We began by taking a close look at laboratory notes from the classic demonstrations of the power the situation. As we will show, these notes provide a different and more complex picture of these studies than the one featured in standard accounts. Rather than being passively buffeted about by powerful situational pressures, participants worked to find a way to pursue their own agendas within the limitations imposed by those pressures.
Revisiting classic evidence that people lack agency
Many believe that the early classic studies provide evidence of near-universal lack of agency. This is inaccurate. Consider the
Asch (1955) line-judgment study. Although most participants in the
Asch (1955) study displayed some evidence of conformity, 57% thwarted the majority more than half the time. Given that the dominant response was dissent, it is factually incorrect to characterize this study as evidence of widespread conformity. Furthermore, when participants did conform to the majority, most did not see themselves as passively yielding to the majority. To the contrary, postexperimental accounts of participants in the Asch study suggested that they were strongly influenced by a feeling of connection with, and obligation to, the other (ostensible) participants. Asch himself pointed out that when they declined to report their true perceptions of the length of the lines, it was often because they worried about embarrassing other participants (
Asch, 1955). These participants were not ignoring their own perceptions; they merely prioritized their feelings of connection to, and empathy with, other individuals over the desire to be correct. Rather than being overwhelmed by the power of the situation, they agentically pursued their desire to be communal, even though it meant rendering judgments that they knew to be incorrect (e.g.,
Higgins, 2012;
Jetten & Hornsey, 2012).
Acceding to the power of the situation was also far from universal in the Milgram study. In fact, although 65% of the participants in Milgram’s “standard paradigm” complied with the teacher’s injunctions all the way to the maximum level of shocks, the designation of the “standard paradigm” was arbitrary; in reality, the 65% condition was no more standard or representative of people’s responses than any of the other variations (
Russell, 2011). Compliance in the 30-odd variations of the paradigm ranged from 0% to 100%. Importantly, the vast differences in rates of compliance appear to have more to do with the connections that participants felt for other people in the experiment than with the power of the situation over the agency of participants. In particular, strengthening the connection with the experimenter (by decreasing psychological or physical distance to him) increased compliance with the experimenter; strengthening the connection with the leaner (by decreasing psychological or physical distance to him) decreased compliance with the experimenter (
Haslam & Reicher, 2012).
Obedience in the Milgram paradigm thus reflected an individual’s capacity to act by connecting with others rather than capitulation to the demands of a powerful authority figure. Further support for this possibility comes from analyses of reactions to the prompts that the experimenter delivered to the teacher (
Burger, Girgis, & Manning, 2011). The experimenter begins politely (“Please continue”) and becomes increasingly forceful and demanding, culminating in a direct order (“You have no choice, you must continue”). Contrary to the assumption that the effects were mediated by obedience, the more order-like the prompt (and thus the more the connection with the experimenter is challenged), the
lower the rate of compliance. Instead, the most effective prompts asked for the participants’ assistance in advancing science, and this pattern also emerged in simulations that varied the order in which different prompts appeared (
Haslam, Reicher, & Birney, 2014). These findings make it easy to understand why, in an unpublished laboratory note, Milgram himself wondered whether it might be more accurate to label the phenomenon he had uncovered “cooperation” rather than “obedience” (
Haslam, Reicher, Millard, & McDonald, 2015). They also support the conclusion that obedience in the Milgram study had more to do with misplaced trust than obedience (
Perry, 2013) and that the results say more about the power of relationships than the power of the situation.
Further evidence for the importance of the connections people form in studies of social influence comes from a later investigation in which participants had an opportunity to form small coalitions against an authority figure (
Gamson et al., 1982). In this study, rates of rebellion were quite high and compliance with the authority figure was relatively low. In fact, half of the groups completely refused to comply with the demands of the authority figure, and a mere 12% complied completely. Such evidence suggests that it is hazardous to use the results of Milgram’s studies as a basis for concluding that people routinely knuckle under to the power of the situation. Instead, the nature of the coalitions people form with other human agents is determinative of the extent to which people defy or give in to the power of the situation. As such, one key to understanding conformity and obedience is identifying the conditions that promote agentic responding through coalition formation.
