In the paper, the researchers identified neurons that selectively responded to one of four pictures of celebrities (i.e. a neuron that fires to Marilyn Monroe, but not Michael Jackson). Then, participants were given a target image (e.g. Marilyn Monroe) and then presented with images that consisted of two of the celebrities overlapped with 50% transparency, one was the target and the other was one of the three other celebrities. The neuronal activity from the four neurons of interest was recorded, and in near-real time, the transparency of the image could be adjusted to show more of the preferred image of the most active neuron. The patients were instructed to do whatever necessary to make the display turn into the target celebrity.
The patients were able to consciously control the neural firing to change the display to the target image about 70% of the time. The even cooler thing is that they were able to do this on the very first trial over half of the time! What were the subjects doing to control brain activity? Many reported that instead of engaging in mental imagery, they formed conceptual associations.
Overall, I think this paper is cool for two reasons: 1. it shows strong top-down modulation of high-level perceptual areas (and without needing massive amounts of training) and 2. using these regions and decoding algorithms might help new brain-computer interfaces for rehabilitation.
Cerf, M., Thiruvengadam, N., Mormann, F., Kraskov, A., Quiroga, R., Koch, C., & Fried, I. (2010). On-line, voluntary control of human temporal lobe neurons Nature, 467 (7319), 1104-1108 DOI: 10.1038/nature09510