Analyzing Event-Related Brain Dynamics in Continuous Compensatory Tracking Tasks Ruey-Song Huang 1,2 , Tzyy-Ping Jung 2 and Scott Makeig 2 1 Department of Cognitive Science, University of California, San Diego, USA 2 Swartz Center for Computational Neuroscience, Institute for Neural Computation, University of California, San Diego, USA Abstract-- The dynamics of electroencephalographic (EEG) activity in continuous compensatory tracking tasks were analyzed by independent component analysis (ICA) and time-frequency techniques. In one-hour sessions, 72-channel EEG was recorded while a healthy volunteer attempted to use a trackball to keep a drifting disc in a bulls-eye in the center of screen. Disc trajectory was converted into a moving measure of disc error. Local minima (perigees) indicated moments when the disc started to drift away from the center. Subject performance was indexed by root mean square disc error in a 20s epoch centered on each perigee, high error generally indicating drowsiness. Maximally independent EEG processes and their equivalent dipole source locations were obtained using the EEGLAB toolbox (http://sccn.ucsd.edu/eeglab). Component activations were epoched in 5s time intervals time locked to perigees. Following disk perigees during (drowsy) periods of high disk error, significant spectral changes were observed. One of the 70 independent components was located in or near primary visual cortex. During periods of poor (drowsy) performance, it had increased mean tonic alpha/theta activity, with a further phasic alpha/theta increase following perigees [1, 2]. At the same time, low alpha activity of a second component located in or near cingulate gyrus increased, and 10-30 Hz EEG activity of a third component in the left somatomotor cortex increased briefly. The alpha activity of the somatomotor component persisted through the following distance maximum. These spatiotemporal phenomena were consistently observed across three sessions within subjects. Thus, event-related EEG brain dynamics can be detected and modeled in a continuous behavioral task without impulsive event onsets. Index Terms -- EEG, ICA, ERSP, continuous task, performance, brain dynamics, theta, alpha, beta. I. INTRODUCTION Sensory event-related potentials (ERP) index mean electroencephalographic (EEG) activities following onsets of visual or auditory stimuli. In many ERP paradigms, participants respond to stimuli with discrete button presses. ERPs are usually obtained by time-domain average of EEG epochs precisely phase-locked to stimulus or to response onsets. In real-life, many tasks require more or less continuous efforts to maintain appropriate behavior, instead of discrete (button) choice responses. During the course of truly continuous performance paradigm, on the other hand (e.g., in driving simulations), participants can receive continuous visual and/or auditory stimuli as well as continuous performance feedback. The ERP averaging technique is limited to tasks with discrete stimulus events and may require a silent baseline period preceding stimulus onsets, making it of little use for assessing brain dynamics in continuous performance tasks. In this study, we applied event-related spectral perturbation (ERSP) methods to study event-related brain dynamics in a continuous compensatory tracking task (CTT) during which participants attempted to use a trackball to keep a randomly drifting disc in a bulls-eye on the center of screen [3]. Independent component analysis (ICA) was applied to continuous EEG data collected in each of the 1-hour CTT sessions. Maximally independent EEG processes and their dipole source locations were obtained by the EEGLAB toolbox (http://sccn.ucsd.edu/eeglab) [4]. During high-error periods, three clusters of independent components exhibited significant spectral perturbations following disc trajectory ‘perigees’ (moments when the disc began to escape from the center). Within subjects, the component activations and scalp topographies of these clusters were stable across sessions. This study also demonstrates that brain dynamics linked to changes in human performance can be assessed on the sub-second time scale in a continuous performance task. II. MATERIALS AND METHODS A. Participants and Tasks Six right-handed adults (3 males, 3 females, mean age= 27.8, SD = 6.0) with normal or corrected to normal vision volunteered to participate in this experiment. Informed consent was obtained from all participants. Participants arrived after lunch and sat on a cozy chair in the EEG booth in which lighting was dim. Each subject participated a three one-hour sessions of continuous visuospatial compensatory tracking tasks in which they attempted to use a trackball to keep a drifting (‘wind-blown’) disc in a bulls-eye as near as possible to the center of screen (Fig. 1). Participants were instructed to continue to perform the task as best as they could even when they began to feel drowsy. No intervention was made when participants occasionally fell asleep and stopped responding; participants resumed task performance themselves after non-responding periods. The coordinates of the drifting disc and trackball activities were recorded about 14 times per second while a synchronous pulse marker train was sent to the EEG recording system. Proceedings of the 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference Shanghai, China, September 1-4, 2005 0-7803-8740-6/05/$20.00 ©2005 IEEE.