ScienceDaily has an item (here) on memory retrieval by K. K. Taylor and others in Current Biology, Reactivation of Neural Ensembles during the Retrieval of Recent and Remote Memory.
It has been assumed that the hippocampus stores an event by storing a key to the group of cortical neurons that are active during the event. Retrieving a memory is then the hippocampus reactivating that particular set of neurons using its key. But, this was difficult to demonstrate experimentally. This research has a method for showing this picture of memory.
Tayler used a genetically modified mouse that carries a gene for a modified green fluorescent protein. When nerve cells in the mouse are activated, they produce a long-lived green fluorescence that persists for weeks, as well as a short-lived red fluorescence that decays in a few hours. However, the whole system can be suppressed by dosing the mouse with the antibiotic doxycycline, so Tayler and Wiltgen could manipulate the point at which they started tagging activated cells.
Using this system, they were able to track the formation of memories and their retrieval. In indeed appears to confirm the theory.
About 40 percent of the cells in the hippocampus that were tagged during initial memory formation were reactivated (during retrieval), Wiltgen said. There was also reactivation of cells in parts of the brain cortex associated with place learning and in the amygdala, which is important for emotional memory.
Here is their abstract:
Background:
Episodic memories are encoded within hippocampal and neocortical circuits. Retrieving these memories is assumed to involve reactivation of neural ensembles that were established during learning. Although it has been possible to follow the activity of individual neurons shortly after learning, it has not been possible to examine their activity weeks later during retrieval. We addressed this issue by using a stable form of GFP (H2B-GFP) to permanently tag neurons that are active during contextual fear conditioning.
H2B-GFP expression in transgenic mice was increased by learning and could be regulated by doxycycline (DOX). Using this system, we found a large network of neurons in the hippocampus, amygdala, and neocortex that were active during context fear conditioning and subsequent memory retrieval 2 days later. Reactivation was contingent on memory retrieval and was not observed when animals were trained and tested in different environments. When memory was retrieved several weeks after learning, reactivation was altered in the hippocampus and amygdala but remained unchanged in the cortex.
Retrieving a recently formed context fear memory reactivates neurons in the hippocampus, amygdala, and cortex. Several weeks after learning, the degree of reactivation is altered in hippocampal and amygdala networks but remains stable in the cortex.