4.1. Shared representations
Perhaps behavioral coordination—from body posture to gaze direction—is yoked to representational coordination. For example, the activation of the motor system while watching others perform actions suggests a common neural mechanism for motor control and action understanding (cf. Liberman, 1974). Sebanz et al. (2006) suggest that the hypothesized shared representations would serve to facilitate the understanding of the actions of another (e.g., Blakemore & Decety, 2001; Richardson & Dale, 2005), predict the actions of others (e.g., Ramenzoni, Riley, Shockley, & Davis, 2008; Bosbach, Cole, Prinz, & Knoblich, 2005), and to know what joint actions are possible (e.g., Richardson et al., 2007b). Another possibility is that communication can be understood as the alignment of conceptual representations (Garrod & Pickering, 2004; Pickering and Garrod, in press). This alignment can occur in words, syntax, gaze, or posture—but alignment at all levels fuels the alignment of conceptual representations, and furthers the goal of communication. In this characterization, cognitive representations are the motivation and mechanism of coordination. We offer a different interpretation building from the principles of dynamical systems theory. Here, cognition is not assigned a primary status and the phenomena in question are characterized in terms of broader physical principles.
4.2. Coordinative structures
Interpersonal coordination can be conceptualized without an appeal to the cognitive representation of one’s own or another’s actions. It can be thought of as a coordinative structure—a self-organized, softly assembled (i.e., temporary) set of components that behave as a single functional unit (Bernstein, 1967). The power of a coordinative structure is that it is an abstract structure: different components of the body can play different roles under different constraints while preserving the overall collective pattern dictated by the task constraints (e.g., Kelso, Tuller, Vatioksis-Bateson, & Fowler, 1985) (cf. Giveans et al., 2008 with Shockley et al., 2003). Coordinative structure is consistent with our understanding of the principles of nonlinear, self-organizing systems that are well known in physics (e.g., Babloyantz, 1986), in that coordinative structures may emerge naturally from the interaction among the components under certain constraints (e.g., Kelso, 1995; Turvey, 1990).
Proponents of dynamical systems accounts of human performance have offered similar explanations of human motor coordination. For example, self-organized, coordinative structures have been hypothesized to explain interlimb rhythmic coordination (e.g., Fuchs, Jirsa, Haken, & Kelso, 1996). Interlimb rhythmic coordination conforms to a model (generally referred to as the Haken, Kelso, & Bunz [HKB, 1985] model) of coupled oscillators (i.e., oscillating limbs) whose behavior is readily understood in terms of a relational, low-dimensional order parameter (relative phase) that captures the stable relation between the two limbs’ oscillation cycles. The model derives from models of physical oscillators and makes no assumptions about executive control systems. Of particular importance is that the dynamics of the coordinative structure emerge from the constraints imposed upon the coordinated units. For example, interlimb coordination dynamics are influenced by physical constraints such as the relative sizes and masses of the limbs and functional constraints such as handedness (e.g., Treffner & Turvey, 1995), perceptual anchors (e.g., Fink, Foo, Jirsa, & Kelso, 2000), and cognition (e.g., Pellecchia, Shockley, & Turvey, 2005).
The power of the HKB model of interlimb coordination becomes evident when considering the fact that 10 nonintuitive predictions of the model have been confirmed experimentally (Schmidt & Turvey, 1995). In fact, this same model can capture interlimb coordination across individuals. For example, Schmidt, Carello, and Turvey (1990) investigated interpersonal coordination of rhythmic limb oscillations. They found that when participants viewed the limb oscillations of another person, the coordination of the two limbs exhibited the same stable coordination modes as those observed within an individual (see Marsh, Richardson, & Schmidt, in press). Schmidt et al. further demonstrated that transitions between stable coordination modes across individuals exhibited the same hallmark dynamics found in individual interlimb coordination (e.g., critical fluctuations, critical slowing down, all predicted by HKB). Interpersonal coordination, therefore, need not entail representations of the other person in any classical sense. Rather, it may reflect the spontaneous organization (i.e., coordination) that emerges naturally from the interaction of two (or more) people under certain constraints (e.g., intentions, task constraints, physical constraints).
