Yoichi Miyawaki and colleagues interpreted signals from fMRI scans of test subjects brains to generate representations of images at which the subjects were looking. The processed images look something like this:
Sure, the image is grainy. But, they are trying to read your mind. Tricky business that. The original image looked like this:
The researchers used a computer to compare each letter of the first image with a subset of all possible random images (there are 2^100 possible random images), including the correct letter from the second image, which turned out to be the best match*. Laid out like this, the "Holy Crap! It says NEURON!" result is pretty obvious**. But, what if a person was trying to read the output? What if we didn't know the correct answer ahead of time?
Maybe I can't read minds. Boy, those images are compelling, though. This got me worrying. Human brains, like my own, can be very clever when it comes to reading words; eevn wehn the wdors are msesed up. There was also
pareidolia. Pareidolia is the psychological phenomenon of pattern detection in random noise.
It is responsible for the Face on Mars (shown to the right), the Man in the Moon, constellations, and innumerable Jesus sightings in crackers, Cheetos, etc.
With the future of my mind reading skills on the line, I decided to conduct a little test
with some help from my good friends*** at the Skeptalk Email Discussion List****. The volunteers were asked to look at the images of each letter individually (to eliminate the word recognition cleverness alluded to above) and to identify what letter they thought was represented. The volunteers, without their knowledge, were also asked to identify the some images more than once to see how consistent their guesses were.On an individual basis, only 20% of my volunteers were able to identify all the images correctly. As a group, however, they produced a response that looked like this:
My volunteers did not do very well on the "O"; but, overall, "Holy Crap! It says NEURON!". The "O" turns out to be irrelevant to this particular word. The only six-letter word in the English language that fits "NEUR_N" is "NEURON" (trust me I checked the Scrabble dictionary). Leaving out the "O', only 40% of my volunteers would have been successful. Although not up to the computer's standard, as a psychic hive mind they do pretty well for first time mind readers.
There are limitations to this technology. I can see imagine immediate applications neuroscience research. I tend to agree with Bob Novella from the Skeptics' Guide to the Universe that practical, frivolous uses, such as recording dreams is a long, long way off. fMRI has limited resolution. This method requires extensive calibration (try looking at 400 different images for 12 seconds each). Eventually, the technology will get better and our SciFi fantasies may come true.
Once that technology comes online, however, I highly recommend either joining your local hive mind or bowing down to our mind-reading, computer overlords.
NOTES:
*I would like to see them try this with katakana symbols instead of Latin characters.
**By an amazing coincdence this paper was also published in the journal Neuron.***How cool is it to know a set of people, who are jump at the chance to help when asked for help conducting an experiment.
****Generously hosted by Brian Dunning and Skeptoid.com, home of the must listen Skeptoid podcast.
