Abstract
Previous research demonstrated performance costs in search for two targets compared with two independent single-target searches (Menneer et al., in press). This dual-target cost indicates that simultaneous maintenance of two target representations is difficult. The current study measured whether extensive search practice leads to generalization across a complex class of targets, and whether that generalization eliminates the dual-target cost.
Search for two differently-colored target classes (x-ray images of guns/knives and IEDs/bombs) was initially as accurate as single-target search for the more complex and unfamiliar target (IEDs) but less accurate than single-target search for the easier target (guns/knives). After 11 sessions (5280 trials) of training, accuracy had improved in all conditions, and dual-target search was less accurate than either single-target search. Testing with novel target images from these same classes revealed a drop in accuracy compared with the final training session, showing generalization was limited. A separate experiment showed that simultaneous search for two similarly-colored targets (x-ray images of guns and knives) was as accurate as single-target search for the harder target class (knives), and remained so with practice (20 sessions, 9600 trials). There was a trend towards a drop in accuracy between the final training session and test session. Experiments were not conducted with screening personnel.
Implications: (1) When members of a target class are sufficiently similar, generalization to novel target exemplars occurs during training, indicating that robust target representations have been learned, but there is a slight loss in accuracy compared with the training stimuli. (2) Simultaneous maintenance of two very different target representations cannot be learned, and results in poorly guided search. (3) Accuracy improves with learning, but when targets are very different, dual-target search accuracy is lower than in single-target searches. (4) Dual-target search is reasonably effective when targets share color but have complexly varying shapes.
Supported by the Human Factors Program of the Department of Homeland Security's Transportation Security Laboratory.