Neural Circuits, Vision, and Computation

## From light to spikes A **neuron** is a cell that receives signals, integrates them, and sends signals on. Neurons connect at **synapses**. At an **excitatory** synapse, the sending neuron makes the receiving neuron more likely to fire; at an **inhibitory** synapse, it makes firing less likely. Many neurons communicate with brief electrical pulses called **spikes** (action potentials). A group of connected neurons is a **neural circuit**. ## What "compute" means here When neuroscientists say a circuit *computes*, they do not mean it runs software. They mean it **transforms** its input into an output in a systematic way. For vision, the chain looks like this: visual stimulus → neural circuit → transformation → spike pattern → representation The output spike pattern is not a copy of the input. It emphasizes some features (for example, edges or sudden changes) and discards others. That selective transformation *is* the computation. ## Three levels of explanation A useful habit is to ask three different questions about any circuit: 1. **Function** — what problem does the computation solve for the animal? (e.g., detect a moving predator against a moving background.) 2. **Computation** — what transformation of input to output is performed? (e.g., respond to local motion that differs from global motion.) 3. **Mechanism** — how do specific cells and synapses implement it? (e.g., which inhibitory neurons suppress responses during global motion.) These levels are linked but not interchangeable. Knowing the function does not tell you the mechanism, and two different mechanisms could implement the same computation. Much of modern neuroscience is about connecting these levels with evidence. ## Why this matters Throughout this primer you will see the same question asked repeatedly: *what is the evidence that links this computation to this mechanism?* Keeping the three levels separate helps you avoid a common mistake — treating a description of what a circuit does as an explanation of how it does it. ## Further reading - [Baccus Lab research overview](https://baccuslab.github.io/research/) - [Baccus Lab publications](https://baccuslab.github.io/publications/)