Oxford university press how the br ain evolved language

Figure 3.9. 
Long-term potentiation. (Bliss and Lømo 1993. Reprinted by permis­
sion of the Physiological Society.) 
tion. It turns out that glutamate also attaches to NMDA gates. These gates are 
initially blocked by Mg
2+ 
ions, but as the membrane depolarizes through con­
tinued opening of non-NMDA receptor gates, the Mg
2+
 ions are dislodged. 
Once opened, the larger NMDA receptors admit doubly charged calcium (Ca
2+
) 
into the cell interior. There the calcium also interacts with calmodulin and CAM 
kinase II, which phosphorylates the non-NMDA glutamate receptors, causing 
them to be more readily activated in the future. This process is believed to 
be the basis of the physiological process of long-term potentiation and of the 
psychological phenomenon of LTM (Grossberg 1968; Lynch 1986; McGlade-
McCulloh et al. 1993). 
GABA, on the other hand, is inhibitory. It is released by cortical interneu­
rons like basket cells and chandelier cells (figure 3.10). Whereas glutamate 
opens gates for positive ions, GABA opens channels for the entry of negative 
ions, principally Cl
–
, into the cell. This influx of chloride ions hyperpolarizes the 
postsynaptic membrane, from –70 mV to perhaps –80 mV, thereby making it 
more difficult for excitatory neurotransmitters like glutamate to depolarize the 
membrane and initiate spiking. This hyperpolarization of cells does not have 
to match impedance in order to propagate a spike, so inhibitory cell synapses 
do not exhibit spines, and they are more commonly found on cell bodies than 
in dendritic arbors. 
For example, GABAergic chandelier cells synapse preferentially on the 
initial axon segments of neocortical pyramidal cells. When stained as in fig­
ure 3.10, the terminal knobs of their axon collaterals look like so many candle­
sticks, hence their name. The axon terminals of basket cells, on the other 
hand, form a “basket” of synapses on the bodies of neighboring excitatory 
cells (see figure 4.8). 

THE  COMMUNICATING  CELL 
•
49
 
Figure 3.10. 
A chandelier cell. (A camera lucida drawing from Peters and Jones 1984, 
365. The bar equals 25
µm. Reprinted by permission of Plenum Publishing Corp.) 
By some reports, inhibitory neurons compose less than one-fifth of the total 
number of CNS neurons, but Crick and Asanuma (1986) observe that even if 
inhibitory cells are outnumbered, they may exercise a disproportionate “veto” 
power over excitatory signals by exerting their inhibitory influence on prime 
real estate like cell bodies and initial axon segments. 
As I have noted, numerous other neurotransmitters are present in the CNS 
in lesser quantities. The most important among these are acetylcholine, norad-
renaline, and the G-protein neurotransmitters serotonin and dopamine. 
Acetylcholine is the neurotransmitter with which motor neurons cause 
muscles to contract and is apparently the principal neurotransmitter of the 
parasympathetic (smooth-muscle) nervous system. Acetylcholinergic fibers are 
also found widely distributed in neocortex and the midbrain. These seem to 
mostly arise from the reticular formation of the brain stem, part of the system 
for perceiving pain. Because it is present and easily studied at the neuromus­
cular junction, acetylcholine was the first neurotransmitter to be identified. 
For his discovery of the neurotransmitter role of acetylcholine, Loewi received 
the 1936 Nobel Prize. 
Serotonin has been traditionally classed as an inhibitory neurotransmit­
ter, although it has recently been found to also exert an excitatory effect upon 
cerebral cells (Aghajanian and Marek 1997). Since serotonin levels are selec­
tively elevated by antidepressives like fluoxetine (Prozac), serotonin has become 
famous for its role in controlling various mood disorders. 
Noradrenaline is another widespread neurotransmitter. It is identical to 
what elsewhere in the body is called the “hormone” adrenaline, but adrenaline 
does not cross the blood-brain barrier. The brain must manufacture its own 
supply of adrenaline, which it does from dopamine. In the brain, this adrena­
line is called noradrenaline or norepinephrine. In the hippocampus of the brain, 
and presumably also in neocortex, which develops from the hippocampus, 

