From “Ridiculous” to “Glad to Have Helped”: Debriefing News Delivery and Improved Reactions to Science in Milgram’s “Obedience” Experiments

Our data are 56 audio recordings from the Stanley Milgram Archive maintained by the Manuscripts and Archives Department at Yale University Library. The recordings preserve the pre-experimental discussion, experiment, and post-experimental interview for 56 subjects from three experimental conditions, all receiving full debriefing (see Table 1).

The larger project in which this article originated involved the purchase of 117 recordings—from conditions 2 (“Voice-feedback”), 3 (“Proximity”), 20 (“Women as Subjects”), 23 (“Bridgeport”), and 24 (“Relationship”)—that had received prior digitization and sanitization (erasure of participants’ names: Kaplan 1996).⁴ Selection was haphazard rather than random due to the archive’s incompleteness, the relative availability of digitized and sanitized recordings in each condition, and a limited budget. Since debriefing was deceptive rather than full in conditions 2 and 3, we drew the 56 recordings analyzed in this article solely from conditions 20, 23, and 24.

To study Milgram’s debriefing procedures using conversation analysis, we first listened to the entire experimental performance and interview of all 117 participants. During this process, we took detailed notes on the interviews, paying particular attention to full debriefings in conditions 20, 23, and 24. We then listened again to the entire interview for the 56 full-debrief participants, edited our notes, transcribed all 56 full debriefings, and wrote detailed analyses of the transcripts. Finally, we listened once again to all 117 debriefings and participation assessments (see component 9, Figure 1 below), noting Experimenter practices that elicited improved initial assessments and comparing and contrasting the two debriefing procedures. Overall, our conversation-analytic strategy was thus, first, to assemble a collection of as large a number of instances as possible of the same interactional phenomenon (news delivery in Milgram’s full debriefing) and, second, to study the collection to gain insight into the phenomenon’s general features (Sacks 1992).

As noted, Perry (2012, 2013) has shed important new light on Milgram’s debriefing procedures. However, missing is a detailed contextualization of both types of debriefing within the interview. Here, we show how our archival research significantly revises Perry’s findings on Milgram’s debriefing practices. This section also provides necessary context to our later analysis of full debriefing and discussion of engaged followership.

Figure 1 illustrates the positioning of both debriefing procedures with respect to the entire interview, conducted by the Experimenter immediately after the experiment.⁵

Although interviews were highly structured, the Experimenter allowed participants ample time to develop responses. Interviews therefore ranged in length, with many reaching more than 20 minutes.⁶ The main differences between deceptive-debrief conditions 2 and 3 versus full-debrief conditions 20, 23, and 24 are that, by spring 1962, components 3 and 8 had disappeared, 12 replaced 11, and 9 followed 12. Moreover, the Experimenter performed full debriefing, like debriefing overall, in two ways: either more extensively (condition 20) or less extensively (conditions 23 and 24). As with the introduction of full debriefing itself, the difference seems primarily due to Milgram’s concern to ensure that female participants (condition 20) left his lab relaxed.⁷Perry (2013:84) argues that a contributing factor was mounting ethical criticism at his funding agency (National Science Foundation) and among psychologist peers. Whereas the more-extensive full debriefings featured the perspective display sequence (PDS) and the panoply of other “reconciliation practices” discussed further in this article, the less-extensive later ones lacked this sequence, featured fewer reconciliation practices, and took less time.

Deceptive debriefing was conducted as follows. Teachers completed a demographic questionnaire (component 10) while the Experimenter retrieved the Learner from an adjacent room. The Experimenter then performed a script indicating the Learner’s “problem was that [he] was too nervous. … This machine has previously only been used to shock animals like mice and small rats. … Learner, the amperage is adjusted so the shocks you received were only slightly more painful than the sample shock.” Following conciliatory scripted talk by the Learner showing him unharmed and unresentful (“If I’d been the Teacher, I would have done the same thing. … No harm done.”), the Experimenter confirmed the Teacher had his check for $4.50 and led him outside.

