Science Fail Monday: How a dead salmon taught us about statistics
Any scientist knows the importance of a good negative control. A negative control in an experiment is a group of samples or subjects in which no response is expected to an experimental treatment. The experimental group can then be compared to the control group. Such negative controls are gold standards in science and are supposed to provide confidence in experimental results. However, occasionally, a negative control gives unexpected and hilarious results worth of an Ig-Nobel Prize, the highest honor for scientists who publish the silliest research. Such was the case in an experiment involving fMRI, human emotions, and an Atlantic salmon.
fMRI stands for functional magnetic resonance imaging. If you’ve ever had a knee injury or a concussion, you have likely experienced a normal MRI scan, which uses radio waves and a magnet to take a structural picture of the organ of interest. The “functional” in fMRI means that researchers can use MRI images to measure brain activity and take a snapshot of changes over time. When a strong magnet is turned on over the brain, the hydrogen atoms in all of the water molecules in the blood point in the same direction, like a compass needle next to a refrigerator magnet. When the magnet is turned off, the hydrogen atoms relax back to their original positions, which releases a signal. This signal changes based on how much oxygen is in the blood, so the end result is a picture of the brain with information about which regions have more oxygenated blood. Regions needing more oxygen are generally assumed to be more active. Researchers can even have study participants perform a task during an fMRI scan, such as viewing particular images or listening to music, and use the fMRI data to determine which areas of the brain are active during the task. These types of studies can tell us a lot about which brain regions are involved in everything from social situations to processing fear.
In the Ig-Nobel Prize-worthy experiment, researchers wanted to use fMRI to determine which parts of the brain were active in response to seeing human faces displaying different emotions. However, they needed a negative control for their human subjects just to make sure that any brain activity they saw in response to the faces wasn’t just due to chance. The ideal candidate for such a negative control? A four-pound Atlantic salmon, purchased by one of the researchers at the local fish market.
The researchers put the dead salmon in their fMRI scanner and, for the sake of science, asked it what emotions it thought the humans were displaying in pictures flashed up on the screen in the scanner. The authors do not comment on the salmon’s responses, but it can be assumed that the salmon was not a model experimental participant and did not comply with the study directions. Expecting to see nothing, the authors analyzed the fMRI signal in the salmon’s brain before and after the salmon “saw” the photos of the faces. Imagine the shock in the room when a few spots in the salmon’s itty-bitty brain lit up like a Christmas tree, suggesting that it was thinking about the faces it saw. Duh duh duuuuuhh….zombie salmon?
Obviously, the salmon was not alive, nor was it thinking about the emotional state of humans. Luckily for the field of fMRI, instead of publishing a paper telling everyone they should use dead salmon to study human response to the emotions of others, the authors of this study delved deeper into why they were seeing “brain activity” in this very dead fish. In their original data, the researchers failed to correct for multiple comparisons: basically, because you are comparing so many brain regions to so many other brain regions, you’re much more likely to find a spot with significant activity in fMRI purely by chance (for more info on multiple comparisons, click here). The authors applied the appropriate statistical corrections to their data, and voila, no more zombie salmon. And then, because scientists have a funny sense of humor, they wrote up and published these results as a lesson to all on the importance of having a good statistician.
Peer edited by Claire Gyorke.
Follow us on social media and never miss an article: