Thursday, March 22, 2007

Grandmother cells

I was watching UCSD-TV last night, a presentation from last year by Christof Koch of Cal Tech, with his pink hair, matching tie, German accent, and occasional inability to find the obvious word in English. It didn’t remind me at all of Mike Myers playing “Dieter”, no. It was interesting to watch him demonstrate a phenomenon I’ve seen before where someone was fluent in scientific jargon in English, but got stuck on describing ordinary things in English. Science is indeed its own culture, a global one at that.

Koch described experiments involving the placement of depth electrodes into the temporal lobes of epilepsy patients at UCLA, in order to plan surgery to best cut out their seizures. Researchers used the opportunity to record from single cells in the amygdala or hippocampus. They found cells that responded to many different pictures of particular famous people, such as Bill Clinton, Halle Berry, or Jennifer Anniston, or famous places, such as the Sydney Opera House. A cell they found that was excited by Bill Clinton didn’t care if Hillary was in the picture or not. Another cell that fired away with a picture of Jennifer Anniston was silent when Brad Pitt was in the picture with Jennifer (This was from a couple of years ago). A cell that fired away at various pictures of Halle Berry, including one as Catwoman, was silent to another woman as Catwoman.

I discovered that John Horgan, a much better writer than I am, wrote about this at the time, so I’ll leave the rest to him, except for my memory of how this wasn’t supposed to be the way the brain works. When I was in neuroscience, the smart guys said it would be too limiting to have a single cell stand for something as specific as one’s grandmother. It might sit idle for years. How would it be able to recognize grandmother in every context in which she might appear or as she grew older? They thought it more likely that the brain extracts features from a scene and makes judgments about those features in a broad, cooperative effort, one’s grandmother being recognized by many associations to specific features, using cells that could recognize many people that way.

Horace Barlow (Barlow, H.B. 1972 "Single Units and Sensation: A Neuron Doctrine for Perceptual Psychology?", Perception 1, 371-394.) wrote the first publication I know that tagged this problem as being a matter of grandmother cells. His idea was that one would have to show not only that a cell is selectively activated by one person’s image, but that stimulating that cell brings that person’s image to consciousness. Researchers have yet to report on that second part.

Then there’s the question of how are all these cells organized in the temporal lobe. What are they mapping? The amygdala is about signaling us that there is an emotionally significant stimulus around, something to eat, someone to have sex with, or something to run away from. Does it really contain cells corresponding to everyone we know or are these results about archetypes of men and women we’d have sex with, spend time with, or want to be like? Or places we’d like to visit like the Sydney Opera House? How long a wish list like this might we have in our temporal lobes and how specific does it get?

Neuroscience will remain limited by how clever we can be at challenging the brain with a stimulus and then recording a reliable response. People keep pushing ahead with both of those, though. I expect that the most predictable outcome of this is that there will be something to surprise anyone.

No comments: