Oxford university press how the br ain evolved language

Figure 4.4. 
Brodmann areas. Lateral (top) and medial (bottom) views of the left 
cerebral hemisphere. (Brodmann 1909. Reprinted by permission of J. A. Barth 
Verlag.) 
Before the discovery of genetics, it was believed that human sperm cells 
contained a homunculus, a little man, which grew up into a bigger baby. This 
belief is now regarded as a preposterous example of medieval pseudoscience, 
so it was quite surprising when Wilder Penfield’s preoperative surveys discov­
ered just such a homunculus in the brain! Penfield showed that the primary 
motor area, which is the gyrus just anterior to the fissure of Rolando (Brodmann 
areas 4 and 6 in figure 4.4), sends signals to body parts as if a little homuncu­

60  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
lus were laid upside down on the gyrus (figure 4.5; Penfield and Rasmussen 
1950). 
In figure 4.5, mouth, lips, and tongue of the motor homunculus are drawn 
disproportionately large because they are controlled by disproportionately large 
areas of motor cortex. Apparently, eating well is as important to humans as it 
was to early coelenterates. But when the human homunculus’s mouth and lips 
are compared with similar studies of apes and other animals, we find they are 
still comparatively very large. For humans, speaking well may be even more 
important than eating well: the large area of motor cortex that the human 
motor homunculus devotes to mouth and lips must correspond to the com­
plex motor planning required by human speech. And indeed, it is in this fron­
tal region that Broca’s area is located, just anterior to the primary cortex for 
mouth, lips, and tongue. In fact, the entire forebrain extending frontward 
beyond Broca’s area is comparatively greatly enlarged in Homo loquens. In fig-
Figure 4.5. 
The motor homunculus. (Penfield and Rasmussen 1952. Reprinted by 
permission of Simon and Schuster.) 

THE  SOCIETY  OF  BRAIN 
•
61
 
ure 4.4 (top) Broca’s area is in the vicinity of Brodmann areas 44 and 45. Be­
cause Broca’s area is located in motor cortex, aphasics with damage to Broca’s 
area are often called motor aphasics. 
Sensory cortex and touch 
Neocortex in the gyrus just posterior to the central sulcus responds first to 
sensation and is often called primary sensory cortex. In this gyrus lies a second, 
twin homunculus (figure 4.6). This one maps the receptive, sensory regions 
of the cerebrum. 
This somatosensory homunculus receives afferent nervous signals from 
touch receptors throughout the body. But touch is of secondary interest to our 
investigation of cognition. Vision and hearing are much more important—so 
important in the chordate brain that they occupy not just parts of a small ho­
munculus but whole lobes of the brain. 
Figure 4.6. 
Somatosensory homunculus. (Penfield and Rasmussen 1952. Reprinted 
by permission of Simon and Schuster.) 

