Showing Leadership by Not Showing Your Face

We examined experimentally whether a person unknown to potential followers could be seen as showing leadership. Based on the social identity analyses of leadership, we predicted that would-be leaders pursuing group-oriented goals would be seen as showing leadership to a greater degree when they were anonymous than when they were identified. We predicted this pattern would reverse when would-be leaders pursued personal, self-oriented goals. Support for this hypothesis was found for all but the most highly identified group members. For extremely highly identified group members, a would-be leader’s pursuit of group-oriented goals was all that mattered to produce relatively high levels of leadership perceptions. For all other participants, an anonymous, in comparison with an identifiable, group-motivated target was seen as showing relatively high levels of leadership. These data provide support for the social identity analysis of leadership, and help explain otherwise counter-intuitive and naturalistic observations of followership of anonymous leaders.


2001;
. Despite its complexity and nuance, the fundamental outcome of this work is the recognition that leadership is a process that is made possible through a shared sense of social identity (Tajfel & Turner, 1979) among potential followers and potential leaders. In this manner, being seen as "one of us" is crucial to being seen as a leader.
Empirically, Gaertner, Mann, Murrell, and Dovidio (1989) observed a greater number of people willing to lodge leadership votes for fellow in-group than out-group members, whereas Duck and Fielding (2003) observed that people were more satisfied with identically behaving in-group rather than out-group leaders. Many other authors (e.g., Grace, David, & Ryan, 2008;Platow et al., 2005Platow et al., , 2007 have shown that social influence, a key feature of leadership, emerges primarily, if not solely, from fellow in-group members; indeed, the use of contingent resource power is likely to be both a symptom and a cause of intergroup relations between would-be leaders and would-be followers (Reynolds & Platow, 2003;Turner, 1991). Moreover, the more a potential leader is seen as a member of the in-group, the greater is his or her ability to lead. Fielding and Hogg (1997), for example, showed that group members' perceptions of a leader's effectiveness increased with increasing levels of perceived in-group prototypicality (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987), whereas Platow, van Knippenberg, Haslam, van Knippenberg, and Spears (2006) observed a positive causal relationship between relative ingroup prototypicality and subsequent attributions of leader charisma.
Indeed, one way that potential leaders can demonstrate and even enhance their in-group credentials is by pursuing and/or achieving in-group causes and goals instead of their own personal ones. This claim features prominently in a variety of leadership models (e.g., Cartwright & Zander, 1968;Hollander, 1964;House, 1971), so that leaders working in the name of the group rather than their own personal interests will garner greater support. Supporting this claim, Yorges, Weiss, and Strickland (1999) found that self-sacrificing leaders exerted stronger influence over their followers and were seen as having more charisma than self-benefiting leaders. Similarly,  observed that leaders associated with dramatic group success were attributed greater charisma than those associated with dramatic group failure; this was particularly pronounced for leaders with a history of in-group promotion. In these and similar instances (e.g., de Cremer, 2002;Wit & Wilke, 1988), group members who are associated with achieving group goals, as opposed to pursuing their own personal self-interest, are most likely to be seen as leaders. As Haslam et al. (2011) recently observed in their summary of this literature, leaders who take care to promote the group interest (more colloquially, those who are in-group champions, those who "do it for us") reap many benefits. They receive endorsements from followers, they are likely to be seen as charismatic, they influence the opinions of their followers, and they are able to enlist the efforts of their followers in bringing their visions of the future to fruition. (p. 133) An important moderator of this fundamental process, however, resides not within potential leaders, but within potential followers. If the social group in question lacks subjective meaning or value to potential followers-if their relative levels of social identification with the group are low-these group processes are likely to be attenuated, if not completely dissipated. For example, in the Fielding and Hogg's (1997) study described above, the positive relationship between perceived leader effectiveness and leader ingroup prototypicality was stronger among participants with higher levels of social identification (with the salient ingroup) than among those with lower levels of social identification. Platow and van Knippenberg (2001) also showed that highly in-group prototypical leaders were given greater leeway to engage in out-group favoritism than were non-ingroup prototypical (in-group) leaders. Again, however, this was only among highly identified group members. Finally,  recently showed that group members' levels of social identification positively predicted not only their perceptions of a personal bond with their in-group leader, but their attributions of charisma to that leader. In particular, those group members who identified highly with their group perceived the greatest personal bond and attributed the greatest amount of charisma to their ingroup leader. Consistent with a social identity analysis of leadership, these effects did not emerge when the leader was an out-group member.

The Current Study and the Potential Role of Leader Anonymity
In light of the above arguments, the current study examined group members' judgments concerning the degree to which a fellow group member showed leadership as a function of (a) the target's personal versus in-group-oriented behaviors and (b) participants' own levels of in-group social identification. Critically, the target leader in our research was either personally identified or anonymous. Conceptually, two different outcomes could reasonably follow from the anonymity of a potential leader. On one hand, anonymity could be seen as breaking any potential relationship between a would-be leader and fellow in-group members. Indeed, the actual presence of a clear interpersonal relationship between followers and would-be leaders assumes an important role in leader effectiveness within some leadership analyses (e.g., Graen & Uhl-Bien, 1995;Scandura & Graen, 1984). In contrast, anonymity research and theory conducted within a social identity framework suggests a potentially different process. In this instance, individuating fellow in-group members serves to break down perceptions of a shared sense of "us" by, inter alia, highlighting individuality and individual differences (Lea, Spears, & de Groot, 2001;Postmes, Spears, Sakhel, & de Groot, 2001). In the context of a salient group membership, therefore, anonymity works to maintain a shared collective orientation and, subsequently, allows leadership processes to unfold.
Together, these opposing perspectives led to the prediction of an interaction between the would-be leader's identifiability versus anonymity and his or her personal versus group-oriented behavior. From traditional relationshipbuilding analyses, we expected that anonymity would reduce perceptions of leadership, though primarily in a context that de-emphasized any a priori shared collective identity (i.e., a situation that was more interpersonal than intragroup in nature). A behavior that would achieve this, of course, is one that is geared more toward the would-be leader's own personal interests than toward those of the group. From a social identity analysis, in contrast, we expected that anonymity would enhance perceptions of leadership in a context that emphasized a priori shared collective identity. Such a context would be one in which the would-be leader pursued more group-oriented behaviors. We, therefore, hypothesized the following: Hypothesis 1: Whereas would-be leaders will be seen as showing lower levels of leadership when they behave anonymously in a personal-oriented manner, they will be seen as showing higher levels of leadership when they behave anonymously in a group-oriented manner.
As noted above, we also considered the role of group members' levels of social identification with the relevant ingroup. Given that group processes are, almost by definition, more important among group members with higher than lower levels of social identification, we expected that social identity processes would be enhanced among high identifiers (relative to low identifiers). In contrast, we expected that interpersonal processes would be enhanced among low identifiers (relative to high identifiers). In this manner, an overall three-way interaction was predicted.
Of course, we recognize that both leaders and would-be leaders can pursue their personal-oriented and group-oriented goals through a variety of means, each of which may vary in its broader normative support (e.g., the differences between ethical and non-ethical leadership; Brown & Treviño, 2006). Some may wonder, for example, how Subcomandante Marcos' revolutionary behaviors became accepted by his followers, as these behaviors, by definition, entail counter-normative actions within a particular frame of reference. While pursuing counter-normative behaviors, would-be leaders may find it particularly difficult to be seen by potential followers as actually showing leadership (e.g., Hollander, 1964). For this reason, we currently included one final independent variable in which the would-be leader supposedly engaged in either counter-normative behavior (in this case, the illegal behavior of graffiti writing) or more normative behavior (hanging banners). At the same time, however, we did not have specific hypotheses regarding this variable. On one hand, we might reasonably expect that enactment of counter-normative behaviors would make it particularly difficult to be seen as showing leadership, even among actors pursuing group-oriented goals. On the other hand, the actual pursuit of group-oriented goals may lead potential followers simply to ignore the counter-normative actions. As such, we formulated no a priori hypothesis with regard to this variable.
Finally, we included in our methods a series of items that would provide additional information about the participants' attributions regarding the target's behavior. Our inclusion of these items allowed us to explore empirically more of the thought processes underlying any effects we may observe on our primary dependent variable of perceptions of showing leadership. These additional measures included ratings of intelligence, initiative, and blame (particularly in light of the counter-normative behavior). We expected that potential leaders would be seen as more intelligent and having more initiative when pursuing group-oriented goals than when pursuing personal-oriented goals. We also expected that potential leaders would be seen as more blameworthy when engaging in counter-normative than more normative behaviors. Once again, however, we were unclear how these measures might be affected by our anonymity manipulation. On one hand, these measures may be understood as leading to highly individuating attributions, so judgments on them may be particularly high under conditions of identifiability; on the other hand, intelligence and initiative, for example, may be attributed simply to those who pursue group-oriented goals, regardless of their relative anonymity. In the end, however, our primary purpose for including these measures was to understand better potential followers' thinking when making their judgments of the target's showing of leadership.

