R1. Optimal group functioning

A central feature of our theory is that differentiation of selves promotes optimal group functioning by increasing system gain. What do we mean by that? We wrote that system gain can help members of a systematically organized group achieve “better results than the same number of individuals working together but without a system” (sect. 1.1, para. 5). Smaldino raised the question of what constitutes “better” results. He asked whether we meant that which would aid group survival and enable groups to attain resources or whether we meant other outcomes, such as subjective enjoyment of group membership. Most groups are formed for reasons and purposes, which means that they have functions. Culture itself likely originated because togetherness and coaction created benefits reflected in the biological outcomes of survival and reproduction (Baumeister Reference Baumeister2005; Boyd & Richerson Reference Boyd and Richerson1985; Mead Reference Mead1934). We therefore consider optimal group outcomes those that enable a group to compete effectively with other groups and accomplish group objectives (as defined by the group).

Belzung, Billette de Villemeur, Grivin, & Iorio (Belzung et al.) suggested that our focus on system gain may carry implicit value judgments and that not all groups actually value system gain (and certainly not all individuals within the group). We focused our review on task groups, that is, groups formed to work together to accomplish some function. Not all collections of individuals or categories of people can be considered a group according to this definition. Religious groups or other types of groups that are not task oriented may not value group survival or system gain. An example of such a group is the Shakers, a religious group that promoted celibacy and experienced a stark reduction in population as a result of that practice. Even among task-oriented groups, the benefits of system gain may vary in importance. System gain should be most important when it can produce the greatest benefits, such as when there is competition among groups for scarce resources. System gain may be important in the context of intergroup competition because that competition may threaten a group's survival. Suddendorf (Reference Suddendorf2013) argued that human ancestors competed against other hominids and prevailed, and this presumably occurred because of system gain (especially including warfare). When such competitive pressures are low and life is easy, there may be less need for system gain and hence less interest in cultivating it. Nijstad & de Dreu also added the helpful point that differentiation of individual selves is much more helpful with some kinds of tasks (especially those involving cognitive complexity and deliberate information processing) than with others.

R2. Differentiation of selves

The target article argued that differentiation of selves is a key feature that allows for optimal group outcomes. Several commenters expressed the need for clarification about what constitutes differentiated selves or argued that the concept includes conceptually dissimilar ideas that do not belong together. In this section, we aim to clarify the term differentiation of selves and to discuss mechanisms that can promote differentiation. Differentiation of selves occurs when group members contribute their distinct skills, knowledge, or opinions to a group task (whether performance-related, informational, or moral). People exist first as bodies, and in that sense they are inherently separate and different. Group systems present ways of organizing these disparate bodies into larger, multiperson units. The central issue for us is how much groups retain and capitalize on differences among selves. Groups may even increase differences among selves, such as when division of labor creates specialized expertise. Alternatively, groups may treat members as essentially similar, interchangeable parts (e.g., cannon fodder). The central argument of the target article is that groups benefit by using systems based on differences among members rather than training members to be more or less interchangeable.

Kruger, Vigil, & Stith (Kruger et al.) noted that there are several potential interpretations of the term differentiation of selves. To be clear, differentiation of selves does not refer to surface characteristics, such as whether a group member is male or female. Differentiation of selves also does not refer to perceptions of identities, such as the extent to which a person views him or herself as a prototypical group member. Instead, differentiation of selves refers primarily to contributing a distinct skill, special knowledge, or key opinion to the group. Differentiation of selves could be construed as an umbrella term for different types of role differentiation. That is, differentiation of selves involves taking on a unique role in the group by contributing skills or knowledge that are different from the contributions made by other group members. In group performance or moral tasks, role differentiation promotes personal responsibility for contributing to group outcomes and thereby promotes effort. In informational tasks, role differentiation can mean playing devil's advocate, arguing for a non-conventional viewpoint, or more simply contributing one's opinion without undue influence or pressure from others. Role differentiation in informational tasks promotes independent thought and thus frees the individual from pressure to conform.

Several commenters mentioned that our definition of differentiation of selves as stated in the target article seemed to involve two conceptually distinct ideas. Dar-Nimrod & Gonsalkorale noted that we discuss differentiation of selves as resulting from both identifiability (e.g., being publicly identified) and specialization through role differentiation. Talaifar & Swann similarly pointed out that we defined the term differentiation of selves both as personal regulation, a result of being individually identified and responsible, and social regulation, a result of role differentiation. Levine also differentiated two constructs within our concept of differentiation of selves. Subjective differentiation, as described by Levine, occurs when people feel pressure to behave in accordance with group goals; and objective differentiation involves contributing differing skills, knowledge, and opinions.

