An electroencephalographic investigation of the impact of eye movements in a change detection task

In studies involving Event-Related Potentials (ERPs), ocular artifacts such as blinks and saccades can compromise the quality of the recorded neural signals. To address this issue, researchers often manually reject epochs (that is a specific time-window extracted from the continuous EEG signal) containing these artifacts. However, this procedure consistently reduces the number of epochs that can be used for extracting ERPs. An alternative solution is to use Independent Component Analysis (ICA), which can preserve more epochs for analysis by removing only the artifact from the EEG recording. However, the reliability of ICA in neurocognitive studies of lateralized ERP components, such as the Sustained Posterior Contralateral Negativity (SPCN) related to visual working memory load, remains unclear, particularly in contexts where subjects are more likely to make saccades during the task. Furthermore, by using ICA, we are assuming that ocular movements do not interact with the neural signal, which has yet to be confirmed. For this reason, in the present experiment, all the participants were asked to perform a change detection task under two conditions: a ‘free gaze/saccade’ condition, where they were allowed to move their eyes to look at the lateralized stimuli, and a ‘fixation’ condition, where they were required to maintain the gaze on the center of the monitor. The subjects were also split into two groups, each performing the same experiment but with different stimulus presentation times (100 ms and 500 ms) to investigate whether saccades could differently affect the ERP in these conditions. The SPCN components were then extracted using both the Independent Component Analysis (ICA) correction and epoch-rejection methods. The results revealed that ICA correction is a robust and reliable method for experimental paradigms with a short presentation time of the stimuli (100 ms). By removing only the saccades, the features of the SPCN are preserved, suggesting that with this method we can retain a higher number of epochs for the ERP extraction with the certainty that saccades do not alter the neural signal.