How Search Performance is Affected by the Guiding Representation

Joseph Schmidt
Department of Psychology
Stony Brook University, New York

Search is an activity that we engage in countless times every day as we
interact with objects in our environment, for example, localizing your
coffee mug in peripheral vision requires a “mini” search process.
Humans typically locate search targets far better than chance would
predict; this can only occur if we maintain a target representation in
memory and match it to features in peripheral vision.  My research
focuses on the availability of target information in the guiding
representation and the resulting search performance.  Models of search
predict that a precise target representation containing many target
features should produce stronger search guidance than an imprecise
target representation containing few target features.  I tested how the
precision of a categorical target representation (boots vs footwear vs
brown boots vs brown footwear) affects search performance.  I further
investigated how the amount of target features affects search
performance by measuring target related contralateral delay activity, an
ERP measure of visual working memory load, when search guidance was
strong vs trials in which search guidance was weak.  I also investigated
what types of information can be used to guide search.  Features of
likely target regions can be used to actively restrict the search space
in the absence of global scene context, suggesting that a process of
rapid scene segmentation occurs which can aid search performance. I also
have shown that relational information between objects (knowing that my
coffee mug is to the right of my keyboard) can be used to guide search.
  It has recently been suggested that peripheral vision, in the absence
of attention, can be represented by summary statistics.  If true, search
should be restricted to the use of this information, as search is often
assumed to arise from preattentive processes across the visual field.
To test this idea we compared search guidance to an unmodified target to
search guidance to a summary statistic representation of the target.
Lastly, I will discuss when we sample information from the world around
us within static objects as well as continuous dynamic spatial-temporal
objects.