Depth, Shape, Rigidity, from Cast Shadows

Together with Pascal Mamassian, David Knill at U. Rochester., and Isabelle Bülthoff and Heinrich Bülthoff at the (Max-Planck Institute, Tubingen, Germany), we have been investigating human perception of depth from cast shadows.

For 12 minutes worth of assorted shadow demonstrations, you can download the 57Mbyte Shadow Demos quicktime movie.

Changing size vs. shadows


There are many cues to changes in depth. Size change and moving shadows are just two. Size change is well-known and has been studied. The cue is straightforward: if the image of an object gets bigger, the object is probably getting closer. If the image doesn't change, the object must be staying put.

The moving shadow cue is also simple: the farther a shadow moves from the object casting it, the farther the object is from the background. Moving shadows have not been systematically studied as an important cue to depth change. One reason may be the presumption that cast shadow information is not strong, because the cue is indirect--vision has to figure out which is the object and which is the shadow in order to use the cue. If so, we might expect that a moving cast shadow is probably a weak cue which wouldn't overide the changing size cue.

With computer graphics, we can test this expectation, by holding one cue fixed (size change = 0), and varying the other to see which one wins (Kersten, Knill, Mamassian & Bülthoff, 1996).

By playing the QuickTime movie below, you can observe and verify for yourself the following conclusion. For the green square, the shadow wins. Moving shadows are a strong cue to a change in depth.

This effect works best with changing blur of the shadow penumbra. The apparent motion in depth is less likely to be experienced if the shadow is sharp, or if the shadow is above, rather than below the object.

Try downloading the above simple demonstration described in the Nature paper (Kersten et al., 1996):

(Quicktime 160x120, compressed Cinepak) - shadow movie (299K). If you have an account with Nature, you can also download the movie from Nature, by going to Supplementary Information.


The ball-in-a-box shadow illusion

In the next demonstration ( ball-in-a-box, 1.3Mbytes), we hold image size fixed, but allow the position of the the object to change in the image. The ball follows a diagonal trajectory inside a box. The ball's shadow first moves diagonally in a trajectory parallel to the ball, then it moves horizontally.

The ball's trajectory is the same for both segments of the animation. The apparent motion in depth of the ball is strikingly different in the two cases. When the shadow is diagonal, the ball appears to slide along the floor to the back of the box. When the shadow trajectory is horizontal, the ball appears to rise above the floor of the box.

In contrast to the square-over-checkerboard, the ball-in-a-box shadow illusion is quite effective even when the shadow has a sharp edge.


For more information on the ball and shadow illusion see:

A related neuropsychological result

At the 1996 Cognitive Neuroscience meeting this year, François Michel and Marie-Anne Henaff reported results of showing the ball-in-a-box shadow demonstration to an occipito-parietal patient and an occipito-temporal patient. The occipito-parietal patient did not report seeing the "ball-in-the box" shadow illusion--the ball did not appear to "fly" above the floor of the box. In contrast to the parietal subject, the occipito-temporal patient repored seeing the ball fly above the floor of the box. This is what a normal observer would see. (

Daniel Kersten
Psychology Department
University of Minnesota
75 East River Road
Minneapolis, MN 55455


Back to Vision Lab Page