Do You Know What You're Tracking? 

Todd S. Horowitz, Sarah B. Klieger, Jeremy M. Wolfe, George
A. Alvarez, and David E. Fencsik

In laboratory studies of multiple object tracking (MOT), observers see
a set of identical objects, track a subset of target items for several
seconds, then report which items were targets. Capacity is 3-5
objects. However, real world observers might monitor several quite
different objects and wish to determine the location of one specific
object (e.g. "Where is my child?").

In two experiments, observers tracked four of eight unique cartoon
animals moving over a backdrop of cactuses. After a random tracking
duration, all animals "hid" behind cactuses. Accurate responses
required observers to track continuously.

In a series of experiments, observers were either asked to click on
all cactuses occluding target animals ("full report") or to click on
the cactus occluding a specific target (e.g. the zebra - "target
location").

Capacity was derived from accuracy corrected for guessing. In all
experiments, "target location" capacity (1.44 to 2.15 objects) was
significantly lower than "full report" capacity (2.84 to 3.54
objects). However, observers could individuate more than one item,
suggesting featural or identity information is tracked. Tracking
ability in MOT may reflect two systems, one tracking individual object
representations, and one discriminating targets from nontargets.