Abstract
We introduce conversation analysis (CA) as a methodological innovation that contributes to studies of the classic Milgram experiment, one allowing for substantive advances in the social psychological “obedience to authority” paradigm. Data are 117 audio recordings of Milgram’s original experimental sessions. We discuss methodological features of CA and then show how CA allows for methodological advances in understanding the Milgramesque situation by treating it as a three-party interactional scene, explicating an interactional dilemma for the “Teacher” subjects, and decomposing categorical outcomes (obedience vs. defiance) into their concrete interactional routes. Substantively, we analyze two kinds of resistance to directives enacted by both obedient and defiant participants, who may orient to how continuation would be troublesome primarily for themselves (self-attentive resistance) or for the person receiving shocks (other-attentive resistance). Additionally, we find that defiant participants mobilize two other-attentive practices almost never used by obedient ones: Golden Rule accounts and “letting the Learner decide.”
Keywords
Despite the recent renaissance of interest in social psychologist Stanley Milgram’s classic “obedience to authority” experiments (Brannigan, Nicholson, and Cherry 2015; Burger 2009; Haslam et al. 2015; Russell 2011, 2014a), the role of resistance to directives in these groundbreaking studies remains unclear. 1 Starting with Milgram (1963) himself, the dominant interpretation of participants’ behavior has always been conformity to an authority figure’s inducements to continue. Operationalizing experimental behavior as either “obedient” or “defiant” (“disobedient”), Milgram believed he had demonstrated that situational variables under experimental control cause heightened or lowered rates of obedience of a research subject (the “Teacher”) to an authority (the confederate “Experimenter”) directing the electroshocking of a peer volunteer (the confederate “Learner”). The variables include proximity of the Learner to a Teacher, the number of Teachers and their hierarchical relationship, and the gender of a Teacher (Blass 2009; Milgram 1974). As Miller (2009:22) notes, however, “[i]ndividual differences within many of the experimental variations are one of the most striking features of [Milgram’s] results, but they are an exceedingly underreported and ignored feature as well.” That is, although two-thirds of participants across all conditions ended “obediently,” virtually all resisted continuation to some extent. And, of course, many (over 250) did so in ways that successfully stopped the experiment and earned Milgram’s “defiant” categorization.
In light of these facts, a crucial theoretical question arises about studying social behavior in Milgramesque situations, one that has received little attention in the Milgram literature, even of recent years (cf. Gibson 2013b). Given that Milgram’s findings have proven so difficult to satisfactorily interpret, what could be gained from making a fundamental shift in our questions about the experiments: bracketing for the moment explanations of why Milgramesque behavior happens in order to seek explanations of how it happens? This shift would require novel methodological means to address the empirical questions it raises. For instance, what social psychological dynamics allow some participants to heroically succeed (defiance) in a situation in which most others fail (obedience)? What features of resistance are common to both groups? How could successful resistance techniques be applied to real-world situations to challenge authoritarian abuses of power and empower potential victims? Answers to questions about the detailed organization of social interaction in the experiments, in turn, could prove essential to ongoing attempts to explain why Milgram found what he found.
Modigliani and Rochat (1995:120) were among the few to make this theoretical shift, albeit incompletely, with their important work on “the psychological and interpersonal dynamics that unravel over the course of encounters with an authority.” Applying content analysis to 34 transcribed audio recordings from Condition 23 (“Bridgeport”), they identified 6 types of increasingly strong participant resistance. Overall, they developed a top-down, phase-based, quantitative model whereby Teachers’ comments were scored according to the degree of resistance. The main variable they examined was the timing of resistance; the earlier in the shock series that firm resistance occurred, the more likely a subject was to end defiantly.
As pioneers, Modigliani and Rochat (1995) explored territory that our study reframes and then surveys more completely. In contrast to content analysis, our use of conversation analysis (CA) constitutes a methodological innovation in studies of Milgramesque behavior. Since CA is bottom-up, qualitative, and attentive to real-time actions and reactions, our analysis includes not only the behavior of the Teacher but also that of the Experimenter and Learner. We thus make the aforementioned theoretical shift by focusing on the detailed structure of Milgramesque interaction rather than simply on what the Teacher does. As Peräkylä (2004:10) has observed in contrasting CA with Bales’s interaction process analysis—similar in its coding and quantifying to content analysis—“Actions do not have meaning in themselves, but they become meaningful through their place in the self-organizing continuum of actions.”
We propose, then, a radical theoretical shift regarding Milgramesque behavior and use CA methodology to answer neglected empirical questions in the social psychology of “obedience to authority.” We thus demonstrate how CA can raise new phenomena for social psychology: here, the structuring of practices for resisting authoritarian directives. More specifically, the paper shows how both defiant—and obedient—outcome participants resist experimental progressivity by orienting to ways that continuation would be troublesome primarily for themselves, on the one hand, or the person receiving shocks, on the other. Before presenting our empirical analysis, however, we first provide further background on our theoretical and methodological innovation in the social psychology of “obedience to authority.”
