Tuesday, September 03, 2013
A closer look at the Milgram Experiments
The other day, I heard this NPR interview with Gina Perry, the author of a new-ish book, "Behind the Shock Machine: The Untold Story of the Notorious Psychology Experiments," and was intrigued. My Criminal Law professor, Joe Goldstein, used the experiments in our unusual (but really fun) introductory course, as part of a discussion about consent and human-subjects research.
I have not read Perry's book (yet), but it sounds like she's established that Milgram was pretty set all along on reaching his "regular people will do really bad things if told to by an authority figure" (or, as this reviewer put it, his "most of us are potential Nazis") conclusions and troublingly uninterested in the possibility that his subjects could have been harmed by their experiences. Here's a bit, from a review in MacLean's:
To start with, Milgram was—in layman’s terms—nuts. He began the shock tests without any clear theory of what he was aiming to prove, and had to cobble it together afterwards, some of which he gleaned from a pamphlet entitled, “How to Train Your Dog.” He refused to consider that many people took it as a given that the stated aim of any psychological test was never its true purpose: A large proportion of the volunteers simply didn’t believe Yale would allow people to administer potentially fatal shocks. Among those who did accept what Milgram told them, far fewer than the 65 per cent he claimed actually continued to up the voltage. Worst of all, for fear the truth would leak out to other prospective volunteers, Milgram refused to fully debrief his subjects, many of whom were haunted for years by guilt at what they thought they had done.
If any readers have had a chance to read the book, I'd welcome and appreciate reactions.
TrackBack URL for this entry:
Listed below are links to weblogs that reference A closer look at the Milgram Experiments:
I have not read the book in question, just the excerpt and your comments. I will say I'm' highly suspicious that this is a serious piece of work, as opposed to a hatchet job. The Milgram results have been obtained in multiple experiments, not just by Milgram. They have turned out to be quite robust, and along the dimensions situationaists in social psychology like to emphasize: e.g., there is more compliance with the demand to turn up the voltage when the alleged experimenter is face-to-face with the subject, than when the experimenter communicates from another room over a loudspeaker. Milgram, himself, never would have claimed "we are all Nazis." Indeed, unlike the situationists, Milgram remained committed to the explanatory role of character, and concluded from his studies (consistent with the results) that some minority of the population always has the characterological fortitude to resist doing heinous things, even when ordered. One of the most striking results from the post-Milgram repetitions of the basic experiment is that there are strong national differences: e.g., German, alas, comply with the order to turn up the voltage at much, much higher rates than Australians!
Posted by: Brian | Sep 3, 2013 7:03:29 PM
Hi Brian - again, I am relying only on NPR and reviews (and what I remember of my own skepticism), but my sense is that the reviewers don't regard this as a hatchet job, and that the author has solid evidence of manipulation of the data and of unethical treatment of subjects. Best, R
Posted by: Rick Garnett | Sep 3, 2013 7:17:23 PM
There was a substantial scholarly conference on Milgram this summer, with Gina Perry as a central speaker. I didn't attend, but would be interested to hear from anyone who did. Here's the link: http://www.obediencetoauthority.com/
Posted by: anon | Sep 3, 2013 8:50:57 PM
I, too, have read only this post (I gather her book is out only today in the U.S.), but I echo Brian's skepticism, on a number of fronts.
First, there seems to be quite a bit of tension between the claim that "Milgram was pretty set all along on reaching his 'regular people will do really bad things if told to by an authority figure'...conclusions," on one hand, and the claim that "He began the shock tests without any clear theory of what he was aiming to prove, and had to cobble it together afterwards," on the other hand. When Milgram was a grad student at Harvard, Solomon Ashe was visiting faculty. Milgram was assigned to be his TF and was deeply influenced by Ashe and his famous "line judgment task" in which subjects showed remarkable conformity to other subjects' stated perceptions of the relative length of line segments. As Brian suggests, as early as 1954, Milgram announced that his dissertation would apply Ashe's conformity hypothesis (but asking subjects to identify the length of auditory tones rather than lines) to a study of differences among nationalities within Europe in social pressure and conformity. (Conclusion: Norwegians were more conforming than the French.) So it's true that Milgram alluded to Arendt's theory of the banality of evil (itself an unfortunate inference from her very partial observation of the Eichmann trial) as a way to explain his results (which, by the way, greatly surprised him, and which he resisted in his pilot study of Yale students and only accepted as real when they replicated in a general New Haven population), but banality doesn't necessarily mean lack of individual (or national) differences.