62  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
The temporal lobe: auditory cortex 
A second large sulcus, the Sylvian fissure, divides each hemisphere horizontally. 
In figure 4.4 (bottom), this is the black estuary running from seven o’clock 
toward two o’clock. Ventral to this fissure and beneath the skull bone for which 
it is named lies the temporal lobe (see also figure 4.1). The auditory nerve erupts 
into cortex here, within the fissure, in Brodmann area 34. This area is variously 
identified as koniocortex, Heschl’s gyrus, or primary auditory cortex (see also figure 
12.1). Wernicke’s area is found posterior to koniocortex, in the posterior part 
of the superior temporal gyrus, in the vicinity of Brodmann areas 41, 42, and 
22 (figure 4.4, top). 
The parietal lobe: association cortex 
The parietal lobe is situated above the Sylvian fissure and behind the fissure of 
Rolando. It, too, is named for the skull bone above it. The parietal lobe is vis­
ibly distinct from the occipital lobe behind it only when the cerebrum is viewed 
from within (as in figure 4.4, bottom, or figure 4.1). As noted above, the ante­
rior gyrus of the parietal lobe is primary sensory cortex, specifically dedicated to 
reception of primary sensory inputs from touch. The posterior parts of the 
parietal lobe, however, yield only diffuse responses to tactile stimulation or focal 
lesions. In consequence, parietal cortex is called association cortex because it 
diffusely associates primary percepts. 
The occipital lobe and vision 
The occipital lobe, again named for the skull bone under which it lies, is the most 
caudal (toward the tail) lobe of the cerebrum. It is clearly defined only when 
the brain is viewed from beneath or from within, where it is clearly delimited 
by the parieto-occipital fissure (as in figures 4.1 and 4.4). The main visual path­
ways terminate here in primary visual cortex, or striate cortex, after running from 
the retina through the thalamic LGN. It at first seems odd that the occipital 
lobe, located at the very back of the head, should process the visual percepts 
from the eyes in front, but when we recall that frogs and other prey animals 
have eyes where humans have ears, the better to watch their backs, the loca­
tion is not so strange. Indeed, we can learn much by close examination of 
animals. So even though this book is mostly interested in language, much of 
what we know about complex neural processing is derived from studies 
of vision in lower animals, and we must devote some space to a discussion of 
the vast scientific literature on vision. 
In 1967, Hartline and Granit received the Nobel Prize for their work on 
excitation and inhibition in animal visual systems (e.g., Hartline and Graham 
1932; Granit 1948; Hartline 1949; Hartline and Ratliff 1954; see also Ratliff 1965). 
In particular, Hartline studied the connection patterns in the retina of the “primi­
tive” species Limulus polyphemus, the horseshoe crab. Hartline found that the cells 
of the crab’s eye were arranged in an off-center off-surround anatomy (figure 4.7). 

THE  SOCIETY  OF  BRAIN 
•
63
 
Figure 4.7. 
An off-center off-surround anatomy. 
This anatomy suggests that Limulus perceives its world like a photographic nega­
tive. That is, when light strikes a cell in the Limulus retina (L in figure 4.7) that 
cell inhibits its corresponding postretinal cell, creating a black percept in response 
to light. But the cell also inhibits its surrounding retinal cells. These cells, being 
inhibited, no longer inhibit their corresponding postretinal cells. Being thus 
disinhibited, the postretinal cells become active, creating an “on” percept in 
response to light stimulation being “off.” 
We shouldn’t be too sure that Limulus perceives black as white, however. 
Just as (–1)(–1) = +1, inhibiting an inhibitory neuron can lead to the excita­
tion of another neuron. This kind of sign reversal can confuse our attempts to 
simply relate neurotransmitters to behavior. So, for example, adrenaline has 
an excitatory effect upon behavior, but in the brain, as the neurotransmitter 
noradrenaline, it has been found to have an inhibitory effect, hyperpolarizing 
the postsynaptic cell membrane (see figure 3.10).
4 
In mammalian vision, similar networks occur, but they are on-center off-
surround networks. This on-center off-surround anatomy is a significant evolu­
tionary advance. In on-center off-surround anatomies like that in figure 4.8, 
some center cells, say in the LGN of the thalamus, are excited by a stimulus. 
Those cells relay information up to cells in striate cortex. These, in turn, echo 
excitation back to the center cells, keeping them on. In the cerebrum (fig­
ure 4.8, top), afferent axons branch into the dendritic arbor of pyramidal cells 
(in the background) and a large basket cell (in the foreground). The basket 
cell sends axon collaterals to surrounding pyramidal cells, inhibiting them. At 
the same time, similar inhibitory cells in the thalamic reticular formation (in 
figure 4.8, bottom) inhibit surrounding thalamic relay cells. 
In contrast to the black-is-white world of Limulus, we can think of the on-
center off-surround anatomy as creating a world of sensation in which white is 
white and black is black. This does not confer any particular evolutionary vi­
sual advantage (Limulus has been around for a long, long time), but on-center 

