University of the Basque Country; University of Hull; Highfield House Consultancy abstract This paper brings together research into and using the team role model developed by Belbin (1981, 1993a) in an attempt to provide an exhaustive assessment of construct validity in light of the conflicting evidence so far produced. Role theory is used to contextualize the origins of the model. The psychometric properties of the Team Role Self-Perception Inventory used to assess a person’s likely behaviour in a team are examined along with 43 empirical studies that have tested theoretical associations between team roles and other cognitive or behavioural traits.
While the evidence is mixed, we conclude that, on balance, the model and its accompanying Inventory have adequate convergent validity. However, strong associations between some team roles are observed, indicating weak discriminant validity among some scales in the Inventory. Through its coverage of important areas of teamworking, the paper contributes to the practitioner and research communities by providing fresh insights into aspects of teamworking and by suggesting new research agendas.
Effective teamworking has become a basic concern for most organizations. While many factors influence a team’s performance, considerable attention has been given to the influence of team member diversity in terms of roles played in a team. The team role model made popular by Meredith Belbin in relation to management teams (Belbin, 1981, 1993a) and available commercially through Belbin Associates (1988) is widely used in practice and has featured extensively in research on teams at work. The model is used by many organizations including FTSE-100 companies, multinational agencies, government bodies and consultants and has been translated into 16 languages.
This paper therefore reviews the published research and assesses to what extent the model is supported by the available evidence. Through its coverage of important areas of teamworking (conflict management, personality traits, team performance, control and power) the paper contributes to the practitioner and research communities by providing fresh insights into aspects of teamworking and by suggesting new research agendas. We first consider the theoretical context for the team role model. Second, all substantive studies that provide psychometric evidence, relationships to personality factors and evidence for predictive validity are summarized, evaluated and contrasted. Finally, we discuss the validity of the model and consider the wider implications of our findings.
Prior to the development of Belbin’s team role model (1981, 1993a) other role theories had been put forward (Benne and Sheats, 1948; Graen, 1976; Graen and Scandura, 1987; Holland, 1985) although the model’s links to these and other role classifications (e.g. Davis et al., 1992; Margerison and McCann, 1990; Parker, 1990; Spencer and Pruss, 1992; Woodcock, 1989) are unclear. While a comprehensive theoretical examination of the many alternative role theories and models is beyond the scope of this paper, it is important to establish a theoretical context for the team role model. The role concept can be viewed from two different perspectives. From an anthropological-sociological perspective it can be defined as a combination of values, attitudes and behaviour assigned to an individual who occupies a social position (a location in a social network) associated with a specific social status (the functions assigned to that person).
From this perspective, a role can be defined as the behaviour that a person displays in relation to his/her social position and social status (Linton, 1945). Secondly, from a psychosocial perspective, a role can be defined as the behaviour expected from an individual occupying a specific position (Biddle, 1979) such that the cognition and expected behaviour associated with the position are fundamentally important to success in the role (Katz and Kahn, 1978). This psychosocial perspective is adopted for the purposes of this review.
Since Lewin created the Research Centre for Group Dynamics in 1944, two types of groups have been studied: groups created to solve problems and groups preoccupied with individual development. This duality has brought about a distinction between so-called ‘task roles’ and ‘socio-emotional roles’. In this light, Bales and Slater (1955) studied laboratory groups and concluded that there were significant differences between individuals concerned with solving tasks and individuals concerned with the social and emotional needs of group members. People concerned with solving tasks were called ‘task leaders’ whereas those concerned with emotional needs were called ‘maintenance or socio-emotional leaders’. Similarly, Benne and Sheats (1948) proposed a role behaviour classification describing 12 task roles and seven maintenance roles.
Task-centred roles were concerned with the coordination of group problem solving activities, whereas maintenance roles were concerned with promoting group-centred behaviour. Both role types were thought necessary for a team to perform well. These theoretical antecedents formed the pillars of the development of the team role model (Belbin, 1981) as its general framework and the names of some team roles connect to these and other theories (Fisher et al., 2001a).
Among theoretical models explaining how roles are acquired, a two-part classification can be made (Ilgen and Hollenbeck, 1991). First, there are ‘role taking’ models that consider individuals as passive acceptors of the roles assigned to them by others (Graen, 1976). An example is the ‘role episode model’ (Katz and Kahn, 1978) where the role is defined by an interaction process between two people; the person performing the role (the focal person) and another who holds a set of beliefs that constitute the role (the role sender). The role sender communicates a set of beliefs and the focal person assumes them. The second classification of role models sees subjects actively participating in the definition and development of their role. These models assume that individuals are much more active and motivated to possess roles that they can perform successfully.
