Zaccaro et al. (1991), the full SRM cannot be estimated (i.e., the model is not identified). If we can assume that the dyadic reciprocity correlation equals zero, the model is identified with as few as three persons per group.
Thus, the dyadic correlation is beneficial in that, if we assume it to be zero, we are making a valid assumption. Second, the SRM may be estimated using a SEM approach, as described later in this chapter. That model is very complicated, so adopting the assumption that some correlations are zero would greatly simplify an otherwise overly complicated model.
studies that employed a single measure of leadership and found 49% of this variance to be target variance. Also, in the five studies of leadership ranking, Livi et al. found that 48% of the variance was target variance. Clearly, leadership is not totally ‘‘in the eye of the beholder,’’ and the obstinacy that many authors demonstrate in continuing to analyze target variance has a strong empirical basis.
Livi et al. did find that there is less target variance and more relationship and perceiver variance when leadership is socioemotional than when it is task oriented. Thus, task-oriented leadership is more clearly consensual, in the sense that members agree on task-oriented leadership rather than on socioemotional leadership. Because leadership consensus on transforma-tional and laissez-faire leadership have been shown to moderate the emotional exhaustion–work commitment relationship (Cole & Bedeian, 2007), it would be interesting to determine whether more transformational leadership styles (e.g., charismatic leadership) are less consensual and laissez-faire style is more consensual.
Given that the target component represents the leadership perception of a member, controlling for other sources of variance, it can be correlated with external criteria; in this way, we may examine the behavior and trait correlates of consensually perceiver leadership. We might first examine which behaviors leaders, rather than followers, exhibit. We might then take that behavioral profile and try to match it with the group prototype of leadership (Epitropaki & Martin, 2004; Hogg, 2001) or leadership stereotypes (Lord, Brown, Harvey, Hall, 2001). Alternatively, we could examine which personality characteristics of the leader predict the perception of leadership (Lord et al., 1986). We could then examine whether behaviors or alternative prototype or stereotype matches mediate the effect of traits on the perception of leadership.
Another interesting question is the extent to which agreement about leadership reflects appearance-based (not behavior based) stereotypes.
In zero acquaintance (Albright, Kenny, & Malloy, 1988) and thin-slices (Ambady & Rosenthal, 1992) studies, perceivers with minimal information have been shown to arrive at consensual and accurate judgments of targets.
To date, no studies have specifically examined the perception of leadership using minimal information, although a thin-slices study did focus on the effectiveness of negotiation (Curhan & Pentland, 2007). It is also known that baby-faced targets are seen as low in dominance, making it unlikely that they would be perceived as leaders (Zebrowitz-McArthur &
Apatow, 1983–1984). Finally, stereotypes about gender and leadership
(Ridgeway, 2001) have been extensively studied. (The effect of gender on the target effect is discussed later in this chapter.)
Livi et al. (2008) found that target variance increases with group size.
Perhaps, leadership becomes both more necessary and more evident in larger groups, much in the same way that extroversion becomes more apparent in groups than in dyads (Kenny, 2004). These results provide further empirical support for the claim that the emergence of a more hierarchical structure within groups is determined in part by a larger group size.
Finally, just because members agree about a target’s leadership in their group, it does not necessarily imply that the same person would be a leader in another group. Although information on this topic is limited, data from rotation designs (Foti & Hauenstein, 2007;Zaccaro et al., 1991) and other sources (Albright & Forziati, 1995) indicate that leaders in one group are very likely to be leaders in other groups.
Perceiver Variance
We also find evidence for perceiver variance that indicates the extent to which some perceivers see others as leaders, and other perceivers see others as followers, though this component explains only 10% of the variance.
In addition, a small, positive generalized reciprocity (correlation between perceiver and target effect) is present. This result may reflect the fact that leaders believe leadership is important and are then biased to see leadership in others. This correlation is not very large, however, averaging only 0.144.
Although the level of perceiver variance is small, it is important to underscore, as other authors have done (Lord et al., 2001), that the perceiver effect could be an important factor in leadership perception. AsLivi et al.
(2008)show, much more perceiver variance is present in the perception of socioemotional leadership.
Relationship Variance
Relationship variance reflects idiosyncratic perception of leadership. For example, perhaps Peter sees Paul as a leader, but no one else does, and Peter does not just see everyone as a leader. Oddly enough, although relationship variance is the second largest source of variance (after target variance), it has traditionally received very little attention from researchers in this field.
Future research might, therefore, focus on within-group dyadic dynamics, as proposed by leader–member exchange theory (Yukl, 1994) and by vertical dyad linkage theory (Dansereau et al., 1975).
An alternative explanation for relationship variance might also involve the presence of triadic effects (Hoff, 2005). For instance, a group might have two factions, such that each faction has a different leader. In this case, subgroups might disagree about the identity of the larger group’s leader.
Ideally, future leadership research studies will investigate such phenomena.
Categorization and attribution analysis of leadership perception (Lord et al., 2001) might benefit by focusing more intensely on relationship variance. In particular, to avoid confounding factors in relationship and error variances, a study would need to employ multiple measures of leadership. Intragroup dyadic dynamics – which are especially important in long-term groups, such as organizations and teams (Tsui, Porter, & Egan, 2002) – could be intensively investigated in this manner.
The results in Table 2 show a weak dyadic reciprocity effect, and we expect this result might be likewise found in other studies of leaderless groups, such as those involving college students. Members in leaderless groups, where status is not predetermined and where interaction is limited to a short discussion session, have less of a chance to make transactions or exchanges among members (e.g., leader–member exchange theory; Yukl, 1994). Perhaps in more long-standing groups, dyadic reciprocity of leadership might be found. To examine this hypothesis, we reviewed the Montgomery (1986) study, which did examine groups overtime. Her data do, indeed, show evidence of increasing positive dyadic reciprocity overtime.
Group Variance
We failed to find any evidence of group variance in the perceptions of leadership. Although most prior SRM studies of person perception in college students have not attempted to find group variance, studies of families that have used the SRM have looked for variance in family effects (i.e., group effects). For instance,Cook (2001)studied influence and control in families and found that only 4% of the total variance was at the level of the group. It may be especially difficult to find group variance in leadership because it may operate in zero-sum fashion. In other words, leadership, much like status and power, may be a finite resource that is divided among group members: Not everyone in the group can be a leader.
Our failure to find group variance may be somewhat surprising, given that various theories propose that groups could have significant differences relative to one another (e.g., Brown & Lord, 2001; Hogg &
van Knippenberg, 2003). All of the studies that we reanalyzed involved leaderless, largely college student groups, however, and it is likely that there are only small differences between groups owing to their shared stereotypes and socialization histories. Also, our results were obtained in laboratory experimental settings, where there is low acquaintanceship and likely shared prototypes that do not differ from group to group. Moreover, many theories have explicitly pointed out that groups should have differences in leadership perception as a function of culture (Chemers, 2000) or group context characteristics (Kerr, 1977). Ideally, to properly assess these factors, the research design should manipulate group composition to measure the effect of these variables.