When Teams Agree While Disagreeing: Reflexion and Reflection in Shared Cognition

When Teams Agree While Disagreeing: Reflexion and Reflection in Shared Cognition

When teams agree while disagreeing: reflexion and reflection in shared cognition

Mark P. Healey

University of Manchester

Timo Vuori

Aalto University

Gerard P. Hodgkinson

University of Warwick


We thank former associate editor Neal Ashkanasy, three anonymousAMRreviewers, and Martin Kilduff, Tomi Laamanen, and Natalia Vuori for their helpful comments and guidance on earlier versions of this article. The article also benefited from the helpful comments and suggestions of three anonymous reviewers in the Managerial and Organizational Cognition Division at the 2011 annual meeting of the Academy of Management (San Antonio, Texas). We are also grateful to the Centre for Organizational Strategy, Learning and Change at the University of Leeds for supporting the development of this work. Timo Vuori acknowledges the support of the Foundation for Economic Education, Finland; the Finnish Work Environment Fund (110026); and the Finnish Funding Agency for Innovation (EV-ACTE).


Drawing on dual-systems theory, we propose a new typology for analyzing shared cognition in workgroups and teams that differentiates reflective (i.e., C-system) mental models formed through reasoning and deliberation from reflexive (i.e., X-system) representations that are more automatic, intuitive, and affective in nature. Our analysis demonstrates how team members’ X-system representations pertaining to the team’s task and its members can compete with shared C-system mental models of the task and team in terms of their respective effects on team functioning. We highlight the consequences for intrateam coordination when team members have similar C-system mental models but dissimilar X-system representations (illusory concordance) and when team members have similar X-system representations but dissimilar C-system mental models (surface discordance). Finally, we consider the implications of our arguments for extending current team cognition theory predicated on reflective cognition and suggest new directions for research on group cognition in organizations more generally.

In the past two decades there has been an explosion of interest in the cognitive mechanisms driving the processes and outcomes of workgroups and teams (DeChurch & Mesmer-Magnus, 2010a; Hodgkinson & Healey, 2008; Kozlowski & Ilgen, 2006; Salas & Fiore, 2004). However, this literature has focused on deliberative, conscious cognition and tended to overlook several important types of less mindful cognition, including implicit attitudes (Greenwald & Banaji, 1995), subconscious goals (Latham, Stajkovic, & Locke, 2010), and implicit stereotypes (Banaji, Hardin, & Rothman, 1993).

This oversight is particularly noticeable in the literature on shared cognition in teams. A construct that occupies an increasingly central position in this literature is shared mental models.In much of the research pertaining to this construct there is an assumption that when team members possess shared (i.e., similar) mental models of their tasks and the team’s attributes, they coordinate their activities more effectively (Cannon-Bowers & Salas, 2001; Cannon-Bowers, Salas, & Converse, 1993; Marks, Sabella, Burke, & Zaccaro, 2002; Mathieu, Heffner, Goodwin, Salas, & Cannon-Bowers, 2000). However, as observed by Mohammed, Ferzandi, and Hamilton(2010: 902), “mixed and contradictory findings have plagued research” on shared mental models in teams.

One explanation for these shortcomings is that by focusing on mental models that are conscious and deliberative, teams researchers have assumed that cognitive sharedness is unimodal—that is, it operates at a single level of deliberative cognition. In this article we develop an alternative approach to conceptualizing shared team cognition based on dual-systems theories of cognition (Epstein, 1994; Smith & DeCoster, 2000; Strack & Deutsch, 2004). Specifically, we posit that cognitive sharedness reflects processes operating at two distinct levels—namely, X-system (i.e., reflexive) and C-system (i.e., reflective) levels. Following advances in social cognitive neuroscience (Lieberman, 2007), we use the term X-system processes to refer to implicit cognitive processes that are automatic and spontaneous and occur without conscious awareness. The X-system is distinct from the C-system, which is responsible for cognitive processes that are reflective—that is, controlled, deliberative, and conscious. Our argument is consistent with basic evidence showing that X-system processes drive a great deal of social interaction (Bargh & Chartrand, 1999) and that much of the information people bring to bear on tasks is encoded in knowledge not available with conscious introspection (Sloman, 1996).

