141 — Spatial Information in a Non-retinotopic Visual Cortex

Most of the visual cortex studies I’ve read have been about mammalian vision. But mammals are not the only organisms with cortex, or at least tissue resembling cortex. This paper explores turtle visual cortex, in an area known as dCx. This dorsal region plays the role of the turtle primary vision center — as close as you can find to the homolog of human V1.

But unlike V1, Fournier & Müller et al confirm the existing knowledge that while dCx receives direct inputs from LGN, its layout is not retinotopic. They expand upon this understanding by recording from both individual cells (with electrophysiological recording) and populations of cells. They show that, aside from some slight preferences of caudal dCx for rostral LGN inputs (and vice versa), dCx doesn’t have any clear layout pattern.

This is interesting because the simple, three-layer cortex of a reptile is a much simpler substrate for understanding vision than the complex, deep cortex of mammals. What, if not “pixel-level” information, is dCx responding to?

While the turtles watched movies, the researchers watched the turtles. In particular, they noticed that certain cells of dCx responded to scene-level stimuli: When there was a cut, or the film started or stopped, there was a burst of activity. When the same stimulus was played repeatedly, these responses died away.

This style of scene-level parsing in mammals is reserved for “later”, down-stream brain areas. Does this mean that turtles are using dCx as both early signal processing as well as higher-level object recognition and interpretation? If so, does this mean that mammalian brains, with discrete areas for low-level and high-level vision, evolved from an earlier system where all of these tasks were colocated?