They are called ‘role making’ models because the focal person actively attempts to influence the role sender as they try to build a role that will be acceptable to both of them. Graen and Scandura (1987) proposed the ‘theory of dyadic organizing’ which integrated and extended Graen’s first proposal (1976). This theory describes how members of a team coordinate their activities to accomplish tasks that are not prescribed in their positions but fundamental for the effective functioning of the team.
When a job role involves very predictable tasks, assigning individuals to roles is relatively easy. However, as work becomes more complex then so do the abilities required by individuals. The question is no longer about the abilities and knowledge a person should have for a specific job but is about predicting how a person will behave in the work unit where the work will be performed. In this sense, Holland (1985) proposed one of the first models that accounted for this individual context adjustment, suggesting that individuals and job environments can be classified into six different types: ‘realistic’, ‘conventional’, ‘entrepreneur’, ‘social’, ‘artistic’ and ‘intellectual’. Each type is associated with specific activities and abilities possessed by individuals. A set of adjectives characterizes each type.
For example, the intellectual type is described as analytical, cautious, critical, inquisitive, independent, pessimistic and reserved. For individuals to be successful and satisfied in a job, their personal abilities, interests and personality traits should adjust with the requirements, rewards and interpersonal relations offered by the job consistent with individual job adjustment theory. Holland (1985) proposed that an individual may display attributes of more than one type and also that there are compatible and incompatible types; for example, ‘intellectual’ and ‘artistic’ types are more compatible than ‘artistic’ and ‘conventional’ types. Belbin’s team role model can be linked to these role theories and role classifications.
We now turn to review the literature on the team role model, drawing upon studies using the Team Role Self Perception Inventory (TRSPI) through which it is operationalized. We also review team role assessment using personality questionnaires and empirical studies that have explored the theoretical network of team role constructs in an attempt to better understand how individual team role preference is related to the behavioural definition of team roles as well as to other areas of teamwork behaviour.
As with most role theories, Belbin’s model is not preoccupied with the roles (behavioural patterns) per se but with the ways in which the roles develop, change and interact with other patterns of behaviour over time. The model was proposed after a nine-year study of team building and team effectiveness with management teams taking part in an executive management exercise (Lawrence, 1974). Prior to participating in the exercise, individuals completed Cattell’s 16PF personality questionnaire and Watson Glaser’s Critical Thinking Appraisal. For each management team an observer recorded group processes based upon Bales’ (1950) interactive process analysis and reported their observations.
Successful and less successful teams were analysed in terms of their members’ personalities and in terms of their critical thinking abilities. Analyses were then crossreferenced with observers’ reports and, as a result, eight team roles were proposed. The initial categorization of team roles was therefore based on assessments of team members’ personalities, critical thinking abilities and a behavioural checklist. The only empirical evidence of the early analysis showed a positive correlation between performance predictions based on team role composition and actual performance across 22 teams (Belbin et al., 1976, p. 26).
The eight role model was introduced (Belbin, 1981) and a team role was defined as a pattern of behaviour characteristic of the way in which one team member interacts with another in order to facilitate the progress of the team as a whole. Names and descriptive adjectives for each of the eight team roles were also included. In 1993 some team roles were renamed and a ninth role added. Descriptions of each role are given in Appendix 1. In this model a role is defined by six factors: personality, mental ability, current values and motivation, field constraints, experience, and role learning. However, Belbin did not show how much of the variance in a team role is explained by each factor.
In keeping with others (Benne and Sheats, 1948; Torrington et al., 1985), Belbin defends the idea that high performing teams need to have a balanced representation of all team roles. The team role balance hypothesis assumes that if all team roles are present in a team then it will perform better than other teams without the balance. Belbin also considers that the team role concept (a preference to behave in a particular way with other team members while performing tasks) should be distinguished from the concept of functional role which refers to the technical skills and operational knowledge relevant to the job. Consequently, several people may have the same functional role but vary greatly in their natural team role(s).
Belbin also stresses the link between the stages of a team’s development and the need for different team roles to dominate at different stages. Six different stages of development are proposed: (1) identifying needs; (2) finding ideas; (3) formulating plans; (4) making ideas; (5) establishing team organization; and (6) following through. In the early stages team roles like Shaper and Co-ordinator will be most needed, whereas in the later stages Completer-Finishers and Implementers make higher contributions.
Operationalizing the Model
The team role model is ideally operationalized through a self-perception inventory and through observers’ assessments to give a rounded assessment of a person’s team role. The original Team Role Self Perception Inventory (TRSPI-8R) was hand-scored such that respondents computed their own profile. This version was later modified to embody the nine role model (TRSPI-9R) and for this version respondents’ profiles are generated by the Interplace computer package. Since it was never intended that the TRSPI should be the only input to exploring a person’s team role, an Observer Assessment Sheet (OAS) was also designed to be used by work colleagues who could make an informed judgement based on their knowledge of the person. The OAS should be used alongside the TRSPI although in many situations only the inventory is used. Details of the scoring procedures for these instruments are given in Appendix 2.
The second way of assessing team roles is derived from personality questionnaires; equations to derive team roles have been developed in conjunction with personality questionnaire publishers. In particular, Cattell’s Sixteen Personality Factor Questionnaire (16PF; Cattell et al., 1970) and the Occupational Personality Questionnaire (OPQ; Saville et al., 1992) have been used (see Dulewicz, 1995).
Reviewing the Evidence
This review draws upon 43 substantive studies of the team role model using the TRSPI, OAS and personality inventories. A table showing the purpose of each study, its aims, instruments and sample used along with the key findings is available from the first author. Psychometric evidence. Eight studies have analysed the psychometric properties of the TRSPI and two have reported results from the OAS. Initial evaluations were critical (Furnham et al., 1993a, 1993b; Broucek and Randell, 1996) and one study arrived at mixed conclusions (Beck et al., 1999). Recent studies have been more supportive of the TRSPI’s reliability and structure (Swailes and McIntyre-Bhatty, 2002, 2003). Since the first criticism of the TRSPI (Furnham et al., 1993a), other researchers have raised concerns about the statistical properties of the original inventories as well as their theoretical basis (Broucek and Randell, 1996). An important issue affecting psychometric evaluation of the TRSPI stems from its ipsative nature which is outlined in Appendix 2. Evidence for the TRSPI. Furnham et al. (1993a) reported low reliability values for three different versions of the TRSPI.
Correlations between team roles were different for a normatively scored (Likert scale) version (M = 0.36) and the original ipsative version (M = -0.29). Factor structures were also different for normative values (two well-defined task and socioemotional factors) and for ipsative scoring (four bipolar factors). Both Senior (1998) and Beck et al. (1999), in their respective exploratory factor analyses, also reported an underlying four factor structure for the ipsative version of the TRSPI. However, the ipsative design of the TRSPI was deliberate and any comparison of forms should recognize that transforming the ipsative structure of the instrument may alter its nature. (See Belbin (1993b) for a rebuke of the normative version.) In the ipsative form the average interscale correlation will be negative (Meade, 2004) whereas in a normative form scales are allowed to correlate freely. In this context, Furnham et al. (1993a) raised concerns about the theoretical basis of the inventory and a lack of evidence for its psychometric properties, noting that the test was ‘neither theoretically nor empirically derived as Belbin developed his team role typology based on observatory and inductive, rather than theoretically deductive means’ (p. 247) with a limited sample of 78 managers.
Similarly, Broucek and Randell (1996) raised concerns about the internal consistency and discriminant validity of the TRSPI and the OAS. They also noted that both tests could not be considered as parallel forms of the same construct. The average correlation between team roles was 0.27 for ipsative scoring and 0.42 for normative scoring; higher correlations were expected from the self-reported data collected by both tests. Similarly, Senior and Swailes (1998) also reported that both TRSPI and OAS did not show high convergent validity as only five team roles showed significant correlations with an average of 0.27. Broucek and Randell (1996) also reported that different correlations were found between the normative and ipsative versions of the TRSPI and the NEO-PI-(R) personality scale although 8 out of 19 predictions for the ipsative version and 14 out of 19 for the normative version were correctly hypothesized.
Different correlation values were taken as ‘dramatic evidence of the type of distortion which use of an ipsative instrument produces’ (p. 401). Similarly, Fisher et al. (1996) looked at the correspondence between the TRSPI and 16PF and found low correlation values on the validity diagonal. Broucek and Randell also tested the discriminant validity of the OAS against the NEO-PI (R) Big Five personality factors, although Fisher et al. (2001a, pp. 125–6) noted that such analysis was dependent on the orthogonality of the personality factors and, as far as the factors have been found to be oblique (Costa and McCrae, 1992), any conclusion regarding the discriminant validity of the OAS should be taken cautiously.