50  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
noradrenaline has a complex effect. It inhibits neuronal response to brief 
stimuli but increases neuronal response to prolonged stimuli. 
Figure 3.11 illustrates this differential effect of noradrenaline on inputs 
to a hippocampal pyramidal neuron. In the first case shown in figure 3.11A, a 
small, ramped input (below) elicits a spike response (above). By contrast, a 
longer, more intense “pulse” input elicits a burst of seven spikes in response to 
the pulse onset. However, under the influence of noradrenaline, the ramped 
input elicits no response, while the pulse elicits fourteen spikes throughout 
the duration of the pulse. That is, noradrenaline inhibits small inputs but makes 
the neuron hyperexcitable and responsive to the more intense pulse stimulus. 
Madison and Nicoll (1986) observe that this effectively enhances the signal-
to-noise ratio of the neuron’s response. I will elaborate on such effects later. 
Dopamine is a CNS neurotransmitter which is chemically transformed into 
noradrenaline (both are catecholaminergic). Dopamine deficiency has been 
found to be a symptom of Parkinson’s disease. Treatment with L-dopa, a dopam­
ine precursor, with subsequent increases in dopamine levels in the subcortical 
Figure 3.11. 
Effect of noradrenaline on evoked postsynaptic potentials. (Madison 
and Nicoll 1986. Reprinted by permission of the Physiological Society.) 

THE  COMMUNICATING  CELL 
•
51
 
basal ganglia, described in the following chapter, has ameliorated the symp­
toms of Parkinson’s disease for many patients. 
Reverse Messengers 
At the same time that neurotransmitter released from the presynaptic knob 
acts upon the postsynaptic cell, retrograde neurotransmitters such as nerve growth 
factor (NGF; Thoenen 1995) and nitric oxide (Kandel and Hawkins 1992; Snyder 
and Bredt 1992) are released from the postsynaptic cell and act upon the pr­
esynaptic cell, opening Ca
2+
 gates and otherwise facilitating the metabolism of 
the presynaptic knob. A kind of intercellular free-trade agreement is set up, 
and the economies of both the presynaptic cell and the postsynaptic cell begin 
to grow (Skrede and Malthe-Sorenssen 1981; Errington and Bliss 1982).
5 
Thus, the cell membrane, life’s first defense in the primordial soup, evolved 
to allow cooperative commerce among friendly neighbors. There are many 
examples of symbiosis in nature, but the cooperative exchange of ions between 
two protoneurons was certainly one of its first and most significant occurrences. 
Once two neurons could communicate between themselves, it was relatively 
easy for 200 or 2,000 to organize themselves in patterns of self-similarity. It took 
something like a billion years for life to progress to the two-celled brain. In 
half that time, the two-celled brain evolved a million species, including Homo 
loquens with his ten-trillion-celled brain, to which we now turn in chapter 4. 

52
•
HOW THE BRAIN EVOLVED LANGUAGE
• 
F 
O 
U 
R
• 
The Society of Brain
In time, the primitive brains that had developed by the end of chapter 2 sur­
rounded themselves with yet another self-similar membrane. Called the noto-
chord, this cartilaginous tube houses the primitive central nervous system of the 
phylum Chordata. In ontogeny as well as phylogeny, this notochord further 
develops into a hard, tubular backbone, so the members of Chordata have come 
to be popularly known as “vertebrates.” The long, hollow notochord/backbone 
with its bulbous skull at one end is self-similar to the long, hollow axon and its 
cell body, and it serves the self-same purpose of communication in these larger, 
more complex organisms. In time, the four-celled model brain of chapter 2 
evolved into the trillion-celled human brain. 
Elephants and whales have larger brains than humans, and they are prob­
ably more intelligent, too. At least they have never engaged in such colossal 
stupidity as World War I. But if we must find a way to assert humankind’s sup­
posedly superior intellect, we can observe that elephants and whales seem to 
devote a good portion of their prodigious brains simply to moving their prodi­
gious bulks, and we can note the oft-repeated fact that per kilogram of body, 
Homo loquens has the largest brain in the world. But even this measure does 
little to support our vanity. Among mammals, birds, and reptiles, there is only 
a 5% variation in the ratio of brain size to body size (Martin 1982). The proper 
question therefore seems to be, not how big an animal’s brain is or how “intel­
ligent” the animal is, but what it does with the brain it has. 
Subcerebral Brain Structures 
If we peer beneath the surrounding cerebrum, we see the older structures which 
the cerebrum has overgrown (figure 4.1). These subcerebral (or “subcortical”)
1 
structures are usually not considered specialized for cognition or language, but 
52 