By contrast, full debriefing differed in several key ways. Unlike the deceptive procedure, which typically lasted two minutes or less, full debriefing took longer, especially the more-extensive version in condition 20 (see endnote 6). After demographic component 10, the Experimenter started removing central planks of the cover story: the Learner didn’t receive shocks; he was actually working together with the Experimenter; Teacher behavior was the focus; the experiment was about obedience to authority, not punishment’s effect on memory and learning. Then, the Experimenter called in the Learner to meet the Teacher “under different circumstances.” Also, as noted, whereas participation assessment (component 9) had previously appeared before deceptive debriefing (component 11), it was now performed after full debriefing (component 12), immediately prior to the Teacher’s exit. Despite these differences, however, the second debriefing procedure was only relatively “full”: Milgram refrained from total disclosure in all conditions and even in the summer 1962 questionnaire, mailed after all experiments had ended (Perry 2013).

Though contrasting, the deceptive-debrief interviews and the full-debrief ones both dealt with two issues of moral import highly relevant to participants’ assessments. First, the interviews addressed responsibility, or the problem of assigning blame for the Learner’s receiving shocks against his will. For instance, during full debriefing the Experimenter typically highlighted “obedient” participants’ resistance (“I had the feeling you were reluctant, that you didn’t really wanna do this.”). By mentioning their lack of enthusiasm and/or active resistance, he invited responses of having been concerned “all along” about the Learner and as being less culpable than the Experimenter himself. Second, the interviews emphasized reconciliation among the three parties. In both types of debriefing the Learner reappeared, claiming responsibility for the problems with continuation (e.g., overreacting to the shocks—“I was nervous.”) and lack of resentment (“No hard feelings.”). As noted, Milgram especially emphasized reconciliation in condition 20 due to his concern about “panicked” women. As shown in the next section, a crucial feature of full debriefing was the Experimenter’s arsenal of what we call “reconciliation practices” (e.g., “Do you feel better now?”). In sum, by addressing issues of responsibility and reconciliation, the interviews could improve participants’ initial evaluations of the experiment, especially in full debriefing. Inconsistent assessments, in turn, are directly relevant to theories of behavior in Milgram’s lab such as engaged followership, as discussed later.

Considerable interactional organization of news delivery characterizes Milgram’s full debriefing. In condition 20, the Experimenter delivers scripted news using what Maynard (1989, 1997, 2003) calls the (1) perspective display sequence and (2) news delivery sequence. This section uses conversation analysis to examine the recurrent and orderly practices of talk and social interaction (Schegloff 2007) by which the Experimenter elicits Teachers’ assessments of the experiment (PDS) and tells them its true purpose (NDS).⁸ Although Perry (2013:84) presents a partial interview transcript of “defiant” subject 2316, claiming the example is “representative” of full debriefing, missing is discussion of the more extensive version in condition 20. A more complete analysis would also articulate its sequential structure, which participants co-produce (notwithstanding potentially relevant inequalities in authority, expertise, and power between Experimenter and Teacher). The picture that emerges of Milgram’s full-debriefing procedure differs in fundamental respects from that painted by Perry.

In connecting current archival research on Milgram with contemporary theorizing of “obedience,” we have built on Perry’s (2012, 2013) work while significantly revising her conclusions about Milgram’s debriefing procedures. Full debriefing in particular should be seen in terms of its local context of two- (Experimenter and Teacher) and three-party (Learner included) social interaction—as a multiparty achievement rather than unilateral “monologue.” Although deceptive debriefing was more common than full, the latter was also a major feature of Milgram’s project overall (conditions 20, 23, and 24 included 100 participants). Also, although he indeed intended the interview to reconcile participants with the Experimenter (and Learner), Milgram’s team sought to do this with reconciliation practices that seem, in themselves, less morally dubious to us than to Perry. Rather than Machiavellian justifications for helping to harm the Learner, the Experimenter offered “obedient” participants mildly sanctimonious support for any resistance they showed or reluctance they felt. Below, we discuss in further detail the contribution this article makes to conversation analysis and discursive psychology, Milgram literature and “obedience to authority,” and the study of public attitudes toward science.

