1 British Journal of Social Psychology (2017) © 2017 The British Psychological Society www.wileyonlinelibrary.com Normalizing trust: Participants’ immediately posthoc explanations of behaviour in Milgram’s ‘obedience’ experiments Matthew M. Hollander1* and Jason Turowetz2 1 2 Department of Sociology, University of Wisconsin, Madison, Wisconsin, USA School of Media & Information, University of Siegen, Germany We bring an ethnomethodological perspective on language and discourse to a data source crucial for explaining behaviour in social psychologist Stanley Milgram’s classic ‘obedience’ experiments – yet one largely overlooked by the Milgram literature. In hundreds of interviews conducted immediately after each experiment, participants sought to justify their actions, often doing so by normalizing the situation as benign, albeit uncomfortable. Examining 91 archived recordings of these interviews from several experimental conditions, we find four recurrent accounts for continuation, each used more frequently by ‘obedient’ than ‘defiant’ participants. We also discuss three accounts for discontinuation used by ‘defiant’ participants. Contrary to what a leading contemporary theory of Milgramesque behaviour – engaged followership – would predict, ‘obedient’ participants, in the minutes immediately following the experiment, did not tend to explain themselves by identifying with science. Rather, they justified compliance in several distinct and not entirely consistent ways, suggesting that multiple social psychological processes were at work in producing Milgram’s ‘obedient’ outcome category. Although many considered assessments of the ethics of Stanley Milgram’s classic and controversial research on ‘obedience to authority’1 have been offered (Baumrind, 1964; Kaufmann, 1967; Nicholson, 2011), little is currently known about how the hundreds of research participants themselves oriented to the moral significance of their actions. With the ongoing resurgence of interest among social psychologists in interpreting and explaining social behaviour in Milgram’s experiments (Burger, 2009, 2014; Haslam, Reicher, Millard, & McDonald, 2015; Haslam, Loughnan, & Perry, 2014; Haslam, Reicher, & Birney, 2014; Hollander, 2015; Perry, 2012; Rochat & Blass, 2014; Russell, 2011, 2014a), it is striking that so little is currently known about how participants justified their actions. In one of twentieth-century social psychology’s most influential projects, Milgram had an authority (the confederate ‘Experimenter’) direct a research participant (the ‘Teacher’ *Correspondence should be addressed either to Jason Turowetz, 132 State Street, Newburyport, MA 01950, USA (email: jason.turowetz@gmail.com) or Matthew M. Hollander, 523 Wilson St, Madison, WI 53703, USA (email: mholland@ ssc.wisc.edu). This paper is equally co-authored. 1 ‘Milgramesque behaviour’ (our paper’s neutral term) may not result from a single social psychological process of ‘obedience to authority’ (Reicher & Haslam, 2011). Although we below describe participants as ‘obedient’ and ‘disobedient/defiant’, we use Milgram’s terms simply to refer to members of the two outcome categories, and do not mean to imply support for his interpretation of their actions. DOI:10.1111/bjso.12206 2 Matthew M. Hollander and Jason Turowetz subject) to deliver putative electroshocks to a peer volunteer (the confederate ‘Learner’).2 Operationalizing participants’ responses to this situation of constrained choice as ‘obedient’ or ‘defiant’ (‘disobedient’), Milgram (1963) claimed to demonstrate that situational variables under experimental control can cause changed rates of obedience. However, most commentators now agree that Milgram’s attempt to formulate a comprehensive theory explaining the phenomenon – that of ‘agentic state’ (Milgram, 1974) – was unsuccessful (Blass, 2004). As Haslam et al. (2015) note, part of the problem with this theory was its failure to ‘. . . provide[] any evidence that allows us to understand how participants justified their behaviour or why they supported his research’ (p. 3). ‘To get a clearer picture of participants’ reflections on their experiences, these therefore need to be examined in the round rather than by looking [as Milgram (1974) does] at selected statements in isolation’ (Haslam et al., 2015, p. 6). We wholeheartedly agree: Theories seeking to explain Milgramesque behaviour must take into consideration how, in the context of their situated involvement, participants themselves made sense of their actions. This study aims to provide just such an analysis of participant-produced accounts for their conduct. We do so by exploiting a data source hitherto virtually untapped in the Milgram literature: the hundreds of immediately post-experimental interviews from Milgram’s archived audio recordings. In analysing accounts for their just-completed experimental behaviour produced by 91 participants from both ‘obedient’ (n = 46) and ‘defiant’ (n = 45) outcome categories, we find that participants tended to explain themselves in recurrent ways. Accounts for continuation come in four distinct types: following instructions, Learner was not really being harmed, importance of the experiment, and fulfilling a contract. Additionally, we find three accounts for discontinuation used by ‘defiant’ participants. After discussing the interview situation and the seven accounts, we spotlight Learner was not really being harmed, a justification voiced by most (33/46, or 72%) of the ‘obedient’ participants in our collection. These participants claimed they continued because they did not think the electroshocks posed a danger of injuring L, despite strong indications to the contrary. Given this account’s high frequency, it offers an important clue towards better understanding Milgramesque compliance. In addition to sheer frequency, we focus on the L not being harmed account for another crucial reason. It illustrates particularly well this paper’s theoretical claim about Milgramesque behaviour, one informed by our analysis of all the interview accounts for (dis)continuation. Participants’ explanations often embody sense-making practices of social interaction that sociologist Harvey Sacks (1984), student of Erving Goffman and founder of conversation analysis, memorably termed ‘doing being ordinary’. Drawing on Harold Garfinkel’s ethnomethodology (Turowetz, Hollander, & Maynard, 2016), Sacks observed that the routine perception of living in a familiar and recognizable world can be empirically shown to be an ongoing accomplishment by society’s members, one akin to an active, ongoing ‘job’. The world usually seems ordinary because our non-conscious, skilled practices are ceaselessly rendering it as such. Such practices constitute everyday situations (driving to the store, seeing a movie with a date, participating in a psychological study), which in turn reflexively act back on our ongoing practices for navigating these situations as they develop. We perform constitutive practices as a matter of course, and ‘trust’ (Garfinkel, 1963) other people populating the situation to do the same. 2 Henceforth, we abbreviate the parties as ‘T’ (participant Teacher), ‘L’ (Learner receiving shocks), and ‘E’ (Experimenter directing continuation). Normalizing trust 3 Trust in this sense is action in accordance with the presumption that most situations and people are basically as they appear to be (cf. Goffman, 1959) – that a traffic intersection ‘really’ is such a situation, that salespeople and parents and scientists (such as Milgram’s confederate playing the Experimenter role) ‘really’ are such persons. Culturally competent social actors can only detect deception and illusion against the perceptual background constituted by our everyday normalizing practices: Trust (certainty, normality) is a condition of possibility of scepticism (doubt, abnormality). When commonsensical expectations are not met – for example, your boss suddenly starts treating you as a complete stranger – we perform interactional work to preserve our sense of everyday reality, normalizing the disrupted situation as ‘ordinary after all’ – she was only kidding (Garfinkel, 1967: ‘breaching tutorials’). In the Milgram setting, participants initially displayed trust (in Garfinkel’s sense) by treating the experiment as a benign study of learning for which