“You go to war with the army you have, not with the army you wish you had.” That oft-repeated military saying also can be applied to team formation: You compose a team with the people you have, not with the people you wish you had. There is both wisdom and danger in that statement. It is wise because it encourages leaders to adjust their decisions about team design to reality rather than to delay launching a team until they can corral exactly the right mix of members with exactly the right qualifications. The statement is dangerous because it can encourage mindless expediency—such as when a leader composes a team of those individuals who happen to be readily available even if they do not have the competencies the work requires.
Finding the right balance between reality and expediency in forming a team requires thought, initiative, and occasionally a bit of political maneuvering. The first priority, of course, is selecting the right people. Does each candidate for team membership have specific capabilities or experiences that can help with the team’s work? Do all prospective members have basic teamwork skills, the demonstrated ability to work well with others on a shared task? If not, what might be done to find and recruit members who do have what is needed? And how can individuals who are known to be team derailers be kept off a team?
Since a team is more than just an assemblage of individual members, leaders also must consider the properties of the team as a whole. Will it be the right size—neither too small to get the work done nor so large that merely coordinating among members will consume too much time and energy? And will it have a good mix—people who are neither so homogeneous that they have little to learn from one another nor so diverse that communication and coordination are fraught with difficulty?
This chapter provides an overview of what is known about such matters, with special attention to how leaders can use that knowledge to properly compose their teams. We first examine how individual members are selected, and then turn to leaders’ decisions about how to form them into a team that works. Finally, we put team composition in context: How much of a difference does member expertise actually make in team performance?
A great deal of research has been conducted to explore the effects of member characteristics—personality, interpersonal style, task skills, and more—on group dynamics and performance.1 Here is what is known about those attributes that bear directly on leaders’ decisions about team composition—the task capabilities, teamwork skills, and previous experiences of prospective members.
Research in the field of industrial-organizational psychology has generated a great deal of knowledge about individual differences at work, including the means by which they are most appropriately conceptualized, measured, and used in making personnel decisions.2 Our Group Brain project at Harvard took a different tack, guided by a simple and rough analogy: A group of people is an emergent entity, akin to a brain. Thus, each member can play the role of a distinct brain system, with the different systems working together in what one hopes is an integrated way. We assumed that a team would be more effective if its members had complementary brain-based abilities—but only if the team was able to integrate those abilities appropriately.3
In one study, we focused on two specific aspects of the visual processing system—individuals’ capabilities to visually process objects versus spatial arrangements.4 As is well known to anyone who has explored a new city with a partner, these are distinct capabilities. Some people are great at navigating the streets but fail to notice and remember landmarks, whereas others have the opposite capabilities. We composed dyads whose members had either the same or complementary abilities, and gave each pair the task of negotiating a computer-displayed maze and “tagging” certain objects located in that maze. We expected that the dyads would perform better when a member with spatial ability was in charge of navigating the maze (using a joystick) and a member with object ability was in charge of tagging the proper objects (using a mouse button). The two-person team, then, was construed as a single cognitive unit charged with maximizing performance on the overall task.
Teams that had the right mix of abilities and, importantly, whose members were assigned to the roles that matched their abilities (i.e., navigator vs. tagger) did perform best, confirming the benefit of proper group composition. These teams outperformed not just those whose members did not have a high standing on task-critical capabilities but also those that did have the right capabilities but whose members were assigned to roles incongruent with their special talents (for example, an “object” person assigned to the navigator role).
We also examined the effects of members’ spontaneous collaboration. When dyads were properly composed (the right people in the right roles), there was relatively little collaboration and it had no effect on performance. They didn’t need it, and when they did it, it didn’t help. But when individuals occupied roles incongruent with their abilities (i.e., the object person in the navigation role, or the spatial person in the tagging role), then spontaneous collaboration helped a great deal. By working together they were able to figure out how to compensate for the initial misplacement of their capabilities.
What surprised us were the findings for dyads whose members were high on the same ability. When both members had high object ability, or both had high spatial ability, collaboration actually impaired performance. Why? Because no amount of talking and planning could compensate for the fact that the dyad was missing a critical capability, and members’ fruitless discussions accomplished nothing other than to waste time and induce frustration.
The maze study is but one small contribution to understanding about how the attributes of individuals are assembled into group-level realities. Yet the findings reinforce the importance of getting the right team members on the team, having them in the right roles, and then encouraging them to fine-tune their performance strategies to fit as well as possible with both task requirements and members’ capabilities.5
When running an experiment in the laboratory, it is possible to design the task to require certain abilities and then to select members who do or do not score high on those abilities. But what about selecting members of intelligence teams in complicated organizational settings? One should, of course, pick members who have specific task-relevant abilities whenever possible—for example, hand-eye coordination for helicopter pilots or language capability for a team that analyzes foreign-language materials. When that is not possible, it is quite reasonable to fall back on general intellectual ability (often referred to as “g”), which turns out to be a surprisingly good predictor of performance for many different types of tasks.6 Composing teams whose members have high “g” is not as challenging in the intelligence community as in some other organizations, since the selection practices of most intelligence organizations give considerable weight to overall intellectual ability in selecting employees.7
What is not a good idea is to compose teams without explicit consideration of members’ task capabilities. Leaders may finesse the member selection task and just turn to the usual suspects, mindlessly picking people who are around the office and available. Or they may worry more about representation than capability, putting on the team one (or even two) people from each unit that has a stake in the team’s work. Or, perhaps worst of all, they may assess the political or ideological leanings of the team’s client and then staff the team with like-minded (or, perhaps, contrary-minded) members. Even if a team is able to work around one or two members who have little to contribute, which welldesigned teams generally can do, the absence of the basic capabilities the task requires is certain to cap its potential.8
Not everybody can work in a team. Just because someone has superb task skills does not mean that he or she will be able to collaborate with others to bring those skills to bear on a collective task. Indeed, having just one destructive member can reduce a team to rubble—it really is true that a single bad apple can spoil an entire barrel, and do so in relatively short order.9 Such individuals are unable (or unwilling) to understand other members’ perspectives, they undermine their teammates, they say one thing in team meetings but do the opposite later, and they bring out the worst in other members. Although they may have much to contribute to the work itself, those contributions are better made as a solo player than as a team member.
Team leaders and members appear less willing these days to tolerate such behaviors. In his remarkably titled book The No Asshole Rule, psychologist Robert Sutton argues that people who poison the work environment and alienate their teammates should simply be moved out.10 That is easier said than done. It is emotionally demanding to exclude from a team a person who has expertise the team needs, or to disinvite from team meetings or projects someone who is already a member. It is hard even if you are a formal leader with full authority to make decisions about team composition. The challenge may seem insurmountable if you are just a regular team member.
Yet sometimes it is done. In one large business, for example, the chief executive explicitly excused his chief financial officer (CFO) from leadership team meetings because the problems he created outweighed the expertise he brought. The CFO really was superb at his individual work, so he was not fired. Instead, the chief executive found ways to leverage his experience and expertise that did not require him to attend senior team meetings. Thereafter, the leadership team no longer had to deal with the problems that invariably developed whenever the CFO was present.11
It is one thing to exclude from a team someone who consistently is disruptive. But how about the opposite: identifying in advance people who will be especially constructive team members? There are lots of tools and tests available for doing just that, ranging from self-report personality tests such as the Myers-Briggs Type Indicator (MBTI) to various devices for assessing the social or emotional “intelligence” of prospective team members.12 I have my doubts about such tests. The psychometric properties of the MBTI are, to say the least, worrisome. And neither that instrument nor those assessing social and emotional intelligence have been shown to strongly predict work behavior or performance.13 The enormous popularity of such instruments appears to stem more from the apparent importance of the qualities they seek to assess, and from the fact that respondents often find in their scores affirmation of their self-perceptions, than from scientific studies of their validity.
If the most popular measures of teamwork skills provide a shaky basis for selection, then what is a team leader to do? One possibility is suggested by what may be the oldest saw in psychology: “The best predictor of future behavior is past behavior.” Does it make sense to base selection decisions in part on how prospective members have behaved on their previous teams? We explore that possibility next.
A great deal of data about the kind of team member a person will be already exists in any organization. All you have to do is ask those who have worked with the person in the past. That is what Robert Ginnett did when he needed an assessment of aircraft captains’ team leadership capabilities for his research on flight-deck crews. He just asked pilots who had flown with the captains to nominate individuals who, independent of their technical piloting skills, were exceptionally good or poor as team leaders. There was remarkable agreement among those who were asked.14 Pilots know about the capabilities and behavioral patterns of their teammates. Members of intelligence teams know, too.
Data about organization members’ teamwork capabilities are not recorded in human resource systems, although it might be good if they were. But such data can readily be obtained the same way that Ginnett got them—by asking other members of the teams on which a person has served. Janice is being considered for membership on a joint terrorism task force. How relevant to the team’s work is her technical knowledge and skill? Has she had plenty of task-relevant training and experience? Has she demonstrated the ability to work collaboratively with others? Does she have a good network of outsiders on whom she can call if information or expertise is needed that is not available within the team itself?
It is surprising how much can be learned about someone’s likely contributions from just a few inquiries made to the right people. And it is even more surprising how infrequently those who are forming a team make such inquires. Note, however, that the “right people” generally are those who actually have worked with the candidate. They are more likely than managers to be deeply knowledgeable about the candidate—and may be less tempted to issue a glowing report in hopes that a mediocre performer will move on.
Two cautions should be kept in mind when using data about someone’s training and previous experience to make selection decisions. One has to do with perceptual distortions, the other with styles of working that are rooted in the intellectual and technical disciplines in which prospective members were trained.
Here is the perceptual problem. Team members sometimes deal with their uncertainties and ambivalences about group dynamics by unconsciously splitting their positive and negative feelings into separate parts, assigning the positive feelings to one much-admired member and the negative feelings to someone else, the group scapegoat.15 Who becomes the scapegoat is not random, nor does it necessarily reflect the individual’s actual teamwork skills. Instead, it commonly is the person who is most different from the majority of team members—the one woman in an all-male group (or vice versa); or the one African-American in an all-Caucasian group (or vice versa); or the one member from law enforcement in a team of intelligence professionals (or vice versa). Although splitting is pervasive in group dynamics, the phenomenon is rarely acknowledged, or even recognized, by members. It therefore is essential to solicit specific examples of actual behavior rather than global impressions when making inquiries to assess a prospective team member’s teamwork capabilities.
The second caution has to do with members’ characteristic styles of working. Both fieldwork and our observations of PLG simulations (Chapter 1) strongly suggest that people who have different occupational identities and experiences exhibit markedly different ways of operating—and that those differences sometimes impair the ability of members to work well together. We confirmed these observations empirically in the Group Brain research program.
To see if people who choose different occupations also have different cognitive capabilities, we gave members of four different occupational groups—humanists, engineers, scientists, and creative artists—a battery of online tests that assessed a range of cognitive capacities and personal dispositions and styles.16 Included were measures of specialized abilities in reasoning, attention, memory, perceptual reaction time, and level of verbal fluency, as well as several personal dispositions and preferences.
It turned out that humanists and creative artists scored higher than the other groups in verbal fluency, they had slower and more error-prone spatial working memory, and they were better able to switch between cognitive frames. Engineers and scientists exhibited the opposite pattern—weaker verbal fluency and more cognitive frame-switching errors but stronger spatial working memory and reasoning. The tests also assessed both the accuracy and speed of participants’ responses. As is seen in Figure 6-1, different occupational groups managed the trade-off between speed and accuracy quite differently—especially the creative artists and the scientists. Artists responded more slowly but made somewhat fewer errors, whereas scientists responded quite quickly but at a cost in accuracy.
FIGURE 6-1 Speed-Accuracy Trade-Off by Occupational Group
Differences in how individuals from different professional backgrounds prefer to reason (verbally versus visually, for example), together with how quickly they respond to task demands, can significantly affect cross-functional collaborations. Sometimes members with different occupational identities have trouble understanding one another, for example. Or a member from one occupational background may become highly frustrated with how quickly teammates from other occupational groups are moving.
To further assess the implications of these disciplinary differences on team functioning, Heather Caruso experimentally assessed the kinds of problems encountered in three-person teams composed of one engineer, one visual artist, and one person whose work mainly involved verbal analysis.17 The task was to build three structures from Lego blocks: one that required close attention to structural issues, another for which aesthetic criteria were important, and a third that required compliance with a dense and complex set of building codes. All three disciplinary perspectives, therefore, were critical to a team’s overall success.
Members with different disciplinary backgrounds had strikingly different reactions to the prospect of collaborating with their teammates. The visual artists and verbal analysts responded positively to the opportunity to work together interdependently. The engineers did not, perhaps because they viewed themselves as uniquely qualified to handle “construction” tasks. Indeed, engineers behaved as if exchanges with the other members were mainly a frustrating waste of time. And that, in turn, may have undermined the potential advantages of collaboration even though the team task had been specifically designed to benefit from members’ diverse disciplinary perspectives.
Differences in team members’ training, experience, and network memberships are among the most valuable resources any team has, and therefore merit careful attention by those who select team members. Indeed, the greater the diversity of members’ training histories and disciplinary backgrounds, the greater the potential for individual and team learning—two of the three criteria of team effectiveness previously discussed (Chapter 3). But teams do not harvest those benefits automatically. It requires not just recognition of each member’s special capabilities but also avoidance of the common tendency to denigrate the contributions of people whose experiences and disciplines differ from those of the rest of the team.
Who is on a team makes a big difference in how well members work together and, ultimately, in how well a team performs—especially when the team must deal with a rapidly changing external context, as is characteristic of many intelligence teams.18 But individual attributes are not the whole story. Even if individual team members have more than enough knowledge, skill, and experience, they still must be properly assembled into a team with the right number of members (neither too many nor too few) and a good mix of members (neither too similar nor too different from one another).
Team size of course depends on the requirements of the work to be done—a bigger task, a bigger team. But there are limits. Anthropologist Robin Dunbar many years ago suggested that there is a wired-in upper limit to the number of people with whom one can maintain mutual relationships, which he computed to be roughly 150. The human brain, he said, cannot handle the cognitive processing required to maintain more relationships than that.19
Dunbar’s number raises some intriguing questions about the size and dynamics of social communities and networks—at what size, one wonders, do members begin to drop out of an online community of practice? Work teams have stiffer requirements than online communities. Team members do not merely need stable relationships, they also must maintain members’ active engagement and coordinate their activities in real time. That means that the upper limit for the size of work teams will be smaller, much smaller, than Dunbar’s number.
My own observations suggest that it is not the number of members but the number of links among members that is critical. And, as Figure 6-2 shows, the number of links among members increases at an accelerating rate as group size grows. A team with six members must manage 15 links; one with eight members must manage 28 links; and one with a dozen members must manage 66 links—a near impossibility.20 Groups whose size is in the double digits are almost certain to encounter free-riding or “social-loafing” problems. And they run a real risk of spending nearly as much time coordinating among members as actually performing the team’s work. A former intelligence officer suggested that some foreign intelligence partners perform better than we do in some areas precisely because they do not have as many people available for the work. Another manager reported that his practice was to figure out how many people were needed to accomplish a piece of work and then to compose a team with one or two members fewer than that. Understaffing, he said, motivated members and focused their attention; overstaffing brought nothing but problems.21
FIGURE 6-2 Links Among Members as Team Size Increases
Work teams in organizations generally are substantially larger than they really need to be and therefore encounter problems that they do not need to have. Why does this happen? For one thing, leaders often put too many people on the team in the first place, either to make sure the team has enough members to accomplish the work or to include at least one representative of every constituency with a stake in the outcome. The perverse result can be such an excess of members than the team has little chance to perform at a level that will please those same constituencies.
Moreover, teams sometimes grow as the work progresses. It is common for a team to experience problems and to fall behind, especially in the early stages of its work. At that point, the person who formed the team may get worried and decide to add some people to help the team make up the lost ground. That common organizational phenomenon provided the title for Frederick Brooks’s book, The Mythical Man-Month, a collection of essays about what he learned from leading IBM’s OS/360 systems programming effort. When a project falls behind schedule, Brooks says, the temptation is to compute how far behind it is and then to add staff to get it back on track: If a project is six months behind schedule, six people are added for one month each. Because of the need to bring the newcomers up to speed and reconfigure roles to accommodate them, this strategy almost always has the opposite of its intended effect, prompting the formulation of Brooks’s Law: “Adding manpower to a late software project makes it later.”22
Intelligence community leaders understandably worry that their teams will be under-resourced, especially when the team task is a large, challenging, and consequential piece of work—as it should be (see Chapter 5). So they create teams that are so large that they cannot operate efficiently or well. It does not have to be that way. Sometimes it is possible to create what was described in Chapter 2 as a sand dune team—a larger unit within which smaller groups form and re-form as necessary to perform specific parts of the overall job. Other times it is appropriate to form a small core team that has responsibility for managing the recruitment and coordination of what are, in effect, associate members who are brought aboard only when needed. Other options also can be invented by leaders who are willing to take the time to think carefully and creatively about how best to compose their teams. Large size is not the only option, and it rarely is the best option.
Well-composed teams also have a good mix of members, people who are neither so similar to one another that they duplicate one another’s resources nor so different that they are unable to communicate or coordinate well.23 Imagine, for example, the internal dynamics of an operations team consisting entirely of white male intelligence officers who were trained at the same military facility. That team may operate quite smoothly, but it runs some risk of planning and implementing a course of action that generates unintended negative consequences, outcomes that might have been foreseen by a member from a different background.
Homogeneous groups also risk performing poorly on nonroutine tasks that require original ideas or approaches. Although it assuredly is more pleasant to be in a group of similar others who get along swimmingly well, congeniality and like-minded thinking can compromise critical thinking, suppress learning, and even polarize member opinions. This was neatly illustrated by groups of liberal- and conservativeminded citizens assembled by legal scholar Cass Sunstein and asked to discuss a number of controversial political issues. Like-minded groups, whether liberal or conservative in orientation, became more extreme in their views after group discussion, with pre-discussion differences in members’ opinions “evaporating.”24
Now compare the all-white, all-male, all-military operations team just described with a joint counterterrorism task force composed of members who vary in gender, ethnicity, and disciplinary training—and whose back-home organizations have chronically strained relationships. There is essentially no chance that this group will fall into a groupthink syndrome, but neither is it likely that members will efficiently develop an integrated strategy for executing the team’s work. The standard operating procedures brought by members are so different that disagreements and misunderstandings are almost certain to exacerbate their coming-in worries about the competence and constructiveness of their teammates.
It is difficult for highly diverse groups to draw upon the full complement of member knowledge, skill, and experience precisely because members think and act so differently. Disagreements in such groups risk devolving into interpersonal and intergroup conflicts that alienate members from one another rather than foster original thought and well-coordinated action. Only when group members personally believe in the value of diversity are they likely to identify strongly with a diversely composed team.25
The best state of affairs, of course, is a balanced team, one whose members have a wide variety of resources but who are similar enough to be able to communicate and coordinate well. It takes careful thought by those who create intelligence teams to find the right balance between the poles of homogeneity and heterogeneity.
How much of a difference does having the right people on the team make in how well an intelligence team performs? Even a single member who is a “team destroyer,” someone whose behavior is egregiously disruptive and destructive, can cause a team to fail. Other compositional problems are almost never fatal, however. Teams generally can work around, carry, or find external resources to compensate for a member or two with shaky task or interpersonal skills. Although problems with the size or the mix of members assuredly do cap a team’s upside potential, they rarely are directly responsible for outright team failures.
Even so, a leader’s decision about team composition can strongly shape team dynamics, either smoothing the path toward effectiveness or placing roadblocks in the team’s way. The problem is that teams in the intelligence community too often are formed merely by finding people who happen to be available, assembling them into a group, and assigning them the work that needs to be done. If members happen to have the knowledge, skill, and experience that the work requires, so much the better. But no attempt is made to use what is known about the attributes of prospective team members, or about how well people with different attributes are likely to work together, in deciding who will be put on a team. “You don’t get to pick and choose,” one intelligence community manager told me. “You just take whoever you can get.”
That strategy of team composition, as common as it may be, is far more expedient than effective. It means that team performance will be driven at least as much by the luck of the compositional draw as by informed deliberation about the mix of members that actually is needed for the work to be done. A perhaps more common strategy, and certainly a better one, is to identify explicitly the knowledge and skills that the work requires and then to compose the team so that no critical area of expertise is left uncovered. But that is about as far as it goes and possibly about as far as it ever will go. At least these days, intelligence team leaders do not have a free hand in composing their teams. To find and recruit members who have solid task and interpersonal capabilities, and to assemble them into a team that has the right size and mix of members, can require a leader to overcome nontrivial obstacles rooted in personnel regulations and political realities. It can take some ingenuity and negotiating skill, therefore, to put together a team that has not just the right people, but also the right number and mix of people.
For all of these reasons, those who form teams can find it difficult to take into account the full range of individual differences explored in this chapter when they set up teams in the real world of intelligence organizations. It is, nonetheless, important to make the effort, because getting the right people on a team creates a solid foundation for teamwork. Good composition, however, is only the foundation. The next chapter explores what else must be done to help even a well-composed team recognize and use well the full complement of its members’ capabilities.
