The Milgram obedience experiments
I want to see what we can learn about wrongful obedience from the most celebrated effort to study it empirically, Stanley Milgram’s experiments conducted at Yale thirty-five years ago.
Even though these experiments are very well known, it is useful to review the details of what Milgram did and what he discovered.[444]Imagine, then, that you answer Milgram’s newspaper advertisement, offering $20 if you volunteer for a one-hour psychology experiment.[445] When you enter the room, you meet the experimenter, dressed in a gray lab coat, and a second volunteer, a pleasant, bespectacled middle-aged man. What you don’t know is that the second volunteer is in reality a confederate of the experimenter.
The experimenter explains that the two volunteers will be participating in a study of the effect of punishment on memory and learning. One of you, the learner, will memorize word-pairs; the other, the teacher, will punish the learner with steadily increasing electrical shocks each time he makes a mistake. A volunteer, rather than the experimenter, must administer the shocks because one aim of the experiment is to investigate punishments administered by very different kinds of people. The experimenter leads you to the shock-generator, a formidable-looking machine with thirty switches, marked from 15 volts to 450. Above the voltages, labels are printed. These range from “Slight Shock” (15-60 volts) through “Danger: Severe Shock” (375-420 volts); they culminate in an ominous-looking red label reading “XXX” above 435 and 450 volts. Both volunteers experience a45-volt shock. Then they draw lots to determine their role. The drawing is rigged so that you become the teacher. The learner mentions that he has a mild heart problem, and the experimenter replies rather nonresponsively that the shocks will cause no permanent tissue damage.
The learner is strapped into the hot seat, and the experiment gets under way.The learner begins making mistakes, and as you escalate the shocks he grunts in pain. Eventually he complains about the pain, and at 150 volts he announces in some agitation that he wishes to stop the experiment. You look inquiringly at the man in the gray coat, but he says only, “The experiment requires that you continue.” As you turn up the juice, the learner begins screaming. Finally, he shouts out that he will answer no more questions. Unflapped, the experimenter instructs you to treat silences as wrong answers. You ask him who will take responsibility if the learner is injured, and he states that he will. You continue.
As the experiment proceeds, the agitated learner announces that his heart is starting to bother him. Again, you protest, and again the man in the lab coat replies, “The experiment requires that you continue.” At 330 volts, the screams stop. The learner falls ominously silent, and remains silent until the bitter end.
But it never actually gets to the bitter end, does it? You may be excused for thinking so. In a follow-up study, groups of people heard the Milgram experiment described without being told the results. They were asked to guess how many people would comply all the way to 450 volts, and to predict whether they themselves would. People typically guessed that at most one teacher out of a thousand would comply - and no one believed that they themselves would.[446]
In reality, 63 percent of subjects complied all the way to 450 volts. Moreover, this is a robust result: it holds in groups of women as well as men, and experimenters obtained comparable results in Holland, Spain, Italy, Australia, South Africa, Germany, and Jordan; indeed, the Jordanian experimenters replicated the 65 percent result not only among adults but among seven-year-olds. Originally, Milgram had intended to run his experiments in Germany, to try to understand how so many Germans could participate in the Holocaust; his American experiments were merely for the purpose of perfecting his procedures.