Although we appreciate the difference between public and private aspects of self, we sought in our analysis to straddle the two for an important reason. Our investigation began with an attempt to understand the roots of human selfhood. Differentiation is not so much a need originating from inside a person but rather in the social system. People become different not because of some mysterious instinct for uniqueness but rather because differentiated selves make groups function better and so people evolved and learned the capacity to perform differentiated roles in these groups.

The issue of sharpening the definition of differentiation also was raised by Mojzisch, Schultze, Hüffmeier, & Schulz-Hardt (Mojzisch et al.). They suggested that the concept of differentiation of selves as we described it could refer to three different constructs, including (1) distinct roles, knowledge, or expertise, (2) metaknowledge about other differentiated identities, and (3) perceiving the self as autonomous and independent. Our intended meaning of the construct differentiation of selves fits most closely with what Levine called objective differentiation and the first definition of differentiation of selves provided by Mojzisch et al. Mojzisch et al.'s latter two constructs are also quite real, we think, but they are there to facilitate the first.

In reviewing our target article, we can understand the source of confusion. In the fifth paragraph of the introductory section, we wrote: “By submerged in the group, we mean any of the following: People are held neither accountable nor responsible, they are not in competition or playing a distinct role, and they are not publicly identified or rewarded.” This statement equates the definition of differentiation of selves with the mechanisms that can be used to promote or discourage differentiation of selves. Here, we differentiate the definition of the concept of differentiation of selves (contributing a distinct skill, distinct knowledge, or distinct opinion to the group) from the mechanisms that promote or undermine differentiation of selves. Our review focused on the benefits of differentiation of selves in three domains: group performance, group decision processes, and moral group behavior. The mechanisms that can affect differentiation of selves include public identifiability, competition, reward, and accountability. None of these features universally increases differentiation of selves. Instead, the effect of each mechanism on differentiation of selves depends on the task domain (e.g., group performance, group decision processes, moral group behavior).

In the domains of group performance and moral behavior, public identifiability, competition, reward, and accountability motivate group members to exert effort on behalf of the group and to successfully execute their role in the group. In other words, these mechanisms promote differentiation of selves in group performance tasks and moral behavior by serving as a form of group control. The tools of group control may backfire in informational tasks. As Budescu & Maciejovsky indicated, competition can undermine willingness to share information, which could hurt group performance on informational tasks. In a range of group informational processes, public identifiability, competition, reward, and accountability may undermine differentiation of selves by providing incentives for conforming to the dominant opinion and keeping unshared information private. Thus, differentiation of selves is not the same thing as public identifiability, competition, reward, or accountability. These factors are mechanisms that can encourage or stifle differentiation of selves in different contexts.

The context-dependent nature of the mechanisms that can promote or hinder differentiation of selves is evident when considering the example of accountability. Whether accountability promotes differentiation of selves is in part dependent on to whom one is accountable. In general, accountability promotes careful thought and action because, by definition, people who are held accountable must justify their feelings, beliefs, or actions to others (Lerner & Tetlock Reference Lerner, Tetlock, Lerner and Tetlock2003). Nonetheless, accountability may produce conformity rather than differentiation of selves if group members are accountable to an audience that prefers a certain conclusion, outcome, or course of action (Tetlock et al. Reference Tetlock, Skitka and Boettger1989). Haslam & Ellemers asserted that accountability produces moral behavior only when the norms and goals of the group support moral behavior. This point raises a broader question about the definition of moral behavior. Originally, perhaps, moral behavior consisted of behavior that benefited the group and enabled it to survive and flourish (starting, presumably, with cooperation and reciprocity). This utilitarian definition of morality would condone behaviors, such as intergroup violence, that benefit one group at the expense of another. The later introduction of notions of higher, more abstract levels of moral reasoning would allow people to raise moral objections to their group's perceived interests as a whole. In the context of intragroup relations, one interpretation is that accountability increases moral behavior because it reduces selfish behavior and leads people to be more likely to act in accordance with group goals. But yes, if one classifies the group's welfare or tactics as immoral, then increasing group control of individuals, such as by accountability, will push to increase immoral behavior.

