Running head : ATTENTION & MOTOR CONTROL The Role of Attention in Motor Control

Research on the focus of attention (FOA) in motor control has found a consistent advantage for focusing externally (on the effects of one’s actions) compared to focusing internally (on one’s body mechanics). However, most of this work has concentrated on movement outcomes, leaving open the question of how external attention changes the movement itself. Somewhat paradoxically, recent research has found external attention also increases trial-by-trial movement variability. To explain these findings, we propose a theory of attention in motor control, grounded in optimal control theory, wherein variability is minimized along attended aspects of the movement. Internal attention thus reduces variability in individual bodily dimensions (positions and velocities of effectors), whereas external attention minimizes variability in the task outcome. Because the goal of a task defines a dimension in the movement space that is generally oblique to bodily dimensions, external attention should increase correlations among bodily dimensions while allowing their individual variances to grow. The current experiment tests these predictions in a dart-throwing task. External FOA led to more-accurate performance and increased variability in the motion of the throwing arm, concomitant with stronger correlations among bodily dimensions (shoulder, elbow, and wrist positions and velocities) in a manner consistent with the task kinematics. These findings indicate a shift in the control policy of the motor system, consistent with the proposed theory. These results suggest an important role of attention as a control parameter in the regulation of the motor system, and more broadly illustrate the importance of cognitive mechanisms in motor behavior. Attention & Motor Control 3 The Role of Attention in Motor Control “I feel like I'm throwing three different kinds of tosses, thinking about what to do with my arm, what to do with my legs, am I leading with my shoulder, those kinds of things. I just need to stop thinking about that so much and do what I need to do.” – Tim Lincecum, San Francisco Giants Baseball Club (Haft, 2011) One of the most important features of human movement is variability. Variability is important because it allows for movement patterns to be effectively adapted to the environment, to the specific requirements of a task, or to endogenous variables (like motivation and fatigue), while the goal of the task remains invariant (Bernstein, 1967; Davids, Bennett, & Newell, 2006). However, variability can be both promising and problematic. From a motor-control perspective, humans have many more degrees of freedom than are needed to accomplish any single task. Thus, the same movement outcome can be achieved in many different ways (Todorov, 2004). Recently, optimal control theories of motor learning and control have quantified and modeled how the nervous system takes advantage of these redundancies to optimize performance (Latash, Scholz, & Schöner, 2002; Todorov & Jordan, 2002). These theories account not only for measures of performance on average, but also trial-by-trial variability in performance (Loeb, Brown, & Cheng, 1999), which has received less emphasis in previous theories of motor control. The current study investigates the role of movement variability in mediating the effects of attention on motor performance. Previous research on attention in motor learning and control has found that when subjects are instructed to focus externally on the goal of a task, they reliably Attention & Motor Control 4 perform better than when instructed to focus internally on their own body mechanics (Lohse, Wulf, & Lewthwaite, 2012; Wulf, 2012). The benefits of an external focus of attention (FOA) with respect to the outcome of movement have been demonstrated in a variety of dynamic and isometric tasks, including golf (Bell & Hardy, 2009; Wulf & Su, 2007), basketball free-throw shooting (Zachry, Wulf, Mercer, & Bezodis 2005), dart throwing (Lohse, Sherwood, & Healy, 2010), volleyball serves and soccer kicks (Wulf, McConnel, Gärtner, & Schwarz, 2002), and force production (Lohse & Sherwood, 2012; Marchant, Grieg, & Scott, 2009). However, only recently have studies begun examining how attention affects properties of the movement itself, such as muscle recruitment (Lohse & Sherwood, 2012; Vance, Wulf, Töllner, McNevin, & Mercer, 2004; Zachry et al., 2005), energetic cost (Schücker, Hagemann, Strauss, & Völker, 2009), and movement kinematics (Lohse et al., 2010). We suggest that analyzing movement variability is critical to understanding the effects of attention, because it provides insights into what aspects of the movement are being controlled (Wolpert & Ghahramani, 2000). One finding from recent research on attention and motor variability is that external FOA actually increases variability of the movement pattern across trials, even though it reduces error in the movement outcome (Lohse et al., 2010). Although this finding may seem paradoxical, it is consistent with findings of functional variability in research on expertise effects in motor control, whereby experts often exhibit greater movement variability than novices, concomitant with better performance. Functional variability can be explained within optimal control theory as a consequence of coordination among effectors, whereby effectors compensate for perturbations in each other’s dynamics to reduce overall error (Todorov & Jordan, 2002). Thus, there is a tradeoff between minimizing variability of the outcome and of the dynamics of individual effectors. When the goal of the motor system is to control some external outcome variable (e.g., Attention & Motor Control 5 the landing position of a dart), the optimal control strategy produces increased correlations among effectors, at the expense