What is long-term potentiation (LTP)? How synapses get stronger and stay that way

Long-term potentiation (LTP) is a lasting increase in the strength of a synapse caused by brief, intense or well-timed activity. After LTP, the same input produces a larger response in the receiving neuron for hours or longer. First reported by Bliss and Lømo in 1973, long-term potentiation is the most studied candidate mechanism for how neurons store learning and memory.

Key takeaways

• LTP is one form of synaptic plasticity: a persistent, input-specific strengthening of a connection between neurons.
• Bliss and Lømo found it in the rabbit hippocampus in 1973, where potentiation lasted from 30 minutes to 10 hours after brief stimulation trains.
• At many synapses, induction depends on NMDA receptors acting as coincidence detectors, and expression involves adding AMPA receptors.
• LTP has properties that make it a plausible memory mechanism: input specificity, cooperativity and associativity.
• For organoid intelligence, LTP is the clearest biological analogue of a weight update, and inducing it on demand is a core training challenge.

What is long-term potentiation and how was it discovered? #

The idea that learning changes connections between neurons predates LTP by decades. Donald Hebb proposed in 1949 that a connection strengthens when one cell repeatedly helps fire another, the principle now called Hebbian learning. What was missing was a direct physiological demonstration.

Bliss and Lømo supplied it in The Journal of Physiology (1973). In rabbits anaesthetized with urethane, they stimulated the perforant path, the main input into the dentate gyrus of the hippocampus, and recorded the population response of granule cells with extracellular electrodes. After one or more conditioning trains (10 to 20 pulses per second for 10 to 15 seconds, or 100 per second for 3 to 4 seconds), responses to single test pulses stayed potentiated for periods from 30 minutes to 10 hours in 15 of 18 rabbits. They suggested two contributing mechanisms: more efficient synaptic transmission and increased excitability of the granule cells.

The hippocampus was a meaningful place to find this, because damage there was already known to impair the formation of new memories. Twenty years later, Bliss and Collingridge (1993, Nature) reviewed the evidence under the title “A synaptic model of memory”.

How does long-term potentiation work? #

The best-characterized form of LTP occurs at excitatory synapses in the CA1 region of the hippocampus, which use glutamate as their transmitter. The process is usually split into induction (what triggers it) and expression (what physically changes).

Induction: the NMDA receptor as a coincidence detector #

At rest, the NMDA type of glutamate receptor is blocked by magnesium ions. It only passes current, including calcium, when two things happen together: glutamate is released from the sending neuron, and the receiving neuron is already depolarized enough to expel the magnesium. This makes the receptor a molecular detector of coincident pre- and postsynaptic activity, which is the cellular version of Hebb’s rule. A strong, brief rise in postsynaptic calcium activates enzymes, notably the kinase CaMKII, that set potentiation in motion (Malenka & Bear, 2004).

Expression: more and better receptors #

The main expression mechanism at CA1 synapses is an increase in the number and function of AMPA receptors, the glutamate receptors that carry most fast excitatory current. More AMPA receptors means a larger response to the same release of glutamate. For years researchers argued over whether LTP was expressed presynaptically (more transmitter released) or postsynaptically (larger response). Nicoll’s 2017 history in Neuron describes how the evidence at CA1 moved largely toward a postsynaptic account, while noting that other synapses use different mechanisms.

Early and late phases #

LTP is commonly divided into an early phase, which relies on modifying proteins already at the synapse, and a late phase that requires new gene expression and protein synthesis. The late phase is often linked to structural changes such as enlarged dendritic spines.

What properties make LTP a model for memory? #

Bliss and Collingridge (1993) highlighted three properties that make LTP a good fit for an information-storage mechanism:

1. Input specificity. Only the synapses that were active during induction are potentiated, so a single neuron can store many independent associations.
2. Cooperativity. A threshold number of inputs must be active together; weak stimulation does not induce LTP.
3. Associativity. A weak input can be potentiated if it is active at the same time as a strong input onto the same cell, which resembles associative learning.

Timing adds a fourth dimension. In spike-timing-dependent plasticity, a presynaptic spike that arrives shortly before a postsynaptic action potential tends to produce LTP, and the reverse order tends to produce long-term depression (LTD).

How do LTP, LTD and other forms of plasticity compare? #

Malenka and Bear (2004) titled their Neuron review “an embarrassment of riches” because LTP turned out to be not one mechanism but a family. Different synapses, cell types and brain regions use different molecular routes.

A further layer is metaplasticity, a term popularized by Abraham and Bear (1996, Trends in Neurosciences): prior activity changes how easily a synapse can later undergo LTP or LTD. In other words, the history of a network shapes its future learning rules.

How is LTP measured? #

In brain slices, LTP is classically measured by recording the field excitatory postsynaptic potential with an extracellular electrode and comparing its slope before and after an induction protocol. Single-cell patch clamp recordings give a more direct view of synaptic currents in one neuron. Both approaches are part of the wider toolkit of electrophysiology.

Cultured networks and organoids on a microelectrode array are harder. Electrodes record spikes and local field potentials from many neurons, so researchers usually infer potentiation from lasting changes in evoked responses rather than measuring one synapse. Spontaneous activity such as network bursts can also mask or reverse changes, which makes careful baselines and controls essential.