We hope this response is useful in clarifying the distinction between the concept of differentiation of selves and the mechanisms that may facilitate or hinder the process. When considering other mechanisms that may affect differentiation of selves, it is useful to consider why these mechanisms have their effect. Faber, Savulescu, & Van Lange (Faber et al.) argued that reputational concerns may underlie many of the mechanisms that we suggested affect differentiation of selves. Similarity, Levine argued that almost all of the factors that influence differentiation of selves can be tied together because they all evoke evaluation apprehension. We agree that reputational concerns or evaluation apprehension are likely responsible for the effect of public identification, competition, reward, and accountability on differentiation of selves. Indeed, these comments underscore our assumption that the inner mechanisms of self were developed to enable groups to function effectively and efficiently. We reiterate, however, that reputational concerns and evaluation apprehension are useful primarily for promoting effort and good behavior in group performance and moral tasks. Reputational concerns and evaluation apprehension can actually be counterproductive in informational tasks, insofar as they create pressure to conform to majority views and thereby suppress the gathering and exchange of information. But even with informational tasks, reputational concerns can help, especially when people gain status in the group by contributing new insights or information. Thus, Levine's and Faber et al.'s point is mainly correct, with the caveat that sometimes reputational concerns and evaluation apprehension can be counterproductive – and mainly when they decrease differentiation of selves.

Healey raised another aspect of the problem of differentiating selves. His contention was that each person may contain multiple selves, or different versions of it, specifically conscious/explicit and implicit structures. In our view, the notion that each person has different selves violates the definition and purpose of selfhood (see Baumeister Reference Baumeister2011), so it is best to think of the conscious and unconscious aspects as different parts of the same self (see also commentary by Forsyth). Terminology aside, Healey's point is instructive. Our target article was in fact motivated by the broad question of how the human self came into being. A solitary person would not need much of a self because things such as ownership, moral reputation, social rank, and even name and address would lose all value, and things like self-esteem and interpersonal appeal (including mate value and job qualifications) would also be irrelevant, if not impossible. The point is that selfhood emerged not out of the needs of the solitary psyche but as something useful to make group systems function better. The different parts of mind and brain thus gradually coalesce to work together to operate an identity in the social system. Healey's comment reminds us that this process is likely incomplete. Making a commitment, such as a marriage or a mortgage, implicates the full self as a unity, even though one may have had inner conflict and misgivings at the time. Inner conflict and disunity can even come back to haunt the person and undercut role performance.

R3. Identification and differentiation of selves

In the target article, we proposed that group formation may occur in two complementary steps. People group together because groups provide benefits to members that ultimately help them survive and reproduce. These benefits can include sharing of resources and information and competitive advantages over other groups. When groups form it is important that individual group members adopt a shared identity and sense of belonging with other group members. As pointed out by Haslam & Ellemers and by Reicher, Spears, Postmes, & Kende (Reicher et al.), a major function of social identity is to promote adherence to group norms (which can even include the norm of being nonconformist, a point raised by Hornsey & Jetten). Promoting adherence to group norms is useful for coordinating activity and developing shared group goals that can act as a guide for individual behavior. The second step, according to our theory, is differentiation of selves. This is accomplished primarily through role differentiation, such as when group members contribute unique skills, knowledge, or opinions to group tasks. We note that these two steps are not necessarily inevitable or governed by a concrete rule. Instead, the steps are meant to have heuristic value and may apply to many but certainly not all groups.

If differentiation of selves involves people contributing distinct skills and expertise, does this mean that differentiation is incompatible with group identification? The commentaries provided a range of interpretations concerning the relationship between step one and step two of our model. Haslam & Ellemers and Hornsey & Jetten, for example, interpreted the target article as arguing that group identification and differentiation of selves are mutually exclusive. We are sorry for the misunderstanding: Again, the steps are intended as complementary, not contradictory.

As another revealing instance, Nijstad & de Dreu characterized our argument as asserting that “members need to differentiate themselves from the group.” Differentiating the self from the group would indeed make the differentiation step the opposite of the first (group identity) step. But that is not what we meant. Instead, we would say that what makes a group effective is that members differentiate themselves within the group – not from it.