As for the claim that Milgram was "troublingly uninterested in the possibility that his subjects could have been harmed by their experiences," this strikes me as extremely uncharitable. Milgram was, in fact, well ahead of his time in research ethics -- he was, to my knowledge, both the first to engage in structured debriefing of his subjects and also the first (and, even today, near-only) researcher to survey his subjects about how they felt, post-debriefing, about having participated in his studies. He did this precisely because of the ethical issues his work raised. He asked the Norwegian subjects in his 1950s dissertation research, for example, and 70 out of 96 said they were glad or very glad, 1 said he was sorry, and none said he was very sorry. He did the same for at least some of his shock study subjects, where he said "I believe that the subject's viewpoint is of extreme importance, perhaps even paramount." The results: 83.7% reported that they were “glad” or “very glad” to have participated; 15.1% reported feeling neither glad nor sorry to have participated; and only 1.5% said they were “sorry” or “very sorry” to have participated. (Interestingly, these responses don't differ significantly when they are disaggregated according to whether subjects were “obedient” or “defiant.”) See Milgram, Subject Reaction: The Neglected Factor in the Ethics of Experimentation, HASTINGS CTR. REP. 19, 22 (Oct. 1977). Even today, subjects are debriefed, but are almost never asked to report their perspectives of a study's costs and benefits; that's a practice that federal regulators are only now seriously considering. I will certainly be interested to read about Milgram's subjects' perspectives as reported in Perry's book, but I would also be cautious in drawing too many conclusions from their memories of their participation 50 years after the fact.
There are, of course, continuing debates about what, exactly, Milgram's results mean, about how they should be explained, and about their generalizability. But these debates are largely due to the way that the most influential experiments in the history of social psychology (and, some would argue, in all of psychology) have been (mis)appropriated over the decades, and because of the fact that replications and variations on his experiments that might flesh out the nuances of his results can no longer be conducted under the IRB regime, which regards the obedience studies as the quintessential example of unethical social science research, subjects' overwhelmingly non-negative responses notwithstanding. (An IRB or two has allowed a clever work-around that allows researchers to approximate the Milgram experiment. Sort of.)
There has already been quite a bit of archival research by various scholars on Milgram and the obedience studies. It's certainly possible that Perry has uncovered additional archival sources that seriously complicate the picture I've offered above, but I think she's got an uphill road to climb to support some of the stark claims you describe her making here. FWIW.
Posted by: Michelle Meyer | Sep 3, 2013 11:27:04 PM
Thanks, Michelle - I gather from interviews that she was able to access new archival material. And I do not mean to be uncharitable, but it does not appear that he took much care for his subjects' experiences in the New Haven experiments. I would like to learn more about the work-arounds you describe.
In one review (in the Chronicle, I think) some additional work by SM with teacher-learner pairs who knew each other is described. Some differences in results, apparently.
Posted by: Rick Garnett | Sep 4, 2013 8:48:16 AM
Hi Rick -- I didn't mean to call you uncharitable; I was referring to Perry, whose claim about Milgram's lack of concern for his subjects' welfare I took you to be reporting. Having now listened to the NPR podcast and read both the CHE and MacLean's reviews you link to, I remain skeptical that she's uncovered new evidence that reveals serious new concerns about his "methods and conclusions," as one of the reviews put it. Still, since I haven't yet read her book, I suppose it's I who am being uncharitable.
Milgram's obedience studies involved some 20 conditions. Subjects were more or less obedient depending on those conditions. (For instance, when the experimenter was in the room with the teacher, the teacher was more likely to be obedient, compared to conditions in which the experimenter instructed the teacher in real time over the phone or in advance by a tape recording. Prestigious Yale researchers identifying as such similarly commanded more obedience from teachers than did researchers from the fake "Bridgeport Research Group," who ran the Bridgeport studies. And so on.) It's not surprising that Milgram's work yielded different results in different conditions -- figuring out why and when and how people obey or not was part of the point of the research. It's true that the 65% obedience number comes from the most favorable condition (Experiment 5), but so what? As far as I know, that number is accurate for that condition, and most people find it pretty astonishing. If there's been too much emphasis on Experiment 5 at the expense of others involving less obedience, then that's the fault of textbooks and the media, and not very likely to be the fault of Milgram. (Usually third parties, especially the media, hype science results. Sometimes, though, researchers are all too eager to help, and perhaps that was somehow the case here. But the bare fact that other conditions existed were obedience was less prevalent hardly seems damning to me.)
The CHE review of Perry's book reports that she discovered one "secret" -- by which I gather she means unpublished -- condition in which, as you say, teachers and learners knew each other and refused to shock much sooner (but not, notably, immediately) than in other conditions. I'm a big proponent of publishing null results -- I think it's one of the biggest issues in research ethics; publication bias skews our evidence base, which leads us to make bad decisions. But if one condemns researchers who fail to publish null results, then one had better condemn not just Milgram but the other 99.99% of researchers who do the same to this day -- usually less because they're trying to hype their positive results or manipulate data than because journals aren't terribly interested in null results and department chairs and hiring, tenure and promotion committees aren't terribly interested in what's not been published (or has been published in a less well-regarded journal). In others words, it's entirely plausible that Milgram was utterly ordinary in his decision not to publish the results of one of his conditions. (It's also possible that Milgram could not secure a sufficiently large sample size for this condition, which required pairs of related subjects or subjects who were otherwise known to each other.)