We argue thateven when team members have similar C-system mental models, X-system representations will not necessarily be similar across team members because the X-system representations individuals bring to a given task can contradict their corresponding C-system representations (i.e., conflict within the person). Hence, teams will often obtain the putative coordination benefits of shared cognition only when individuals’ X-system representations are aligned with their C-system counterparts, whereas misalignment within and, thus, across team members will yield the contradictory effects observed in the extant literature.

Despite growing appreciation, in organizational research, of X-system processes at the individual level (Ashkanasy & Humphrey, 2011; Dane & Pratt, 2007; Hodgkinson & Healey, 2011), teams research has not kept pace with these developments. We propose that incorporating the basic distinction between X-system and C-system cognition provides a more complete understanding of shared cognition and its influence on team processes and outcomes. Moreover, our analysis extends team cognition theory to include X-system processes that are more visceral and affective in nature. Such “hot” cognition is integral to human interaction but often overlooked in research on teams (Barsade & Gibson, 2007).

We begin by briefly reviewing the literature on shared mental models because these models are increasingly posited as the main cognitive coordination mechanisms underpinning the performance of workgroups and teams (DeChurch & Mesmer-Magnus, 2010a,b; Mohammed et al., 2010). Next we argue the case for distinguishing between X-system representations and C-system mental models at the individual level. We then move up to the group level to outline a new taxonomy for understanding configurations of X-system and C-system shared cognition that identifies four types of cognitive concordance/discordance (hereafter concordance). Our taxonomy describes how team members can agree while disagreeing—that is, they can hold similar C-system mental models while possessing dissimilar X-system representations, and vice versa. We then examine the consequences of these alternative states for intrateam coordination and develop a series of testable propositions that extend current team cognition theory predicated on reflective cognition. Finally, we consider the implications of our arguments for future research on group cognition in organizations more generally.


In this section we contrast X-system processes with C-system mental models and examine how shared cognition at the two levels interacts. In so doing, we seek to broaden understanding of team functioning,beyond shared C-system mental models per se,by incorporating reflexive processes as simultaneous and competing influences on team coordination processes.

Shared Mental Models

Although there are multiple definitions of the shared mental models construct, we adopt the seminal one offered by Cannon-Bowers and colleagues:

knowledge structures held by members of a team that enable them to form accurate explanations and expectations for the task, and in turn, to coordinate their actions and adapt their behavior to demands of the task and other team members (1993: 221).

Based on studies of mental models in systems control (Rouse & Morris, 1986), team cognition researchers assume that mental models involve conscious deliberation. For example, Salas, Rosen, and DiazGranados argue that mental models provide the basis for mentally simulating different courses of action, noting that “mental simulation is the conscious and deliberate process in which decision makers engage” (2010: 957).The assumption that shared mental models reflect explicit knowledge that is amenable to verbalization is also evident in the widespread practice of operationalizing them by using direct measures—that is, measuring beliefs concerning causal relations among task attributes with rating scales that require deliberative reasoning and effortful judgment (Cooke, Salas, Cannon-Bowers, & Stout, 2000; Mohammed et al., 2010).

Extant research emphasizes the importance of two particular types of shared mental models: task-related mental models and team-related mental models (Cannon-Bowers & Salas, 2001; DeChurch & Mesmer-Magnus, 2010a; Mathieu et al., 2000). Shared mental models of the task contain knowledge of the key task components and their interrelationships, possible scenarios, and the actions necessary for effective task performance. In contrast, mental models of the team contain information pertaining to the knowledge, skills, and abilities of fellow team members. We posit that although these two types of cognitive content are important, sharing deliberative mental models of the task and team is often insufficient for effective team performance; to function effectively, teams must also share task and team cognition at the reflexive level.