they had volunteered. However, as interactional tension between L (projecting discontinuation) and E (continuation) ratcheted up the ambiguity of the situation, some participants came to experience it as one in which L ‘really’ was being harmed. Of these Gestalt-switch participants, some resisted forcefully enough to stop the experiment and earn Milgram’s ‘defiant’ categorization. Given the frequency of the L not really being harmed account, we argue that everyday normalizing practices go far towards explaining why many failed to take such decisive action. Engaged followership This paper builds upon, while also challenging, important recent contributions to Milgram scholarship. For instance, Haslam et al. (2015) provide an incisive analysis of Box 44 of the Yale archive, consisting of participants’ largely positive, mailed questionnaire responses returned to Milgram in the weeks and months following participation. These authors see these data as evidence supporting engaged followership, a social identity theory (Reicher & Hopkins, 2001) currently widely accepted as explaining Milgramesque behaviour. According to S. Alexander Haslam and Stephen D. Reicher, its main authors (Haslam, Loughnan et al., 2014; Haslam, Reicher et al., 2014; Reicher & Haslam, 2011; Reicher, Haslam, & Smith, 2012), ‘people are able to inflict harm on others not because they are unaware that they are doing wrong, but rather because – as engaged followers – they know full well what they are doing and believe it to be right’ (Haslam et al., 2015, p. 25). Such engagement involves social ‘identification with a “noble” cause’ (p. 25) such as ‘science as a source of human progress and welfare’ (p. 23). In the Milgram context, engagement is an attitude that, although compliance may be difficult, the experiment is important, worthwhile, and/or scientifically or practically valuable. In short, participants are motivated to overcome the difficulties of compliance by E’s appeals to science. In examining participants’ interview accounts, we too are using the Yale archive to achieve more insight into Milgramesque behaviour and its real-world relevance.3 We see this paper as directly responding to Haslam et al.’s (2015) important call for research accomplishing ‘. . .a more structured analysis of the experiences reported by Milgram’s participants which seeks to establish the relative merits of agentic state and engaged followership explanations. It is here that the Milgram archive at Yale proves invaluable’ 3 Gibson (2013a,b, 2014), though not necessarily accepting engaged followership, likewise uses Yale’s archived audio recordings in his important research in discursive psychology on negotiation and knowledge work in T’s interactions with E. 4 Matthew M. Hollander and Jason Turowetz (p. 8). However, unlike these authors, we use as data participants’ accounts provided immediately after the experiment. This is critical, insofar as the subsequent debriefing process often changed participants’ minds about the experiment, as we discuss below. Our findings thus bear directly on the theory of engaged followership, and we examine the implications below. Data and methodology This paper examines Milgram participants’ immediately post-hoc accounting for their experimental behaviour. As with all post-hoc reports about social behaviour, whether immediate or not, such materials furnish, at best, only indirect explanations regarding the actions described. Nevertheless, just as participants’ mailed questionnaire data, as used by Haslam et al. (2015), illuminate (albeit indirectly) social psychological processes occurring in the experiment, so too do participants’ interview accounts. And, insofar as the interviews occurred immediately (vs. weeks and months) afterwards, and prior to debriefing, this data source is even more indispensable for interpreting Milgramesque behaviour. In studying the interactional production of interviewees’ accounts, as well as their forms and general features, our theoretical and methodological orientation was ethnomethodology. This paper should be seen as building on prior ethnomethodological research on trust conditions and social practices for normalizing ‘disrupted’ situations (Duck, 2015; Garfinkel, 1967; Jefferson, 2004; Liberman, 2013; Rawls, 2000; Sacks, 1984).4 Like conversation analysis, ethnomethodology is a research tradition that brackets questions about ‘why’ a social action occurred, instead highlighting the detailed accomplishment (the ‘how’, ‘just-this-ness’) of its production (Liberman, 2013). This perspective helps us explicate just how participants made sense of their actions, which in turn provides a circumstantial basis for explaining why they acted as they did.5 Accordingly, our data are 91 recordings from the Stanley Milgram Archive maintained by the Manuscripts and Archives Department at Yale University Library.6 The recordings document the experimental performances and immediately post-experimental interviews and debriefings of 91 different participants (46 ‘obedient’; 45 ‘defiant’). Owing to circumstances such as the archive’s incompleteness and a limited budget for acquisition, we first selected 117 recordings from the relatively complete Conditions 2 (‘Voicefeedback’), 3 (‘Proximity’), and 20 (‘Women as Subjects’) – as well as the less complete Conditions 23 (‘Bridgeport’) and 24 (‘Relationship’) – that had already received preparatory digitization and removal of participants’ names (Kaplan, 1996).7 Of these 4 In particular, this research shows that, when social expectations are violated, participants work to normalize or rationalize the violation; indeed, even when they conclude that the situation is not ordinary after all, they nonetheless emphasize, in post-hoc reports, how they initially tried to normalize it (Jefferson, 2004). 5 Thus, we draw on ethnomethodological insights and tools to explicate participants’ sense-making practices. Of course, this does not provide a direct purchase on what occurred during the experiment – a limitation that all attempts to explain Milgramesque behaviour face. Nonetheless, we believe it is possible, and necessary (given the constraints of available data), to make inferences on the basis of post-hoc reports, particularly when such inferences are informed by an established body of research findings – for example, identity theory (Haslam et al., 2015) or, in the present case, ethnomethodology. 6 Recording copies were purchased with U.S. National Science Foundation Grant #1103195. 7 Milgram ran 40 participants in each condition, with some exceptions: for example, Condition 24 (T and L have pre-existing family or friend relationship) placed only 20 participants in the T role (Russell, 2014b). According to Haslam, Loughnan et al. (2014), Haslam, Reicher et al. (2014) and Perry (2012), the total number of participants in all conditions was 780. Miller (1986) summarizes the experiment: Conditions 2 (L and T in adjacent rooms) and 3 (in same room) vary the physical/psychological proximity of T to L. Condition 20 uses women, whereas all other conditions use men. Condition 23 moves the setting from Yale to a business office in Bridgeport, CT. Normalizing trust 5 117, seven featured recording problems with the interview (was not recorded, inaudible). After identifying a further 19 cases in which participants provided no account for experimental behaviour, we had 91 interviews featuring at least one account: 36 (of 40 total) participants from Condition 2, 17 (of 40) from Condition 3, 30 (of 40) from Condition 20, 6 (of 40) from Condition 23, and 2 (of 20) from Condition 24. To study accounting practices, we listened carefully at least once to each participant’s entire experimental performance, and to the entire recording of each interview and debriefing. During this process, we made detailed notes on each interview, paying particular attention to accounts. We then listened again to the entire recording of each of the 91 interviews featuring at least one account. During this process, we edited our notes and refined our criteria for distinguishing account types.8 Our analytic strategy was thus, first, to assemble a collection of as large a number of instances as possible of the same interactional phenomenon – participants’ post-hoc accounts for experimental behaviour. And second, to study the