A comparison of the study of Giveans et al. (2008) with that of Shockley et al. (2003) may illustrate the functional nature of the cross-person organization. Giveans et al. (2008) found that the shared activity of head movements was influenced by visual manipulations, while Shockley et al. (2003) found no influences on head movements but the shared activity at the waist was absent. In other words, different body segments were coordinated in the studies of Shockley et al. (2003) and Giveans et al. (2008) for the same task under different constraints. This may suggest that the body segments involved in postural coordination during conversation changes flexibly with the specific task constraints. The coordination among body segments to both maintain upright stance and to achieve the joint task appears to be a task-dependent organization of the segments supporting stable performance of the task itself (e.g., Oullier, Bardy, Stoffregen, & Bootsma, 2004). This type of functional organization is seen in within-person coordinative structures. For example, vocal tract configuration for a given utterance changes with changes in task constraints. Kelso et al. (1985) demonstrated that when the jaw is perturbed during the articulation of a particular utterance, an immediate (20–30 ms) compensation by the upper and lower lips occurs to preserve the integrity of the intended utterance. Importantly, this modulation occurs far more quickly than could be controlled by a central executive, suggesting that the modulation is occurring outside of executive control. If joint cognitive tasks are similarly achieved via a functionally defined cross-person organization, then perturbing/constraining the actions of one component of the cross-person coordinative structure of one member (e.g., a relevant body segment, optical information, cognitive constraints) should result in rapid compensatory changes in other components of the cross-person structure (e.g., changes in movement patterns of a body segment, looking patterns, or cognitive kinematics in the other member of the pair). In other words, if the cross-person coordinative structure consists of a certain relation among body segments and cognitive states, then constraints on the (action) effectors of one person should affect the movements and/or cognition of the other member of the pair as readily as cognition can affect one’s own effectors.
Interpersonal postural coordination as a functionally defined cross-person coordinative structure was explicitly suggested in a recent study by Ramenzoni, Baker, Riley, and Shockley (2007). Ramenzoni et al. had participant pairs who perform a joint precision pointing task while standing. One participant was required to maintain the position of a pointer (held in the hand) within the boundaries of a target of a particular size (held by the other participant). The motion of the torso, forearm, arm, and hand were recorded. The time series of this collection of body segments of pairs of participants were submitted to principal components analysis to evaluate how many principal components were required to account for the coordination in the various conditions. Fewer components were required to account for 90% of the variance in the behavior of an individual when the task was performed with another participant than when the precision task was performed without a partner. Ramenzoni et al. argued that this intriguing finding of a reduction in the number of modes required to account for the collective behavior of the body segments may reflect an overall reorganization of components (within and across individuals) into a two-person, lower-dimensional (i.e., fewer modes) system that embodies the contributions of both participants.
Given the recent insights into the embodiment of cognitive processes in eye movements (Richardson & Dale, 2005; Richardson et al., 2007a) and the constraints of eye movements on interpersonal postural coordination (Giveans et al., 2008), a speculative hypothesis is that conversational coordination reflects a functional reorganization of body segments and eye movements to support the joint goals/actions of interacting individuals. However, to understand the organization of this joint-action system, the assumption that each participant can be studied in isolation must be relaxed. Moreover, characterizing interpersonal coordination as a cross-person coordinative structure may require a different conceptualization of cognition. Cognition, from a dynamical systems perspective, may be more usefully understood as a set of constraints on action. For example, more knowledge in a particular situation will constrain the looking patterns of an individual, just as more perceptual anchors constrain oscillating limb patterns (Fink et al., 2000). Likewise, more shared knowledge across two individuals will similarly constrain their respective looking patterns to be more similar to one another (Richardson et al., 2007a). Moreover, if one’s actions embody cognitive processes, then two people’s actions would be expected to be more similar to one another with similar cognitive constraints when they are interacting in a common environment.
In many ways, this conceptualization of cognition is consistent with the interactive alignment model (Garrod & Pickering, 2004), as both assert that forms of coordination promote communication. We differ mostly in the status or emphasis that is accorded to cognitive representations. Garrod and Pickering (2004) argue that communication, in a broad sense, is the alignment of conceptual representations. When conversing, people will (automatically or intentionally) align their behavior at many levels, from word choice to syntactic structure to body sway. Alignment at each level serves to bring conceptual representations closer together. They offer a plausible account of how “higher” lexical or syntactic representations become aligned through priming and promote linguistic processing. However, in their framework, it is harder to see how alignment of behavior at “lower” levels, such as body sway, can bring about alignment at the loftier level of conceptual representation.
The coordinative structure account, in contrast, does not accord a special status to cognitive representations. They are constraints upon coordination just as any other and can be analyzed in the same language used for physical systems. For example, the type of spontaneous alignment described by Garrod and Pickering is not special to cognitive linguistic systems, and can be observed in many places in physics and biology where there is a tendency for units to coordinate (i.e., pulling coupled units into coordinated modes). This physical principle is known as the magnet effect (von Holst, 1973), and the HKB model of interlimb coordination is based on our understanding of this tendency in physically coupled oscillators (see Turvey, 1990, for a detailed discussion). As these principles already exist in nature, and behavioral coordination has already been shown to conform to these principles in many instances, we argue that they lead to a more parsimonious account than proposing a new principle of alignment that is special to cognition.