Participants and Design
Three-hundred forty-one undergraduate psychology students (73 males, 265 females, and 3 who did not indicate their gender) participated in the experiment. Ages ranged from 18 to 56 years (median age = 20.50 years). Each participant was randomly assigned to one condition of a 2 (target anonymity: anonymous/identified) × 2 (target motive: group motive/personal motive) × 2 (type of behavior: writing graffiti/hanging banners) between-participants factorial design. The relevant social identity we currently employed was that of students' university identity.

Materials and Procedure
The experiment was administered via a two-page questionnaire. The first page of the questionnaire, labeled "Attitude Scales," presented four items to measure social identification with the group "university students" (Doosje, Ellemers, & Spears, 1995;Luhtanen & Crocker, 1992): (a) "I often regret being a university student" (reverse scored), (b) "I am pleased to be a university student," (c) "I feel good about being a university student," and (d) "I feel that being a university student is not worthwhile" (reverse scored).
The second page of the questionnaire, titled "Understanding Motivations for Actions," instantiated the three manipulated variables. This began by stating, "As you may be aware, there has been": (a) "a recent spate of graffiti," or (b) "an upsurge in the number of banners hung" on the university campus. This served as the manipulation of the type of behavior, with the former being more counter-normative than the latter. This was followed by the manipulation of target anonymity. In the target identified condition, the text read, "The graffiti artist [person hanging the banners] has been identified as John Barker, aged 21, who lives at 48 Canyon Road." In the target anonymous condition, the text read, "At this time, no-one knows the identity of the graffiti artist [person hanging these banners]; he/she is completely anonymous." Finally, this text was followed by the target motive manipulation. In the group motive condition, the text read that the content of the graffiti [banners] "has focused solely on criticisms of the government's reduction of funding to universities throughout the country. The graffiti artist [person hanging the banners] is clearly acting in the name of student causes." In the personal motive condition, the text read that the content of the graffiti [banners] "as focused solely on topics" about the "personal interests" of the graffiti artist (or person hanging the banners) "unrelated to university student issues." Participants were then asked to respond to a series of questions on a 7-point scale (1 = strongly disagree, 7 = strongly agree). The first question was a manipulation check of target anonymity, stating, "The identity of the graffiti artist [person hanging the banners] is known." The second question was a manipulation check of the target motive, stating, "The graffiti artist [person hanging the banners] is acting in the name of a specific ideal rather than his[/her] own personal interests." These questions were followed by the primary dependent variable, "The graffiti artist [person hanging the banners] is showing leadership in his[/her] behavior." Finally, a series of questions provided additional information about the attributions participants made regarding the target's behavior. Three of these were as follows: (a) "In all likelihood, the graffiti artist [person hanging the banners] is an intelligent person," (b) "The graffiti artist [person hanging the banners] showed initiative," and (c) "The graffiti artist [person hanging the banners] is completely to blame for his or her actions." Given the counter-normative nature of graffiti writing, this last item also served as a manipulation check for type of behavior. These questions were followed by three additional items asking whether it was "something about the" (d) "personality," (e) "group affiliations," and (f) "environment" of the graffiti artist [person hanging the banners] that "caused him[/her] to do what he[/she] did." Participants' age and gender were recorded, and a complete debriefing was provided upon completion of the experiment.

