European Journal of Neuroscience

European Journal of Neuroscience, Vol. 35, pp. 1–9, 2012 doi:10.1111/j.1460-9568.2011.07923.x

REVIEW
How many neurons do you have? Some dogmas of
quantitative neuroscience under revision

Roberto Lent,1,2 Frederico A. C. Azevedo,1,* Carlos H. Andrade-Moraes1 and Ana V. O. Pinto1

1
Instituto de Ciências Biomédicas (ICB), Centro de Ciências da Saúde Bl. F, Universidade Federal do Rio de Janeiro (UFRJ), CEP
21941-902, Rio de Janeiro, Brazil
2
Instituto Nacional de Neurociência Translacional (INNT), Brazil

Keywords: brain development, brain evolution, glial cell number, neuron number

Abstract
Owing to methodological shortcomings and a certain conservatism that consolidates wrong assumptions in the literature, some
dogmas have become established and reproduced in papers and textbooks, derived from quantitative features of the brain. The first
dogma states that the cerebral cortex is the pinnacle of brain evolution – based on the observations that its volume is greater in more
‘intelligent’ species, and that cortical surface area grows more than any other brain region, to reach the largest proportion in higher
primates and humans. The second dogma claims that the human brain contains 100 billion neurons, plus 10-fold more glial cells.
These round numbers have become widely adopted, although data provided by different authors have led to a broad range of 75–
125 billion neurons in the whole brain. The third dogma derives from the second, and states that our brain is structurally special, an
outlier as compared with other primates. Being so large and convoluted, it is a special construct of nature, unrelated to evolutionary
scaling. Finally, the fourth dogma appeared as a tentative explanation for the considerable growth of the brain throughout
development and evolution – being modular in structure, the brain (and particularly the cerebral cortex) grows by tangential addition
of modules that are uniform in neuronal composition. In this review, we sought to examine and challenge these four dogmas, and
propose other interpretations or simply their replacement with alternative views.

Introduction: Four dogmas of quantitative neuroscience

Quantitative neuroscience should not, perhaps, be properly defined as common relationship with the evolution and development of the
a discipline, within the broad field of the neural sciences, as the whole nervous system. They are probably not the only ones, but are perhaps
field employs quantitative methods to investigate the morphology, those most commonly quoted in the literature.
physiology, cell biology and molecular interactions in the nervous The first dogma states that the cerebral cortex is the highest
system. It should rather be considered as an important, normative achievement of brain evolution, on the basis of the observations that
collection of quantitative methods and data that can be used by brain volume is greater in more ‘intelligent’ species, and that cortical
neuroscientists to approach evolutionary and developmental issues, surface area seems to grow more than any other brain region, to reach
and to ask new questions about these and other aspects of brain the largest proportion in higher primates and humans. According to
structure and function. this view, the traditional concept of evolutionary encephalization
However, over time, because of methodological limitations (Box 1) could actually be translated as ‘corticalization’, to express the
and strong adherence to tradition, some beliefs derived from volumetric predominance of the cerebral cortex as the brain has
quantitative features of the nervous system have been extensively evolved in vertebrates. The second dogma targets the human brain and
reproduced in papers and textbooks, becoming undisputed dogmas of its computational ‘units’ – our brain would contain 100 billion
neuroscience. For instance, if asked how many neurons the human neurons, plus about 10 times more glial cells. These round, ‘magic’
brain has, the reader will probably answer – 100 billion. However, this numbers have become largely adopted in the literature, despite the fact
figure is not soundly supported by empirical evidence, and its origin in that estimates by different authors vary as much as 10-fold in the
the literature is unknown. cortex and 1.5-fold in the cerebellum, leading to a broad discrepancy
In this review, we analyze and question four dogmas of quantitative range of 75–125 billion neurons in the whole brain (a 50% range
neuroscience, and propose new interpretations or alternative views. around the adopted number). As a sort of corollary, the third dogma
These dogmas were chosen and brought together because they have a states that the human brain is structurally and functionally special, and
exquisitely sophisticated as compared with those of other primates. Its
Correspondence: R. Lent, as above. large and highly convoluted structure would give it the status of a
E-mail: rlent@icb.ufrj.br unique construct of evolution, somehow unrelated to comparative
scaling. Such a concept places the human brain apart from mammalian
*Present address: Max Planck Institute for Biological Cybernetics, Tubingen, Germany. evolutionary rules, in contradiction to most, if not all, empirical
Received 22 May 2011, revised 18 September 2011, accepted 27 September 2011 evidence. Finally, the fourth dogma connects evolution with

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
2 R. Lent et al.

