Dreams may hold key to memory transfer

By VIRGINIA WINDER
IT’S 5am and I’m accessing a massive database including fresh information downloaded overnight.
For the past week, I have been transferring “need-to-know” stuff from temporary files into this database.
The transfer always happens in the deep, dark night.
For several hours yesterday, those temporary files were crammed with scientific research on the brain’s ability to learn, retain and retrieve information.
With ease, all that data is coming back to me now.
The interviews, the articles in magazines, online and in books, are all there waiting to be accessed.
All because I slept on it.
I am, of course talking about memory.
Mine.
And yours – although what you choose to download won’t be the same as mine. Even if we watched the same movie or listened to the same piece of music, we would consciously and unconsciously store different information.
That’s why court witnesses almost always have different versions of the same event, even though they swear (on the Bible) that they are being honest and accurate.
While we may glean different information, we all go through the same memory transferral process.
This involves moving information from our temporary files, a kind of short-term memory stored in our hippocampus, to our massive neuro-database, called long-term memory found in our cortex.
To do this job, we need to sleep.
More specifically, we need to dream, says Otago University’s Anthony Robins (left).
Just exactly why we dream has been a puzzle to scientists, philosophers and great thinkers for hundreds of years.
Sigmund Freud believed dreams were our subconscious way of expressing fears and the suppressed sexual urges we prefer not to face or act on in waking life.
Robins has another theory, one he’s discovered because of memory loss.
The associate professor of computer science is a leading figure in the university’s Memory: Mechanism, Processes and Applications research team, made up of 30 academic staff and about 150 students.
This in-depth study, which encompasses many of the university’s departments, has been running for a decade now.
But the more Robins continues his research using artificial neural networks (ANN), built to simulate how our brains work, the more he believes he’s discovered something startling.
It’s this: When the artificial network is fed new data it deletes the old. This is called “catastrophic forgetting”.
However, when a “rehearsal” of that existing information is played through the network while new data is added, none of the old is lost.
But Robins realised that rehearsing or playing every piece of information stored in the network was impractical and meant having to have a second memory system.
So he tried another tack. “I have proposed a mechanism, pseudorehearsal, which is similar to rehearsal but does not require the storage and access of old information.”
To explain further, Robins uses a musical metaphor.
“The ANN (network) is like an orchestra that can learn lots of pieces – say 50 all at once. But if the orchestra learns another piece tomorrow, it forgets the first 50 pieces.”
But if the orchestra “jams” or “makes stuff up” in a pseudorehearsal of the old, it can retain music already learnt and add the fresh piece with no loss of information.
He surmises that brains have evolved to overcome the problem faced by the artificial network. This is where dreams come into play, because they are our own “pseudorehearsal” of stored information or our own jamming sessions. This explains why they are a bizarre mixture of fact and fiction.
“In order to learn new things, the brain has to have the time to jam and to wander randomly over bits and pieces of old information or made-up fantasy stuff,” Robins says.
Another piece of information Robins has learnt from his research is that recollections are not set in stone.
“Memory is not based on any one fixed structure – it has to keep re-encoding itself,” he says.
Every time we access a memory, we are re-recording it. “A memory is a piece played live by an orchestra.”
That’s opposed to being like a digital or tape recording, which is the same every time it’s played. Think of a single you hear on the radio – often you see it performed live and it’s annoyingly or pleasantly different to the recorded version. The live version may be sung by the same person, but they often change a word here and there, add an extra emphasis on a note, or simply forget the lyrics and improvise. Also, the band may be different, because there’s a new lead guitarist or drummer.
“Our memories are houses of sand, not cement,” Robins says.