STATS ARTICLES 2012
Math Pain: It’s Possibly Not Even In Your Head
Rebecca Goldin, Ph.D., November 20, 2012
“When people worry about math, the brain feels the pain,” claimed the University of Chicago News recently. And no, that’s not merely the kind of metaphorical pain you might have when thinking about a calculus exam in the past or around the corner. We’re talking about a tangible “ouch.” As CBS News warned, “worrying about a math test can quite literally hurt.” Not to be outdone, MSN Now asked whether you have “ever strained your brain trying to tally up your shopping list or calculate a tip?” Wonder no more: “Math really does hurt.”
The study behind what turned out to be an orgy of masochistic news reporting (presumably writing about thinking about math hurts too, but in a good way) was conducted by University of Chicago psychologists Ian Lyons and Sian Beilock and just published in PLOS One.
What they found was that pre-identified math-phobic people, when anticipating an upcoming math-task, experienced activity in the dorso-posterior insula (INSp) – a region of the brain associated with pain. People who didn’t fear mathematics didn’t experience activity in this region. And it was specifically math – a similar word test didn’t produce the same effect.
What all of this reveals is not the silent torture of the innumerate in a sadistically quantitative world, but how easily we – or rather the media – misinterpret these kinds of experiments – and how little, in fact, we – that is, scientists – know about how our brains respond to tasks we’re asked to do. Consider one of the key mechanistic elements in the media’s explanation of how math pain is supposed to happen, the repeated claim that thinking about math activated the brain’s “pain network.” First, and let’s be painfully clear here: there was no actual pain involved in this study; second, it’s not even clear that the pain network was activated.
What the authors of the study wrote was that “Our data suggest that pain network activation underlies the intuition that simply anticipating a dreaded event can feel painful.” And that painful bit of prose hedging underlies the reality that fMRI signals do not indicate activation of the “pain network;” rather, they reflect oxygen changes in blood.
At best, the particular oxygen changes seen by Lyons and Beilock are a surrogate for neuronal activity, but no one really knows. It may be that the activity is happening somewhere else in the brain in a less intense way – but that a central location for the receipt of signals is showing levels of activity above normal. The suggestion that fMRI signals imply activity is not one that has been settled in neuroscience; indeed, many scientists are skeptical.
There are other, important reasons to be skeptical of his study. Consider how the authors found the region with all the activity. Fourteen math-phobic people were put into an fMRI machine and given math and word cues – these cues indicated that a problem (either math or word-related) was about to be given. Their brain responses were recorded. The authors then picked the regions of the brain with the most difference between math-cue activity, and word-cue activity. One of these regions in which a greater difference was observed was the INSp.
But there’s a problem with this set up. By identifying the most active regions for people with high anxiety, we are almost assured that people with low anxiety will have a weaker response in those areas. In other words, the authors set up the data analysis so that we would expect the lower anxiety people to have less activity in these identified regions, regardless of the relationship of anxiety to neural firing.
This is a spin-off of a Texas sharpshooter fallacy Two terrible shots are competing to see who can shoot more accurately. The first shoots at the side of the barn randomly and then draws a bulls-eye around the best “cluster” of bullet-holes. He then invites the other person to shoot at the bulls-eye. The second Texan also discharges randomly, resulting (generally speaking) in fewer shots in or near the bulls-eye. He is declared a worse shot than the first one, even though they both shot randomly.
In neuroscience, this problem has been dubbed “double dipping”– using the data to decide which statistics to do, and then using these statistics on the same data to claim the data show a relationship for which they were chosen. It doesn’t mean that there isn’t a correlation at the end of the analysis, it just means that the correlation could be a product of circularity.
Even if there were a correlation between anticipating a math problem and activity in this region of the brain, we have to wonder what this means. Does being asked to confront any fear activate the same region of the brain? Is there anything specific to mathematics as opposed to any other kind of anxiety? Unfortunately the “controls” were words versus mathematics, and not “being primed for anxiety” versus “being primed in the context of math anxiety.”
Now back to pain, which is easier to grasp, but, notoriously, hard to quantify. At the very least, it typically involves a self-reported experience of pain. Once again, these experiments did not involve actual pain. They involved media-inferred pain: since the lit up brain regions are also involved in responding to pain, ergo math-phobic people are experiencing pain when anticipating a math problem if these regions of the brain light up. But as a matter of their subjective experience, pain doesn’t seem to be the real problem. Anxiety is.
Unfortunately the media coverage ended up signaling an all too typical response to studies waving brain scans, which is to treat them as holy writ in terms of causality. But at best, fMRI is a god of small things. And those small things might not mean very much. The only hard evidence here is that journalism about neuroscience continues to be painful to read.
Rebecca Goldin, Ph.D. is the Director of Research at STATS.org.