How Leaders Foster Self-Managing Team Effectiveness ... · How Leaders Foster Self-Managing Team Effectiveness: Design Choices Versus Hands-on Coaching Ruth Wageman Dartmouth College,

How Leaders Foster Self-Managing Team Effectiveness: Design Choices versus Hands-onCoachingAuthor(s): Ruth WagemanReviewed work(s):Source: Organization Science, Vol. 12, No. 5 (Sep. - Oct., 2001), pp. 559-577Published by: INFORMSStable URL: http://www.jstor.org/stable/3085999 .Accessed: 30/04/2012 20:59

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Organization Science.

http://www.jstor.org

http://www.jstor.org/action/showPublisher?publisherCode=informs

http://www.jstor.org/stable/3085999?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp

How Leaders Foster Self-Managing Team

Effectiveness: Design Choices Versus

Hands-on Coaching

Ruth Wageman Dartmouth College, The Amos Tuck School, 100 Tuck Hall, Hanover, New Hampshire 03755-9023

ruth. wageman @ dartmouth. edu

Abstract This multi-method field study examines the relative effects of two kinds of leader behaviors-design choices and hands-on coaching-on the effectiveness of self-managing teams. Find- ings show that how leaders design their teams and the quality of their hands-on coaching both influence team self- management, the quality of member relationships, and member satisfaction, but only leaders' design activities affect team task performance. Moreover, design and coaching interact, so that well-designed teams are helped more by effective coaching- and undermined less by ineffective coaching-than are poorly designed teams. (Team Effectiveness; Team Leadership; Self-Managing Teams; Team Coaching)

Considerable research has examined the effects of imple- menting self-managing teams on team performance outcomes and member satisfaction (e.g., Cohen and Ledford 1994, Cordery et al. 1991, Cummings and Griggs 1977, Goodman et al. 1988, Jackson et al. 1994, Macy et al. 1991, Wall et al. 1986). The difficulties of fostering self- management teams-particularly in organizations with histories of individualistic, manager-directed work-have been well documented. Attempts to create self-management teams have often resulted in poor performance, individualistic behavior, and avoidance of the decision making necessary for effective, cooperative teamwork (Cohen and Ledford 1994, Cordery et al. 1991, Cummings and

Griggs 1977, Hackman 1998). These difficulties have been attributed to deficits in the motivation and ability of

managers to create the conditions that foster self-

management (Goliembiewski 1995, Hut and Molleman 1998), as well as to resistance from team members in

taking on self-management (Balkema and Mollerman 1999, Wellins et al. 1991). Much less is known about the kinds of leader activities that surmount these difficulties

1047-7039/01/1205/0559/$05.00 1526-5455 electronic ISSN

to create self-managing teams that both perform well and sustain their self-managing character over time. This paper investigates the separate and joint effects of two quite different kinds of leader activities-team design and hands- on coaching-on the degree to which teams become self- managing and on their performance effectiveness.

The domain of the research is restricted to self- managing teams that have a designated team leader (sometimes called a manager) who is not a regular member of the team. Excluded are teams that have no authority for managing their own performance processes (that is, whose only responsibility is to follow procedures specified by others in pursuit of those others' objectives), and teams with no designated leader or whose leadership is

exclusively informal.

Team Self-Management and Effectiveness Four general functions need to be accomplished when- ever work is performed in a purposive organization (Hackman 1987). First, a person or group must actually execute the work. Second, a person or group must monitor and manage work processes, initiating changes in pace or procedure as needed. Third, a person or group must structure the performing unit and its context, setting up the task of the unit, staffing it, and arranging for organizational resources and supports. And fourth, a person or

group must specify the goals or objectives that are to be

accomplished. A self-managing team, by definition, has authority and

accountability for the first two functions-executing and

managing the work-but within a structure and toward

purposes set by others. Thus, a team's level of formal

authority determines whether or not it falls within the

present domain-that is, whether it is a "self-managing team." The degree to which self-managing team members

actively use their authority to manage their work processes, however, varies from team to team and is a key

ORGANIZATION SCIENCE, ? 2001 INFORMS Vol. 12, No. 5, September-October 2001, 559-577

RUTH WAGEMAN How Leaders Foster Self-Managing Team Effectiveness

dependent variable in this research. Specifically, three behavioral indicators of self-management identified by Hackman (1986) are examined: (1) the degree to which team members take collective responsibility for the outcomes of their work; (2) the degree to which the team monitors its own performance, actively seeking data about how well it is doing; and (3) the degree to which the team manages its own performance, making altera- tions in work strategies when circumstances change or feedback indicates that a new approach may be needed.

Self-management is a behavioral process, and it is entirely possible for a team to be highly self-managing but relatively ineffective, or to be manager-led and to be highly effective. Team effectiveness is therefore also assessed for each team studied. Effectiveness is defined as having three components: (1) task performance-the degree to which the team's product or service meets the needs of those who use it; (2) group process-the degree to which members interact in ways that allow the team to work increasingly well together over time; and (3) individual satisfaction-the degree to which the group ex- perience, on balance, is more satisfying than frustrating to team members (Hackman 1990).

Team Leaders' Actions The two kinds of leader activities explored in this research flow directly from how authority is partitioned for self-managing teams. Self-managing teams, as defined here, do not have the authority to set or alter their purposes, structures, or organizational contexts. One type of leader activity, therefore, is to establish those features (that is, to design the team) in a way that fosters self- management and performance effectiveness. On the other hand, self-managing teams do have the authority to monitor and manage, as well as to execute, their work. A second potentially valuable leader activity, therefore, is to provide hands-on coaching that helps a team manage itself and its work well. These two kinds of leader activities are explicated in detail next.

Design Choices. When a leader designs a team, he or she always uses some model that specifies how the team ought to be set up and what organizational resources and supports ought to be provided. Sometimes the leader's model is explicit, and its implementation is deliberate; other times the model is implicit, and implementation is relatively mindless. Sometimes the leader is proactive, exercising influence with peers or senior managers when he or she does not have sufficient authority to create the desired design; other times the leader just accepts existing organizational conventions and arrangements. And sometimes the design the leader comes up with is a good one, in accord with what is known about the conditions that

foster team effectiveness; other times the leader's model may be significantly flawed.

A growing body of evidence specifies the structural and contextual features that promote team effectiveness. In the present research, I draw upon both that literature and Hackman's (1987) conceptual model of work-team effectiveness to identify four general conditions that, when present, foster self-managing team effectiveness.

(1) A real team. Organizations sometimes attempt to achieve the purported benefits of teamwork by, for example, merely calling people who have similar job res- ponsiblities a "team." Real teams are defined for present purposes as bounded social systems with clear membership that is reasonably stable over time, thereby providing the capability for members to behave as a collective. This design condition, although often violated in practice, is the prerequisite for all the rest.

(2) Clear direction. This is the degree to which the purposes of a team are stated clearly, few enough in number to be memorable to both team members and leaders, and focused on the ends to be achieved rather than on the details of the means to be used in pursuing those ends (Atkinson 1958, Bennis and Nanus 1985, Cohen et al. 1996, Locke et al. 1981, Walton 1985, Zander 1971).

(3) An enabling team structure. Team structure in- cludes five basic design features: appropriate team size- no larger than the minimum required to accomplish the work (Druskat 1995, Hackman and Vidmar 1970, Jackson 1992, Steiner 1972); optimal skill diversity, with substantial heterogeneity of task-relevant skills among members-but not so much that members have trouble coor- dinating their efforts (Ancona and Caldwell 1992, Campion et al. 1993, Druskat 1995, Goodman and Shah 1992, Jackson 1992); task interdependence, such that members are dependent upon one another to accomplish the collective work of the team (Campion et al. 1993, Wageman 1995); challenging task goals with "stretch" performance targets (Cohen and Ledford 1994, Hackman and Oldham 1980, Saavedra et al. 1993); articulated strategy norms, which legitimize and support active strateg- izing and long-term planning by the team, rather than the mindless or reactive execution of the work (Hackman et al. 1976, Woolley in press).

(4) A supportive organizational context. A supportive context provides (a) a reward system that recognizes and rewards excellent team performance (Cohen et al. 1996, Hackman 1987, Rosenbaum et al. 1980, Shea and Guzzo 1987, Wageman 1995), (b) an information system that provides information members require to competently plan their collective work (Abramis 1990, Goodman 1986, Hackman 1987), (c) an education system that is available to provide training or technical consultation for

ORGANIZATION SCIENCE/Vol. 12, No. 5, September-October 2001 560


any aspects of the work that members are not themselves competent to handle (Klaus and Glaser 1970, Liang et al. 1995, Prince et al. 1992, Hackman 1987, Salas et al. 1993), and (d) the mundane material resources that are needed to carry out the work (Hackman 1987).