As for her discussion about how many teachers knew what was really going on, I'd need to know more, but it seems to me that Milgram has a point: subjects who were obedient do indeed have an incentive after the fact to claim that they knew all along that the shocks they were administering weren't real. Perry and Milgram can disagree about how many subjects actually knew what was going on, and there's likely no way to resolve that dispute. (Note, too, that if she's right that more subjects than Milgram admitted were hip to the ruse, then the argument that the studies were monstrously unethical because they tormented subjects for years is undercut.)
As for her claim that Milgram refused to fully debrief subjects, my guess is that she's referring to the fact that Milgram debriefed in two stages: immediately following the study, he told them that the shocks were really at levels that were appropriate for smaller animals, but not humans. My interpretation of this is that subjects understood that confederates were never in any serious danger, which ought to have mitigated any psychological effects of participation. Once Milgram finished all of his conditions and was done recruiting subjects, he fully debriefed them. (Of course, subjects who obeyed will forever live with the unpleasant self-knowledge that they were willing to give what they thought were potentially lethal shocks to another human being, but we didn't need Perry to tell us that.)
The work-around I mentioned comes from J. M. Burger, Replicating Milgram: Would people still obey today? 64 American Psychologist 1-11 (2009). See also the excellent symposium on Milgrim and the Burger "replication" in the same issue. Burger noticed that in Milgram's studies, something like 75% of subjects who shocked at 150 volts continued until the bitter end, i.e., to the "potentially lethal" 450 volts. The Santa Clara IRB allowed him to run Milgram's famous 65% experiment (plus one variation on it, for a total of two conditions) provided that he end the experiment when the subject refused to continue or when she shocked at 150 volts, whichever came first. Subjects who shocked at 150v he deemed to have shocked up to 450v. He fully and immediately debriefed subjects, and he screened out subjects who indicated on various psychology batteries that they might be negatively affected by the study. He found 63% and 70% obedience rates in the two conditions (the difference wasn't statistically significant), effectively replicating Milgram.
It's not clear to me that this makes the ethics substantially different, and note that it would not have been possible without the original Milgram experiments. Burger's "replication" has also been doubted since he screened out many of the subjects who may have been most likely to have refused to obey (thus artificially inflating the obedience rate) -- see Elms's contribution to the symposium noted above.
Posted by: Michelle Meyer | Sep 4, 2013 3:52:15 PM
Hi Michelle -- thanks very much for these engaged, illuminating, and thoughtful comments. I am (obviously!) not qualified to referee your challenges to (what appear to be) Perry's claims, and I should not try. I hope that, if you do read the book, you'll "come back" (or write off-line) and let me know what you think. Again, the NPR thing caught my attention primarily because it "took me back" to first-year (and to a paper I wrote for Robert Burt, on the ethics of human experimentation), and I was, I admit, grabbed (hooked?) by the "a new book suggests that one of the best known psychological experiments might not be what we thought!" line.
I'm looking for more reviews -- I've found one at Scientific American and Psychology Today (http://www.psychologytoday.com/blog/fulfillment-any-age/201301/the-secrets-behind-psychology-s-most-famous-experiment) and will keep looking for more. Thanks again.
Posted by: Rick Garnett | Sep 4, 2013 4:01:57 PM
Thanks, Rick, and will do.
Posted by: Michelle Meyer | Sep 4, 2013 8:32:30 PM
Well, I must say, this WSJ review of the book by Carol Tavris is pretty satisfying: http://online.wsj.com/article/SB10001424127887323324904579040672110673420.html (paywalled, alas).
Here are some choice selections:
Such important revelations [e.g., about deviations from the protocol in which the experimenter sometimes prodded the teacher numerous times rather than 4] add complexity to the Milgram story, and Ms. Perry is to be commended for reporting them. Yet she wants her findings to be seen not as the revisions that this classic study warrants but as a fatal stake through its heart. In this goal, she fails. Whatever modifications of the findings are called for, its fundamental results have been replicated in other countries, and most recently here, in an ethically acceptable version.
Ms. Perry is so angry about the experiments that her hostility permeates the narrative, making her an unreliable guide for general readers. . . . She accuses Milgram of bending his data to shape his story, yet she does the same for hers. Thus she approves of one interviewee who says that "the only evil in the obedience research . . . was 'the unconscious evil of experimenters' " but admits that she stopped listening to interviewees who defended Milgram's work.
. . . Ms. Perry insists that people's personalities and histories influence their actions. But Milgram never disputed that fact; his own research found that many participants resisted.
. . . At one point, Milgram's biographer reports to Ms. Perry that, although many subjects suffered at the time of the study, only 1.5% said later in a survey that they were sorry to have taken part. Because she doesn't believe this claim, she mishears it. "How could people that Milgram had described as having agonized, sweated . . . and groaned through the experiment later say that it had no effect on them?" But that isn't what they said. They said it did have an effect—not always a salubrious one, but one they learned from.
Posted by: Michelle Meyer | Sep 6, 2013 8:12:57 PM