After the American dry run, however, Milgram remarked: “I found so much obedience, I hardly saw the need of taking the experiment to Germany.”8In my view, we should regard the radical underestimates of subjects' willingness to inflict excruciating shocks on an innocent person as a finding just as important and interesting as the 65 percent compliance rate itself.9 The Milgram experiments demonstrate not only that in the right circumstances we are quite prone to destructive obedience, but also that we don't believe this about ourselves, or about our neighbors - nor do we condone it.10 Corroborating this final conclusion, subjects in another experiment had the Milgram set-up described to them, and were shown the photograph of a college student who had supposedly participated in the experiment as a “teacher.” They were asked to rate the student in the photograph (weak- strong, warm-cold, likable-not likable), based on appearance. Unsurprisingly, the ratings varied drastically depending on what level of shock the student had supposedly proceeded to - the higher the shock, the weaker, colder, and
8 Quoted in Robert B. Cialdini, Influence: Science and Practice 176 n. 2 (3rd edn., 1993). In 1970, David Mantell repeated some of the Milgram experiments in Munich, obtaining an 85 percent compliance rate in the basic experiment. David Mantell, The Potential for Violence in Germany, 27 J. Social Issues 101 (1971). So perhaps destructive obedience is a German pathology after all! - except that a similar 85 percent compliance rate appeared in an American replication as well (17 compliant subjects out of 20 - David Rosenhan, Some Origins of Concern for Others, in Trends and Issues in Developmental Psychology 143 [P. Mussen, J. Langer, & M. Covington eds., 1969]). Interestingly, Mantell introduced still another variation, in which the subject would see a prior “teacher” - a confederate of the experimenter - refuse to proceed with the experiment and indignantly confront the experimenter.
At that point, the experimenter revealed that he was actually an unsupervised undergraduate, and not a member of the institute where the experiment was conducted. In this version, more than half the subjects nevertheless complied, even after having observed the melodramatic scenario just described; and in subsequent interviews many of them criticized the previous teacher who had broken off the series of shocks. In his American replication, Rosenhan also obtained a compliance rate over 50 percent when it was revealed the experimenter was an unsupervised undergraduate.9 The fact that those who hear the experiments described vastly underestimate compliance may result in part from the “false consensus effect,” the well-confirmed tendency to exaggerate the extent to which others share our beliefs. Lee Ross & Richard E. Nisbett, The Person and the Situation: Perspectives of Social Psychology 83-85 (1991). That is, once a subject in the follow-up surveys has concluded that she would defy the experimenter in Milgram’s set-up, she is also likely to conclude that most people would. A sophisticated subject aware of the false consensus effect should compensate for it by upping her initial estimate of how many people would comply, say from 1 percent to 5 percent. Yet even this 500 percent compensation would drastically underestimate what Milgram actually found. Something more than false consensus is evidently at work here. Moreover, the follow-up subjects' belief that they would defy the Milgram experimenter is itself an unwarranted prediction, since we know that in the basic experiment two-thirds of them would comply to the 450-volt maximum. That is, the very premise for their false consensus - their prediction of their own behavior - is itself most likely false.
10 Miller, The Obedience Experiments, supra note 7, at 28-29. less likable the subject. The natural explanation of the “likability” finding is that subjects found the teacher unattractive to the degree that they found her behavior unattractive - from which it follows that they disapproved of her compliance.
In short, Milgram demonstrates that each of us ought to believe three things about ourselves: that we strongly disapprove of destructive obedience; that we think we would never engage in it; and that the odds are almost two to one that we are fooling ourselves to think we would never engage in it.
Milgram was flabbergasted by his findings. He and other researchers ran dozens of variations on the experiment, which I won’t describe, although I’ll mention some of them shortly. His battery of experiments, which lasted for years and ultimately involved more than 1,000 subjects, stands even today as the most imaginative, ambitious, and controversial research effort ever undertaken by social psychologists.
The Milgram experiments place moral norms in conflict. One is what I will call the performance principle: the norm of doing your job properly, which in hierarchical work-settings includes the norm of following instructions. The other is the no-harm principle: the prohibition on torturing, harming, and killing innocent people. In the abstract, we might think, only a sadist or a fascist would subordinate the no-harm principle to the performance principle. But the Milgram experiments seem to show that what we think in the abstract is dead wrong. Two out of three people you pass in the street would electrocute you if a laboratory technician ordered them to.
The question is why. At this point, I am going to run through several explanations of the Milgram results. None of them fully satisfies me. After exploring their weaknesses, I turn to the explanation that seems to me most fruitful.