Abstract
Working memory (WM) and attention are intimately connected, but the precise relationship between the two processes is unclear. Here we investigated whether attention load during maintenance had an effect on precision of WM recall. In separate experiments, participants viewed a memory array of either random dot motion kinematograms or oriented bars, of different colours. Following a blank interval, a visual search task with varying levels of difficulty (attention load) was presented. Participants were asked to respond as fast and as accurately as possible. After a further delay, they were probed on their memory: adjusting a coloured probe to match either the motion direction or the orientation of the bar with the same colour in the memory array. In both experiment, participants were slower and made more errors on the visual search task under high compared to low load, confirming that attention load was successfully manipulated. Precision of WM recall was significantly lower in high compared to low load and baseline (no visual search) conditions. Furthermore, the proportion of responses around the target decreased under high load compared to the other two conditions. Crucially, this was accompanied by an increase in the proportion of responses around non-target features (i.e., directions or orientations of other items in the memory array that were not probed), demonstrating feature misbinding in WM. A probabilistic model of performance used to examine possible sources of error in WM also revealed that the loss in precision under high load was purely explained by an increase in non-target responses. Thus, under high attention load, features of other items in the array contaminated recall of the direction of motion or orientation of the probed item. These findings suggest the resources used for attention (visual search) are also employed for maintenance of feature binding in WM representations.
Meeting abstract presented at VSS 2012