June 17, 2014 | 1
Spanish surrealist painter Salvador Dali was a reader of Scientific American, and created one of his most iconic pieces based on a Scientific American article on face perception.
Stephen Macknik and I were thrilled to serve as scientific consultants for the “Marvels of Illusion” exhibit at the Dali Museum in St. Petersburg, FL. The exhibit, on the role of illusions in many of Dali’s masterpieces, will run through September (and may become a traveling exhibit afterwards… we’ll let you know). Go see it if you can, it’s an incredible collection!
We have excerpted a sample of the neuroscience write-up and images featured on the exhibit for today’s post:
Gala Contemplating the Mediterranean Sea, which at Twenty Meters Becomes the Portrait of Abraham Lincoln, is one of Dali’s finest ambiguous illusions.
Dali painted this piece after he learned about the pioneering work of Leon D. Harmon published in the November 1973 issue of Scientific American magazine, entitled, “The Recognition of Faces.” In it, Harmon had taken an image of Abraham Lincoln from a $5 bill and produced “block averaging” renderings of it. Block averaging means that the image is broken down into blocks of a grid and each block is filled in with the average gray-scale value for that block; essentially a single tone per pixel.
Harmon (et al) found that the minimum number of blocks needed for facial recognition was 16 x 16 blocks (256 total). Considering that many of those blocks were the background, essentially, we can recognize faces from about 150 blocks of information!
The homage to Mark Rothko (1903-1970) paid respect to a leading Abstract Expressionist who had recently committed suicide. Dali’s multiple blocks of color in varying progression of hues are evocative of the meditative “color field” paintings of Rothko.
Researchers refer to the fine details and abrupt changes of contrast in an image as high spatial frequencies. Low spatial frequencies correspond to the coarse details in an image and can convey global information about shape, orientation, and proportion.
The Gala Contemplating … illusion works because Gala is composed of high spatial frequencies, whereas Lincoln’s portrait is made of low spatial frequencies. Standing close, we focus on the high-spatial frequency details, fine differences of hue and value, and perceive Gala looking out a cruciform window at the sea and a crucifixion rendered in heavy impasto in the sky. The checkerboard framing the painting allows us to register the fine gradients of color and value that describe the form of Gala. The high-spatial frequencies at the edges of the large “pixels”, multiple dark and light blocks, dominate our perception at close distance, masking Lincoln’s face.
Standing far away (20 meters) from the painting, our vision mostly sees the low spatial frequencies: we now perceive crude, general elements about the scene, rather than fine–high frequency–details such as the edges of the colored blocks and Gala’s contour. We no longer see Gala, because the high spatial frequencies that delineate her body fade into the surrounding area (which holds comparable values of light to Gala’s figure), leaving you with only the overall shapes and shadings that make up Lincoln’s face (low spatial frequencies). Squinting your eyes at close distance to the painting also serves to blur and soften the edges, thus removing the high spatial frequency information and revealing the face information “hidden” in the low frequencies. Dali’s choice of hues, values, tones, textures, and saturation for the sea, clouds, and Gala’s body now become the appropriate shading for our perception of the skin on Lincoln’s face.
Once you perceive Lincoln’s face, your face neurons will provide additional details to help ‘fill-in” the image. One you associate Lincoln’s face with that particular group of squares, it becomes difficult to stop seeing it (a visual form of hysteresis, if you will). Dali’s (and Harmon’s) choice of Lincoln was not accidental: we recognize better familiar faces than unfamiliar ones. Young children may not recognize Lincoln per se, but perceive just a face.
When you move closer, the images are once again dominated by their fine details: Lincoln vanishes and Gala re-appears.
The perceptual processes at play are similar to those that allow you to see individual photographs when you approach a contemporary computer-generated photo-mosaic, but a single large image when you view it from a distance. Indeed, your whole visual experience is derived from integrating across a mosaic, because the photoreceptors in your retinas are an actual mosaic of individual light detectors!
Get 6 bi-monthly digital issues
+ 1yr of archive access for just $9.99