development to explain the considerable growth of the brain along First dogma – the cerebral cortex is the highest
both ontogenetic and phylogenetic timelines – the existence of achievement of brain evolution
structural modules in the brain (and particularly in the cerebral cortex)
‘The cerebral cortex holds two-thirds of the brain’s neurons and
allows size increases simply by the addition of modules that would be
thus appears to be a promising candidate for determining the
uniform in neuronal composition.
primary neuroanatomical correlates of intelligence’ (Luders et al.,
2009)
Box 1. Simple methods, complex science Most theories on brain evolution and development assume that the
Most of the mentioned dogmas derive from gross volume and ⁄ or size of the brain or the sizes of particular regions are sufficient
weight measurements of the brain or its parts, and from regional cell correlates of cognitive abilities (Hofman, 1985; Luders et al., 2009),
counts in brain sections. Gross measurements, however, albeit simple although neurons, and more recently glial cells, have been recognized,
and straightforward, tell us very little about the functional abilities of in essence, to be the minimal computational units of nervous systems
brains. This led to measurements designed to quantify the compu- (Williams & Herrup, 1988; Roth & Dicke, 2005).
tational ‘units’ of the nervous tissue – cells, especially neurons – and Indeed, there is a great wealth of evidence that neurons are involved
relate them to size, function, and other parameters. in different aspects of cognition, either because of their properties as
To count cells in the brain, however, is not a simple task, owing to
single units, or because of emergent functions deriving from their
the pronounced anisotropy of the nervous tissue. In a typical brain
association in circuits and networks that include glial cells (see
sector (Fig. 1A), cell density can vary from highly populated, cell-
Dehaene & Changeux, 2011; Sporns, 2011; and Wig et al., 2011 for
packed layers, to neuropil and white matter sectors, which are fiber-
rich and cell-scarce. The main methods used to estimate cell numbers
recent reviews). Perhaps two good examples are the ‘gnostic cells’,
are stereological techniques (reviewed by Schmitz & Hof, 2005) – cell discovered in the monkey inferotemporal cortex (Gross et al., 1972),
densities are quantified in sample sectors, and the average is multiplied and the mirror neurons of the premotor cortex (Rizzolatti et al., 1996).
by the measured volume or mass of the region or of the whole brain. Both have been reported in humans as well (Quiroga et al., 2005;
When the variance is large, however, the resulting absolute number Mukamel et al., 2010). The functional role of these cells in complex
becomes highly uncertain (von Bartheld, 2001; Benes & Lange, 2001; perception and other cognitive operations (reviewed by Gross, 2009;
Farel, 2002), which explains the broad range of results in the literature and Rizzolatti & Sinigaglia, 2008) was interpreted initially in a rather
(see main text for references). Stereological techniques have the reductionist way as self-sufficient (that is, they would be able
advantage of allowing quantification of more restricted cytoarchitec- individually to code complex stimuli), and has more recently been
tonic regions, such as cortical areas and subcortical nuclei, as well as interpreted as ensemble-dependent (that is, they would work cooper-
their subdivisions, e.g. layers. They remain powerful tools for atively in interconnected circuits), forming face-recognition and mirror
quantifying brain sectors of functional relevance. systems, respectively (Freiwald & Tsao, 2010; Heyes, 2010).
In order to increase the precision of high-scale measurements, a new Whatever one’s view on the way in which neurons operate, it is
method was proposed – the isotropic fractionator (Herculano-Houzel & undisputed that they play the most crucial role in determining the
Lent, 2005) – by which the anisotropic tissue is rendered isotropic functions of the nervous system, increasingly so for the larger brains
[compare (A) with (B) in Fig. 1], so that cell densities quantified from of mammals that display highly complex behavior, and that have
tissue samples yield very representative averages. The technique starts recently appeared in evolution. Possibly, in the future, new techniques
with homogenization of the tissue sufficient to rupture cell membranes that quantify absolute numbers of synapses, circuits and even multi-
but not nuclear membranes. A nuclear suspension is obtained (Fig. 1C),
circuit networks will be developed, thus allowing us to approach more
DNA-stained for better visualization (Fig. 1D), and immunostained to
closely the computational operations of the nervous system.
reveal a neuron-specific antigen (Fig. 1E), enabling discernment of
Nevertheless, despite the evidence for neuronal prevalence over
neuronal from non-neuronal nuclei. From the nuclear suspension,
aliquots are taken, placed into a hemocytometer, and counted under a
brain size, evolutionists have often correlated cognition (which equals
fluorescence microscope. Similar results have been obtained by running ‘intelligence’ in many authors’ words) with the absolute volume or
the suspension through a flow cytometer (Collins et al., 2010). mass of the brain as a whole or the relative proportions of its main
The isotropic fractionator has the advantage over stereology of sectors (Hofman, 1985; Luders et al., 2009), implying that larger brain
yielding less variable (and therefore more precise) absolute cell regions will necessarily contain more neurons. Furthermore, they have
numbers, especially for large brains or brain regions. However, it used these structural variables to derive scaling relationships that
depends on the possibility of standardizing the dissection of target brain would predict how complex a brain or brain region will become in
structures for different specimens, restricting its use mostly to evolution (Clark et al., 2001). Along these lines, the relative size of
anatomical regions, and usually not for their histological, cytoarchitec- the cerebral cortex reveals a clear increase (Fig. 2A, green dots) – in
tonic sectors. For instance, the cerebral cortex as a whole can be easily rodents, the cerebral cortex relative representation in the whole brain
dissected apart from subcortical regions in almost all brains of different goes from about 40% in mice to 60% in capybaras (Herculano-Houzel
species. For this reason, estimates of its global cell composition, as et al., 2006); in primates, it represents about 67% of the galago
obtained by use of the isotropic fractionator, are statistically precise and (Otolemur) brain (Herculano-Houzel et al., 2007), and approaches
reliable. In contrast, this is not so for each of the cortical layers, because 82% in humans (Azevedo et al., 2009). On the other hand, the relative
they cannot be perfectly separated by dissection from neighboring size of the cerebellum remains constant across phylogenetic groups,
layers. In this case, stereological methods are recommended. occupying about 10–15% of the total brain mass in different orders
Isotropic fractionation and stereology, in fact, are complementary (Hofman, 1988) (Fig. 2B, green dots).
techniques. The first is best applied to large, dissectable anisotropic On the basis of these relationships, the evolutionary predominance
regions (e.g. the cerebellum and the cerebral cortex), and the latter to
of the cerebral cortex became a dogma, and because its relative
small anisotropic brain sectors with imprecise borders (e.g. cytoarchi-
volumetric proportion in humans reaches the peak among mammals,
tectonic cortical areas), or to relatively isotropic regions of any
this brain region came to be considered as the pinnacle of evolution
dimension (e.g. individual subcortical nuclei).
(Luders et al., 2009). However, if neurons, glial cells and their

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
European Journal of Neuroscience, 35, 1–9
 Neuroscience dogmas and brain cell numbers 3

A B

D E

Fig. 1. It is possible to count absolute cell numbers in the brain, by use of the isotropic fractionator (Herculano-Houzel & Lent, 2005). (A) A sector of the human brain,
with its high anisotropy. Cell counts within the three red circles give largely different values. (B) If this same sector is rendered isotropic, cell counts within the same circles
give very similar values. (C) The brain can be rendered isotropic by fractionation in potters, to arrive at a nuclear suspension, aliquots of which can be counted under the
microscope in hemocytometers. (D and E) The same field, where some of the 4¢,6-diamidino-2-phenylindole (DAPI)-labeled nuclei do not appear to be labeled by NeuN
(yellow arrows) – these are non-neuronal nuclei.

