What stores a memory? The obvious answer is a neuron. One cell, one fact. Like a file in a directory.
Alexander Bain saw it differently. In Mind and body (1873), he proposed that the information is not in the neurons. It is in the connections.
The same set of neurons can store completely different information depending on how they are wired. A neuron does not mean anything by itself. Its meaning comes from which other neurons it talks to, and how strongly.
Bain described what would later be called neural groupings: neurons that excite and stimulate each other. Different combinations of inputs to the same group produce different outputs. Even the intensity of activation matters — a weak signal to neuron might activate but not , while a strong signal activates both.
The pattern of connections, not the identity of the neurons, determines the result.
Bain also proposed a learning mechanism, three-quarters of a century before Donald Hebb:
“When two impressions concur, or closely succeed one another, the nerve-currents find some bridge or place of continuity, better or worse, according to the abundance of nerve-matter available for the transition.”
This is Hebbian learning in everything but name: neurons that fire together wire together. The connection strengthens when both sides are active at the same time. The bridge is the memory.
Bain estimated the brain would need one million neurons and five billion connections to account for 200,000 acquisitions. By 1883, he worried he had underestimated — partial associations and learning mechanisms would require far more. By 1903, near the end of his life, he recanted entirely. The numbers seemed impossible.
He was wrong to doubt. The human brain has roughly 86 billion neurons and over 100 trillion synapses. The connection count he found overwhelming was, if anything, an underestimate.
Connectionism begins with this shift. Before Bain, associations were abstract mental links with no physical story. After Bain, the question was no longer whether ideas are connected, but how those connections are physically realised. Learning is the process of rewiring them.