10 KiB
Table of Contents
include_toc
| include_toc |
|---|
| true |
Complete set of general guidelines
DAY — Behavior
Behavior is determined by structure modulated by NIGHT. Every DAY behavior operates within the ceiling set by the previous NIGHT's structural commit. The neuron does not decide how sensitive its spines are, how large its active zones are, or how tightly its astrosynapse wraps the cleft — it simply operates within whatever architecture NIGHT left behind. Structure is the memory of past validated experience, and it silently shapes every moment of current processing.
Behavior has an energy cost. Every behavior draws from the local budget. No behavior is free. The budget has a hard ceiling set by the vascular glucose supply through the astrocyte, and by the soma's own mitochondrial capacity. When the budget is exhausted, behavior is suppressed regardless of the input — the bouton goes silent, the spine cannot reset, the branch cannot propagate the bAP. Energy scarcity is not a failure mode — it is a regulatory mechanism that prevents runaway activity from destabilizing the system.
Behavior has a material cost. Fast behaviors consume biological materials — D-serine precursors, vesicle membrane proteins, actin monomers for transient spine changes. These are distinct from energy: a behavior can be energetically affordable but materially limited. D-serine release depletes the astrocyte's serine pool. Sustained vesicle release depletes the readily-releasable pool faster than it can be refilled. Material costs impose a second independent constraint on behavior that energy alone does not capture.
Behavior leaves traces. Every behavior deposits a graded, decaying record of its occurrence in the local fast trace variable. The trace is not a passive record — it actively biases the next behavior in the same context. Residual calcium in the bouton makes the next release more probable. Calcium in the spine encodes the instruction for future structural change. The trace is the system's short-term memory of what just happened, and it decays automatically so that only sustained or repeated behaviors accumulate enough trace to reach the tagging threshold.
Behavior happens in a specialized fashion determined by structure modulated by NIGHT. The form of the behavior — not just its ceiling but its quality — is shaped by structure. A bouton with high pre_structure has tightly clustered calcium channels beneath its docking slots, making each AP more reliably coupled to vesicle fusion. A spine with high post_structure has more anchoring slots, making each glutamate pulse more faithfully converted into AMPA current. The astrosynapse with high astro_structure delivers D-serine tonically, keeping the NMDA gate chronically primed. Structure does not just set the maximum — it shapes the transfer function between input and output at every moment.
Behavior happens in a time window determined by accumulation and elimination of traces. The eligibility window for tagging is not externally imposed — it emerges from the decay dynamics of the fast trace. A bouton is taggable only while its residual calcium is still elevated. A spine is in its Hebbian anticipation window only while its NMDA calcium is still above the tagging threshold. The window opens when the fast trace crosses the eligibility threshold and closes when it decays below it. Dopamine or the bAP must arrive within this window or find nothing to validate. Timing is enforced by chemistry, not by a clock.
Behavior happens in a context determined by local and global circumstances. Local context is the current state of the compartment itself — whether an AP just arrived, whether the membrane is depolarized, whether the RRP has content. Global context is the state of the organism — whether dopamine is elevated, whether norepinephrine is driving arousal, whether acetylcholine is signaling attention. Neither alone determines behavior. The local context determines what the compartment can do. The global context determines whether what it does gets validated. Both must align for a behavior to leave a lasting trace.
NIGHT — Structural Rewriting
Based on traces from DAY there is competition to restructure. The NIGHT scope is not passive recovery — it is an active competition for finite structural resources. Tags accumulated during DAY represent the claims that different synapses are making on the shared material and energy pools. Tagged synapses draw first, in proportion to their tag magnitude. The strongest tags — those that accumulated the most coincident local activity and global validation — draw the most resources and achieve the largest structural expansion. Weaker tags draw proportionally less. The competition is not adjudicated by any central mechanism — it emerges automatically from the shared pool dynamics.
What is not potentiated decays for lack of resources. After potentiation has drawn its share, what remains is distributed as maintenance to all synapses. Structural maintenance requires a continuous material and energy allocation just to resist the baseline decay of molecular components. When the maintenance allocation falls below the decay rate — because potentiation has consumed too much — untagged synapses drift toward lower structural states. They are not told to weaken. They simply do not receive enough to stay where they are. Depotentiation is the shadow of potentiation, enforced by conservation of resources rather than by an active depression signal.
