One of the big revelations from the last few years worth of focus on glia is shining a light on how we learn.
Astrocytes create non-overlapping local domains, and each local domain is essentially stored information is the core of the learning process. Astrocytic local domains have a maximum amount of discrete information available (capped by a metabolic limit), after which point no other information can be stored in that local group.
The constant effect of pruning away unused information allows astrocytes to continue storing information, but when that pruning no longer keeps up with information input learning can no longer happen. This has the side effect of also blocking all information transfer through that particular local group. Information can still be transferred through the local group as long as it matches (closely enough) the information the local group has stored.
That’s pretty much it, lol.
Let’s take some other concepts under this framework.
“Fire together, wire together” or Hebbian mechanics are currently (I think) the most popular explanation for how engrammatic information is stored. Under this framework, what’s actually happening is that an astrocyte is creating a local group by configuring the synapses of the neurons around it to provide it a differential between the incoming signal and cell body information, which it integrates, processes internally, and passes forward. Post Hebbian, unnecessary/unused information is pruned away to make room for more information.
Under this framework, “plasticity” is a measure of how efficiently local groups can discard information in order to make room for new information. “Plasticity” is reduced when astrocyte local groups can no longer store new information because they’ve reached their metabolic information limit, and can no longer prune information to make more room. As we age, our astrocytes store more discrete information, making non-destructive pruning more difficult and reducing “plasticity”.
On the extreme end of this dementias are the result of overaggressive pruning, which destroys the local group configurations signalling pathways. The more connections a local group has, the more ways it’s information can be accessed and transmitted. Looking at this through AD for instance, the mechanic occurring is that over time these connections are starting to degrade, requiring more and more specific (or energetic) “queries” in order to access the information.
Most education systems are designed around this particular learning concept, where we establish a base concept, then build little bits of linked information on top of it. In configurations like “Asperger’s”, where brains are extra-ordinarily good at creating links but with relatively lower metabolic maximums, it creates a heavy dependence on a consistent base concept to lower the energy necessary to create differentials, but once the chain of differentials is made a large amount of information can be readily accessed off that base due to the high linking capability. This results in a knowledge base which is relatively narrow due to the difficulty (energetic requirements) in establishing new base concepts, but high associativity (differential links) once that base engram set is established.
On the other extreme are “autistic” configurations which have amazing local energy, but relatively low linking energy. These types of phenotypes are along the “schizophrenia” arm of the “autism/schizophrenia” dichotomy. The high local energy provides extremely flexible local processing, but makes it difficult to correctly transmit differentials due to the energy imbalance. Interestingly, all of this would be invisible to tools like MRI and somewhat invisible to EEG due to the localization effect of the energy as these tools are generally only sensitive to (assumed) flows of energy.
Summarizing a bit, each astrocyte stores information within itself and configures cells around it to provide it a differential of external information. It integrates this differential and passes that information onward. Each local group has a metabolic maximum that determines how much information it can store. The metabolic maximum also determines how “strong” the passing signal to other cell groups is. Once a maximum information limit is reached, astrocytes can no longer ingest new information, it can only process differentials of it’s existing information. More space for new information is created by stripping old/unnecessary information or refining the incoming differential. As multiple cell groups work together on a regular basis, they create more and more refined differentials between each other. It is less expensive to access information via this differential than it is to establish new information in a local group.
Looking at constructs like g (which is mostly bullshit as measured), tests are generally focused on testing the general maximum energy of astrocyte local populations. Crystallized IQ is a measure of how efficiently “linking” occurs in brains, while fluid intelligence is how much energy overhead the local group has to process incoming information.
In neurodegenerative conditions, high “fluid” intelligence/or local group processing energy survives a bit more gracefully than high linkers.
The experience of an “Aha” moment, is when a differential between two previously energetically incompatible local groups becomes efficient enough to allow linking together those domains of information each had access to.
What other concepts do I need to conform against this?
Edit: So here’s what’s still unclear, what’s the mechanic of creating the differential? Do neurons work as sort of signal breaks/modifiers to calcium signals? Or is the morphology of the dendrite/axon enough to create it? For example, is all information transferred exclusively over calcium/particle mechanics and “pared down” depending on incoming neuronal connections? Or do the neurons themselves have information directly encoded on the ends which gets interpreted by the local group? Maybe a combination of both, with outbound “shouts” being transmitted over Ca waves, and then modified by neuronal return properties.