Afterimages: Their neural substrates and their role as short-term memory in the human brain’s computation.

Charles Q. Wu

Keywords: neural computation, blind spot, afterimage, short-term memory

TL;DR: This paper establishes a neural theory for afterimages: pinpointing their neural substrates to layer 4s in the human brain's cortical areas and proposing their computational role in the human brain's visual computation.

Abstract: Afterimages are seemingly simple yet very intriguing visual phenomena. Presently, essentially all the textbooks in vision science and in perceptual psychology introduce these phenomena; meanwhile they also ubiquitously subscribe to an incorrect view that afterimages are due to some peripheral adaptation mechanisms occurring in the retina of the eye. The contrasting view is that afterimages originate in the brain: This view is not new at all, but only recently there has been accumulating a multitude of evidence pointing to its truthfulness. Two recent and critical lines of advances related to afterimages in vision science are as follows: 1. LeVay et al. (1985) discovered a representation of the physiological blind spot in Layer 4 of the cortical area V1 (hereafter, V1-L4) in the macaque monkey’s brain, and Adams et al. (2007) discovered the same in the human brain; 2. Wu (2024) re-discovered the phenomenon of an observer seeing their own blind spot as an afterimage and correlated this phenomenon to the above neuroanatomical findings. Together, these advances essentially pinpoint the first-stage neural substrate for afterimages to V1-L4. Here we build upon these advances and establish a neural theory of afterimages consisting of the following tenets: 1. Positive and negative afterimages share the same neural substrate; 2. Afterimages should be viewed as short-term memory (STM) in the brain instead of as peripheral adaptation in the retina; 3. In terms of the neural computational architecture of any cortical area, STM is sandwiched between a feedforward neural network and a feedback counterpart—it may play a computational role for variable binding. Finally, we discuss potentially fruitful bi-directional interactions between perceptual & neuroscientific researches in biological vision on the one hand and computational & engineering endeavors on artificial vision on the other.

Primary Area: Neuroscience and cognitive science (e.g., neural coding, brain-computer interfaces)

Flagged For Ethics Review: true

Submission Number: 20418