Dead Fish Tell Tales

When A Dead Salmon Changed NeuroImaging

Studying the brain or the mind can have many adverse effects on the brains and minds of those studying them. It can make one do weird things. A half-century ago, it made us cut parts of the brain out to see what effect it would have. Sometimes it makes us electrocute some poor mice's brains to see how they would change.

And sometimes, just sometimes, it makes us put a salmon fish inside an fMRI machine.

That is exactly what Dr Craig Bennett did.

He went to the nearest fish market and bought a full salmon. Why? For research purposes. Why else?

The salmon, let’s call it Richard, was not supposed to be anything more than a test for the machine, to ensure it was working well.

Instead, Richard changed how neuroimaging was used across the world. But before we get into the heroic tale of Richard; here is some backstory on how an fMRI works exactly.

fMRI: The Polaroid For The Brain

fMRI machines work pretty simply. They detect the levels of oxygen in blood in various parts of the brain. Oxygen is our fuel. If some area is consuming more oxygen, that means it is working more than the other areas of the brain.

The brain is divided into various parts through a grid and each section is called a ‘voxel’. The fMRI machine measures the amount of oxygen in each voxel.

But there is a problem.

Our brain cells, much like other cells, cannot live without oxygen. So at all times, there is some degree of oxygen in every voxel. How do we know which voxel is actually consuming a lot more oxygen than another?

Comparing numbers? Sounds like the task for the greatest fear strength of each psychology student, STATISTICS!

Statistical analyses are carried out by computer software to analyse which voxel has significantly higher levels of oxygen and which ones don’t. This means thousands of statistical analyses are carried out for each image in the fMRI.

At its core, fMRI combines physiology and statistics to give us an idea about the mechanisms of the brain.

Enter Richard, The Salmon King

When Craig Bennett stuck our beloved Richard inside the fMRI machine, I doubt he could have known what he was about to find.

Richard is described as 18 inches long, weighing 3.8lbs and in a state of unaliveness at the time of the test. After being put under the machine, Richard was shown images of humans in social interactions. Richard’s task was to guess the emotion the people in the photos might have been experiencing.

The whole scan lasted 15 minutes.

Now you might say, Arjun, have you gone nuts? How can a dead fish guess the emotions of humans?

To you I say, how dare you question the ways of science dear sir/madam?

The results had spoken. Richard, our dear dead fish showed increased activity in the spinal cord and hindbrain region upon being shown the images.

Maybe Richard was confused. Human emotions can confuse even the best of us.

The study was published as a poster which made people really question things. Can dead fish think? Is there something wrong with how we do fMRIs? Is this proof of life after death?

Metaphysics apart, it did make people think about how data is analysed for fMRIs.

Was Richard Talking From The Beyond?

No.

He was not. He was dead. Permanently.

But the fMRI still picked up some activity because when you compare so many voxels with each other, you WILL eventually get a positive result just by chance!

When 130,000 things are being compared to each other individually, the scope for a false positive increase with each comparison. This requires a statistical correction known as ‘multiple comparisons correction’

Bennett and his colleagues did not use the correction to make the point that not using it will eventually lead to absurd results. Such as a dead fish being confused by human emotions.

Before the study was published, about 30-40% of studies using fMRI did not use this correction. Since then, the number has dropped to below 10%.

The salmon study was not meant to bash the use of fMRI to understand mental states. It was supposed to make people more careful about analysing the data we collect from fMRIs and contextualising them the right way.

Richard may not have had active participation in this study. It could be argued that he was studied against his will. Nevertheless, we shall remember his contributions to the field of neuroscience forever. (Hopefully)

Don’t let his sacrifice be in vain. Use multiple comparisons correction when you can.

---

Hey hey everyone! Sorry about the radio silence for the last couple of weeks. I had to take some time off to focus on my mental wellness. I had been running on empty for quite a while but now I am back with a lot more stories to tell!

Until next time,

Arjun

🛣🖊🎯