Abstract
Much has been learned about the architecture of visual working memory, and its capacity limit, by measuring and modelling the errors that participants make on visual working memory tasks. Yet, debates persist over fundamental aspects of this memory system. In the present study, we investigated whether event-related potentials (ERPs) could clarify what happens when participants make errors on these tasks, and in particular the role of guessing. Participants were asked to remember colors for a brief delay, and memory was probed by having participants report the location of a probed colour. Unbeknownst to participants, on some trials the memory probe was a novel color not from the memory array, forcing participants to guess their answer. Thus, participants’ responses could be classified as correct, incorrect, or forced guesses. Response-locked ERPs on forced-guess trials revealed a negative going potential followed by a positivity with scalp topographies characteristic of the ERN and Pe error-related potentials. The same ERN and Pe were observed when participants responded correctly, suggesting no difference in error monitoring for these two trial types. Incorrect trials were associated with the same ERN, and a larger amplitude Pe; coupled with a larger amplitude P3a time-locked to the memory probe. We conclude this increased error-monitoring reflects awareness of errors due to inattentiveness, rather than failures of memory. These findings suggest that error monitoring was unrelated to task performance; the same ERN and Pe were observed when participants correctly recalled by the probed item and when recall was impossible. While this finding might seem counterintuitive from the perspective of the conventional discrete-capacity memory architecture, it is consistent with other architectures, such as the target confusability model, which posit that all response options are polled at retrieval, potentially producing equivalent response conflict on all trials even when the correct response is made.