Additional General Points
The system has one primary drive and one emergent consequence. The entire machinery — fast traces, possible tagging, tag stabilization, NIGHT commits — is oriented toward potentiation. There is no symmetric machinery oriented toward depression. Depression is what happens to everything that the potentiation machinery did not select. This asymmetry is not accidental — it reflects the organism's fundamental orientation toward learning from significant experience, with forgetting as the natural cost of that selectivity.
Validation requires signals from progressively larger spatial scales. A behavior at the synapse leaves a local fast trace. The fast trace creates local eligibility. Eligibility alone is not sufficient — the bAP from the soma must confirm coincidence at the spine level. The soma's own firing must coincide with nuclear calcium for the gene expression mandate. And the organism's dopamine broadcast must coincide with local eligibility at every level for the tag to stabilize. Each spatial scale adds a confirmation that the previous scale cannot provide for itself. The synapse cannot know whether the soma fired. The soma cannot know whether the organism's outcome was rewarding. Both are required. Neither is redundant.
The system integrates across time through a hierarchy of decay timescales. The fast trace decays in milliseconds to seconds. Possible tagging decays in seconds to minutes. The tag decays in hours. The structural variable decays over days to weeks unless maintained. Each timescale feeds the next: a single spike leaves a trace that can accumulate into possible tagging if spikes keep arriving, which can accumulate into a stable tag if dopamine validates the pattern, which can accumulate into structural change if NIGHT commits are fulfilled. The system does not decide at any single moment what to remember — it integrates continuously across all timescales simultaneously.
The system is conservative — resources are redistributed, not created. Every structural gain at one synapse is paid for by material and energy drawn from pools that all synapses share. LTD at one synapse returns material to those pools, partially funding LTP at another. The total structural capacity of the system is bounded above by the vascular glucose supply and the soma's protein synthesis ceiling. No activity can increase those ceilings — only sleep-driven CREB expression and vascular adaptation can. Within those ceilings, the system continuously redistributes its fixed resources toward whatever the organism has most recently found significant.
Structure is both the memory and the prior. The structural variables written in each NIGHT become the starting conditions for the next DAY. A potentiated synapse enters the next DAY with a larger active zone, more receptors, tighter astrosynaptic wrapping — making it more likely to respond strongly to the same pattern and more likely to reach the tagging threshold again. A depotentiated synapse enters the next DAY with reduced capacity, making it less likely to respond and less likely to be tagged. Structure is not just a record of past experience — it is a prediction about what patterns are likely to matter in the future, continuously updated by what the organism actually encounters. This is the system's implementation of a prior: a bias toward re-experiencing what has proven significant, encoded in the physical architecture of the synapse itself.
The astrosynapse is the gain control of the entire system. Unlike pre and post, whose structural variables set a ceiling on occupancy, astro_structure reshapes the input itself — controlling how much glutamate reaches effective concentration and whether the NMDA gate is chronically open or chronically closed. This makes the astrosynapse the only component whose structural change directly modifies the operating point of the other two components rather than just their range. And because its structural variable is self-reinforcing in both directions, it amplifies whatever trajectory the synapse is already on — making potentiated synapses progressively easier to potentiate and depressed synapses progressively harder to rescue. The astrosynapse is not just a third component of the synapse — it is the component that determines whether the synapse as a whole is in a learning-permissive or learning-resistant state.
The organism's rest period is the execution window for structural memory. The DAY scope fills the system with evidence — traces, tags, eligibility records — but commits nothing permanently. The NIGHT scope executes on that evidence — writing structure, replenishing budgets, clearing traces. A day without sleep would leave the tags decaying without being committed and the structural decay proceeding without being compensated. The system is architecturally dependent on the alternation between DAY and NIGHT: neither scope alone can accomplish what both together achieve. DAY without NIGHT produces learning that cannot consolidate. NIGHT without DAY produces structural replenishment without new content to consolidate.