View Full Version : Bug in fMRI means 40,000 studies might be invalid


Fortune
08-23-16, 08:10 AM
So there's a bug in the fMRI that may invalidate up to 40,000 studies of the human brain.

That is to say, we might know a lot less about the human brain than we think we know, and what we think we know is already not very much.

Article here. (https://curiosity.com/topics/about-40000-fmri-studies-might-now-be-useless-curiosity/#how-does-fmri-brain-scanning-work-alan-alda-and-dr-nancy-kanwisher-mit-brains-on-trial)

Little Missy
08-23-16, 08:21 AM
So there's a bug in the fMRI that may invalidate up to 40,000 studies of the human brain.

That is to say, we might know a lot less about the human brain than we think we know, and what we think we know is already not very much.

Article here. (https://curiosity.com/topics/about-40000-fmri-studies-might-now-be-useless-curiosity/#how-does-fmri-brain-scanning-work-alan-alda-and-dr-nancy-kanwisher-mit-brains-on-trial)

And somehow that does not surprise me at all.

Unmanagable
08-23-16, 08:41 AM
I've learned we know much less about most things than we think we know, as well as much more than we are allowed to believe, regardless of our levels of training and piles of research offered up as our main options to choose from. Thanks for sharing!

Fortune
08-23-16, 02:35 PM
It might mean, for example, that all of the neurological evidence for ADHD gathered via fMRI is inaccurate.

SB_UK
08-23-16, 04:06 PM
So there's a bug in the fMRI that may invalidate up to 40,000 studies of the human brain.

That is to say, we might know a lot less about the human brain than we think we know, and what we think we know is already not very much.

Article here. (https://curiosity.com/topics/about-40000-fmri-studies-might-now-be-useless-curiosity/#how-does-fmri-brain-scanning-work-alan-alda-and-dr-nancy-kanwisher-mit-brains-on-trial)

Reminiscent of one of my favourite quotes from here.

Stabile.
The scientist knows to calibrate all of his instruments, at least all bar the one that really matters - his/her mind.

Scientists are extremely dependent on correctly calibrated instrumentation - but mostly on a mind.

TygerSan
08-23-16, 05:22 PM
I haven't had time to really delve into the full implications of this paper. From what little I've read, it seems like the reality is somewhere between the sensationalist "the sky is falling and all research is crap" and "nothing to see here."

First of all, the bug does artificially raise false positive threshold, but that does not mean that 70% of all studies have false positive rates (which seems to be how some reporters have interpreted the results??)

Secondly, fMRI encompasses a couple of different techniques, and while this bug affects one of the most commonly used techniques, it does not affect others like functional connection studies.

That said, I'm not entirely surprised that the bug was perpetuated as long as it has been. Usually what happens is someone writes home-cooked software and distributes it. Loads of people use it and modify it, and eventually someone finds a problem with it.

I found a small bug in the software I used during grad school. It wasn't huge, and didn't really affect the conclusions drawn from the calculated values, but it had been perpetuated for years before I found it; including in published papers. So, yeah, things happen.