Stanley Milgram Experiment: Obedience, Authority, Ethics, and What It Really Showed
Author: Ukrainian Psychological Hub · Published: September 24, 2026 · Editorial Policy
The Stanley Milgram experiment was a series of laboratory studies of destructive obedience conducted at Yale University in the early 1960s. In the best-known condition, 26 of 40 men continued to the final 450-volt-labeled switch after being instructed by an experimenter to punish a “learner” for errors. The learner was an actor and received no real shocks, but the participant was led to believe the apparatus was genuine. The result showed that an authority structure can lead many ordinary people to continue an action they experience as disturbing. It did not establish that 65 percent of humanity will obey any authority, that participants became mindless, or that one psychological mechanism explains obedience in every setting (Milgram, 1963).
Milgram's studies became famous because they converted an abstract question about obedience into an observable conflict: a participant could hear signs of another person's suffering, could object, could hesitate, and still sometimes continue when a legitimate-seeming experimenter insisted that the procedure go on. Yet the classic story became simpler than the evidence. Milgram ran many conditions, not one. A later synthesis of 21 conditions involving 740 participants found an overall maximum-voltage rate of 43.6 percent and substantial variation with the legitimacy and directiveness of the experimenter, peer behavior, interpersonal distance, and other features of the situation (Haslam, Loughnan, & Perry, 2014).
Current scholarship therefore treats the Milgram program as both a landmark and a contested historical object. Its core behavioral finding remains important: under some structured authority conditions, substantial numbers of participants continued a harmful-seeming task despite conflict and distress. Its broader interpretation is still debated. Archival research, interaction analysis, modern replications, research on sense of agency, and renewed ethical analysis have complicated the old language of “blind obedience” and the idea that the experiment proved a single “agentic state.” For the broader psychological phenomenon rather than this experiment as a historical and methodological object, see Obedience to Authority: Milgram, Agency, Responsibility, and Social Influence. For a direct comparison between Milgram's authority paradigm and Asch-style group conformity, see Conformity vs Obedience: Asch, Milgram, Social Pressure, and Authority.
The Milgram Experiment at a Glance
• Research question: how far would a person continue administering apparently painful electric shocks when instructed to do so by an experimenter?
• Classic published sample: 40 men, ages 20 to 50, recruited from New Haven and surrounding communities for what they believed was a study of memory and learning (Milgram, 1963).
• Apparatus: a shock generator with 30 switches increasing in 15-volt steps from 15 to 450 volts. The participant received a sample shock, helping make the apparatus credible; the learner was a confederate and was never actually shocked.
• Classic result: 26 of 40 participants reached the maximum 450-volt switch, producing the famous 65 percent figure (Milgram, 1963).
• Critical qualification: 65 percent belongs to one condition. Across 21 Milgram conditions that could be synthesized on the same maximum-voltage outcome, 43.6 percent reached the maximum, and rates varied sharply with the social structure of the experiment (Haslam, Loughnan, & Perry, 2014).
• Evidence status: the behavioral observations are historical empirical findings. Milgram's agentic-state explanation is a historical theory. Engaged followership, interactional pressure, legitimacy, uncertainty, credibility, gradual commitment, and competing-role conflict are contemporary explanatory approaches with different levels of evidence; none functions as a universally accepted single explanation.
What Was the Stanley Milgram Experiment?
“The Milgram experiment” is convenient shorthand, but it can mislead. Milgram conducted a program of related obedience studies in which he systematically changed features of the experimenter, learner, setting, peer context, and procedure. The 1963 article reported the condition that became iconic; the larger research program was later assembled in his 1974 book Obedience to Authority: An Experimental View.
The central design placed the participant between two incompatible demands. One came from the experimenter, who represented scientific authority and instructed the participant to continue. The other came from the learner, whose errors were followed by progressively stronger apparent shocks and whose reactions communicated pain, objection, or refusal. The dependent measure was behavioral: how far along the shock scale the participant went before refusing.
This structure made the study psychologically powerful because “obedience” was not inferred from an attitude questionnaire. Participants had to act, hesitate, question, continue, or stop in a live social interaction. It also made the study ethically controversial because the experimental realism depended on deception and on participants believing that their decisions could be causing severe suffering.
Historical Context and Milgram's Question
Milgram framed destructive obedience as a problem of social organization: how can people perform actions they would not ordinarily choose when those actions are embedded in a hierarchy and presented as duties? His writing was shaped by the twentieth-century problem of organized atrocity and by public discussion of perpetrators who defended their conduct as obedience to orders. This historical framing explains why the experiment was never merely about laboratory compliance; Milgram wanted a model for studying how authority can reorganize responsibility and action (Milgram, 1963; Milgram, 1974).
The laboratory model should not be treated as a complete explanation of genocide, war crimes, or state violence. Those phenomena involve ideology, institutions, group identities, material incentives, coercion, dehumanization, career structures, sustained participation, and historical conditions that cannot be reproduced by a short experiment. Later scholarship has therefore challenged the popular move from “Milgram found obedience in a Yale laboratory” to “Milgram explained large-scale atrocity” (Haslam & Reicher, 2017).
How the Classic Experiment Worked
Recruitment and participants
Milgram's 1963 report described 40 male participants between 20 and 50 years old, drawn from New Haven and nearby communities through newspaper advertising and direct-mail solicitation. Their occupations and educational backgrounds varied. They were paid for coming to the laboratory and were told the research concerned memory and learning (Milgram, 1963). The sample was therefore not a random sample of all adults, and it contained no women in this published condition.
Teacher, learner, and experimenter
The real participant was assigned the role of “teacher.” The “learner” was a confederate of the research team. A rigged role-allocation procedure ensured that the real participant became the teacher. The learner was placed in another room and appeared to be connected to electrodes. The experimenter, wearing a laboratory coat in the Yale setting, directed the procedure.
The participant's task was to present word-pair questions. Each incorrect answer was supposed to be followed by a shock, with the intensity increasing one step at a time. The learner's responses were scripted. No electrical shocks were actually delivered to the learner.
The shock generator
The generator displayed 30 switches progressing from 15 to 450 volts in 15-volt increments, with verbal labels suggesting increasing danger. The apparatus transformed the social conflict into a visible sequence of small steps. The participant did not make one isolated decision to administer a maximum shock. Instead, the procedure asked for repeated continuation from one level to the next.
That incremental structure matters. Each new action was only one step beyond the previous action, while stopping required the participant to establish a boundary after having already accepted earlier steps. This makes gradual commitment a plausible part of the situation, although it should not be treated as a complete explanation by itself.
The learner's protests and the experimenter's prods
As the apparent shocks intensified, the learner's scripted behavior increasingly signaled pain and eventually refusal. When participants hesitated or objected, Milgram's published description represented the experimenter as using a sequence of standardized prompts, beginning with variants of “Please continue” and escalating toward stronger insistence (Milgram, 1974).
Later analysis of archival recordings complicates the idea of a perfectly standardized command sequence. Gibson's rhetorical analysis showed that the experimenter's interventions occurred within negotiation rather than as simple one-way commands, and that participants actively argued, questioned, and sought ways to define what they were obligated to do (Gibson, 2013). A 2026 analysis of 136 sessions likewise found that “Please continue” was unusually pervasive compared with other nonprocedural interventions, highlighting the importance of implicit interactional pressure rather than reducing the procedure to four mechanically delivered orders (Kaposi, Kacsuk, & Hegarty, 2026a).
What Were the Results?
The famous 65 percent
In the condition reported in 1963, 26 of 40 participants continued to the maximum 450-volt switch and 14 stopped before the end. That is the origin of the famous 65 percent figure (Milgram, 1963). The result was striking partly because observers had expected much lower levels of completion.
The behavioral outcome did not mean that the 26 participants calmly endorsed harming the learner. Milgram described marked nervous tension, including sweating, trembling, stuttering, nervous laughter, and other signs of conflict (Milgram, 1963). Continuing the procedure and approving of it are different psychological facts.
Why 65 percent is not a universal obedience rate
The most common statistical error in popular accounts is to turn 26 of 40 people in one condition into a general statement that “65 percent of people obey authority.” Milgram's own research program contradicts that interpretation because he varied the social situation and obtained different outcomes.
Haslam, Loughnan, and Perry synthesized 21 conditions with a common maximum-voltage outcome, covering 740 participants. Across those conditions, 43.6 percent reached the maximum. Eight coded features were associated with the probability of reaching 450 volts, including experimenter directiveness, legitimacy and consistency, peer pressure toward disobedience, the distance between teacher and experimenter, and properties of the teacher–learner relationship (Haslam, Loughnan, & Perry, 2014).
This meta-Milgram analysis is especially important because it changes the scientific question. The question is not “What percentage of humans obey?” It is “How do particular authority relations, social identities, interpersonal distances, peer models, and situational structures change the probability of continuing?” That is a more precise formulation and a more useful one.
