Simulated single-unit recording · retina
One retinal ganglion cell, one patch of visual field. Put light where you like and watch what the axon sends to the brain — the cell reports contrast across its receptive field, not how bright the world is.
Every readout on this page comes from the same calculation: how much light falls on the excitatory centre, minus how much falls on the antagonistic surround.
The rate climbs while the spot fills the centre, peaks when it just covers it, then falls away as light spills into the surround. Bigger is not better.
Centre and surround cancel and the cell sits near its maintained discharge. Switch the contrast to negative — darkness does the same. The cell reports differences, not levels.
Nothing much happens until the border crosses the receptive field. An ON cell fires hardest with its centre on the bright side and part of its surround in shadow — the neural basis of Mach bands.
Every response inverts. Turn on the 1 Hz flash: the ON cell fires at light onset, the OFF cell stays quiet during the light and discharges when it goes out.
Coarse gratings put one whole bar over centre and surround alike — little response. Fine gratings put both a light and a dark bar inside the centre, which cancels. In between, the cell is at its best.
Without horizontal- and amacrine-cell feedback the cell becomes a plain light meter: diffuse light now drives it hard and size tuning disappears.
How it's modelled. The receptive field is a difference of Gaussians — an excitatory centre of width σ and an antagonistic surround of width 3.2 σ, each normalised to unit volume so that the surround-strength slider sets the balance directly. Net drive is the stimulus contrast integrated against that profile, passed through a mild transient/sustained adaptation (τ ≈ 0.22 s) and a saturating nonlinearity, then added to a 12 spikes/s maintained discharge; spikes are drawn from a Poisson process at the resulting rate. It is a teaching caricature of a cat X-cell or a primate parasol cell — real cells add non-linear subunits, light adaptation and ON/OFF asymmetries this page leaves out.