Methodological Advance: CA and Interactional Organization in the Milgram Experiment
CA studies have a long history of representation in Social Psychology Quarterly (see e.g., Maynard 1987). Since their methodology, however, still may be unfamiliar to many SPQ readers, we outline four main features of the approach. We then specify this paper’s methodological contribution to the Milgram literature, suggesting the continuing importance of CA to social psychology.
First, CA views social interaction as an orderly domain in its own right—not in the first instance reflecting exogenous structures but having a participant-produced in situ organization (Maynard 2013). The systematics of turns of talk exemplify this; by displaying interpretations of another’s talk through this mechanism, speakers routinely achieve mutual understanding. Second, nonverbal gestures and speech details are important sites of orderly actions. Silences, overlapping talk, laughter, pitch, and other aspects of prosody can all contribute to the design and intelligibility of social actions. To capture these details, CA researchers use audio- or video-recordings of naturally occurring interaction in “everyday” and “institutional” settings. Third, CA views utterances as doing social actions but not as necessarily formulating them as such. When a telephone caller says, “I was trying you all day and your line’s been busy for like hours!,” the practice is one of “telling my side” as a way of “fishing” for an account (Pomerantz 1980). The caller is requesting an explanation without overtly asking, “Who were you talking to?” Speakers rarely identify their actions as such with a semantic verb form (“I request that you tell me who you were talking to”) or by the syntactic structure of an utterance. Rather, we use features of the immediate social context, especially an utterance’s place in an organized sequence of talk, to respond. Thus, for CA, participant-produced sequence organization is the primary resource for analyzing talk and social interaction. Fourth, responses to initiating actions (questions, directives, complaints, etc.) display indigenous organization in terms of “preference,” a structural feature of turn sequencing. For example, when speakers produce assessments—whether positive or negative—the structurally preferred response is agreement, whereas the dispreferred response is disagreement (Pomerantz 1984).
CA tools such as these can be used to develop new understandings of Milgramesque behavior. Specifically, our methodological advance has two components. First, we reframe Milgram’s description of experimental behavior (obedience to authority) in terms of directive-response sequences and complaint-remedy sequences. These two everyday types of sequence organization—studied in family settings (Craven and Potter 2010; Hepburn and Potter 2011) and in children’s games (Goodwin 1990)—have not previously been investigated in experimental situations. This is despite the fact that in Milgram’s sessions, these sequences enable and constrain much of what happens. The Experimenter directs the Teacher to deliver increasingly severe electroshocks to the putative Learner when the latter gives errant answers to word pairs he was supposed to memorize. The Learner in turn displays increasing agitation as he complains about the shocks and demands to be released. (2) We are thus able to rethink the dilemma Milgram created for the Teacher in terms of sequence organization: as the experiment progresses, the situation becomes increasingly organized in terms of mutually competitive action proposals. The structurally preferred response to an initiating directive is compliance (here, experimental continuation), whereas the preferred response to an initiating complaint is remedy (discontinuation). Caught in the middle of such opposing proposals for action, Teachers must align with one or the other; they cannot align with both. 2 Whether Teachers end as obedient or defiant depends on whether their alignments ultimately favor the sequential outcome made relevant by directives or by complaints.
The innovative methodological perspective we bring to Milgramesque behavior builds on the insights of important recent scholarship. Like studies using archival research (Perry 2012; Russell 2014b), experimentalism (Burger 2014), and meta-analysis (Haslam, Loughnan, and Perry 2014), we are revisiting social psychology’s single most famous project to shed new light on its situational context and challenge Milgram’s notion of obedience to authority. Unlike other studies, however, ours is especially concerned with concrete and detailed interactional practices for coping in Milgram’s lab. In this, our approach resembles Gibson’s (2014) important use of discursive psychology to analyze transcripts of participants who resisted the Experimenter. Gibson’s analysis, however, includes only defiant participants, and his methodology does not fully engage the sequential organization (Schegloff 2007) of the interactions. By contrast, our use of CA is fully comparative and sequential since we analyze the detailed and sequentially organized practices of both groups.
One such practice, hitherto unnoticed by Milgram scholars, is the object of this paper’s empirical analysis. In resisting the Experimenter’s directives to continue delivering shocks, participants may display self- or other-orientation. Resistance is self-attentive when Teachers egoistically attend to how continuation is difficult and troublesome primarily for themselves. Heritage (1998:313) characterizes self-attentiveness as a feature of talk that “focus[es] on how the world looks” from the perspective of speakers rather than from that of copresent interlocutors or absent third parties. Alternatively, resistance is other-attentive when Teachers altruistically orient to how continuation is problematic primarily for the Learner and thereby to how the world looks from his perspective. In sum, then, the paper uses innovative methodology to analyze self- versus other-attentive resistance in the Milgram experiment in local contexts of directive-response and complaint-remedy sequences. We advance our methodological imagination regarding the understanding and interpretation of Milgram’s findings, shed new light on what the obedient and defiant participants share, and report findings about what members of the latter group do to successfully stop the experiment.