networks are responsible for cognitive abilities, relying on brain cell Analysis of 19 species of four mammalian orders has recently
numbers to calculate relative proportions among brain regions and revealed that the absolute neuronal composition in the cortex covaries
define ‘cerebrotypes’ among taxa (Clark et al., 2001) would be an significantly with that of the cerebellum (Herculano-Houzel, 2010),
advance relative to relying simply on size. The picture that emerges showing that these two brain structures display coordinated growth
from this exercise tells us that it is the cerebellum that underwent during phylogenesis in mammals (Fig. 2C). According to these data,
greater growth in phylogeny (Fig. 2B, blue dots). In rodents, each neuron in the cerebral cortex corresponds to about four in the
cerebellar neurons represent about 60% of all brain neurons in the cerebellum, irrespective of the species considered.
small-brained mouse, increasing to about 70% in the capybara Such a coordinated evolution of the cerebral cortex and cerebellum fits
(Herculano-Houzel et al., 2006). In primates, cerebellar neurons well with the recent clinical and experimental evidence suggesting an
represent about 83% in the macaque (Herculano-Houzel et al., 2007) important role of the cerebellum in cognitive and affective functions, in
and about 80% in humans (Azevedo et al., 2009). close connection with cortical associative areas (reviewed by Schmah-

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
European Journal of Neuroscience, 35, 1–9
4 R. Lent et al.

A The internal circuitry of the cerebral cortex and that of the

cerebellum have been thoroughly studied for a long time (Eccles
et al., 1967; Ito, 1972), and display differences that certainly explain
their different functional strategies. Despite some controversy (e.g.
Horton & Adams, 2005; Da Costa & Martin, 2010), both are
considered to be organized in modules – neocortical modules
(columns, as they are often called) have been regarded either as
varying considerably (Da Costa & Martin, 2010) or being homoge-
neous and consistent in their cell composition and internal circuitry
(Douglas et al., 1989; Lund et al., 2003; Innocenti & Vercelli, 2010).
On the other hand, cerebellar modules (or microzones, as they are also
called) are considered to display a homogeneous circuit performing
similar types of computation that differ only in input and output
information (Andersson & Oscarsson, 1978; Apps & Hawkes, 2009;
B Dean et al., 2010). Synaptic plasticity in both the cerebral cortex
(Kirkwood & Bear, 1994; reviewed by Feldman, 2009) and the
cerebellum (Ito et al., 1982; Carey, 2011) provides the biological
substrate for their involvement in complex cognitive operations.
However, the developmental control of neuronal populations is
known to depend on functional interactions between connected regions
that influence neurogenesis and programmed cell death (Piñón &
Linden, 1996; Sherrard & Bower, 1998; Davies, 2003; Madalosso et al.,
2005). It is therefore conceivable that the coordinated regulation of cell
numbers between the cortex and cerebellum took place in evolution as a
consequence of the abundant connections that formed between both
structures, rather than resulting from their intrinsic differences, which
would tend to drive them into independent evolutionary trends.
It seems more appropriate, therefore, to question the view that the
C mammalian cerebral cortex was selected in evolution for greater
growth, as a result of growing environmental pressure for more
sophisticated abilities to orient behavior. Instead, coordinated evolu-
tion of both the neocortex and the cerebellum should be viewed as a
more realistic evolutionary ‘investment’ that resulted in the cognitive
computations of higher primates, including humans.

Second dogma – the human brain has one hundred billion

neurons and 10 times more glial cells

‘The mature brain is composed of 100 billion to 200 billion

neurons and perhaps 10 times as many glial cells’ (Hubel, 1979)
The ‘magic number’ of 100 billion neurons in the human brain has
been widely sustained in papers (Hubel, 1979; Fischbach, 1992; Noctor
et al., 2007) and textbooks (Kandel et al., 2000; Bear et al., 2007; Purves
et al., 2008), although a broad range is arbitrarily adopted, from 10 billion
to 1 trillion (reviewed by Soper & Rosenthal, 1988). However, little
Fig. 2. Concerted increase in the proportion of neurons between the cerebral direct evidence for it has been produced. In fact, stereological estimates
cortex and the cerebellum. (A and B) Whereas the proportion of cerebral cortex have yielded numbers of 3 billion, 7 billion, 14 billion, 19–23 billion,
mass increases more than that of the cerebellum in rodents (green dots in top 21–26 billion and 28–39 billion neurons for the cerebral cortex
and middle graphs), the proportion of neurons shows a greater increase in the
latter (blue dots). Data from Herculano-Houzel et al. (2006). (C) The absolute (Pakkenberg, 1966; Pakkenberg & Gundersen, 1997; more extensively
numbers of neurons in both structures, however, show a significant correlation reviewed by Azevedo et al., 2009). The same has been the case for
among many mammalian orders (bottom graph; data from Herculano-Houzel, the cerebellum, for which counts have produced numbers from 70 bil-
2010). ag, agouti; gp, guinea pig; ca, capybara; ha, hamster; mo, mouse; ra, rat. lion to 109 billion neurons (Lange, 1975; Andersen et al., 1992, 2003).
Statistical parameters: a = slope; r2 = regression; q = Spearman correlation
coefficient.
Using the isotropic fractionator, we contributed to reducing the
uncertainty of these numbers (Azevedo et al., 2009) – absolute counts
yielded an average of 86 billion neurons in male human brains
mann, 2010). Much as motor dysmetria is caused when the anterior lobe 50–70 years old (Fig. 3), about 15% less than the ‘magic number’.
of the cerebellum is damaged, dysmetria of thought appears when the Interestingly, it is noticeable that the cerebral cortex contains only
posterior lobe becomes lesioned (Schmahmann, 1998). Besides clinical, 19% of this total neuronal number, despite occupying 81% of the brain
physiological and neuroimaging data, the proposal is substantiated by mass. On the other hand, the cerebellum contains the impressive
findings of extensive connections between the associative cortex and proportion of 80% of all cerebral neurons, packed within only 10% of
cerebellum (Middleton & Strick, 1994; Schmahmann & Pandya, 1997). the total brain mass.

