This is a comment I made on Joi Ito's blog. He's reacting to the report of a study that suggests a correlation between a deficit in metaphor recognition and a particular anatomical segment of the brain:
Joi, confusing correlation with causality is a common fallacy when clinical study results are reported in the press. The key word here is "linked." SciAm is reliable for its reporting but you have to appreciate their careful wording. The scientists "believe" they've found a correlation, and with only 4 patients. But since there is no animal model for testing abstract thinking, and barring human experimentation, this is the best we can expect.
To show you what happens in the blogosphere, boingboing.net takes your post and says, "Brain's metaphor center identified." This is not a knock against Cory whom I respect as a writer and an activist. It's a common misconception especially when you're led to think that this is a significant scientific finding.
It could be a nerve pathway that passes through this region but connects a number of sections of the brain. Or maybe they're just wrong.
If you want to take it to the extreme, you could have 4 people walk out of donut shops, with each getting hit by a car, and then read somewhere, "Donuts Are Deadly."
AKMA: Every medical student is taught how to conduct a mental status exam (MSE) of a patient. You start asking if they know where they are, what the date is, who's the President. Not being able to answer doesn't automatically point to an organic cause, unless you consider tequila organic. Asking (quoting from a published MSE): "What would I mean if I said I were feeling blue? seeing red? I had a chip on my shoulder? or was hot under the collar?" is meant to test abstract thinking. Again, this is used for lack of any other reliable measure of higher cortical thinking.
But what it really comes down to is, why don't we approach any news report with a high level of skepticism, rejecting any claim until it has met our standard of proof or believability? In statistics it's called failing to reject the null hypothesis, i.e. the reported results are most likely due to coincidence.