What Changed Obedience Across Milgram's Variations?
Milgram's variations altered the relationship among the experimenter, teacher, learner, institution, and peers. The later empirical synthesis supports the conclusion that these manipulations mattered rather than functioning as decorative details (Haslam, Loughnan, & Perry, 2014).
Proximity to the learner
When the learner was psychologically or physically closer, the participant had less distance from the apparent consequences of the shocks. Across the historical conditions, properties of the teacher–learner relation were associated with maximum-voltage responding. Contemporary work using modified Milgram paradigms also finds that physical proximity can matter: a 2024 study reported greater obedience when the experimenter was physically present and lower obedience when the learner was physically present (Dolinski & Grzyb, 2024).
Proximity and legitimacy of the experimenter
Authority was not simply a voice issuing words. Its influence depended on how the experimenter was positioned in the social situation. The Meta-Milgram analysis found effects associated with experimenter legitimacy, directiveness, consistency, and distance (Haslam, Loughnan, & Perry, 2014). This is one reason authority should be distinguished from raw power or coercion: recognized standing and institutional meaning can change how an instruction is interpreted.
Peers and models of resistance
Milgram also varied what other people appeared to do. The synthesis found that group pressure toward disobedience was associated with lower maximum-voltage responding (Haslam, Loughnan, & Perry, 2014). Social influence therefore operated in more than one direction. The presence of authority could support continuation, while visible resistance by peers could make refusal more socially available.
Institutional prestige
Milgram treated the prestige of Yale as one possible source of legitimacy and conducted studies outside the university. Recent work has revisited the issue. A 2025 “obedience lite” study manipulated university versus non-university settings and scientific versus commercial framing and reported similar levels of obedience across its tested conditions, suggesting that the effect of prestige is not adequately summarized by the idea that an elite university automatically causes obedience (Grzyb et al., 2025b). This contemporary result does not erase historical setting effects; it shows that institutional prestige is one variable within a larger configuration.
What Did Milgram Think the Experiment Showed?
Milgram's most influential theoretical explanation was the agentic-state model. In Obedience to Authority, he distinguished an autonomous mode, in which a person experiences themselves as directing their own behavior, from an agentic mode, in which the person experiences themselves as an instrument carrying out another person's wishes. Responsibility can then be experienced as located upward in the hierarchy.
This was an attempt to explain a paradox visible in the sessions: a participant could be distressed, protest, and still continue. In Milgram's account, conflict remained because moral standards had not vanished, but the structure of authority altered how the participant understood agency and responsibility.
Agentic-state theory is historically important, yet the original experiments did not directly measure a discrete psychological state called the agentic state. The theory should therefore be presented as Milgram's explanatory model rather than as an empirical fact established by the 65 percent result.
What Does Modern Psychology Think Milgram Really Showed?
Modern scholarship supports a narrower and stronger conclusion than the popular “people blindly obey” story: structured authority can powerfully influence behavior, but the effect is contingent, interactive, and psychologically heterogeneous. Participants interpret legitimacy, scientific purpose, the learner's suffering, the experimenter's persistence, their own role, previous commitments, and the behavior of others. Continuation can emerge from several processes rather than a single switch from autonomy to obedience.
Engaged followership
Reicher, Haslam, and Smith proposed “engaged followership” as an alternative to passive surrender of agency. Their account argues that participants may continue when they identify with the experimenter and with a valued collective project—most notably the advancement of science—rather than merely because an authority commands them (Reicher, Haslam, & Smith, 2012). Archival work on participant reactions has been interpreted in support of this approach (Haslam et al., 2015).
Engaged followership is influential, but it is not a settled replacement for every other explanation. Hollander and Turowetz, drawing on archived interactions, argued that Milgram's behavior is better understood through multiple compliant processes and raised specific objections to treating identification-based followership as the dominant mechanism (Hollander & Turowetz, 2018). The disagreement itself is scientifically useful: it shifts interpretation from a morality tale about passive humans toward testable questions about what participants believed, valued, negotiated, and experienced.
Interaction, negotiation, and the pressure to continue
Audio-based and archival analyses show that hesitation did not simply trigger a clean, scripted command-and-response sequence. Participants asked questions, challenged the experimenter, proposed alternatives, sought reassurance, and tried to renegotiate obligations. Gibson's analysis therefore treats the sessions as rhetorical encounters in which the legitimacy of continuing was actively constructed and contested (Gibson, 2013).
