You Just Read my Mind…

STOP PRESS!Scientists decode words from brain signals, fueling hopes for mind reading!” “Mind-reading device could become reality!” “Scientists make telepathy breakthrough!” “Secrets of the inner voice unlocked!” – just some of the sensational headlines that greeted a PLOS Biology research article a couple of years ago. OK, so I might’ve put the exclamation marks in myself, but the science behind the claims was indeed pretty extraordinary.

Spectrograms of the original stimulus (top) and reconstructed speech (bottom). doi:10.1371/journal.pbio.1001251.g002

Spectrograms of the original stimulus (top) and reconstructed speech (bottom). doi:10.1371/journal.pbio.1001251.g002

I’ll talk more about the study itself in a moment, but how do you know when you’ve published a great paper? Traditionally it would be a matter of counting journal citations, but that captures only one strand of the complex set of influences that a paper can have. Now we can separate out these strands and follow them, each with its own characteristic. The quick spike of Twitter, the steady spread on Facebook, the rash of saves in Mendeley, of citations in F1000Prime, the complex dynamics of html page views, PDF downloads, secondary coverage in scientific and popular media and in the blogosphere, and – yes – the slow-burn of journal citations. Each has its own time course, its own demographic (age, field, background) and its own texture. And each speaks to a subtly different aspect of the work’s appeal.

There are methodological papers that emerge unnoticed and become citation classics, and there are papers on intriguing animal behaviour that are splashed across the tabloids but have single-figure citations. Two very valid forms of impact, but back in 2012 we published a truly fascinating paper, and two years of metrics show that it managed to hit both buttons very firmly.

Much of neuroscience arguably involves subjecting an animal to a stimulus and then trying to find out how the brain responds. This paper describes a spookily successful attempt to achieve the reverse – looking at the brain’s activity and trying to reconstruct the stimulus that must have caused it. By placing electrodes directly in contact with the auditory cortex, they were able to “mind-read” the words that the person had heard. Listen to this incredible recording of spoken words paired with their “mind-reading” reconstruction:

 

Red circles show the position of electrode arrays on the surface of the superior and middle temporal gyrus. doi:10.1371/journal.pbio.1001251.g001

Red circles show the position of electrode arrays on the surface of the superior and middle temporal gyrus. doi:10.1371/journal.pbio.1001251.g001

Normally the human cerebral cortex lies nicely protected by the cranium, but in this study the authors were able to exploit an extraordinary opportunity. Medics removing tumours or epileptic tissue from highly sensitive parts of the brains of fifteen patients were eager to keep their patients awake so that they could check that they didn’t affect anything crucial. This gave the authors a 10-12 minute window of time when the cortex of an alert patient was exposed to the atmosphere.

During this period an array of electrodes was placed on the surface of the auditory cortex and recordings were made while single words were spoken to the patient. The signals received were used to construct a computational model of the relationship between the stimulus (“Waldo”) and the representation in the brain. This relationship could then be flipped to reconstruct a passable (and recognisable) reproduction of the original spoken word (“Woodor”).

Read the full paper for the details of research, listen to the astonishing audio file of the reconstructed speech, and for a lively discussion with authors Brian Pasley and Bob Knight, I strongly recommend that you listen to Ruchir Shah‘s excellent PLOS Biology podcast (some more remarkable recordings of the speech stimuli and their paired reconstructions appear at 11:15-12:10). In their discussion, they reveal that a long-term translational aim is to use the brain activity elicited by people’s imagined words to drive a prosthetic speech device for those who are unable to speak.

To date the paper has received nearly 85,000 page views, including over 9,000 PDF downloads, and has been cited 66 times already (see more metrics here). It’s the 48th most saved PLOS article in Mendeley of all time, is cited in two Wikipedia entries (“Thought Identification” and “2012 in Science“), and has respectable Facebook activity (21 likes, 467 shares, 33 posts). Twitter looks a bit thin on the ground because we only started collecting stats 5 months after the paper was published, so missing the main spike. But as well as the healthy academic attention, the article attracted massive press coverage that taps liberally into the memes of mind-reading and telepathy, and has inspired bloggers from Wales to Brazil. Now that’s a great paper.

This entry was posted in Biology, Computational biology, Debate, Neuroscience, PLOS Biology, Publishing, Research. Bookmark the permalink.

2 Responses to You Just Read my Mind…

  1. Pingback: Making Metrics Count - ALM Article Feature Series - PLOS Tech

  2. Pingback: “Low T” and Prescription Testosterone: Public Viewing of the Science Does Matter - EveryONE

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>