A Passing Sunshine

Alarming, fascinating news concerning the persistence of memory: it doesn’t. Or rather, the mechanism appears to be different than we collectively assumed. This Ars Technica summary of the Science magazine article explains:

Although transient memories are largely the product of chemical processes in the neurons of the brain, it has been thought that the long-term consolidation of memories involves permanent changes.

[…]

Normally, the rats remember the taste and avoid anything with saccharine in it. Injections of a PKMζ inhibitor called ZIP into the brain’s cortex a few days after, while memory consolidation might still be in progress, blocked the formation of long-term memories.

[…]

But then the authors performed an experiment that I can only assume was expected to act as a control: they injected ZIP at one week and at 25 days. The surprising result was that these later injections worked just as well as the earlier one had.

[…]

It’s hard to tell what the limits of ZIP-mediated memory erasure are, given that the test subjects are rats. There were some things it did not affect, though. Rats are normally hesitant about unfamiliar tastes, a hesitancy that goes away with familiarity. Hitting the cortex with ZIP was unable to eliminate this sense of familiarity.

So in rats, within the few-weeks time period, there doesn’t appear to be a permanent storage of memory. Is it possible that the “formation of memory” process is effectively also the form in which memory is stored?

To use a computer science analogy (always risky when describing neurobiology), there are two ways to represent a stored piece of information such as an image: a semi-permanent representation of the image, or an algorithm that is capable of generating the image upon request. The latter approach, called Procedural Generation, is a hot topic in computer science at the moment, as it promises to reduce the cost of creating media content significantly.

So, to extend this idea beyond media content, what algorithmic approach could possibly be abstract enough to generate any human memory? In her book Passage, Connie Willis makes a strong (albeit laborious) case that metaphor is the fundamental building block of human memory – and, perhaps, human experience. Memory of events, then, is repeatedly telling oneself the narrative of one’s own past experiences – cease to tell yourself the story and you begin to forget. Memory of images is re-describing the visual experience, showing it again and again to your own inner eye…

All of this is a conceptual leap – all we know for certain about ZIP is that it erases memory of tastes in rats. Oh, and that it leaves behind a haunting sense of familiarity, as all forgetting does (“I knew the answer to that!”). This study shows just how far we have to go before we truly understand the functioning of the mind, and the dangers of assuming that our existing computational models (such as permanent data storage) apply to it.