Archival Data of Recorded Interactions
Data are 117 recordings from the Stanley Milgram Archive maintained by the Manuscripts and Archives Department at Yale University Library. These recordings, documenting the performances of 117 different subjects, were selected from five of the original experimental conditions: 2 and 3 (Voice-Feedback and Proximity, from the Proximity Series), 20 (Women as Subjects), 23 (Institutional Context, aka “Bridgeport”), and 24 (Relationship, featuring pairs of subjects with a preexisting relationship). 3 Milgram ran 40 subjects in each condition, with some exceptions (e.g., condition 24 placed only 20 subjects in the role of Teacher, as noted in Russell 2014b). The first author also transcribed two partial sessions from Milgram’s Obedience film (Milgram 1965), one obedient and one defiant. 4 To ensure that the transcripts did not omit any unusually early instances of resistance to directives, the first author listened carefully to the entire recording for each subject, starting with the Experimenter’s directive (“Ready, begin.”) to initiate the “practice lesson” preceding the “second lesson” (the experiment proper).
Owing to circumstances such as the archive’s incompleteness and a limited budget for acquisition and transcription, the first author primarily selected recordings from the relatively complete conditions 2, 3, and 20, where the recordings had received preparatory digitization and the removal of subjects’ names (Kaplan 1996). 5 The analytic strategy involved assembling a collection of as large a number of instances of the same interactional phenomenon as possible—here, directive-response sequences (Craven and Potter 2010; Ervin-Tripp 1976; Goodwin 2006) and instances of resistance to such directives. Working with such a collection affords insight into an interactional practice’s general features (Maynard 2013). To the best of our knowledge, this is the largest corpus of detailed transcripts of talk-in-interaction, in particular directive-response and complaint-remedy sequences, in Milgram’s experiments yet assembled.
CA Methodology and New Phenomena for Social Psychology: Self- and Other-Attentive Resistance
CA can raise new phenomena for social psychology. Our use of this methodology reveals that both obedient and defiant subjects can marshal self- and other-attentive resistance to the Experimenter’s directives to continue the experiment. Extensive analysis of the two styles of resistance yielded several initial findings. First, many participants from both groups use both styles. Second, in both groups, explicitly other-attentive instances are more common than self-attentive ones. Third, however, self-attentive resistance is actually twice as prevalent among the defiant (18/53, or 34 percent, display at least one instance) as the obedient (10/64 total instances, or 16 percent) subjects. This finding makes sense in light of the fact that defiant participants resist more than do obedient ones; they tend to perform more varied and more explicit forms of nonsilent resistance, and they repeat previously used forms of resistance (Hollander 2015). And fourth, although each group uses both self- and other-attentive forms of resistance, there are at least two kinds of other-attentive resistance used by defiant participants—what we analyze in the following as “Golden Rule accounts” and “letting the Learner decide”—that almost never appear in the obedient group.
Starting from these initial findings, this paper analyzes patterns of self- and other-attentive resistance first in the obedient-outcome cases and then in the defiant-outcome ones. At the outset, we wish to introduce a distinction between “positional” and “compositional” other-attentiveness, one that builds on Heritage’s (1998) analysis of self- and other-attentiveness in social interaction. Any form of resistance to the Experimenter’s directives is other-attentive in a positional (sequential) sense. That is, by orienting to problems with continuation due to the Learner’s (hereafter, “L’s”) reactions to receiving shocks or to the fact of his receiving shocks, resistance disrupts the progress of the experiment at least in part on his behalf. Proceeding to the next higher shock level is temporarily postponed; the respite represents a momentary benefit to the Learner. So when we discuss self- and other-attentiveness, this sense of positional other-attentiveness is assumed while we focus on self- and other-attentiveness in the compositional design of the Teacher’s (“T’s”) talk.
Developing this point, it is common for there to be positional other-attentiveness—that is, resistance to the Experimenter’s (“E’s”) directives—but no display of self- or other-orientation in the compositional design of T’s utterances. This point also addresses our methodological concerns to explicate the Milgramesque setting as one of three-party interaction with an interactional dilemma for T. In example 1, T announces the shock level (line 1), then administers it (line 3):
(1) [0219 disobed, 135V, 26] 6
At line 5, L issues a “pain cry” (cf. Goffman 1981 on “response cries”). This is a form of complaining and as such requires remedy. T, however, is also operating under the auspices of the experimental mandate to follow protocol and therefore to provide a correct answer and continue with further items. But T remains silent for over two seconds (line 6) and thus deals with the pain cry by this momentary delay in continuation. 7 He nevertheless shows his orientation to the experimental mandate by addressing E and asking for confirmation (line 7), which suggests an orientation to E’s “epistemic status” (Heritage 2012) or domain of knowledge regarding what should occur next and to E’s “deontic” authority (Stevanovic and Peräkylä 2012) or right to determine the Teacher’s next action. E, after clearing his throat (lines 8–9), provides a confirming response at line 10, directing T to continue, which T then does after some delay (lines 11–13). In sum, T’s query at line 7 orients to L’s negative reaction to receiving shocks as problematic for continuation and for that reason is resistive and other-oriented in a positional sense. But in terms of its compositional design, the query does not exhibit self- or other-orientation with an explicit demonstration of how the world looks from either perspective.
