Long-term memory is thought to exist as structures within synapses, which is why the destruction of synapses in the earlier stages of neurodegenerative conditions such as Alzheimer's disease causes memory loss. If memory doesn't exist in the synapses, however, then there is more of a possibility of restoration through effective treatments for the condition. At this point models of long-term memory that put the data somewhere other than the synaptic connections between neurons has an uphill road to travel: past years have seen the accumulation of good evidence for synaptic memory storage, such as experiments in which memory in rats is erased and restored. Nonetheless, this is interesting work:
The new study provides evidence contradicting the idea that long-term memory is stored at synapses. "Long-term memory is not stored at the synapse. That's a radical idea, but that's where the evidence leads. The nervous system appears to be able to regenerate lost synaptic connections. If you can restore the synaptic connections, the memory will come back. It won't be easy, but I believe it's possible."
"If you train an animal on a task, inhibit its ability to produce proteins immediately after training, and then test it 24 hours later, the animal doesn't remember the training. However, if you train an animal, wait 24 hours, and then inject a protein synthesis inhibitor in its brain, the animal shows perfectly good memory 24 hours later. In other words, once memories are formed, if you temporarily disrupt protein synthesis, it doesn't affect long-term memory."
The scientists added serotonin to a Petri dish containing a sensory neuron and motor neuron, waited 24 hours, and then added another brief pulse of serotonin - which served to remind the neurons of the original training - and immediately afterward add the protein synthesis inhibitor. This time, they found that synaptic growth and memory were erased. When they re-counted the synapses, they found that the number had reset to the number before the training. This suggests that the "reminder" pulse of serotonin triggered a new round of memory consolidation, and that inhibiting protein synthesis during this "reconsolidation" erased the memory in the neurons.
If the prevailing wisdom were true - that memories are stored in the synapses - the researchers should have found that the lost synapses were the same ones that had grown in response to the serotonin. But that's not what happened: Instead, they found that some of the new synapses were still present and some were gone, and that some of the original ones were gone, too. There was no obvious pattern to which synapses stayed and which disappeared, which implied that memory is not stored in synapses.