Similar issues arise when applying the power-of-the-situation narrative to the bystander intervention studies reported by
Darley and Latane (1968). Recent studies have highlighted the crucial importance of the sense of connection participants have to the bystander in such studies. That is, when the victim is a member of one’s own group, increasing the number of bystanders actually leads to
more rather than less helping (
Levine & Crowther, 2008;
Slater et al., 2013). Apparently, when there is a connection between the bystander and victim (e.g., when a
fellow group member is imperiled), bystanders are expected to lend a hand, and the pressure to do so increases with increments in the number of group members who are observing. This finding suggests that helping was relatively low in the original Darley and Latane study because participants felt no connection to the victim. Unfamiliarity with the victim may also explain the (greatly exaggerated—
Manning, Levine, & Collins, 2007) lack of helping in the Kitty Genovese incident that inspired the Darley and Latane research.
These relatively recent findings suggest that the Darley and Latane findings should not be regarded as providing a highly generalizable baseline for rates of helping behavior. In addition, they also show that to understand bystander behavior, we need to focus on the potential of coalition formation and shared group membership. The decision to help should be seen as an expression of agency whereby people help because they care about the victim (for an overview, see
Fischer et al., 2011).
A distinctive set of concerns cloud interpretations of the Stanford prison study. First, questions have recently emerged regarding the prevalence of compliance in that study. Although textbook accounts of this study generally imply that compliance with the assigned roles was near universal, recent accounts of the study suggest otherwise (
Haslam & Reicher, 2012). For example, in the first phase of the study, prisoners challenged the guards, refused to obey their orders, and mocked their authority (
Zimbardo, 2007, p. 54). When guards responded by punishing such resistance, insubordination among prisoners escalated. The culmination occurred when two prisoners removed their caps and prison numbers and barricaded themselves in their cell, shouting “[T]he time has come for violent revolution!” (
Zimbardo, 2007, p. 61). Precisely how this rebellion was crushed has never been spelled out, but Zimbardo himself (who designated himself the leader of the guards) may have been instrumental in the process. For example, he offered one of the rebellious prisoners preferential treatment to act as a “snitch” and somehow convinced him that it was impossible for prisoners to leave the prison. When the prisoner conveyed the news that “You can’t get out of here!” to the other prisoners, it had a “transformational impact on the prisoners” (
Zimbardo, 2007, p. 71). Henceforward, the prisoners stopped acting as a collective, and guards subsequently had little trouble crushing the resistance of individuals acting alone.
And prisoners were not the only ones who resisted the roles to which they were assigned. Despite Zimbardo’s prodding, only “about a third” of the guards “became tyrannical in their arbitrary use of power” (
Zimbardo, 1971, p. 154). Of the two thirds who refused to fall into the autocratic mode, some strove to be “tough but fair” while others were actually friendly to the prisoners, performing small favors for them (
Haslam & Reicher, 2012).
