Vision by prediction
Chapter 4 endnote 12, from Lisa Feldman Barrett.
Some context is:
You are always being bombarded by sensory input. One human retina transmits as much visual data as a fully loaded computer network connection in every waking moment...
Not only that, but the sensory input that reaches your brain from the outside world is woefully incomplete. Think of it this way: there is light reflecting from every point in your immediate physical surroundings. Your retina is not capable of taking in the immensity of this visual input. Evolution has tuned the human retina to accept particular patterns of input, such as edges and corners, and discard others, such as infrared wavelengths. For efficiency, your retina also throws away input that’s redundant, or that your brain is capable of reconstructing from past experience. Scientists disagree on how much of the world the retina filters out—estimates vary by a factor of 10—but the input is trimmed down enough to travel efficiently along your optic nerve to your brain. The connections between neurons (synapses) are noisy, and so up to half the remaining information is lost as it makes its way along the optic nerve to the brain. Yet somehow, you see the world in fine detail and vivid color. If someone throws a baseball to you, somehow you can catch it.
How is this possible with such incomplete sensory information? The secret ingredient is prediction. Your brain is always predicting which sensory inputs will arrive from the world, and which actions are required to deal with them, filling in the absent information. Prediction is the reason why you can see objects like baseballs, bees, and apples even though complete visual input from the world never reaches your primary visual cortex.
Notes on the Notes
- Balasubramanian, V. 2015. Heterogeneity and efficiency in the brain. Proceedings of the IEEE, 103(8): 1346-1358.
- Vijay Balasubramanian, personal communication, October 17, 2015.
- Sterling, Peter, and Simon Laughlin. 2015. Principles of Neural Design. Cambridge, MA: MIT Press.