The four general conditions summarized above are defined in this research by a total of eleven specific design features: real team, clear direction, appropriate size, skill diversity, task interdependence, challenging task objectives, core strategy norms, team excellence recognized/ rewarded, information for planning available, training/ technical consultation available, and material resources available (see Table 1 for a summary of conceptual def- initions of these features). While Hackman's (1987) conceptual model of work team effectiveness was derived from extensive empirical work and is heavily cited in the teams literature, the present research represents the first direct test of the full model. Here, I predict that to the degree that team leaders have put these features in place (or arranged for others to do so), self-managing teams will do better than when these features are absent.

HYPOTHESIS 1. Well-designed teams exhibit more self- management and are more effective than teams whose designs are flawed.

Hands-on Coaching. Both scholars and practitioners who write about self-managing team effectiveness focus a great deal of attention on what leaders do in their day- to-day interactions with team members-that is, they focus on hands-on coaching (e.g., Barry 1991, Bass 1957, Berkowitz 1953, Fiedler 1958, Jackson 1953, Likert 1958, Lippitt 1940, Manz 1986, Manz and Sims 1987, Schlesinger et al. 1960). Coaching refers to direct interaction with the team that is intended to shape team processes to produce good performance. Pervading the literature on team coaching is the view that leader coaching behaviors can directly affect team members' engagement with their task, their ability to work through interpersonal problems that may be impeding progress, and the degree to which members accept collective responsibility for performance outcomes.

Some research evidence supports the view that coaching can improve both the quality of group processes (Kaplan 1979, Schein 1988) and the level of member satisfaction (Cohen et al. 1996, Yukl 1989). Findings regarding the effects of coaching on performance outcomes, however, are, at best, mixed. Some studies have found that operant-based coaching does facilitate team performance (Komacki et al. 1989, Smoll and Smith 1989). Operant coaching refers to the direct reinforcement (generally through positive feedback) either of particular task behaviors when exhibited by the team members or of good performance directly. Operant studies of

coaching often do not specify what team processes are most beneficial for performance, however; one study of operant coaching in which coaches attempted to affect team performance through positive reinforcement of co- ordinating behavior failed to find any relationship between such coaching and performance outcomes (Ko- macki et al. 1989).

Further, Kaplan (1979) reviewed research on the effects of process consultation-a form of coaching aimed specifically at improving the interpersonal relations of team members-on team task performance and found no studies to support the hypothesis that process consultation improves performance (see also Woodman and Sherwood 1980). More recently, Cohen et al. (1996) found that "en- couraging behavior" (providing feedback intended to en- hance team motivation) from supervisors was negatively associated with team performance as assessed both by managers and by customers, and Beekun (1989) found that self-managing teams that had no coaches significantly outperformed those that did.

In sum, existing research evidence suggests that leaders' coaching in some circumstances fosters team self- management, the quality of members' interpersonal relationships, and member satisfaction with the team and its work. But coaching alone (that is, without reference to the quality of a team's design) may make little or even a negative difference in how well a team actually performs. Moreover, it remains an open question as to just what "competent" coaching is. Rather than adopt an a priori conceptualization of coaching behavior, this research captured all nontrivial interactions between teams and their leaders, grouped those interactions into substantive categories, and then empirically assessed the relationships between those emergent categories and measures of team self-management and effectiveness.

HYPOTHESIS 2. Teams that receive coaching exhibit more self-management, higher quality interpersonal relationships, and higher member satisfaction-but not higher task performance-than do teams that receive no coaching at all.

The Interaction of Team Design and Hands-on Coach- ing. There is reason to believe that design and coaching may interact in shaping how, and how well, a self- managing team performs. Specifically, Hackman (1987) has suggested that a leader's influence comes mainly from his or her design choices, with his or her coaching activities making a difference only at the margins through small adjustments in what is an already well-determined trajectory. In this view, leaders have the opportunity to coach a team to higher levels of self-management and superior performance only when the team is relatively well designed. If design conditions are stacked in favor

ORGANIZATION SCIENCE/VOl. 12, No. 5, September-October 2001 561


Table 1 Coding of Team Design Variables

1. A real team Definition: Group has clear membership and is stable over time. Data from: Team interview, manager interview, organizational archives True if: (1) Team membership is clear (true for all teams studied).

(2) Team membership is stable-one or no membership changes in past year.

2. Clear direction Definition: Statement of purpose for the team is clear and is about ends and not means. Data from: Team interview, manager interview True if: (1) Both team and manager can articulate direction.

(2) Three or fewer objectives are stated. (3) Is about end states, not means.

3. Enabling structure

A. Appropriate team size Definition: Number of members is adequate for the work and no larger. Data from: Manager interview, organizational archives True if: (1) Size is between 4 and 7 (from findings of Wageman 1995).

B. Optimal skill diversity Definition: Team is composed of members with substantial heterogeneity of task-relevant skills among members. Data from: Team interview, team survey True if: (1) Skill diversity is within one scale point of survey optimum.

C. Task interdependence Definition: Team has a group task such that members must work together to accomplish it. Data from: Manager interview True if: (1) All five of the following task elements are present:

(a) collective responsibility for customers, (b) team responsibility for parts expenses, (c) collectively designed maintenance practices, (d) frequent required team meetings, and (e) cross-training of members for different machine types (for details, see Wageman 1995).

D. Challenging task goals Definition: Team has objective performance targets that are "stretch" goals. Data from: Team interview, team survey True if: (1) Present objectives require performance exceeding previous levels, with specific deadlines and hard data about

whether or not objectives are accomplished. E. Core strategy norms

Definition: Articulated norms that represent expectations of strategic planning. Data from: Team interview, manager interview, team survey True if: (1) Stories elicited from team and manager about expectations for how team should respond to a change or

unfamiliar problem include (a) convening a problem-solving meeting, and/or (b) initiating change in work practices, and/or (c) soliciting outside assistance.

(2) The score is above midpoint on all four survey items assessing the team's normative expectations of members, regarding (a) experimenting with new work procedures, (b) seeking best practices from other teams, (c) initiating actions to solve problems, and (d) discussing what different members have to contribute to the work.

4. Supportive organizational context

A. Group reward system Definition: Group-level excellence is recognized and rewarded. Data from: Manager interview, organizational archives True if: (1) 80% or more of available rewards are contingent on the team rather than individual performance.



Table 1 (cont'd.) Coding of Team Design Variables

B. Available information Definition: Information system provides the data that the group needs to manage its work. Data from: Team interview, manager interview True if: (1) Task relevant information (e.g., about trends in machine performance, customer feedback, etc.) is given to

team. (2) Information is received once a week or more frequently.

C. Available education Definition: Educational system provides training and technical consultation. Data from: Team is interview, manager interview True if: (1) Team is trained in both technical and interpersonal skills.

(2) Technical and interpersonal training is readily available to team.

D. Material resources Definition: Group is provided with the basic material resources it needs to accomplish the work. Data from: Team interview, manager interview True if: (1) No material resources are needed for the work reported as unavailable by either team or manager.

of good performance, then the team coach can help the team exploit its favorable circumstances. If the team is poorly designed, on the other hand, attempts to foster team effectiveness through hands-on coaching may be fu- tile or even backfire. In such cases, the flawed design may create dysfunctional processes so strong and preoccupy- ing that coaching interventions risk merely adding to the disarray (Wood 1990). Because the present research setting has significant variation both in leaders' design choices and in their coaching behaviors, it offers a unique opportunity to test this as yet untested line of reasoning.

HYPOTHESIS 3. Leaders' design activities and hands- on coaching interact in affecting team self-management and effectiveness, with coaching having a greaterpositive impact for well-designed teams than for poorly designed teams.

Method Thirty-four self-managing teams, split between consistently superb performers and consistently poor performers, participated in the research. Multiple measures of the design features of the teams, team leaders' behaviors, and level of team self-management were obtained using both structured interviews with the teams and their leaders and surveys completed by all team members. Quantitative measures of team performance were obtained from organizational archives.

Research Site The research required that there be variation in the design of teams. It was possible to locate an organization in which a number of design elements of teams were known,

based on prior research (Wageman 1995), to vary widely-namely, the U.S. Customer Services Division of Xerox Corporation. The service organization is divided into nine geographical areas that are in turn subdivided into districts. Each district consists of five to ten subdis- tricts, formed either on the basis of geography or type of machines serviced. Each subdistrict is headed by a field manager to whom 20 to 30 technicians report. The technicians are organized into work teams of between three and nine individuals. The basic task of the teams is to respond to customer calls about machine breakdowns and to initiate visits to customer sites for preventive maintenance.

Team Selection Process Teams were nominated for inclusion in the research by field and district managers. Managers in 12 districts were asked to nominate both superb teams and ineffective teams in their districts. Superb teams were defined as those that (1) consistently meet the needs of their customers, (2) have been operating increasingly well over time, and (3) whose members are engaged in and satisfied with their work. Ineffective teams were defined as those that (1) frequently fail to meet the needs of their customers, (2) have been operating increasingly poorly over time, and (3) whose members are alienated from or dissatisfied with their work. A clerical assistant sorted nominations by district and identified the teams within each district that met two criteria for inclusion: First, that the team had received at least three manager nominations (of a possible nine, on average) as either superb or ineffective, and second, that no team be included that received nominations in both categories (no teams, in fact, did receive conflict- ing nominations).