CA Methodology and Obedient Participants: Self- and Other-Attentive Resistance
Our use of CA methodology shows that obedient as well as disobedient participants use both self- and other-attentive forms of resistance in their talk. In this section, pursuing another of our methodological aims, we show how obedient subjects use such forms and discuss what happens in the subsequent interaction. That is, we will decompose Milgram’s categories of obedience and defiance into routes by which these outcomes are achieved through concrete displays of resistance.
Self-Attentive Resistance among Obedient Participants
By self-attentive resistance, we mean that participants provide an ego-involved, “my world” reason for postponing experimental continuation or projecting discontinuation. The practice typically accounts for their just prior resistance, which is often a stop try (an utterance that explicitly projects imminent experimental discontinuation). In the following example, L has been refusing to answer the questions, and E has directed T to treat no response as a wrong answer to be punished with the next shock. At line 1, T reads the next word list:
(2) [0216 obed, 315V, 390]
As indicated by the long silence (a strong positional type of resistance) at line 2, no response is forthcoming from L. T then sighs and addresses E with an account for the silence and his own noncontinuation that references his “nerves,” and he issues an apology and stop try formulated in terms of disinclination to administer more shocks (lines 3–4). The format of the stop try is account + apology + marker of action disaffiliation (“I don’t want to”) such that T could be characterized as building the vocal resistance here as a dispreferred type of action. After a silence (line 5), E issues a directive to continue, which, as it does not respond to T’s account as such and although prefaced with a politeness term (“please”), displays high entitlement to have the directive complied with. T responds with silent delay (line 7), then sustains the earlier resistance with an offer to return the check in a further stop try that states his preference to “go home” (lines 8–9). After more delay (line 10), E counters T’s resistance with prod 3 (line 11, “It is absolutely essential that you continue”). T performs further resistive silence (line 12), and E delivers prod 1, “Please go on” (line 13). 8 T postpones a bit further (line 14), then complies with E’s directives (line 15). Although a concern for L’s suffering may be implicit, particularly in a positional sense, T’s displayed orientation to the troublesomeness of continuation emphasizes how that course of action is difficult for himself: he formulates his own “want” and “desire” rather than articulating something about L’s situation. Although T’s introduction of stop tries here resembles defiant participants’ practices, when E insists on continuation, he backs down (line 15). Also, the episode occurs after 315 volts have been delivered; most defiant participants mobilize repeated stop tries earlier in the progression of shocks and never reach this point in the experiment. Although T occasionally offers further resistance in the subsequent interaction, example 2 is his last use of stop trying.
T’s self-attentiveness in example 2 initially refers to his psychological state (e.g., “My nerves . . . ”). Another form of self-attentiveness can refer to the social consequences T may face for shocking L. For instance, subjects sometimes wonder in whispers to E whether after L is released he will seek vengeance. And a number of participants resist self-attentively by expressing concerns about being held responsible for harm inflicted on L. In example 3 in the following, after receiving 195 volts L protests loudly and complains about his heart (lines 7–9); E then urges T to “go on” (line 10). Regarding our earlier methodological points, by virtue of the CA focus on sequence organization and the competing relevancies posed by directive and complaining types of action, the example well illustrates the contradictory situation Milgram’s subjects face. T initially resists the directive with silence (line 13), then upgrades her resistance at lines 14–15 with the prefacing marker of action disaffiliation coupled with an embedded reference to her own potential responsibility and turn-ending recognition of L’s loud complaining (“anything happening to him”):
(3) [2012 obed, 195V, 72]
Immediately following the excerpt, E counters T’s resistance with a version of prod 2 (“the experiment requires that you go on, Teacher”). T then backs down (“oh, I see”) and resumes experimental tasks, presenting herself as being informed by E about the necessity of continuation (Heritage 1984). Although she resists intermittently throughout the remainder of the experiment, especially with a stop try at 405 volts (“I’m afraid I ~
Other-Attentive Resistance among Obedient Participants
Resistance may be compositionally other- rather than self-attentive by displaying an orientation to ways in which continuation is troublesome primarily for L rather than for T. As discussed earlier, other-attentiveness as a feature of resistance is more common than self-attentiveness in both outcome groups. This fact may indicate a favored, asymmetrical ordering of altruistic over egoistic accounts for providing remedy, an ordering attended to by both groups. Here, we discuss other-attentive resistive practices as used in the obedient group and their relation to compliant outcomes.
