Two recent neuroscience news items in The Independent represent exactly the problem with bad science journalism today and the tendency to overhype incremental studies.

Brain-Machine Interface

Here’s the first:

Device that can literally read your mind invented by scientists. An ‘easily operated’ machine linked to a smartphone could be ready within five years.

Um, no.  I have be writing about this technology for years, because it is genuinely interesting and I think is a technology to watch. Several labs have made significant progress in brain-machine interfaces. The idea is that you read the electrical activity of the brain with either scalp electrodes or brain surface electrodes. Scientists have developed software that interprets the EEG patterns and learns to correlate them with the thoughts or intentions of the subject. The subject, in turn, learns to control their mental activity to affect the EEG output.

Here is where the technology stands: With brain surface electrodes, you get a much greater resolution of EEG activity. The software has progressed to the point that monkeys can control a robotic arm with sufficient subtlety to feed themselves.

With humans we have mostly used scalp electrodes, which have a more blurry signal. Even with these people have learned to control robots or control a cursor on a computer.

Further, using fMRI and EEG analysis some labs have been able to “read” the minds of subjects in a very crude way, by telling what letter they are looking at, for example.

The main limiting factor to this technology right now is the electrodes. Scalp surface electrodes do not give enough detail, and brain surface electrodes are too invasive. The electrodes move when the brain pulses, and eventually scar over and stop working. They are also at risk of infection. Labs are working on developing electrodes that you insert into the veins of the brain, or more flexible brain surface electrodes that reduce movement and scarring.

The software is already pretty robust, but is improving incrementally as we get better at designing learning algorithms.

It’s hard to say when all of this will get to that tipping point where medical applications become viable and even routine. Five years is too ambitious. Something like 10-20 years is more realistic. I do think it’s coming, however.

With this background, let’s look at what the team from Toyohashi University of Technology in Japan accomplished. they used a cap covered in scalp electrodes to read EEG activity, and used “deep learning” softward to analyse the activity. They had subjects say out loud numbers from 0 to 9. The software was eventually able to distinguish among these options with 90% accuracy.

They also tested with 18 monosyllabic words, and were able to achieve 61% accuracy.

OK, that’s interesting. That is a nice incremental study. It is not fundamentally different than prior research, however. At best it is a small advance in the software.

The Independent article, however, takes this tiny advance and goes entirely off the rails. First, scientists did not invent a “mind reading device.” They didn’t invent any new technology, and it is a real stretch to call this mind reading. It is using EEG signals to match a very limited number of targets. Also, in this study the subject were saying the number or word out loud. It’s unclear how that affected the information going to the EEG (muscle movements also generate electrical activity).

Most of the article, however, was spent in wild speculation. They were throwing out applications that may be 50 years away, without making that clear. The notion that a mind-reading device will be ready in a smartphone app in five years is absurd. First, they claim they can reduce the number of electrodes. What is that based on? The main limiting factor here is that surface electrodes give a blurry signal, so you have to use a lot of them. I see no advance that changes this fundamental limitation.

Further, they are probably compensating for using scalp electrodes with computer crunching power. It sounds optimistic to say this will be a simple smartphone app in five years.

Everything You Know is Wrong

Here’s the next article:

‘Mind-blowing’ discovery could revolutionise our understanding of how brain works. ‘The implications, if this interpretation is correct, are massive’

Again, the study is interesting, but the implications fall far short of “mindblowing” and I highly doubt it will have “massive” implications.

The researchers used fMRI scanners to look at brain activity in people born with only one hand. They found that when the subjects were doing tasks they would normally do with the missing hand, this often involved multiple body parts. Those body parts, of course, lit up. But also the part of the cortex that would normally serve the missing hand also lit up. That part of the brain now maps to the body parts substituting for the missing hand.

They conclude:

Because the same body parts used for compensatory purposes are those showing increased representation in the missing hand’s territory, we suggest that the typical hand territory may not necessarily represent the hand per se, but rather any other body part that shares the functionality of the missing hand.

Again, interesting. However, the conclusion is pure speculation, and in my opinion is a bit of a stretch. But listen to how this gets translated in the Independent article.

This suggests that the brain is not organised so that each area is responsible for an individual body part, but that different areas are responsible for different functions.

Dr Makin said: “If true, this means we’ve been misinterpreting brain organisation based on body part, rather than based on function.

“It’s kind of mind blowing for me to think we could have been getting this wrong for so long.

“The implications, if this interpretation is correct, are massive.”

Sorry, but this sounds like utter dreck.

First, the motor cortex is clearly organized by body part. There is what we call the motor homonculus, which is a physical representation of the body on the motor strip. It is organized anatomically, not functionally. The size of the body part on the motor cortex is directly proportional to the number of motor units needed to control that body part. So, the hands are large because our hands have a lot of small motor units. The deltoids are small because they have fewer (but larger) motor units, with correspondingly cruder control.

The fallacy here is the false dichotomy, the notion that the motor cortex is either anatomically or functionally organized. It is almost certainly both. The basic layout of the primary motor neurons are clearly anatomical. However, we do stuff with our muscles, and these actions require networks that are functionally organized.

These new findings, while interesting, I don’t find surprising at all. It does not imply that the hand region of the motor cortex maps only to what the hand does, not the hand itself, and the article suggests.

There are two other factors that limit how we can interpret this study. First, the hands are different than any other body part because we do so much with them. It makes sense that the motor cortex for the hands would have lots of functional connections to other parts of the cortex.

Second, the subjects in this study are not normal. They are missing a body part, so the usual program of somatotopic organization breaks down. There is, essentially, a large part of the motor cortex that would normally map to the hand that now has nothing to do. We know from prior studies that unused brain cortex will get recruited.

So, there is part of the brain with nothing to do, and the person has to learn new motor techniques to compensate for the missing limb. Hmm…. I wonder if the unused part of the brain will get recruited to do the new tasks.

So the fundamental conclusion of this study may not be true, and may just be an artifact of the subjects they are studying. That would be, in my opinion, the more parsimonious interpretation.

Further, even if true (that the hands have some functional mapping) it may be limited to the hands, and further it probably does not replace but supplements anatomical mapping.

At best we have a totally incremental and entirely unsurprising (although interesting) new bit of information. However, we cannot be sure how to properly interpret the results of this study without a lot more information.

Mind not blown. Paradigm not shifted.

What is sad is that there is a lot of interesting neuroscience to talk about here. The study is genuinely interesting, and when put into its proper context is newsworthy. It’s simply not a revolution.

Conclusion

When I read such articles it is always my question as to who is mostly to blame here. Was the hype coming from the researchers, the press office, or the journalists? This question has actually been studied, and it seems that statistically the press office is most commonly to blame. However, all three usually contribute to some degree.

I understand why scientists want to emphasize what could be exciting implications of their research, and I am not suggesting they should all be boring sticks-in-the-mud. But they should not be naive about science journalism. They have to be very careful when discussing the implications of their research and anticipate how it can be overhyped.

There also needs to be a better working relationship between scientists and their press offices.

Finally, we need dedicated science journalists who can put a story into a proper context. The Independent did a terrible job on both articles. They did not appear to talk to other scientists to help them put the study into context. They took the hype offered to them by the scientists or their press representatives and ran with it.

In both cases an opportunity to learn some fascinating neuroscience was missed and replaced with a distraction of unwarranted hype.