There are two existing theories--proposed by Harmon and Julesz in an influential 1973 Science article--for why the face is obscured by the blocks when seen from near, and revealed when seen from far. Pelli's results refute both theories.
The first theory proposes that a certain number of blocks per face are required to adequately represent the information that our visual system uses to see the face. According to this "critical band" theory, coarser blocks interfere by introducing visual "noise" into the "critical band." This theory predicts that faces with fewer than, say, 10 blocks across the face (from cheek to cheek), should never be recognizable as faces, and that faces with more blocks will always be recognizable. But, in fact, Close's portraits have variable numbers of blocks across the face, and every one is recognizable at some distances and not at others. The second theory, as mentioned above, assumes that backing away reveals the face because our eye optically blurs the blocks together. But this would predict a much smaller critical angle, about 0.1 deg, rather than the 0.3 deg Pelli found.
Instead, Pelli's results suggest that when the marks exceed the critical 0.3 degree size they are perceived as discrete objects, preventing the observer from seeing the face.
In the Science article, Pelli argues strenuously that Close deserves credit for
these findings, "One might suppose that he was a nave artist, obsessed by
grids, who innocently produced the coar
'"/>
Contact: Josh Plaut
josh.plaut@nyu.edu
212-998-6797
New York University
5-Aug-1999