The assistant gave me a list of teams that met the nomination criteria, with categories removed, so that I could remain blind to condition. I then selected a final sample of 33 teams, of which 18 were superb teams and 15 were ineffective teams. For efficiency of data collection, I included only teams from the seven districts where at least three teams met the nomination criteria. These districts were located in a variety of urban, suburban, and rural settings across the nation.

Data Collection Three research associates (graduate students) and I collected and coded the data. At the time of entry, the researchers knew only which teams in the district they would be studying, and not their nomination category. Two researchers collected data for each team, spending two hours or more conducting a structured team interview that elicited collective descriptions of significant events in the team's life, the team's structure and organizational context, and typical daily interactions with the team's manager. Interviews with managers provided further data about team history, structure, and context, as well as about the manager's own view of his or her interactions with the team. The interview questions, adapted from Hackman (1982), asked for straightforward descriptions of various organizational features, not for interviewees' opinions about whether a team was well or poorly designed. The three research associates and I test-coded the first six interviews, and discrepancies were discussed and resolved to increase interrater reliability for the remaining coding. All subsequent interviews were coded by two members of the research team.

Team members also completed a 108-item survey that assessed team design, quality of group interaction processes, perceived task interdependence, use of problem- solving tools, and member satisfaction with the team and its work. The survey was sent to members by mail and collected at the team interview. Each survey was given a unique code to identify the respondent's team, and teams sealed their completed surveys together in an envelope before handing them to the researchers. Thus, individual team member responses were not identifiable either to the team's leader, the team members, or the researchers. Re- sponse rate to the survey was 92%, and I had at least three-fourths of the team members' responses from each team in the research. Finally, organizational records were consulted for information about team membership, dis- tribution of financial rewards, and objective team performance.

Measures Data from multiple sources using different data collection methodologies was drawn upon in constructing all measures. The measure development strategy, measure descriptions, and psychometric data are reported below; the

team survey, team and manager interview protocols, and details of coding procedures can be obtained from the author.

Team Design. The 11 theory-specified design features previously described were coded from the team and manager interviews, supplemented by data obtained from the team survey and from organizational archives. Each of the design features was coded dichotomously as either being in place or not.' Coding procedures were highly conservative, with a feature coded as "present" only if there was convergence about its presence across multiple indicators. In the event of a disagreement between the codes assigned from team and manager interview data, the team's survey score for the variable in question was consulted. In the six cases that fell into that category, the design feature was considered to be present for the team only if its survey score was above the mean, and if descriptive data from either the manager or the team interview unambiguously indicated that the feature was, in fact, in place. Intraclass correlations were computed for all team-level variables to ensure that they could appro- priately be aggregated ; ICCs for survey measures of the team design features ranged from 0.18 to 0.41, all significant at p < 0.05. Overall quality of team design was computed as the total number of design conditions in place for a team, from zero (no conditions present) to 11 (all conditions present).2 Coding conventions are summarized in Table 1, and descriptive statistics and inter-rater reliabilities of design variables are included in Table 2.

Correlations among the individual design conditions from the dichotomous coding range from -0.14 to 0.48, with the highest being between task goals and strategy norms, and between clear direction and task interdependence (see Table 6). These correlations probably are not attributable to confounds introduced by data collection methods because each condition was coded independently from descriptive accounts of work practices rather than from self-report ratings. It is unlikely, for example, that team and managerial descriptions of the amount of interdependence in specific task elements would be spu- riously affected by whether the team also had a clear and engaging direction. Two explanations for these correlations are more likely. First, teams may not achieve excellence unless many of the design features are simultaneously present; consequently, the selection process may have yielded only groups with either many or few of the design features in place. Second, managers who create favorable designs for their teams may tend to put multiple features in place rather than just one or two. This pattern of cooccurence of design features is consistent with Hackman (1990).



Table 2 Team Design Descriptive Statistics

Data Source

Manager Interview Group Interview Survey

r with other r with other r within r with other Team Designttt M SD IRR* design features** M SD IRR* design features** M SD scale (o)t design featurestt

Direction 0.61 0.50 0.94 0.37 0.54 0.51 0.92 0.20 4.96 1.12 0.44(0.77) 0.21 Optimal diversity 0.28 0.49 0.84 0.08 2.04 0.98 0.09 Appropriate size 0.48 0.31 1.00 0.22 5.41 1.32 0.17 Stability of membership 0.34 0.52 1.00 -0.01 Task interdependence 0.56 0.50 1.00 0.29 2.40 0.51 0.29(0.61) 0.11 Task goals 0.47 0.50 0.89 0.04 0.56 0.39 0.90 0.07 4.09 1.13 0.58(0.78) 0.33 Core strategy norms 0.62 0.41 0.91 0.11 0.55 0.51 0.86 0.41 3.69 0.61 0.40(0.87) 0.19

Group rewards 0.63 0.49 1.00 0.20 0.52 0.48 1.00 0.20 4.99 1.06 0.32(0.69) 0.25 Information supports 0.33 0.48 1.00 0.06 0.29 0.50 1.00 -0.08 5.30 1.00 0.49(0.83) 0.26 Education supports 0.71 0.46 0.99 0.12 0.57 0.50 0.95 0.09 5.19 1.05 0.49(0.80) 0.22 Material resources 0.70 0.50 0.98 0.20 0.73 0.51 0.94 0.13 5.30 1.04 0.46(0.77) 0.30

*Interrater reliabilities **Average correlations with all other team design variables measured using same source (manager or team interview)

tAverage correlation of items within scale and (internal consistency reliability, i.e., Cronbach's alpha) ttAverage correlations with items from other scales within the same survey section tttFor overall team design, which combines both across all design dimensions and across sources, M = 6.12, SD = 1.39, and IRR = 0.90.

Leader Coaching. Coaching behavior was coded from several questions asked in both the team and the manager interviews. The team versions of these questions were: "How often is your manager with the whole team? With individuals? How does s/he use time with the team versus with individuals? How frequently does s/he do those things?" Rather than use an a priori coding scheme, I allowed categories of behaviors to emerge from the behavioral descriptions that leaders and teams provided in

response to these questions. First, each behavior mentioned in the transcripts of the

team and managerial interviews was marked. These behaviors were then sorted into related categories-e.g., "showing us performance shortfalls in reports" and "tell-

ing the team when reliability is slipping" were catego- rized as "identifying the team's problems." Ten such behavior categories were identified. Four that had little or no variance-that is, for which all managers engaged in the behavior or only one or two managers engaged in the behavior-were dropped.3

Each behavior category was then coded for frequency on a scale from 1 = low to 3 = high for each team leader, based on team and manager descriptions of how often that behavior occurred. If a particular behavior category was not identified either by the team or the leader as some-

thing the manager did, it was coded as low. If both the

team and the leader identified a behavior as something the leader did at least once per month, it was coded as

high. If the behavior was identified by the leader or the team but not both, or if it was identified as happening less

frequently than once per month, it was coded as moderate. The final behavior categories used in the analyses are listed below; their means, standard deviations, and reliabilities are presented in Table 3.

(1) Providing informal rewards and other cues that the

group-as-a-whole is responsible for managing itself (e.g., rewarding the group for solving a problem, spending more time in interaction with the group-as-a-whole than with individual members).

(2) Broadening the group's repertoire of problem- solving skills through appropriate problem-solving consultation (e.g., teaching the group to use a problem-solving process; facilitating problem-solving discussions).

(3) Dealing with interpersonal problems in the team

through team process consultation (e.g., bringing up intra- team conflicts for discussion).

(4) Signaling that individuals (or the manager her/him-

self) are mainly responsible for managing the team's work (e.g., by spending more time with individuals than with the group; by running group meetings rather than

teaching the group how to run its own meetings). (5) Intervening in the task (e.g., monitoring call rates



and asking a specific member to take a particular call; dealing directly with a team's customer without involving the team).

(6) Identifying the team's problems (e.g., pointing out overexpenditures on parts; showing the group a down- ward trend in its response time).

Even though these dimensions of coaching behavior were derived entirely from what teams and their leaders in this organization identified as common coaching behaviors, they bear a strong resemblance to others in the literature on self-management-and especially to those of Manz and Sims (1987). These authors also identify the "leader behaviors" of positive feedback, problem-solving consultation, intervening as a "foreman," and process consultation. Thus, this method of assessing leader coaching appears to be both ecologically valid for this research setting and consistent with other research on the coaching of self-managing teams.

