Abstract
Prior work has shown that observers are often poor at providing detailed information concerning the visual properties of a briefly presented stimulus when this stimulus is part of a larger set of stimuli. On the contrary, observers appear to be very capable of reporting on statistical regularities that describe a whole set of stimuli, such as the mean size of a set of circles. Furthermore, there is some evidence that the way an observer perceives such statistical regularities is partially dependent on top-down attention, such that the perceived mean is often biased in the direction of a voluntarily attended stimulus. However, it is unclear to what extent other sources of attentional bias, such as bottom-up and value-driven attention influence how an observer perceives and encodes statistical regularities. To investigate this, we conducted a series of experiments in which an observer was asked to adjust the size of a test circle to the mean size of a previously displayed set of circles that all varied in size. Crucially, we manipulated the saliency and value of one or more members of the set of circles, such that bottom-up and/or value-driven attention would be involuntarily allocated to these circles. Our results show that the way observers perceive statistical regularities is predominantly biased by saliency/bottom-up attention, whereas value-driven attention appears to play a lesser role in this process.
Meeting abstract presented at VSS 2017