Decoding trial by trial variations in VWM performance from oscillatory activity during maintenance.

Visual short term memory (VSTM) allows us to maintain a handful of representations in a highly accessible state. While this capacity limited system is well known to vary systematically between individuals, the within-individual variability of memory capacity is not well characterized. This stems from the prevalent use of partial report tasks such as change detection and cued recall tasks, in which only a single item from an array is probed. These procedures make it challenging to quantify the amount of information available in memory from individual trials. Here we examined within-subject variability using a whole report task, in which subjects reported the color and orientation of each item of the array after a memory delay. Our behavioral results indicated that while the modal number of items correctly recalled was three items, there was a substantial proportion of trials in which subjects recalled fewer or greater than this mode. We also recorded EEG from subjects during the task to examine whether the observed within-subject variability in performance reflected trial-by-trial variation in neural activity during the maintenance period. Time-frequency analyses of the EEG indicated that average power in the 8-22 Hz band during maintenance decreased as a function of the number of items in the array, reaching an asymptote at approximately 3 items. We used a linear classifier on the bandpassed individual trial EEG during the retention period and found that it could reliably distinguish the upcoming behavioral outcome on that trial (1,2 or 3 correct out of 6 items). Our results suggest that memory capacity is not fixed across trials and that this trial-by-trial variability is driven in part by variations in the sustained 8-22hz desynchronization during memory maintenance. Meeting abstract presented at VSS 2013