Kaposi, Kacsuk, and Hegarty's 2026 analysis strengthens the interactional point. Among the experimenter's nonprocedural interventions, “Please continue” appeared with exceptional regularity across obedient sessions and obedient phases of sessions that later ended in disobedience (Kaposi, Kacsuk, & Hegarty, 2026a). This suggests that apparently mild language can function within a persistent interactional structure without being equivalent to an explicit threat.
Obedience was not procedural perfection
A further 2026 archival analysis asked a different question: even when participants kept administering shocks, did they actually follow the full set of experimental instructions? Kaposi and Sumeghy reported that no participant classified as entirely “fully obedient” complied perfectly with all procedures, and that procedural violations were common within sessions traditionally categorized as obedient (Kaposi & Sumeghy, 2026b).
This does not erase the central fact that participants continued administering apparent shocks. It changes what the label “fully obedient” should be allowed to mean. Reaching the endpoint and complying with every instruction are not identical behaviors.
Did Participants Really Believe the Shocks Were Real?
The learner was never actually shocked, so the experiment depended on participants believing the deception. The important historical question is therefore not whether the shocks were real—they were not—but how credible the setup was to different participants.
Archival reanalysis by Perry, Brannigan, Wanner, and Stam found that belief varied. In 18 of 23 variations they analyzed, participants who fully believed the learner was receiving painful shocks had lower mean shock levels than participants who did not fully believe it (Perry et al., 2020). That finding matters because behavior partly depended on the participant's interpretation of the experimental reality.
The archival evidence does not justify the opposite myth that “everyone knew it was fake.” Some participants clearly treated the learner's situation as serious; others were skeptical; belief could change during a session. Credibility is therefore a methodological variable, not a binary verdict that either saves or destroys the entire research program.
The 65 Percent Result and the Problem of Classification
Milgram's outcome measure was operationally clear: the highest switch reached before refusal. That clarity helped make the experiment memorable. It also compressed a complicated interaction into one endpoint. Two participants could both reach 450 volts while differing sharply in hesitation, protest, attempts to help the learner, belief in the apparatus, identification with the research, and compliance with procedural instructions.
Contemporary archival work makes this compression visible. The endpoint remains a real behavioral datum, but it should not be allowed to stand for an entire psychological state. “Reached 450 volts” is not synonymous with “endorsed the experiment,” “lost moral awareness,” “believed every element of the deception,” or “obeyed every instruction.”
Ethics: Why the Milgram Experiment Became a Landmark Controversy
The ethics debate began almost immediately. Diana Baumrind argued in 1964 that researchers owe participants respect, welfare protection, and attention to the possible effects of deception, humiliation, and loss of trust (Baumrind, 1964). Milgram responded that the research had major scientific value, that participants were not physically endangered, and that follow-up responses indicated substantial retrospective approval of participation (Milgram, 1964). The ethical dispute is therefore part of the experiment's original history rather than a criticism invented decades later.
Deception and informed consent
Participants consented to a study presented as research on memory and learning, not to the true purpose of studying obedience under apparent harmful pressure. The learner's role, the shock delivery, and the real research question were concealed. Deception was essential to Milgram's design because advance disclosure would have changed the behavior being studied.
Under the current American Psychological Association Ethics Code, research deception requires significant prospective scientific, educational, or applied value and the absence of feasible effective nondeceptive alternatives. The code also states that psychologists do not deceive prospective participants about research reasonably expected to cause physical pain or severe emotional distress, and it requires timely explanation of deception and debriefing. These modern standards make a direct recreation of the classic procedure ethically difficult.
Distress and the right to withdraw
Milgram's own 1963 report is the strongest primary evidence that the procedure could generate acute distress: he described sweating, trembling, stuttering, nervous laughter, and, in some cases, severe tension (Milgram, 1963). Ethical evaluation therefore cannot be reduced to the fact that the learner was safe. The participant's psychological experience was itself part of the risk.
A second concern is whether participants experienced withdrawal as genuinely available when the experimenter continued pressing them to proceed. Modern informed-consent standards emphasize the participant's right to decline or withdraw. In U.S. federally regulated human-subjects research, the HHS Common Rule framework also places informed consent and institutional oversight within a formal protection system. Specific legal requirements depend on the research and institution, so the Common Rule should not be treated as a universal rule for every psychological study everywhere.