No doubt, some of the participants assigned to the role of guards did indeed behave in a punitive and authoritarian fashion. Yet there are reasons to question whether situational pressures associated with mere role assignment were actually responsible for these effects. For example, Zimbardo did not simply assign participants to the role of guard; he provided them with instructions about how to implement their roles. Not surprisingly, those assigned to the role of guards in Zimbardo’s prison study reported feeling obligated to do the bidding of Zimbardo (their self-assigned leader). This likely reflects the fact that he told guards to deprive prisoners of their sense of agency and autonomy—extraneous elements that went well beyond role assignment:
You can create in the prisoners feelings of boredom, a sense of fear to some degree, you can create a notion of arbitrariness that their life is totally controlled by us, by the system, you, mean they’ll have no privacy. They have no freedom of action they can do nothing, say nothing we don’t permit. We’re going to take away their individuality in various ways. In general what all this leads to is a sense of powerlessness. (
Zimbardo, 2007, p. 55)
Indirect evidence suggests that the extraneous elements in the instructions that Zimbardo gave to his guards were crucial determinants of the outcome of his study. Consider the conceptual replication of the Stanford study filmed by the BBC. Participants reported to a mock prison where they were randomly assigned to the role of guards or prisoners. In contrast to the Stanford study, guards did not become authoritarian nor did prisoners become submissive. Although the BBC study differed in numerous ways from the Stanford experiment, at the very least it shows that role assignment does not inevitably produce the behaviors observed in the Stanford study. Of particular relevance here, it also provides further evidence of the importance of coalition formation in such settings. In the original experiment, a powerful experimenter formed coalitions with the guards and encouraged them to adopt an authoritarian stance against the hapless prisoners who were discouraged from forming coalitions. In the BBC study, the experimenter offered neither guidance nor support to either guards or prisoners. Despite this, prisoners spontaneously formed coalitions and organized a rebellion. The “powerful situation” in the two studies—role assignment—was identical; what made all the difference was the coalitions that emerged in the two studies either by design (in the Stanford study) or spontaneously (in the BBC study).
In short, several decades after the publication of landmark studies by Milgram, Asch, Darley and Latane, and Zimbardo, it is clear that most participants in these experiments did not perceive themselves to be obeying, conforming, failing to help, or merely acting in accordance with assigned roles. Instead, participants seemed strongly motivated to navigate an unfamiliar situation by forming coalitions with other actors in the setting. In their eyes, at least, they remained true to their own convictions. It is thus misleading to conclude that participants in these experiments lacked agency or conformed to situational pressures mindlessly (e.g.,
Langer, Blank, & Chanowitz, 1978). Instead, these classic studies suggest that people were active agents who attempted to form connections they hoped would help them see their way through the perplexing situations in which they found themselves.
Of course, the foregoing analysis relies, in part, on participants’ explanations of the reasons for their behavior. One could discredit such reports by pointing out that retrospective reports of the causes of behavior are prone to bias (e.g.,
Nisbett & Wilson, 1979;
Wilson, 2002). Although retrospective reports are indeed unreliable at times, at other times they are quite valid (
Ericsson & Simon, 1980). Moreover, there is independent evidence that agentically forming connections was on the radar of participants in the original classic experiments as well as follow-ups to those experiments. As noted above, the results of the many variations of the Milgram experiments illustrate that the key determinant of the behavior of participants was the relative strength of the connections they felt with the experimenter versus the learner (e.g.,
Reicher, Haslam, & Smith, 2012). Similarly, follow-ups to the Darley and Latane studies (
Levine & Crowther, 2008;
Slater et al., 2013) revealed much higher levels of intervention when participants felt connected to the victim. Collectively, these studies support the idea that, in response to powerful situational influences, an agentic desire to form connections, rather than passive capitulation, best explains the responses of participants in the classic studies.
Revisiting recent support for the people-lack-agency theme
If an agentic desire to form connections motivated the responses of participants in the classic people-lack-agency demonstrations, may it also have operated in more recent demonstrations that people lack agency? Probably not. As we noted earlier, whereas participants could express agency by forming a connection with someone in the classic studies, participants in the more recent studies had no opportunity to connect with anyone. This reflects the fact that because the more recent studies focused on individual task performance, the experimenters prevented them from interacting with others.
We begin with research on learned helplessness. This work highlighted the ways in which people’s experiences with prolonged lack of control may dampen their subsequent motivation to master the environment, pursue goals, and realize their personal potential (e.g.,
Hiroto & Seligman, 1975;
Seligman, 1975). Such motivational deficits are worrisome, as they might produce motivational deficits in contexts in which perseverance is necessary for success. This argument was later amplified by research indicating that uncontrollable situations tend to foster inefficient investment of cognitive effort that culminates in cognitive exhaustion (
Kofta & Sedek, 1999;
Sedek & Kofta, 1990).