collection to gain insight into the interactional practice’s general features (Maynard, 2013, p. 12). This research fills an important gap by contributing insight into an area of the Yale archive largely overlooked by the Milgram literature: the hundreds of post-experimental interviews preserved on Milgram’s audio recordings. Milgram’s interview situation: Candid or self-exculpatory accounts? The relationship of behaviour to post-hoc accounting for that behaviour is a classic topic in the social sciences. Historians of the Holocaust, for instance, when examining perpetrators’ post-war testimony, rightly treat such material with circumspection, acutely aware of the possibility that it is self-serving rather than sincere (e.g., Hoess, 1996; for the Soviet GULAG, see Mochulsky, 2010). Even when the behaviour of interest is not ideologically or morally charged, and the time lapse between behaviour and account is short, people may nevertheless consciously or non-consciously report about prior behaviour in unreliable ways. Some scholars have recently raised this issue in the context of the Milgram Experiment.9 For instance, Russell and Gregory (2011) argue that most ‘obedient’ participants knew they were engaged in wrongdoing, but that (after 150 V) it was psychologically easier to continue than to stop. When members of this group later seek to explain themselves, their justifications may be consciously self-serving, akin in this respect to Holocaust perpetrator testimony. After prolonged study of Milgram’s recordings, it seems to us that the interview accounts of ‘obedient’ participants should not simply be dismissed as defensive or evasive.10 It is not obvious that many of them clearly sensed that what they had done, in a highly ambiguous situation, was wrong (pace Russell & Gregory, 2011). Rather, it seems more accurate to say that as they conversed with E they were sincerely struggling, to a greater or lesser extent, to make sense of what had happened since entering the laboratory some thirty minutes previously. After all, the interviewer is almost always E himself (occasionally it is Milgram or his assistant, Elms), the very man who – a few minutes back – had been insisting they continue. They are alone with him and, unlike us, do not know the interview is being recorded. Milgram’s interview situation thus stands in marked contrast 8 See below for criteria. Cf. Orne and Holland (1968) and Rosenthal and Rosnow (1969) on ‘demand characteristics’: Social psychological participants may produce the kind of behaviour they think is expected of them. For Milgram’s response, see Milgram (1972). 10 For example, in the ‘pie of responsibility’ question 7 (see below for interview schedule), asking about proportional ‘shares of responsibility’ for L being shocked against his will, few ‘obedient’ participants assign themselves minimal blame. 9 6 Matthew M. Hollander and Jason Turowetz to paradigmatic Holocaust testimony, in which perpetrators – often in situations of confinement and trial, knowing some type of punishment was likely – were interrogated and cross-examined by their captors, representatives of Allied powers or of post-war liberal states such as West Germany (e.g., Sobib or trial, 1965–1966) and Israel (Adolf Eichmann trial, 1961). We therefore think it reasonable to take participants’ selfjustifications seriously, while remaining aware of the possibility that not all are being perfectly candid. Participants’ post-hoc explanations In our collection, participants justify themselves with seven distinct types of account. Four of these account for continuation, and are more frequently used by ‘obedient’ than by ‘defiant’ participants. Though defiant, members of the latter group commonly voice explanations for why they continued as far as they did before stopping.11 The remaining three account for discontinuation and are used by the defiant group (a few obedient participants also use them, to account for reluctance to continue). In this section, we first discuss the structure of the interview itself. Then, we present the seven accounts, showing how the ‘obedient’ and ‘defiant’ groups differ in their use. Since this paper’s concern is primarily with ‘obedience’ rather than ‘defiance’, we focus on the accounts for continuation rather than discontinuation. E performed the interview using the following series of questions and statements composed by Milgram (Table 1).12 Interviews were highly structured; nevertheless, E allowed participants ample time to voice reactions to the experiment. The interviews therefore ranged in length, with many reaching over 20 min. The major differences between interviews in Conditions 20, 23 and 24 vs. Conditions 2 and 3 were that component 7 was dropped, and component 8 replaced 12; that is, the women in Condition 20 were the first to be told that L had not actually received any shocks (Perry, 2013).13 E often directly asked interviewees why they continued or stopped. However, participants also frequently voiced justifications while answering E’s other questions.14 Across both contexts (solicited vs. volunteered accounts), participants used seven types of explanation.15 Their accounts for continuation appear in Table 2. The most common ‘obedient’ explanation is L not being harmed, as 33 (of 46 total) ‘obedient’-outcome participants used it at least once (72%). Moreover, as our collection includes all 40 interviews from Condition 2 (Voice-feedback; four participants provided 11 With ‘defiant’ participants, E tells them they continued for at least one shock after L indicated he wanted to be released. Their accounts for continuation are thus for shocking L at least once against his will. 12 Table 1 is our reproduction of the series of questions Milgram instructed the actor playing E, John Williams, to ask, based on what we hear in the audio recordings. It is not a paper document in the Milgram archive. Occasionally, Milgram or Alan Elms, his psychology graduate student assistant, conducted the interview. 13 As E explained, for example, to subject 2006 (disobedient): ‘We’re giving much more elaborate uh dehoax, if you will, to uh the women because we don’t want them going out of here panicked, you see. . . .Just the women. The men were never told’. Nevertheless, the men in the later Conditions 23 and 24 received the same, fuller debriefing as the women in Condition 20. 14 Accounts often appear in interview components 1, 2, 7, and 8, but also in other components. For reasons of space and scope, this paper refrains from more detailed discussion of accounts’ interactional context, and of many organizational features of talk-ininteraction displayed by the data excerpts below. Our primary goals are to document forms of participants’ accounts and discuss implications for engaged followership theory. 15 Other account types are rare (used by no more than 3 of the 91 T’s), and are not included in the tables. For example, inability to stop due to E’s pressure: 2035 claimed to be fearful of repercussions from E (obedient, 27:36): ‘I thought you’d h(h)it me if I st (h)opped! ((laugh))’. Normalizing trust 7 Table 1. Reconstruction of post-experiment interview schedule (1) (2) (2a) ‘Explain to me what happened and why it happened, treating me as if I didn’t know’. ‘Tell me, did this experiment have any other purpose that you could think of?’ If ‘defiant’ outcome: ‘Do you recall when this experiment was supposed to come to an end? What then do you propose I tell my supervisor, since we did not end when we were supposed to?’ (3) Schoolroom picture featuring angry teacher, shamefaced boy, and onlooker. ‘Make up a story about the picture. Who is the most likeable figure, and least likeable?’ (4) ‘Indicate with an x on this scale how painful was the sample shock you received’. (5) ‘Indicate with an x how painful to the Learner were the highest shocks he received’. (6) ‘Were you at all tense or nervous during the experiment? At the point when you were most tense or nervous, how tense or nervous were you?’ (7) Pie of responsibility: ‘With the use of this pie, indicate the proportional shares of responsibility, moral responsibility, that we three had for this man being shocked against his will’. (8) Debriefing [Conditions 20, 23 and 24]: ‘Actually this man was not really being shocked’. L enters room and restores relations with T. ‘We were actually interested in how you would behave’. (9) ‘All things considered, how do you feel about having participated in the experiment?’ Likert scale ranging from very sorry to very glad. (10) ‘What is your name? your age? your occupation? your marital status?’ (11) ‘Fill out this questionnaire’. Questions 6–18: more demographic questions including education, military service, political party affiliation, etc. (12) Debriefing [Conditions 2 and 3]: to ensure participants leave in a good frame of mind. ‘Before you leave I’d like to tell both you gentlemen [T and L] that this machine has previously only been used to shock small animals like rats. Learner, the amperage is adjusted so the shocks you received were only slightly more painful than the sample shock’. Table 2. Accounts for continuation overall (91 accounting participants) ‘obedient’ outcome (n = 46) ‘defiant’ outcome (n = 45) Total Following instructions L not being harmed Importance of experiment Contract Total 27 16 43 33 5 38 11 4 15 5 2 7 76 27 103 Interviewees may provide multiple types of account. Thus, the total of Tables 2 and 4, 167 (103 + 64), does not equal 91, the total number of interviews. Also, if a participant provided the same type of account more than once, we only counted it once. no account), we can report the frequency distribution for this condition by itself. These results are noteworthy due to the prominent role that Condition 2 has played in the Milgram literature, as shown in Table 3. Condition 2’s distribution is quite similar to that of the collection overall. Again, L not harmed appears more frequently than does any other account in the ‘obedient’-outcome interviews. Fifteen of 22 total (68%) ‘obedient’ interviewees use it at least once. We also find three accounts for discontinuation (or for reluctance to continue, for ‘obedient’ participants) principally used by the ‘defiant’ group as shown in Table 4. These participants portrayed themselves as unwilling to continue (e.g., to risk harming L; for religious reasons; against L’s will; because L was suffering; to avoid responsibility for harming L). Or as unable to continue (against L’s will; due to 8 Matthew M. Hollander and Jason Turowetz Table 3. Accounts for continuation in Condition 2 (36 accounting participants) Following instructions L not being harmed Importance of experiment Contract Total 12 5 17 15 4 19 6 2 8 4 1 5 37 12 49 ‘obedient’ outcome (n = 22) ‘defiant’ outcome (n = 14) Total Table 4. Accounts for discontinuation overall (91 accounting participants) Faulty design Unwilling to continue Unable to continue Total 9 17 26 0 25 25 0 13 13 9 55 64 ‘obedient’ outcome (n = 46) ‘defiant’ outcome (n = 45) Total nervousness; unable to make L suffer). Or they referred to the experiment’s faulty design (criticizing the experiment as ineffective for studying learning). Having documented these three accounts for discontinuation, we now resume our focus on the four accounts for continuation and our criteria for distinguishing them. Following instructions Participants may justify compliance by pointing out that E kept telling them to continue, and they merely followed his ‘instructions’ (‘orders’, ‘commands’, ‘requests’).16 (1) [0322, obedient, 55:38, component 7]17 1. T: Uh:, (0.4) You you were uh (.) >In other words< I would’ve 2. 3. stopped if (anything ) (1.2) 4. E: But you didn’t.=You were- (.) you [↑did go on:. 5. T: [We:ll, 6. T: I go- I went ↑on. 7. (0.5) 8. T: Because I was ↑told to go on. L was not really being harmed With this account (discussed at length below), participants indicate they continued because they did not think the situation was dangerous, despite appearances to the contrary (e.g., L’s pain cries and demands for discontinuation, E’s insensitivity to L’s well- 16 In broad terms, the ‘following instructions’ account resonates with Milgram’s agentic state explanation for participants’ behaviour. To the extent that this is so, it may provide some support for that explanation. Indeed, among obedient participants, the ‘following instructions’ account (59%) is more common than the ‘importance of experiment’ account (24%) – although neither is as common as ‘L not being harmed’ (72%). 17 Excerpt heading refers to ‘0322, obedient’ (Participant 22, Condition 3, ‘obedient’ outcome), and ‘55:38, component 7’ (excerpt starts at 55 min and 38 s in the recording, in the context of the ‘pie of responsibility’ interview component; Table 1). See above for Milgram’s experimental conditions, and see Hepburn and Bolden (2013) for transcription conventions. Some excerpts contain more than one account type; the type under discussion appears in bold. Normalizing trust 9 being). Such participants treat E as sufficiently competent to have kept L from harm. They may distinguish ‘pain’ from ‘harm’ here, echoing E’s Special Prod 1 (‘Although the shocks may be painful, there is no permanent tissue damage, so please go on’). However, they give no indication of approving of the way E conducted the experiment, or of how it investigates learning, or of how he handled L’s dramatic resistance. Other than the sheer fact of continuation, participants show no sign of sympathizing with science or of identifying with E’s goals. (2) [0216, obedient, 56:49, component 7] 1 T: And uh:, (0.3) I felt that uh (1.3) when I saw that other (guy) 2 go out aheada me,=And then (0.5) Now: I figured (uh) geez there 3 must’ve been a hell of a lot ↑more that’ve been doin’ this same 4 ↑thing. 5 (0.5) 6 E: There ↑have been.=°(Many more.)°= 7 T: =And uh:, (0.7) ↑there:fore what I am doing (1.3) uh:: is ↑probably alright. 8 9 (.) 10 T: As long as ↑you: were over there to- askin’ me to continue. 11 (.) 12 E: ↑I: see. 13 T: Because I: (0.8) didn’t (0.3) (eh) take ya to be an inhuman 14 $↑monster.$=Eh [heh!=.h heh!=Heh! 15 E: [h(h)! (3) [0221, obedient, 41:29, component 7] 1 E: Why do we share equally in in this man receiving (.) °uh shocks against his will.°=In the responsibility for this man? 2 3 4 (1.3) T: ↑Well I was (0.8) $>I(h) was just doin’ it$< cuz you ↑said 5 so?=And uh:=You heard him ↑yelling,=You coulda ↑sto:pped.=You 6 coulda stopped ↑me. 7 (0.6) 8 E: (Well) you could[ve stopped ( 9 T: 10 ) [I figured if it wasn’t ↑serious? (0.9) 11 T: If it was ↑THAT serious you woulda ↑stopped me. 12 (.) 13 T: (°D’ya see that?°) Importance of the experiment In contrast, some participants do orient to the importance of the experiment, or of science more generally. Although continuation may have been difficult, they indicate that the experiment was worthwhile or valuable in some way, even if they do not clearly understand its purpose. In so doing, they may present themselves as engaged followers of E’s leadership (Reicher et al., 2012). Due to this account’s relevance to engaged followership theory, we construe the category very broadly. For instance, we include any account that explicitly avows scientific values: 10 Matthew M. Hollander and Jason Turowetz (4) [2024, obedient, 26:42, component 8] 1 E: ((E has dropped cover story)) Why ↑did you do it. 2 3 (1.2) T: I don’t ↑know:.=As I said because: (0.2) I mean this is- (0.4) I 4 felt- (0.4) >In the back of the head< this is research and this 5 is important?=An’ .h= 6 E: =Mhm 7 8 (0.8) T: And you were running the show, 9 (0.3) We also include accounts in which T’s orientation to the importance of the experiment or scientific values is more implicit: (5) [0201, obedient, 53:02, component 5] 1 M: ((Milgram conducts interview)) Why did you listen to that man 2 ((E)) rather than (.) the man who was getting hurt. 3 4 (0.9) T: Apparently he knows what he’s ↑do:ing, 5 6 (0.7) M: Well the man in there knows that he’s getting hurt. 7 (7.8) 8 T: Well ↑I was uh: $right in th(h)e ↑m(h)iddle$=I didn’t know which 9 way to (.) turn.=°I:° (5.2) I felt sorry for the man out↑side, 10 an’ (2.6) I thought the experiment (.) uh:: (0.5) depended on my 11 going a↑head. 12 (0.3) 13 M: Mhm. And we include accounts indicating that T believed E had an important reason for insisting on continuation, even if T did not know or understand the reason: (6) [0207, obedient, 43:11, component 2] 1 E: And yet you continued in sp[ite of his uh [protest. 2 T: 3 [I (continued) [( ) (.) 4 E: [((clears throat)) 5 T: [Yeah 6 T: Well,=Eh the way I feel about it=I come here,=An’ yer $payin’ me 7 the m(h)oney,$ .h=For ↑my time, an’- (1.4) ((sigh)) There musta 8 been ↑some reason, (.) for ↑you: to want me to continue. 