How sparse are engrams for concepts?

I missed this one when it came out earlier this year in Scientific American. One of the authors, Christof Koch is Chief Scientific Officer at the Allen Brain Institute, which collaborated with Krasnow on the DIADEM challenge.

The key point is that certain scientific questions can only be asked in a particular methodological context–in this case, epileptic surgical patients. I’m not sure I agree with the conclusion however–there is an awful lot of alternative evidence supporting the notion of widely distributed engrams. Koch would argue (I think) that the methods that support the distributed hypothesis are flawed. Functional brain imaging using MRI for example has both poor spatial and temporal resolution compared to the electrophysiological data from these patients.

But if functional brain imaging is flawed in this way, it’s also conservative–very sparse representations would be invisible. In fact, the representations that show us are widely distributed. See the work of Tom Mitchell and colleagues, here.