One practice that shows other-attentiveness is for T to report to E something problematic about L’s reaction to receiving shocks:
(4) [2310 obed, practice 75V, 2]
Seated in the adjoining room, L utters a pain cry upon receiving 75 volts in the practice lesson (line 4). T monitors the situation with resistive silence (line 5). At line 6, reporting on what he “heard,” T explicitly brings the cry to E’s attention in an other-attentive way. The report suggests the relevance of remedying action by E, who is seated in the same room as T and presumably heard it just as clearly (participants wear no headphones). E does not, however, immediately respond (line 7). At line 8, E directs T to continue, thereby treating the pain cry as unproblematic for continuation. Following further resistive silence and hesitation, T complies (line 10). Although the excerpt occurs earlier (75 volts in the practice lesson) than most other instances of nonsilent resistance in both the obedient and defiant collections, this T offers no further nonsilent resistance until 150 volts in the experiment proper. Whereas many defiant participants upgrade resistance to stop tries at 150 volts in the shock series, this T queries E as to what he should do next (“Continue?”). When E then directs him to go on, T does so after a pause, offering only minimal resistance as he ascends the shock series. He does not perform a stop try until 300 volts, well past the middle range of shocks where almost all the defiant participants successfully stop the experiment. So, although example 4 occurs relatively early in the shock series, this T’s overall performance conforms to the obedient pattern: resistance is relatively unassertive, and if assertive practices such as stop tries are mobilized, they come relatively late and finally are abandoned as the subject complies with E’s directives.
Both outcome groups may also express sympathy for L, an other-attentive practice similar to—but distinct from—the characteristically defiant practice of empathetic Golden Rule accounting (discussed in detail in the following). In extract 5, T announces to L at line 10 a correct word-pair answer. T does not immediately continue, and at line 12, E, referring to line 4 where T had shocked L, directs T to hold down the shock switch longer than he has been doing. E trails off, and after a period of silence (line 13), T accounts for the way he has been delivering the shock in terms of feeling “sorry” for L (line 14). The account postpones immediate continuation and by referencing “him” (L) is compositionally other-attentive:
(5) [0215 obed, 225V, 123]
The expression of sympathy is mitigated with “a little bit,” although L has repeatedly issued pain cries, complaints, and demands to be released. The account, which follows up on T’s prior silent resistance (lines 11, 13), tacitly invites E to take remedial action on L’s behalf. E, however, counters L’s invitation by completing at line 15 his unfinished directive from line 12. T expresses willingness to comply (line 16) and after a bit of further delay (line 17), does so at line 18. Past this point, T’s further resistance is similarly mitigated until he offers a stop try at 360 volts—“I can’t (.) keep on (givin’ enough of this voltage)”—a point well past the middle range of shocks. Following T’s accounts for the stop try, E directs him to continue; T does so and reaches 450 volts and becomes categorized as obedient. As in example 4, T is slow to mobilize assertive resistance, and when he does so, it is unsuccessful.
Expressions of worry (example 6) and “imprecations” such as oaths and sighs (example 7) can also show sympathetic other-attentiveness to L’s plight:
(6) [2017 obed, 285V, 177]
(7) [2024 obed, 420V, 335]
As in the previous examples, these participants are offering resistance that—although other-oriented insofar as they reference L’s situation—is relatively mitigated and late in the shock series (285 and 420 volts, respectively). Subsequent to excerpt 6, T resists several more times before reaching 450 volts and becoming obedient, although none of these instances are stop tries. The participant in example 7, who at 420 volts has almost reached the end of the shock series, offers almost no resistance past this point. Accordingly, despite variation within the obedient group, the evidence suggests that obedient resistance is typically less assertive than that of defiant participants, it is slower to mobilize, and even when obedient subjects engage in assertive resistance through stop tries, they back down or comply with subsequent directives rather than sustain or renew their resistance.
CA Methodology and Disobedient Participants: Self- and Other-Attentive Resistance
We will now use CA methodology to show how disobedient participants use self- and other-attentive resistance to experimental continuation. In contrast to the situation with resistive but ultimately compliant participants, when E counters resistive responses to his directives to continue, defiant subjects offer counters of their own. They sustain noncompliance with accounts for their earlier resistance and with iterated stop tries (again, utterances that explicitly project imminent experimental discontinuation as described earlier). As the experiment progresses, these defiant subjects rapidly upgrade to the most explicit forms of resistance found in the corpus such as the stop try. By contrast, obedient subjects upgrade more slowly and use the most explicit form far less frequently (19 percent in the corpus use at least one stop try, as compared to fully 98 percent among the defiant). Although defiant accounts do include self-attentive statements of the sort analyzed with regard to obedient subjects, they are more frequently other-oriented, demonstrating concern for L’s welfare and projecting experimental discontinuation as the appropriate remedy for his complaining. To further our methodological goal of decomposing Milgram’s categorical outcomes in terms of constitutive practices performed in real-time interaction, we now examine examples of both self- and other-oriented forms of resistance in the trajectories toward disobedience among the defiant subjects in order to draw further conclusions about what distinguishes noncompliant from compliant practices.
