← Reference library

REFERENCE LIBRARY

Brain Maps & ParcellationBrain Maps

"Prefrontal cortex," "superior temporal sulcus," "V1" — these sound like street addresses, as if there were a real border in the brain separating them. In fact the border was drawn by people, and it moves when you change how you draw it.

One sheet of cortex, three ways to cut it

The surface of the brain is a wrinkled sheet about 2.5 mm thick, called the cerebral cortex, packed with roughly 16 billion neurons (out of about 86 billion in the whole brain — most of the rest are in the cerebellum). Seen from outside, that sheet is continuous: no dotted lines, no seams. So "parcellation" is, from beginning to end, something humans do: pick a property you can measure, find where it changes abruptly, and cut there.

Historically three properties have been used, and they cut at wildly different grain:

① Lobes

The coarsest cut, based on the sulci and gyri — the folds you can see with the naked eye: frontal, parietal, temporal, occipital, plus the insula hidden deep in the lateral sulcus. The upside is that anyone can point them out; the downside is that the folds sit differently in every person, and cortex on either side of one sulcus may well be doing the same job.

② Brodmann areas

In 1909 Korbinian Brodmann sliced and stained the cortex and looked at how the neurons stack up in layers (called cytoarchitecture), drawing a border wherever the stacking changed. He numbered them 1–52. This is still the most-cited scheme going — BA17 is primary visual cortex, BA4 primary motor, BA44/45 roughly Broca's area. It has lasted a century because how cells are arranged really does tend to track what a patch of cortex does.

③ HCP multimodal parcellation

In 2016 Glasser and colleagues used Human Connectome Project data to look at four properties at once — cortical thickness and myelin content, which patches light up during tasks, who they synchronize with at rest, and the spatial gradients of all of these — and only counted a border where several changed together. The result: 180 areas per hemisphere (360 in total), 97 of which had never been named on their own before.

Coordinate spaces (MNI / Talairach)

A completely different way of locating things: forget parcels, just stretch every brain onto one standard template and report three numbers (x, y, z). "Activation at (−42, 18, 4)" in a paper means exactly this. It lets labs talk to each other, at the price of pretending every brain is the same shape.

Why the line is so hard to draw

Three very real problems, stacked on top of each other:

Different measurements give different answers. Draw a border by cell layering, another by what connects to what, another by what activates together — and the three often land millimetres apart, sometimes running in different directions entirely. The reason the HCP scheme insists on "four at once" is precisely the admission that any one of them is not enough.

Individual variation is larger than you'd think. The same functional area can sit a centimetre or two apart in two people. Averaging across a group smears it into a blur — especially for highly individual networks like language, which look "diffuse" on the group map while being perfectly sharp in each person, just in a different spot each time.

One name is not one area. Broca's area is the classic case: the name originally referred to the patch damaged in a single patient (Leborgne, 1861), and over a century different people have used it for different extents of cortex — and today we know language is far from confined to it. See → language network.

Counterintuitive: the map is for talking, not for locating

The easiest thing to get backwards: a parcellation is not a progressive approach to the anatomical truth of the brain. It's an accounting system — it lets what I saw over here be reconciled with what you saw over there. So "this activation is in BA10" really means "it falls in box number 10 of a grid drawn from stains a hundred years ago," not "there is an organ in the brain called number 10."

Which is also why parcellations keep getting revised: not because our predecessors measured wrong, but because another measurable property became available. The count went from 52 to 360 and will go higher — what's growing isn't the brain, it's the number of dimensions we can see it in.