Abstract
In many surveillance situations, observers monitor for an event (e.g. Did anyone abandon a bag?). In multiple object tracking (MOT), humans can track ~4 pre-specified objects among distractors. What is our capacity for "event monitoring" and how does it relate to MOT capacity? In four experiments, observers tracked 2,4,6,8 moving items. Their task was to detect a state change in a one, unspecified object. Observers monitored all items. In the first experiment, items were Ts or Ls. The target event was a change in one letter from T to L or vice versa. In the second experiment, all items were unique objects, and the target would experience a state change (e.g. open book shuts). In Experiment 3, half of the items were disks "carrying" black round 'bags' while the other half were identical disks having no bag (Experiment 3). In Experiment 4, half the disks carried black square bags and the other half carried blue round bags. The target event was either one disk passing a bag to another (Exp3) or two disks swapping bags (Exp4). In all experiments, observers responded immediately once they found the target. Targets were counted as missed if observers did not respond within 2 seconds after the change occurred. Performance was better when there was only one type of change (TvsL, Exp1, capacity 3.3) than when each item and change was unique (Exp2, capacity 2.5, p = 0.01). The bag swap experiments produced similar capacities (2.4 items). Interestingly, RTs were markedly faster in Experiments 3&4, suggesting a different strategy for monitoring for part changes than for identity or state changes. However the results are explained, it seems that human observers have a very limited ability to monitor groups for critical events – a limit seemingly related to the working memory and multiple object tracking limits.
Meeting abstract presented at VSS 2016