There has been a huge backlash in the scientific community and blogosphere over a recent New York Times op-ed piece by Martin Lindstrom discussing a functional MRI study of the brain response to hearing and seeing an iPhone ringing.
Purported finding: insula activation.
Interpretation: we love our iPhones, literally.
Why this interpretation?: because insula activation has previously been associated with feelings of love.
Obviously a dubious interpretation and certainly a highly questionable piece of editorial decision making on the part of the Times. Not surprisingly, the response has been vigorous.
Russ Poldrack, a respected UT Austin prof, called it "complete crap" and wrote a letter to the editor of the Times to such effect. The letter, which was co-signed by 44 neuroscientists, was published recently. Poldrack correctly pointed out the flawed logic of the claim and further noted that the insula activates for all kinds of things. I would have signed it too and I'd like to extend my thanks to Russ for taking the time to write the letter.
Tal Yarkoni, a UC Boulder post doc, wrote, "the New York Times blows it big time on brain imaging."
The Neurocritic blog was all over this one too, as was science writer and blogger David Dobbs who wins the prize for the most unrestrained Wired.com headline: fMRI Shows My Bullshit Detector Going Ape Shit Over iPhone Lust
It might surprise you that I'm not going to jump on the bandwagon here. Yes, I agree the claim is complete crap and yes my bullshit detector when ape shit and yes I think the Times editorial staff clearly could use an education on functional imaging. But I'm not too worried about this op-ed piece or the blathering of a pseudoscientist like Lindstrom. Why? Because it is so clearly ridiculous that the most harm it will do is discredit the content of the NYT, stir up some debates about iPhones vs. Androids, and maybe cause the public to question functional MRI a bit more (not a bad thing). Importantly, it will have no impact on scientific progress.
What worries me more, a lot more, are claims by serious, respected scientists that sound reasonable, but are based on the same flawed or weak logic. These claims fly under the radar, go unchallenged and DO impact scientific progress.
Consider another NYT piece published in 2006 called "Cells that Read Minds" by respected science writer Sandra Blakeslee. The article is an excellent summary of the state of scientific thought regarding the function of mirror neurons (you KNEW this was going to come back to mirror neurons, didn't you?!). Let me be clear, what follows is not a critique of Blakeslee, who very accurately summarized the field, but of the logic of the claims made by her sources, respected scientists like Giacomo Rizzolatti, Vittorio Gallese, Marco Iacoboni, and others.
To illustrate my point I'll quote from the NYT piece which quotes Iacoboni:
"When you see me perform an action - such as picking up a baseball - you automatically simulate the action in your own brain," said Dr. Marco Iacoboni, a neuroscientist at the University of California, Los Angeles, who studies mirror neurons. ... "you understand my action because you have in your brain a template for that action based on your own movements.
"You automatically simulate the action" -- this claim comes from the observation that when you watch (some!) actions (not all!) you activate motor-related areas. This is an inference, not a fact. Yes, motor areas do activate during perception, in some experiments, under some conditions. But does this mean that actions are "automatically simulated"? Or are there other possibilities? Iacoboni's own highly cited paper in the journal Science, which showed activation of the motor system during observation of actions also activated just as robustly during the observation of grey rectangle with a dot in it. Does this mean that the motor system "automatically simulates" grey rectangles with dots in them? And the region that was activated was Broca's area, long known to activate during motor action, particularly speech, but also under a variety of other behaviors and tasks, just like the insula.
"...you understand action because you have in your brain a template for that action based on your own movements." To use Poldrack's words, "this kind of reasoning is well known to be flawed." Just because a region previously shown to be associated with a given function (action execution) also activates for another function (action perception), doesn't mean it is doing the same thing for both functions or that the activation is causing the behavior under investigation (understanding). More to the point, the activation of a brain region in such a study tells us nothing directly about what is causing the activation. Example: in the early days of fMRI we were piloting a visual perception task and found very highly correlated and wildly significant activity in the frontal pole during visual stimulation. Using standard logic, this would indicate that the frontal poles were critically involved in low-level vision, an odd finding. What we later discovered was that the mirror that allowed the subject to view the screen was titled down too much so that every time we presented a stimulus the subject had to look up, which moved their head just enough to generate a perfectly correlated change in the signal in the portion of the brain that moved the most, the frontal pole.
Notice that the original interpretation of mirror neurons based on observations in the monkey, that cells fire both during action observation and action execution, is no less flawed logically. There is a correlation, but correlation does not imply causation.
