Assessing Temporal Variability in Fixation-Locked P300 Responses during Free-Viewing Visual Search

Abstract

The P300 is an evoked, neural response commonly studied in the visual neurosciences. However, this response has been almost exclusively studied under highly controlled laboratory conditions and not in unconstrained free-viewing visual search paradigms. This is due to a number of limiting factors associated with free-viewing visual search, such as noise induced by behavioral artifacts, lack of repeatability at the individual trial level, and the temporal uncertainty associated with the onset of the perceptual “event”. Thus, it is still an open question whether evoked responses, such as the P300, occur in these environments as readily as they occur in controlled laboratory settings. The current work builds on prior efforts that used deep convolutional neural networks to decode EEG-based P300 signals in cross-domain applications. While the prior work established that convolutional networks could decode the variability in the P300 response in evoked paradigms, here we apply this approach to decode P300-like responses during free-viewing visual search in unconstrained environments in order to estimate the temporal variability in the underlying responses. When measured with respect to fixation onset, the results show that the temporal variability for the onset of the P300-like response is nearly an order of magnitude larger than what is often observed in laboratory settings. Armed with this knowledge and approach, we believe that future work can then focus on identifying factors impacting the temporal variability in order to make analysis of such data manageable.