Long-Term Potentiation: Part Three: The Non-Synaptic Alternative

Abstract

Parts One and Two described the synaptic theory of memory: learning strengthens connections, and a molecular machine holds the strengthening in place. It descends from Hebb and it fits a great deal of data.

In 2024, Christian Hansel and Rafael Yuste proposed in Frontiers in Cellular Neuroscience that a second mechanism deserves equal standing — one that changes not the connections between neurons but how readily each neuron fires at all.

Their argument is that intrinsic excitability is not a modulator supporting synaptic plasticity, but a partner to it, and that the reason it has been dismissed is a claim about duration that turns out to be wrong.

1.Two Ways to Change a Network

Groups of neurons that fire together during a task form ensembles — the operational units of perception, movement and memory, called an engram when the task is remembering.

The standard account of how an ensemble forms is synaptic: connections between its members strengthen through long-term potentiation, binding them into a unit that reactivates together.

The alternative concerns the cell rather than the connection. A neuron's intrinsic excitability — how much input it takes to make it fire — is itself plastic, and enhanced excitability is widely observed across brain areas after behavioural learning.

The difference matters because the two do different work. Synaptic change is specific to a pair of cells. Excitability change applies to everything arriving at that neuron.

2.Why It Was Dismissed

Excitability changes after learning have generally been reported as transient — lasting tens of minutes to a few days in cortical structures and the amygdala.

That is the basis for the standard interpretation, known as the memory allocation hypothesis: raised excitability is a temporary state that determines which neurons get recruited into an ensemble, by making them more likely to undergo potentiation. Useful, but preparatory. The memory itself remains synaptic.

Hansel and Yuste accept the observation and reject the inference.

3.Two Models

Their first argument they call the iceberg model. Enhanced excitability is a signature of active ensembles, and because of it, synaptic connections that were subthreshold become suprathreshold — in the absence of any synaptic plasticity.

The same synapse, unchanged, becomes functionally effective because the cell receiving it now needs less to fire. Raise the water level and the same rock breaks the surface.

Their second is the permissive gate model. Enhanced excitability promotes propagation of dendritic potentials toward the cell body, which couples the size of a synaptic response more tightly to whether the neuron actually fires — and therefore to whether it participates in the ensemble.

On this account potentiation could occur at a dendrite and still fail to matter, if the cell's excitability does not permit that signal to reach the soma. The synaptic change is necessary and not sufficient.

4.The Duration Objection, Answered

The permissive gate model requires excitability to stay elevated, which appears to contradict its reputation as short-lived. The authors address this in two ways.

First, they propose that longer modifications are possible because the threshold for inducing intrinsic plasticity is low — so excitability might be switched on and off repeatedly at short intervals, producing a persistent state out of repeated brief ones.

Second, and more directly, they point out that duration is not universally short. In cerebellar Purkinje cells, excitability changes last days to weeks, which shows that in at least some circuits duration is not a limiting factor at all.

A property assumed to be transient because it was measured in the structures where it happens to be transient is a familiar error, and worth noting as a general reading habit.

5.A Division of Labour

The proposal they end on is not that synaptic plasticity is wrong but that it does a different job than assumed.

Synaptic plasticity defines the information content that a neuron receives, through the connectivity network it sits in. Plasticity of cell-autonomous excitability regulates something else — which of those neurons participate, and when.

Content and participation, in two separate systems.

That maps onto the engram findings covered elsewhere in this section with uncomfortable neatness. Amnesic engram cells retained their specific connectivity — the information content — while losing their synaptic potentiation. If potentiation governs access and participation rather than content, a memory could survive its loss and become unreachable, which is precisely what was observed.

We are not presenting that convergence as established. Two independent lines of argument pointing the same way is a reason to take a hypothesis seriously, not a demonstration that it is right.

Editorial Comment

MindHeaven® makes no claim relating to neuronal excitability, synaptic plasticity, ensembles or memory, and nothing in our formulations has been shown to influence them.

Across three parts: the phenomenon is real and bidirectional, the molecular account of its persistence is detailed and openly incomplete, and a serious alternative proposes that the thing everyone has been measuring governs access rather than content.

Fifty years of work on a single mechanism, and the central question — whether a specific memory is stored as a specific set of strengthened synapses — is still open. That is not a failure of the field. It is what a hard problem looks like when it is being worked on properly, and it is the standard against which any claim about "supporting plasticity" should be read.