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
European Journal of Neuroscience, 35, 1–9
 Neuroscience dogmas and brain cell numbers 5

Fig. 3. Absolute cell composition of the human brain. Values represent mean ± standard deviation. Glia stands for non-neuronal cells, for simplicity. g ⁄ n,
glia ⁄ neuron ratio; neur, neurons; GM, gray matter; WM, white matter. Modified from Azevedo et al. (2009).

A precise determination of the actual number of neuronal cells in the axons (Chotard & Salecker, 2004). In adults, they play important roles in
adult human brain is important for many reasons. It stands as a synapse physiology (Eroglu & Barres, 2010), neurovascular interac-
reference figure for comparisons with diseased brains (e.g. Alzhei- tions (Nedergaard et al., 2003), immune mechanisms and circuit
mer’s), for a determination of the developmental oscillations until the stabilization (Graeber, 2010), signal conduction (Yamazaki et al.,
mature number is reached, for a better understanding of the quantitative 2010), fiber maintenance and regeneration (Nave, 2010), and higher
impact of aging in different brain regions, and for estimates of the information processing (Pereira & Furlan, 2010). As a whole, the
number of neurons in different hominin brains (Box 2). The ‘magic evidence indicates that glial cells do participate actively in the functional
number’ of 100 billion neurons has to be considered as an arbitrary computations performed by neurons, circuits, and networks.
midpoint between the broad range of estimates inferred from regional Given the increasing importance of these cells, it is crucial to tackle
stereological counts that should now be replaced with the experimental the issue of what determines their quantitative relation with neurons.
numbers as determined by use of the isotropic fractionator. Why is it that the cerebellum needs fewer glial cells than the cortex,
For glial cells, the prevalent dogma poses that the glia ⁄ neuron ratio even though it contains the majority of brain neurons? Is this an
is approximately 10 : 1 in the brain (Hubel, 1979; Nauta & Feirtag, indication that the glial cells therein are more supportive than
1986; Soper & Rosenthal, 1988; Nishiyama et al., 2005; reviewed by computational? On the other hand, could it be the case that glial cells
Hilgetag & Barbas, 2009). However, with the isotropic fractionator, the in the cerebral cortex acquired more sophisticated computational
actual glia ⁄ neuron ratio for the whole human brain was shown to be functions, thereby increasing in numbers to the extent of being more
close to 1 (Fig. 3) (Azevedo et al., 2009). The cerebellum, which has a numerous than neurons? There are no clear answers to these questions
huge neuronal population, contains a much lower proportion of glial – the next few years will probably clarify this issue.
cells, with a glia ⁄ neuron ratio of 0.23, whereas the cerebral cortex has a
higher number of glial cells, with a ratio of 1.48 for the gray matter Third dogma – the human brain is exceptionally complex,
alone, and a ratio of 3.76 for the associated gray and white matter. The as compared with those of other primates
remaining regions, altogether, have the highest glia ⁄ neuron ratio –
‘There is a long tradition that ascribes properties to humans that
11.35. Concerning the cerebral cortex, in particular, stereological
are supposedly not found in other animals … It is assumed that
methods provided similar (Dombrowski et al., 2001; Lidow & Song,
animals with larger brains are more intelligent than those with
2001) or even lower proportions in monkeys (O’Kusky & Colonnier,
smaller ones’ (Roth & Dicke, 2005)
1982; Christensen et al., 2007) and humans (Pakkenberg & Gunder-
sen, 1997; Pakkenberg et al., 2003; Pelvig et al., 2008). The belief that the human brain displays exceptional properties as
Unfortunately, so far there is no universal marker for the nuclei of compared with the brains of most other animals possibly derives,
specific glial types. For this reason, the absolute numbers and besides from our own anthropocentric tendencies, from brain–body
proportions of astrocytes, oligodendrocytes and microglial cells relationships as represented by encephalization quotients (EQs)
remain unknown. (Jerison, 1955, 1973; Marino, 1998) and other measures (reviewed
Glial cells cooperate with neurons in the proper function of the by Deaner et al., 2007). EQs are mathematical indices relating brain
nervous system. Their importance has increased recently, well beyond size with body size, and became widely used to compare animals of
the early conception of their role as a structural ‘glue’ for the tissue different orders and to draw evolutionary trends from this compar-
(reviewed by Kettenmann & Ransom, 2005). During development, glial ison. Under this logic, humans are positioned as outliers among all
cells act as stem cells (Kriegstein & Alvarez-Buylla, 2009) and as animals, displaying an unusually great EQ (Marino, 1998; Klein,
guidance scaffolds for migrating neurons (Rakic, 2003) and growing 2009), closely followed by some cetaceans and non-human primates.

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
European Journal of Neuroscience, 35, 1–9
6 R. Lent et al.