9 (1.4) However, we exclude accounts in which T’s displayed attitude towards the experiment or E is neutral.18 Despite our inclusiveness, importance of experiment appears far less frequently than the first two accounts (Table 2). We find that, in word searches of our detailed notes of both ‘obedient’ and ‘defiant’ interviews, relevant words are rare: only six (of 91 total) participants mention ‘science’ or ‘scientific’; only four 18 For example, we categorized 0209 (obedient, 38:44) as following instructions: E: ‘Why did you proceed to the end’. T: ‘Because I wasn’t in charge of the experiment’. Normalizing trust 11 mention ‘help’, ‘helpful’, or ‘cooperative’; three ‘value’ or ‘valuable’; one ‘important’ or ‘importance’; and one ‘research’.19 Fulfilling a contract Finally, some participants indicate they continued out of a sense of contractual obligation towards E. For instance, consider T’s account in Excerpt 6 above: “I come here, =An’ yer $payin’ me the m(h)oney,$ .h=For ↑my time,.” Normalizing trust Participation in Milgram’s experiments was undoubtedly a confusing and often stressful experience. Action centred on T delivering putatively real electroshocks while E calmly directed continuation in the face of L’s increasingly vociferous resistance. This section examines in more detail L was not really being harmed, the account for continuation most frequently used by our ‘obedient’ participants (72%) to explain how they made sense of this ambiguity – what E, in the interviews, repeatedly calls a ‘seemingly senseless situation’. After briefly developing our theoretical discussion of everyday sense-making practices, we show how these are exemplified by participants’ claims that, despite appearances to the contrary, the shocks posed no danger to L. As noted above, commonsensical practices for ‘doing being ordinary’ (Sacks, 1984) embody ‘trust’ (Garfinkel, 1963), action in accord with the presumption that others will cooperate in co-producing the social objects constitutive of our shared reality. Via trust, members of society sustain the ‘natural attitude’ of everyday life (Sch€ utz, 1970). ‘For the person conducting his [sic] everyday affairs, objects, for him as he expects for others, are as they appear to be’ (Garfinkel, 1967, p. 50). Our compliance with trust conditions, then, is a prerequisite for constituting the ‘natural facts of life’. And since, for society’s members, these facts are ‘through and through moral facts of life’ (Garfinkel, 1967, p. 35), their ongoing production is not simply a practical imperative, but a moral one. As Garfinkel demonstrated in his classic breaching tutorials (Garfinkel, 1963, 1967, p. 38), violations of trust produce confusion, distress, and anger. Faced with a breach, members of a shared culture work to normalize the situation, often in emotionally laden ways. For instance, if you tell people about your UFO abduction, they will hear you as ‘making a joke’; only if such accounts fail (you persist, or UFOs take over the world) will they resort to ‘serious’ explanations, such as mental illness or the reality of UFOs. To put the point differently, when facing a new situation, we make sense of it in terms of what we are already familiar with. New situations are assimilated to familiar situations, populated with familiar types of people and things acting in expectable ways. If such sense-making practices are put to the test, we first try to assimilate what is happening to the ‘ordinary’, despite significant appearances to the contrary. It is only when ordinary accounts fail us that we come to experience the situation as radically different than what we at first presumed. Milgram’s participants initially took for granted that the situation was ‘normal’. E was who he appeared to be; the experiment was a benign study of a typical psychological topic, learning and memory; and appropriate action would be taken if anything untoward happened. However, the situation unfolded in unexpected ways that challenged these 19 Moreover, sometimes it was the same participant using more than one of these words, and sometimes not an obedient but a defiant one. Word counts exclude a small number of instances of the word clearly deprecating the experiment’s importance. e.g., 2425 (defiant, 30:35): T: ‘It may be important to you [E] but it ain’t important for me’. 12 Matthew M. Hollander and Jason Turowetz presuppositions. Although participants’ numerous post-experimental claims that L was not really being harmed all worked to make sense of the ambiguous situation, they did so in different ways. Nevertheless, in none of them did participants display an orientation to the importance of the experiment – by sympathizing with science, identifying with E’s goals, or approving of E’s management of L’s resistance or of how the experiment studied learning. Below, we analyse thematic differences among the L not really being harmed accounts, which normalize the situation by (1) trusting E’s judgement, (2) treating L as overreacting, and (3) doubting the cover story. Though contrasting in theme, these accounts all justify T’s continuation in terms of the perception that L faced no serious danger. Trusting E’s judgement Some participants accounted for continuation by claiming they had relied on E’s judgement as to whether the experiment should continue. As 0223 (obedient, 48:10) put it, ‘I am trusting your judgment to know when to stop’. As a presumably competent scientist with ordinary (benign) motives and character, E was expected to be monitoring L’s well-being. Participants reasoned, commonsensically, that had L ‘really’ been in danger E would have stopped the experiment. Consider the following reports: (7) [0204, obedient, 51:44, component 5] 1 T: °Uh° (0.3) He ((L)) acted as if they were: uh (0.6) extremely 2 (0.3) painful. 3 4 (0.7) T: I assume they’re only ↑moderately painful. 5 6 (.) E: Why do you assume they’re only moderately painful. 7 8 (0.6) T: Well:,=I assume this because- ((swallow)) >t there was no 9 hesitation upon your part< that we con↑tinue, (0.5) And because 10 also he was in .h (0.4) the roo:m without anybody being w:ith 11 him. 12 (0.3) 13 E: Mhm, 14 (0.2) 15 T: >And I knew that< uh I couldn’t see him, (0.7) An::d uh if there 16 was any question that this would affect his heart?=Or that he 17 would faint? (0.6) You wouldn’t have (0.4) allowed me to go o:n. 18 (0.4) 19 E: I see. 20 T: So I assume for ↑that reason that it was only ↑moderately 21 painful. (8) [0218, obedient, 40:08, component 6] 1 E: Why did you then.=Why did you [go on. 2 T: 3 4 5 [.hh (.) T: Well,=I: used discretion in the sense that I didn’t think that (0.7) eh::: (.) that this uh .hh experiment would be:: as you 6 said in the beginning nothing would- actually ↑come of it,=It 7 wouldn’t- leave the man .hh (.) with any extreme (.) e↑ffects of 8 it.=So [I (felt) Normalizing trust 9 E: [There would be no permanent [tissue damage. 10 T: 11 13 [No permanent tissue.=So I thought (.) they were just sudden shocks, (9) [2004, obedient, 27:55, component 6] 1 E: $Why did you think$ it was funny ((T’s name deleted)) 2 T: Eh heh heh heh! .h (.) $I don’t know.$=I knew it wouldn’t ↑hur:t 3 4 5 6 7 8 9 him. (.) E: Uh huh, (0.4) T: ↑That was one thing.=That I knew it wouldn’t hurt him. (0.8) E: Well they’re not dangerous,=No:.= Because I knew you ↑wouldn’t give it to us if it was. 10 T: =No:.