Good question. Ultimately, in fMRI we are looking for changes in the local ratio of oxygenated to deoxygenated blood. As neurons begin firing in a brain region more oxygen-rich blood is dumped in by the circulatory system. If the salmon really was responding to the pictures there would need to be blood flow for these changes to be detectable using fMRI. Since the salmon had been prepared by the supermarket there was nothing to circulate blood around.
TLDR; researchers at UC Santa Barbara, intending to show how statistics can be used misleadingly, received statistically significant results when scanning a salmon with fMRI to show that it's brain responded differently to photos of different social situations. The salmon was dead.
This belongs to "pure gold dust" category of my neurofeedback-related readings. Makes a good point about the prevalence of data dredging in scientific (and esp. medical) communities.
As we're talking about serious (though not entirely un-ludicrous) experiments concerning dead fish, I'm not going to miss an opportunity to post this video of a dead fish placed in "fast flowing water" (it's actually a tank with a jet and an outlet, but the effect is much the same).
Which way do you think the dead fish will move, upstream or downstream? Place your bets now.
