1. Situation · The Terrain
Milgram’s experiments are the field’s founding demonstration. Participants cast as “teachers” were instructed to deliver what they believed were escalating electric shocks — in 15-volt steps to a labeled 450 volts — to a “learner” (a confederate; no shocks were delivered). In the baseline, about 65% continued to the maximum despite scripted protests, a result that shocked Milgram and the professionals he surveyed, who predicted almost no one would fully comply (Milgram, 1963, 1974) [C4].
Two cautions must travel with this finding. First, the studies drew serious ethical criticism — extreme stress under deception, contested debriefing (Baumrind, 1964) [C4]. Second, and defensively more important, the “65%” headline flattens the data: Milgram ran more than twenty variations with obedience swinging enormously, and many who continued did so in visible anguish — compliance was contested and effortful, not automatic (Milgram, 1974) [C3]. Archival re-examination shows the picture was messier than the summaries: procedures varied, the standardized prods were often departed from, some participants suspected the shocks were unreal (Perry, 2013) [C3, contested]. Honest baseline: a substantial fraction of ordinary people can be brought to inflict serious harm under authority — but the rate is highly conditional, the process is resisted, and “blind obedience” overstates a more complicated result.
Obedience is engineered, not innate
The most useful legacy of the variations is the map of conditions that raise and lower obedience — each a lever a manipulator pulls and a defender recognizes. As the victim became more proximate, obedience fell sharply. Orders by telephone rather than in person, or from a rundown office rather than a prestigious university, cut it markedly — authority depends on presence and perceived legitimacy. The 15-volt increments are the mechanism, not a detail: each step is only slightly worse than the last, none large enough to trigger refusal — foot-in-the-door escalation of commitment (Ch 6; §2.12–§2.13) as the engine of obedience. And when participants saw other “teachers” refuse, obedience collapsed — the vertical mirror of Asch’s ally effect (Ch 15).
Table 16.1 — Situational levers of obedience: exploit and defense
| Situational lever | How it is exploited | Detection indicator | Defensive countermeasure |
|---|---|---|---|
| Victim distance | Consequences kept abstract or out of sight | You cannot see who is harmed by what you’re asked to do | Make the consequence concrete; picture the specific person harmed |
| Authority presence / legitimacy | Uniforms, titles, settings, insistent in-person pressure | Pressure eases when the authority is remote or its credentials are checked | Verify legitimacy for this specific act; disengage from physical pressure |
| Graduated commitment | Trivial harmless ask, then incremental escalation | Each request “only slightly worse” than the last; far from where you started | Compare each step to the original baseline; treat escalation itself as the warning |
| Disobedient peer (absent) | Isolation from anyone who might refuse | No visible model of refusal; you feel alone in doubting | Seek or be the refusing peer; find one ally who will say no |
2. Enemy Forces · Engaged Followership
Milgram explained his results with the agentic state — under legitimate authority a person feels like an instrument carrying out another’s will, responsible to the authority rather than for the act (Milgram, 1974) [C3]. This has not held up as a complete account, and the correction is central to the defense. If obedience were a passive surrender, the wording used to secure it shouldn’t matter — but it matters enormously. Re-analysis shows the four standardized prods were not equally effective: the more a prod resembled a direct order (“you have no other choice, you must go on”), the less likely participants complied; prods framing continuation as serving the scientific goal (“the experiment requires that you continue”) secured more compliance (Gibson, 2013; Burger, Girgis & Manning, 2011) [C3]. People did not obey best when ordered — they complied best when persuaded that continuing served a worthwhile collective enterprise they identified with.
This is the engaged-followership account (Haslam & Reicher): destructive obedience is not the switching-off of moral agency but its misdirection — participants actively work toward what they take to be a valuable goal, led by an authority they identify with, brought to see harm as regrettable necessity. Compliance drops when the request is stripped of legitimizing purpose, and identification with the enterprise predicts continuation (Reicher, Haslam & Smith, 2012; Haslam, Reicher & Birney, 2014) [C3]. An ethical partial replication stopping at 150 volts found obedience broadly comparable to Milgram’s (Burger, 2009) [C3]. The defensive payoff is large: because obedience runs on identification with a cause and a legitimate-seeming authority, the defense is to interrogate exactly those — is this authority legitimate, and for this? Is the “worthwhile purpose” real, or a wrapper for harm?
The Stanford Prison Experiment — hold at arm’s length
No obedience-adjacent study is more famous, or more misleading as usually taught. The standard telling — that ordinary people spontaneously became sadistic simply by donning a role — does not survive scrutiny (Haney, Banks & Zimbardo, 1973; Zimbardo, 2007) [C3]. Archival material indicates guards were briefed and encouraged toward toughness — cruelty was in part instructed (Le Texier, 2019) [C2–C3, contested]; there was no control group and the experimenter played an in-role part. A controlled replication, the BBC Prison Study, found participants did not mindlessly conform to a brutal role — tyranny required active work, consistent with engaged followership (Reicher & Haslam, 2006) [C3]. The defensible lesson is almost the opposite of the legend: cruelty is more readily produced when an authority actively encourages it, supplies a legitimizing purpose, and removes accountability — not that roles mechanically corrupt.
Table 16.2 — Obedience mechanisms as techniques: exploit, detection, defense
| Domain (Category) | Obedience mechanism | Detection | Defense |
|---|---|---|---|
| Impersonation fraud (Cat 2, Cat 14) | Staged legitimacy + graduated asks + isolation | Unverifiable “official” demanding escalating actions by phone; keeps you on the line | Hang up and verify the authority through an independent channel; refuse isolation |
| Institutional / workplace (Cat 2) | Legitimate authority + “the process requires it” | Asked to do wrong “because you’re told,” purpose invoked to override doubt | Invoke the right to refuse illegitimate orders; demand accountability; four-eyes check |
| Radicalization / cults (Cat 22, Cat 23) | Higher purpose + escalation + isolation from refusers | Escalating demands wrapped in a cause; cut off from outside dissent | Compare to baseline; test the “purpose” openly; keep outside contact |
| Coercive relationship (Cat 22) | Claimed legitimacy + incremental control + removed accountability | Commands obeyed to avoid conflict; each demand slightly larger; isolation | Name the escalation; rebuild outside support; seek professional/safety help |
The general exploit pattern
Across the authority families one structure recurs: establish a legitimate-seeming authority and a worthwhile-seeming purpose → secure a small, easy, harmless first compliance → escalate in increments each only slightly larger than the last, enlisting prior compliance to justify the next → keep the harm distant and accountability diffuse, and isolate the target from any peer who might model refusal → the target, now an engaged follower of a purpose they were led to accept, carries out acts they would have refused if asked directly. The big compliance is assembled from small ones under a legitimacy and purpose installed first — which is why the defense is uniform: test legitimacy, expose the purpose, measure against baseline, make harm concrete, restore a disobedient peer.