THE  SOCIETY  OF  BRAIN 
•
53
 
Figure 4.1. 
Medial section of the human brain. (CC) Corpus callossum; (CG) 
cingulate gyrus; (T) Thalamus; (Cblm) cerebellum; (P) pons; (Med) medulla; (Hc) 
hippocampus; (Hy) hypothalamus. (After Montemurro and Bruni 1988. Reprinted 
by permission of Oxford University Press.) 
they are nonetheless indispensable to higher cortical functioning, so at least a 
quick survey is in order. 
At the base of the subcerebral brain structures is the brain stem. Like the 
primitive notochord, it can be viewed as the source of all further brain devel­
opment. The lower part of the brain stem, the medulla (Med in figure 4.1) 
conducts afferent (ascending) signals from the periphery to the brain and effer-
ent (descending) signals from the brain to the periphery. These peripheral 
connections can be either relatively local (e.g., to the eyes, ears, and mouth) 
or quite distant (e.g., to the limbs via the neuromotor highway of axons called 
the pyramidal tract).
2 
Several diffuse catecholaminergic neurotransmitter systems arise from the 
medulla that have important connections to the limbic system and the hypo­
thalamus. These include important noradrenaline (norepinephrine) systems 
arising from the locus coeruleus, and dopamine systems arising from the substantia 
nigra. A serotonergic system arises from the raphe nuclei of the medulla, also 
with diffuse connections in the limbic system and cerebrum. These systems 
modulate behavior, but in globally autonomic and emotional ways rather than 
through cognition. 
Rostral to (i.e., above, toward the head of) the medulla is the pons (P in fig­
ure 4.1). The pons relays ascending and descending motor signals and is the 
primary relay site for signals to and from the cerebellum. Above the pons is the 
midbrain, a short section of the brain stem containing the superior and inferior 
colliculi. The superior colliculus is a major relay point in the visual system, while 
the inferior colliculus is a major relay point along auditory pathways. 

54  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
Rostral to the midbrain is the diencephalon. We divide the diencephalon into 
three parts: the thalamus, the basal ganglia, and the limbic system. The thalamus (T 
in figure 4.1) receives afferent (ascending, incoming) sensory signals and relays 
them to the cerebrum. Within the thalamus, the lateral geniculate nucleus (LGN) 
is an important relay point for visual signals, while the medial geniculate nucleus 
(MGN) is the thalamic relay point along auditory pathways. Most afferent sen­
sory circuits rise through the dorsal thalamus into the cerebrum. The dorsal 
thalamus is enveloped by a sheet of inhibitory neurons. This inhibitory envelope 
is called the reticular nucleus of the thalamus (RNT, RTN), even though it is not 
nuclear in shape. It should be distinguished from the reticular formation of the brain 
stem, which is involved in the more primitive sensation of pain. The RNT and 
dorsal thalamus together (sometimes called paleocortex) may be seen as the phy­
logenetic precursor of cerebrum. Both paleocortex and cerebrum exhibit on-center 
off-surround circuitry, a neuroarchitectural design that will become very impor­
tant in subsequent chapters. 
The basal ganglia comprise a group of structures including the caudate 
nucleus, the putamen, and the globus pallidus. Not visible in figure 4.1, these struc­
tures lie behind the plane of the medial section, lateral (alongside) and ante­
rior to (to the front of) the thalamus. They have occasionally been linked to 
language behavior (Metter et al. 1983; Ullman et al. 1997), but we will view this 
link largely in terms of their better-documented involvement with posture and 
gross motor control. The basal ganglia are particularly influenced by dopami­
nergic signals arising just below in the substantia nigra of the midbrain, and dete­
rioration of this dopaminergic pathway is an immediate cause of Parkinson’s 
disease. 
The Limbic System 
Ventral to (beneath) and surrounding the thalamus are the various structures 
of the limbic system. The hypothalamus (Hy in figure 4.1) is the seat of physiologi­
cal drives (hunger, sex, fear, aggression). Often classified as a gland, it secretes 
many hormones, such as vasopressin and oxytocin, and exerts direct influences 
on the autonomic nervous system, which in turn controls such functions as heart 
rate and breathing. The hypothalamus also directly affects the brain’s pituitary 
gland, which secretes a wide range of other hormones. As noted in chapter 3, 
these hormones are sometimes classed as neurotransmitters, but since they are 
diffusely circulated with effects beyond the nervous system, I prefer to class 
them simply as hormones, reserving the term “neurotransmitter” for chemi­
cals with more cognitive synaptic effects. 
Lateral to the hypothalamus is the amygdala. The amygdala sends pro­
jections directly to the caudate nucleus and is sometimes therefore counted 
among the basal ganglia. The amygdala is quite directly connected to the 
frontal and the motor cortex and has been implicated in the modulation of 
emotions. 