In contrast to the early investigations of learned helplessness, subsequent perspectives offer a more nuanced and optimistic picture of responses to control deprivation. In particular, it appears that people’s initial and most common reaction to control deprivation is to
amplify efforts to exert control in contexts that offer the possibility of reasserting control (for a review, see
Bukowski & Kofta, 2017). The early studies in this tradition examined the impact of control deprivation on subsequent attributional activity, with the assumption that attributions represent a means of making the social environment more predictable and controllable (for even earlier work on resistance against restrictions of freedoms, see
Brehm, 1966;
Wortman & Brehm, 1975). The researchers discovered that participants who were deprived of control were subsequently more sensitive to information regarding the causes of another person’s behavior (
Pittman & Pittman, 1980). Follow-up studies demonstrated that control deprivation also bolstered and refined subsequent information seeking (
D’Agostino & Pittman, 1982;
Swann, Stephenson, & Pittman, 1981) and inferential processing (
Pittman & D’Agostino, 1985,
1989).
A related line of work suggests that, contrary to common understandings of depression, people who are mildly depressed may intensify their efforts to restore control by processing available information more carefully. Support for this possibility comes from research using the same outcome measures utilized in research on control deprivation by Pittman and colleagues. For example, depressed persons were particularly sensitive to information regarding the causes of another person’s behavior (
McCaul, 1983), displayed exceptionally high levels of interest in diagnostic information about an interaction partner (
Hildebrand-Saints & Weary, 1989), and were less apt to display the correspondence bias (
Yost & Weary, 1996). Consistent with our analysis, it appears that participants in these studies attempted to regain control (and agency) by connecting with others who found themselves in similar situations.
More recent research has demonstrated that control deprivation can actually stimulate approach motivation. In particular, participants who were deprived of control were energized by the experience in that they were subsequently more inclined to actively pursue goals (e.g.,
Greenaway et al., 2015). Moreover, the opportunity to respond to experiences with control deprivation by taking active steps to restore control had palliative effects in that it eliminated negative effects of deprivation on subsequent information processing (
Bukowski, Asanowicz, Marzecová, & Lupiáñez, 2015).
Also consistent with this reasoning is evidence that threats to personal control can trigger compensatory efforts to restore control through allegiance to an agentic ingroup (
Stollberg, Fritsche, & Bäcker, 2015). In particular, threatening university students’ feelings of control (by having them contemplate aspects of their lives that induce feelings of helplessness) increased their support for educational innovations that were consistent with an ingroup’s agenda. Presumably, in this instance, conforming to ingroup norms allows people to restore personal control. Here again, the evidence suggests that people are decidedly more resilient when they experience a loss of control than the early research on learned helplessness implied. Rather than responding to threats to control by giving up, people engage in active efforts to regain control through connecting with other individuals or groups.
The notion that people’s experiences in the situation may induce them to give up is also featured in ego depletion theory (
Baumeister & Heatherton, 1996;
Muraven & Baumeister, 2000). This formulation proposed that self-control is a limited resource that can be exhausted by attempting to control oneself. Such mental exhaustion theoretically lowers glucose levels, which in turn causes people to suspend further efforts to control themselves.
Although early explorations of this phenomenon seemed strongly supportive (
Baumeister & Vohs, 2007), recent accounts have suggested that the mechanism underlying these effects has little to do with decrements in glucose or a decision to give up. For example, early evidence that ego depletion caused diminutions in glucose (
Gailliot & Baumeister, 2007) failed to replicate when more precise measures of glucose were used (
Molden et al., 2012; see also
Beedie & Lane, 2012;
Kurzban, 2010). More telling, further studies and conceptual analyses indicated that ego-depletion manipulations do not sap motivation in any broad sense; instead, they sour people on the unpleasant “ego-depletion” task (for reviews, see
Inzlicht & Schmeichel, 2012;
Inzlicht, Schmeichel, & Macrae, 2014). For example, consider the evidence that goading people to complete an undesirable task (eating radishes rather than sweets) reduced later efforts to solve a second undesirable task (an unsolvable anagram task;
Baumeister et al., 1998). Subsequent research revealed that this pattern failed to replicate when the second activity was valued in some way: when participants were rewarded for the activity (
Boksem, Meijman, & Lorist, 2006), when they have a modicum of control over the activity (
Hockey & Earle, 2006), or when they were personally invested in it (
Legault, Green-Demers, & Eadie, 2009). In addition, the ego depletion effect fails to emerge if the initial task is pleasant or valued. The underlying mechanism here may be that when experimenters encourage participants to eat bitter foods like radishes, it degrades their relationship with them and undermines participants’ motivation to persevere on subsequent tasks. In contrast, providing participants with chocolates improves the relationship and motivates them to persevere.