Debriefing and retrospective approval
Milgram emphasized debriefing and later participant reactions in defending the research. Archival analyses have shown that participant responses were more complex than a simple choice between “traumatized” and “glad to participate.” Haslam and colleagues found substantial retrospective expressions of satisfaction and interpreted them through identification with the scientific project (Haslam et al., 2015). Kaposi's later review of the second wave of Milgram scholarship emphasizes that ethics, affect, credibility, reporting, and interaction all require separate analysis rather than one global verdict (Kaposi, 2022).
Would the Original Milgram Experiment Be Approved Today?
A verbatim recreation of the classic procedure would face serious ethical barriers because its scientific value would have to be balanced against deception, acute distress, ambiguity about withdrawal, and the availability of less harmful designs. Current APA standards on deception, informed consent, debriefing, and harm are substantially more explicit than the conditions under which Milgram worked (American Psychological Association, 2017).
This does not mean obedience can no longer be studied. Researchers have developed modified paradigms that stop earlier, reduce the apparent severity of harm, screen participants more carefully, strengthen withdrawal protections, use simulations or virtual environments, or study coercion and agency with different tasks.
Modern Replications and Extensions
Burger's 2009 partial replication
Jerry Burger designed a well-known partial replication that stopped the procedure at 150 volts and introduced additional participant protections. The logic was that behavior at this decision point could provide information about whether participants would continue without exposing them to the full classic escalation. Burger found substantial continuation under the modified procedure and argued that the basic social influence phenomenon remained observable decades later (Burger, 2009).
Burger's study is not a literal replication of the 450-volt experiment. That distinction matters. It supports continuity in the phenomenon under a safer design; it does not recreate the full ethical and psychological conditions of Milgram's original sessions.
Cross-cultural evidence
Milgram-like procedures have been conducted in multiple countries and historical periods. Blass's cross-cultural review concluded that obedience within the paradigm was not unique to the United States, while also documenting variation across studies (Blass, 2012). Cross-cultural findings should not be converted into a universal human percentage because sampling, procedures, historical context, institutional credibility, and operational definitions differ.
Recent obedience-lite research
Recent researchers continue to modify the paradigm rather than simply repeating the 1960s design. In a 2025 study, Grzyb and Dolinski compared a condition in which participants had first occupied a genuine teaching role with one in which they functioned mainly as examiners. Refusals were more frequent in the teacher condition, although the difference in the primary stopping outcome was not statistically significant; experimenter prompting did differ significantly (Grzyb & Dolinski, 2025a).
Another 2025 study varied institutional setting and the stated scientific versus commercial purpose while using a modernized obedience-lite apparatus. The tested manipulations did not produce the simple prestige effect one might expect from textbook summaries (Grzyb et al., 2025b). Together with proximity research, these studies reinforce a general point: the Milgram paradigm remains useful when researchers treat its social structure as a set of variables rather than as a timeless script.
Does Modern Research Support the Agentic State?
Modern sense-of-agency research provides limited support for one component of Milgram's theory: coercion can change how people experience the relation between their own actions and outcomes. Caspar and colleagues found weaker intentional binding—a commonly used implicit measure related to sense of agency—when participants acted under coercion than when they chose freely, together with differences in neural processing of action outcomes (Caspar et al., 2016).
Akyüz and colleagues later found that forced choices increased perceived action–outcome intervals relative to free choices, a pattern interpreted as reduced agency under obedience (Akyüz et al., 2024). This line of evidence is important because it tests agency more directly than the original Milgram studies.
It still does not prove Milgram's full agentic-state theory. An implicit reduction in experienced agency is not the same as entering one discrete psychological state, surrendering all choice, transferring all responsibility, or losing moral understanding. Modern evidence therefore supports a narrower proposition: coercive or commanding contexts can measurably alter the experience of authorship over action. The wider questions of responsibility and obedience remain more complex. The dedicated Obedience to Authority article develops this distinction in greater depth.
What About “Blind Obedience”?
“Blind obedience” is a memorable phrase and a poor description of much of the archival record. Many participants were visibly conflicted. They questioned the experimenter, worried about the learner, attempted to renegotiate what they were required to do, hesitated, violated parts of the procedure, or eventually stopped.
The fact that many still continued is precisely what requires explanation. Their conflict is part of the phenomenon, not evidence against it. The experiments are psychologically interesting because people could remain morally and emotionally engaged while behaviorally continuing. A theory that assumes participants stopped thinking misses that tension.
