Chapter I
From Cells to Circuits
By the 1950s neuroanatomy had mapped the brain's regions, and electrophysiology could record the impulses of a single fibre. Joining the two meant asking what the neurons of a particular region respond to, and what goes wrong when the region is lost. The answers came first from a patient.
In 1953 William Scoville, a surgeon in Hartford, Connecticut, removed the inner parts of both temporal lobes of Henry Molaison, known in the literature as H.M., to relieve his epilepsy. The operation took out much of the hippocampus on both sides. Afterwards Molaison could hold a conversation and remembered his childhood, but anything new was lost within minutes of his attention moving elsewhere. Brenda Milner of the Montreal Neurological Institute studied him from 1955. She found that he could learn a new skill, drawing while watching his hand in a mirror, improving day by day, while having no memory of having practised. Memory is not one thing, and the hippocampus is needed to lay down new memories of facts and events. Molaison was studied until his death in 2008.
Chapter II
What the Eye Tells the Brain
In 1958, at Johns Hopkins, David Hubel and Torsten Wiesel were recording from single neurons in the visual cortex of anaesthetised cats. Spots of light, which excite cells in the retina, did little. Then one neuron fired a burst as they slid a glass slide into the projector, and its faint edge swept across the screen. The cell responded to a line at one particular angle. Over the next years they found that the primary visual cortex breaks the image into short edges of every orientation, with neighbouring cells sharing preferences, and that later stages combine them. They also found that a kitten with one eye closed during an early critical period never recovers vision in that eye, because its connections in the cortex are lost.
Chapter III
Maps of Space
In 1971 John O'Keefe and Jonathan Dostrovsky at University College London recorded from the hippocampus of rats running freely and found place cells, each firing only when the rat was in one part of its enclosure. The hippocampus that H.M. had lost seemed to hold a map. In 2005 May-Britt Moser and Edvard Moser found the coordinates behind it. Grid cells in the entorhinal cortex fire at many places, arranged in a triangular lattice across the floor, like the points of a honeycomb pattern. Grids of different spacings together can specify a position much as digits specify a number. Other cells signal the direction the head faces and the distance to walls.
Recording electrodes see only a few neurons at a time. In 1990 Seiji Ogawa found that magnetic resonance imaging can detect the changes in blood oxygen that follow neural activity, and in 1992 several groups used it to watch the working human brain. Functional MRI showed regions for faces, places and words, and it became the main tool for studying the human brain in action.
Chapter IV
A Closer Look: What a Voxel Sees
A functional MRI scan divides the brain into small cubes called voxels, typically 1 to 3 millimetres on a side, and measures the blood-oxygen signal in each. How much brain is in one voxel?
The human cerebral cortex holds about 16.3 billion neurons in roughly 500 cubic centimetres of grey matter, or 500,000 cubic millimetres. On average that is
| Voxel side | Volume | Neurons (approx.) |
|---|---|---|
| 1 mm | 1 mm³ | 33,000 |
| 2 mm | 8 mm³ | 260,000 |
| 3 mm | 27 mm³ | 880,000 |
With about synapses in the same volume, a 3 mm voxel also holds about billion synapses. A single signal from a voxel is an average over nearly a million neurons, which may include cells doing opposite things.
Time is blurred as well. A nerve impulse lasts about a millisecond. The blood-flow response it triggers peaks about five seconds later, some times longer, and a typical scanner takes one image of the whole brain every two seconds. A ten-minute scan yields only 300 images.
This is why the field uses several methods at once. Hubel and Wiesel's electrodes record one neuron at a time with millisecond precision, but see almost nothing of the whole. MRI sees the whole brain, in blocks of hundreds of thousands of neurons and seconds of time. Modern silicon probes record from hundreds of neurons at once, and calcium imaging with glowing proteins from cell biology watches thousands, filling part of the gap.
Chapter V
Cause and Effect
Recording shows which neurons are active during a behaviour, not whether they cause it. In 2002 Gero Miesenböck made neurons fire with light using proteins from the fly's eye. Georg Nagel and Peter Hegemann had found channelrhodopsin, a light-gated channel from a green alga, and in 2005 Karl Deisseroth and Edward Boyden put it into mammalian neurons. Flashes of blue light made them fire, one impulse per flash. With genetic targeting, researchers could activate or silence one cell type in one region of a behaving mouse, and in 2012 reactivate neurons that had been tagged while a memory formed. Theories of how such circuits compute became computational neuroscience. Why any of this activity is accompanied by conscious experience is still unexplained.