Self-Attentive Resistance among Disobedient Participants
Self-attentive resistance among defiant subjects can look very much like self-attentive resistance among the obedient ones. Yet one kind of self-attentive statement that disobedient subjects make far more commonly is that which refers to legal repercussions of continuation. Of the 53 defiant participants in the corpus, 11 (21 percent) resist by self-attentively referring to legal consequences; in contrast, only 4 of the 64 obedient participants (6 percent) do so.
Example 8 in the following is from condition 3, in which Milgram has moved L from an adjacent room (condition 2) to the same room as T and E. L has been refusing to continue, and at lines 1–2, T asks E about having “any responsibility.” This formulation, with its reference to “any responsibility,” is negatively polarized (Heritage 2002:78; cf. Horn 1989) and in that way structurally projects disconfirmation (a “no”). A silence develops (line 3), and T adds an increment to this turn (line 4), showing how he may be implicated. After a slight delay (line 5), E at line 6 asks to clarify the “responsibility” as “legal” (which interpretation T confirms in overlap at line 7). Beyond the overlap (line 6), E denies that T’s responsibility would ensue:
(8) [0310 disobed, 195V, 126]
The self-attentiveness is palpable in the “responsibility” issue, and T does not immediately continue after E completes the question-answer sequence but sustains resistance with silence (line 8). At line 9, T’s resistance becomes more explicit: whereas the prior query had presumed E’s epistemic entitlement or stance (Heritage 2012)—his knowledge and right to pronounce on experimental matters—this one downgrades or critiques it. The laugh tokens (line 9) resemble those that have been documented in patients’ responses to a physician’s directive, which “are recurrently produced with laughter” (Haakana 2001:191). 9 Following the excerpt, T sustains resistance in several more exchanges with E. Though in time he decides to continue, he later achieves a defiant outcome by using stop tries.
Other-Attentive Resistance among Disobedient Participants
Members of both outcome groups tend to initiate resistance upon first hearing L cry out in pain at 75 volts. Noncompliance at this point is not very forceful, in both groups typically taking the form of silent monitoring of the situation that delays continuation and/or reporting to E that something is problematic about L’s reaction to the shocks. Between 75 and 120 volts, both groups also commonly address queries to E that similarly display other-oriented concern for L’s welfare. For instance, example 9 shows a defiant participant performing a first nonsilent attempt at resistance with an other-attentive query (line 3):
(9) [0235 disobed, 105V, 7]
T’s negatively polarized (“not gonna hurt him”) question asks for confirmation and shows concern for L’s welfare (Jefferson 2004; Raymond 2003). After slight delay (line 4), E provides confirmation. Following more conferral with E that further delays continuation, T resumes his experimental tasks (transcript not shown). He soon, however, achieves a defiant outcome at 150 volts with a series of other-attentive accounts (e.g., “I believe him, they hurt”; “He’s tired, he wants to stop”) and stop tries (“Well, I won’t continue with endangering a man’s life”).
Although both outcome groups use such other-attentive inquiries, defiant participants tend to swiftly mobilize additional forms of other-attentive resistance. Two such additional forms—(1) Golden Rule accounts and (2) letting L decide—are almost never used by obedient participants. Both practices account for prior or accompanying stop tries and are crucial means by which defiant subjects distinguish themselves from obedient ones.
Golden Rule accounts
Both in everyday and theoretical ethics, a typical formulation of the Golden Rule is “Do unto others as you would have them do unto you.” The maxim enjoins moral actors to behave charitably and beneficently by empathetically identifying with how the world looks to others (cf. Heritage 1998:313). When T shows such empathetic identification, we refer to the practice as a Golden Rule account: T justifies (accounts for) resistance in terms of empathetic identification with L. The practice thus contrasts with the merely sympathetic other-attentiveness of examples 6 and 7. Golden Rule accounts occur in the dilemmatic situation we described earlier, in which T is caught in the middle of two opposing and competing projected courses of action. L has typically emitted some kind of protest, which suggests remedying the putatively painful administration of shocks, but E nevertheless directs T to continue. Of 17 instances of the practice in the corpus, all account for prior stop tries. Moreover, all but 2 occur with ultimately defiant participants, such that the technique is one of the signatures of this outcome group. The 2 Golden Rule obedient cases occur at 300 volts (subject 0216) and 330 volts (0220), whereas the defiant cases typically occur below 270 volts—that is, earlier in the progression of shocks.
One of the obedient instances, though highly unusual for that outcome group, is nevertheless a perspicuous example of the practice:
(10) [0216 obed, 300V, 348]
Here, T, responding to earlier directives to pursue experimental tasks, tries to stop the experiment at lines 7–8 with a “look”-prefaced turn addressed to E, which redirects the course of action toward the alternative of discontinuation (Sidnell 2007). The stop try has the marker of action disaffiliation (“I don’t wanna”) and features a post-positioned apology. No response from E is forthcoming at line 9, and T (line 10) pursues a response with a Golden Rule account for his just prior stop try. The turn justifies the prior try by empathetically identifying the speaker with L and projects discontinuation. Golden Rule accounts have a proverbial flavor; for this reason, the account may, like idiomatic expressions accompanying complaints (Drew and Holt 1988), solicit alignment in a particularly strong way after it has not been otherwise forthcoming. Nevertheless, after some delay, E directs him to continue (line 12). The directive treats T’s empathetic identification with L as irrelevant to continuation and renews the relevance of that course of action. Though T does not immediately comply, at line 14 he does so grudgingly with a sigh.