Self-Managing Behaviors. The level of team self- management, a key dependent variable in the research, was coded from the team interview. Three theory- specified components of self-management (collective responsibility for work outcomes, monitoring own performance, and managing own performance) were coded separately (as high, medium, or low) and then averaged to form an overall measure of a team self-management (Cronbach's alpha = 0.94). The behaviors coded for each component, along with means, standard deviations, and intercoder reliabilities, are shown in Table 4.

Groups that scored high on collective responsibility (n = 14) were those that described multiple instances of initiative on a customer's or team member's behalf and reported no instances of members avoiding responsibility-such as one group whose members came in on days

Table 3 Leader Coaching Behavior Descriptive Statistics

Data Source Manager and Group Interviews

r with Coaching behaviors M SD IRR* others**

Cues/rewards for self-management 1.82 1.49 0.96 0.15 Problem-solving consultation 1.48 1.63 0.86 -0.20 Process consultation 2.41 1.02 0.91 0.13 Negative signals 1.93 1.04 0.94 -0.06 Intervention in the task 1.55 1.06 0.92 0.13 Identifying team problems 1.38 1.09 0.98 0.06

*Interrater reliabilities **Average correlations with all other coaching behaviors measured

off to take calls when the call rate became too high for remaining members to handle. Groups that scored low in responsibility (n = 12) described multiple instances of members avoiding responsibility-such as a team that was unable get its members to take calls from particularly difficult customers. Groups that described instances of both high and low responsibility were coded as moderate (n = 8).

Measurements of active monitoring were coded principally from responses to the question, "What kinds of data does your group receive about its performance? And how do you use these data, if at all?" Groups that were coded high on monitoring (n = 12) actively sought (or generated) data about their performance-such as one team that requested regular reports about its 20 poorest performing machines to identify common causes of problems. Groups that scored low on monitoring (n = 16) ignored or refused performance-relevant data-such as one whose members reported that company reports about machine reliability were routinely discarded unread. Groups moderate in monitoring (n = 6) reported that they did sometimes use company-provided data, but never sought additional information.

Finally, measurements of managing own performance were coded mainly from responses to three interview questions: (1) "Describe a time when your group delib- erately took time out from its regular work to discuss better ways of operating;" (2) "Describe a time when your group should have taken time out but did not do so;" and (3) "Describe a time when you felt you could not act as a team." Groups coded as high on managing (n = 11) described multiple instances of autonomously convening problem-solving meetings and altering their performance strategies-such as a team that redesigned standard maintenance practices to fix a machine reliability problem. Groups coded as low (n = 10) tended to let chronic problems ride unaddressed. Groups coded as moderate on managing (n = 13) were those that neglected some on- going problems but altered their performance strategies to deal with other problems or opportunities. One group, for example, had redesigned its call-queuing strategy to improve response time, but had not addressed (and had no active plans to address) a long-standing reliability problem.

Team Effectiveness. Team effectiveness was assessed using (1) archival data about group performance, (2) survey measures of the degree to which members interact in ways that over time increase their ability to work together, and (3) survey measures of individual work satisfaction. Performance data were collected from company records for the six months prior to and six months after field data



Table 4 Measures of Level of Team Self-Management

r with other

Components M SD IRR SMBs Behaviors Coded

Collective 2.07 0.98 0.94 0.78 (1) Took time and initiative to help other members or to solve customer

Responsibility problems not in the group's purview.

(2) Avoided work that was formally the team's responsibility (reverse scored).

Monitoring Own 1.88 0.98 0.95 0.85 (1) Sought or generated data about aspects of team performance. Performance (2) Ignored or discarded data about aspectsof team performance (reverse

scored).

Managing Own 2.03 0.95 0.92 0.85 (1) Altered performance strategies to solveproblem without managerial Performance directive.

(2) Had long-term problem still not addressed (reverse scored).

(3) Had never taken time from regular work to discuss better ways of operating(reverse scored).

Composite Measure 1.99 0.75 0.92

collection.4 These 12 monthly assessments of team performance were averaged to generate summary measures for the following five dimensions.

(1) Customer satisfaction with service, derived from a customer survey distributed periodically by the organization. Survey ratings range from 1 ("very dissatisfied") to 5 ("very satisfied"). The customer service measure was not computed for the 11% of teams that had received fewer than 10 surveys during the 12-month assessment

period. (2) Parts expense, computed as the expected cost of

parts (based on machine type and level of usage) divided

by actual expenditures for parts. (3) Response time, computed as the percentage of calls

taken by the team within specified time limits. This measure reflects how quickly a team gets to customers who have malfunctioning machines.

(4) Repair time, computed as the amount of repair time a machine is expected to require, based on historical data, divided by the actual number of minutes spent repairing it. Repair time assesses technical proficiency, how

quickly teams can fix problems. (5) Machine reliability, computed as the number of re-

pair calls expected on a machine, based on historical data, divided by the actual number of repair calls. Machine

reliability assesses how well the machines are maintained. An overall performance measure was generated by tal-

lying the number of these five dimensions for which a team scored above Xerox norms. This measure ranges from 0 (low performing on all criteria) to 5 (high performing on all criteria).

Quality of group process was assessed from survey data, using seven items adapted from Allmendinger et al.

(1992). Team members rated, on a scale ranging from 1

("strongly disagree") to 7 ("strongly agree") their agree- ment with statements describing positive and negative group processes (e.g., "Every time someone tries to

straighten out a work group member whose behavior is not acceptable, things seem to get worse rather than bet-

ter"). Scores on the seven items were combined into a

composite measure of process quality for each team. Individual work satisfaction also was assessed from

survey data, using three job satisfaction items from the Job Diagnostic Survey (Hackman and Oldham 1974). Each item (e.g., "Generally speaking, I am very satisfied with this job") used a seven-point agree-disagree scale. Scores on the three items were combined into a composite measure of individual work satisfaction for each team.

Psychometric data for measures of team performance, quality of group process, and individual work satisfaction are presented in Table 5. All variances and internal con-

sistency reliabilities are acceptable. In addition, to ensure that the composite scores are meaningful group-level measures, intraclass correlations were computed for both the quality of group process and the work satisfaction

measures; these coefficients were 0.23 and 0.29, respectively, both significant at p < 0.01.

Intercorrelations among all measures used in the research-team design, leader coaching, self-management, and team effectiveness-are presented in Table 6.



Results To assess the separate and joint effects of the six measures of leader coaching, a preliminary analysis regressed each of the three measures of team self-management (collective responsibility, monitoring own performance, and managing own performance) on overall quality of team design, the six coaching measures, and their interactions, with all variables entered simultaneously. Two types of coaching (providing cues and informal rewards for self- managing behaviors, and problem-solving consultation) contributed positively and significantly to self-management, whereas two other types of coaching (identifying team problems, and leader task intervention) contributed negatively to self-management. Consistent with previous findings, leaders' interpersonal process consultations had no

significant effects. Therefore, the two coaching measures with positive effects on self-management were combined into a measure of "positive coaching," (alpha = 0.77) and the two measures with negative effects were combined into a measure of "negative coaching" (alpha =

0.56). Principle components analysis confirmed that these two summary indices represented two orthogonal dimensions (Eigen values >1.0). Further, preliminary inspec- tion of patterns across the six leader coaching behaviors in interaction with design suggested that there were two distinct interaction patterns for these indices. Conse- quently, these summary measures are used in all subsequent analyses.

The main analyses, in which dependent variables are regressed on team design, positive coaching, negative coaching, and the two-way interactions of positive and negative coaching with team design are presented in Ta- ble 7. According to the adjusted r-squares, the design and coaching variables together account for 77% of the variation in overall self-management, 39% of the variance in objective team performance, 50% of the variance in the quality of member relationships, and 38% of the variance in individual work satisfaction. These are large effects, especially given that the predictors and the dependent variables were based on data from different sources and collected using different methods.

Effects of Design Conditions Hypothesis 1, that well designed teams exhibit more self- management and are more effective than teams whose designs are flawed, is supported. The summary measure of team design contributed positively and significantly to overall level of team self-management (adjusted r-square = 0.42), to objective group performance (adj. r-square = 0.37), and to quality of group process (adj. r-square = 0.12), but not to individual work satisfaction (p = 0.13). Regression of the 11 specific design features on

Table 5 Team Effectiveness Descriptive Statistics

Variable Data Source

Archival Data

r with other perf.