This line of thought explains the view that the human brain is seven growth in the number of neurons, maintaining both brain and body
times larger than expected for any mammal, and about 3.5 times size within dimensions more adapted to on-land locomotion.
larger than expected for an anthropoid primate of its body size We conclude that the human brain is not exceptional in the absolute
(Marino, 1998). composition of neurons and glial cells, the main operators of its
However, not only scaling rules may be different among orders computational functions. It is, rather, a result of the linear scaling rule
(indeed, they are – see below), but also body size may not represent characteristic of primates. We are not special in nature, but only big-
an appropriate parameter with which to explain the cognitive ⁄ affec- brained primates. Having big brains and being primates (Box 2), we
tive achievements of the different species. Absolute brain measures, have acquired a gigantic number of computational units that have
on the other hand, have been reported as being significantly correlated made us capable of superior cognitive performance (Deaner et al.,
with cognitive measures (Deaner et al., 2007). Additionally, when the 2007).
absolute number of neurons is employed to study the relationships
between brain evolution and development, the conclusion is that
orders, in fact, differ in scaling rules. Whereas neuronal number
Box 2. The long road to the current number of neurons
correlates with brain size on the basis of a power equation in
in humans
rodents (Herculano-Houzel et al., 2006), the same correlation is
linear for primates (Herculano-Houzel et al., 2007; Azevedo et al., The linear scaling law determined for primates allows one to estimate
2009). the number of neurons of human ancestors, using brain volumes as
The power equation of rodents means that, to reach the number of inferred from fossil cranial endocasts (Klein, 2009).
neurons that humans have, a rodent would need a brain weighing This exercise shows that ancestral primates living between 35 and
about 40 kg, in a body of about 100 tons! Humans of 75 kg, if built 20 million years ago – arboricole and quadruped – did not have more
than 20 billion neurons in their brains (Fig. 4). By the end of the
under rodent scaling rules, would have brains no bigger than 150 g,
Miocene period, between 7 million and 2 million years ago, neuronal
with only about 3 billion neurons. On the other hand, when the linear
numbers may have increased to about 40 billion in Ardipithecus and
equation of primates is applied to a human with a brain of about
Australopithecus, just above the estimated 30 billion neurons of
1500 g, one arrives at a figure of 90 billion neurons, very close to
chimpanzees. These hominins became bipedal, and produced the first
what has been found experimentally (Azevedo et al., 2009). It can be flaked stone tools. Another increase took place at the end of the
concluded, therefore, that a change in scaling from rodents to primates Pliocene, about 2 million years ago, with the appearance of the genus
(from allometric to isometric) has made it possible to achieve a great

Fig. 4. The number of neurons of human ancestors (left ordinates) can be derived from endocranial volumes (right ordinates; taken from Klein, 2009). Time (in
millions of years, at bottom) is segmented into paleontological periods (top). The different hominin species are represented as colored circles or bars; their main
cognitive achievements are shown on the right, and their main motor behaviors on the left.

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
European Journal of Neuroscience, 35, 1–9
 Neuroscience dogmas and brain cell numbers 7

Box 2. Continued under 1 mm2 at the surface. In addition, the prediction that the total
surface area of the cortex would increase linearly with the number of
Homo. The number of neurons in the brain grew to about 50 billion in neurons (provided that uniform columns were added across species)
Homo habilis, reaching about 70 billion in Homo erectus, and finally did not prove true – on the contrary, surface area was found to increase
our 86 billion. It is still a puzzle that, on the basis of the estimated more slowly than neuronal number (Herculano-Houzel et al., 2008).
brain volumes, Homo neanderthalensis would contain about 100 bil- Uniform modules would also require an inverse relationship between
lion neurons. density and thickness, but, in fact, these two variables did not show
With such a great number of neurons, bipedal locomotion
any correlation (Herculano-Houzel et al., 2008).
consolidated, and hand ⁄ finger movements acquired sophisticated
Questioning the presumed uniformity of cortical columns does not
abilities, which allowed Homo to produce more and more elaborate
imply questioning either the radial unit hypothesis or the concept that
tools, dominate fire, and improve social interactions.
the brain is organized in modules. In the course of development, the
number of neurons that will form each module in the mature cortex
will be necessarily attained by regulation of the number of cell cycles
Fourth dogma – brains grow in evolution and develop- within each radial unit in the germinative layers. Such regulation,
ment by the addition of uniform modules however, would be controlled and influenced by local variables of
‘In fact, one of the puzzling dogmas in comparative studies on the unknown nature, rather than being homogeneous across different
mammalian cerebral cortex is the constant number of neurons in cortical areas and species.
an arbitrary unit column’ (Abdel-Mannan et al., 2008)
Concluding remarks
A rising trend in neuroscience is to explain evolution by means of
Dogmas are fundamental to science (Kuhn, 1970). They serve the
changes in developmental mechanisms (often called the evo-devo
purpose of being challenged, and eventually replaced by other
approach). Under this rationale, brain growth during phylogenesis is
transient truths. Neuroscience is no different. In addition to the four
explained mainly by increases in the numbers of cell cycles of
dogmas examined in this review, others can be mentioned, for future
neuronal and glial precursors. The cerebral cortex, in particular, is a
analysis. Is it true that the number of neurons in the brain declines with
favorable structure with which to test this idea, not only because it
aging? To what extent would this be influenced by dementia? Do
grows considerably in evolution (about five orders of magnitude
males have higher number of neurons than females? The near future
across mammals), but also because this growth takes place mainly in
will perhaps reveal the answers to these and other questions. Other,
surface area rather than in thickness.
newer, dogmas will be generated for later challenge.
As the cortex is reportedly organized in modules (Douglas et al.,
1989; Lund et al., 2003; Da Costa & Martin, 2010; Innocenti &
Vercelli, 2010), as mentioned above, the prevalent view has been that Acknowledgements
precise developmental mechanisms control the number of modules
The authors’ research is supported by the Brazilian Council for the
that will later populate the cortical surface. A strong hypothesis has Development of Science and Technology (CNPq), the Rio de Janeiro
been put forward in this respect (Rakic, 1988), suggesting that the Foundation for Research (FAPERJ), and the Ministry of Education (CAPES-
ventricular zone of the embryonic cortex was composed of radial units MEC). The authors’ laboratory is a member of the Institute of Translational
that would then, by centrifugal migration along a radial glial scaffold, Neuroscience, Ministry of Science and Technology.
give rise to the cortical columns. According to this hypothesis, a
protomap of radial units would reside in the germinative layers, giving
References
rise to the adult cortical map of different cytoarchitectonic areas.
A cortical column, therefore, should be composed of a given Abdel-Mannan, O., Cheung, A.F.P. & Molnár, Z. (2008) Evolution of cortical
neurogenesis. Brain Res. Bull., 75, 398–404.
number of neurons, predetermined by the number of cell cycles Andersen, B.B., Korbo, L. & Pakkenberg, B. (1992) A quantitative study of the
undergone by ventricular precursors. Of course, neurons therein may human cerebellum with unbiased stereological techniques. J. Comp. Neurol.,
associate according to their functional and output properties to form 326, 549–560.
dendritic bundles (Vercelli et al., 2004), and afferents may cluster in Andersen, B.B., Gundersen, H.J.G. & Pakkenberg, B. (2003) Aging of the
human cerebellum: a stereological study. J. Comp. Neurol., 466, 356–365.
specific layers in consonance with the modular architecture (White & Andersson, G. & Oscarsson, O. (1978) Climbing fiber microzones in cerebellar
Peters, 1993; Sato et al., 2007). vermis and their projection to different groups of cells in the lateral vestibular
Investigation of the numerical composition of cortical columns nucleus. Exp. Brain Res., 32, 565–579.
(Rockel et al., 1980) has led to the dogma that a constant number of Apps, R. & Hawkes, R. (2009) Cerebellar cortical organization: a one-map
neurons per column compose any cortical area in any mammalian hypothesis. Nat. Rev. Neurosci., 10, 670–681.
Azevedo, F.A.C., Carvalho, R.B., Grinberg, L.T., Farfel, J.M., Ferretti, R.E.L.,
species (except for area 17 of primates). According to this view, a Leite, R.E.P., Jacob-Filho, W., Lent, R. & Herculano-Houzel, S. (2009)
constant number of about 147 000 neurons populate each cortical Equal numbers of neuronal and non-neuronal cells make the human brain an
cylinder of 1 mm2 at the surface. Despite the fact that different authors isometrically scaled-up primate brain. J. Comp. Neurol., 513, 532–541.
provided evidence in favor (Schüz & Palm, 1989) or against von Bartheld, C.S. (2001) Comparison of 2-D and 3-D counting: the need for
calibration and common sense. Trends Neurosci., 24, 504–506.
(Stolzenburg et al., 1989; Poth et al., 2005) this idea, it became Bear, M.F., Connors, B.W. & Paradiso, M.A., eds. (2007) Neuroscience.
rooted in the literature (Zhang & Sejnowski, 2000; Cheung et al., Exploring the Brain. Lippincott Williams & Wilkins, Baltimore, USA, p. 24.
2007; but see Rakic, 2009 for a critical appraisal). Benes, F.M. & Lange, N. (2001) Two-dimensional versus three-dimensional
This dogma has been approached by use of the isotropic fraction- cell counting: a practical perspective. Trends Neurosci., 24, 11–17.
ator, to avoid the pitfalls of an arbitrary, histological definition of the Carey, M.R. (2011) Synaptic mechanisms of sensorimotor learning in the
cerebellum. Curr. Opin. Neurobiol., 21, 1–7.
counting module, as was the case in the work by Rockel et al. (1980). Cheung, A.F., Pollen, A.A., Tavare, A., DeProto, J. & Molnár, Z. (2007)
A variation as great as 300% was found in primates (Herculano- Comparative aspects of cortical neurogenesis in vertebrates. J. Anat., 211,
Houzel et al., 2008) for the number of neurons within a unit volume 164–176.

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
European Journal of Neuroscience, 35, 1–9
8 R. Lent et al.

Chotard, C. & Salecker, I. (2004) Neurons and glia: team players in axon Ito, M., Sakurai, M. & Tongroach, P. (1982) Climbing fibre induced depression
guidance. Trends Neurosci., 27, 655–661. of both mossy fibre responsiveness and glutamate sensitivity of cerebellar
Christensen, J.R., Larsen, K.B., Lisanby, S.H., Scalia, J., Arango, V., Dwork, Purkinje cells. J. Physiol. (Lond.), 324, 113–134.
A.J. & Pakkenberg, B. (2007) Neocortical and hippocampal neuron and glial Jerison, H.J. (1955) Brain to body ratios and the evolution of intelligence.
cell numbers in the rhesus monkey. Anat. Rec., 290, 330–340. Science, 121, 447–449.
Clark, D.A., Mitra, P.P. & Wang, S.S.-H. (2001) Scalable architecture in Jerison, H.J. (1973) Evolution of the Brain and Intelligence. Academic Press,
mammalian brains. Nature, 411, 189–193. New York, USA.
Collins, C.E., Young, N.A., Flaherty, D.K., Airey, D.C. & Kaas, J.H. (2010) A Kandel, E.R., Schwartz, J.H. & Jessell, T.M., eds. (2000) Principles of Neural
rapid and reliable method of counting neurons and other cells in brain tissue: Science, 4th edn. McGraw-Hill, New York, USA, p. 19.
a comparison of flow cytometry and manual counting methods. Front. Kettenmann, H. & Ransom, B.R. (2005) The concept of neuroglia: a historical
Neuroanat., 4, 5. perspective. In Kettenmann, H. & Ransom, B.R. (Eds), Neuroglia, 2nd edn.
Da Costa, N.M. & Martin, K.A.C. (2010) Whose cortical column would that Oxford University Press, Oxford, pp. 1–8.
be? Front. Neuroanat., 4, 16. Kirkwood, A. & Bear, M.F. (1994) Hebbian synapses in visual cortex.
Davies, A.M. (2003) Regulation of neuronal survival and death by extracellular J. Neurosci., 14, 1634–1645.
signals during development. EMBO J., 22, 2537–2545. Klein, R.G. (2009) The Human Career, 3rd edn. Chicago University Press,
Dean, P., Porrill, J., Ekerot, C.-F. & Jörntell, H. (2010) The cerebellar Chicago, USA, p. 197.
microcircuit as an adaptive filter: experimental and computational evidence. Kriegstein, A. & Alvarez-Buylla, A. (2009) The glial nature of embryonic and
Nat. Rev. Neurosci., 11, 30–43. adult neural stem cells. Annu. Rev. Neurosci., 32, 149–184.
Deaner, R.O., Isler, K., Burkart, J. & van Schaik, C. (2007) Overall brain size, Kuhn, T.S. (1970) The Structure of Scientific Revolutions, 2nd edn. enlarged.
and not encephalization quotient, best predicts cognitive ability across non- University of Chicago Press, Chicago, USA.
human primates. Brain Behav. Evol., 70, 115–124. Lange, W. (1975) Cell number and cell density in the cerebellar cortex of man
Dehaene, S. & Changeux, J.P. (2011) Experimental and theoretical approaches and other mammals. Cell Tissue Res., 15, 115–124.
to conscious processing. Neuron, 70, 200–227. Lidow, M.S. & Song, Z.M. (2001) Primates exposed to cocaine in utero display
Dombrowski, S., Hilgetag, C.C. & Barbas, H. (2001) Quantitative architecture reduced density and number of cerebral cortical neurons. J. Comp. Neurol.,
distinguishes prefrontal cortical systems in the rhesus monkey. Cereb. 435, 263–275.
Cortex, 11, 975–988. Luders, E., Narr, K.L., Thompson, P.M. & Toga, A.W. (2009) Neuroanatom-
Douglas, R.J., Martin, K.A. & Whitteridge, D. (1989) A canonical microcircuit ical correlates of intelligence. Intelligence, 37, 156–163.
for neocortex. Neural Comput., 1, 480–488. Lund, J.S., Angelucci, A. & Bressloff, P.C. (2003) Anatomical substrates for
Eccles, J., Ito, M. & Szenthagothai, J. (1967) The Cerebellum as a Neuronal functional columns in macaque monkey primary visual cortex. Cereb.
Machine. Springer, Berlin. Cortex, 13, 15–24.
Eroglu, C. & Barres, B. (2010) Regulation of synaptic connectivity by glia. Madalosso, S.H., Pérez-Villegas, E.M. & Armengol, J.A. (2005) Naturally
Nature, 468, 223–231. occurring neuronal death during the postnatal development of Purkinje
Farel, P.B. (2002) Trust, but verify: the necessity of empirical verification in cells and their precerebellar afferent projections. Brain Res. Rev., 49, 267–279.
quantitative neurobiology. Anat. Rec. (New Anat.), 269, 157–161. Marino, L. (1998) A comparison of encephalization between odontocete
Feldman, D.E. (2009) Synaptic mechanisms for plasticity in neocortex. Annu. cetaceans and anthropoid primates. Brain Behav. Evol., 51, 230–238.
Rev. Neurosci., 32, 33–55. Middleton, F.A. & Strick, P.L. (1994) Anatomical evidence for cerebellar
