Obedience in Social Psychology

Obedience is a form of social influence where an individual acts in response to a direct order from another individual, who is usually an authority figure. It is assumed that without such an order, the person would not have acted in this way.

Obedience occurs when you are told to do something (authority), whereas conformity happens through social pressure (the norms of the majority).

Obedience involves a hierarchy of power/status. Therefore, the person giving the order has a higher status than the person receiving the order.

Milgram (1992) separated the two on three grounds:

• Explicit order: Obedience follows a direct order, whereas conformity follows an implicit norm.
• Status hierarchy: Obedience runs from a higher-status authority to someone of lower standing, whereas conformity operates between equals.
• Responsibility: Obedience is typically excused by denying responsibility (“I was only doing what I was told”), not by denying that influence occurred.

Brown (1986) summarised the contrast neatly. Peers shape conformity by example, whereas a superior shapes obedience by direction.

What is the difference between compliance and obedience?

Compliance refers to adjusting one’s behavior due to a request or suggestion from others, often without internal agreement. Obedience involves following direct orders or commands from an authority figure.

The key difference lies in the source: compliance stems from peers or general social cues, while obedience arises from perceived authority.

Explanations for Obedience

Psychologists have proposed several explanations for why ordinary people obey destructive orders. The main accounts are the agentic state, the perception of legitimate authority, and situational factors, alongside a modern challenge to the first of these.

The Agentic State and Agentic Shift

Milgram argued obedient participants enter an agentic state: they see themselves as an instrument carrying out someone else’s wishes, not as a free moral agent. This is the opposite of the autonomous state, in which a person acts on their own principles and feels responsible for the outcome.

“The essence of obedience consists in the fact that a person comes to view himself as the instrument for carrying out another person’s wishes, and he, therefore, no longer regards himself as responsible for his actions” (Milgram, 1974).

Milgram believed we are taught to enter the agentic state from childhood. We are socialised from an early age to obey authority figures, until doing so feels automatic.

Two conditions make this shift stick, Milgram argued. The person must see the authority as qualified to give the order. They must also believe the authority, not themselves, will answer for what follows (Milgram, 1974).

The clearest support comes from the two-teacher variation. Obedience rose to 92.5% once a confederate threw the switch and the real participant only read out the word pairs (Milgram, 1974).

Some participants who had refused also resumed once the experimenter said, “I’m responsible for what happens” (Milgram, 1974).

Three secondary mechanisms then keep a distressed person in that state.

• Situational obligation: Participants had agreed to take part.
• Awkward withdrawal: Stopping mid-procedure felt socially awkward.
• Task absorption: Absorption in the narrow technical task of reading word pairs and throwing switches crowded out reflection on the victim’s suffering.

Milgram (1974) saw the agentic state as the essential feature linking laboratory obedience to obedience in Nazi Germany, despite differences in scale and political context.

The Perception of Legitimate Authority

A related question is whether obedience depends on the authority looking genuinely legitimate. Milgram tested this directly.

Authority figures also rely on visible symbols of power. Milgram’s experimenter always wore a grey lab coat, which made refusal harder for participants.

Many participants still showed real distress and hesitation, so diffusion of responsibility cannot be the whole story. They were caught between two competing demands: the external authority of the experimenter, who says “Shock”, and the internal authority of conscience, which says “Don’t shock”.

Obedience ended where conscience won. In the original experiment, 35% reached that point before 450V. For many, the crucial prod was “You have no other choice, you must go on”, which reminded them they did have a choice.

Kelman and Hamilton (1989) make a related point: people comply with an authority only for as long as they regard it as legitimate.

A synthesis of 21 Milgram conditions supports this.

Across 740 participants, Haslam, Loughnan and Perry (2014) found that the experimenter’s legitimacy was one of eight factors predicting whether participants went on to 450V. The others included the experimenter’s directiveness and consistency, group pressure to disobey, and how close the participant was to the learner and to the experimenter.

Overall, 43.6% of participants continued to the maximum.

Situational Factors

Beyond legitimacy, three further mechanisms help sustain obedience once it has started.