Team performance M SD measures*

Customer satisfaction 4.31 1.35 0.16 Parts expenses 106.82 22.49 0.30

Response time 86.67 9.59 0.35

Repair time 104.03 11.19 0.01 Machine reliability 101.90 11.81 0.22 Overall group performance 2.21 1.38 0.19

Survey

r within rwith M SD scale (alpha) other scales**

Quality of group process 4.51 1.33 0.51 (0.85) 0.25 Member satisfaction 5.06 1.25 0.59 (0.75) 0.28

*Average correlations with all other variables within the same category **Average correlations with items from other scales within the same survey section

overall self-management show that the major effects of design are due to clear direction (t = 2.64, p < 0.05), task interdependence (t = 2.51, p < 0.05), group rewards (t = 2.31,p < 0.05), and strategy norms (t = 2.01,p <

0.05). Regression of the individual design features on task

performance show that this relationship is principally due to clear direction, task interdependence, and group rewards (t values of 2.67, 2.56, and 2.15, respectively, all p < 0.05). A parallel analysis for quality of group process shows that this relationship is mainly due to the effects of clear direction (t = 2.67, p < 0.05), group rewards (t = 2.35, p < 0.05), strategy norms (t = 2.14, p < 0.05), and challenging task objectives (t = 1.80, p = 0.07). Chi-square analyses were conducted assessing the relation of the dichotomous selection criterion (effective vs. ineffective) on key dependent variables and are shown in Table 8. These analyses confirm the regression findings.

Effects of Leader Coaching Hypothesis 2 states that well-coached teams exhibit more self-management, higher quality interpersonal relationships, and higher member satisfaction-but not higher


0 o 0 z

0

CO Z

n m Table 6 Correlations Among Design Factors, Coaching Behaviors, and Dependent Measures*

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Team design 1. Direction 2. Optimal diversity 0.08 3. Appropriate size -0.03 0.10 4. Stability of memb. 0.11 0.33 0.21 5. Task interdependence 0.46 -0.13 0.11

-0.04 6. Task goals 0.12 0.00 0.06 -0.14 -0.03 7. Core strategy norms 0.05 0.10 -0.01 -0.14 0.19 0.48 8. Group rewards 0.22 0.41 -0.07 -0.12 0.22 0.30 0.19 9. Available information 0.37 -0.12 0.06 -0.11 -0.11 -0.03 0.17 -0.07

10. Available training -0.03 0.22 0.26 0.02 -0.02 -0.14 0.01 0.01 0.17 11. Material resources 0.34 -0.05 0.04 -0.01 0.11 0.03 0.27 0.32 0.36 0.11

Coaching behaviors 12. Cues/rewards for s-m 0.23 -0.04 0.02 0.09 0.27 0.43 0.46 0.04 0.19 -0.09 0.32 13. Problem-solving cons. 0.33 0.07 0.08 0.11 0.35 0.53 0.61 0.30 0.18 0.18 0.24 0.62 14. Process consultation -0.02 -0.08 0.03 0.00 0.06 0.26 0.30 0.23 -0.26 0.25 0.08 0.34 -0.40 15. Negative signals -0.15 -0.45 -0.27 -0.12 0.09 -0.20 -0.25 -0.31 -0.08 0.21 -0.08 -0.05 -0.49 0.36 16. Intervention in the task -0.06 0.50 -0.02 0.12 0.01 0.02 -0.14 -0.07 -0.02 -0.02 -0.02 -0.06 -0.43 -0.03 0.02 17. Identifying problems -0.16 0.32 0.22 -0.01 -0.23 0.23 0.16 -0.04 0.09 0.11 0.02 -0.12 -0.18 0.13 0.21 0.19

Dependent measures 18. Collective responsibility 0.22 0.01 0.11 -0.02 0.38 0.09 0.36 0.55 0.08 0.26 0.51 0.32 .021 -0.18 -0.11 -0.22 -0.09 19. Monitoring performance 0.36 0.08 0.10 0.06 0.44 0.26 0.49 0.40 0.31 0.23 0.23 0.40 0.28 0.11 0.02 -0.44 -0.39 0.79 20. Managing performance 0.41 0.06 0.12 -0.01 0.47 0.21 0.57 0.57 0.19 0.11 0.30 0.21 0.58 0.12 -0.11 -0.42 -0.29 0.77 0.87 21. Overall self-management 0.29 0.04 0.12 0.00 0.44 0.19 0.56 0.52 0.23 0.19 0.48 0.35 0.55 0.07 -0.32 -0.32 -0.34 0.78 0.85 0.85 22. Overall performance 0.56 0.25 0.19 0.15 0.43 0.16 0.45 0.52 0.22 0.05 0.41 0.07 0.20 0.17 -0.34 -0.36 -0.09 0.32 0.44 0.58 0.44 23. Quality of process 0.09 -0.02 0.12 0.17 0.25 0.31 0.70 0.33 0.16 0.05 0.28 0.47 0.40 0.05 -0.34 -0.07 -0.30 0.72 0.55 0.78 0.72 0.19 24. Member satisfaction 0.43 0.10 -0.12 0.11 0.41 0.40 0.24 0.52 0.29 0.04 0.31 0.27 0.51 0.06 -0.14 -0.37 -.09 0.39 0.49 0.49 0.48 0.60 0.32

*AII correlations are based on group-level data. Correlations greater than (0.37) are significant at p < 0.05.

p

C)

t)

("I

n

D

_t

L~

(,/

O\ 11D


task performance-than do teams that receive poor coaching. This hypothesis also is supported. Positive coaching contributes positively and significantly to overall level of team self-management (r-square = 0.08) and to quality of group process (r-square = 0.18), but not to either group performance (p = 0.75) or work satisfaction (p = 0.87). Negative coaching contributes negatively to overall self-management (r-square = 0.11) and to work satisfaction (r-square = 0.30), but not to either group performance (p = 0.37) or quality of group process (p = 0.36).

Interaction of Design and Coaching So far we have seen that team design conditions are strongly and positively related to level of team self- management, how well teams perform, and the quality of teams' interpersonal processes. By contrast, the findings for leaders' coaching behaviors differ in that (a) level of self-management is relatively weakly predicted by coaching (average r-squares of 0.46 and 0.06 for design and coaching, respectively), (b) negative coaching is the only predictor to influence member satisfaction, and (c) neither positive nor negative coaching is related to objective team performance. These findings corroborate Hackman's (1990) qualitative findings exploring similar relationships.

We turn now to Hypothesis 3, which states that leaders' design activities and hands-on coaching interact in affecting team self-management and effectiveness. Specif- ically, it is predicted that competent coaching has a greater positive impact for well-designed teams than for poorly designed teams. The interaction terms reported in Tables 7 and 8 show that this hypothesis is also supported for level of self-management and for quality of group process, but not for objective task performance or member work satisfaction.

For level of team self-management, the interaction between design and coaching has a markedly different form for positive and negative coaching behaviors. As is seen in Figure 1, leaders' positive coaching behavior enhances team self-management more when teams are well designed than when they are not. Figure 2, by contrast, shows that negative coaching behavior undermines team self-management more for poorly designed than for well- designed teams.

Findings for measures of team interpersonal processes exactly parallel those for level of self-management. Posi- tive coaching has a stronger positive effect on process quality in well-designed groups than in poorly designed groups, and ineffective coaching undermines the interpersonal processes of poorly designed teams more than those of well-designed teams.

Table 7 Effects of Design and Coaching on Self-Managing Behavior, Group Performance, Group Process, and Member Satisfaction*

Overall Self-Management Overall Adj. R2 = 0.77 adj. r-

Independent Variable B t p square

Team design 4.492 4.09 0.00 0.42 Positive coaching 0.523 2.24 0.01 0.08 Negative coaching -0.353 -2.42 0.02 0.11 Design x positive coaching .606 1.99 0.05 0.06 Design x negative coaching - 0.687 - 2.33 0.03 0.09

Group performance Overall Adj. R2 = 0.39 adj. r-


Team design 0.601 2.59 0.01 0.37 Positive coaching 0.129 0.32 0.75 0.00 Negative coaching -0.238 -0.91 0.37 0.00 Design x positive coaching 0.712 0.73 0.47 0.00 Design x negative coaching -1.37 - 1.53 0.13 0.02

Quality of group process Overall Adj. R2 = 0.50 adj. r-


Team design 0.305 2.33 0.03 0.12 Positive coaching 0.437 3.08 0.01 0.18 Negative coaching -0.138 -0.94 0.36 0.00 Design x positive coaching 0.603 1.88 0.05 0.07 Design x negative coaching -0.805 -2.47 0.02 0.13

Member Satisfaction Overall Adj. R2 = 0.38 adj. r-


Team design 0.183 1.00 0.13 0.02 Positive coaching 0.033 0.17 0.87 0.00 Negative coaching -0.441 -2.71 0.01 0.30 Design x positive coaching 0.015 0.03 0.97 0.00 Design x negative coaching -0.598 - 1.37 0.18 0.04

For all analyses, n = 33

In sum, the effects of leaders' coaching behaviors depend substantially on how well they have designed their teams: Effective coaching helps well-designed teams more than poorly designed teams, and ineffective coaching undermines poorly designed teams more than well- designed teams.