First, in the course of showing that Milgram’s full debriefing could improve participants’ evaluation of the experiment the course of showing that Milgram’s full debriefing could improve participants’ evaluation of the experiment, we have argued that social psychological debriefing procedures should be viewed as interactional accomplishments. Although other conversation-analytic studies have examined research activities such as survey interviewing and informed consent acquisition (Maynard and Schaeffer 2000; Speer and Stokoe 2012), none to our knowledge has investigated debriefing in real time. A conversation-analytic or discursive psychology approach would allow analysts to identify concerns to which the research participants themselves orient as relevant and consequential (Schegloff 2007).

As we’ve seen, Milgram intended his debriefing procedure, especially in its full version, to reconcile Teachers with the Experimenter and Learner after the stresses of participation. The Experimenter delivered full debriefing news with a panoply of practices that collectively worked to produce the outcome of “reconciliation”: participants’ positive assessment of the experiment, participation, or personal well-being. Although especially relevant to Milgram’s project, given the extreme nature of the experimental task, similar issues of reconciliation and positive participation assessment arise in any social psychological experiment involving deception. Though the gain in knowledge may be considerable, such research relies on dishonesty—about the nature of the research, the motives of the confederates, and so forth—creating situations that lend themselves to potential abuses of power, authority, and expertise and that must be explained candidly in the debriefing interview. Future studies of the detailed interactional organization of actual debriefings, as captured by video- or audio-recordings, would usefully contribute to ongoing debates in social psychology and elsewhere about the use of deception (see, e.g., Cook and Yamagishi 2008). Such studies could fruitfully investigate the ways that researcher and subject negotiate their relative statuses within, and stances toward, the research, along with the power dynamics this may entail. In this vein, despite the long-standing ethical controversy surrounding Milgram (Baumrind 1964), our conversation analytic approach has provided evidence that power asymmetries—in the form of opportunities to talk—in Milgram’s full debriefing may have been less stark than Perry (2013) claims. As shown above, the Experimenter provided numerous turn-transition relevance spaces during news delivery, actively pursuing participant responses. Likewise, we found that full debriefings lasted significantly longer than Perry’s “minute and a half” (2013:85). In such ways, the Experimenter took a more conversational, egalitarian approach to full debriefing than to its deceptive counterpart.

Second, our finding of improved assessments has significant implications for engaged followership. Did most “obedient” participants comply because they identified with Milgram’s scientific values and putative research on punishment’s effect on learning and memory, as maintained by Haslam and Reicher? Granted that this article has examined initial PDS assessments from a single condition, 20, it is nevertheless striking that so few participants here, regardless of outcome category, initially assessed the experiment positively (see Table 2). Missing are the indications of close identification with the Experimenter, engagement with his perceived goals, and approval of his management of the Learner’s dramatic resistance that engaged followership theory would predict. Given that Milgram designed condition 20 (“Women as Subjects”) to differ from the more famous condition 2 (“Voice-feedback”) only by participant gender, and that to this point in the interview (prior to debriefing) little distinguishes the full debriefing version from the deceptive one, doubts arise as to the theory’s ability to comprehensively explain “obedience.” If, as Haslam et al. (2015) claim, participants’ generally positive questionnaire responses should be regarded as evidence supporting the theory, shouldn’t initially ambivalent and negative responses prior to debriefing be taken to cast doubt on it?¹¹ Whereas engaged followership treats positive post-experimental assessments as displaying a static commitment “all along” to the Experimenter’s leadership and importance of the experiment, we have found that the interviews could have a dynamic, ameliorative effect on initial reactions—by changing, in full debriefing, the assessments’ object from an experiment about “punishment and learning” to one about “obedience to authority.” The basic definition of the situation proved fluid, with participants’ evaluations shifting accordingly. It seems to us that insufficient attention to full-debrief participants’ dynamically changing experience of, first, the putative experiment on “punishment” and, second, the actual experiment on “obedience” has created a mistaken impression: that their sympathy for the latter typically extended to the former, such that they positively assessed the experiment before the cover story was dropped.

