Interacting with the Real World Visual Environment: How Do We Search, Learn, and Remember?

© 2021 Weijia ChenOur rich and complex visual world provides us with more information than our visual system can handle at any given time. To navigate in this world, we need to constantly search for relevant information, learn to identify objects, and be selective in the information that we memorise. In a series of papers, I investigate how people interact with their visual environment, in particular, how they search, learn, and remember the things that they see. The first paper explores the boundaries of attribute amnesia, a counter-intuitive phenomenon where people often "forget" an aspect of an object despite having attended to it mere seconds ago. Results showed that while people had difficulty recalling a colour or a repeated animal, they could accurately recall the identity of a novel animal. This suggests that attribute amnesia may only occur for repeated, familiar stimuli, hence explaining why we do not often observe it in the real world. The second paper addresses the low prevalence effect in real world visual search tasks where target rarity can lead to disproportionately high miss error rates. This is especially problematic in medical image interpretation as certain targets, such as cancerous lesions, can be very rare. In an attempt to lower miss errors while avoiding an increase in false positive identifications, a novel protocol was developed for mammogram screening. Unfortunately, miss errors were only reduced through a shift in response bias, and no improvement was observed in participants' ability to distinguish target from noise. As a result, the false alarm rate escalated substantially, hence we cannot recommend this protocol be trialled in a clinical setting. The final paper examines whether perceptual training should supplement the education program for radiology students that currently relies heavily on conceptual training, and how it can be optimized. It is demonstrated that novices improved rapidly in a hip fracture identification task and the more able students achieved the same level of accuracy as expert radiologist in less than an hour of training. Surprisingly, training novices on an image set that over-represented the more difficult images was detrimental to learning outcome. But repeating training images was as effective as showing an equivalent number of unique images. Consequently, perceptual training should be implemented in the current education programme as an effective way of teaching novices