Discussion The data presented here demonstrate both strengths and weaknesses for addressing team leader activities and their



Table 8 Relationships of Design Conditions with Team Effectiveness

Nomination Category

Design Condition Superb Ineffective X2 p

Clear direction Yes No

Task interdependence Yes No

Group rewards Yes No

Strategy norms Yes No

Material resources Yes No

Challenging task goals Yes No

Appropriate Size Yes No

Information Yes No

Member diversity Yes No

Real Team Yes No

Education Yes No

17 1

1 14

15 3

15 3

16 2

16 2

2 13

4 11

5 10

7 8

13 5

9 9

8 10

5 13

7 11

5 10

7 8

5 10

4 11

5 10

12 6

9 6

Figure 1 Interactive Effects of Quality of Team Design and Positive Leader Coaching on Team Self- Management

High 3.0

c

19.90 <0.01 S

? Moderate 2.0

12.46 <.01 I

40

8.92 <0.01 Low 1.0

1.0 Low

8.74 <0.01

Well-designed teams

Poorly designed teams

2.0 Moderate

Positive Leader Coaching

3.0 High

4.77 <0.05 Figure 2 Interactive Effects of Quality of Team Design and Negative Leader Coaching on Team Self- Management

2.99 =0.08

1.87 n.s | S

0.160 n.s. ct

0.099 n.s. 0 0.099 n.s.

High 3.0

Moderate 2.0

Well-designed teams

Poorly designed teams

Low

0.082 n.s.

0 n.s.

Note. Cell contents are the numbers of superb vs. ineffective teams for which each design condition was in place vs. not in place. x2

Reported are continuity-adjusted.

effects on self-managing teams. Among the strengths is

the fact is that the measures of design features are derived from coded descriptions of actual organizational features, rather than from team-member ratings of design quality. This method of data collection is quite rare in the litera-

ture, and significantly adds to the trustworthiness of these

1.0 2.0 3.0 Low Moderate High

Negative Leader Coaching

measures. In addition, independent data sources provided assessments of team performance, quality of interpersonal processes, and team member satisfaction (archival and

survey sources, respectively). As a consequence, the find-

ings of this research were protected from problems per- vasive in teams research: Namely, relations between assessments of team effectiveness and assessments of

design features and leader coaching are potentially attributable to overall positivity or negativity biases on the part of respondents. Moreover, the findings present the first direct test of Hackman's (1986) model of work team ef-

fectiveness, and provide confirmation for the model. This research has several weaknesses, as well, that call



for caution in the interpretation of findings. First, the sample of teams was small, and thus interpretations of findings, especially analyses involving the numerous design conditions entered separately, must be drawn with caution. Also, measures of the frequency of effective and ineffective coaching behaviors may not fully capture the overall quality of coaching, by leaders, that teams receive. Finally, the selection process in this research necessarily produced teams that tended to be extremely effective or extremely ineffective, limiting the opportunity to explore curvilinear effects, and also limiting the degree to which conclusions might be drawn about teams that are only moderately effective on all criteria.

Nonetheless, the findings of this research have implications for three issues about the dynamics of self- managing teams: (1) the degree to which self-managing behaviors by team members mediate the effects of design and coaching; (2) how team design, leader behavior, self- management, and team performance can evolve into a system whose parts are mutually reinforcing; and (3) the conditions under which team leaders can have the greatest influence on team effectiveness. Below, I draw upon the pattern of findings in this research to explore each of these questions in turn.

Self-Managing Behaviors as Mediators Self-management may-or may not-mediate the relationship between team design and coaching, on the one hand, and team effectiveness on the other. It could be, for example, that team design and leader coaching are con- sequential only to the extent that they affect the degree to which members take on high levels of responsibility for the work, engage in monitoring performance over time, and develop high-quality work strategies. Alterna- tively, design and coaching may influence team effectiveness only partly (or not at all) through the team's self- managing behavior. For example, team composition, a design feature, may have positive effects both on how members work together and on the team's capability in actually executing the work (Ancona and Caldwell 1992, Druskat 1995, Goodman and Shah 1992). In this case, performance effects associated with composition would be due in part to the effects of composition on team self- management and in part to the quality of members' task execution.

The present data allows an exploratory test of the degree to which self-managing behavior does mediate the impact of design and coaching on team effectiveness. Ac- cording to Baron and Kenny (1986), three conditions must be met to demonstrate mediation: (1) The independent variables (in this case, team design and leader coaching) must significantly predict the proposed mediator

(self-managing behavior); (2) the mediator must significantly predict the dependent variable (performance, group process, and satisfaction); and (3) when the effects of the independent variable and mediator are entered simultaneously, the contribution of the independent variable should drop substantially for partial mediation and to nonsignificance for full mediation.

The first of these three conditions is met (see Table 7). The strong relationship between team self-management and team effectiveness shown in Table 9 (Model 1) estab- lishes that the second condition is also met. To assess the third condition, team design and leader coaching are entered into the regression simultaneously with level of self- management. Comparison of Model 2 with Model 1 in Table 9 shows that the relative contributions of the predictors do change substantially when this is done. For team performance, the contribution of self-management becomes nonsignificant, with only team design predicting performance. This shows that team design influences performance independently of its effects on self-management. For quality of group process, however, self-management remains a significant predictor-but coaching does not.5 This pattern suggests that the effects of leader coaching on quality of group process are mediated by their effects on self-management, in contrast to team design, which affects process quality independently of its influence on team self- management. Finally, for work satisfaction, level of self- management becomes nonsignificant, but negative leader coaching remains marginally significant. This pattern suggests that ineffective coaching undermines work satisfaction in part through its tendency to reduce team members' self-management-but that ineffective coaching also undermines work satisfaction directly.

In sum, self-managing behaviors do appear to mediate the effects of team design and coaching to some extent- but differently for the three components of effectiveness. Self-management fully mediates the relationship between coaching and the quality of members' interpersonal processes; it accounts partially for the effect of coaching on work satisfaction; but it plays no mediating role at all in the relationship between team design and objective group performance. Indeed, self-management never fully accounts for the effects of quality of team design on any of the three components of effectiveness. This finding suggests that overall quality of team design is equally important for teams, regardless of the level of behavioral self-management.

Dynamic Relations Among Design, Coaching, Process, and Performance The present findings may call into question conventional understanding about the relationships among leader



Table 9 Test of Self-Management as a Mediator of the Effects of Team Design and Leader Coaching Behavior on Team Effectiveness

Model 1 Model 2

Team performance Overall Adj. R2 = 0.19 Team performance Overall Adj. R2 = 0.44

Independent variable B t p adj. r-square Independent variable B t p adj. r-square

Self-management 0.795 2.90 0.01 0.19 Self-management 0.224 0.41 0.68 0.01 Team design 1.691 2.03 0.05 0.39 Positive leader behavior 0.246 0.49 0.63 0.02 Negative leader behavior -0.318 -0.57 0.34 0.02

Quality of group process Overall Adj. R2 = 0.52 Quality of group process Overall Adj. R2 = 0.57

Independent variable B t p adj. r-square Independent Variable B t p adj. r-square

Self-management 0.723 5.45 0.00 0.52 Self-management 0.717 3.47 0.00 0.33 Team design 0.289 1.50 0.04 0.09 Positive leader behavior 0.134 0.99 0.33 0.04

Negative leader behavior 0.055 0.49 0.63 0.01 Member satisfaction Overall Adj. R2 = 0.24 Member satisfaction Overall Adj. R2 = 0.44

Independent variable B t p adj. r-square Independent variable B t p adj. r-square

Self-management 0.559 3.02 0.01 0.24 Self-management 0.146 0.55 0.59 0.01 Team design 0.521 1.97 0.05 0.17 Positive leader behavior 0.009 0.04 0.96 0.00

Negative leader behavior -0.295 -1.97 0.06 0.13

*For all analyses, n = 33.

coaching behavior, team design, team processes, and team performance effectiveness.

Team and Leader Behavior. When teams in this study were well designed, effective coaches tended to have a more positive influence on team processes than they did when teams were poorly designed. Moreover, well-

designed teams appeared more robust-that is, ineffective coaching behavior did not undermine them nearly as much as it undermined teams with flawed designs. Thus, the impact of leaders' coaching on their teams is condi- tioned by the way in which they set the team up in the first place.

However, not only does the impact of leader coaching behavior depend upon team design, but particular leader

coaching behaviors may also be elicited by quality of team design. Recall that when leaders pointed out work

problems to a team and when leaders intervened in the

task, teams were less likely to manage their own performance. These patterns raise the question of whose behavior is influencing whose. One possibility-the one that

predominates in the team leadership literature-is that the

coaching behavior of the leader drives the self-managing behavior of the team. Thus, in the examples cited above, the leaders' behaviors may prompt team members to at- tribute authority not to the team, but to managers, and thereafter they may take responsibility only for task ex-

ecution, leaving the managing to managers. An alternative explanation is that leaders monitor the

performance and manage the tasks of poorly designed teams because members themselves are not doing so. Leaders may respond to low team self-management and

poor performance by monitoring team performance closely, by increasing their own interventions in the work of their teams, and by providing fewer cues and rewards for team self-management. By contrast, well-designed teams are highly likely to take on management functions

themselves, making it unnecessary for the leader to do so. Thus, the degree of effective self-management by teams, itself influenced by the design of the team, may be shaping the coaching behavior of team leaders. This



possibility, which is consistent with other research on how subordinate behavior can shape leader style (e.g., Farris and Lim 1969, Lowin and Craig 1968), merits additional investigation in the specific case of self- managing teams.