Results
Less than 1% of the data were missing; missing values were replaced using an expectation likelihood method. All main effect means (and associated standard errors) not reported in the text below are presented in Table 1.

Manipulation Checks
A 2 (target anonymity) × 2 (target motive) × 2 (type of behavior) between-participants factorial ANOVA was conducted on the target-anonymity manipulation check. The only statistically significant effect was the main effect for target anonymity, F(1, 333) = 371.89, p < .001, η p 2 = .53. Participants agreed more strongly that the target's identity was known in the identified condition than in the anonymous condition. A similar three-way ANOVA was then conducted on the target motive manipulation check. A statistically significant main effect occurred for target motive, F(1, 333) = 82.01, p < .001, η p 2 = .20. Participants agreed more strongly that the target was working for a higher ideal rather than his or her own personal interests in the group motive condition than in the personal motive condition. Unexpectedly, an interaction occurred between target motive and type of behavior, F(1, 333) = 9.49, p < .01, η p 2 = .03. This interaction indicated that the overall main effect was stronger in the banners condition (M group motive = 4.34, Standard Error of the Mean (SEM) = .17; M personal motive = 2.67, SEM = .18) than in the counter-normative, graffiti condition (M group motive = 4.27, SEM = .18; M personal motive = 3.21, SEM = .18). Importantly, pairwise comparisons within each type-of-behavior condition were statistically significant, ps < .001.

Social Identification
The four social identification items formed a reliable scale (α = .80); hence, a mean of these items was calculated for each participant. Overall levels of social identification were found to be high, with the mean level (M = 5.96, SEM = .05) statistically significantly greater than the mid-point of the response scale (i.e., 4), t(340) = 40.14, p < .001. The distribution of these means was then centered and used as a continuous predictor in the remainder of our analyses below.

Attributions
Personality, group affiliations, and environment. Attributions of behavior to personality, group affiliations, and the environment were treated as a single within-participants variable as "type of attribution." We analyzed this through a five-way, mixed analysis using the univariate General Linear Model module of SPSS 21 with an ordinary least squares estimation procedure (for similar analyses, see Peate, Platow, & Eggins, 2008;Platow, Hunter, Branscombe, & Grace, 2014). Each of the manipulated independent variables was entered as a twolevel categorical variable, centered social identification was entered as a continuous covariate, and type of attribution was entered as a three-level within-participants variable. We then specified all main and interaction effects. This analysis produces results equivalent to a linear regression but becomes particularly useful in analyzing models, such as the current one, that contain within-participants factors.
One statistically significant effect including the withinparticipants variable emerged: the interaction between type of attribution and target motivation, F(2, 650) = 19.09, p < .001, η p 2 = .06. Under group-motivation conditions, participants made greater group-affiliation attributions (M = 5.38, SEM = .08) than either personality (M = 4.91, SEM = .10) or environment attributions (M = 5.05, SEM = .10). Whereas the latter two did not statistically significantly differ from each other, both statistically significantly differed from the first, ps < .01 with a Bonferroni correction. Under personalmotivation conditions, participants made greater personality attributions (M = 5.02, SEM = .08) than ether group-affiliation (M = 4.53, SEM = .09) or environment attributions (M = 4.66, SEM = .09). Again, the latter two did not statistically significantly differ from each other, and both statistically significantly differed from the first, ps < .01 with a Bonferroni correction. 1 These differences provide further confirmation of the success of our target motivation independent variable.