In sum, recent research suggests that ostensibly depleted participants remain quite motivated to perform activities that they construed as expressions of personal agency. Hence, ego depletion does not produce mental exhaustion that impairs ability to perform all subsequent tasks; it merely encourages people to divert their agentic resources away from disagreeable tasks onto activities that interest them. Ironically, it appears that when ego depletion effects are observed, they demonstrate people’s efforts to agentically divert their limited motivational resources onto activities that they deem worthy of their efforts.
Another contemporary approach that purports to capture a general tendency for people to withdraw effort in response to strong situational pressure is stereotype threat (
Steele, 1997; for an integrative review, see
Schmader, Johns, & Forbes, 2008). Here, merely recognizing the existence of negative social stereotypes may undermine performance in the short term even if victims themselves do not believe in the stereotypes (
Steele, 1997). Over time, such performance decrements may lead targets of such stereotypes to withdraw effort. This message, of course, is quite commensurate with the power of the situation narrative that has been so influential in the field of social psychology.
To be sure, some have questioned the strength and replicability of stereotype threat effects (
Flore & Wicherts, 2015;
Ganley et al., 2013;
Sackett, Hardison, & Cullen, 2004; but see
Walton & Spencer, 2009). One reason why stereotype threat effects may be weaker than originally proposed is that people may actively resist stereotypic beliefs that are not “self-verifying”—that is, stereotypes that clash with enduring beliefs about themselves (
Swann, 1983). Consider, for example, evidence that when people receive appraisals that challenge their self-views, they actively and agentically work to set the record straight by bringing those appraisals into harmony with their self-views (e.g.,
Swann & Ely, 1984;
Swann & Hill, 1982). In fact, even denying people the opportunity to behave in an authentic (
Harter, 2002), self-verifying manner (e.g., inducing those who see themselves as assertive to behave submissively) triggers compensatory self-verification strivings (
Brooks, Swann, & Mehta, 2011).
Whether motivated by self-verification strivings or other processes, defiance of stereotypic appraisals can thwart stereotype threat effects. In one line of work, researchers explored the implications of collectively challenging or affirming performance stereotypes (
Smith & Postmes, 2010). Female participants in a group discussion session were prompted to question the stereotype that men outperform women on math tests. Later these participants outperformed participants who had been prompted to affirm the stereotype. Together with the earlier work on self-verification, this finding suggests that stereotype threat effects are less likely to occur when people are free to express and harness their personal or group-based agency. This qualifier is important. That is, although strong situational pressures may sometimes deprive lone individuals of opportunities to exercise agency in laboratory studies of stereotype threat, in naturally occurring contexts people can often exercise agency by resisting, or compensating against, the threats that they confront.
The foregoing research suggests an alternative to conventional strategies for counteracting stereotype threat effects. That is, standard remedies for countering stereotype threat effects (e.g.,
J. Aronson, Fried, & Good, 2001;
Lewis & Sekaquaptewa, 2016) have focused on supporting individuals when stereotype threat arises (e.g., through mind-set interventions, social support, or self-affirmation) or reducing contextual triggers of stereotype threat (by reframing the task, removing threat cues). All of these interventions involve restructuring the situation so that people are either less likely to suffer from, or quicker to recover from, agency deficits. Although this approach has borne fruit, it fails to consider that targets of negative stereotypes are not mere passive victims of unavoidable “threats in the air.” Rather, individuals may actively work to devise ways to agentically cling to the views they have of themselves or to directly challenge the negative stereotype personally or collectively—especially when their companions join with them in questioning the accuracy of the negative stereotype.