This article, then, has made the important, yet relatively modest, argument that Milgram’s full-debriefing interview could improve participants’ initial assessments. We took this approach because PDS assessments, in which participants provide explicit evaluations of the experiment, are relatively rare in the interviews overall, being mostly confined to condition 20, and our strategy has been to compare explicit initial assessments with later ones provided during and after full debriefing. Our approach has thus been conservative and has not examined additional practices by which participants displayed their perspective during and immediately following the experiment. Our immersion in the larger collection of 117 recordings has familiarized us with many instances of negative reaction to participation, and we have little doubt that Perry is right that “distress and anger” were initially widespread.¹² By contrast, Haslam and Reicher, whose archival research relies solely on the questionnaires sent out weeks or months later, argue that “participants’ post-experimental responses provide little evidence that they were stressed or angry” (2017:71). As discussed earlier, in this article we have sought to explain this tension: the full debriefings could improve assessments. Here we note further that, if we and Perry are correct about the initially negative reactions, and granted Haslam and Reicher’s findings about the questionnaires, then our argument’s scope widens accordingly: the interview often improved assessments, in both deceptive and full debriefing conditions. All this provides compelling reasons for doubting that engaged followership is uniquely adequate for explaining behavior in Milgram’s lab.

Finally, our analysis also contributes to the study of public attitudes toward science. In examining naturally occurring debriefings, we have shown how assessments of science can be affected by participation in research and how such participation can involve researchers and subjects in concerted negotiations about the meaning and value of scientific work. Of course, Milgram’s project constitutes an extreme case in many respects: not only did the experiments go far beyond what would be allowed in current research, but the participants themselves may have differed in important ways from the general population; on average, they were probably somewhat more sympathetic to science—people with a “scientific habitus” (Gauchat 2011)—and were thus more likely to respond to Milgram’s newspaper ad or mail solicitation in the first place. Further, the Milgram experiments of 1961–1962 occurred during a high point of public trust in science, the 1950s and 1960s, which has since declined. In an era in which trust in experts is waning (Giddens 1991), it is increasingly important to understand how public participation in science, whether in clinical trials (Roberts 2002) or social scientific studies (Maynard and Schaeffer 2000; Speer and Stokoe 2012), affects people’s dispositions toward the scientific enterprise. Analysis of language and social interaction in scientific research activity, as performed in the present article, provides a valuable complement to meso- and macro-level investigations of trends in public understanding of, attitudes toward, and trust in science (Gauchat 2011, 2012; Miller 2004).

In conclusion, this article’s findings on debriefing news delivery in Milgram’s laboratory raise fundamental questions not only about prior archival research on his research practices, but also about the meaning of participants’ actions. Although social psychological processes of social identity such as engaged followership do seem likely to operate in many real-world contemporary and historical settings of compliance with malevolent authority, we have offered a compelling, empirically grounded challenge to that theory as a general model of Milgram participants’ “obedience.” It would seem that no single social psychological process uniquely suffices to explain their actions but rather that compliance resulted from multiple processes involving a complex interplay of situational forces and individual dispositions (Overy 2014; Staub 2014) operating in concert with organized practices of social interaction (Hollander 2015; Hollander and Turowetz 2017). The best starting point for future interpretations of “obedience” and its real-world relevance may perhaps be acknowledgment of such multiplicity and interplay.

We thank the Social Psychology Quarterly anonymous reviewers for very helpful feedback on an earlier draft of this article. Doug Maynard and Ceci Ford offered crucial guidance and criticism of the dissertation in which this project originated. We also benefited from jointly presenting this research at the July 2017 International Institute for Ethnomethodology and Conversation Analysis conference at Westerville, Ohio. Cynthia Ostroff at Yale University Library’s Manuscripts and Archives Department helped arrange purchase of Milgram recordings in 2012. Research was funded by National Science Foundation grant number 1103195. This article features transcribed excerpts from audio-recordings of the Milgram Obedience Experiment; permission to publish these is granted by Alexandra Milgram. This article is equally coauthored, with both authors making equal contributions.