A funny thing about memory is that it bleeds. It takes on schematic properties that allows other concepts to associate with it, notoriously as free association, but also during learning, trauma, metaphor, and creativity. PTSD is a problem where too many things trigger bad memories and those bad memories generalize to mean far more than the bare incidents in question. If the memory of one distasteful event is to be useful, its core elements need to be recognized and generalized to enable avoidance not just in that exact setting again, but of future patterns that are similar to some regulated degree. So how similar is similar?
Surprisingly, memory generalization can be studied, and a recent paper described connections between the hippocampus, frontal cortex, and an intermediary structure called the nucleus reuniens, showing that some of these connections affect the degree of memory generalization when either destroyed or enhanced. The nucleus reuniens is part of the thalamus which is smack in the middle of the brain, with important roles in consciousness and alertness, among much else.
This is all done with mice, using cutting edge techniques. The test was to have the mice experience a shock in combination with an alert sound in one cage, and then to test response to the same sound (tone condition) or to no sound in various settings, like the same cage, or similar cages, or a somewhere completely different. Normal mice do not show fear outside the original cage context, unless you ring the same tone that was associated, Pavlov-style, with the original fear training. But mice that have had their medial prefrontal cortex areas ablated (mPFC), or their nucleus reuniens ablated (N. reuniens), show continued fear in new cages. I could not tell how universal this fear expression was ... were the mice just totally frozen now, fearful of all conditions? That was not clear.
At any rate, the coup de grace was for the researchers to use some clever molecular technology to ablate only the neurons connecting the PFC with the N. reuniens, by injecting the mice with one genetic element of a lethal cocktail (a gene for tetrodotoxin hooked up backwards to a promoter of expression) in the mPFC, and the other element (a recombinase that flips the toxin cassette into the expressing orientation). Each was labeled with a fluorescent protein to show where the injections took place:
Neurons transport this kind of expressed protein through their cell bodies, even out to the farthest axons. They also let some into the synaptic vesicles that send neurotransmitters across the gap between neighboring neurons (axons and dedrites). This means that a small amount of the recombinase would be transported from the injected region into the target area where its projections go, and be able to induce expression of the toxin in those target neurons that received the complementary injection, killing them. The figure shows the result, where control mice with non-expressing injections show normal fear-context dependence, while mice that got the full treatment in the nucleus reuniens and prefrontal cortex show more generalized fear, like mice that had their whole nucleus reuniens ablated.
They even were able to do a converse experiment, using the recombinase to induce, instead of a toxin, a repressor of neuroligin 2 which represses neuronal transmission, essentially enhancing transmission between the two injected areas (the blue bars, above). These show the (modest) opposite response of lower fear response in the altered context (than control mice with sham treatment), with similar response in both the original cage and the trained tone. This provides quite a strong argument for the specificity of what they are seeing- that the nucleus reuniens is a critical way-point for signals from the prefrontal cortex that tell the hippocampus that a memory is relevant to more specific conditions than it might be inclined to apply them to otherwise. As in most things, the prefrontal cortex refines and inhibits our deeper brain processes.
Lastly, they used the very latest technology to convert the nucleus reuniens neurons to be light-inducible in their firing (optogentics), by injecting a gene for channel rhodopsin, which converts light into membrane potential which in the neuron can induce action potentials. This method allows researchers to run the affected cells at any firing rate they wish, up to 50 times per second, driven by a light guide directed right into the structure in the brain they are interested in.
|Optogenetic experiment, where light is conducted into the site of an injected channel/rhodopsin gene which allows researchers to use strobed light to fire neurons at will, and also fluoresces red.|
They used two driving methods, one a constant 4Hz (4 firings per second; tonic) and the other bunched in fifteen firings within a half-second, at five second intervals (phasic). The experimenters evidently came up with these patterns through trial and error, and they had opposite effects, when run while training the mouse to the fear condition (shocks to the feet in a cage). Running the tonic pattern during training yielded lower fear later in altered contexts (lower generalization), while the phasic pattern yielded higher fear. Neither one affected the response of the mice to the true fear condition of the original training context and/or tone.
|Optic driving of the N. reuniens during memory training yields opposite effects on ultimate memory generalization, depending on the pattern of neuronal firing driven- phasic or tonic.|
The idea from this work is that while the hippocampus stores specific memories, other connected areas allow those memories to be generalized to *similar situations, accounting for allegory and much else about our mental operations. There is a two-way circuit from the hippocampus to the nucleus reuniens and on to the medial prefrontal cortex, where memories exist in more abstract form. Here, the researchers show that for at least one type of memory, the obscure nucleus reuniens links the two, with an active role in whether a mouse's memories flood in during distantly related situations, or only in the more restricted context of the original experience.
Related mechanisms are likely to be relevant to the binding problem of how separate features of an experience or scene are linked by our minds into a unitary experience with various possible abstractions and composite or derived properties, i.e. consciousness. The way we associate new experiences with prior memories and knowledge, either freely or sparingly, should work similarly and be intimately connected with the degree and quality of our creativity.