We aim to clarify the relationship between group identification and differentiation of selves in this section. The root of the various interpretations of our point can perhaps be tied back to our use of the phrase “submersion of the self in the group.” We intended to use this phrase to indicate that individual selves within groups are not differentiated (e.g., in different roles). A lack of differentiation of selves does not imply anything about group members' social identification with a group. We view social identity and differentiation of selves as orthogonal concepts. Group members can identify strongly with their group and yet show a differentiation of selves, such as when a person adopts a group's goals and yet uses individual agency, thought, and skill to help accomplish those goals. That is what we meant when we wrote that groups flourish when members differentiate themselves within the group rather than from it.

Hodges & Packer indicated that people who lack a social identification with the group may be most likely to benefit from external mechanisms that can promote differentiation of selves, such as accountability and incentives. This excellent point has several implications. The first is that not all group members in large organizations have developed a sense of shared identity with the group, and identification with the group is more likely a continuum than a dichotomy. With group performance tasks, those people who lack strong identification with the group may be more likely to exert effort on its behalf when they are individually identified and can be held accountable for their behavior. (To be sure, the group must control some rewards that the person cares about, or else accountability lacks motivational force.) For people who do identify with the group, differentiation of selves may promote excellent performance because differentiation of selves made possible through role differentiation would enable people who really care about the group to receive credit for their effort. Blanton's comment elucidated this point very well. Meanwhile, people who care less about the group would be unable to hide their lack of effort within the crowd.

Differentiation of selves is likely most beneficial when a person is also socially identified with the group, as several commentators pointed out (Budescu & Maciejovsky; Haslam & Ellemers; Healey; Hornsey & Jetten; Nijstad & De Dreu). This insight improves the analysis of the two steps as complementary. In performance tasks, for example, rewards, competition, and accountability are unlikely to promote effort if group members care little about maintaining their membership in the group, Faber et al. noted. The idea that identification is needed to reap the benefit of differentiation is consistent with our heuristic model of two complementary steps. Group members who identify with their group will likely be motivated to behave in a manner that benefits the group. Differentiation of selves allows groups to achieve maximum benefits through roles, development of unique skills, and willingness to share privileged information. Consistent with these comments, we predicted that the best outcomes occur both when group members achieve a sense of belonging and identity in the group and then go a step beyond that by differentiating themselves within the group.

Nijstad & De Dreu argued that the contribution of differentiation of selves may have been overestimated while the contribution of identification underestimated. In their analysis, they suggested that many problematic group behaviors can be tied back to individuals acting in accordance with their own self-interest rather than in line with the interests of the group. In particular, Nijstad & De Dreu raised the issue of anonymity (one factor we proposed has an effect of differentiation of selves) allowing people to act according to their own self-interest. In the domain of group performance, anonymity could enable group members to act out of self-interest by slacking off on effortful tasks. In the domain of informational tasks, anonymity again may allow a person to act out of self-interest, which could result in a reduction in conformity. One of the main points of Nijstad & De Dreu's commentary is that differentiation of selves can be harmful to group performance if this differentiation leads people to act out of self-interest. Possibly this again suggests the misunderstanding we noted earlier: They thought we were talking about differentiating the self from the group, whereas we focused on differentiating the self within the group. Still, the broader point involves the value of combining both steps, or the need for identifying with a group before cultivating role differentiation. Group identification is needed to reap the benefits of differentiation because it helps to align self-interest to the interests of the group and to avoid negative outcomes mentioned in their commentary, such as deception and power struggles. This insight also reinforces the order of our model's two steps. If group identification is not in place before people act in accordance with differentiated selves, then problematic outcomes driven by self-interest could well occur.

Several commentaries (Belzung et al.; Forsyth; Healey; Hodges & Packer) pointed out that the two steps may be more continuous and fluid than we depicted. Forsyth noted that previous models of group formation, such as Tuckman (Reference Tuckman1965), have identified four stages of group development. Forsyth's own work shows that groups cycle through different levels of cohesion, productivity, and conflict (Forsyth Reference Forsyth2014). Healey raised the issue of whether social identity may need to be reinforced in an ongoing manner. In general, if something about the group (such as being large and loose) leads to decreased social identity, then it may be necessary, as Healey suggested, to reinforce group identity. We recognize that groups are not static and that groups may shift their focus at times, from promoting differentiation of selves to social identity, as goals and members change. These ideas imply opportunities for future empirical work.

