This is fast turning into a blog about movies and skiing, which shouldn’t be the case. But when your life is centered around neuroscience, movies and skiing, and the neuroscience is still unresolved, movies must take up some space. However, I recently had the pleasure of meeting Prof. Christoph Koch [go explore his website here, it has lots of cool stuff] when he visited Dartmouth about a week ago. He sat in on the graduate class I am taking on Consciousness, and it was a pleasure to be able to discuss some of Koch’s own papers on the subject, with the man himself. Koch also gave a talk on his work on single- and multi-unit deep brain recordings in relation to visual representations such as faces and landmarks. I had the chance to read a paper presenting some of these findings, and I must say that the findings, although not necessary surprising, are still extremely cool.
Basically, Koch and his co-authors showed their participants (who had been implanted with depth electrodes to localize the focus of seizure onset) pictures of a number of famous faces, landmark buildings, as well as animals and objects. What they found was that there were units that responded quite selectively to the faces of particular celebrities, such as Jennifer Aniston. What the picture above is showing, is that this particular unit is responding only to pictures of Jennifer Aniston, and not to pictures of Aniston with Brad Pitt, or to other blond actresses, or other famous people, and so on [if you have trouble seeing the picture, try clicking it - the red bars are showing neuronal responses]. Furthermore, Koch and his co-authors found units that also responded to the name of the preferred celebrity presented as a letter string, suggesting that this is related to the concept, more than some simple features of that person. The authors found similar units that responded to specific landmark buildings, animals and objects. Perhaps the most amazing finding of them all, in my opinion, was that neurons that responded, for example, to the Sydney Opera House, also responded when the subjects were in fact not seeing the Sydney Opera House, but another landmark building that the subjects thought was the Sydney Opera House (in this case, the Baha’i Temple in India).
Based on these results, Koch and his colleagues are trying to figure out a model for how this recognition might take place. What they are suggesting is in fact not grandmother cells but a sparse representation where each concept might be represented by a few hundred neurons, and where there might be representational overlap between neurons, so that one neuron might be a part of the representation of several seemingly disparate concepts (they actually did find units that became active both for Jennifer Aniston and the Sydney Opera House). This research is apparently pretty difficult to undertake (it takes a dedicated surgeon, as Koch remarked), but I find it immensely exciting. Just imagine, each of my friends have a specific neural population that will fire when I look up pictures of them on Facebook…
[P.S. The authors did apparently find cells that were responding specifically to Mr. T - pretty bad ass cell!]
EDIT:
Reference: Quiroga, R. Q., L. Reddy, G. Kreiman, C. Koch & I. Fried (2005): Invariant Visual Representation by Single Neurons in the Human Brain. In Nature, vol. 435 (7045) (pp. 1102-1107). More recent articles on the same subject can be found on Pubmed.