A few numbers can help to define the challenge. The human brain is thought to have close to 86 billion neurons, each making on average about 10,000 connections. In contrast to most animals, our brains are largely made up of a heavily folded cortex, accounting for 80 percent of brain mass and about 100,000 miles of axons that provide the highways between neurons.1How many different kinds of neurons are there in the brain? We really don’t know. Unlike the heart or kidney, which have a small, defined set of cell types, we still do not have a taxonomy of neurons, and neuroscientists still argue whether specific types of neurons are unique to humans. But there is no disputing that neurons are only about 10 percent of the cells in the human brain. Most of our brain cells are glial cells, once thought to be mere support cells, but now understood as having a critical role in brain function. Glial cells in the human brain are markedly different from glial cells in other brains, suggesting that they may be important in the evolution of brain function. As one hint to their function, astrocytes, which are one form of glial cell, have been reported recently to “eat” synapses in the brain, providing a critical new mechanism for brain plasticity.2
How does the brain work? Again, we really don’t know. We have a very detailed understanding of how the heart pumps and the kidney filters, but how the brain encodes, stores, and retrieves information is still largely a mystery. We have known for over a century that most of the cortex is organized horizontally into six precise layers, and much of the cortex has vertical mini-columns, but how this matrix of horizontal and vertical structures computes information is not really clear.
Neuroscientists talk a lot about brain circuits. In fact, the word “circuit” is probably misleading. We do not know where most circuits begin and end. And unlike an electrical circuit, brain connections are heavily reciprocal and recursive, so that a direction of information flow can be inferred but sometimes not proven. We believe there are “emergent properties” of the brain that convert electrical signals into memories or dreams, but how this happens is still a mystery. Recent studies have shown that diffuse waves of synchronization across the brain may be critical for attention or learning, but we are just learning about these slow waves of activity, and whether they occur at the “speed of thought” is still debated.3
Of course, the spectacular images from MRI and PET scans have already given us maps for perception and fear and language and many other functions. As scanners have improved their resolution from 1.5T (tesla) to 3T to recent 7T magnets, and the protocols and analytic approaches have evolved, we now can map the cortical real estate associated with complex tasks like decision-making and face recognition. But these approaches, even with the best current technology, are still a 30,000-foot view of the action. Jay Giedd here at NIMH estimates that each gray matter voxel—the individual 3D pixels of 1 cubic mm that make up the scan—contains about 90,000 neurons, 400 meters of dendrites, and 4.5 million synapses. Each scan has over 650,000 voxels. And the actual measure is not neural activity per se but local blood flow, which changes slowly relative to the speed of thought.