Intelligent.
A four-way analysis was conducted on the attribution of intelligence item, in which all main and interaction effects for the three manipulated variables and measured social identification were examined. The only statistically significant effect to emerge was the main effect for target motivation, F(1, 325) = 9.25, p < .01, η p 2 = .03. Participants more strongly agreed that the target was intelligent when the target was working for group rather than personal causes.
Initiative. The same four-way analysis was conducted on the attribution of initiative item. A statistically significant main effect was obtained for target motivation, F(1, 325) = 6.18, p < .05, η p 2 = .02. Participants considered that the target showed more initiative when the target was working for group than personal causes. A statistically significant main effect also occurred for type of behavior, F(1, 325) = 112.79, p < .001, η p 2 = .26. Participants considered that the target showed more initiative when the target hung banners than when he or she engaged in the counter-normative behavior of graffiti writing. This latter main effect was qualified by a type-of-behavior by target-anonymity interaction, F(1, 325) = 4.50, p < .05, η p 2 = .01. The overall difference revealed in the type-of-behavior main effect was greater under anonymous conditions (M hanging banners = 5.35, SEM = .15; M writing graffiti = 3.44, SEM = .15) than identified conditions (M hanging banners = 5.18, SEM = .15; M writing graffiti = 3.91, SEM = .15). Finally, there was a statistically significant interaction between type of behavior and level of social identification, F(1, 325) = 4.90, p < .01, η p 2 = .02. There was a negative relationship between social identification and target showing initiative when the target engaged in counter-normative behavior (i.e., graffiti writing; β = −.12), but a positive relationship when the target engaged in the more normative (banner-hanging) behavior (β = .14).

Blame.
A four-way analysis was then conducted on the attribution of blame item. This analysis revealed two statistically significant effects. The first was a main effect for type of behavior, F(1, 325) = 21.72, p < .001, η p 2 = .06. As expected, participants placed greater blame on the target for the counter-normative behavior of graffiti writing than for the more normative, banner hanging. This was qualified, however, by

Showing Leadership
A four-way analysis was conducted on the primary dependent variable of "showing leadership." A statistically significant main effect was found for type of behavior, F(1, 325) = 53.97, p < .001, η p 2 = .14. Participants agreed more strongly that the target showed leadership when engaged in the more normative behavior of hanging banners than when he or she was engaged in the counter-normative behavior of graffiti writing. A statistically significant main effect also occurred for target motivation, F(1, 325) = 37.30, p < .001, η p 2 = .10.
Participants agreed more strongly that the target showed leadership when working for group causes than when working for personal causes. Finally, a statistically significant three-way interaction occurred between target motivation, target anonymity, and level of social identification, F(1, 325) = 6.01, p < .05, η p 2 = .02. This interaction was explored by examining the effects of target motivation and target anonymity among participants with higher and lower levels of social identification. Upon inspection of the distribution of social identification, the median (on a 1-7 scale) was 6.25. Whereas approximately 50% of participants had levels of social identification between 1 and 6, the remaining 50% had levels of social identification between 6.25 and 7. Consequently, we chose not to examine the effects of target motivation and anonymity at one standard deviation above and below the mean as suggested by Aiken and West (1991). 2 Instead, we conducted separate analyses among participants above and below the median. For participants above the median, whom we call "extremely high identifiers," the main effect for target motivation remained statistically significant, To examine the responses of participants below the median level of social identification, we recognized that these were not simply "low identifiers"; indeed, they covered the range of most of the social identification scale. Hence, we first re-centered the distribution of social identification among this latter sub-group of participants, then included it in a further analysis to see if it entered into any significant moderating effects. It did not. It was thus treated simply as a covariate so as to control statistically for the known heterogeneity of social identification. In this analysis, the main effect for target motivation was statistically significant, F(1, 158) = 31.59, p < .001, η p 2 = .17. Once again, participants agreed more strongly that a group-motivated target showed leadership (M = 4.17, SEM = .15) than a personally motivated target (M = 2.96, SEM = .15). Moreover, the interaction between target motivation and target anonymity was statistically significant, F(1, 158) = 5.82, p < .05, η p 2 = .04.
As displayed in Figure 1, participants agreed more strongly that a group-motivated target who was anonymous (rather than identifiable) showed leadership. By contrast, participants agreed more strongly that a personally motivated target who was identifiable (rather than anonymous) showed leadership. Interestingly, the only mean above the mid-point of the response scale was of the anonymous target performing group-motivated behaviors; this difference, however, was only marginally significant, p < .06.

