Homepage of the Action Laboratory of the PenState University with information on the following research projects: Our research examines dynamical principles responsible for the generation of perceptually controlled movements in biological and artificial systems. The focus of our research agenda is on complex minimally constrained movements, involving all mechanical degrees of freedom of the moving limb in a three-dimensional task. The investigation follows a three-tiered strategy consisting of analytic work developing a model, empirical work with human subjects, and implementation on a robot "subject". A nonlinear dynamic model is developed that is able to generate movement trajectories on the basis of a network of coupled dynamic systems. Each unit consists of a nonlinear oscillators with an embedded discrete dynamic system and is thus capable of generating both rhythmic and discrete movements in a unified framework. There are seven unit dynamic systems, each corresponding to one of the seven biomechanical degrees of freedom of a human arm. Behavior is captured in kinematic trajectories of the endeffector and the individual degrees of freedom at the joints as well as their phase relationships. The experimental work proceeds in two stages. First, a series of human experiments which use constrained laboratory tasks test the model's validity by examining specific trajectory characteristics, both of the endeffector movements and the individual joint movements. Second, the model's scope is tested by studying the complex task of bouncing a ball on a racket. A major goal is to explore the integration of perception and action variables in the trajectory generating dynamic system within a task context. Complementing both series of experiments, the model equations will be implemented on an anthropomorphic robot arm with seven degrees of freedom in order to synthesize movements on the basis of the proposed organizational dynamics. Interactions Between Rhythmic and Discrete Movements In the literature on coordination and control of limb movements there appears to be a dividing line between studies that focus on discrete pointing movements, mostly rooted in a control theoretical perspective, and studies that investigate rhythmic single or interlimb actions, many of them persuing a dynamic systems perspective. Most daily activities, however, consist of a sequence or combination of discrete and rhythmic elements. Current investigations search for a unified account for the coordination of discrete and rhythmic actions, extending from the dynamic systems perspective. We suggest that the effector is controlled by two interacting dynamic pattern generators, one point attractor and one rhythmic limit cycle attractor. To determine the nature of this interaction, we performed a single-joint elbow movement in the horizontal plane, in which participants oscillated their forearm at a prescribed frequency around a specified target. Upon hearing a metronome signal, they shifted their center of oscillation to a second specified target as fast as possible, without stopping the oscillation. Analyses focussed on extracting the mutual influences of the rhythmic and the discrete component of the task. (1) The rhythmic oscillation was perturbed by the discrete movement as it showed phase resetting, typical for perturbed nonlinear oscillations. (2) The onset of the discrete movement was confined to a limited phase window in the rhythmic cycle; From these interaction effects we conclude that there are two control systems which are simultaneously present and are coupled with each each other. A dynamic model has been developed which simulates these effects. Studies that expand this model into multi-joint movements are now in priliminary stages. Control of Force and Timing in Rhythmical Tapping The control of timing and force have each been the subject of many studies, yet, comparatively little attention has been directed at the interaction of these two task requirements. A series of experiments investigating rhythmic single-joint wrist movement are underway, in which participants tap their fingertips on a force plate simultaneously satisfying explicit timing and force requirements. Tapping frequencies bracketing the natural frequency of the wrist (1-3 Hz) and peak impact forces ranging between 5 and 15 Newtons are used. Analyses focusses on the interrelation between the peak forces (PF) and intertap intervals (ITI), both in terms of their magnitudes and variabilities. Previous results revealed that tapping at frequencies lower than the wrist's natural frequency was associated with higher PF variability. This effect highlighted the pivotal role of the limb's natural frequency on variability in both force and time, in line with results by Inui and Ichihara (1998), Turvey, Schmidt and Rosenblum (1987), and Rosenbaum (1991), the latter showing that selection of limb segments is guided by the natural frequency. These results are modeled by a driven mass spring impacting a surface. Within-trial analyses examined the relationship between successive PFs and ITI s. A positive correlation was obtained between PF and its following ITI, an effect seen across all target forces and frequencies. Learning and Transfer in Multifrequency Coordination To further understanding of multifrequency bimanual coordination from a dynamic perspective a long-term learning study was conducted with the objective to test whether: 1) 1:3 coordination can be understood as coupled oscillators in extension of Sternad et al. (1995); 2) a continuous measure of relative phase and its fluctuations capture changes occurring in the learning process; and 3) transfer effects can be understood as a result of a changed dynamics. Using the wrist-pendulum paradigm (Kugler & Turvey, 1987), participants (n=5) practiced the task over 20 learning sessions (15 trials each), holding identical pendulums in both hands, with the dominant hand performing three times the frequency of the non-dominant hand. Prior to, during, and after the 20 sessions, participants¹ performance was also assessed at 1:1, 1:2, 1:4 and 2:3 coordination. Performance was successfully quantified by relative phave and its standard deviation indicating that all subjects learned to perform 1:3 coordination. Each individual developed his/her preference for one particular mode, despite receiving demonstrations only for 0 rad. Decreasing standard deviation of relative phase across practice reflected improved performance. Results from the other multifrequency tasks showed both positive and negative transfer with spontaneous appearance and loss of ratios as a function of the accompanying practice of 1:3. In sum, acquisition of multifrequency skill is discussed as stabilizing an oscillatory regime with interference effects across related tasks. Dynamic Stability in Bouncing a Ball The skill of rhythmically bouncing a ball on a racket is investigated from a dynamic systems' perspective. In previous research a model for the racket's and ball's motions was derived. Linear stability analysis and Lyapunov stability analysis led to two major predictions: (1) A solution is dynamically stable if the racket's acceleration at the moment of impact is negative. (2) The degree of stability is a nonlinear function of the acceleration at impact. These predictions had been verified when participants performed a one-dimensional task. The goal of present inquirey is to investigate the role of perceptual information utilized by subjects to attune to this dynamically stable state. In one experiment, three perceptual conditions involved excluding haptic information about the impact, excluding visual information, and a control with full perceptual information. Results showed that (1) when only haptic information is available participants found the system's stable regime, accompanied by higher variability compared to the control; (2) when only visual information is accessible subjects perform in a dynamically unstable pattern. In another experiment the task constraints were relaxed and racket and arm movements were performed in 3D. Participants bounced a ball with their eyes open and closed, and at three different amplitudes. The model's predictions were confirmed under all task conditions. The acceleration values showed no difference in the two perceptual conditions, although variability was higher in the closed-eyes condition. Additionally, in three successive trials the acceleration at impact changed towards the values for which a higher degree of stability was predicted.