What are the debates and open questions? #

• LTP versus memory. LTP is usually induced with artificial stimulation. Many studies link LTP-like changes to learning, but showing that a specific memory is stored by a specific potentiation remains difficult.
• Locus of expression. The presynaptic versus postsynaptic debate was largely settled for CA1, but the answer differs between synapse types (Nicoll, 2017).
• Stability. Synaptic proteins turn over constantly, so how potentiation persists for days or years is still being studied.
• Interaction with homeostasis. Unchecked LTP would push networks toward saturation. Stabilizing mechanisms must work alongside it without erasing what was learned.

Why does long-term potentiation matter for organoid intelligence? #

In machine learning, training means changing numerical weights. In biocomputing with living neurons, nobody can type in a new synaptic strength. The only lever is experience: patterns of stimulation and feedback that trigger plasticity. LTP is the best-understood way that experience becomes a lasting change in a circuit, so it frames several practical questions for anyone training cultures or brain organoids:

• Which stimulation patterns, delivered through electrodes, reliably induce lasting potentiation in a given culture?
• Can closed-loop stimulation, where feedback depends on the network’s own output, steer potentiation toward task-relevant pathways?
• How long do changes persist, and how quickly do spontaneous activity and homeostatic processes erode them?
• What readout counts as evidence of LTP when electrodes see network activity rather than single synapses?

Smirnova et al. (2023, Frontiers in Science) identified learning and memory in organoids as a central goal for organoid intelligence. Reaching it will require organoid synapses mature enough to support robust plasticity, and measurement methods that can confirm it. The experiments covered in how neurons learn in a dish show the closed-loop approach in cultured neurons, and microelectrode arrays explained describes how the recordings are made.

Turning induction protocols, baselines and readouts into versioned code that can be rerun and compared across cultures is part of what we are building at Organoid Programming.

Conclusion #

Long-term potentiation is the lasting strengthening of synapses after patterned activity, and it remains the most studied model of how neurons store information. Its NMDA receptor-based coincidence detection gives Hebb’s idea a concrete mechanism, and its specificity and associativity explain why it has been linked to memory for half a century. For living biocomputers, learning to induce and verify long-term potentiation on demand is one of the steps between a responsive culture and one that genuinely learns.

Frequently asked questions

What is long-term potentiation in simple terms?

Long-term potentiation is what happens when a connection between two neurons gets stronger after it has been used intensely, and stays stronger afterward. Once potentiated, the same input signal produces a bigger response in the receiving neuron. Because the change can last hours or much longer, LTP is widely studied as a possible cellular mechanism for storing memories.

Who discovered long-term potentiation?

Tim Bliss and Terje Lømo described long-term potentiation in a 1973 paper in The Journal of Physiology. Working in anaesthetized rabbits, they stimulated the perforant path into the hippocampus with brief repetitive trains and found that responses of dentate granule cells stayed enhanced for periods ranging from 30 minutes to 10 hours in most animals tested.

What is the difference between LTP and LTD?

LTP is a lasting increase in synaptic strength, while long-term depression (LTD) is a lasting decrease. At many synapses both depend on calcium entering through NMDA receptors: strong, brief calcium signals tend to produce LTP and weaker, prolonged signals tend to produce LTD. Having both lets networks strengthen useful connections and weaken others, so they do not simply saturate.

Is LTP the same as memory?

Not exactly. LTP is a laboratory phenomenon, usually induced by artificial stimulation, that shares key properties with memory: it is fast to induce, long lasting and specific to the inputs involved. Many experiments link LTP-like changes to learning, but whether natural memories are stored by the same mechanisms remains an active research question rather than settled fact.

Can brain organoids show long-term potentiation?

Brain organoids contain neurons that form functional synapses, so LTP-like changes are plausible, but measuring them is harder than in classic brain slice experiments. Organoid synapses are developmentally immature, and extracellular electrode arrays record network activity rather than individual synaptic strengths. Showing reliable, lasting potentiation in organoids is a goal that reviews of organoid intelligence have identified as important.

Why does LTP matter for biocomputing?

In biocomputing there are no numerical weights to edit. If a living network is to learn, experience must change its synapses, and LTP is the best-understood way that happens. Knowing which stimulation patterns induce lasting potentiation, and how to confirm it, helps researchers design training protocols for cultured neurons and organoids.

Sources

1. Bliss & Lømo (1973). Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. The Journal of Physiology.
2. Bliss & Collingridge (1993). A synaptic model of memory: long-term potentiation in the hippocampus. Nature.
3. Malenka & Bear (2004). LTP and LTD: an embarrassment of riches. Neuron.
4. Nicoll (2017). A brief history of long-term potentiation. Neuron.
5. Hebb (1949). The Organization of Behavior: A Neuropsychological Theory. Wiley. Reissued by Psychology Press.
6. Abraham & Bear (1996). Metaplasticity: the plasticity of synaptic plasticity. Trends in Neurosciences.
7. Smirnova et al. (2023). Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Frontiers in Science.