Given that we identified the two complementary steps as important to group formation, it is worthwhile to consider factors that may lead groups to move on from Step 1 to Step 2 of the model. Cabeza de Baca, Garcia, Woodley of Menie, & Figueredo (Cabeza de Baca et al.) offered an ecological analysis of factors that may lead to differentiation of selves within groups. They used the Strategic Differentiation-Integration Effort hypothesis (SD-IE) to argue that differentiated roles within groups may be driven in part by pressure due to environmental or ecological conditions. In particular, their work has focused on the question of why some groups are highly differentiated and specialized while others remain undifferentiated and unspecialized. SD-IE argues that high population density in combination with low resource availability should promote “niche-splitting,” which means specialization or role differentiation. Niche-splitting reduces competition for scarce resources by increasing labor productivity and the ability to use resources efficiently and effectively. SD-IE offers support for the prediction that differentiation of selves is particularly relevant to the functioning of large groups and suggests that this differentiation is useful for promoting system gain due to the benefits it confers in making use of scarce resources.

R4. Theoretical connections and extensions

The commentaries offered several opportunities to connect our work to that of other theorists and to consider potential extensions. Blanton provided a particularly useful connection between the target article and his Deviance Regulation Theory (Blanton & Christie, Reference Blanton and Christie2003). Deviance Regulation Theory proposes that groups have two main goals. The first goal is increasing social order, which involves group members adhering to certain conduct codes. The second goal is social complexity, which is the idea that groups benefit from diversity of thought and order. These two group goals approximately mirror the two complementary steps we proposed in the target article. Deviance Regulation Theory centers on how groups employ rewards and punishments to enforce behavior. Instead of focusing on differentiation of selves in terms of performance, knowledge, or opinions, Deviance Regulation Theory defines deviation as differentiation from descriptive and injunctive social norms. According to Deviance Regulation Theory, groups can promote social order by punishing members who deviate from the social norm and can promote social complexity by rewarding group members who excel and differ from the group in a desirable way. As Blanton noted, Deviance Regulation Theory offers a framework for predicting contingencies or mechanisms that might best serve the function of differentiating individual selves. Punishment may promote conformity to group norms, and reward may be especially useful for promoting differentiation within groups.

Possibly related to the regulation of deviance is the encouragement of dissent. Hodges & Packer made the important point that identifying with the group can increase trust, thereby making members feel more comfortable and willing to express dissenting views. Actually, the term they used was “solidarity,” which captures not just the member's individual identification with the group but also the confident sense of being accepted by it. Regardless, the point is that the person who feels strongly included in the group can express dissent without fearing being ejected from the group. Further research may profitably test and build on this insight.

Several commentaries offered thoughtful ideas for expanding our theory to other aspects of group functioning. Zlatev, Halevy, & Tiedens (Zlatev et al.) asserted that rank differentiation may be needed in addition to role differentiation. Indeed, they elaborated this by pointing out that social rank or status can be used as a type of reward to incentivize cooperation and presumably other behaviors that benefit the group (such as when high-performing employees receive promotions). This is a great point that we had overlooked (hence the value of exchanges such as BBS provides!) Rank differentiation is needed in order to direct group activities and to make use of mechanisms that promote differentiation of selves (Halevy et al. Reference Halevy, Chou and Galinsky2011; Reference Halevy, Chou, Galinsky and Murnighan2012b; Simpson et al. Reference Simpson, Willer and Ridgeway2012).

With group performance tasks, rank differentiation is needed to know who is responsible for making decisions, enforcing punishments, or delivering incentives (all of which rely on differentiation of selves). In informational tasks, group leaders or enforcers of some kind are needed to assign people to play the role of devil's advocate or to ensure that perspectives can be expressed without outside influence. In terms of tasks in the moral domain, rank differentiation may be needed to help establish group goals and norms, not least by allowing leaders to emerge. Rank differentiation is therefore a special form of role differentiation that could foster further differentiation of selves and help the group capitalize on the advantages of differentiation. Even in the absence of explicit rank differentiation, a form of rank differentiation may nonetheless be possible. As Hogg specified, prototypical group members have a greater influence over the group and often lead more effectively. Thus, prototypicality may be a key determinant of intragroup differentiation in the absence of an explicit hierarchy.

In addition to rank differentiation, Forsyth argued, subgroups may help coordinate group action. As Haslam & Ellemers pointed out, we did not devote space in the target article to discussing subgroups, and so again we appreciate the insightful contributions emerging from this exchange of views. We will attempt to address this issue briefly here, but further theoretical and empirical work would be most welcome.

