The right-hand side of the picture is uncorrelated image noise,
yet many observers report seeing apparent ring structure on the right.
One could argue that perceptual sensitivity to deviations should decrease when the relevant information is grouped with a familiar model object (cf. depth examples in Sinha et al., 1996 and Mitchison and McKee, 1985). But perception would also seem to have the job of detecting small deviations from perceptual models that may or may not fit the data, and thus should have heightened sensitivity for detecting data that don't quite fit. This tension can be seen in some symmetry pictures I describe below.
Imagine image noise on the right-hand side of a picture, and the left-half of a symmetric object with added noise on the left-hand side. On the one hand, noise on the right might be expected to be grouped to fit the picture (or "model") on the left hand part of the picture. On the other hand, the noise on the right could be rejected as being part of the left-hand figure--especially in the context of objects like faces, for which detecting asymmetry is functionally important. In fact, putting noise on the right side and a half of a symmetric object on the left doesn't work very well --i.e. the concentric rings on the left are not very effective at inducing structure on the right. Also, half of a human face on the left seems not to recruit noise on the right. However, I seemed to have to achieved some small success at getting perception to organize noise on the right-hand side, with target structure on the left, by "bootstrapping" symmetry grouping with symmetric noise as illustrated below.
In the pictures below, random noise was generated on the right hand side of the pictures. The noise was then reflected about the vertical midline generating a symmetric noise pattern. (Actually, in this example, the noise symmetry isn't perfect, the noise having a one pixel diagonal displacement).
Then the left half of a source-image (half of a bull's eye target of concentric rings) was added to the noise on the left making a test picture (also shown at the top of the page).
Although the right-hand side of the image is still just noise, do you see additional apparent structure on the right side induced by the source-image on left-hand side of the picture? People often report seeing apparent rings on the right hand side--possibly perception's attempt to complete the bull's eye target.
Here is a version of Joanne's face (which she would not be pleased with if she knew!).
The next picture was generated the same way, except that the noise, rather than having a flat amplitude spectrum, has a low-pass spectrum matching that of the target object (Eli's face).
And finally, here is ...well judge for yourself.
Symmetric white noise + left half of a face source-image
Symmetric low-pass noise + left half of a face source-image Do you see illusory facial structure on the right side?
The original left half of face