Upon close scrutiny, then, it becomes apparent that putative evidence that powerful situational forces routinely deprive people of agency has been overstated at best. In the classic studies, when participants faced powerful situational forces, they expressed agency by striving to form connections with other individuals who were present in the experiment. In the more recent investigations in which the capacity to form connections was unavailable, participants surrendered to situational pressures only when it was clearly unreasonable to do otherwise. In fact, in none of the studies included in our review did we find clear evidence that situational forces stripped people of their capacity to exercise agency. To the contrary, in some studies, we encountered evidence that situational forces sometimes
bolstered agency (e.g., research on compensatory reactions to control deprivation and self-discrepant evaluations). This conclusion is also supported by recent explorations of phenomena such as resilience, grit, and growth (e.g.,
Duckworth, Peterson, Matthews, & Kelly, 2007;
Seligman, Steen, Park, & Peterson, 2005). Here again, when people confront powerful situational forces that clash with their salient agendas, they often resist and, if such forces prove intransigent, they try to circumvent them.
If the notion that situations systematically deprive people of agency is not supported by the empirical evidence, then how has this assumption continued to flourish in the social psychological literature? We suggest that social psychology’s metatheoretical assumptions have played a major role in producing this state of affairs. But could these assumptions also influence the topics that researchers study and cover in major textbooks? To test this possibility, we conducted Study 2. A research assistant (who was blind to our hypotheses) counted the proportion of pages devoted to topics related to lack of agency versus agency in seven major social psychology texts (the same ones examined in Study 1). We then tallied the average proportion of pages devoted to the three most common forms of lack of agency (conformity, compliance, bystander effect) and the three most common forms of agency (dissent, deviance, resistance). The means displayed in
Figure 2 reveal that far more space was devoted to topics focused on lack of agency (topics on the left side) than agency. Moreover, this tendency to favor evidence of lack of agency over agency was apparent in each of the texts we examined. Hence, it likely reflects the state of the art of social psychological research rather than bias on the part of the particular textbook authors.
Skeptics could point out that despite the pervasive emphasis on lack of agency in the social influence literature, many studies have been designed to identify
moderators of lack of agency effects rather than lack of agency per se. In fact, many of the authors of the classic studies envisioned a key component of the balanced approach that we are advocating here. For example, in an attempt to understand how situational pressures moderate obedience, Milgram conducted roughly 30 studies designed to identify the boundary conditions of obedience to authority. Collectively, Milgram’s studies offer a remarkably complete picture of the impact of situational pressures on obedience. Similarly,
Asch (1955) conducted multiple variations of the original line-judgment task that revealed the conditions under which people rebel versus conform. Likewise, in the wake of the publication of his landmark paper on the bystander effect, Darley wrote an article entitled “Do groups always inhibit individuals’ responses to potential emergencies?” in which he challenged simplistic readings and interpretations of the effect (
Darley, Teger, & Lewis, 1973). Nevertheless, the manner in which a phenomenon is framed initially can have considerable impact on how it is understood and studied. Note, for example, that researchers generally construe moderator studies as demonstrating “a limit on the effect” rather than a distinctive phenomenon. In reality, however, rebellion is more than “a limit” on conformity; it is a phenomenon itself with a unique set of causes, mediators, moderators, and consequences (
Jetten & Hornsey, 2014). Each of these phenomena requires independent conceptual and empirical study. Before recognition and analysis of these processes can occur, researchers must first recognize the existence of both sets of phenomena. Only then will the field achieve greater balance in the phenomena examined in
Figure 2. We accordingly consider strategies for developing a more balanced and comprehensive vision of social psychological phenomena in the next section.