|
|
|
23 July 2005 |
|
|
A little while ago the New Scientist reported research which apparently revived the discredited theory of the 'grandmother neuron'. The piece attributed this theory to Jerry Lettvin, and unfortunately this idea was picked up and widely echoed elsewhere. In fact, although I think Jerry Lettvin probably is the author of the phrase 'grandmother neuron' (it is also sometimes attributed to Horace Barlow), I believe he devised it as a rhetorical aid to his rejection of the idea. Alas, I dare say the meme that he endorsed the idea is now unstoppable. The grandmother neuron is actually one of those ideas that no-one really wants to advance, but which many people would like the opportunity of refuting. I suppose this makes the new research especially interesting, at least in principle. What is a grandmother neuron? The idea is about the encoding of memories in the brain, and the idea is that one memory, or the memory of one thing, is represented by one neuron. There is, for example, a neuron that stands for your grandmother: that neuron is the only one which invariably fires when you think of your grandmother, and never fires when you think of anything else. Other neurons fire in a similar way for your grandfather, your cousin, your house, an oak tree, and so on. There is a neatness about the idea which appeals, and some suggestive empirical support - we've known for some time that in experimental animals there are neurons that fire only in response to certain movements in the visual field, for example, and Wilder Penfield's classic research showed that electrodes applied to specific areas of the brain may evoke particular tunes or other remembered items. | ||
|
| ||
|
But there are some problems. If your memory of your grandmother is constituted by a single neuron, it's rather precarious. Losing that neuron would suddenly render you unable to think of your grandmother. Intuitively, moreover, it seems odd to suppose that thinking of your grandmother is an on/off matter, like the presence of a 1 or a 0 in some grandmaternal register somewhere. Surely when you think of your grandmother you entertain a mental image of her, or recall her voice, her habits, or some significant events in which she was involved? For these reasons, and because the research has generally suggested that many neurons are active when thinking about even simple things, it is almost universally agreed that mental representations are in some sense distributed: that to think of Granny requires the firing of a set or pattern of neurons. Some would suggest that the idea of Granny is built up out of a set of neurons representing various qualities and characteristics the old lady possesses - one set of neurons representing the shape of her face, one set of neurons which fire for relations, one that fires for all old ladies - but this seems to me almost as unsophisticated as the grandmother neuron idea itself. The truth is that we are on the fringes of the mystery of intentionality here, where there are few satisfactory answers. | ||
|
| ||
| What about the new research then? It seems the researchers identified a neuron which fired only in response to images of Jennifer Aniston; the images were different, so it must have been some general Aniston-ness (at least, of a visual kind) which caused the response. While this is not uninteresting, it seems to me to fall well short of reviving Granny Neuron. The researchers did not establish that no other neurons fired in response to Jennifer Aniston (nor, as a matter of fact, that the same neuron would respond to Jennifer Aniston pictures on another occasion). Of course, as with many other experiments, this one was possible only because it could be fitted in alongside a therapeutic investigation, and its scope and nature were naturally constrained by the medical requirements of the exercise. But what exactly was established? If thinking of Jennifer Aniston involved the simultaneous firing of a distributed set of neurons, one would rather expect that there would also be at least one particular neuron which fired on every occasion. It might yet be that grandmother neurons exist, or that they don't; these experiments seem to sit equally comfortably with either hypothesis. | ||
|
| ||
|
|
|
|
|
|