Design, Self-Management, and Performance. Over time, design factors, team self-management, and team performance may become so interdependent that they set in motion a self-reinforcing spiral. In the present research setting, team design is largely in the hands of the teams' immediate managers. These managers could redesign rewards, alter tasks, articulate direction, and provide resources to teams at their discretion. Many team leaders did so-and the better the design conditions they provided, the more their teams were self-managing and the better they performed. But leaders' decisions to provide better design conditions may themselves be influenced by prior team performance. For example, teams well designed enough to perform adequately are more likely to be given additional authority over their work, more support resources, and/or more challenging goals (e.g., Ancona and Caldwell 1992). By contrast, teams with few support conditions tend not to use the authority they do have-nor do they perform well. Leaders may be under- standably reluctant to bestow even more resources on those teams, even though that might be just what is needed to remedy their performance problems.

Thus, design, self-management, and performance may operate as self-reinforcing spirals, wherein already well- designed teams manage themselves effectively, receive even more organizational support, and thus become better self-managing performing units over time (Lindsley et al. 1995). By contrast, poorly designed teams that manage themselves ineffectively may receive from their leaders fewer of the very supports and resources that contribute to team effectiveness-and even risk having withdrawn some of the positive design features that they presently enjoy.

Under What Conditions Do Leaders Affect Team Outcomes? The findings of this research suggest that team behavior and performance may be most affected by structural, technological, and contextual factors-factors that often lie beyond team leaders' direct control. These exogenous factors can significantly constrain the variance in how teams are managed within organizations and within single industries. For example, Hackman (1993) shows how the factors that most strongly influence the behavior of aircraft flightdeck crews are themselves shaped by three exogenous factors: standard cockpit technology (determined by engineers at corporations that design and

manufacture aircraft), standard operating procedures (determined by regulatory agencies), and the deeply rooted and highly individualistic culture of flying. Thus, envi- ronmental and institutional forces (DiMaggio and Powell 1983, Meyer and Rowan 1977) determine the structures that are proximal to the team (such as the team's purpose, the design of its work, how rewards are allocated, and so on), which in turn shape actual team behavior.

Because the major influences on team behavior and performance often have little variation within any given organization, single-organization (or even single-industry) investigations of influences on team outcomes can explain relatively little variance in how, and how well, teams operate. And because the effects of leaders' day-to-day coaching of teams-indeed, coaching behavior itself- may depend to a considerable extent on how those teams are designed, leader coaching will appear to have only a modest influence on the overall trajectory of a team.

This view suggests two fruitful avenues for future investigation-both at the organizational level-for understanding how and when leaders influence team self- management and effectiveness. In the present research setting, front-line managers had real impact because they had a great deal of discretion in determining their teams' basic structure, how rewards and information were distributed, and overall team direction. This setting, then, is an exception to the usual strong exogenous influences on team design features. In this exception lie two leads to researchable questions about when team leaders can re- ally make a difference.

The first feature of this organization was an unusual degree of authority invested in front-line managers due to decentralization. Individual districts were in the process of becoming profit centers. Within each district, both district and field managers had the authority to alter major design features of their business units. For example, while constraints existed on total funds available for rewards, how those rewards were distributed to technicians as autonomously determined by field managers-and often differently by different field managers. Moreover, it was left to the managers to communicate to their teams their role in the new units, allowing them to shape their teams' direction.

Second, a number of design features came under the control of team leaders because the organization was un- dergoing major change. The organizational structures and systems that ordinarily operate as tightly interconnected and inert components of an organization were therefore open to change. For example, service territories had to be redetermined as some districts acquired new geography. How territories were defined and staffed directly affects



team composition-especially the number of group members and the mix of skills among them. Changes of staffing and territory responsibilities also created demands for training in some teams-demands that team leaders could decide to accept or to deny.

These two organizational features-decentralization and major structural change-represent two conditions under which team leaders have the latitude to make real differences in team effectiveness. They are also the same conditions under which researchers are likely to find real variation in team design features. A focus on leaders' behavior-both as team designers and as coaches-in circumstances in which leaders have such latitude may be more fruitful in the long term than continuing to search for the best kinds of day-to-day styles for leaders to use in interaction with their teams.

Acknowledgments The author is grateful to Ann Burress, Carlton Crowley, David Hyatt, and

Cathy Sirett for assistance in collecting and coding the data, and also to David for helping to design the interview protocols. Many thanks to Tom

Ruddy and to the teams and managers at Xerox who made this research

possible. Thanks especially to Richard Hackman for invaluable comments on earlier versions of this work.

Endnotes 1Supplementary analyses conducted on several variables that could be construed as continua (e.g., group rewards and strategy norms) showed that treating them as dichotomous variables did not misrepresent their effects. In no case, for example, were curvilinear effects obtained for the continuous versions of these variables. 2This straightforward additive model assumes that all elements of de-

sign are equally important. A fine-grained analysis of the differential effects of individual design features on the dependent variables is re-

ported in the results section of the paper. 3The behaviors not analyzed were "providing organization-level information to the team," and "attending team meetings," which all man-

agers did at least monthly; and "overriding group decisions" and "pro- ducing computer analyses of performance data," which very few

managers did. 4The use of averages of multiple performance indicators across such a substantial time range also serves to defend the findings of this research

against validity threats from regression to the mean.

5Although self-managing behavior and quality of interpersonal processes are highly correlated, this relationship cannot be accounted for

by method variance because the two constructs were assessed using different methods. Moreover, they operate differently in the mediation

analyses, suggesting that the relationship is a substantive one.

References Abramis, D. J. 1990. Semiconductor manufacturing team. J. R. Hack-

man, ed. Groups That Work and Those That Don't. Jossey-Bass, San Francisco, CA, 449-470.

Allmendinger, J., J. R. Hackman, L. Kowal-Wolk, E. V. Lehman. 1992. Methods and measures for the cross-national study of symphony

orchestras. Report no. 4, Cross-National Study of Symphony Or- chestras, Harvard University, Cambridge, MA.

Ancona, D. G., D. F. Caldwell. 1992. Demography and design: Pre- dictors of new product team performance. Organ. Sci. 3, 321- 341.

Atkinson, J. W. 1958. Towards experimental analysis of human motivation in terms of motives, expectancies, and incentives. J. W. Atkinson, ed. Motives in Fantasy, Action, and Society. Van Nos-

trand, Princeton, NJ.

Balkema, A., E. Molleman. 1999. Barriers to the development of self-

organizing teams. J. Managerial Psych. 14 134-149. Baron, R. M., D. A. Kenny. 1986. The moderator-mediator variable

distinction in social psychological research: Conceptual, strategic and statistical considerations. J. Personality and Soc. Psych. 51 1173-1182.

Barry, D. 1991. Managing the bossless team: Lessons in distributed

leadership. Organ. Dynamics 20 31-47. Bass, B. M. 1957. Behavior in Groups. Louisiana State University,

Baton Rouge, LA.

Beekun, R. I. 1989. Assessing the effectiveness of sociotechnical interventions: Antidote or fad? Human Relations 47 877-897.

Berkowitz, L. 1953. Sharing leadership in small, decision-making groups. J. Abnormal and Soc. Psych. 48 231-238.

Bennis, W., B. Nanus. 1985. Leaders: The Strategies for Taking Charge. Harper & Row, New York.

Campion, M. A., G. J. Medsker, A. C. Higgs. 1993. Relations between work group characteristics and effectiveness: Implications for de-

signing effective work groups. Personnel Psych. 46 823-850.

Cohen, S. G., G. E. Ledford, Jr. 1994. The effectiveness of self-

managing teams: A quasi-experiment. Human Relations 47 13- 43.

-, G. Spreitzer, G. E. Ledford Jr. 1996. Developing effective self-

managing work teams in service organizations. Working paper. Cordery, J. L., W. S. Mueller, L. M. Smith. 1991. Attitudinal and be-

havioral effects of autonomous group working: A longitudinal field study. Acad. Management J. 34 464-476.

Cummings, T. G., W. H. Griggs. 1977. Workers' reactions to autonomous work groups: The impact of control. Organization and Ad- min. Sci. 7 87-100.

DiMaggio, P. J., W. W. Powell. 1983. The iron cage revisited: Insti- tutional isomorphism and collective rationality in organizational fields. Amer. Sociological Rev. 48 147-160.

Druskat, V. U. 1995. A team competency study of self-managed manu-

facturing teams. Doctoral dissertation, Boston University, Boston, MA.

Fiedler, F. E. 1958. Leader Attitudes and Group Effectiveness. Uni-

versity of Illinois Press, Urbana, IL.