Kaposi's review describes a “second wave” of critical engagement in which researchers returned to archives, recordings, ethics, experimental interactions, and historical documentation rather than treating Milgram's published narrative as the final word (Kaposi, 2022). The result is a more difficult but more informative picture: obedience was produced through interpretation and interaction as well as formal hierarchy.
Milgram, Authority, Power, and Legitimacy
The experiment is often described as a study of “power,” but its canonical psychological object was obedience to an authority figure. The experimenter had influence because the role carried perceived standing within a scientific institution, because the participant had accepted participation in a structured procedure, and because the situation supplied reasons to treat the experimenter's instructions as legitimate or knowledgeable.
This is why power and authority should not be used as synonyms. Power is the broader capacity to affect outcomes; authority concerns a recognized or accepted right to direct conduct. Milgram's laboratory combined authority, informational dependence, institutional legitimacy, social pressure, and control over the immediate procedure. Separating these components prevents the experiment from becoming a vague demonstration that “power changes people.”
For the wider scientific construct of authority—including legitimacy, expertise, trust, and obedience—see Authority in Psychology: Legitimacy, Expertise, Trust, and Obedience. For the broader cluster connecting social power, status, influence, control, relationships, and institutions, see Psychology of Power.
What the Milgram Experiment Did Not Prove
The experiment did not prove that a fixed percentage of people will obey immoral orders. The 65 percent result belongs to one historical condition, and Milgram's own variations produced different outcomes.
It did not prove that personality is irrelevant. The design powerfully demonstrated situational influence, but showing that situations matter is different from showing that individual differences never matter.
It did not prove that authority erases agency. Participants argued, hesitated, resisted, continued for different reasons, and sometimes violated instructions while still advancing through the shock sequence. Modern coercion research finds reductions in implicit sense of agency, not disappearance of agency.
It did not prove that participants believed every element of the deception. Archival evidence shows meaningful variation in credibility, and that variation was related to behavior (Perry et al., 2020).
It did not establish engaged followership as the one correct replacement theory. Identification with scientific goals is supported by important evidence, while interactional research and theoretical debate indicate multiple compliant processes (Reicher, Haslam, & Smith, 2012; Hollander & Turowetz, 2018).
It did not reproduce the psychological, ideological, institutional, or temporal structure of genocide. The paradigm can illuminate destructive obedience while remaining an incomplete model of mass violence.
What the Experiment Still Supports
A durable conclusion survives the criticism: people can continue actions they experience as deeply troubling when those actions are embedded in a credible authority relationship and a structured sequence of demands. The probability of continuation changes when the structure changes. That finding makes social context causally important rather than merely decorative.
A second durable conclusion is that resistance is part of the same phenomenon. Fourteen of 40 participants in the classic condition stopped before the final switch, and disobedience became more likely in some variations. Studying obedience therefore also means studying how people identify a boundary, reinterpret legitimacy, use social support, and transform hesitation into refusal.
A third conclusion concerns method. Milgram's work demonstrates both the scientific power and the danger of immersive behavioral experiments. The paradigm generated behavior that questionnaires might never have captured, yet the same realism produced deception, distress, and interpretive ambiguity. The experiment's methodological legacy is inseparable from its ethical legacy.
Why the Milgram Experiment Still Matters
Milgram remains important because modern life is saturated with delegated action. People work inside hospitals, corporations, laboratories, militaries, schools, governments, platforms, and technical systems in which responsibility is divided across roles. Instructions often arrive through procedures rather than overt threats. The psychological question—how people interpret what they are authorized, expected, or required to do—remains current.
The strongest modern lesson is therefore not a cynical slogan about human nature. It is a structural lesson about behavior. Authority is more influential when it is credible, coherent, proximate, institutionally meaningful, and supported by the situation; resistance becomes more available when legitimacy fractures, peers dissent, the consequences become harder to distance, or participants redefine what their role requires. These are probabilistic relations, not guarantees.
The experiment also remains a warning about labels. “Obedient,” “disobedient,” “agentic,” and “resistant” are analytical categories. They can clarify behavior when operationalized carefully, and they can hide complexity when treated as complete descriptions of a person.
Common Myths About the Milgram Experiment
Myth 1: The learner received real 450-volt shocks
The learner was a confederate and received no actual shocks. The participant was deceived about this. The ethical issue arose from what participants believed they might be doing and from the distress produced by that belief.
Myth 2: Milgram proved that 65 percent of people always obey authority
The 65 percent figure is 26 of 40 participants in one published condition. Across 21 comparable Milgram conditions, the synthesized maximum-voltage rate was 43.6 percent, with systematic variation across conditions (Haslam, Loughnan, & Perry, 2014).
