Abstract
In classic Multiple Object Tracking and Multiple Identity Tracking (MOT, MIT), observers attempt to track N of M identical or unique items respectively as they move around a display. Typical tracking capacity is 2-4 items. Noting that we usually seem to know about the location and identity of more than 2-4 items in the world, Wu and Wolfe (2018) introduced the Multiple Object Awareness (MOA) paradigm. In MOA, observers are asked to track all unique items on screen. Like MIT, when items are occluded, observers are asked to localize one randomly selected item. Unlike MIT, if the first answer is wrong, observers keep clicking on locations until they find the target. If they were guessing, observers would need to click on half of the locations on average. However, they are not guessing. They have imperfect, but real knowledge about multiple items. From the distribution of clicks, it is possible to derive a capacity, based on this partial knowledge. This yields much larger estimates of 8-10 items. Is MOA capacity trainable and, if so, is any improvement distinct from improvements in classic tracking capacity? Five observers were trained for two hours per day for 10 days within three weeks. Observers tracked 16 unique items. Items were from three categories: animals, tools and food. After 7-20 seconds all items were occluded. Observers were asked to locate one specific item. Data consists of the number of clicks required to uncover the target. Three observers significantly improved from capacity of 8.3 to 11.5 items (averaged over final three sessions). Two observers started poorly and got worse (capacity 0-3 items). MIT capacity improved proportionally suggesting that this improvement may not be MOA specific training. Future research will determine if the stark differences between learners and non-learners reflect “real” individual differences or merely badly behaved observers.