Farris, G. F., F. G. Lim. 1969. Effects of performance on leadership, cohesiveness, influence, satisfaction, and subsequent performance. J. Appl. Psych. 53 490-497.

Goliembiewski, R. T. 1995. Managing diversity in organizations. Uni-

versity of Alabama Press, Tuscaloosa, AL.

Goodman, P. 1986. Leading groups in organizations. Designing Effec- tive Workgroups. Jossey-Bass, San Francisco, CA.

, R. Devadas, T. L. Hughson. 1988. Groups and productivity: An-

alyzing the effectiveness of self-managing teams. J. P. Campbell,



R. J. Campbell, and Associates, eds. Productivity in Organiza- tions. Jossey-Bass, San Francisco, CA.

- , S. Shah. 1992. Familiarity and work group outcomes. S. Wor-

chel, W. Wood., and J. Simpson, eds. Group Process and Pro-

ductivity. Sage, London, U.K.

Hackman, J. R. 1982. A Set of Methods for Research on Work Teams. Technical report no. 1, Research Program on Group Effective-

ness, Yale School of Organization and Management, Yale Uni-

versity New Haven, CT. - . 1985. Doing research that makes a difference. E. E. Lawler,

A. M. Mohrman, S. A. Mohrman, G. E. Ledford, and T. G. Cum-

mings, eds. Doing Research That Is Usefulfor Theory and Prac- tice. Jossey-Bass, San Francisco, CA.

- . 1986. Psychology of self-management in organizations. M. S. Pollack and R. O. Perloff, eds. Psychology and Work: Productiv-

ity Change and Employment. Amer. Psych. Assoc. - . 1987. The design of work teams. J. W. Lorsch, ed. Handbook of

Organizational Behavior. Prentice-Hall, Englewood Cliffs, N.J. - . 1990. Groups That Work (and Those That Don't): Creating Con-

ditions for Effective Teamwork. Jossey-Bass, San Francisco, CA.

-, G. R. Oldham. 1974. The job diagnostic survey: An instrument for the diagnosis of jobs and the evaluation of job redesign pro- jects. JSAS Catalogue of Selected Documents in Psych. 4 148.

-- , . 1980. Work Redesign. Addison-Wesley, Reading, MA. , N. J. Vidmar. 1970. Effects of size and task type on group performance and member reactions. Sociometry 33 37-54.

- , K. R. Brousseau, J. A. Weiss. 1976. The interaction of task de-

sign and group performance strategies in determining group effectiveness. Organizational Behavior and Human Performance 16 350-365. .1993. Teams, leaders, and organizations: New directions for

crew-oriented flight training. E. L. Wiener, B. G. Kanki, R. L. Helmreich, eds. Cockpit Resource Management. Academic Press, Orlando, FL.

-- . 1998. Why teams don't work. R. S. Tindale, J. Edwards, E. J. Posavac, eds. Applications of Theory and Research on Groups to Social Issues. Plenum, NY.

Hut, J. A., E. Molleman. 1998. Empowerment and team development. Team Performance Management J. 4 53-66.

Jackson, J. M. 1953. The effect of changing the leadership of small

groups. Human Relations, 6 25-44. Jackson, P. R., S. Mullarkey, S. Parker. 1994. The implementation of

high-involvement work teams: A four-phase longitudinal study. Paper presented at BPS Occupational Psych. Conf. Birmingham, U.K.

Jackson, S. 1992. Team composition in organizations. S. Worchel, W. Wood, and J. Simpson, eds. Group Process and Productivity. Sage, London, U.K.

Kaplan, R. E. 1979. The conspicuous absence of evidence that process consultation enhances team task performance. J. Appl. Behavioral Sci. 15 346-360.

Komacki, J. L., M. L. Deselles, E. D. Bowman. 1989. Definitely not a breeze: Extending an operant model of effective supervision to teams. J. Appl. Psych. 74 522-529.

Klaus, D. J., R. Glaser. 1970. Reinforcement determinants of team proficiency. Organ. Behavior and Human Performance 5 33-67.

Liang, D. W., R. Moreland, L. Argote. 1995. Group vs. individual

training and group performance: The mediating role of transactive

memory. Personality and Soc. Psych. Bull. 2, 384-393.

Likert, R. 1958. Effective supervision: An adaptive and relative process. Personnel Psych. 11 317-332.

Lindsley, D. H., D. J. Brass, J. B. Thomas. 1995. Efficacy-performance spirals: A multilevel perspective. Acad. Management Rev. 20 645-678.

Lippitt, R. 1940. An experimental study of the effect of democratic and authoritarian group atmospheres. University of Iowa Studies in Child Welfare 16 43-95.

Locke, E. A., L. M. Soari, K. N. Shaw, G. D. Latham. 1981. Goal

setting and task performance: 1969-1980. Psych. Bull. 90 125- 152.

Lowin, B., J. R. Craig. 1968. The influence of level of performance on

managerial style: An experimental object-lesson in the ambiguity of correlational data. Organ. Behavior and Human Performance 3 440-458.

Macy, B. A., P. D. Bliese, J. J. Norton. 1991. Organizational change and work innovation: A meta-analysis of 131 North American field experiments-1961-1990. Annual Convention of the Acad.

Management, Miami, FL.

Manz, C. C. 1986. Self-leadership: Toward an expanded theory of self- influence processes in organizations. Acad. Management Rev. 11 585-600.

, H. P. Sims. 1987. Leading workers to lead themselves: The ex- ternal leadership of self-managing work teams. Admin. Sci. Quart. 32 106-128.

Meyer, J. W., B. Rowan. 1977. Institutionalized organizations: Formal structure as myth and ceremony. Amer. J. Sociology 83 340-363.

Prince, C., T. R. Chidester, J. A. Cannon-Bowers, C. S. Bowers. 1992. Aircrew coordination: Achieving teamwork in the cockpit. R. W.

Swezey and E. Salas, eds. Teams: Their Training and Perfor- mance. Ablex, Norwood, NJ.

Rosenbaum, M. E., D. L. Moore, J. L. Cotton, M. S. Cook, R. A. Hieser, M. N. Shovar, M. J. Gray. 1980. Group productivity and process: Pure and mixed reward structures and task interdependence. J.

Personality Soc. Psych. 39 626-642.

Saavedra, R., P. C. Earley, L. Van Dyne. 1993. Complex interdependence in task-performing groups. J. Appl. Psych. 78 61-72.

Salancik, G. R., J. Pfeffer. 1977. Constraints on administrator discretion: The limited influence of mayors on city budgets. Urban Af- fairs Quart. 12 475-498.

Salas, E., J. A. Cannon-Bowers, E. L. Blickensderfer. 1993. Team performance and training research: Emerging principles. J. Washing- ton Acad. Sci. 83 81-106.

Schein, E. H. (1988). Process Consultation. Addison-Wesley, Reading, MA.

Schlesinger, L., J. M. Jackson, J. Butman. 1960. Leader-member interaction in management committees. J. Abnormal and Soc.

Psych. 61 360-364.

Shea, G. P., R. A. Guzzo. 1987. Groups as human resources. G. R. Ferris and K. M. Rowlands, eds. Research in Personnel and Hu- man Resources Management, Vol. 5. JAI Press, Greenwich, CT 323-356.

Smoll, F. L., R. E. Smith. 1989. Leader behaviors in sport: A theoretical model and research paradigm. J. Appl. Soc. Psych. 19 1522-1551.



Steiner, I. D. 1972. Group Process and Productivity. Academic Press, New York.

Wageman, R. 1995. Interdependence and group effectiveness. Admin. Sci. Quart. 40 145-180.

Wall, T. D., N. J. Kemp, P. R. Jackson, C. W. Clegg. 1986. Outcomes of autonomous work groups: A long-term field experiment. Acad. Management J. 29 280-304.

Walton, R. E. 1985. From control to commitment: Transformation of workforce management strategies in the United States. K. B. Clark, R. H. Hayes, and C. Lorenz, eds. The Uneasy Alliance: Managing the Productivity-Technology Dilemma. Harvard Busi- ness School Press, Boston, MA.

Wellins, R. S., W. C. Byham, J. M. Wilson. 1991. Empowered Teams: Creating Self-Managing Working Groups and the Improvement

Accepted by M. Scott Poole.

of Productivity and Participation. Jossey-Bass, San Francisco, CA.

Wood, J. 1990. J. R. Hackman, ed. Groups that Work (and Those That

Don't): Creating Conditions for Effective Teamwork. Jossey- Bass, San Francisco, CA.

Woodman, R. W., J. J. Sherwood. 1980. The role of team development in organizational effectiveness: A critical review. Psych. Bull. 88 166-186.

Wooley, A. W. 2001. Effects of intervention type and timing on task performance. J. Appl. Behavioral Sci.

Yukl, G. 1989. Managerial leadership: A review of theory and research. J. Management 15 251-289.

Zander, A. 1971. Motives and Goals in Groups. Academic Press, New York.
