I probably still need to more clearly illustrate the “dorsal” vs. “ventral” streams first, however I’ve been looking at these function groups through the lens of the more/less/modulate framework, and going circuit by circuit it appears the function of these two streams is segregated along this split.
My expectation was that dorsal streams could generate their own “salience” and ventral streams could generate their own “valence”, but I can’t find find any evidence that’s true. Instead, we consistently see these two streams running parallel through integration points like the hippocampus and putamen. This is a pretty dramatic simplification of brain function and fits almost too perfectly within the MLM model.
If we assume that valence (More/Less) is generated primarily on the feed forward systems stimuli, this provides a mechanic for quick adaptation to a completely asynchronous environment. By removing salience calculations, it allows the allocentric view of the world to be equally weighted among all of the objects being tracked.
I’ve posited before that in order to create a state, we need both a valence and salience signal, so in order for this to work, there must be a default valence and salience being calculated for each path, and we alert to novelty when valence or salience diverge from our default prediction.
I’m beginning to think the hypothalamus might be best described as the storing the default state information for the valence side of the equation. This is consistent with the understanding of the hypothalamus being the homeostatic regulation center, it essentially is always manipulating this more/less signal to maintain balance.
Now the question is, does a default state generator for salience exist, or is it always calculated? If it does exist, where is it?
Edit: Pushing this concept a bit further, do we have three independent salience channels, with circuit analogs to a hippocampal CA region? D1/D2&3/D4 would map to CA1-3 pretty neatly as a salience signal. This would give us serotonin, acetylcholine, and ? (oxytocin maybe) as global valence signals. Valence is then channelized among lower level functional groups. This gives us a pretty good description of how the hippocampal region is able to modify the specific content of sensory information, by splitting context specific items into a series of more/less modulations until desired target is reached.
So continuing from here, that makes the nucleus accumbens, habenula complex, and striatum as the top level valence calculation centers. Indeed most “treatments” seem very targeted toward manipulating one side of the “external/internal” calculation in these centers. An interesting bit of consilience is that most optogenetic manipulations of “choice” in animals come down to manipulating the output of these areas.
Lets assume for the sake of argument that dopamine and acetylcholine are a common valence/salience pair and integrate in the striatum, dopamine integrates with serotonin in the habenula, and dopamine integrates with (?) in the NAcc.
Another interesting convolution here is that some valence centers use two particular receptor groups (e.g. D1 and D2) to modulate each side of the equation. Is this consistent throughout each of the major valence centers? This would appear that dopamine is establishing valence, however more appropriately what’s happening is a binary choice between two different streams.
Do brains determine more/less based on salience differentials?
If this is so, this reduces valence itself down to a salience calculation. A “how much do I want this vs. that” mechanic, and this returns us back to the brainstem for pretty much everything. Okay, so at root, this gives us Raphe, Locus Coeruleus, and Pedunculopontine as our core association centers in the brain stem.
So the brains use GABA/Glu balance to “remember” state between iterations of the more/less evaluation.
Okay, at the very least we’ve come to a framework which is both testable and negateable. I need to keep my eyes out for anything along this path. This might be why brains require both inputs to create a state, what Ive been describing as “novelty” is the feeling of this more/less calculation occurring. The salience channel sets the target, or what the brain is trying to reach, the valence channel is comparing whether it is getting closer or further to this target, and GABA/Glu store current iteration state information between iterations. Other transmitters are more specific valence calculation networks, but these may be the big ones.
So this makes the Red Nucleus the equivalent of the hypothalamus, and dentate nucleus the hippocampal equivalent. Does this make the olives the master control center, or is it one of the pontine nuclei? Interestingly, we do go through an inversion here, with ventral dentate circuits establishing against dorsal cerebral circuits. Hah, exactly like the hippocampus actually! Hrm… Could also be the interposed nuclei?
So is salience simply a target signal? And how does the brain generate these discrete target signals?
So engrams are a collection of “valence/salience” combinations. Oh crap. So brains don’t know what “red” is after all.
The way this must work is that brains create engrams which are valence/salience combinations that represent an object. All data in the brain is comprised of these combinations which represent specific objects. “Memories” are bundles of these units across various domains which get reconstructed into a state in the hippocampus and dentate gyrus[nucleus]. What we are conscious of are the hippocampal ones, as they hijack our sensory system in order to create the differentials.
So this must mean that salience is a little bit more specific than just want/don’t want, it has to encode a target valence as well. Lol, again which is why state requires both a salience and valence. This is starting to make a lot more sense.
I need to look at the dentate gyrus a bit more. It’s pretty clear that it works as a sort of master comparator, but I’m wondering how it mechanically determines sub valence/sub salience.