Discussion
The focus of the current study was whether a person who is unknown to potential followers can be perceived as showing leadership. Perhaps unsurprisingly, the strongest effect from our dependent measures concerned the actual behavior of the target. Would-be leaders were seen as more intelligent, and as showing more initiative and, indeed, leadership when motives for their actions were group-oriented rather than personal-oriented. "Doing it for us," rather than doing it for one's own personal self-interest, as Haslam et al. (2011) and others noted, is crucial in garnering leadership attributions from potential followers. Indeed, by pursuing group-oriented goals, would-be leaders were seen as more intelligent and having more initiative. In addition to the group-oriented versus personal-oriented behavior, however, was our central question regarding the impact of target anonymity versus identifiability on potential followers' leadership perceptions. It was hypothesized that anonymous (compared with identifiable) targets would be perceived as showing lower levels of leadership when they behaved in personal-oriented manner. This pattern was expected to reverse, however, among anonymous targets behaving in a group-oriented manner. This basic interaction prediction was, indeed, supported, although not among group members whose levels of identification were extremely high.
In the current sample, participants with extremely high levels of social identification seemed to be solely concerned that the would-be leader's behavior was group-rather than personal-oriented. This was not consistent with our prediction. Rather, the central aspect of our predicted interaction with regard to participants' levels of social identification was that social identity processes for highly identifying group members would be strongest in a context emphasizing social identity. It was assumed that a would-be leaders' group-oriented behavior would constitute such a context. However, the extremely high levels of social identification expressed by half our sample seemed to act as a form of perceiver readiness (McGarty & Grace, 1999), to the extent that group-oriented behaviors were all that mattered. Personal-oriented behaviors were simply unacceptable and, therefore, likely to be seen by extremely high identifiers as antithetical to leadership. Consistent with the social identity analysis from which our predictions were derived, leadership for these participants was likely to be always, and only, a group process. The identifiability or anonymity of a would-be leader thus became superfluous.
Among the remaining sample, however, our key interaction hypothesis was supported. As with extremely high identifiers, the group-oriented versus personal-oriented behavior displayed by the target was a powerful determinant of group members' leadership perceptions. However, this was moderated by the anonymity versus identifiability of the would-be leader. When targets acted in ways that served to break down shared collective identities by pursuing self-serving goals, being identifiable seemed to temper the withholding of leadership ascriptions. Having some knowledge of the would-be leader as a person-simply by being able to attribute a name and address to this person-appears to have been sufficient to begin reconstructing the social relationship necessary for successful leadership to emerge (e.g., Graen & Uhl-Bien, 1995). In contrast, when the target was already pursuing group-motivated ends, identification served only to individuate him or her, potentially separating him or her from potential followers and breaking down shared social identity (cf. Haslam et al., 1998). Anonymity, it seems, allows a grouporiented leader to be seen purely as that: group oriented. It is precisely this group orientation that, as Haslam et al. (2011) concluded, is the enabler of leadership.
At this point, it is worth noting how our current work dovetails with, yet differs from, Hickman and Sorenson's (2014;Hickman, 2004) independent work on invisible leadership. Much like the social identity analysis of leadership, their work focuses on the power of collective orientationsin their case, the power of common purpose. This common purpose entails "deeply meaningful shared experiences, beliefs, values or goals" that motivate group members "beyond self-interest to focus on the well-being of a group" (Hickman & Sorenson, 2014, p. 4). This is very similar to the social identity and self-categorization analyses of social norms and motivation (e.g., Haslam, Powell, & Turner, 2000;Turner, Oakes, Haslam, & McGarty, 1994). At the same time, invisible leadership is not isomorphic with our current usage of anonymous leadership. While leaders can be anonymous under conditions where invisible leadership is operative, they need not be. This is because it is the "common purpose [that] is the invisible leader" (Hickman & Sorenson, 2014, p. 12) and not the person or persons actually fulfilling the leadership role. Indeed, our analysis provides experimental evidence in support of Hickman and Sorenson's claims that leaders can, indeed, be anonymous. The essential process involves potential followers having high levels of social identification with the psychological group, and the wouldbe leaders must be acting for "us" and not for themselves personally.
