Estimates for the human brain range between 10 billion and 1 trillion. The imprecision in these estimates is due almost entirely to uncertainty about the number of granule cells in the cerebellum, a problem that can be traced back to a study by Braitenberg & Atwood (1958). More recent work by Lange makes a reasonably accurate estimate possible: The average human brain (1350 gm) contains about 85 billion neurons. Of these, 12 to 15 billion are telencephalic neurons , 70 billion are cerebellar granule cells and fewer than 1 billion are brainstem and spinal neurons.
In a beautiful quantitative analysis of human cortex using the optical disector, Pakkenberg and Gundersen have shown that the number of neocortical neurons ranges from 15 to 31 billion and averages about 21 billion. Other forebrain structures—primarily the hippocampal region, basal ganglia, and thalamus—are likely to contain an additional 5–8 billion neurons. Total neuron number in humans therefore probably averages 95–100 billion. What is perhaps more remarkable is the normal two-fold difference in neocortical neuron number among healthy adults of normal intelligence.
The number of neurons and their relative abundance in different parts of the brain is a determinant of neural function and, consequently, of behavior. Phyla whose members have larger brains and more neurons respond to environmental change with a greater range and versatility of behavior . Orders of mammals with big brains, such as cetaceans and primates, are more clever than those with little brains, such as insectivores and marsupials. However, the correlation breaks down as we narrow the focus and compare allied species and even individuals within species—the exceptions obscure any trend. There is no generally valid equation that relates neuron number to behavioral complexity. For instance, humans of normal intelligence may have brains that weigh only half the average of 1350 gm, and in which there is no evidence for any compensatory increase in neuron density, can have normal intelligence.
Despite this reservation, there is some experimental evidence of a relation between neuron number and intelligence. Triploid and tetraploid newts have the same brain mass as diploid controls, but their neurons are larger and there are only 50-70% of the normal number . Although the locomotion of these polyploid newts is indistinguishable from normal, they take two to three times as many trials to learn a maze as do normal newts. Thus, a reduction in neuron number in this case lowers performance markedly. Seasonal oscillations in neuron number in the song nuclei of canaries correlates well with singing ability ,this observation lends further support to a notion that many of us have been willing to take on faith.