Demos

Perceiving persistence: Can multiple objects be tracked predictively?

Objects go in and out of view all the time, either because of eye blinks, saccades, or temporary occlusion.  How do we continuously track such objects?  Do our visual systems come equipped with visually predictive tracking devices?  The question was addressed with a multiple object tracking paradigm in which subjects attempt to track four briefly disappearing targets that move among four distractors.  Participants tracked best when objects reappeared where they disappeared, and worse when objects reappeared further from the points of disappearance. Test it for yourself!

Here is a movie where all objects reappear where they disappear. Track the blinked objects:

Here is a movie where all objects reappear as if they had moved during the disappearance interval:

For further evidence for the non-predictive nature of parallel tracking, see Keane & Pylyshyn, 2006.

When the above elements are rotated by 48 degrees (relative to the display above), illusory contours are removed, and tracking becomes easier:

Illusory contours are not the only way that targets and distractors automatically group during tracking (Erlikhman et al., 2013).  Here is a movie demonstrating the distracting effect of color grouping:

Reduced depth inversion illusions in schizophrenia

 In depth inversion illusions (DIIs), the visual system perceives a hollow object as convex, presumably from having internalized the statistical regularity that most objects in the world protrude outward. People with schizophrenia are less likely to rely on this convexity assumption and therefore experience depth inversion illusions to a lesser extent (Keane et al., 2013).  We found that resistance to DIIs was especially pronounced among those with severe psychotic symptoms.  An example of the illusory stimulus is provided below, courtesy of my collaborator Thomas Papathomas. One side of the object is an ordinary convex mask and the other concave side produces an illusory appearance of convexity.