Blogger Comments:
I — The Myth of the Stable Neuron
Why neuroscience cannot decide what its basic units are
Neuroscience still quietly inherits a metaphysics in which the brain is composed of stable units—neurons—that reliably “do things.” This assumption is rarely defended; it is simply embedded in experimental design, analysis pipelines, and interpretation.
The logic is familiar:
- a stimulus is presented
- a neuron responds
- repeated trials reveal a stable tuning curve
- therefore, the neuron “encodes” something
This produces a deceptively clean picture: a world of features mirrored by a world of neural tokens.
But this picture is not discovered—it is constructed by the experimental frame itself. It depends on:
- discretising continuous activity into unit responses
- holding behavioural context fixed
- assuming that temporal variability is noise around a stable mapping
The problem is not that this model is wrong in detail. The problem is that it defines correctness in advance of what counts as variation.
Once long-term recordings become possible, the supposed stability begins to dissolve—not dramatically, but systematically. The neuron stops behaving like a unit of function and starts behaving like a participant in a shifting coordination field.
At that point, neuroscience does not revise its ontology. It renames the anomaly.
Which brings us to “representational drift.”
II — Representational Drift is Not an Explanation
It is a label for the failure of a model that insists on stability
“Representational drift” is often treated as a discovery: neurons change their tuning over time even when behaviour is held constant.
But notice what must already be assumed for this to be a “problem”:
- that there exists a stable representational mapping
- that neural populations are meant to preserve it
- that time should not affect identity of encoding
Drift is then defined as:
deviation from a presumed invariant code
But this is not an explanation. It is a diagnostic term for when a representational model stops fitting its own output.
The deeper issue is that “representation” does not degrade gracefully under empirical stress. It either holds, or it becomes a residual category for everything that no longer behaves as expected.
So the field responds by multiplying distinctions:
- drift
- noise
- plasticity
- remapping
- context dependence
Each term is an attempt to protect the original assumption:
that there must still be a stable code somewhere underneath.
But what if that assumption is optional?
What if the system never contained a stable code in the first place?
III — Edelman Already Solved the Problem (And Was Ignored Anyway)
Population dynamics without the need for representational units
Neural Darwinism (Neuronal Group Selection Theory) already displaced the unit-based picture decades ago.
Its core move is simple but radical:
- neural function is not assigned to fixed units
- it emerges from selection over dynamic populations
- stability is a property of recurrent selectional outcomes, not enduring elements
In this frame:
- variability is not a deviation from function
- it is the medium through which function stabilises
Which makes a crucial difference:
what mainstream neuroscience calls “drift” is not a problem in Edelman’s ontology, because there is no requirement for fixed representational identity to begin with.
Yet instead of adopting this shift, the field largely absorbed population language while retaining representational expectations.
This produces a hybrid structure:
- population-level measurements
- representational-level interpretation
- unit-level intuition
And hybrids of incompatible ontologies do not produce synthesis—they produce recurring paradox.
“Drift” is one such paradox.
IV — The Stability Problem is Misframed
Why the question “how does behaviour remain stable?” assumes its own answer
The central puzzle in drift literature is:
If neural representations change, how does behaviour remain stable?
This question appears deep, but it quietly assumes:
- behaviour is stable in a way that requires neural invariance beneath it
- neural activity is responsible for preserving that stability
But this reverses the dependency.
Behavioural stability is not something the brain “maintains” against variability. It is an emergent regularity of:
- repeated coupling
- constrained environments
- embodied action
- and re-instantiated neural dynamics
Stability is not stored. It is enacted repeatedly under changing internal configurations.
The confusion arises because neuroscience keeps looking for:
a stable object inside the system that guarantees stability outside it
But there is no such object.
V — The Real Crisis: Neuroscience Does Not Know Its Ontology
Why “drift” keeps returning no matter how often it is explained
The persistence of “representational drift” as a concept is not evidence of mystery in the brain. It is evidence of instability in the explanatory framework.
Neuroscience currently oscillates between three incompatible commitments:
- Neurons as stable encoders (classical tuning curve model)
- Populations as dynamic systems (Edelman-compatible structure)
- Representation as explanatory currency (still dominant language)
These cannot all be true at once.
So when data violates (1), the field shifts to (2) but interprets it through (3). When (2) resists (3), the mismatch is labelled (drift, noise, remapping, context dependence).
Nothing ever resolves because the underlying issue is not empirical—it is ontological.
The system is attempting to explain:
dynamic, continuously re-instantiated coordination fieldsusing:a vocabulary built for stable symbol-like mappings
That mismatch is the real drift.
Coda — What Would Have to Change
A coherent neuroscience would not necessarily need a new discovery. It would require a refusal:
- refusal of neurons as privileged carriers of function
- refusal of representation as default explanation
- refusal of stability as hidden substrate rather than emergent regularity
This is not a call for vagueness or mysticism. It is a call for consistency with the very data neuroscience already produces.
The irony is that the field is not lacking evidence.
It is lacking permission to reinterpret its own evidence without preserving its inherited metaphysics.