64  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
Figure 4.8. 
A thalamocortical on-center off-surround anatomy. ( Jones 1981. 
Reprinted by permission of MIT Press.) 
off-surround anatomies are not restricted to vision. They are found at all lev­
els of anatomy from our six-celled brain in chapter 2 and the brain stem on up 
to the thalamic reticular formation and the cerebrum, where they are ubiqui­
tous. As we shall see, what especially confers an evolutionary advantage on the 
on-center off-surround anatomy is that it can learn, something for which Limu-
lus is not renowned. 
“Modules” and intermodular connections 
The lateralization of language to the left hemisphere, the localization of speech 
in Broca’s area and speech understanding in Wernicke’s area, and the pres­
ence of homunculi in primary sensory and motor cortex suggested that there 
are places in the brain for different processes, just as there are places or mod-

THE  SOCIETY  OF  BRAIN 
•
65
 
ules in a computer program for different subprograms. Perhaps the most spec­
tacular evidence for such modularity has come from Sperry’s work. 
The underside of the cerebral cortex is covered with tendonlike “white 
matter.” Until the discovery of nerve cells, such white matter was thought to 
be simply a ligature, holding the lobes and hemispheres of the cerebrum to­
gether. Now we know that the white matter actually consists of bundles of 
pyramidal cell axons, neural highways across which brain cells communicate. 
In patients suffering from severe epilepsy, Sperry bisected the corpus callosum, 
the massive bundle of nerve fibers connecting the right and left cerebral hemi­
spheres (Sperry 1964, 1970a, 1970b, 1967; see figure 4.10 and CC in figure 4.1). 
Sperry’s reasoning was that since massive, grand mal seizures exhibited nervous 
signals reverberating out of control back and forth between the hemispheres, 
severing the corpus callosum would stop this pathological resonance. The pro­
cedure was dramatically successful, but almost equally dramatic were some of 
the patients’ postoperative sequelae. 
For example, when such a “commisurectomized” patient was blindfolded 
and an apple placed in her left hand, she salivated and otherwise recognized 
that the object was food, but she could not say she was holding an apple. This 
happened because the sense of touch projects contralaterally from the left body 
to the right brain and from the right body to the left brain. As Broca noted, 
most people’s language is lateralized to the left hemisphere. So the patient’s 
left hand sent sensory touch signals up the brain stem to her right hemisphere, 
where the apple was behaviorally recognized as food, but since the corpus cal­
losum had been cut, the right hemisphere could not relay this information to 
the language modules of the left hemisphere, and the patient could not say 
she was holding an apple. Fortunately for commisurectomized patients, the 
cognitively important senses of vision and hearing are not as strongly lateral­
ized as touch. 
Vision is strictly contralateral in that the left visual field projects to the right 
hemisphere, and vice versa, but both eyes have both a left and a right visual 
field, so as long as the commisurectomized patient has both eyes open, he can 
see normally. Hearing is also less lateralized in the sense that each ear sends 
not only contralateral signals but also ipsilateral (same-side) signals. As it hap­
pens, the contralateral connections are stronger, allowing a dichotic listening 
test (Kimura 1967) to identify the language-dominant hemisphere even in 
normal subjects without localizable brain lesions. In a dichotic listening test, 
minimally contrasting words like bat and pat (minimal pairs) are simultaneously 
presented in stereo, one to each ear. Asked what word they hear, most subjects 
will most often report the word presented to the right ear. This indicates that 
the contralateral left hemisphere is dominant for language. 
More recently, another region of “intermodular” interaction has been iden­
tified in the angular gyrus, which lies at the intersection of the parietal, tempo­
ral, and occipital lobes, in the vicinity of Brodmann area 40 (figure 4.4, top). 
Focal lesions to this area of the cerebrum have resulted in rather pure alexia— 
the inability to read. This observation suggests that the modules for hearing, 
language, and vision send their outputs to the angular gyrus for processing 

