University of Washington

Department of Rehabilitation Medicine

Search Again


Cortically-Derived Error-Signals During BCI Use


Wander JD, Olson JD, Ojemann JG, Rao RPN



Publication Info:

Proceedings of the Fifth International Brain-Computer Interface Meeting: Defining the Future, June 2013:


Standard training regimens for decoders used in brain-computer interfaces (BCI) typically involve a calibration phase where training data is collected a priori within the context of a known task. Such a paradigm is not applicable in the general case of BCI use where the tasks are not known a priori. A potential solution to this problem is to derive a "reward" or "punishmen" signal from the neural data itself. One example of such a signal is an error signal generated whenever the BCI fails to reach a goal the subject seeks to achieve. In this study we trained five human subjects to use a simple, one-dimensional electrocorticographic (ECoG). We demonstrate that invariant to multiple task parameters, there exist robust cortical signals that are correlated with failure to successfully complete the task. Further, we show that these error signals can be used to infer the outcome of new executions of the task. Our results suggest that such signals could potentially be utilized as reinforcement signals in the general case where task structure is unknown. ISBN 978-3-85125-260-6 and DOI 10.3217/978-3-85125-260-6

Link to Article:

© Copyright 2000-2018 University of Washington