Fischbach, G.D. (1992) Mind and brain. Sci. Am., 267, 48–57. and basal ganglia involvement in higher cognitive function. Science, 266,
Freiwald, W.A. & Tsao, D.Y. (2010) Functional compartmentalization and 458–461.
viewpoint generalization within the macaque face-processing system. Mukamel, R., Ekstrom, A.D., Kaplan, J., Iacoboni, M. & Fried, I. (2010) Single
Science, 330, 845–851. neuron responses in humans during execution and observation of actions.
Graeber, M.B. (2010) Changing face of microglia. Science, 330, 783–788. Curr. Biol., 20, 750–756.
Gross, C.G. (2009) The genealogy of the ‘grandmother cell’. In Gross, C.G. Nauta, W.J.H. & Feirtag, M. (1986) Fundamental Neuroanatomy. W.H.
(Ed.), A Hole in the Head. More Tales in the History of Neuroscience. MIT Freeman, New York, USA, p. 33.
Press, Cambridge, USA, pp. 263–283. Nave, K. (2010) Myelination and support of axon integrity by glia. Nature,
Gross, C.G., Rocha-Miranda, C.E. & Bender, D.B. (1972) Visual properties 468, 244–252.
of neurons in inferotemporal cortex of the macaque. J. Neurophysiol., 35, Nedergaard, M., Ransom, B. & Goldman, S.A. (2003) New roles for astrocytes:
96–111. redefining the functional architecture of the brain. Trends Neurosci., 26,
Herculano-Houzel, S. (2010) Coordinated scaling of cortical and cerebellar 523–530.
numbers of neurons. Front. Neuroanat., 4, 12. Nishiyama, A., Yang, Z. & Butt, A. (2005) Astrocytes and NG2-glia: what’s in
Herculano-Houzel, S. & Lent, R. (2005) Isotropic fractionator: a simple, rapid a name? J. Anat., 207, 687–693.
method for the quantification of total cell and neuron numbers in the brain. J. Noctor, S.C., Martı́nez-Cerdeño, V. & Kriegstein, A.R. (2007) Contributions of
Neurosci., 25, 2518–2521. intermediate progenitor cells to cortical histogenesis. Arch. Neurol., 64, 639–
Herculano-Houzel, S., Mota, B. & Lent, R. (2006) Cellular scaling rules for 642.
rodent brains. Proc. Natl. Acad. Sci. USA, 103, 12138–12143. O’Kusky, J. & Colonnier, M. (1982) A laminar analysis of the number of
Herculano-Houzel, S., Collins, C., Wong, P. & Kaas, J.H. (2007) neurons, glia, and synapses in the adult cortex (area 17) of adult macaque
Cellular scaling rules for primate brains. Proc. Natl. Acad. Sci. USA, 104, monkeys. J. Comp. Neurol., 210, 278–290.
3562–3567. Pakkenberg, H. (1966) The number of nerve cells in the cerebral cortex of man.
Herculano-Houzel, S., Collins, C.E., Wong, P., Kaas, J.H. & Lent, R. (2008) J. Comp. Neurol., 128, 17–20.
The basic nonuniformity of the cerebral cortex. Proc. Natl. Acad. Sci. USA, Pakkenberg, B. & Gundersen, H.J.G. (1997) Neocortical neuron number in
105, 12593–12598. humans: effect of sex and age. J. Comp. Neurol., 384, 312–320.
Heyes, C. (2010) Where do mirror neurons come from? Neurosci. Biobehav. Pakkenberg, B., Pelvig, D., Marner, L., Bundgaard, M.J., Gundersen, H.J.G.,
Rev., 34, 575–583. Nyengaard, J.R. & Regeuer, L. (2003) Aging and the human neocortex. Exp.
Hilgetag, C.C. & Barbas, H. (2009) Are there ten times more glia than neurons Gerontol., 38, 95–99.
in the brain? Brain Struct. Funct., 213, 365–366. Pelvig, D.P., Pakkenberg, H., Stark, A.K. & Pakkenberg, B. (2008)
Hofman, M.A. (1985) Neuronal correlates of corticalization in mammals: a Neocortical glial cell numbers in human brains. Neurobiol. Aging, 29,
theory. J. Theor. Biol., 112, 77–95. 1754–1762.
Hofman, M.A. (1988) Size and shape of the cerebral cortex in mammals. II. Pereira, A. Jr & Furlan, F. (2010) Astrocytes and human cognition: modelling
The cortical volume. Brain Behav. Evol., 32, 17–26. information integration and modulation of neuronal activity. Prog. Neuro-
Horton, J.C. & Adams, D.L. (2005) The cortical column: a structure without a biol., 92, 405–420.
function. Philos. Trans. R. Soc. Lond. B Biol. Sci., 360, 837–862. Piñón, L.G. & Linden, R. (1996) Target and afferents interact to control
Hubel, D.H. (1979) The brain. Sci. Am., 241, 44–53. developmental cell death in the mesencephalic parabigeminal nucleus of the
Innocenti, G.M. & Vercelli, A. (2010) Dendritic bundles, minicolumns, rat. J. Neurosci. Res., 45, 174–182.
columns, and cortical output units. Front. Neuroanat., 4, 11. Poth, C., Fung, C., Güntürkün, O., Ridgway, S.H. & Oelschläger, H.H.A. (2005)
Ito, M. (1972) Neural design of the cerebellar motor control system. Brain Res., Neuron numbers in sensory cortices of five delphinids compared to a
40, 81–84. physeterid, the pigmy sperm whale. Brain Res. Bull., 66, 357–360.