The main benefit of subgroups is that they help coordinate complex action or large-scale operations (Kozlowski & Bell Reference Kozlowski, Bell, Schmitt and Highhouse2013). In a sense, subgroups may create an additional level of differentiation by providing members of that subgroup a distinct job, problem, or task. We conceptualize subgroups as functioning much in the same way as larger, umbrella groups in terms of the benefits of social identification and role differentiation – but also functioning within the larger group like differentiated individuals, in that they can focus on specialized tasks and improve group outcomes. Extending our theory, we predict that identification with the subgroup would confer basic benefits not only to the subgroup, but also to the overall group (Hornsey & Hogg Reference Hornsey and Hogg2000). On a football team, for example, identifying with the defense will help the defense but also help the whole team. Identifying both with the subgroup and overall group should reduce conflict between subgroups insofar as those groups view themselves serving complementary roles aimed at attaining a superordinate goal, rather than as competing groups. Beyond the benefits identification offers, our theory predicts that a subgroup in which individual selves were differentiated would perform better than a subgroup in which individual selves were not differentiated, such as a subgroup that assigned each member the same role rather than differentiating roles. Meanwhile, large groups may gain benefits by having differentiated, specialized subgroups that perform distinct tasks, contribute a particular kind of information, are accountable, and so forth.

R5. Bridges and opportunities

Several commentaries applied aspects of our theory to other areas or suggested outstanding empirical questions in manners that we had not anticipated but were quite thought-provoking. In this section, we discuss these bridges and opportunities.

One point made by Kruger et al. and by McDermott is that it is useful to consider just how specialized roles should be to achieve maximal group functioning. Kruger et al., for example, argued that division of labor may require some redundancy in order to be effective. This is a great comment and suggests that differentiation can be overdone, to the point that it is counterproductive. Thus, it can be useful to have more than one person who knows how to complete a particular aspect of a task. If only one person knows how to perform the task, then the group cannot move forward if that person becomes unavailable. Hence, it is beneficial to have some redundancy of skills when using division of labor. Differentiation of selves does not necessarily mean that each group member is assigned to a completely nonredundant role. In many groups, it is necessary to have more than one person perform the same role. When roles are somewhat overlapping, then additional mechanisms are needed to bring out fully differentiated selves (e.g., accountability, public identifiability, reward). McDermott also mentioned that nonredundancy in roles can be highly problematic in high-functioning groups such as the military if a person in an extremely specialized role is killed during warfare. Future empirical work could profitably explore the optimal level of differentiation within a group. At which degree is a group too differentiated to function optimally?

Brown noted that many findings reported in our paper involved so-called WEIRD samples (Henrich et al. Reference Henrich, Heine and Norenzayan2010a; the acronym stands for Western, educated, industrialized, rich, and democratic) and could be considered Western-centric. He argued, for example, that depersonalization actually may lead to many positive group outcomes, perhaps especially in less WEIRD samples (such as for groups in collectivistic cultures). We are intrigued by this possibility and welcome methodologically strong research showing how groups can function better without differentiation of selves than with differentiation. This research would be useful in illuminating boundary conditions and might contain lessons that could be incorporated into our theory.

Another question that could be used to establish boundary conditions: Under what circumstances might a group not benefit from differentiation of selves? Perhaps whether differentiation of selves benefits groups depends in part on the goals of the group, such as discussed by Talaifar & Swann. In particular, they argued that differentiation may not be needed to achieve some group goals, such as in cases of identity fusion, in which people are strongly identified with the group and on that basis are willing to make extreme sacrifices for the group. People who are strongly identified with the group may be willing to make extreme sacrifices, even if there is no role differentiation. This is compatible with McDermott's suggestion that the military may provide an exception to the idea that differentiation helps group functioning. McDermott argues that military groups often need people to be somewhat interchangeable (this echoes the redundancy point, above; the potential danger of impairing group function because a specialist is killed is obviously greater in combat units than in most other groups). We see the logic behind that statement and acknowledge that historically, military groups with more soldiers were generally more successful in battles (e.g., Morris Reference Morris1965). Nonetheless, it is useful to note that military units have evolved to be more and more differentiated and specialized. This process is presumably driven by pressure for the group to be as effective as possible. Factors that promote group identification and cohesiveness, such as uniforms, certainly are beneficial, but from our perspective there is a case to be made for the usefulness of differentiation, even in military groups, Zlatev et al.'s point about rank differentiation is obviously highly applicable to military groups; it is doubtful that a fully egalitarian army (i.e., one without ranks or commanders) would function effectively in battle.