THE  SOCIETY  OF  BRAIN 
•
55
 
The hippocampus (Hc in figure 4.1) is the amygdala’s sensory counterpart. 
In 1966, Milner described a patient, HM, who had suffered from severe tempo­
ral lobe epilepsy (see also Milner et al. 1968). In a standard effort to control the 
epilepsy, surgeons removed large areas of temporal cortex. In HM’s case, they 
also removed substantial portions of the hippocampus. HM’s epilepsy was brought 
under control, but a new problem was created. HM developed anterograde amne-
sia: he “forgot the future.” HM was unable to form new memories. He was un­
able to learn. He remembered well his wife, friends, and family and his old 
neighborhoods and haunts. He could not, however, remember people he met 
since his surgery. Each day, he would meet with his doctor and say, “Have we met?” 
Milner’s report caused an instant sensation in the neuroscience commu­
nity and inspired numerous studies of the hippocampus. The most popular 
theoretical explanation of hippocampal function has been a consensus “buffer” 
model (see the four-volume series edited by Isaacson and Pribram for the evo­
lution of this model). In this consensus model, the hippocampus is viewed as 
“working memory,” rather like a computer coprocessor or RAM cache which 
performs operations on sensory input or briefly stores data before it is trans­
ferred to long-term storage in “declarative memory.” This model has the at­
traction of computational metaphor, but it accords a large cognitive role to a 
small structure in a brain region otherwise not found to be particularly “cogni­
tive.” It is also unable to explain many facts. For example, HM was able to de­
velop long-term memory for certain unemotional forms of knowledge like the 
solution to the Tower of Hanoi problem.
3
 More plausible from our perspec­
tive is Gray’s theory of the hippocampus as a “comparator” (Gray 1975, 1982). 
Gray’s theory also appeared in the Isaacson and Pribram series (Gray and 
Rawlins 1986), but it represented a minority opinion and has been slower to 
gain popularity. 
Situated posterior to the amygdala and lateral to the hypothalamus, the 
hippocampus is also widely connected to the cerebrum and the cingulate gyrus 
(sometimes called the “limbic lobe”; (CG in figure 4.1) via the cingulum. The 
cingulum is a massive bundle of axons that originate in the parahippocampal 
lobe of the temporal lobe and arch up and around the diencephalon, behind 
the cingulate gyrus in figure 4.1. (In figure 4.10, the left cingulate gyrus has 
been dissected, exposing the cingulum.) Indeed, the hippocampus is so mini­
mally differentiated from cerebral cortex that it is sometimes called the “hip­
pocampal lobe” of cerebrum. But the hippocampus is also connected to the 
mammilary body of the hypothalamus via the fornix. In Gray’s theory, the limbic 
system—and the hippocampus in particular—can be seen as moderating be­
tween the cognitive information of the cerebrum and the physiological and 
emotional drives of the more primitive brain. 
We will return to this issue several times in subsequent chapters. For the 
present, we simply observe how the tears and tantrums of any two-year-old 
demonstrate that learning can be a very emotional business, and that the case 
of HM suggests that learning can fail if it is disconnected from primitive sur­
vival instincts and drives. 

56  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
The Cerebellum 
Dorsal to the pons is the cerebellum (Cblm in figure 4.1). The cerebellum, like 
the cerebrum, is composed of a rind, or cortex, wrapped around several deep 
nuclei (CN in figure 4.2) and divided into two hemispheres. The cerebellum 
is especially notable for its control and coordination of fine motor behaviors 
like knitting, playing a musical instrument, or speaking a language. This con­
trol does not, however, include the planning or initiation of behavior, so as 
we shall see, the cerebellum’s role is considerably less “cognitive” than the 
cerebrum’s. 
The distinctive cerebellar architecture (figure 4.2) is characterized by large 
output cells, the Purkinje cells (PC in figure 4.2), which send signals to cerebel­
lar subcortical nuclei (CN). Purkinje cells are innervated by climbing fibers (CF), 
nonspecific inputs that arise from the inferior olive of the medulla, and by long 
by parallel fibers (PF) arising from granule cells (GC), which are the cerebrum’s 
principal input of sensory and motor information. Granule cells are innervated 
by afferent mossy fibers (MF). 
The Purkinje cells are embedded in the regular grid of the cerebellum’s 
parallel fibers, forming a matrixlike architecture which is schematized in fig­
ure 4.3. The fact that this architecture can be efficiently modeled by using well-
known shortcuts like matrix algebra made it easy to quickly program complex, 
putatively cognitive, computational neural networks. But unlike cerebral pyra-
Figure 4.2. 
Cerebellar cortex. (Fox 1962. Reprinted by permission of Appleton and 
Lange.) 

