How do neurons use context from their neighbours?

Contextual Modulation

A neuron doesn't work alone — it "talks" to neighbouring neurons detecting the same orientation.

• A 45°-detecting neuron also communicates with neighbours tuned to 45° via horizontal connections — this builds context about the surroundings
• If neighbouring neurons aren't firing (random noise), contextual modulation tells the neuron to fire more strongly — reinforcing real edges over noise
• This is how we see the leaves of a tree: you register the outer leaves with clear edges, but inner leaves remain blurry — you can infer their edges but don't actually perceive them sharply
• The common "Attention" mechanism in deep learning is similar to this, yet it still lacks these horizontal connections between neurons at the same level

Inhibitory Surround — End-Stopping

Not every 45° line belongs to the object — it could be background. The brain handles this with inhibitory zones.

• Inhibitory zones sit at both ends of the receptive field along the same axis
• Short lines stay within the excitatory zone → neuron fires strongly (likely an object edge)
• Long lines extend into the inhibitory zones → neuron fires weakly (likely background — backgrounds tend to have uniform, longer lines, plus textures and surfaces)

Moving Stimuli & Saccades

• Visual cortex neurons go silent if the image is perfectly stable on the retina — they need change to fire
• So our eyes make tiny involuntary movements called saccades to keep the image shifting on the retina
• This is fascinating — without constant micro-movements, vision would literally fade to nothing

Next

• Next step: understand how the brain perceives depth