Abstract
The abilities to store and manipulate information in visual working memory (VWM) stand at the cornerstone of intelligent behavior. Demonstration of a 3-4-item storage limit in humans and other species implies a homologous storage mechanism. In contrast, evolutionary bases underlying VWM manipulation remain unexplored. Here, we use a task akin to the "Shell Game" to examine whether the ability to manipulate information is uniquely human. We compared humans' performance on a computerized version of this task with that of a Grey Parrot on a live version adapted for animal testing, administered by a human experimenter. In both tasks, a varying set size (2-4) of colored objects was presented briefly, after which objects were covered by occluders. Occluders either remained stationary (storage: 0 swaps) or pairs of occluders swapped positions a number of times (manipulation: 1-4 swaps). A target color was subsequently presented and participants had to indicate the location of the cued color. Whereas storage trials require only that color-location bindings be stored in VWM, manipulation trials additionally require that these bindings be updated as objects swap positions. Performance accuracy on storage trials exhibited by humans and the parrot were both consistent with a 3-4 –item storage limit; the parrot was also perfect on 5 (humans not tested). Manipulation-related performance was near ceiling for both groups on set size 2 trials. However, humans experienced a systematic decrease in accuracy as swaps increased for set sizes 3 and 4. Interestingly, the parrot generally remained near ceiling for all set size 3 dynamic trials, but experienced catastrophic failure for set size 4 at three swaps. These results suggest that VWM manipulation is not unique to humans, inviting further investigations into whether VWM manipulation is supported by common vs. separate mechanisms across species, and whether differences exist for computer-based vs. live trials.
Meeting abstract presented at VSS 2017