=B These participants normalized the experiment’s ambiguity by highlighting E’s professional judgement and downplaying the importance of L’s resistance. In Excerpt 7, participant 0204 notes that E displayed ‘no hesitation. . .that we continue’ (lines 8–9) and that L was alone and unmonitored in the other room (lines 10–11). Based on such observations, he claims that the shocks must not have been dangerously painful. Likewise, in Excerpt 8, participant 0218 relies on E’s judgement (‘as you said in the beginning’: lines 5–6) that L is not in danger. And in Excerpt 9, 2004 claims to know that the shocks were not harmful, explicitly basing this judgement on the conviction that E would not have exposed participants to danger. Note that participants 0204 (line 17) and 2004 (line 10) use reasoning-by-hypotheticals in accounting for their situated judgements. If L had been in danger, E ‘wouldn’t’ have allowed the experiment to go on, or have performed it in the first place. From the facts that E did perform the experiment and allow it to continue in the face of L’s resistance, it follows that L was not ‘really’ in danger. This commonsensical argument depends, of course, on the belief that E’s professional judgement guarantees his participants’ wellbeing. Relatedly, participant 0218 expresses trust in E by claiming to ground his own account for continuation in something E had said. He interrupts himself (‘experiment would be. . .’: line 5) to insert a report of E’s speech (‘. . .as you said’: lines 5–6). This interactional move, which conversation analysts call an insertion repair (Schegloff, 2013), shows T orienting to what E said as grounds for his reasoning; that is, this participant’s use of ‘discretion’ (line 4) is largely grounded in E’s perceived trustworthiness. Treating L as overreacting Other participants, in claiming that L was not really being harmed, asserted that L’s distress indicated not harm but rather overreaction to the shocks. According to these accounts, something about L made him overly sensitive to shocks that were, ‘in reality’, only moderately painful. (10) [0203, obedient, 49:26, component 5] 1 ((Elms conducts interview in role of E’s 00 assistant0 0 )) 2 T: And I imagine that (.) it may nos- not necessarily have been 3 all: ↑painful except maybe somewhat- in his ((L’s)) mi:nd (.) 4 thoughts. 14 Matthew M. Hollander and Jason Turowetz (11) [0215, obedient, 1:01:30, component 12] 1 L: [( 2 T: [((to L)) In other words you were already building and (.) and ) 3 an:- an↑ticipating something greater (0.5) that what you were 4 ac- actually receiving.= 5 E: =Well actually he did.=He: he started I guess to anticipate 6 7 mo[re ( L: ) [↑Ah:=I was just nervous. These participants suggested that L experienced the shocks as more painful than they ‘really’ were. 0203 and 0215 identify L’s mental state as the source of the trouble: He was anxiously ‘building and anticipating’ (Ex. 11, lines 2–3) something worse than he received, or the pain was simply ‘in his mind’ (Ex. 10, line 3).20 Doubting the cover story Finally, a handful of participants explained themselves via L was not being harmed because they doubted E’s cover story. Although Milgram went to great lengths to maximize experimental realism, he was not entirely successful with every participant. Clearly, the gestalt of the situation was altered for strong doubters, and with it the implications of their actions. Like many other interviewees, they justified themselves to E in terms of L not being harmed; however, the sense they made of E’s ‘seemingly senseless situation’ differed from those who displayed no doubt. Nevertheless, doubting was not an all-or-nothing affair of total credence versus total scepticism.21 Rather, of those who displayed suspicion of some aspect(s) of the cover story (often near the beginning of the interview), nearly all believed other aspects: That L had indeed received the lower-level shocks, that the experiment’s purpose was indeed to study learning and memory through punishment, that L’s cries were spontaneous and not pre-recorded, or that E was in fact a research scientist. In cases of scepticism, E defended the cover story or, by sidestepping the issue, avoided confirmation of doubts.22 These strategies usually succeeded, it seems to us, in preventing doubts from crystalizing into hardened certitude of total deception. Ex. 12 below illustrates both E’s tactics for handling doubt and participants’ practices for making sense of situational ambiguity. E had previously asked T to assess how painful to L were the highest shocks he received (interview component 5). At 44:34, T prefaces his suspicion that L had not actually received shocks with the ‘honesty phrase’ (Edwards & Fasulo, 2006) ‘Well, to be extremely candid’, suggesting that the forthcoming response is contrary to appearances or expectations.23 Several minutes later, E reasserts the cover story: 20 Similarly, a female participant (2030, obedient, ff. 19:28) proposed L’s male gender as a reason he overreacted to the shocks. She argued that women are more capable than men of withstanding pain, using as an example childbirth. 21 That is, when T’s express doubt in the cover story, it is typically far from clear how certain they are of deception. For this reason, we do not exclude doubters from our collection of 91 interviews. 22 Moreover, in doubting the cover story, participants displayed trust in the benign character of the experiment – that is, they assumed that E would not really allow L to get hurt, even if this meant constructing a fictitious cover story (which, incidentally, is not uncommon in ordinary experiments). 23 Cf. 0205 (obedient, 46:40): T: ‘Well to be perfectly frank I don’t believe the intensity of the shock (.) uh:: of any of those was any greater than the one I received’. Here, the prefacing honesty phrase is ‘to be perfectly frank’. Normalizing trust 15 (12) [0225, obedient, 47:20, component 5] 1. E: Well he actually was receiving shocks.=Now you uh (0.7) 2. indicated to me I (.) I believe uh: (0.6) two: or three 3. times=At least ↑once that you did not want to con↑tinue 4. 5. (0.7) E: If you (.) believed that he was not getting shocked °uh° why 6. did you continue 7. 8. (1.7) E: Or if you (re-) if you believe that uh (1.5) °uh:° (0.3) these 9. shocks were not painful or:: (0.3) whatever 10. (1.5) 11. T: Why did I wanna stop, 12. E: Yeah.=Why did you: uh: (0.7) wanna stop at that point 13. (2.6) 14. T: Um:, (1.6) Well uh 15. (0.3) 16. E: (And//I mean) I don’t understand the ↑reason.= 17. T: =>And [I I-< There’s a sort of a threshold.= 18. E: [(I don’t) 19. E: =((clears throat)) 20. (0.8) 21. T: Or a twilight zone where uh (0.8) uh: (1.4) although I- I felt 22. >I had a< good- (0.3) idea that that per↑haps he (wasn’t)=> 23. perhaps he wasn’t receiving the shock. (0.6) Uh (3.2) That 24. there was a bare possibility. (0.2) That he was. (0.7) An:d 25. (.) if there was no (0.6) physical response from him (in a way 26. of) (0.6) just leaving the roo:m,=Or pulling out the wires or 27. what have you, (0.9) and then (0.4) on your command I was to 28. continue I would go ahead until I had passed that threshold. 29. (0.6) of: (0.2) doubt 30. (0.4) 31. E: Hm: 32. (0.5) 33. T: At a certain point,=When >I dunno maybe about three hundred 34. an’ (1.2) some uh (0.7) volts °I was° (0.4) pretty contin (0.3) 35. convinced that uh (0.9) he wasn’t receiving anything ↑th:at 36. (1.1) strong:. (0.7) Or I would be getting a (3.2) uh::: (4.7) 37. different type of response.=A more uh (0.6) uh::: (1.2) more 38. violent (0.4) response. Participant 0225 orients to the discrepancy between his expectation of how an ordinary person would respond to shocks and L’s response. As people would normally react in a ‘more violent’ way than L did to 300 volts of electricity, he reasons that this feature of the cover story is false. In contrast, another participant accounting for continuation with L not really being harmed (2304, obedient, 23:07) claims that ‘when he [L] stopped hollerin’ I didn’t know what happened. . . I just figured that somebody had let him out’. Both 0225 and 2304 are normalizing the ambiguity of L’s reactions to the shocks. But whereas 0225 decides that L is not actually receiving shocks – and thereby calls into question the cover story – 2304 arrives at the less sceptical conclusion that L has been released. Discussion: Implications for engaged followership Our empirical ‘. . .structured analysis of the experiences reported by Milgram’s participants’ (Haslam et al., 2015, p. 8) has important implications for the engaged followership 16 Matthew M. Hollander and Jason Turowetz model of Milgramesque behaviour. Engaged followership is rightfully one of the leading contemporary Milgram theories, with substantial plausibility and empirical support. However, our extensive study of Milgram’s experimental sessions and post-experimental interviews has led us to dissatisfaction with it in its current version. The Milgram literature has always faced a ‘black box’ problem: wanting to see the phenomenon of compliance comprehensively explained, but having relatively little access to published, primary data sources. As more researchers open the black