66  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
during reading. The angular gyrus lies on a diffuse intermodular pathway con­
necting all of these modules, the arcuate fasciculus. 
The arcuate fasciculus 
Wernicke, noting the functional correlation of language understanding with 
the brain region that bears his name, was the first to propose the modular theory 
of brain structure. In the Wernicke-Lichtheim model (figure 4.9), a lesion at 1 
disrupts input from the ear to Wernicke’s area and corresponds to hearing 
impairment or deafness. A lesion at 2, pure Wernicke’s aphasia, allows sounds 
to be heard, but word perception and recognition are impaired. A lesion at 3, 
receptive aphasia, allows words to be recognized, but the comprehension of 
speech by association cortex is impaired. 
On the productive side, a lesion at 6 disrupts output from Broca’s area to 
the mouth. This corresponds to dysarthria, the motoric (as opposed to cogni-
tive/aphasic) inability to articulate speech. A lesion at 5, pure Broca’s aphasia, 
allows speech sounds to be uttered, but word production is impaired. A lesion 
at 4, productive aphasia, allows words to be uttered and repeated, but connected 
speech is impaired.
5 
The arcuate fasciculus (figure 4.10; 7 in figure 4.9) connects Broca’s area 
and Wernicke’s area. Wernicke predicted that a lesion to the arcuate fascicu­
lus would produce a conduction aphasia that would present the quite specific 
inability to perform verbatim repetition. The conduction aphasic would be able 
to understand speech because the pathway 1-2-3-4-5-6 would remain intact. He 
would be able to produce speech because the same pathway would also be in­
tact in the opposite direction. Following these pathways, he would be able to 
accurately paraphrase what is said, but he would be unable to repeat language 
Figure 4.9. 
The Wernicke-Lichtheim model of language cortex. (After Lichtheim 
1885.) 

THE  SOCIETY  OF  BRAIN 
•
67
 
Figure 4.10. 
The arcuate fasciculus, right hemishphere. (Montemurro and Bruni 
1981. Reprinted by permission of D. G. Montemurro.) 
verbatim along pathway 1-7-6! Although only a few cases of conduction apha­
sia have been unambiguously diagnosed (typically resulting from deep brain 
tumors), they have proved Wernicke right: Broca’s area and Wernicke’s area 
do exchange information across the arcuate fasciculus, via the angular gyrus. 
Broca’s work and Wernicke’s work inspired a century of research into the 
localization of language and other brain functions and led to the first scien­
tific understanding and treatments of aphasia, epilepsy, and other brain and 
language disorders. By the 1950s, however, psychologists and physiologists had 
begun to question localization (Lashley, “In Search of the Engram,” 1950). 
Under the rising influence of “modular programming” in computer science 
(Wirth 1971; Parnas 1972), research into localization persisted under compu­
tational metaphor, but Lashley’s “engram” was not to be found in any one place. 
Language “modules” like Broca’s area and Wernicke’s area were indistinct and 
variable. Indeed, there was even found to exist a small population of other­
wise normal people whose language is localized in the right hemisphere. Nor 
was this simply an isolated anomaly: a much larger minority (most left-handers) 
were found to have language fairly evenly distributed across both hemispheres. 
Then there were cases of aphasic children. For adult aphasics, the prognosis is 
bleak. Once the language hemisphere is damaged, recovery is usually incom­
plete and often minimal. But for child aphasics, the prognosis is miraculously 
good. Within several years, the other hemisphere or a spared gyrus takes over 
the language functions of the damaged region, and recovery is often—even 
usually—complete. 
There was also the case of sign language, mentioned in chapter 2. As it 
became accepted that the sign languages of the deaf were cognitively complete, 
it also became apparent that they are initially processed by manual cortex and 