• Diffusion of Responsibility: When the experimenter said “I’m responsible for what goes on here,” participants visibly relaxed; telling them they themselves were responsible sharply reduced obedience (Hamilton, 1978).
• Entrapment: Gilbert (1981) argued participants were drawn in by small, 15-volt increments, so no single step felt like a big escalation from the last one.
• Socialisation: Brown (1986) notes that obedience is instilled from childhood by parents and teachers, becoming an automatic habit that is hard to suspend even when it feels wrong.

Milgram (1974) also noted further situational cues that deepened participants’ sense of obligation. They believed the study would advance scientific knowledge. They had been paid to take part, and had volunteered.

Hamilton (1978) tested diffusion of responsibility outside Milgram’s paradigm. In a jury simulation, mock jurors blamed an actor less for a harmful act done under direct order than for the same act done voluntarily. Perceived responsibility, not obedience alone, shaped moral judgement.

Gilbert (1981) argued that the graduated series of shocks, not personality, explains why so many ordinary participants went so far.

Entrapment works like the foot-in-the-door technique. Freedman and Fraser (1966) showed that agreeing to a small request raises the chance of agreeing to a much larger one.

The Dispositional Alternative: Authoritarian Personality

Adorno, Frenkel-Brunswik, Levinson and Sanford (1950) offered a rival, dispositional account, which explains obedience through stable personality differences.

They proposed an authoritarian personality: a personality structure, rooted in a harsh, punitive upbringing, that makes some people more likely to submit to authority. Their F scale measures it.

The two accounts make different predictions. A situational account expects high obedience in an unselected sample whatever the personality. A dispositional account expects obedience to track a measurable, pre-existing trait.

This link appeared inside a largely obedient sample, since 65% went to the maximum shock in the baseline study. The accounts therefore work at different levels.

Situation explains why obedience was so high overall.

Elms and Milgram’s finding suggests personality played a smaller, secondary role in who resisted earliest. Milgram’s own point still stands: ordinary people, not a disturbed or unusually obedient minority, produced the 450V outcome.

Engaged Followership

Haslam and Reicher pushed back against the agentic-state account. From the 2010s, they argued obedient participants were active followers who identified with the research, not passive instruments carrying out orders.

A related experiment by Haslam, Reicher and Birney (2014) found the same pattern: participants continued when appeals were made to the value of science, but not when given direct orders.

This changes what needs explaining. The agentic-state account asks how authority disables moral agency. Engaged followership asks how authority recruits it for a cause the follower accepts as legitimate and worthwhile.

Is Engaged Followership Supported?

The evidence is contested.

Hollander and Turowetz (2017) examined 91 archived recordings of interviews held immediately after the experiments. Obedient participants often justified continuing by treating the situation as benign. They did not tend to explain themselves by identifying with science.

Haslam and Reicher (2018) disagreed. They said the interviews never invited talk about science, so they could not test the account fairly. They also argued that trust in the experimenter can itself express a shared identity.

Real-World Applications of Obedience Research

Obedience research explains more than a laboratory procedure. The same mechanisms appear wherever a person acts on a superior’s instruction and assumes the superior carries the responsibility.

Atrocity and Military Contexts

Milgram (1974) saw diffusion of responsibility and the agentic state as central to Nazi atrocities. Eichmann’s defence, that he was “just carrying out orders”, fits the pattern.

Milgram applied the same reasoning to William Calley. Calley was court-martialled for the 1968 My Lai massacre of several hundred Vietnamese civilians.

Hirsch (1995) frames this at the level of the whole system, not the individual. Authorisation supplies the agentic shift. The order seems to come from a legitimate source. Routinisation removes the moment-to-moment deliberation that might trigger conscience.

Dehumanisation neutralises the residual pull of empathy. This system-level view asks what a hierarchy does to its members, not what kind of person joins it.

How far this analysis explains the Holocaust, or oversimplifies it, remains debated.

Healthcare Hierarchies and Patient Safety

Kelman and Hamilton (1989) extend the same “crimes of obedience” framework to professional hierarchies. Harm is committed at the request of a nominally legitimate superior. No threat is needed.

