Neural Circuits, Vision, and Computation

## The retinal circuit Light is detected by **photoreceptors** (rods and cones). Their signals pass to **bipolar cells**, which in turn drive **ganglion cells**. Ganglion-cell axons form the **optic nerve**, the retina's only output to the brain. Two classes of interneurons shape the signal along the way: **horizontal cells** act laterally at the photoreceptor–bipolar synapse, and **amacrine cells** act laterally at the bipolar–ganglion synapse, often through inhibition. Signals split early into **ON pathways**, which respond to increases in light, and **OFF pathways**, which respond to decreases. Different ganglion-cell types each carry a different "view" of the scene in parallel. ## Receptive fields A ganglion cell's **receptive field** is the region of visual space, and the pattern of light within it, that changes its firing. Many ganglion cells have **center-surround** organization: an ON-center cell is excited by light in a small central region and inhibited by light in a surrounding ring. Consequence: a uniform bright field covering center *and* surround produces a weaker response than a bright spot confined to the center, because the surround cancels part of the center's drive. The cell signals **contrast** — differences in light across space — more than absolute brightness. ## Feature extraction Because of these circuits, ganglion cells report *features*: edges, local contrast, changes over time, and, for some cell types, motion in particular directions. The retina therefore **transforms** visual information before it ever reaches the brain. ## Not a camera A camera records each pixel's brightness faithfully. The retina does something different: it discards much of the redundant information in a scene and emphasizes what is informative. That is why the retina is a good model system for studying neural computation — inputs (light) can be precisely controlled and outputs (ganglion-cell spikes) can be recorded directly. ## Further reading - [Baccus Lab research overview](https://baccuslab.github.io/research/) - [Baccus Lab publications](https://baccuslab.github.io/publications/)