Are New Neurons and New Sequences in the Hippocampus Preserved or Abandoned?

As the Hippocampus discovers novel sequences and encodes the associated cortical network are the Hippocampal cells preserved and associated with that memory, or once the memory is encoded (remembered) are they no longer relevant?

In other words does the hippocampus retain associations with the novelties it stores or, once novel sequences are stored, is the new association between the direction, head position, place, space, and location no longer necessary?

I can’t quite tell from the literature if hippocampal rapid-reaction chemistry and cells are disposable so to speak, or whether their association to new memories is retained.

I would guess the former but curious.


1 Like

The cells support the synapses that hold the memory. The cell lives as long as the memory does.

If I understand what is going on in the hippocampus, it holds a mirror of the cortex areas that it is connected to.

During the day the new learning in the hippocampus is the delta between the active recall is in the cortex and the surprise in perception.

The hippocampus is tuned for rapid one-shot learning but this high degree of plasticity has a limited capacity - tunes by evolution to one day.

The rest of the cortex learns more slowly with “normal” Hebbian learning. Obviously- with much higher capacity.

During sleep spindles both areas are interrogated and the cortex is prompted to learn the difference - the new memories.

When both areas respond equally the memories have been transferred.

I see the hippocampus as evolved from the point before sight and hearing- when all the critter had was taste and touch. The chemical signature of the meal (olfaction) was retained until the gut could make the call if it was good or bad. This could be considered the first emotion.

The hippocampus is a vastly enhanced version of this. Sequences are held, along with the emotional values experienced during these episodes. You don’t know if it was good or bad until it is over. What is transferred to cortex is what happened and how you felt about it.


Note that this is a self-repairing system. Since there are two identical copies if one part degrades due to cell death the distributed nature of the representation matched in both systems and a hole in the pattern will be filled in by recruitment of a new memory in that area.


Any biological evidence? sounds like the operating system’s memory?:sweat_smile:
mirror question: How does short term memory work?

These general concepts have been distilled from decades of reading on the topic.

Unfortunately- my notes are poor enough that I can’t point to the source papers where I have read the supporting material.

If nothing else - consider this an organizing trail guide for your own research.

Yes. I understand. However, I can’t find any evidence that those cells preserve information in the hippocampus long term - and between its ability to rapidly mature neurons, and the volume of the area being suspiciously small for the retention of novel stimuli it doesn’t look like it. Once associations are transferred to long term memory, is the ‘temporary’ information (organization of neurons) in the hippocampus disposed of? It looks like it to me, on volume alone.

I guess this should be a challenge for our neuroscientists. Because they must first confirm the unit of memory organization. so they can measure or track the memory.

The hippocampus does not preserve all of the cortex- only the high level representations in the adjacent temporal lobe. The items you mention are all related to the high degree of plasticity. As I suggested above, the new cells “grow into place” to match the contents of the connected temporal lobe. The new memory is layered on top of that.

I am offering a mechanism where a memory in this sheet of cells (hippocampus) can form AND be transferred to a different sheet of cells (cortex). I have not seen anyone else offer a mechanism to do this. If you are aware of one please point it out.

We know from patient HM that he was perfectly capable of perceiving and processing space without hippocampus- but unable to form new memories. My explanation is compatible with these observations.

1 Like

Obviously, there is another possible mechanism: The hippocampus is just a power source, not a memory chip. If you compare the brain to a computer.

I see someone needs to spend some time with neurobiology textbooks.

The brain is not like a computer.

I have taught computer architecture and programming AND studied neuroscience for decades. They are very different in how they get things done.

As far as I know there are two extant possibilities
The hippocampus quickly provisions connections and cells in response to novelty, and those connections and cells persist, and by persistence and rehearsal (repetition) form associations between a novel network and auto-associated network, forming new long term memories (episodes) as well as generalized episodes (concepts).
And then:

  1. The connection, neurons and sequences in the hippocampus remain (unnecessarily) related to the novel association.
  2. The connection, neurons, and sequences in the hippocampus, having served their function, and no longer necessary, deprecate those spines, dendrites, axon terminals and cell bodies, which are rapidly overwritten with the next day’s memories.

I just don’t know the answer. I am not sure anyone else does.

How does the hippocampus know that the activity is novel?

It has to have a baseline to compare it to.

The cortex is building models of the world 24/7 as the perceptions come in - using active recall.

On reflection- there is a novelty detection mechanism in the cortex - bursting. This the the reaction when a mini-column is not able to match a perception.

The busting activity does not signal sequence - that is a local learned property. Bursting signals that the column learned something but we do not know what it learned. This points to the hippocampus learning in parallel with the cortex.

This still has to get transferred back to the cortex. And erased in the process.

The new memory contents of the hippocampus including sequence are available in real time before they are consolidated in the cortex. I don’t understand how this parallel readout might work.

I will have to think on this.

First we need to dig up papers to narrow down the options:

I think your mechanism is how memories are transferred from temporary storage in hippocampus to preeminent storage in cortex. I do not see any need to retain the memory in hippocampus beyond the first night’s sleep. I believe the hippocampal cells invoked are freed to be used again for a new memory.

1 Like

Novel sequences are unmatched in the cortex, but integrated in the EHC, and this discontinuity (i suspect it’s absence of orientation and direction information) is how the Hippocampus identifies novelty. (variation from prediction, and therefore greater attention)

If you were more familiar with the layout of the brain you would know that the HC/EC is at the end (top?) of the hierarchy where it is virtually impossible for sensation to reach the HC/EC complex without being processed by both WHAT and WHERE cortical streams. Any novelty would have been wrung out long before the HC/EC ever gets it.

What the medial temporal lobe gets is the most processed version of what the senses are perceiving. This is shadowed in the HC/EC complex. This includes the spatial and object representation.

If you have been following along with HTM theory you would see that novelty is the driver for learning at all levels of representation in the cortex. Both the temporal and spatial information are processed is some form or another at all levels of the cortex.

I will remind you that patient HM was perfectly able to pay attention, process space in real time and do slow Hebbian learning. Any function that you feel is performed by the hippocampus has to be tested against the losses that HM had with the removal of this area. (mostly the ability to form new episodic memories) If the function is retained then it is not likely to be in the hippocampus. Much about his case is available online.

Note that the losses were temporally graded (over a span of years) which strongly suggests that the HC does not “forget everything overnight” as has been suggested in prior posts on this thread. This is also the reason I propose that the HC/EC complex holds a copy of the contents of the cortex; the complete transfer takes time. We know that it does get transferred because HM can recall older memories.

1 Like

I don’t know of anything. I would be interested in more details.


@rhyolight I have been working up a longish post to address your question and then this dropped in the forum; it does a better job than what I was planning.

1 Like