of increasing their individual variances. These findings lead to the present proposal that attention regulates motor control by helping to determine the control strategy of the motor system. In internal FOA conditions, we hypothesize that bodily dimensions such as muscle activations or joint angles are directly controlled, minimizing their individual variabilities. Under external FOA, we hypothesize that the target of control is the outcome itself. This control strategy leads to improved performance, by allowing individual effectors to compensate for each other in order to reduce variability in the outcome. As a byproduct of this coordination, the variabilities of individual effectors increase, as do their intercorrelations. Thus, the present theory makes predictions for how FOA affects variability in the movement outcome (i.e., traditional measures of performance), variability across trials of individual bodily dimensions (e.g., joint coordinates, angles, or velocities), and the correlation structure among bodily dimensions. This theory of attention in motor control is grounded in optimal control theory and is consistent with models of attention in other domains, including learning and perception. After reviewing these connections, as well as previous research on FOA in motor control, we report an experiment testing the theory in a dart-throwing task. This experiment shows that more-external FOAs produce improved performance as well as increased variability in the angles and angular velocities of the joints of the throwing arm (shoulder, elbow, and wrist). Critically, external FOA also strengthens the correlation structure among joints during the movement in a manner consistent with the kinematics of the task, indicating that their increased individual variabilities are consequences of coordination. These results support the proposal that attention alters the Attention & Motor Control 6 control structure of the motor system, and more broadly, they argue for a central role of cognitive variables in motor control. The Effects of Focus of Attention on Motor Control Research on FOA suggests that instructions or feedback directing subjects’ attention externally (to the effect of an action on the environment) significantly improves performance relative to focusing internally (to the mechanics of the body itself). For instance, when shooting a basketball, subjects do better when mentally focused externally on the back of the rim compared to internally on the motion of the wrist, even though visual attention (i.e., gaze direction) is the same in both conditions (Zachry et al., 2005). Furthermore, previous studies have shown focusing externally improves performance relative to control conditions where no attentional instructions are given (see Wulf, 2007, 2012, for reviews). The advantage of focusing externally also holds in clinical studies of motor performance following stroke (Fasoli, Trombly, TickleDegnen, & Verfaellie, 2002), in Parkinson’s disease patients (Landers, Wulf, Wallman, & Guadagnoli, 2005; Wulf, Landers, Lewthwaite, & Töllner, 2009), or following musculoskeletal injury (Laufer, Rotem-Lehrer, Ronen, Khayutin, & Rozenberg, 2007). Currently, the dominant explanation in the literature of impaired performance resulting from an internal FOA is the constrained action hypothesis (Wulf, 2007, 2012), which posits that an internal FOA increases explicit monitoring of otherwise implicit motor behaviors, slowing processing and hurting performance (see also Beilock & Carr, 2001). The constrained action hypothesis has been criticized, however, for not being integrated with larger theories of motor control (Oudejans, Koedijker, & Beek, 2007) and because the precise mechanisms that constrain action need to be better specified in order to make the hypothesis testable (Raab, 2007). For instance, in its current form, the constrained action hypothesis does not make predictions about Attention & Motor Control 7 the details of movement under internal versus external focus conditions. One reason the constrained action hypothesis does not address movement details is that the majority of studies on FOA have been limited to the effects of attention on motor outcomes (e.g., accuracy, balance, speed), and less work has been done to explore the effects of attention on the kinematic and dynamic properties of movement itself. One recent study on dart throwing that did examine movement kinematics (Lohse et al., 2010) found that accuracy was significantly improved by directing subjects’ attention to the flight of the dart (external focus) compared to the motion of the arm (internal focus). Biomechanical analysis of trial-by-trial variability in the shoulder angle of the throwing arm at the moment of release showed greater variability with external FOA. These changes in movement variability likely play an important role in mediating the influence of attention on performance, but they lie outside the scope of current theories. Thus, the aim of the current study was to develop a more mechanistic theory of attention in complex motor tasks, integrating research on FOA with optimal control theories of motor control and learning. We propose that attention regulates motor control by changing which aspects of the movement are controlled— goal-relevant dimensions with an external focus or bodily dimensions with an internal focus. To motivate how such shifts of the control policy can affect both performance and patterns of movement variability, we next review research on the role of movement variability in skilled and optimal performance. Variability in Expertise and Optimal Control Paradoxically, experts can show increased trial-by-trial variation in movement patterns while simultaneously showing superior performance in the movement outcome. This phenomenon has been referred to as functional variability, to capture the idea