Several commenters (Kruger et al.; Levine; Mojzisch et al.) thought that the review would have benefited from organizing the literature review around an existing task typology (e.g., McGrath Reference McGrath1984; Steiner Reference Steiner1972). This could certainly be done. Mojzisch et al., for example, suggested that the demands of a task may determine whether differentiation is needed. As an example, they noted that in a group of mountain climbers tethered together, the skill of the least-skilled climber determines the group's success. That statement is undoubtedly true, but it also may be true that differentiating selves within the group can improve the outcome relative to not differentiating selves. If the least-skilled climber feels responsible and accountable to the group as an individual, he or she may exert extra effort to climb quickly and accurately, which is consistent with evidence we cited in our review about the least-skilled swimmer in a team relay performing better when in a group but individually identified (Osborn et al. Reference Osborn, Irwin, Skogsberg and Feltz2012).

Levine offered an alternative organization of our literature review around norms. As stated in his commentary, one norm could be “work hard, cooperate with others,” which would encompass the performance and moral domains, and the other norm could be “express opinions regardless of what others say,” which would cover the informational domain. We organized the literature around three broad categories reflecting different group outcomes. Those categories include: (1) performance task outcomes (effortful production of some end product), (2) informational process outcomes (group decision-making, judgement, etc.), and (3) moral control of group behaviors. With these kinds of papers, it is often a challenge to create the best organization for the literature review. We made the decision to organize the review around task outcomes (rather than task type or norm) because it aligns with our primary interest in how differentiation of selves affects various types of outcomes.

Brennan & Enns mentioned the need to distinguish between statistical effects (e.g., statistical facilitation) and social effects. Their commentary described the wise crowds phenomenon as a statistical effect rather than a social phenomenon. That is right, but groups can benefit by organizing their social interactions to capitalize on the statistical effect. As indicated in the example of the wisdom of crowds effect, social groups can improve their decision making by making use of statistical facilitation. Conversely, social interaction does not invariably improve outcomes, such as when it leads to biased decisions.

Several other commentaries noted questions for future empirical investigation. McDermott, for example, asked whether people may self-select into groups that include a certain amount of differentiation. Perhaps people with few unique skills may self-select into a relatively undifferentiated group or into a group with redundancy in roles. Kruger et al. suggested that the mechanisms that affect differentiation of selves may depend on individual differences. Extroverts, for example, may respond more to reward contingencies designed to promote differentiation of selves.

Sezer & Norton discussed the target article in terms of its implications for vicarious processes. Vicarious processes, such as vicarious contagion, occur when a group member acquires or catches attitudes and preferences from another group member. These processes create group members that are increasingly similar over time. As Sezer & Norton highlighted, these processes can bring about negative consequences, especially when the attitudes or emotions being transferred among group members are undesirable. Differentiation of selves within groups may help decrease some of the potentially harmful consequences of vicarious contagion. Although their commentary considered vicarious contagion of negative behavior only, we assume that it also applies to behaviors, attitudes, and preferences that would benefit the group.

Several commentaries discussed applications of the target article to other lines of research. Douven, for example, commented on how agent-based simulation (a type of computational modeling) could be used to test certain aspects of our theory. Barnier, Harris, & Sutton (Barnier et al.), like us, are interested in the question of when groups are more or less than the sum of their parts. Based on the collaborative recall literature from cognitive psychology and the distributed cognition literature from philosophy, their work suggests that knowledge must be integrated and differentiated to achieve optimal group outcomes. Jacobson commented on our assertion that background diversity is not always helpful for forming shared group identity by noting that it would be unethical to select for background homogeneity in hiring decisions. In contexts that value diversity, they are right, though presumably Jacobson was not asserting an ethical imperative to include men on the women's track team. The research we reviewed is descriptive rather than prescriptive. Ben-Ze'ev & Krebs applied our theory to when partners decide to dissolve romantic relationships, noting that partners who take on a unique role in the relationship may be more likely to stay than partners who do not. Spiegel applied our theory to resuscitation teams responding to emergency, finding support for our theory in a team situation involving stress and time pressure.