The term “shared” has multiple meanings in the shared cognition literature (Cannon-Bowers & Salas, 2001; Mohamed et al., 2010). For the present analysis, we define shared in terms of cognitive similarity (Rentsch, Small, & Hanges, 2008). Specifically, we consider that cognition is shared when team members’ individual representations contain similar information elements and the relationships between those elements are depicted similarly by the various team members. We focus on cognitive similarity because it is a key driver of team coordination. Similar beliefs and attitudes pertaining to the task and the team enable team members to hold compatible perceptions of their tasks and environments, draw common interpretations, and approach those tasks in a consistent manner (Cannon-Bowers & Salas, 2001; Mohammed, Klimoski, & Rentsch, 2000). Conversely, dissimilar beliefs and attitudes pertaining to the task and team can lead to failed expectations and confusion.

Shared Reflexive Cognition

While we focus on “shared-as-similar”, we acknowledge that work team members often possess varied knowledge and skills based on their specialist expertise or roles (Guzzo & Dickson, 1996; Hackman, 1987; Kozlowski & Ilgen, 2006). Nonetheless, we maintain that some form of cognitive sharedness isnecessary for team members withcomplementary knowledge and skills to coordinate their activities (see also Mohammed & Dumville, 2001). For instance, in medical teams surgeons and nurses possess different skills but must also share common goals to act cohesively. By unpacking the very nature of cognitive similarity, we seek to understand how different forms of similarity enable teams to maintain harmony,given varied knowledge and skills. To this end, we analyze how(dis)similarity of X-system representations, rather than (dis)similarity of C-system mental models alone, cancontribute to team coordination by acting as a form of glue that holds together the activities of team members with complementary knowledge and skills, enabling them to take actions that are consistent with one another, or, in the case of dissimilarity, acting as a force that pulls them apart.

By focusing on the effects of similarity across individual team members’ representations, we adopt what Kozlowski and Klein (2000) term thecompositional(as opposed to the compilational) approach to conceptualizing team cognition. The compositional approach represents higher-level team constructs by describing the array of individual team members’ attributes (e.g., using the degree of similarity across team members’ personal mental models to represent team mental models). In contrast, the compilational approach emphasizesthe influence of collective states that are more than the sum of individual parts. The compositional approach is appropriate for analyzing shared X-system cognition for two reasons. First, the representations that individuals bring to a given task often influence how they approach that task, thereby influencing coordination at the team level, even if those representations are not entered into collective understanding through some process of group amalgamation (Cronin & Weingart, 2007). Second, it is necessary to establish the individual-level building blocks of shared X-system cognition before considering how teams might compile those components into some collective product. We revisit compilational emergence when discussing the implications of our analysis for future work.

Individual-Level Foundations: DualSystems and Intrapersonal Dissociation

Dual-systems theorists posit that individuals possess two distinct information processing systems (Epstein, 1994; Lieberman, Gaunt, Gilbert, & Trope, 2002; Smith & DeCoster, 2000; Strack & Deutsch, 2004). Although terminology varies among such theorists,there is general agreement that what we term the X-system operates rapidly and below conscious awareness, with minimal cognitive effort. In contrast, the C-system operates more slowly in conscious working memory and requires greater cognitive effort. Researchers distinguish two views of how the systems interact(e.g., seeEvans, 2008). First-generation theories hold to a defaultinterventionist view. The assumption is that the two systems operate in sequence, the C-system shutting off the X-system when (scarce) cognitive capacity is available. In contrast, in parallel competitive modelsthe assumption is that thetwo systems operate simultaneously such that the X-system continues to function even when the C-system is active. For our present purposes, we adopt the parallel competitive view, consistent with the growing weight of neurological and experimental evidence (Lieberman, 2007; Sloman, 1996; Smith & DeCoster, 2000). Theories falling within this class are united by the following key assumptions: (1) the two systems contain distinct cognitive and affective content, and (2) they exert competing influences on behavior.

