Reading focus
Syllabus: How Do Neural Circuits Turn Sensation into Action?
Compare feedforward and recurrent explanations of an organized response, using Hubel and Wiesel's findings as a concrete case.
Why you’re reading it. A cell's selectivity must fit an account of how connected populations operate; it does not choose that account by itself.
A question to consider. Propose a population recording or perturbation that could distinguish two networks producing the same single-cell response.
Why this reading
Understanding a cell's properties and understanding a circuit's activity are related but different tasks. Yuste argues that interacting populations of neurons can generate patterns that require explanations at the network level. The paper links a change in research methods with a proposed change in the unit of explanation.
Before you read
A neuron is a nerve cell; its axon carries signals toward other cells, often through synapses onto their dendrites. Excitation and inhibition alter the likelihood of activity. A circuit is a connected group of neurons. Feedforward connections pass activity onward; recurrent connections feed activity back. A model's connection weight represents the influence of one unit on another. An attractor is a pattern toward which a system's activity tends to return.
Read for
- Separate the claim that neurons are distinct cells from the stronger claim that individual cells are sufficient functional units.
- Use Figure 4 to compare feedforward processing, recurrent activity, and stable network states.
- Ask when an observation merely fits a network model and when it could discriminate between competing models.
Consider afterwards
- What would recording many neurons add if each neuron were still analyzed independently?
- What intervention could test whether a recurring population pattern contributes to behavior?
Editorial context
This is a selective historical and theoretical perspective, not proof that one network model explains the brain. Rubinov's response distinguishes useful artificial networks from biologically validated models. In his reply, Yuste agrees that correspondence with observations falls short of rigorous causal validation. Network explanations must connect with cellular evidence as well as behavior.