fMRI: a very critical review, again (Introduction)

by David Turell , Thursday, May 03, 2018, 19:08 (2397 days ago) @ David Turell

From a expert scientist:

http://nautil.us/issue/60/searches/why-happiness-is-hard-to-findin-the-brain?utm_source...

"That’s not really how fMRI works, or how it should work. Back when fMRI was developed, back in the ’90s, what we call the “bad old days” of neuroimaging, there was a lot of what we called “Blobology”: putting people in scanners and hunting around for “blobs” of activity in the brain.

"One of my favorite examples of this is from one of the very first conferences I went to; there was a study being presented called “The fMRI of Chess vs Rest.” Basically, you had people lying in a scanner, either playing chess, or doing nothing. The whole brain was active, but in different ways for the different scenarios, and in the chess scenario certain brain regions would show up as “more” active. From this, they then claimed these regions are responsible for the processes involved in chess. There was so much inverse inference applied: This part is active, and we do these things in chess, so that must be what those areas are for. It’s working backward. It’s viewing the brain like a car engine; the idea that each brain region must do one thing and one thing only.

"This approach leads to these wrong conclusions; you see activity in a brain region and assign it a specific function. But it’s completely wrong. Multiple functions are subsumed by multiple areas, which are handled by cognitive networks. It’s very complicated. That’s a problem with neuroimaging generally; it goes up a notch further when you’re dealing with anything subjective, like happiness.

***

"It turned out Professor Chambers is a very keen and active individual when it comes to highlighting the issues and problems that afflict modern neuroimaging studies, and psychology in general. He’s even written a book, The Seven Deadly Sins of Psychology,1 all about how modern psychology could and should be improved.

"There are several important issues about fMRI that clarified just how hard it would be for me to use it to set up an experiment to find happiness. Firstly, as stated, it’s expensive. So studies that utilize it tend to be relatively small, using a limited number of subjects. This is an issue, because the fewer subjects you use, the less certain you can be that your results are significant. The greater the number of subjects used, the greater the “statistical power”2 of any results, and the more confident you can be that they’re valid.

***

"The data produced by fMRI aren’t nearly as clear as mainstream reports suggest. Firstly, we talk about which parts of the brain are “active” during a study, but as Professor Chambers pointed out, “This is effectively nonsense. All parts of the brain are active, all the time. That’s how the brain works. The question is how much more active are these certain regions, and is it significantly more active than it usually is.

***

“'fMRI has a huge what we call ‘researcher degrees of freedom’ problem. People often don’t decide how they’re going to analyze their data, or sometimes even which question they’re going to ask, until after they’ve run their study. And they go ahead, and they explore, and they have this ‘garden of forking paths’ problem, where in even the simplest of fMRI studies there are thousands of analytical decisions to make, each one of which will slightly change the outcome they get. So what researchers will do is mine their data at the end to find a result which is useful.”

"This comes about because there are many different ways to analyze complex data, and one combination of approaches may provide a useful result, where others wouldn’t. It may sound dishonest, somewhat like firing a machine gun at a wall then drawing a target around where the most bullet holes are clustered and claiming to be a good shot. It’s not that bad, but it’s heading that way. But then when your career and success depends on hitting the target and this option is available, why wouldn’t you do it?"

Comment: same thoughts as usual. The fMRI follows muscle movement and other application areas fairly easily but thought may be in frontal areas and is influenced by other areas controlling emotions, which are harder to define. Gives us some info on how the s/s/c interfaces to use the brain networks.