Motor cortex and the systems BCIs target

You are skimming: the title, the first figure, and the short version. Switch to Read in the header for the full page, or Deep to open every deep dive.

Primary motor cortex

A strip along the precentral gyrus, just in front of the central sulcus, laid out as a distorted body map: legs at the top near the midline, hand and arm on the lateral surface, face and tongue lower down. Its large layer 5 pyramidal neurons send axons all the way to the spinal cord, and their firing is the most direct cortical correlate of movement. The hand region is a knob-shaped fold you can see on an MRI, and it is where BrainGate and most other human arrays have gone, because hand movement is what people with paralysis want back, because the area is superficial and accessible, and because its neurons are tuned to the direction and speed of intended movement in a way that has decoded well since the population vector.

Tuning here is broad and mixed: a neuron cares about direction, speed, force, posture, and the target, in proportions that shift with context. That mixture is why linear decoders work moderately and why the manifold view has been productive.

Premotor and supplementary motor areas

In front of primary motor cortex. Dorsal premotor cortex plans reaches, especially their direction and target, before movement begins; its activity during a delay period predicts the coming movement, which is useful for a decoder that wants to act early. Ventral premotor cortex, lower down, is involved in grasping and in the mouth and face. The supplementary motor area, on the midline, is about sequences and self-initiated movement.

Posterior parietal cortex

Behind the central sulcus, where vision meets movement. It encodes goals and intended endpoints rather than trajectories, in the coordinates of eye and hand. The Caltech group has implanted here in humans and decoded intended reach targets and even imagined speech, showing that “intention” in a fairly abstract sense is readable. Parietal signals are less dependent on the exact movement, which could make them more stable, and more abstract, which makes them harder to turn into continuous control.

The speech motor area

Speech is movement: lips, tongue, jaw, larynx. The part of ventral precentral gyrus and adjacent ventral premotor cortex that drives those articulators (area 6v in the current shorthand) is where the 2023 speech BCIs put their arrays. The neurons there encode which articulator is moving and how, and a decoder trained on a few hours of attempted speech maps that to phonemes and then, with a language model, to words at 60 to 80 words per minute. No thoughts are read; attempted speech movements are. That distinction is the whole ethical and technical story.

What the location decides

Signal type. Primary motor cortex: continuous kinematics, good for cursors and arms. Premotor: plans and targets, good for discrete selection. Parietal: goals, more abstract. Speech motor: articulators.

Stability. Superficial gyral surface is easier to implant and gives cleaner recordings; sulcal walls, where much of the area actually lies, are harder to reach with rigid arrays, which is one argument for flexible electrodes.

Redundancy. Hand knob is small; arrays of 100 channels sample perhaps a hundred neurons of a population of millions, and that has been enough, because the manifold is low-dimensional and redundantly represented.

Risk. Every implant site trades benefit against the surgery and the tissue response. The field has stayed with sites whose function is well understood and whose loss, should the implant damage them, would not add to what the person has already lost.