An important debate among those who study vision has been settled with the publication of an article in the December 1995 issue of the British journal, Nature. The article also points out the dangers of investigating complex processes like vision with models that may be mathematically sound, but which lack some of the characteristics of real-world objects and surfaces.
"There are significant differences between artificial stimuli and real ones that animals see," said the paper's author, Peter Bex, a postdoctoral researcher in the Department of Brain and Cognitive Science at Rochester. "In itself, that might not be a problem; but the difficulty arises because the use of artificial images to test how the visual system works can produce strange results."
Specifically, two competing camps with alternative explanations for how the vision system detects motion have been able to support their respective models using these artificial images.
One group has contended that animals see motion by determining where a shape's edges are located, and then by noticing when those edges change position.
The second camp explains motion detection using energy-based mechanisms. This camp believes that there are groups of cells in the brains of mammals that specialize in detecting motion within limited spatial ranges. Some cells specialize in seeing only large ranges of motion, like the large structure of a dog walking. Others detect mid-sized ranges of motion, like the dog's tail wagging, and still others see only small ranges, like the fine structure of a dog's nostrils twitching as they catch a scent.
In his study, Bex had subjects view the artificial images (moving random dot patterns) through a sequence of filters that screened out certain energy components so that the patterns more closely resembled real-world images. He asked subjects to say whether or not they saw movement from one image to the next. By analyzing the pattern of responses, Bex was able to show conclusively that the edge-based model is a faulty explanation for how mammalian brains detect motion, and that the energy-based model more accurately reflects how brains actually work.