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
European Journal of Neuroscience, 35, 1–9
 Neuroscience dogmas and brain cell numbers 9

Purves, D., Augustine, G.J., Fitzpatrick, D., Hall, W.C., LaMantia, A.-S., Schmitz, C. & Hof, P.R. (2005) Design-based stereology in neuroscience.
McNamara, J.O. & White, L.E., eds. (2008) Neuroscience, 4th edn. Sinauer Neuroscience, 130, 813–831.
Associates, Sunderland, USA, p. 10. Schüz, A. & Palm, G. (1989) Density of neurons and synapses in the cerebral
Quiroga, R.Q., Reddy, L., Kreiman, G., Koch, C. & Fried, I. (2005) Invariant cortex of the mouse. J. Comp. Neurol., 286, 442–455.
visual representation by single neurons in the human brain. Nature, 435, Sherrard, R.M. & Bower, A.J. (1998) Role of afferents in the development and
1102–1107. cell survival of the vertebrate nervous system. Clin. Exp. Pharmacol.
Rakic, P. (1988) Specification of cerebral cortical areas. Science, 241, 170–176. Physiol., 25, 487–495.
Rakic, P. (2003) Elusive radial glial cells: historical and evolutionary Soper, B. & Rosenthal, G. (1988) The number of neurons in the brain: how we
perspective. Glia, 43, 19–32. report what we do not know. Teach. Psychol., 15, 153–156.
Rakic, P. (2009) Confusing cortical columns. Proc. Natl. Acad. Sci. USA, 105, Sporns, O. (2011) The human connectome: a complex network. Ann. NY Acad.
12099–12100. Sci., 1224, 109–125.
Rizzolatti, G. & Sinigaglia, C. (2008) Mirrors in the Brain. Oxford University Stolzenburg, J.U., Reichenbach, A. & Neumann, N. (1989) Size and density of
Press, Oxford. glial and neuronal cells within the cerebral neocortex of various insectivorian
Rizzolatti, G., Fadiga, L., Gallese, V. & Fogassi, L. (1996) Premotor cortex and species. Glia, 2, 78–84.
recognition of motor actions. Cogn. Brain Res., 3, 131–141. Vercelli, A., Garbossa, D., Curtetti, R. & Innocenti, G.M. (2004) Somatoden-
Rockel, A.J., Hiorns, R.W. & Powell, T.P.S. (1980) The basic uniformity in dritic minicolumns of output neurons in the rat visual cortex. Eur. J.
structure of the neocortex. Brain, 103, 221–244. Neurosci., 20, 495–502.
Roth, G. & Dicke, U. (2005) Evolution of the brain and intelligence. Trends White, E.L. & Peters, A. (1993) Cortical modules in the posteromedial barrel
Cogn. Sci., 9, 250–257. subfield (Sm1) of the mouse. J. Comp. Neurol., 334, 86–96.
Sato, T.R., Gray, N.W., Zachary, F., Mainen, Z.F. & Svoboda, K. (2007) Wig, G.S., Schlaggar, B.L. & Petersen, S.E. (2011) Concepts and principles in
The functional microarchitecture of the mouse barrel cortex. PLoSBiol., 5, e189. the analysis of brain networks. Ann. NY Acad. Sci., 1224, 126–146.
Schmahmann, J.D. (1998) Dysmetria of thought. Clinical consequences of cere- Williams, R.W. & Herrup, K. (1988) The control of neuron number. Annu. Rev.
bellar dysfunction on cognition and affect. Trends Cogn. Sci., 2, 362–370. Neurosci., 11, 423–453.
Schmahmann, J.D. (2010) The role of the cerebellum in corgnition and Yamazaki, Y., Hozumi, Y., Kaneko, K., Fujii, S., Goto, K. & Kato, H. (2010)
emotion: personal reflections since 1982 on the dysmetria of thought Oligodendrocytes: facilitating axonal conduction by more than myelination.
hypothesis, and its historical evolution from theory to therapy. Neuropsy- Neuroscientist, 16, 11–18.
chol. Rev., 20, 236–260. Zhang, K. & Sejnowski, T.J. (2000) A universal scaling law between
Schmahmann, J.D. & Pandya, D.N. (1997) The cerebrocerebellar system. Int. gray matter and white matter of cerebral cortex. Proc. Natl. Acad. Sci. USA,
Rev. Neurobiol., 41, 31–60. 97, 5621–5626.

ª 2011 The Authors. European Journal of Neuroscience ª 2011 Federation of European Neuroscience Societies and Blackwell Publishing Ltd
European Journal of Neuroscience, 35, 1–9