ABSTRACT
Abstract The efficient implementation of breast screening depends on the radiologists' ability to detect and appropriately classify any abnormality on the X-ray mammograms. This is fundamentally a search task and, in common with all tasks involving human visual search, errors occur in this process. We describe a simulated screening task using experts and more naive radiologists to examine the incidence of these errors and their relationship to visual search times. It is argued that visual search behaviour in screening can be adequately encompassed by a model of search found useful in other radiological tasks.