68  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
visual cortex, not the oral and auditory cortex proximate to Broca’s module 
and Wernicke’s module. Although some recent fMRI (functional magnetic 
resonance imaging) studies of deaf signers do show some elevated activation 
in Broca’s area and Wernicke’s areas during signing, the activation is relatively 
small compared with spoken language (Neville and Bavelier 1996). A reason­
able explanation is that these are resonances created by the existence of the 
arcuate fasciculus rather than vestigial activation of a local “module.” 
It can be said that vision is innately hardwired to striate cortex in the occipi­
tal lobe, for this is where the optic nerve erupts into the cerebrum. It can be said 
that hearing is innately hardwired to koniocortex in the temporal lobe, for this 
is where the auditory nerve erupts into the cerebrum. It can be said that touch 
(as in Braille reading) is innately hardwired to the tactile-sensory regions of the 
parietal lobe, for this is where tactile nerves erupt into the cerebrum. It can be 
said that speech articulation is innately hardwired to posterior Broca’s area in 
the motor regions of the frontal lobe, for this is where signals from motor pyra­
midal cells leave the cerebrum, bound for the articulators. But it seems hard to 
say that language itself, like a sixth sense, is similarly hardwired to any particular 
place in the cerebrum. The cerebrum is plastic. (Indeed, Roe et al. [1990] suc­
cessfully surgically rewired a monkey’s optic nerve to auditory cortex!) As recov­
ered child aphasics show, language can be almost anywhere in the cerebrum. 
If there is a “module” for language, then the evidence we have seen would 
suggest that the module is neither Broca’s area nor Wernicke’s area nor 
the angular gyrus, nor even anything specific to the left hemisphere. If 
there is a “module” for language, it would seem that the best physiological 
candidate is the arcuate fasciculus, but the arcuate fasciculus is neither lan­
guage specific nor really a “place” in the brain. Science always progresses by 
ex-pressing the unknown with metaphors of the known, but the localist, com­
putational metaphor of “modularity” seems exhausted. Like calling the tele­
phone company “the building across the street,” metaphor can wrap new 
science in terms which the mass of researchers will find familiar and mean­
ingful (Kuhn 1962), but familiar metaphor can also obscure what is novel, 
distinctive, and essential. 
The laminar structure of neocortex 
Thus far, we have been concerned primarily with the cerebrum as it appears 
to the unaided eye, a wrinkled rind covered with bone. But if we were to take 
the cerebral cortex out of its skull, snip its white-matter ligaments, and then 
unfold it, we would get a rather different picture. We would find that the cere­
brum is actually a 0.5 m
2
 sheet of tissue about 4–5 mm thick. Instead of a dis­
tinctive geography of folds and bulges, we would see only an undifferentiated 
plain of gray matter above and white matter below. To find structure, we would 
need to look inside the sheet with a microscope. 
Anatomical studies (see figures 3.1 and 4.8) count six or seven cytologi­
cally distinct laminae in cerebral cortex. We, however, will focus on only three. 

THE  SOCIETY  OF  BRAIN 
•
69
 
In most areas of cortex, we can distinguish three layers in which (stained) py­
ramidal cell bodies predominate. In figure 3.1, these are in laminae III, V, and 
VI. Anatomical studies number these laminae in the order in which the anato-
mist encounters them, from the outside in. Ontogenetically, however, the first 
pyramidal cells to develop are those in the lowest layer, so unless I specifically 
use roman numerals in citing the anatomical literature, these are the cells I 
shall call “first.” These first pyramidal cells migrate out from the embryonic 
brain to form a first layer of cortex. Phylogenetically, it is as if a mutation oc­
curred in chordate evolution, causing not one but two thalamic reticular for­
mations to develop, the second (neocortex) enveloping the first (paleocortex) 
and becoming the cerebrum. Later, a second and a third self-similar layer of 
on-center off-surround brain developed, each enveloping the former. 
It is reasonable to assume that the first layer of early-evolving and early-
developing pyramidal cells are largest because they are oldest. As the largest, 
they have large dendritic trees which rise and branch high above their cell 
bodies. Their myelinated axons project far and wide below the cortical sheet. 
Many of these axons, notably those in the primary motor strip (the “motor 
homunculus,” figure 4.5), project far down into the basal ganglia and beyond. 
These and many others also send widely branching axon collaterals, forming 
the corpus callosum, the arcuate fasciculus, and the entire web of white mat­
ter beneath the cerebral sheet. Eventually these collaterals rise up again else­
where into the neocortical sheet, to innervate other neurons far from their 
originating cell bodies. 
Later in ontogeny, smaller pyramidal cells develop. They migrate outward 
from the embryonic brain, past the first layer of pyramidal cells, to form the 
second layer. In most of cortex, these are the primary sensory cells of neocor­
tex. Afferent pathways, rising up from the sensory organs through the thala­
mus, synapse preferentially upon these cells. Like the older cells of the first 
layer, their myelinated axons also project into the neocortical white matter, but 
they project more locally into neighboring neocortex, or they return recipro­
cal signals back to the thalamus. 
Finally in ontogeny, a third cohort of pyramidal cells migrates to form a 
superficial layer near the top of the cortical sheet. These are the smallest pyra­
midal cells. Their axons and dendritic trees are small, and their axons are the 
most local of all. 
This pattern is repeated everywhere in cerebral cortex with one major the­
matic variation: wherever cerebral input is concentrated, the second layer is 
especially densely populated with especially large pyramidal cells, and wher­
ever cerebral output is concentrated, the oldest layer is especially densely popu­
lated with especially large pyramidal cells. Thus, in striate cortex, where the 
optic nerve pathway enters neocortex, and in koniocortex, where the auditory 
nerve pathway enters neocortex, the middle layer is densely populated with 
largish pyramidal cells. Similarly, in the bottom pyramidal layer of primary 
motor cortex, where motor signals exit cortex, the pyramidal cells are relatively 
large and dense. 

