Predicting Trial-by-Trial Variation in Oculomotor Behavior Using Multivariate Electroencephalography Theta Phase

Abstract

When we interact with our environment, there is often a significant amount of variations in our behavioral responses to incoming sensory input even when inputs are identical. Variations in sensory-motor behavior can be caused by several factors, including changes in cognitive status and intrinsic neural variations in the brain. The correct identification of neural sources of behavioral variations is important for understanding the underlying neural mechanisms of sensory-motor behavior and for practical applications (e.g., the development of a precise brain-computer interface). However, studies on humans that investigate the neural sources of the trial-by-trial variation of the sensory-motor behavior are scarce. In this study, we explored the neural correlates of behavioral variations in smooth pursuit eye movements. We collected electroencephalography (EEG) activity from 15 participants while they performed a smooth pursuit eye movement task, wherein they tracked randomly selected visual motion targets that moved radially from the center of the screen. We isolated neural components that are specific to the trial-by-trial variation of smooth pursuit latency, speed, and direction using a novel multivariate pattern-analysis technique. We found that the phase of the spatially distributed multivariate theta oscillation was correlated with the trial-by-trial variation of pursuit latency and direction. This suggests that the changing patterns of the theta phase across EEG sensors can predict upcoming behavioral variations.

Keywords

• Brain computer interfaces
• cognition
• cognitive science
• correlation
• electroencephalography
• machine learning