Thursday, October 23, 2008

Erasing memories

Scientists at the Medical College of Georgia just found a protein that can help erase selected memories in mice. After the protein is injected in the mouse, they have the mouse recall the memory, and then after a consolidation period the memory is gone. Related memories don't seem to be affected, so they conclude that only the targeted memory is erased.

In the cognitive model that I use, this can be explained by postulating that the connections composing the target memory schema's composite cell assembly connections are weakened enough so that they don't reach threshold during normal activity. But the related memories are intact because those connections have a different combination of cell assemblies. So while some are weakened, there is enough strength left that the total schema can still reach threshold during normal activity.

But here is the key question. What long term effects would there be during normal and extreme situations for humans? Weakening of the target memory connections, even if they don't erase related memories, would weaken them. This would make them less likely to activate during periods of divided attention and would slow down the activation even during focused attention.

Also, the erased memories would be easy to recreate with a reminder. Erasing a traumatic memory (the hoped for long term goal of this research team) would be possible, but perhaps would only be temporary.

A full extinction of the memory would only be possible if related memories were also erased or significantly weakened or skewed. And even unrelated memories could be affected if they happen to share some connections with the target memory. This was not tested (and would be really hard to test) in the mice used in this study.

Of course, the point of research is not to solve the problem conclusively but to get us one step closer. More research on this protein and others will help us better understand memory and maybe lead one day to practical interventions and therapies. But if the model I use is correct (or close to it), this path will be a lot harder than the Georgia researchers are hoping for.