Wolfgang Köhler praised ‘trespassing as a scientific technique’, on the grounds that what is merely special data in one field may turn out to have much broader significance in another. Now this doesn’t mean the trespasser sees the big picture in a way that eludes everyone else. Trespassing can be helpful by accidentally treading spores from one field into another, where they unexpectedly start fizzing and wriggling into life. Or the trespasser might find fertilizer sacks full of rubble and rusty cogs blocking the entrances to badger setts. Certainly one of the great joys of researching this book has been to disinter fascinating brain science buried under all the reductive bluster.
And then there’s the fact that brain science appears to have arrogated to itself all understanding of human behaviour anyway, which makes it kind of hard to move a muscle without trespassing. In fact, since the brain science fiefdom now includes life, the universe and everything, the question is who is the real trespasser here? In the words of the great comedian Michael Redmond:
People are always saying to me, ‘What are you doing in my back garden?’ To which I reply: ‘What are you doing in my house?’
Let’s go and climb the back steps and see what they are doing in our house.
From the get-go, it is important to remind ourselves that brain-imaging does not actually film your brain in action. There is no live action footage of thoughts or feelings. No one will ever be able to read your mind – except your mum. Brains do not light up during functional magnetic resonance imaging (fMRI) and electroencephalography (EEG). Strictly speaking fMRI and EEG are not techniques of brain imaging but of blood imaging, since they track blood flows to different brain regions on the working hypothesis that active neurons devour more oxygen, and blood is the brain’s oxygen delivery service.
On 17 May 2016 the Proceedings of the National Academy of Sciences of the USA published the first comprehensive review of 25 years of fMRI data*. The conclusions were damning:
* Anders Eklund, Thomas E. Nichols & Hans Knutsson, ‘Cluster failure: Why fMRI inferences for spatial extent have inflated false-positive rates’, PNAS, 2016.
In theory, we should find 5 per cent false positives … but instead we found that the most common software packages for fMRI analysis … can result in false-positive rates of up to 70 per cent. These results question the validity of some 40,000 fMRI studies and may have a large impact on the interpretation of neuroimaging results.
In 2009, the journal Perspectives on Psychological Science published ‘Puzzlingly High Correlations in fMRI Studies of Emotion, Personality and Social Cognition’. (Original title: ‘Voodoo Correlations in Social Neuroscience’).* Two of the paper’s authors, Ed Vul and Harold Pashler first became suspicious when they heard a conference speaker claim he could predict from brain-images how quickly someone would walk out of a room two hours later. This had to be voodoo.
* Ed Vul et al., Perspectives on Psychological Science, 2009.
Imagine a scenario in which roaring floodwaters smash the office windows. Your colleagues flee, but then turn back to see you stranded in the rising water.
‘Save yourself!’ they cry. ‘Run for you life!’
‘You go on ahead’, you holler back. ‘I’m not gonna make it. I decided a couple of hours back on an airy saunter through the doorway.’
‘Well, sashay for your life! Mince like you’ve never minced before!’
Vul et al. set about re-examining the data. They surveyed the authors of 55 published fMRI papers and found that half acknowledged using a strategy that cherry-picked only those voxels exceeding chosen thresholds. These cherry-picked voxels were then averaged out as if they were the average of all voxels, not just the ones that fit the hypothesis they were supposed to prove. This strategy, says Ed Vul, ‘inflates correlations while yielding reassuring-looking scattergrams.’
Voxels are the organisation of statistical correlations into cuboid 3D pixels. Each cube represents a selective sample of billions of brain cells. They provide a computer-generated image of what brain activity would look like if cherry-picked statistics matched raw data. Together the cubes build a Minecraft map of the mind.
To form each cuboid voxel, you collate all the neuronal clusters that have a blood oxygen level of x at split second 0.0000001 with all the ones that have a value x at split-second 0.0000007. Junk all the non-x brain cell activity going on between 0.0000002 and 0.0000006. (Call it ‘noise’). Now amalgamate your cherry-picked voxel with other voxels, (themselves boxes of cherry-picked data), and there you have your fMRI picture showing which region of the brain spontaneously ‘lights up’ when we are thinking about love or loss or buying a house. There you have the murky world of the technicolour voxel.
Ed Vul et al. call this strategy ‘non-independent analysis’. To illustrate how this strategy inflates correlations they used it to show how daily share values on the New York Stock Exchange could be accurately ‘predicted’ by the recorded fluctuations in temperature at a weather station on Adak Island in Alaska. Here’s how it works. Non-independent analysis simply skims the strongest correlations between each of the 3,315 stocks being offered on Wall Street, and finds a handful whose value appears to strongly correlate with the previous day’s temperature drops on the windswept Alaskan tundra.
‘For $50, we will provide the list of stocks to any interested reader,’ wrote Ed Vul. ‘That way, you can buy the stock every morning when the weather station posts a drop in temperature, and sell when the temperature goes up.’
Not long after came the banking crash. It turned out that Wall Street had been using some ‘non-independent analysis’ of its own. In the same way that fMRI false positives are boxed up into cuboid voxels, Wall Street was boxing bad debt into mortgage bonds and ‘collateralised debt obligations’, or CDOs – better known as the wobbly stack of Jenga blocks Ryan Gosling eloquently demolishes in The Big Short.
And yet ponzi voxels were to rescue ponzi bonds. Newsrooms used brain-imaging data to explain the banking crash. Neuroscience helped shift the blame from banks to brains, and from rich to poor. It turned out that the system of short-term greed that caused the banking crash was the limbic system. Neuroeconomists
popped up on the nightly news to explain that sub-prime mortgages were entered into by people who let their limbic system’s urge for instant gratification triumph over the prudence of their prefrontal lobes. Those who lack the mental strength to resist the limbic system’s short-term greed, it turned out, would always make bad property investments. Grotesquely, ponzi voxels rescued ponzi bonds by shifting the blame onto the feckless poor.
A huge jar of sweets in a sweetshop window
To want to understand human behaviour is a human need, but frustratingly the answers are always complex and incomplete. There is no royal road to the truth, just a multiplicity of weakly-acting causal pathways. And so when we are shown a new technology that appears to answer our deepest questions, it is only human for us to want to fill our boots. EEG and fMRI are what we have been looking for all along: shiny machines that produce simple answers to complex questions. Better yet, these answers come in the form of vivid arrangements of 3D voxels, like a huge jar of sweets in a sweet shop window.
In the rush for a quick-fix answer to a complex problem
