In 1949, the Canadian psychologist Donald Hebb — working in Montreal — published The Organization of Behavior with a postulate so concise it would be repeated for the rest of the century: when an axon of cell A is near enough to excite a cell B and repeatedly takes part in firing it, some growth process takes place in one or both cells such that A's efficiency, as one of the cells firing B, is increased. In the workshop slogan: cells that fire together, wire together. The postulate was theoretical. Twenty-four years later, in 1973, Tim Bliss and Terje Lømo in Oslo provided the experimental confirmation: stimulating a presynaptic pathway in the hippocampus of an anesthetized rabbit at high frequency caused a long-lasting increase in the postsynaptic response to subsequent test pulses. They called the phenomenon long-term potentiation — LTP.
Long-term potentiation is a durable strengthening of synaptic transmission produced by patterned activity. The canonical version, studied in the hippocampal CA1 region, requires coincidence between strong presynaptic activity (releasing glutamate) and postsynaptic depolarization, which is what releases the magnesium block on NMDA receptors and lets them flux calcium. The calcium influx triggers a cascade that inserts new AMPA receptors into the postsynaptic membrane, enlarges the dendritic spine, and in late-phase LTP changes gene expression and synthesizes new proteins to make the change durable. Subsequent presynaptic spikes therefore produce larger postsynaptic responses — the synapse has been potentiated. Spike-timing-dependent plasticity refined the rule: the presynaptic spike must arrive just before the postsynaptic spike (within about twenty milliseconds) for potentiation; the reverse order produces depression. The arrangement implements causal learning, strengthening connections only when the presynaptic event plausibly helped cause the postsynaptic event. The brain stores different kinds of memory in different places. Episodic memory for specific events depends on the hippocampus and medial temporal lobe, as the patient H.M.'s case made famous when bilateral hippocampectomy abolished his ability to form new episodic memories while leaving older ones intact. Semantic memory for general knowledge eventually consolidates into cortical storage on a timescale of weeks to years. Procedural memory for how to ride a bike depends on basal ganglia and cerebellum. Working memory runs on persistent prefrontal-cortex activity rather than LTP at all. The system is reconstructive rather than reproductive — each act of recall is partly a re-encoding, and retrieval can modify the memory through reconsolidation. This is the structural reason for Ebbinghaus's forgetting curves, Loftus's false-memory results, and the unreliability of eyewitness testimony.
Alzheimer's disease — the leading cause of dementia — is characterized by amyloid plaques and tau tangles that disrupt synaptic function; new amyloid-clearing antibodies (lecanemab, donanemab) are the first FDA-approved disease-modifying drugs (~2023–24), with modest clinical effect. PTSD is increasingly conceptualized as a consolidation pathology — emotionally salient memories overconsolidating — and treated through reconsolidation-based protocols (propranolol given during recall) and MDMA-assisted psychotherapy. Optogenetic memory manipulation in animal models — implanting and erasing specific memories — has moved from possibility to routine technique. Spaced repetition — distributed practice produces better long-term retention than massed practice — is the basis of every flashcard app (Anki, Duolingo, RemNote, Polymathic itself); the underlying spacing effect is a direct consequence of LTP and consolidation dynamics.