From Neural Mechanisms of Short-Term Memory to the Computational Architecture of the Human Brain

Charles Q. Wu

Keywords: Short-Term Memory, Primary Visual Cortex, V1-L4, Computational Architecture, Human Brain, Color Sensation, Visual Consciousness

TL;DR: This paper utilizes the visual phenomenon of seeing the physiological blind spot as an afterimage to locate the neural site for visual short-term memory and to formulate a computational architecture of the human brain.

Abstract: Since 1840, it has been known that the cerebral cortex of the human brain mainly consists of six layers. What is the computational architecture of this multi-layer computing machinery? This problem is certainly worth investigating and is interesting to many researchers in cognitive science \& neuroscience on the one hand and computer science \& engineering on the other. The present paper tackles this problem from a new perspective—namely, starting from the visual phenomenon of seeing blind spots as afterimages, to determining the neural locus of afterimages, to conceptualizing afterimages as visual short-term memory, and finally to formulating a computational architecture of the brain. First, presently, all the textbooks in vision science and perceptual psychology do introduce afterimages but also ubiquitously assume them as due to some peripheral adaptation mechanisms occurring in the retina of the eye; here, the long-forgotten phenomenon of seeing blind spots as afterimages is revived and is used as an "instantia crucis" to demonstrate the cortical origin of afterimages. Second, correlating this phenomenon with the neuroanatomical fact that the blind spot is represented within V1-L4 (i.e., Layer 4 in the visual cortical area V1), it is deduced that V1-L4 is the neural site for afterimages. Third, it is further deduced that the overall computational architecture of the brain is as follows: In each visual cortical area, the superficial layers (i.e., L2\&3) constitute a feedforward neural network for transforming information from one level to the next, the deep layers (i.e., L5\&6) constitute a feedback one for the reverse transform, and the middle layer (i.e., L4) serves as short-term memory. Finally, it is discussed how the computational architecture of the human brain conceptualized in this manner may help foster interdisciplinary interactions towards a more complete understanding of the human brain.

Primary Area: applications to neuroscience & cognitive science

Submission Number: 8038