Hypothesis

Age‑related cognitive rigidity stems not from synaptic loss but from a shift in the synaptic tenacity set‑point that favors over‑consolidation. Stabilized synapses exhibit slowed protein turnover, elevated silent synapse fractions, and a biochemical bias toward LTP‑like states that occludes LTD. This maladaptive stability predicts that restoring flexibility requires transiently lowering the tenacity threshold—e.g., by inducing small, salient prediction errors—rather than globally boosting plasticity.

Mechanistic Basis

1. Synaptic tenacity increase – With age, cytosolic calcium dysregulation and heightened calcineurin activity dephosphorylate NMDA‑receptor subunits, shifting the phosphorylation balance toward a state that resists both LTP induction and, critically, LTD expression [https://pmc.ncbi.nlm.nih.gov/articles/PMC1693160/][https://pmc.ncbi.nlm.nih.gov/articles/PMC11330810/]. The resulting synapses show reduced turnover of scaffolding proteins (e.g., PSD‑95) and accumulate stable AMPA‑receptor complexes, which raises the energy barrier for synaptic weakening.
2. Silent synapse accumulation – Aged hippocampal circuits retain overall synapse numbers but display a rise in NMDA‑only (silent) synapses that lack functional AMPA receptors [https://pmc.ncbi.nlm.nih.gov/articles/PMC6761599/]. These synapses are poised for potentiation but cannot be depressed because the molecular machinery for LTD (e.g., PP1/PP2A activation, AMPA‑receptor endocytosis) is inaccessible in the over‑consolidated state.
3. LTD ablation and behavioral inflexibility – LTD is essential for reversing learned associations; its loss correlates directly with deficits in spatial reversal tasks [https://www.pnas.org/doi/10.1073/pnas.1404670111]. When LTD is blocked, old memory traces dominate, producing the pattern entrenchment observed in aging.
4. Bias in consolidation during sleep – Age‑related changes in spindle frequency and coupling bias sleep‑dependent consolidation toward strengthening existing ensembles rather than integrating novel inputs [https://pmc.ncbi.nlm.nih.gov/articles/PMC12456079/]. This further entrenches the over‑consolidated map.

Novel Mechanistic Insight

We propose that the brain’s cost‑benefit optimizer interprets the high tenacity state as a sign of environmental predictability, thereby down‑regulating neuromodulatory signals (e.g., acetylcholine, norepinephrine) that normally gate LTD. Introducing controlled uncertainty—brief, low‑stakes prediction errors that violate the brain’s current model—should transiently raise neuromodulator levels, activate calcium‑dependent phosphatases, and lower the tenacity threshold, permitting LTD‑mediated weakening of overly stable synapses. This process would be self‑limiting: once prediction error signals subside, the system re‑stabilizes at a new, more flexible set‑point.

Testable Predictions

• Prediction 1: In aged mice, a single session of mild novelty‑induced prediction error (e.g., unexpected change in a well‑learned maze cue) will increase hippocampal LTD magnitude measured ex vivo within 30 min, as evidenced by enhanced AMPA‑receptor internalization [https://www.pnas.org/doi/10.1073/pnas.2208681119].
• Prediction 2: The LTD rescue will correlate with improved performance on a reversal learning task, but only if the prediction error is neither too weak (no neuromodulatory surge) nor too strong (triggers stress‑mediated LTP).
• Prediction 3: Pharmacological blockade of β‑adrenergic receptors during the prediction‑error window will abolish the LTD enhancement, confirming the role of norepinephrine in lowering synaptic tenacity.
• Prediction 4: Chronic low‑dose administration of a calcineurin inhibitor (e.g., FK506) in aged animals will mimic the effect of prediction error by reducing tenacity and restoring LTD without requiring behavioral novelty.

Falsifiability

If prediction‑error experiences fail to produce a measurable increase in LTD or do not improve reversal learning across multiple intensities and timings, the hypothesis that over‑consolidation is gated by a tenacity set‑point modifiable by uncertainty would be falsified. Conversely, consistent rescue of LTD and flexibility under the specified conditions would support the model and suggest that cognitive aging interventions should focus on re‑introducing calibrated surprise rather than indiscriminate plasticity enhancement.