Myth 3: Participants simply stopped thinking
The record contains anxiety, argument, negotiation, procedural violations, attempts to clarify responsibility, and eventual refusal as well as continuation. “Blind obedience” compresses these processes into a caricature.
Myth 4: Agentic state is a proven fact
Agentic state is Milgram's historical theory. Modern coercion studies provide evidence that commands can reduce implicit measures of sense of agency, but they do not establish the complete original theory (Caspar et al., 2016; Akyüz et al., 2024).
Myth 5: Archival criticism made the experiment meaningless
Archival criticism changes interpretation rather than deleting the historical behavior. Credibility varied, procedures were less uniform than textbook versions imply, and “full obedience” hid procedural noncompliance. The scientifically appropriate response is a more precise model of what the experiment measured (Perry et al., 2020; Kaposi & Sumeghy, 2026b).
FAQ
What was the main purpose of the Milgram experiment?
Its main purpose was to study destructive obedience under laboratory conditions: how far a participant would continue an apparently harmful action when instructed by an experimenter who occupied a position of authority.
What percentage of participants obeyed Milgram?
In the famous 1963 condition, 26 of 40 participants—65 percent—reached the 450-volt endpoint. That number should not be generalized to all people or all Milgram conditions. A synthesis of 21 conditions found 43.6 percent reached the maximum overall (Haslam, Loughnan, & Perry, 2014).
Did anyone actually get electrocuted?
No. The learner was an actor and did not receive the shocks shown on the generator. Participants were deceived into believing the setup was real enough for their behavior to be meaningful.
Why did participants continue?
There is no single established cause. Evidence and theory point to a combination of perceived legitimacy, commitment to the scientific setting, uncertainty, gradual escalation, interpersonal pressure, identification with the research project, distance from the learner, proximity to the experimenter, peer behavior, and individual interpretation. Different participants may have continued for different reasons.
What is Milgram's agentic state?
It is Milgram's theory that under legitimate authority a person can experience themselves as an agent executing another person's wishes and shift experienced responsibility upward. It is historically influential but not established as a single discrete state by the original experiment.
Was the Milgram experiment ethical?
It was ethically controversial when published and remains so. The central issues are deception, acute distress, informed consent, perceived freedom to withdraw, debriefing, and the balance between scientific value and participant welfare. Baumrind criticized the study in 1964, and Milgram defended it the same year (Baumrind, 1964; Milgram, 1964). A verbatim modern replication would face major obstacles under current research-ethics standards.
Can the Milgram experiment be replicated today?
Researchers can and do study related processes, but modern studies generally modify the procedure. Burger's 2009 partial replication stopped at 150 volts and added protections, while later work has used obedience-lite procedures, alternative forms of harm, or different measures of coercion and agency (Burger, 2009).
Did participants know the shocks were fake?
Belief varied. Archival analysis indicates that some participants were skeptical and that credibility was associated with shock levels. The evidence supports neither “everyone believed it” nor “everyone knew it was fake” (Perry et al., 2020).
What did the Milgram experiment really show?
It showed that a structured, credible authority situation can lead substantial numbers of ordinary participants to continue a harmful-seeming action despite conflict, and that the probability of continuing changes when the social situation changes. It also showed that continuation can coexist with distress, negotiation, partial resistance, and moral conflict. Modern evidence does not support turning the study into a universal obedience percentage or a one-mechanism theory of human behavior.