We recognize, of course, that the context of our research was without the vividness and complexities of ongoing group dynamics in which leaders rise and fall with temporal and material interdependencies. Future research would do well to examine this anonymous leader effect in more complex and engaging group settings. A field study, for example, would allow for a critical test of our analyses with issues and group memberships about which potential followers have enduring and urgent concerns. Such a study would also allow for an examination of the external validity of our theoretical analysis. Nevertheless, the current experimental design has provided a clear mechanism by which to test our hypothesis in an unencumbered manner.
Future research could also benefit from expanding the measures of leadership perceptions. For example, a broader range of dimensions drawn from the leadership stereotypes outlined in leader categorization theory (e.g., Lord, Foti, & de Vader, 1984) would help clarify our understanding of just how anonymous, group-oriented leaders are seen to show leadership. Currently, perceptions of both intelligence and initiative were not affected by the target's relative anonymity. However, the leader stereotype contains a substantial list of attributes; exploring the full extent of these attributes would undoubtedly shed more light on the psychological processes underlying our current anonymity effects. More directly related to the social identity analysis of leadership, however, would be to employ the newly developed Identity Leadership Inventory . This inventory, developed after the current research was conducted, presents four sub-scales directly related to social identity-relevant leadership processes: Identity Prototypicality (being "one of us"), Identity Advancement (acting for us), Identity Entrepreneurship (crafting a sense of us), and Identity Impresarioship (embedding a sense of us). In the current study, it may well be, for example, that the group-oriented, anonymous leaders would be perceived as have enhanced identity prototypicality and advancement, but not entrepreneurship or impresarioship. Clearly, this is a direction for future work.
Finally, we note that our current experimental inclusion of two different types of behavior, one of which was counternormative, can be seen to strengthen the findings, precisely because the hypothesized interaction was not qualified by this domain of the leader's behavior. We were unable to predict how this variable might affect perceptions that targets were showing leadership because we did not know how the group-oriented nature of the would-be leader's behavior, as well as followers' own social identification, would interact with the type of behavior. As it turned out, even the counternormative behavior for which the would-be leader was actually seen to be blameworthy allowed an anonymous target to be seen as showing the highest levels of leadership when pursuing group-oriented goals.
Reflecting back on anonymous leaders outside the laboratory then, the followership they attain becomes explicable. Perhaps it is the case that Subcomandante Marcos needed only to maintain his anonymity not simply to avoid arrest by the Mexican government, but to maintain the followership of the people he sought to lead. This, of course, holds true insofar as he continued to champion the causes of the group as a whole, and not his own personal self-interest.

Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) received no financial support for the research and/or authorship of this article.

Notes
1. Two between-participants effects also emerged from this analysis; these effects collapse across the three types of attributions. First, there was a statistically significant main effect for target motivation, F(1, 325) = 14.41, p < .001, η p 2 = .04.
Overall attributions were greater under group-motivation conditions (M = 5.10, SEM = .07) than personal-motivation conditions (M = 4.73, SEM = .07). This was qualified by a statistically significant interaction between target motivation and type of behavior, F(1, 325) = 3.97, p < .05, η p 2 = .01. The difference in the attribution of overall motivations occurred primarily in response to the target hanging banners (M group motive = 5.21, SEM = .10; M personal motive = 4.65, SEM = .10), but not in response to the target writing graffiti (M group motive = 4.98, SEM = .10; M personal motive = 4.81, SEM = .10). 2. When we did follow the Aiken and West (1991) method, the pattern of results was actually identical to that currently presented.