Single Neurons Rival Networks in Solving Complex Puzzles

The brain is often viewed as a vast web where intelligence emerges strictly from networks, but new research suggests individual cells are far more powerful than assumed. An in silico study—conducted via computer modelling—reveals that single striatal projection neurons (SPNs) can solve the 'nonlinear feature binding problem', a complex integration task previously thought to require groups of neurons.

The researchers utilised a biophysically detailed model incorporating a local, calcium-based synaptic learning rule. This mechanism relies on 'dendritic plateau potentials'—sustained voltage elevations in the neuron's branches—alongside dopaminergic reward signals. To maintain stability, the system employs metaplasticity, a self-adjusting method for synaptic weights.

Crucially, the study shows that when dendrites process signals nonlinearly rather than simply adding them up, a single neuron can execute complex computations. Furthermore, an inhibitory plasticity mechanism helps compartmentalise these dendrites, enhancing efficiency. This discovery highlights that single neurons are not merely simple relays, but sophisticated processors capable of intricate decision-making.