Related Articles
References
Akyüz, N., Marien, H., Stok, F. M., Driessen, J. M. A., de Wit, J., & Aarts, H. (2024). Obedience induces agentic shifts by increasing the perceived time between own action and results. Scientific Reports, 14, 16769. https://doi.org/10.1038/s41598-024-66499-8
American Psychological Association. (2017). Ethical principles of psychologists and code of conduct (2002, amended effective 2010 and 2017). https://www.apa.org/ethics/code
Baumrind, D. (1964). Some thoughts on ethics of research: After reading Milgram's “Behavioral Study of Obedience.” American Psychologist, 19(6), 421–423. https://doi.org/10.1037/h0040128
Blass, T. (2012). A cross-cultural comparison of studies of obedience using the Milgram paradigm: A review. Social and Personality Psychology Compass, 6(2), 196–205. https://doi.org/10.1111/j.1751-9004.2011.00417.x
Burger, J. M. (2009). Replicating Milgram: Would people still obey today? American Psychologist, 64(1), 1–11. https://doi.org/10.1037/a0010932
Caspar, E. A., Christensen, J. F., Cleeremans, A., & Haggard, P. (2016). Coercion changes the sense of agency in the human brain. Current Biology, 26(5), 585–592. https://doi.org/10.1016/j.cub.2015.12.067
Dolinski, D., & Grzyb, T. (2024). Obedience to authority as a function of the physical proximity of the student, teacher, and experimenter. The Journal of Social Psychology. Advance online publication. https://doi.org/10.1080/00224545.2024.2348479
Gibson, S. (2013). Milgram's obedience experiments: A rhetorical analysis. British Journal of Social Psychology, 52(2), 290–309. https://doi.org/10.1111/j.2044-8309.2011.02070.x
Grzyb, T., & Dolinski, D. (2025a). Becoming a teacher can reduce obedience compared to being solely an examiner: Agentic state and obedience in the Milgram paradigm. Frontiers in Psychology, 16, 1613507. https://doi.org/10.3389/fpsyg.2025.1613507
Grzyb, T., Dolinski, D., Sudoł-Malisz, M., Kulik, G., et al. (2025b). Doubting the power of prestige: Obedience to authority beyond institutional and research justifications. Scientific Reports, 15, 24911. https://doi.org/10.1038/s41598-025-10331-4
Haslam, N., Loughnan, S., & Perry, G. (2014). Meta-Milgram: An empirical synthesis of the obedience experiments. PLOS ONE, 9(4), e93927. https://doi.org/10.1371/journal.pone.0093927
Haslam, S. A., Reicher, S. D., Millard, K., & McDonald, R. (2015). “Happy to have been of service”: The Yale archive as a window into the engaged followership of participants in Milgram's “obedience” experiments. British Journal of Social Psychology, 54(1), 55–83. https://doi.org/10.1111/bjso.12074
Haslam, S. A., & Reicher, S. D. (2017). 50 years of “Obedience to Authority”: From blind conformity to engaged followership. Annual Review of Law and Social Science, 13, 59–78. https://doi.org/10.1146/annurev-lawsocsci-110316-113710
Hollander, M. M., & Turowetz, J. (2018). Multiple compliant processes: A reply to Haslam and Reicher on the engaged followership explanation of “obedience” in Milgram's experiments. British Journal of Social Psychology, 57(2), 301–309. https://doi.org/10.1111/bjso.12252
Kaposi, D. (2022). The second wave of critical engagement with Stanley Milgram's “obedience to authority” experiments: What did we learn? Social and Personality Psychology Compass, 16(6), e12667. https://doi.org/10.1111/spc3.12667
Kaposi, D., Kacsuk, Z., & Hegarty, P. (2026a). “Please Continue”: Implicit communication and the experimenter's interventions in Stanley Milgram's Obedience to Authority series. European Journal of Social Psychology, 56(1), 292–306. https://doi.org/10.1002/ejsp.70033
Kaposi, D., & Sumeghy, D. (2026b). From legitimate to illegitimate violence: Violations of the experimenter's instructions in Stanley Milgram's “obedience to authority” studies. Political Psychology. https://doi.org/10.1111/pops.70112
Milgram, S. (1963). Behavioral study of obedience. Journal of Abnormal and Social Psychology, 67(4), 371–378. https://doi.org/10.1037/h0040525
Milgram, S. (1964). Issues in the study of obedience: A reply to Baumrind. American Psychologist, 19(11), 848–852. https://doi.org/10.1037/h0044954
Milgram, S. (1974). Obedience to Authority: An Experimental View. Harper & Row. Google Books record
Perry, G., Brannigan, A., Wanner, R. A., & Stam, H. (2020). Credibility and incredulity in Milgram's obedience experiments: A reanalysis of an unpublished test. Social Psychology Quarterly, 83(1), 88–106. https://doi.org/10.1177/0190272519861952
Reicher, S. D., Haslam, S. A., & Smith, J. R. (2012). Working toward the experimenter: Reconceptualizing obedience within the Milgram paradigm as identification-based followership. Perspectives on Psychological Science, 7(4), 315–324. https://doi.org/10.1177/1745691612448482
U.S. Department of Health and Human Services, Office for Human Research Protections. (2025). 45 CFR 46: Protection of Human Subjects. https://www.hhs.gov/ohrp/regulations-and-policy/regulations/45-cfr-46/index.html