counts as threatening. A longitudinal study examining threat and right-wing attitudes offers evidence beyond a single cross-sectional snapshot, while still leaving limits on generalization. Fear is neither a universal excuse nor an inevitable route to authoritarian politics. Three-wave longitudinal study of threat and attitudes

Why the Stanford prison story needs caution

The Stanford Prison Experiment is often paired with Milgram as proof that assigned roles automatically produce cruelty. Archival research by Thibault Le Texier challenged that narrative, drawing attention to instructions, researcher involvement and problems with the study’s presentation. Its dramatic scenes should not function as clean experimental evidence that any ordinary person becomes abusive after receiving a uniform. Le Texier: Debunking the Stanford Prison Experiment

The broader lesson concerns scientific authority itself. A famous experiment deserves the same attention to methods, records and alternative explanations as an unfamiliar one. Criticizing an influential study does not eliminate the importance of social situations. It improves the question: which features of a situation, operating through which relationships and beliefs, influence which people under which conditions?

How can people make resistance possible?

Resistance becomes more plausible when a person can recognize a conflict, find support and act through a usable route. Milgram’s variations and conformity research suggest that the behavior of others can alter the social situation. An ally can interrupt apparent unanimity. But no brief checklist guarantees safety, and a laboratory effect should not be advertised as a universal technique that will defeat coercion. Bond and Smith

A practical application is to make questioning a normal part of a process before a crisis occurs. Teams can require a written rationale for exceptional requests, allow independent review and specify who can pause a dangerous action. These are editorial recommendations based on the mechanisms discussed, rather than outcomes established by a single cited trial. Their value is that they reduce the burden on one isolated person to improvise a heroic refusal.

What should we remember about power?

Authority gains strength when people consider it competent, rightful or representative of something they value. That strength can coordinate care and cooperation, or help conceal harm. The evidence gives no warrant for treating obedience as destiny. It asks us to look closely at institutions that organize trust, explanations that turn commitment into duty, and opportunities for disagreement that remain real when the stakes rise.

The most useful question is not whether we are the kind of people who would always resist. Most of us cannot know that from an armchair. It is whether the systems we build make responsibility visible, preserve access to evidence and allow a doubtful instruction to be examined. Power deserves scrutiny precisely because cooperation is so valuable, and because people can care deeply about a cause while being mistaken about what it requires.

Further reading in this collection: The Lost Library of Alexandria: The Knowledge That Vanished · 20 Rare Natural Phenomena That Seem Impossible.