Korva is a scroll of the internet’s best ideas.
Averaged brains look like nobody's
How averaging misleads
The standard way to make claims about the human brain is to assume the same regions do the same things in everybody, then average the activity at each location across a whole group. The math works and returns an answer. That answer may reveal things about the species, yet it may not resemble a single person in the study.
A 2009 paper by an MIT postdoc and her adviser argued that averaging distorts in two ways. It can lump together functions that are really distinct, and it can miss functions that are there but sit in different places in different people.
Picture two people who both have a patch of cortex that handles word order, sitting a few millimeters apart. Averaged together, the patches blur, and the shared structure of the network gets lost along with the location.
Mapping each person
The alternative is to map each person's own anatomy first and compare people afterward. A recent preprint did exactly that in almost 1,200 people, some scanned more than once, which made almost 2,000 scanning sessions. Each session was treated separately, so every brain got its own language map.
Why does group averaging blur brain maps?
Brains differ in the folding pattern of the cortex, so the same spot in a standard coordinate system lands on different tissue in different people. When two people's active regions sit in slightly different places, averaging spreads the signal across both locations and weakens the peak at each. The result is a smooth blob that can sit between the real regions without matching either one. Even identical twins do not have identical cortical folds.
How are brains lined up across people?
Scans are warped to fit a shared template so that locations can be compared across people. A widely used template, MNI152, was built by averaging the MRI scans of 152 people. The warping lines up large landmarks such as major folds, but it cannot guarantee that small functional regions line up as well. Two voxels with the same coordinates can therefore belong to different functional areas in different people.
Why scan the same person more than once?
The fMRI signal is noisy, and the change it tracks is only a percent or two of the baseline at typical scanner strengths. Repeated sessions let researchers check whether a person's map comes out the same each time. A map that holds up across sessions is evidence that it reflects the individual's brain rather than random noise. Studies of this kind have found that individual functional maps are stable over weeks or months.
More from Wu Tsai Neurosciences Institute, Stanford
