Neuroscience students take it for granted that there are many more glia than neurons in the brain. Neuroscience textbooks state with confidence: “Although there are many neurons in the human brain…, glia outnumber neurons by tenfold” (Bear et al. 2006) or, not to be outdone, even by “10–50 times”, as claimed in another text (Kandel et al. 2000). This fact is happily invoked by gliologists to promote the status of their field.

Given this well-accepted figure, we were surprised when our cell counts in the prefrontal cortex of the rhesus monkey turned up a glia-to-neuron ratio (GNR) of just about 1 (Dombrowski et al. 2001). There was some regional variation, but no prefrontal area had a GNR larger than 1.2. Maybe the proportion of glia is very different in other cortical regions or other parts of the brain, so that the overall ratio for the whole brain is much larger than 1? Classic studies, however, conducted by O’Kusky and Colonnier (1982) in the opposite pole of the brain, the visual cortex, had reported an even lower GNR of 0.5. More recently, Lidow and Song (2001) estimated the GNR across a whole cortical hemisphere of healthy rhesus monkeys to be 0.82. And another recent study of neuron and glia numbers in various frontal, parietal, temporal and occipital cortical regions in the rhesus monkey also found a low average GNR of 0.56 (Christensen et al. 2007). However, the same study also showed an average GNR of 1.72 across hippocampal regions CA1 and CA2-3.

All these estimates were based on stereologic cell counting methods (West 1999), which are considered reliable and are now widely used in neuroscience. The methods are based on the simple and fundamental principle of unbiased counting, with each cell having an equal chance of being counted once and only once. However, the approach requires sampling from isotropic cellular populations, which makes it difficult, or at least labor intensive, to apply across many different cortical regions or the whole brain. Therefore, it is conceivable that the stereological approach may have missed contributions from structures with high concentrations of glia, such as subcortical nuclei or the white matter. In the spinal cord of the rat, for example, the GNR is about 1.6 (Bjugn and Gundersen 1993). The problem of overlooking the true contribution of glia to the neuraxis was circumvented by a straightforward approach for large-scale tissue analysis that was recently used by Herculano-Houzel and Lent (2005). Instead of counting cells in fixed tissue, they homogenized the brain and then counted cellular nuclei in the isotropic suspension. This approach gives a simple and reliable estimate of total cell number, assuming that each cell has exactly one nucleus. Neuronal nuclei were identified by NeuN staining which labels neurons but not glia. The remaining unlabeled cells, which included glia but also other non-neuronal cells, provided an upper estimate on the number of glia in the brain.

With the help of this elegant approach Herculano-Houzel and Lent determined total cell numbers in different large-scale brain regions as well as the whole brain. They estimated that of all cells in the adult rat brain, 60% (or about 200 million) are neurons. The ratio of all non-neuronal cells to neurons across different primate species was also found to be roughly constant, and not much larger than 1 (Herculano-Houzel et al. 2007). Measures for the human brain may yet turn up a higher ratio, but it would be a real surprise to see a fundamentally different proportion of glia, for instance one that was tenfold larger. After all, estimates for the GNR in human cortical tissue (1.7 for healthy young males; Pakkenberg and Gundersen 1997; Pakkenberg et al. 2003) are in line with those of the cortex in other primate species.

The above evidence suggests that across different brain tissues and different species the ratio of glia to neurons is approximately 1. This finding prompts the question, so far unresolved: why are neurons and glia roughly matched in number? Potentially, this relation has its origin in development, through intertwined mechanisms for the generation of neurons from glial progenitor cells (Noctor et al. 2007). In addition, the balanced ratio might also come from competitive adjustments in the number of neurons and glia that are related to their activity. In any case, a better understanding of the glia-to-neuron balance is likely to shed light on fundamental processes of brain development and function.

Glia have diverse and essential functions, although in the past they were often relegated to secondary supporting roles for their more interesting neuron siblings. Such basic glial functions include buffering transmitters, insulating axons, and generally filling brain space, as already suggested by their name (‘glue’). But recently, glia have become much more exciting, as they have been implicated in a wide range of brain processes, such as the regulation of synaptic strength, modulation of cellular signals and neurotransmitters, synaptogenesis, neurogenesis, as well as metabolic signaling cascades, which are important for neuroimaging approaches (Nedergaard et al. 2003). It even seems that there are some traditionally “neural” functions, such as selective tuning for the orientation of visual stimuli, that glia can perform better than neurons (Schummers et al. 2008).

Many of the putative functions of glia point to a strong association with synapses. Since the number of synapses increases faster than the number of neurons in larger brains (Striedter 2005), this affiliation of glia with the multitude of neural connection points may help explain why the ratio of some glia types apparently increases with brain size. For example, in large brains such as the human brain, with an estimated number of more than 10E14 synapses (Pakkenberg et al. 2003), there may be as many as 1.4 astrocytes for each neuron, up from 0.33 in the rodent cortex (Nedergaard et al. 2003). Even that ratio, however, is still a long way from the myth of 10 times more glia than neurons, in any species.