that variability is Attention & Motor Control 8 somehow enabling improved performance (Müller & Loosch, 1999). For instance, Schorer, Baker, Fath, and Jaitner (2007) found that novice and intermediate hand-ball players had only two stable movement patterns, which principally differed in the direction of the throw (viz., one stereotyped pattern for a shot to the high left and another to the low right). In contrast, experts’ throwing motions clustered into roughly four different patterns, none of which could be assigned to a specific throwing direction. This absence of correspondence between throwing direction and movement pattern suggests that experts use varying movement patterns to produce similar flight trajectories. One explanation of these findings is that experts control variation in only goal-relevant aspects of the movement, while allowing redundant dimensions (i.e., aspects that do not directly affect the outcome) to vary. Evidence for this type of selective control is seen in anisotropic patterns of variability, wherein redundant dimensions show greater trial-by-trial variation than goal-relevant dimensions. A classic example in the motor control literature comes from motion analysis of expert hammer swings (Bernstein, 1967), in which the contact point of the hammer on the target is very consistent, but the motion paths of the shoulder and elbow are variable. Such patterns have been observed in a wide range of other tasks, including reaching (Haggard, Hutchinson, & Stein, 1995), grasping (Cole & Abbs, 1986), pointing (Tseng, Scholz, & Schöner, 2002), writing (Wright, 1990), postural control (Scholz & Schöner, 1999), and even skiing (Vereijken, van Emmerick, Whiting, & Newell, 1992). Importantly, anisotropic variability is more pronounced in the movement of experts than novices (Schorer et al., 2007; Vereijken et al., 1992; Wilson, Simpson, van Emmerick, & Hamill, 2008). Scholz and Schöner (1999) offer a formal framework for addressing the relationship between anisotropic variability and motor control strategies. They define the uncontrolled Attention & Motor Control 9 manifold as the subspace, within the space of all possible movements, within which the movement is uncontrolled and hence allowed to vary. When the control strategy of the motor system is to optimize the task outcome, the uncontrolled manifold comprises the subspace of movements that are consistent with the goal (Kang, Shinohara, Zatsiorsky, & Latash, 2004; Scholz & Schöner, 1999). Based on this definition, Scholz and Schöner (1999) proposed that trial-by-trial movement variability should be greater parallel than perpendicular to the uncontrolled manifold (see also Scholz, Schöner, & Latash, 2000) Building on this framework, we define a goal-relevant dimension as any dimension within movement space that affects the task outcome, and a redundant dimension as any dimension that does not. Variability on goal-relevant dimensions is detrimental, whereas variability on redundant dimensions contributes no error. To be clear, by dimension we mean not a spatial direction, but a dimension within the abstract multidimensional space of possible movements (e.g., shoulder angle or elbow angle), analogous to a perceptual dimension within an abstract stimulus space (e.g., size or brightness). Importantly, because the outcome of most motor tasks depends on the combined actions of many effectors, a goal-relevant dimension will tend to lie at some oblique angle in the movement space defined by individual bodily dimensions (e.g., positions and velocities of individual joints). Decomposing the movement space into goal-relevant and redundant dimensions enables contact with optimal control theory, which offers a rational and quantitative basis for the prediction that movement variability should be greater along redundant dimensions than along goal-relevant dimensions. Optimal control theory casts motor behavior in terms of statistically optimal control (for reviews see Latash et al., 2002; Latash, Scholz & Schöner, 2007; Todorov, 2004). According to this perspective, a control rule is defined by a movement variable to be Attention & Motor Control 10 either maximized or minimized (e.g., the goal in a vertical jump is to maximize center of mass displacement, whereas the goal of a balance task is to minimize sway). Lower levels of control (e.g., the activities of individual muscles or joints) then interact to implement the optimal solution to the control rule. Central to optimal control theory is the assumption that motor dynamics are inherently noisy, so that exact movement patterns are not reproducible (Wolpert & Ghahramani, 2000). Thus, the motor system works to minimize expected error in the face of this noise. In cases of closed-loop control (as opposed to ballistic movement), the brain can adapt control signals in response to perturbations that arise during the course of the movement, thus reducing final error. However, because motor noise is positively dependent on muscle activation (Harris & Wolpert, 1998; Schmidt, Zelaznik, Hawkins, Frank, & Quinn, 1979; Todorov, 2004), optimal control conserves the corrective signals it generates, correcting only those perturbations that affect attainment of the task goal. This conservation strategy is referred to as the minimal intervention principle (Todorov & Jordan, 2002). Because there are generally many more degrees of freedom in the space of possible movements than in the constraints defining the task goal, variability in certain directions in movement space will be irrelevant to the goal. Optimal control allows these irrelevant perturbations to accumulate, rather than correcting them at the cost of increasing motor noise. Consequently, optimal control theory predicts greater variability in task-irrelevant than in task-relevant aspects of the movement. +++++++++++++++++++ Insert Figure 1 about here