Milgram on disobedience

Stanley Milgram’s lesser known experiments on conditions for dissent

I recently read Thomas Blass’s intellectual biography of Stanley Milgram, The Man Who Shocked the World. In addition to being a rich behind the scenes look into Milgram’s life and work, the book described multiple variations of his well-known obedience experiments. While most were published in academic journals, some of his important findings are omitted from introductory psychology courses. These include experiments investigating conditions that reduce obedience to harmful orders. Yet, in totality, they offer a much richer picture of how circumstances shape behavior.

Milgram conducted over 20 variations of his famous and controversial obedience experiments. In his baseline study, 65% of participants administered the highest intensity shock to the victim or “learner” when told by an authority figure (the shocks were fake, and the learner was a confederate). Milgram began these experiments in the early 1960s, motivated to explain how ordinary Germans participated in the Holocaust.

The most effective experimental conditions for disobedience included:

• Dissenting peers: When participants witnessed other “teachers” defy authority, only 10% obeyed orders.
• Personal choice and responsibility: When participants could choose shock levels rather than being explicitly ordered to escalate, almost no one delivered shocks.
• Closer contact with the victim: When the participant had to hold the learner’s hand onto a shock plate, obedience fell to 30%.

Other conditions that diminished obedience included contradictory orders from multiple authority figures, reduced legitimacy of the authority (moving the experiment from Yale’s lab to a run-down office off-campus), and remotely delivered orders.

There are valid critiques of Milgram’s obedience experiments related to ethics as and replication. For an excellent overview, see 50 Years of “Obedience to Authority”: From Blind Conformity to Engaged Followership by Haslam and Reicher.

Note: an earlier version of this post was published here in June 2025.