No, They Did Not Find Love in the Brain

On the day I am writing this post, which I have auto-scheduled for publication at a later date, there are two stories in the science news trying to suggest that scientists made some progress in finding a neural basis for love. One article marks the death of neuroscientist Helen Fisher.  We have a headline of "Dr Helen Fisher, MRI maven who showed just how love works, dies at 79," and a subtitle of "It's all about a chemistry."  The article that follows discusses no robust evidence that this researcher found any such thing as a neural basis for love.

The article has a link to some research by Fisher. It's a link to a page on her web site that then has a link to her 2006 study "Romantic Love: An fMRI Study of a Neural Mechanism for Mate Choice."  It's a very poorly-designed study using a too-small study group of only 17 subjects, and no control subjects. The 17 subjects were people who described themselves as being very much in love. They were shown pictures of the person they loved, along with neutral pictures. We have a claim that some areas of the brain showed more activation when the pictures of the loved ones were shown.  We have no mention of any blinding protocol, no mention of any control subjects, and no mention of any sample size calculation to try to determine whether the study group size was adequate.  

The lack of a blinding protocol and the lack of any control subjects are enough to disqualify this study as being any robust evidence for a neural basis for love. An intelligent way to design a study like this would be to have a number of control subjects equal to the number of people who claimed to be in love, the control subjects being people who were not in love. Then "blinded" analysts examining only the brain scans (without knowing whether a particular brain scan belonged to someone who claimed to be in love)  could examine the scans, and attempt to predict whether a particular set of scans belonged to a person in love who was seeing a picture of his loved one. A high predictive success (maybe 90%) might suggest some neural basis for love, although it would be necessary to replicate such a finding. 

Nothing like that is done in the study. Instead, it simply had analysts checking all parts of the brain, looking for some area that could be claimed as an area of "activation." A claim to have found a few such tiny regions is just what we would expect even if the brain has no involvement in love.  Similarly, if you scanned the liver of 17 subjects while they were looking at photos of their loved ones and photos of strangers, and you had freedom to check any of 1000 tiny liver areas, you would no doubt find random variations that would allow you to claim that some area of the liver is involved in love.  What is going on Fisher's study is mere noise mining. Someone is looking at random, noisy data, and trying to find some evidence of something, which the data is not actually presenting. 

In an fMRI study you know that a decent "superior activation" has been found when the paper uses the phrase "percent signal change" to indicate a signal change of more than 2%.  The Fisher paper has no mention of any percent signal change. 

An equally weak piece of research is Fisher's 2012 paper "NEURAL CORRELATES OF MARITAL SATISFACTION AND WELL-BEING: REWARD, EMPATHY, AND AFFECT," presented on Fisher's web site as an example of her work. It has exactly the same flaws as the 2006 paper discussed above: a too-small study group size of only 17 subjects, a failure to do a sample-size calculation (which would have revealed the inadequate statistical power), a lack of any control subjects, a lack of any blinding protocol, and a failure to report any impressive result as a percent signal change of even 1%. No effect size is reported. 

The same flaws are also found in Fisher's 2011 weak science paper "Neural correlates of long-term intense romantic love," another paper presented on her web site as if it were an example of her best work.  It has a too-small study group size of only 17 subjects, a failure to do a sample-size calculation (which would have revealed the inadequate statistical power), a lack of any control subjects, a lack of any blinding protocol, and a failure to report any impressive result as a percent signal change of even 1%. No effect size is reported. 

It's very easy to explain the kind of tiny blips reported in these studies, without any belief that romantic love has a neural basis. In the type of studies done, people were shown photographs that might either be of their loved one, or a stranger. Seeing a photograph of a loved one, a typical subject might have had a smile of recognition. But it well known that even very small muscle movements can cause fMRI blips. 

In the article about Fisher's death, we have this strange quote about the 2006 paper discussed above: " 'I distrust about 95 percent of the MRI literature and I would give this study an 'A'; it really moves the ball in terms of understanding infatuation love,'  Dr Hans Breiter, director of the Motivation and Emotion Neuroscience Collaboration at Massachusetts General Hospital, told The New York Times after the publication."  No, all of the papers discussed have the same kind of flaws that should cause you to distrust 95% of the fMRI literature. 

Two of the three papers by Fisher used the very misleading "lying with colors" technique in which tiny differences in brain activation less than 1% are misleadingly depicted in bright red or bright yellow, giving the incorrect impression of a major difference when there was no such difference. 

Neuroscientists deceive us with such "lying with colors" diagrams, which visually create the impression that very tiny differences in activity  such as 1 part in 200 are very big differences. Another way in which neuroscientists deceive us in studies such as these are by making statements that there was "activation" in some particular region of the brain. Such language gives readers the impression that there was some turning on effect in which an inactive region of the brain started to become active.  The truth is that all regions of the brain are electrically active at all times. All neurons fire at rates averaging about 1 time per second or more. 

So imagine you have found by an fMRI reading a case in which some tiny region of the brain starts to show maybe 1 part in 200 greater activity than other regions. Is it correct to call that "activation"? No, it is not. Activation means to start become active. When all regions of the brain are continually active, and are not varying in activity by more than about 1 part in 200, it is deceptive to claim that "activation" occurred when some little difference such as 1 part in 200 was first seen. This deception occurs abundantly in a large fraction or most brain scan papers. 

On the same day the death of Fisher was announced, we had a link to a news story promoting the paper "Six types of loves differentially recruit reward and social cognition brain areas."  We read of an fMRI study about love and the brain, one using a much larger study group size, one of 55 subjects. A study group size like that may make you realize how inadequate were the study group sizes (only 17) of Helen Fisher's papers discussed above. If scientists felt a need to use 56 subjects for a brain scan study involving love, there was presumably some reason why they thought a study group size such as 17 was way too small. 

Unfortunately the new paper uses a protocol that is nonsensical. It involves scanning people's brains while they were read passages such as these (there were dozens of different paragraphs like these):

• "You are having a candle lit dinner with your partner. You look into their eyes over the table, and you share a mutual understanding without words. You love your partner."
• "Your child runs to you joyful on a sunny meadow. You smile together and the sunrays flicker on their face. You feel love for your child."
• "You need help moving house and you call your friend. They promise to of course come to help out, and soon you are lifting cardboard boxes together in a van. In the middle of the ordinary situation you feel love for your friend."

The authors have assumed that hearing such phrases would cause subjects being brain-scanned to feel different types of love. But there is no reason to think that such phrases would do that. Did you feel any love upon reading the sentences above? Almost certainly you did not. 

The paper provides no evidence that the differences in brain activity recorded were greater than 1% or even half of that, and there is no reason to think that they have picked up anything other than random variations in brain activity, variations having nothing to do with love. The paper makes no mention of any percent signal change detected, which is typically what would happen when the detected percent signal changes are very low and unimpressive.  No robust evidence has been provided to show a brain basis for love. 

A recent fMRI study seems like a rare example of a well-done brain scan experimental study. The study "Resting-State Functional Connectivity of the Amygdala in Autism: A Preregistered Large-Scale Study" involved brain scans of more than 400 people, including more than 200 with autism. Rather than the usual "fishing expedition" allowing investigators to check any of hundreds of places in the brain, like the Fisher studies discussed above,  the study was a pre-registered study that committed itself to examining only differences in one tiny region of the brain, the amygdala region. The study reports finding "no reliable evidence for atypical functional connectivity of the amygdala in autism, contrary to leading hypotheses."  When we look at the pre-registration in a study like this and the use of hundreds of brain scans, the Helen Fisher papers discussed above seem all the more like junk work.