THE  SOCIETY  OF  BRAIN 
•
57
 
Figure 4.3. 
Schematic of cerebellar cortex. (Loritz 1991. Reprinted by permission 
of Oxford University Press.) 
midal cells, cerebellar Purkinje cells are inhibitory. As we shall see, the cerebel­
lum therefore does not have the on-center off-surround architecture of cere­
bral cortex. As one result, the cerebellum is “swept clean” of residual neural 
activity within 0.1 ms of input (Eccles 1977). This is all well and good for the 
processing of fast and fine motor commands, and in later chapters we will see 
how it is also crucial to the fluent pronunciation of language, but a 0.1 ms short-
term memory ill-serves what most scientists would call “cognition.” In compari­
son with cerebral architecture, we will conclude that the cerebellar architecture 
is fundamentally noncognitive. The important point here is that language is 
not and cannot be learned by just any brain cells: it is, with several interesting 
qualifications, really learned only by cerebral cells, and so it is to the cerebrum 
that we turn our major attention. 
The Cerebrum 
After the skull is opened, it is the cerebrum that first presents itself to the sur­
geon. On first viewing, the cerebral cortex looks like a thick placemat (or, yes, 
a thin rind) about 5 mm in thickness, which has been wrinkled and stuffed 
into a too-small cranium. This is the traditional view of the cerebrum, as well 
as the one most familiar to the lay reader, and it is the view with which we shall 
begin. 
But there are more illuminating ways to look at the cerebrum. As we have 
seen, Ramón y Cajal used the microscope to take a quite different view. He 
put the cerebral sheet on edge and studied it from the perspective of its thick­
ness. It was in this view that the cytoarchitecture of cerebral cortex was first 
exposed, and this will be the second perspective from which we shall view 
cerebral cortex. 

58  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
A third view of the cerebrum combines the surgical and cytoarchitectural 
views. This is the “planar” view of the cerebrum, and it is the view that will be 
most important to our subsequent development of adaptive grammar. 
The surgical view of the cerebrum: 
the cerebral hemispheres 
The human cerebrum is fraught with folds and bulges. These (usually larger) 
fissures and lobes or (usually smaller) sulci and gyri were the basis of the earliest 
anatomical attempts to describe the brain by its shape. No two brains are 
wrinkled in exactly the same way, but the larger folds and bulges are common 
to all human brains. Later, the finding that these common lobes often process 
specific types of information gave rise to the computational metaphor of a 
“modular” brain. 
Like most chordate anatomies, the cerebrum (as well as the subcerebral 
brain) is bilaterally organized and divided into left and right hemispheres by a 
deep central sulcus. As originally suggested in figure 2.9, the right hemisphere 
moves and senses the left body while the left hemisphere moves and senses 
the right body, and as noted in chapter 1, language is primarily processed in 
the left hemisphere. The right hemisphere is more “lateralized for” spatial tasks. 
In other major respects, however, the hemispheres are essentially identical. 
Figure 4.4 shows the major fissures of the brain, as well as cytoarchitec­
turally distinguishable areas of cortex known as Brodmann areas. Shown are 
a lateral view (an outside-in view from the side) of the left hemisphere and a 
sagittal (an outside-in side view—not a dead-center, medial view) of the left 
hemisphere. 
Mapping motor cortex 
Each cerebral hemisphere is divided by a long, vertical fissure of Rolando, the 
S-shaped line running down the center of figure 4.4 (top). The frontal lobe, which 
lies anterior to this central sulcus, plans actions and so is often called motor cortex. 
In the 1950s through the 1970s, before the development of radiological 
techniques, surgical removal was the most effective treatment for brain tumors. 
Since the brain itself has no neurons which can directly sense pain, brain sur­
gery only requires a local, scalp anesthetic, and after the skull has been opened, 
the patient can remain conscious during the procedure. This requires great 
courage on the part of the patient, but it is also a great contribution to the 
safety of the procedure. With the brain exposed, the surgeon can stimulate 
different regions of the brain with an electric probe while the patient reports 
what he senses. Using this information, the surgeon can avoid accidentally 
damaging speech areas or other especially critical areas of the brain. The com­
posite of a number of such surveys revealed that many other primary motor 
and sensory functions are localized in the same way that language is localized 
in Broca’s and Wernicke’s areas. 

THE  SOCIETY  OF  BRAIN 
•
59