Neural correlates of human error processing and performancemonitoring have been studied through extensive evaluations with EEG, ECoG or fMRI among others. These studies rely on physiology tasks involving sudden-onset stimuli, which allow to analyze these neural imprints locked and relative to the stimuli onset. However, human perceptual decisions –and in particular performancemonitoring– comprise continuous internal processes where evidence needs to be accumulated before reaching a boundary that leads to a decision. As a result, error evaluations in diverse tasks might reflect a gradual process, rather than suddenly generated from discrete events. This paper studies EEG-measurable correlates of human error processing using a continuous trajectory monitoring task where participants have to accumulate evidence before evaluating the task performance. The task design included two experimental error conditions: first, sudden erroneous changes in the executed trajectory; and second, gradually unfolding erroneous changes from the correct trajectory. Brain-source analyses from eight subjects suggested the existence of brain activity linked to human error processing mechanisms for sudden errors and for gradual errors to a lesser extent. Notwithstanding, gradual errors did not show any discernible scalp EEG pattern, and thus it was not possible to build a machine learning model to detect these signals. Nonetheless, it was possible to use a model trained on the sudden error condition to asynchronously detect gradual errors. Furthermore, a post-hoc scalp and brain-source analysis of the detected gradual errors evidences that there is associated and discernible EEG activity, which is originated in brain areas related to error processing.
Supplementary notes can be added here, including code and math.
