Working memory appears to be based on simple networks of “small world” (maximally connected) neurons in the prefrontal cortex that participate in self-sustaining bursts of electrical activity.

Northwestern University researchers have created a model of these networks, using simple neurons that when activated would activate their neighbours for a brief period: an activating pulse travelled through the network and then disappeared at the fringes. They then added shortcuts to connect distant parts of the network — the key to a small world.

They found that when 10 to 20 per cent of the neurons participated in shortcuts, the network formed self-sustaining loops of activity.

The new work may be an important step towards a theory of how the brain works.