70  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
Neurogenesis 
One of the great mysteries of the brain has been how a trillion-odd cells man­
age to wire themselves together in anything like an orderly fashion. How is the 
brain born? How does it develop? It has long been clear that neurons atrophy 
and die in the absence of nervous stimulation. Autopsies of the spines and 
brains of amputees reveal neuron atrophy and degeneration extending sev­
eral synapses away from the amputation. In animal studies, this natural experi­
ment has been refined to scientific technique, and much of what we presently 
know about the connectivity of nerve pathways has been learned by oblating 
sections of animal nervous systems and following the resulting patterns of at­
rophy and degeneration. 
Such facts, coupled with the discovery of chemical neurotransmission ac­
companying neural stimulation, led to the general theory that neurons depend 
for their very life upon the kind of neural import-export policy proposed at 
the end of chapter 3. But if (excitatory) neurotransmitters of the sort Dale 
discovered in the 1920s were the currency in this neural exchange, it was not 
until the 1980s that the goods were identified. In 1986, Levi-Montalcini and 
Cohen received the Nobel Prize for their discovery of nerve growth factor 
(NGF). 
Although NGF and related neurotrophins first appeared to simply be a new 
but important class of growth-stimulating hormones, it has recently come to 
be understood (Thoenen 1995) that their synthesis and efficacy are critically 
dependent upon neural activity. Simultaneously, nitric oxide (NO) was discov­
ered to be a ubiquitous “retrograde messenger” that is released from the acti­
vated postsynaptic cell and taken up by the stimulus-activated presynaptic axon 
terminal, there to facilitate growth and the production of more neurotrans­
mitter (Kandel and Hawkins 1992). The details are still the subject of cutting-
edge research, but from these pieces we can begin to develop a picture of how 
the brain is built. This emergent picture is very much one of “neural Darwin­
ism” (Edelman 1987), in which the developing brain not only grows but also 
evolves in a complex, neuroecological interplay of competitive and coopera­
tive responses to the environment. 
Columnar Organization 
In Golgi-stained sections (figure 3.1), the apical dendrites of cerebral pyrami­
dal cells stand out as pillars of neural structure, and close examination has 
revealed that inputs to and outputs from cerebral cortex are all perpendicular 
to the neocortical sheet. (Contrast this with the parallel fibers of the cerebel­
lum in figure 4.2.) From such observations arose the “columnar model” of 
cerebral organization. 
In the columnar model, the functional unit of cerebral processing is taken 
to be a multicellular column (Szentágothai 1969; figure 4.11). At the center of 
each such column, we imagine a large pyramidal cell. Specific afferent inputs 

THE  SOCIETY  OF  BRAIN 
•
71