In 2005 Craig Bennett, then a first year graduate student at Dartmouth College in New Hampshire, carried out an equipment test that inadvertently revealed how it might be possible to read just about anything into a brain scan. He and a colleague tried to find the most unusual objects they could fit inside a functional magnetic resonance imaging machine, to help calibrate it before their serious scientific work began. It was a joke that started with a pumpkin and ended with a dead, eighteen-inch-long, mature Atlantic salmon wrapped in plastic.