Two further Golden Rule instances, performed by an ultimately defiant subject (examples 11a and 11b in the following), are remarkable because they happen early in the experiment (90 volts) before L has started to complain and demand to be released. T is seated next to L (condition 3). Prior to E’s directive at line 1, T has twice told E that he doesn’t “wanna hit that switch” (transcript not shown):
(11a) [0311 disobed, 90V, 386]
T counters E’s directive by expressing misgivings about the high amount of voltage (line 3). After a prod from E (line 5) about the “absolutely essential” necessity of continuation, T delays (line 6) and then produces a prototypical form of protest with an agree-disagree format (line 7; Pomerantz 1984) in which he refers to the “man” sitting next to him. Physical and psychological proximity of the man receiving shocks may enhance the likelihood of empathetic responses. After a subsequent prod from E (lines 10–11), T produces another disagreement (line 13), assessing the last punishment as being “enough.” Following a lapse (line 14), he adds a Golden Rule account that conditionally (by way of an implied “if” prefacing “I was sitting over there” and the modal “would hope”) suggests an identification with the role of the other, to put the point in Meadian ([1934] 1990) terms.
Following several more exchanges of directives and resistive responses (28 lines of transcript omitted), E again directs T to continue (lines 17 and 20 in the following). Again T resists, other-attentively describing L’s experience (lines 22–24). E waits (line 25) and then (line 26) produces an account on behalf of his directives to continue, to which T responds in his patterned way of acknowledging E’s talk in an agreeing fashion (line 30) and then showing misalignment with his interlocutor, E (lines 30–32).
(11b) [0311 disobed, 90V, 430]
T’s disagreement (lines 30–32) is formulated as a Golden Rule account for resistance that once again shows his empathetic identification with L’s experience. Following this excerpt, L agrees to “try one more” word pair. But he soon reverts to expressing his earlier wish to discontinue, and after delivering 105 volts, T performs another stop try [“Well, I’m sorry, but (.) I’m not gonna go up anymore on that voltage”]. After T repeatedly counters E’s further directives prodding him to continue, E announces discontinuation.
Letting L decide
In the course of resisting and especially accounting for stop tries, some participants try to negotiate with E the terms under which they would be willing to continue the experiment (cf. Gibson 2014). One other-attentive way this is done is the practice of letting L decide, which is a second distinctive practice by which defiant subjects account for their attempts to stop. In the following, we first show how obedient participants can attempt to negotiate terms and then discuss the characteristically defiant technique of letting L decide.
Members of both outcome groups attempt to set the terms of continuation. The practice resembles an action observed in everyday family interaction in which a child tries to bargain with a parent directing her to go to bed. For instance, a child may stall by “shaking her head no, and then bargaining for delaying going to bed: ‘Will you read me.’” The child is “negotiat[ing] the conditions for her compliance” (Goodwin and Cekaite 2013:7). In the experiment, negotiating compliance may involve stating a condition to be met before T will continue or asking E to act in a certain way if T is to comply. Attempting to set the terms of continuation may take on a primarily other-attentive cast, as in this obedient example:
(12) [2035 obed, 405V, 278]
At line 1, T produces an apology-prefaced declination (cf. Robinson 2004:296), a stop try that projects discontinuation. Then, countering E’s directive at line 3, T sustains resistance by making a self-attentive appeal to her “conscience,” with resistive laugh tokens interspersed and appended (line 5; Haakana 2001). Following a silence, she produces the next word-pair item (line 7). No response is forthcoming from L (line 8), who is seated in the next room and has not responded for some time. At line 9, T uses an interrogative formulation to plead with E to postpone continuation until E has ensured that L is in acceptable condition to continue. The plea prompts E to “
Negotiating the terms of continuation, then, is a resistance practice displayed by members of both outcome groups; but actually setting such terms in a way that is letting L decide is a practice that defiant participants uniquely deploy. With this practice, they treat L, rather than themselves or E, as the party who should decide whether to go on. Letting L decide can account for trying to stop and/or can counter E’s directive to continue. Participants typically follow up the practice not by backing down but with insistence that L be allowed to make his decision known. E, for his part, uses special prod 2 to counter such concerns that L is being forced to continue (“Whether the learner likes it or not, you must go on until he has learned all the word pairs correctly. So please go on.”). 10
Letting L decide is performed in either of two ways. In the less common version, T addresses L with a proposal to discontinue or a statement that L should decide whether to continue:
(13) [0306 disobed, 150V, 205]
In the more common version, T addresses E with a stop try account or other counter to E’s directive that tasks E with demonstrating that L is willing to go on:
(14) [0208 disobed, 120V, 26]
In either version of letting L decide, T accounts for trying to stop by treating L as the relevant authority on the matter of continuation, the party with primary rights and entitlement to decide. By contrast, in Golden Rule accounting, participants typically treat themselves as knowing that L is being forced to act against his will and that it is up to them alone to decide whether or not to continue.
Discussion
We have introduced conversation analysis (CA) as an innovative approach in the study of Milgramesque behavior, one capable of raising new phenomena for social psychology. Methodologically, CA is innovative as a response to Reicher and Haslam’s (2011) call for a rethinking of Milgram’s dichotomization of action in the experiments as either obedient or defiant. CA allows us to see recurrent directive-response and complaint-remedy types of sequence organization occurring in the three-party interactions of the Experimenter, Learner, and Teacher. The juxtaposition of these conversational sequences creates an interactional dilemma for research participants as they find themselves caught in the middle of two mutually opposed and competitive proposals for what next action they should take. Comparative analysis of a large number of obedient and defiant experimental sessions makes possible a better understanding of the lived, temporal experience of Milgram’s participants by illuminating the generic structure of their dilemma and the recurrent practices by which they attempted to resolve it.
Substantively, we have analyzed self- versus other-oriented techniques for resisting experimental directives, practices hitherto undocumented in the Milgram literature. Participants from both outcome groups commonly use both styles, either alone or in combination. The evidence, however, suggests that successful, defiant resistance can involve both styles, with other-attentive resistance appearing more often than the self-attentive variety. Moreover, successful resistance can feature two recurrent other-attentive techniques that are signatures of the defiant group: (1) Golden Rule accounts justify stop tries (prior and ongoing attempts to bring the experiment to an end) in terms of empathetic identification with the Learner. And by (2) letting the Learner decide, participants negotiate with the Experimenter, treating the Learner—the man receiving shocks—rather than the Experimenter or themselves as the appropriate party to decide whether or not to continue. Our analysis of these structured techniques sheds light on generic patterns by which successful defiance in the Milgramesque situation is achieved in real time.
With this methodology and substantive analysis, not only do we understand Milgram’s obedience experiments better. We also suggest that an agenda has been set whereby the methodology of CA, with its detailed explication of interactional practices, can be used in efforts to understand the social organization of other social scientific methods, including experiments (as here) and survey interviews (Maynard and Schaeffer 2000). CA could also improve understanding of authority-subordinate relations as they imbue non-Milgram settings such as police-citizen encounters, military command structures, work relationships, and more.
Further research taking the sequential and comparative approach of the present article is needed to elucidate more fully how Milgram’s two outcome groups—obedient and defiant—take different trajectories from the point of initial resistance onwards. Defiant participants owe their success to their practical competence at resourcefully weaving together diverse forms of resistance. The archived Milgram recordings display the structured practices by which they resist more frequently, more assertively, and with more heterogeneous resistance techniques than do obedient ones. With respect to self- versus other-oriented resistance, defiant participants use both techniques more frequently and earlier in the shock series and may also mobilize Golden Rule accounts and letting L decide. Such findings contribute to what is known about the background of shared resistance practices—what defiant and obedient subjects may have in common. They also add to our understanding of how some people succeed in putting a stop to inducements to engage in immoral actions (Zimbardo 2007) whereas others do not. In short, our methodological innovation in the social psychology of obedience to authority has important substantive implications for the better understanding of Milgramesque situations.
Footnotes
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Research for this paper was funded by National Science Foundation grant #1103195.
1
Social psychologists currently disagree as to whether Milgramesque behavior is in fact best described as obedience to authority. Theories of social identity (Reicher and Haslam 2011) and the web of obligation (
) are among those currently debated. Although the present article contains descriptions of participants as obedient and disobedient/defiant, we do not thereby imply support for Milgram’s interpretation. We use the terms simply as useful descriptors for members of the two outcome groups.
3
Milgram (1974) and Miller (1986) summarize most of the 24 experimental conditions. Recently, Russell’s (2014b) archival research has shed new light on Relationship Condition 24. Haslam, Loughnan, and Perry (2014) and place the total number of Milgram’s subjects in all conditions at 780, namely, somewhat less than what Milgram reported.
4
With a grant from the National Science Foundation (#1103195), the first author purchased copies of the 117 recordings and hired undergraduates to make preliminary transcripts and then edited them according to the conventions of conversation analysis (CA).
5
Conditions 2 (Learner and Teacher in separate rooms) and 3 (in same room) vary the physical and psychological proximity of the Learner to the Teacher. Condition 20 uses women as subjects, whereas all other conditions use men. This paper’s findings are not affected by the situational variations of these several experimental conditions.
6
The data excerpt heading refers to “0219 disobed” (Subject 19, Condition 02, disobedient outcome) and “135 V, 26” (excerpt starts at line 26 of the original transcript; 135 volts is the highest shock delivered at that point). Analogous headings appear in all excerpts that follow. See note 5 for more on each experimental condition.
7
8
It is not uncommon for E to deliver the scripted prods somewhat out of order.