This really is old news but we recently stumbled across a brilliant blog post from Neuroskeptic about a controversy that erupted from a methodology paper about errors in research comparing fMRI brain imaging with aspects of behaviour or experience. The original paper got a lot of media coverage and basically pointed out that some of the methods used in analysing this research are likely to exaggerate the actual relationships seen between brain activity and behavioural measures. This is important to us as the current explosion in fMRI research and pain might also be prone to this error. It is one thing that the message from fMRI research frequently gets distorted along the path from laboratory to media reporting (ladies and gentlemen I give you “acupuncture lessens pain in the brain not the body”) but by understanding these more subtle errors we can at least be sure that the right message leaves the lab. Anyway, enough from us, Neuroskeptic wrote what we think is an incredible review and analysis of the whole issue. There’s nothing useful we can add but to say READ THIS!
Vul, E., Harris, C., Winkielman, P., & Pashler, H. (2009). Puzzlingly High Correlations in fMRI Studies of Emotion, Personality, and Social Cognition Perspectives on Psychological Science, 4 (3), 274-290
{ 4 comments… read them below or add one }
Wonder what you think of this research?
http://home.uchicago.edu/~decety/publications/Decety_NI2010.pdf
[Reply]
Neil O'Connell Reply:
March 23rd, 2010 at 7:51 am
Thanks Anne Marie. I love studies like this. The correlation problem highlighted in the Vul article, and explained by Neuroskeptic does not apply here as they are using EEG rather than MRI and as such are have far fewer comparisons to deal with. From the looks of things (in my humble opinion etc) they have performed a reasonable and fair statistical analysis. Fascinating findings, someone should blog on it!
[Reply]
More from author
Hey – Haven’t had a good look yet but it sure looks interesting. those guys are pretty good too.
thanks for the heads up – will take a good look….
Lorimer
[Reply]
If you want more information, you can start by visiting our research group: The Bergen fMRI Group.
[Reply]
More from author
Thanks for the link!
That EEG study looks kosher to me from a brief look, although the Vul et al non-independence error can certainly be a problem in EEG (and PET, MEG, any neuroimaging technique) as well as in fMRI.
Problems arise whenever you measure activity at many different points in the brain, and then focus in on the parts where the biggest effects seem to be occurring. If you then do statistics on those parts without taking into account that you’ve selected them specifically because you saw big effects there… you might end up in error.
[Reply]
More from author