Hofling et al.’s (1966) hospital study shows how this works in medicine. The great majority of nurses followed a telephone instruction from an unfamiliar “doctor” to give a drug dose well above the labelled maximum. A separate sample of nurses had said they would refuse.

Deference built into a hierarchy can override a person’s own safety judgement.

The same structure recurs in modern accounts of medication and procedural errors in hospitals. A professional accepts an order from someone whose authority is assumed, not verified. The professional also assumes the prescriber carries clinical responsibility.

Workplaces and Coercive Groups

Kelman and Hamilton (1989) argue that crimes of obedience are not confined to war and medicine. They recur wherever someone in a bureaucracy is asked to carry out, or stay silent about, actions they privately doubt.

The superior has given the instruction. The superior also seems ready to take responsibility.

Three laboratory ingredients explain why employees keep going past the point of private objection. The authority seems legitimate. Responsibility diffuses onto it. Small compliances escalate gradually (Gilbert, 1981).

Each extra compromise looks only marginally worse than the last.

The same entrapment, plus lifelong socialisation into deference (Brown, 1986), is invoked to explain why people in cults or coercively controlling relationships struggle to leave.

Each demand is framed as only slightly beyond the last. The group’s authority is treated as legitimate, so it assumes responsibility for outcomes the person would resist if asked to accept them all at once.

Preventing Destructive Obedience

Engaged followership changes how organisations should try to prevent destructive compliance. Haslam and Reicher (2012) argue that harm-doing in Milgram’s paradigm tracked identification with a valued cause, not bare obedience to command. Destructive acts were presented as constructive.

Training that simply tells staff to resist “blind obedience” therefore targets the wrong mechanism. Engaged wrongdoers typically act knowingly and with conviction, not passively.

A better approach trains people to notice how a harmful instruction is framed as necessary, legitimate or for a greater good. The framing, not only the authority, deserves scrutiny.

Kelman and Hamilton (1989) reach the same practical point from organisational ethics. Responsibility for a harmful policy diffuses onto a hierarchy that seems to have accepted it. Whistleblowing structures should therefore help individuals place responsibility back on themselves and the organisation jointly.

One example is a named sign-off. An identifiable decision-maker must approve ethically contested instructions.

Critical Evaluation

The situational, legitimacy-based account of obedience has substantial support, though the agentic state itself has been challenged directly.

• Situational Variables: Burger (2011) links obedience to four features of the situation: small incremental demands, diffused responsibility, a novel situation with no script, and limited time to think, each with empirical support.
• Legitimacy Over Trait: The legitimate-authority account explains the graded drop in obedience when authority is made remote or less legitimate far better than a fixed personality trait could.
• Cross-Cultural Robustness: Smith and Bond’s (1998) review found broadly similar obedience rates across several countries, suggesting the effect is not a quirk of 1960s America.

• Agentic State Unproven: Reicher and Haslam (2011) argue the studies “provide no real evidence of such an agentic state,” and note it cannot explain differences between conditions.
• Little Current Support: Burger (2011) notes that “few serious researchers talk about an agentic state these days.”
• Ecological Validity: Lab findings may not generalise to real life. A scripted confederate and a one-off encounter may trigger different processes from the sustained obedience found in real institutions.
• Situational vs Dispositional: These situational explanations oppose trait accounts like the authoritarian personality. Yet obedient participants scored higher on its F scale (Elms & Milgram, 1966), so the debate remains live.
• No Single Explanation: The agentic state, legitimacy, diffusion of responsibility, entrapment, socialisation and engaged followership work best together, as complementary mechanisms rather than as rival, mutually exclusive theories.

Contemporary Research

Ethical rules rule out an exact replication today. Modern partial replications instead ask whether destructive obedience has declined.

Burger (2009) stopped the procedure at 150V, adding safeguards such as a screening interview, and found obedience up to that point remained close to Milgram’s original figure.

Both modern replications stopped at 150V, so they show a willingness to start harming someone on command, not obedience all the way to 450V. They match Blass’s (1999) earlier meta-analysis, which found no decline between 1963 and 1985.

Together, these results count against the hope that public awareness of Milgram’s findings would by itself protect people from destructive obedience.