Here's another quote from Blakeslee's piece that discusses the work of Christian Keysers and that may sound suspiciously similar to the iPhone claim.
Social emotions like guilt, shame, pride, embarrassment, disgust and lust are based on a uniquely human mirror neuron system found in a part of the brain called the insula, Dr. Keysers said. In a study not yet published, he found that when people watched a hand go forward to caress someone and then saw another hand push it away rudely, the insula registered the social pain of rejection. Humiliation appears to be mapped in the brain by the same mechanisms that encode real physical pain, he said.
Despite similar logical flaws apparent in the 2006 piece, no one seemed to notice, unlike in the recent NYT op-ed case. There was no letter to the editor signed by a couple dozen neuroscientists, no uproar that I can find regarding the ridiculously flawed logic of the claims, only mindless acceptance or indifference, except for a nice piece in Slate.com by Alison Gopnik who called mirror neurons a myth and suggested that they are the "'left brain/right brain' of the 21st century". This critique has been largely ignored if we take the proliferation of mirror neuron claims as evidence.
This should trouble you. How is it that flawed logic in one domain is obvious and creates a dramatic scientific reaction, while goes largely unnoticed or even rubber stamped in another? It comes down to intuition and bias. We intuitively know that a single fMRI study cannot tell us whether or not we love our iPhones. Further, we are biased by the source: one man's claim based on an unpublished study is easy to dismiss. With mirror neurons the claim seems reasonable, intuitive, and (deceptively) simple. And it is grounded in hardcore neuroscience methods, recording from single cells, with findings reported in the best journals by established, respected scientists. Given such a bias, we don't think about it as hard and are more willing to overlook or just fail to notice the logical flaws.
I don't worry about the crazy claims. They will take care of themselves. It's the ones that make sense that worry me the most and that require all of us to think about a little more carefully.
2 comments:
Thanks for the mention, and for another expert take-down of mirror neurons. While there wasn't a Letter to the Editor in the NYT after Blakeslee's Jan 10, 2006 article, there was some critical blog activity around that time, way before Russ and Tal had blogs. My own blog started in late January 2006, and one of my early posts was a link to Mind Hacks, where Vaughan Bell did a superb job of debunking the ridiculous exercise in publicity and hype known as "the Super Bowl Brain Scans":
Super Bowl brain scans with added hype
. . .
Cardinal sins:
1) Not giving the comparison conditions and experimental design. This makes the reported results essentially meaningless.
2) Interpreting brain activity in certain areas to mean a certain response from viewers, even when they actually report something else.
"female subjects may give verbally very low ‘grades’ to ads using actresses in sexy roles, but their mirror neuron areas seem to fire up quite a bit, suggesting some form of identification and empathy."
Mirror neurons tend to fire when anyone else’s actions are viewed, there is no evidence that approval or liking of the person doing the actions has any bearing on the response.
3) Assuming activation in the ‘mirror system’ equals empathy.
. . .
You can also look through the archives of Mixing Memory for early mirror neuron critiques. Here's a quote from a classic, on Mirror Neurons, Language, and Meaning (Oh, My!):
"There are a few topics in cognitive science that are like fingernails on a chalkboard to me. I find the very mention of them irritating, and the irritation can stick in my craw for days. At or near the top of that list are mirror neurons. These little cells have been made so sexy, either by their sheer coolness or massive, overblown publicity (it's hard to tell which), that they seem to have become the solution to virtually every problem in the study of cognition and behavior in the minds of some researchers."
Finally, I have long ago ceded my crown as king of the mirror neuron critics to you but there are 37 unique posts in The Neurocritic archive. An old favorite from July 2006 is Spindle Neurons: The Next New Thing?
But all this blogging pales in comparison to the published literature, where there are at least 6 papers critical of mirror neurons by Hickok. Keep 'em coming!
Well said.
Here's a quote from Henson (2005) paper from his EPS prizewinning lecture, concerning the "structure-to-function induction":
"if condition C2 elicits responses
in brain region R1 relative to some baseline condition C0, and region R1 has been associated
with function F1 in a different context (e.g., in comparison of condition C1 versus C0 in a
previous experiment), then function F1 is also implicated in condition C2"
This seems to license exactly reverse inference that is currently getting (a similar group of) people so hot under the collar.
I complained about this (and much else) in my companion paper (Page, M, Cortex, 2006, http://bit.ly/iUt23i). As far as I remember, nobody wrote to the New York Times to agree.
Quite right about mirror neurons too.
Mike.
Post a Comment