box of the Milgram archive, theorizing becomes increasingly grounded in the relevant primary data. Here, we discuss two of our findings seemingly at odds with engaged followership – call them ‘engagement’ and ‘debriefing’. Engagement As we have seen, this paper offers little empirical support for the claim that participants generally understood themselves as engaged followers of E’s leadership. In our collection, only 24% (11/46) of ‘obedient’ participants used importance of the experiment to explain continuation. In Condition 2 by itself, this figure is only slightly higher (6/22, or 27%). In contrast, much larger proportions of ‘obedient’ participants explained themselves in terms of L not really being harmed (33/46 = 72%) or following instructions (27/ 46 = 59%); that is, while some participants certainly do seem to have identified with the experiment, the majority do not appear to have continued out of commitment to science. One challenge for engaged followership is to explain why so many participants in our collection displayed low engagement. Debriefing It is apparent in listening to the interviews that the debriefing process especially the fuller debriefing provided to the participants in Conditions 20, 23 and 24 improved many of their attitudes about the value of the research. As we have seen, relatively few participants prior to debriefing displayed orientations to the experiment as important or worthwhile. More commonly, attitudes displayed in the interviews, towards E and the way the experiment used punishment to study learning, ranged from polite neutrality to disillusionment, alienation, and hostility. For instance: 0203 (obedient, 51:10): T: ‘. . .I might have had a tendency to stop when I heard the man not willing to go on. . . I felt there was no use possibly in going ahead. . .’ 2001 (obedient, 28:55): T: ‘To me it was punishment that wasn’t serving any purpose’. 2003 (obedient, 33:34): T: ‘I saw no reason why I should painfully shock this poor man. Just simply because he didn’t have the ability to put two words together’. 2006 (defiant, 17:10): T: ‘(I don’t belie:ve) in punishment, I’ll tell you that, I think this is: quite ineffective’. However, debriefing (‘Actually, this man was not really being shocked’) appears to have improved many participants’ views of the experiment’s importance. Most participants responded positively to E’s final interview question (‘All things considered, how do you now feel about having participated?’). In the light of the way debriefing improved attitudes, the challenge for engaged followership is to explain the relationship of positive attitudes after debriefing to more negative attitudes during the experiment and Normalizing trust 17 interview; that is, when Haslam et al. (2015) describe ‘. . .the sense of fulfillment that participants experienced from participating in a scientific study that they considered to be extremely important’ (p. 17), they are referring of course to debriefed participants, veterans of a taxing study of obedience to authority. But we are primarily interested in how participants viewed the experiment prior to debriefing, the putative study of learning and memory. After all, engaged followership is a theory about compliance itself, not behavioural or attitudinal proxies subsequent to the experiment (whether expressed immediately afterwards in the interview or several weeks later in the questionnaire). Engaged followership is currently a global or comprehensive theory of Milgramesque behaviour. To make sense of variation in that behaviour, it overestimates, we think, the identification (esprit de corps) with E that was actually experienced by many participants. A more adequate explanation would allow a middle ground between maximal ‘defiance’ (principled and firm rejection of continuation) and maximal ‘obedience’ (principled compliance owing to the experiment’s importance). Close examination of the interviews suggests that many, if not most, ‘obedient’-outcome participants were only minimally so: grudgingly willing to go through the motions of compliance, and afterwards displaying alienation from E and the experiment. After all, their pre-participation ideological commitment to, training in, or experience with experiments of this type (violent, performed against L’s will) is typically nil. There is, of course, an important difference between engaged followers willing to perform a difficult task in the service of values in which they believe, and compliant followers going through the motions of commitment. Although the Milgram outcome is the same (L suffers in either case), the behavioural explanation should take this difference into consideration. Otherwise, theorizing may overlook differences that participants claimed were important to understanding why they acted as they did. Conclusion This article has introduced ‘normalizing trust’ as an innovative means of gaining insight into the behaviour of many of Milgram’s participants.24 Our analysis uses a data source hitherto virtually untapped in the Milgram literature: the hundreds of immediately postexperimental in-depth interviews. In examining the accounts for their just-completed experimental behaviour produced by 91 of Milgram’s participants of both outcome categories, we have documented four recurrent accounts for continuation: following instructions, L was not really being harmed, importance of the experiment and fulfilling a contract. After discussing in detail how ‘obedient’-outcome participants use the L not being harmed account more frequently than any other explanation (33 of 46 participants, or 72%) to make sense of their just-prior actions, we drew on these findings to point out limitations of engaged followership theory in comprehensively explaining Milgramesque behaviour. Perhaps future scholarship will satisfactorily resolve the challenges, as we see them, that are raised for engaged followership by the interview data. However, we would like to end on a note of caution. It may well be that the goal of identifying a single psychological process – whether ‘obedience to authority’ or ‘agentic state’ (Milgram), or ‘engaged followership’ (Haslam and Reicher) – exhibited by Milgram’s experiments, and 24 We would like to emphasize that this paper has not proposed a comprehensive theory claiming to explain Milgramesque behaviour as such. 18 Matthew M. Hollander and Jason Turowetz responsible for causing his results, is illusory (Brannigan, Nicholson, & Cherry, 2015). As Staub (2014) avers, in Milgram’s laboratory ‘[i]t is unlikely that a single psychological process led to either continued compliance/cooperation, or refusal to cooperate. Varied processes can arise from the interaction of situation and personality’ (p. 508). And, in the context of the Holocaust, ‘. . .any analysis of why they [perpetrators] behaved the way they did must rely on a variety of explanations rather than a generic core’ (Overy, 2014, p. 515). The findings of this paper, showing that many of Milgram’s participants provided several kinds of account for continuation, likewise suggest that multiple social psychological processes (and/or organized interactional practices) may have produced Milgram’s ‘obedient’ outcome category: of which ‘normalizing trust’ is a particularly prominent one. We have provided compelling evidence of this complexity not only across our five experimental conditions, but also within particular conditions such as 2 (Voicefeedback). ‘Certainly, all the evidence suggests that Milgram’s participants were “happy to have been of service”’ (Haslam et al., 2015, p. 26). To the contrary, an essential and hitherto neglected body of evidence – the interviews – suggests that the puzzle of Milgramesque behaviour is more complex than engaged followership at present allows. References Baumrind, D. (1964). Some thoughts on ethics of research: After reading Milgram’s “behavioral study of obedience”. American Psychologist, 19, 421–423. https://doi.org/10.1037/h0040128 Blass, T. (2004). The man who shocked the world: The life and legacy of Stanley Milgram. New York, NY: Basic Books. Brannigan, A., Nicholson, I., & Cherry, F. (2015). Introduction to the special issue: Unplugging the Milgram machine. Theory & Psychology, 25, 551–563. https://doi.org/10.1177/09593 54315604408 Burger, J. M. (2009). Replicating Milgram: Would people still obey today? American Psychologist, 64, 1–11. https://doi.org/doi.org/10.1037/a0010932 Burger, J. M. (2014). Situational features in Milgram’s experiment that kept his participants shocking. Journal of Social Issues, 70, 489–500. https://doi.org/10.1111/josi.12073 Duck, W. (2015). No way out: Precarious living in the shadow of poverty. Chicago, IL: University of Chicago Press. Edwards, D., & Fasulo, A. (2006). ‘To be honest’: Sequential uses of honesty phrases in talk-ininteraction. Research on Language and Social Interaction, 39, 343–376. https://doi.org/10. 