Synaptic Plasticity
Causal strengthening, depressive pruning, and stability brakes
Plasticity is now framed as a family of teaching phenotypes rather than a single STDP curve. You can compare causal potentiation, anti-causal depression, metaplastic restraint, and saturation risk without pretending one local rule explains an entire memory.
Teaching presets
Start from a learning rule phenotype
Choose from fourteen phenotypes that span near-neutral coincidence, narrow and broad association windows, slow consolidation, reversible depression, homeostatic pullback, precision pruning, and full floor-or-ceiling saturation.
Causal potentiation
A clean pre-before-post window where causal timing strengthens the synapse without immediately saturating it.
Use this as the baseline teaching frame for Hebbian strengthening before discussing addiction, recovery, or maladaptive reinforcement.
This is a timing-rule scaffold, not a literal disease model.
STDP curve
Timing window
Delay sensitivity
Equal delays, opposite order
At 10 ms, the same absolute delay can still favor strengthening or weakening depending on spike order and window asymmetry.
Selected timing
Current vs mirrored pairing
The mirrored ordering would actually dominate at this delay because the LTD side is broader or stronger.
Weight evolution
Synaptic strength across pairings
Realized update
What each pairing actually changed
Because the weight stays away from floor and ceiling, each pairing keeps expressing a real update instead of being clipped away.
Trajectory checkpoints
How fast the rule moves the synapse
Checkpoint bars stay in midrange, so the rule remains teachable rather than collapsing into floor or ceiling.
Phenotype
Stability-biased plasticity
The timing window still permits potentiation, but the overall balance leans toward depression, restraining growth and preserving midrange weights.
Final weight
0.7911
Where the synapse finishes after all pairings.
Total change
+0.2911
Net movement from the starting weight after repeated pairings.
Delta per pair
+0.0049
The current local update rule expressed at the selected timing offset.
Window bias
-0.0100
Positive favors LTP overall; negative favors LTD and stability pressure.
Saturation state
midrange
Whether the synapse is pressed toward ceiling, floor, or still lives in a teachable middle range.
Clinical lens
What this rule teaches
Helpful for explaining why real networks need braking forces that prevent every correlated event from becoming runaway strengthening.
Behavioral readout
What learners should notice
- Weight changes are present but moderated.
- The rule stays teachable without rapidly collapsing to floor or ceiling.
- Depression remains available as a stabilizing counterweight.
Differential traps
What not to overclaim
- A restrained rule is not the same thing as weak learning.
- Stability pressure should not be confused with simple inhibition or lack of coincidence detection.
Examples
Where this plasticity phenotype shows up
These are teaching examples, not claims that one whole disease reduces to one synapse. The point is to give students multiple clinical, rehabilitation, developmental, and learning-science settings where this pattern of strengthening, weakening, or restraint is a useful explanatory frame.
Skill learning with preserved flexibility
A motor pattern improves through repetition, but countervailing rules keep the circuit from locking too early into one rigid solution.
This is the kind of example that makes metaplasticity feel useful instead of abstract: the network learns, but keeps room to adapt.
Post-injury recovery with safeguards
Rehabilitation strengthens helpful pathways, yet stability pressures keep the system from amplifying every noisy coincidence that appears during early recovery.
Students can use this to understand why recovery often needs both potentiation and brakes.
Ocular-dominance style recalibration
One stream begins to dominate, but slower counter-rules keep the overall system from permanently saturating after short-lived imbalance.
The point is not exact visual-cortex fidelity, but the principle that stabilizing rules preserve teachable midrange weights.
Sleep-supported renormalization
Wakeful learning leaves some pathways biased toward strengthening, but broader stabilizing processes keep the network from carrying every daytime gain forward at full amplitude.
This helps students connect metaplastic restraint to the idea that useful brains must both learn and re-normalize.
Adaptation without lock-in
A patient improves with prism or gait adaptation, yet the system retains enough counterpressure that the learned shift can still be revised when the environment changes again.
It is a good example of why stable midrange weights are often what make flexible relearning possible.
Classic anchors
Landmark examples to teach alongside the rule
Bi and Poo spike-timing experiments
The same pair of neurons could potentiate or depress depending on spike order and timing lag.
That experiment is the cleanest entry point for showing why anti-causal timing deserves its own teaching weight.
Cerebellar LTD literature
Climbing-fiber and parallel-fiber timing rules in cerebellar models highlighted selective weakening as an error-correction mechanism.
This gives students a concrete example where depression is instructive rather than simply destructive.
Turrigiano homeostatic scaling
Neurons can globally tune synaptic strength upward or downward to keep firing in a workable range.
It anchors the idea that real plastic networks need stability rules alongside Hebbian change.
Plain-language rule
The shortest phrasing worth remembering
What happened
Biological basis
- NMDA receptors behave like coincidence detectors because they need presynaptic glutamate and postsynaptic depolarization together.
- Pre-before-post timing can trigger calcium conditions that favor CaMKII signaling and AMPA receptor insertion.
- Post-before-pre timing changes calcium dynamics enough to favor phosphatases and synaptic weakening instead.
- Making LTD slightly stronger than LTP helps stabilize the network and prevents runaway excitation.
Next questions
Useful follow-up experiments
- How much negative bias is needed before causal timing stops producing net gain?
- What shifts first if tauMinus or aMinus is reduced slightly?
Continue the loop
Use this with excitability, dopamine, and tutoring
Neuron Simulation
Advanced cellular physiology
Dopamine Prediction Error Lab
Computational clinical neuroscience
Neuro Tutor
Cross-module consult reasoning with explicit scoring