Two systems, two distinct forms of knowledge. A large weight of evidence indicates that the X-system stores information in representations that are structured associatively (Gawronski & Bodenhausen, 2006; Smith & DeCoster, 2000; Strack & Deutsch, 2004)—that is, X-system representations capture simple object-attribute relations (e.g., associating a team task with difficulty). In contrast, C-system representations are propositional in nature, going beyond learned associations to enable inferential reasoning and prediction (e.g., mentally simulating how a particular task might develop). A key feature of X-system processes is that they are affectively charged—they automatically invoke physiological arousal and feeling states (Dane & Pratt, 2007; Strack & Deutsch, 2004). Whereas the X-system deals in hot cognition in the form of associations learned through emotionally significant events(Epstein, 1994; Lieberman, 2007), C-system mental models are typically assumed to involve “cold” (i.e., unfeeling) cognition.

Table 1 summarizes three major subtypes of X-system cognition—namely,implicit attitudes, subconscious goals, and implicit stereotypes—and outlines their implications for intrateam coordination. We focus on these three types of X-system content because, as indicated in Table 1, they constitute reflexive analogues of C-system mental models of the task and team.


In dual-systems theories the assumption is that X-system representations are deep seated and stable across time and situations (Epstein, 1994; Smith & DeCoster, 2000; Strack & Deutsch, 2004). Although there is currently some debate concerning the test-retest reliability of X-system representations, compared to their C-system analogues, and whether the two systems are equally amenable to change (Nosek, 2007), the balance of evidence suggests that X-system representations generally are more resistant to external pressures relative to C-system representations. Specifically, X-system representations tend to change only following repeated exposure to an association between objects and/or attributes and are therefore less likely to change based on single instances of new information (Gawronski & LeBel, 2008; Gregg, Seibt, & Banaji, 2006; Joy-Gaba & Nosek, 2010; Rudman, Ashmore, & Gary, 2001). In contrast, C-system mental models can be updated quickly via single instance learning—as, for example, when using feedback to modify mental models of a taskin working memory (Marks, Zaccaro, & Mathieu, 2000).

Because the two systems involve different types of knowledge, it is possible that an individual’s C-system representations can conflict with his or her X-system representations. We use the term intrapersonal dissociation to refer to inconsistencies in the knowledge stored variously in the two systems within a given person, indicatedbythe (often low) degree of correlation between the individual’sX-system representations and his or her C-system representations pertaining to a given object (Hofmann, Gawronski, Gschwendner, Le, & Schmitt, 2005). In so doing, we distinguish conflict within the person from disagreement across individuals at the team level. Evidence for intrapersonal dissociation exists in various domains. People often have dualattitudes toward objects, constructing reasoned evaluations that may differ from automatic attitudes, the latterhaving unidentified origins (Wilson, Lindsey, & Schooler, 2000). Similarly, they regularly pursue implicit motives that conflict with their explicitly stated goals (Kehr, 2004). Furthermore, individuals’ self-reported views of social groups (e.g., ethnic minorities) often differ from their implicit stereotypes revealed through their actions (Dovidio & Gaertner, 2004). Finally, while problem solving, people often experience simultaneous contradictory beliefs when their conscious reasoning conflicts with their intuitive reactions (Sloman, 1996).

Competing influences on behavior. The X-system and C-system influence behavior in distinctive ways. X-system representations tend to guide spontaneous actions that occur with little deliberation (Bargh & Chartrand, 1999; Strack & Deutsch, 2004). In contrast, C-system representations are constructed effortfully in working memory and influence reasoned and planned behavior through conscious intention. Meta-analytic evidence demonstrates that measures of X-system cognition predict significant additional variance in behavioral criteria over C-system measures and are often stronger predictors of behavior overall (Greenwald, Poehlman, Uhlmann, & Banaji, 2009; Johnson, Tolentino, Rodopman, & Cho, 2010; Jost et al., 2009).

Parallel competitive dual-systems theories—epitomized by Lieberman’s (2007) X- and C-system notions—also align with Latham et al.’s (2010) arguments concerning the competing and additive effects of subconscious and conscious goals. When the two systems are aligned within the person, the effect is facilitative: subconscious goals assist the attainment of conscious goals. However, when a person’s subconscious and conscious goals conflict, the effect is competitive: the goals are processed in a relatively independent manner, and there are two distinct main effects on behavior.