1207/s15327973rlsi3904_1 Garfinkel, H. (1963). A conception of and experiments with ‘trust’ as a condition of concerted stable actions. In O. J. Harvey (Ed.), Motivation and social interaction (pp. 381–392). New York, NY: Ronald. Garfinkel, H. (1967). Studies in ethnomethodology. Englewood Cliffs, NJ: Prentice-Hall. Gibson, S. (2013a). Milgram’s obedience experiments: A rhetorical analysis. British Journal of Social Psychology, 52, 290–309. https://doi.org/10.1111/j.2044-8309.2011.02070.x Gibson, S. (2013b). ‘The last possible resort’: A forgotten prod and the in situ standardization of Stanley Milgram’s voice-feedback condition. History of Psychology, 16, 177–194. https://doi. org/10.1037a0032430 Gibson, S. (2014). Discourse, defiance and rationality: ‘Knowledge work’ in the ‘obedience’ experiments. Journal of Social Issues, 70, 424–438. https://doi.org/10.1111/josi.12069 Goffman, E. (1959). The presentation of self in everyday life. New York, NY: Anchor. Haslam, N., Loughnan, S., & Perry, G. (2014). Meta-Milgram: An empirical synthesis of the obedience experiments. PLoS One, 9, e93927. https://doi.org/10.1371/journal.pone.0093927 Normalizing trust 19 Haslam, S. A., Reicher, S. D., & Birney, M. (2014). Nothing by mere authority: Evidence that in an experimental analogue of the Milgram paradigm participants are motivated not by orders but by appeals to science. Journal of Social Issues, 70, 473–488. https://doi.org/10.1111/josi.12072 Haslam, S. A., Reicher, S. D., Millard, K., & McDonald, R. (2015). ‘Happy to have been of service’: The Yale archive as a window into the engaged followership of participants in Milgram’s ‘obedience’ experiments. British Journal of Social Psychology, 54(1), 55–83. https://doi.org/10.1111/bjso. 12074 Hepburn, A., & Bolden, G. B. (2013). The conversation analytic approach to transcription. In J. Sidnell & T. Stivers (Eds.), The handbook of conversation analysis (pp. 57–76). New York, NY: Blackwell-Wiley. Hoess, R. (1996 [1946–1947]). Death dealer: The memoirs of the SS kommandant at Auschwitz. In S. Paskuly (Ed.), A. Pollinger (Trans.). New York, NY: Da Capo Press. Hollander, M. M. (2015). The repertoire of resistance: Non-compliance with directives in Milgram’s ‘obedience’ experiments. British Journal of Social Psychology, 54, 425–444. https://doi.org/ 10.1111/bjso.12099 Jefferson, G. (2004). ‘At first I thought’: A normalizing device for extraordinary events. In G. H. Lerner (Ed.), Conversation analysis: Studies from the first generation (pp. 131–167). Amsterdam, the Netherlands and Philadelphia, PA: John Benjamins. Kaplan, D. E. (1996). The Stanley Milgram papers: A case study on appraisal of and access to confidential data files. The American Archivist, 59, 288–297. https://doi.org/10.17723/aarc 59. 3.k3245057x1902078 Kaufmann, H. (1967). The price of obedience and the price of knowledge. American Psychologist, 22, 321–322. https://doi.org/10.1037/h0037664 Liberman, K. (2013). More studies in ethnomethodology. New York, NY: SUNY Press. Maynard, D. W. (2013). Everyone and no one to turn to: Intellectual roots and contexts for conversation analysis. In J. Sidnell & T. Stivers (Eds.), The handbook of conversation analysis (pp. 11–31). New York, NY: Blackwell-Wiley. Milgram, S. (1963). Behavioral study of obedience. Journal of Abnormal and Social Psychology, 67, 371–378. https://doi.org/10.1037/h0040525 Milgram, S. (1972). Interpreting obedience: Error and evidence: A reply to Orne and Holland. In A. G. Miller (Ed.), The social psychology of psychological research (pp. 138–154). New York, NY: The Free Press. Milgram, S. (1974). Obedience to authority: An experimental view. New York, NY: Harper & Row. Miller, A. G. (1986). The obedience experiments: A case study of controversy in social science. New York, NY: Praeger. Mochulsky, F. V. (2010). Gulag boss: A Soviet memoir. D. Kaple (Trans.). New York, NY: Oxford University Press. Nicholson, I. (2011). ‘Torture at Yale’: Experimental subjects, laboratory torment and the ‘rehabilitation’ of Milgram’s ‘obedience to authority’. Theory & Psychology, 21, 737–761. https://doi.org/10.1177/0959354311420199 Orne, M. T., & Holland, C. H. (1968). On the ecological validity of laboratory deceptions. The International Journal of Psychiatry, 6, 282–293. Overy, R. (2014). Ordinary men, extraordinary circumstances: Historians, social psychology, and the Holocaust. Journal of Social Issues, 70, 515–530. https://doi.org/10.1111/josi.12075 Perry, G. (2012). Beyond the shock machine: The untold story of the Milgram obedience experiments. Melbourne, Australia: Scribe. Perry, G. (2013). Deception and illusion in Milgram’s accounts of the obedience experiments. Theoretical & Applied Ethics, 2(2), 79–92. https://doi.org/10.1353/tha.2013.0018 Rawls, A. (2000). ‘Race’ as an interaction order phenomenon: W.E.B. Du Bois’ ‘double consciousness’ thesis revisited. Sociological Theory, 18, 241–274. https://doi.org 10.1111/ 0735-2751.00097 20 Matthew M. Hollander and Jason Turowetz Reicher, S. D., & Haslam, S. A. (2011). After shock? Towards a social identity explanation of the Milgram ‘obedience’ studies. British Journal of Social Psychology, 50, 163–169. https://doi. org/10.1111/j.2044-8309.2010.02015.x Reicher, S. D., Haslam, S. A., & Smith, J. R. (2012). Working toward the experimenter: Reconceptualizing obedience within the Milgram paradigm as identification-based followership. Perspectives on Psychological Science, 7, 315–324. https://doi.org/10.1177/ 1745691612448482 Reicher, S. D., & Hopkins, N. (2001). Self and nation: Categorization, contestation and mobilization. London, UK: Sage. Rochat, F., & Blass, T. (2014). Milgram’s unpublished obedience variation and its historical relevance. Journal of Social Issues, 70, 456–472. https://doi.org/10.1111/josi.12071 Rosenthal, R., & Rosnow, R. L. (Eds.) (1969). Artifact in behavioral research. New York, NY: Academic Press. Russell, N. (2011). Milgram’s obedience to authority experiments: Origins and early evolution. British Journal of Social Psychology, 50, 140–162. https://doi.org/10.1348/014466610 X492205 Russell, N. (2014a). The emergence of Milgram’s bureaucratic machine. Journal of Social Issues, 70, 409–423. https://doi.org/10.1111/josi.12068 Russell, N. (2014b). Stanley Milgram’s obedience to authority ‘relationship’ condition: Some methodological and theoretical implications. Social Sciences, 3, 194–214. https://doi.org/10. 3390/socsci3020194 Russell, N., & Gregory, R. J. (2011). Spinning an organizational ‘web of obligation’? Moral choice in Stanley Milgram’s ‘obedience to authority’ experiments. The American Review of Public Administration, 41, 495–518. https://doi.org/10.1177/0275074010384129 Sacks, H. (1984). On doing being ordinary. In J. M. Atkinson & J. Heritage (Eds.), Structures of social action: Studies in conversation analysis (pp. 413–429). Cambridge, UK: Cambridge University Press. Schegloff, E. A. (2013). Ten operations of self-initiated, same-turn repair. In M. Hayashi, G. Raymond & J. Sidnell (Eds.), Conversational repair and human understanding (pp. 41–70). Cambridge, UK : Cambridge University Press. Sch€ utz, A. (1970). On phenomenology and social relations. Chicago, IL: University of Chicago Press. Staub, E. (2014). Obeying, joining, following, resisting, and other processes in the Milgram studies, and in the Holocaust and other genocides: Situations, personality, and bystanders. Journal of Social Issues, 70, 501–514. https://doi.org/10.1111/josi.12074 Turowetz, J., Hollander, M. M., & Maynard, D. W. (2016). Ethnomethodology and social phenomenology. In S. Abrutyn (Ed.), Handbook of contemporary sociological theory (pp. 387–410). New York, NY: Springer International. Received 1 December 2016; revised version received 29 April 2017