Oxford university press how the br ain evolved language

How the Brain Evolved 
Language
Donald Loritz 
OXFORD UNIVERSITY PRESS

HOW  THE  BR AIN
EVOLVED  LANGUAGE

This page intentionally left blank 

·
·
·
·
·
 
HOW  THE  BR AIN 
EVOLVED  LANGUAGE 
·
·
·
·
·
 
Donald Loritz 
1

1
Oxford  New York 
Athens  Auckland  Bangkok  Bogotá  Buenos Aires  Cape Town 
Chennai  Dar es Salaam  Delhi  Florence  Hong Kong  Istanbul  Karachi 
Kolkata  Kuala Lumpur  Madrid  Melbourne  Mexico City  Mumbai  Nairobi 
Paris  São Paulo  Shanghai  Singapore  Taipei  Tokyo  Toronto  Warsaw 
and associated companies in 
Berlin  Ibadan 
Copyright © 1999 by Donald Loritz 
Published in 1999 by Oxford University Press, Inc. 
198 Madison Avenue, New York, New York 10016 
First issued as an Oxford University Press paperback, 2002 
Oxford is a registered trademark of Oxford University Press 
All rights reserved. No part of this publication may be reproduced, 
stored in a retrieval system, or transmitted, in any form or by any means, 
electronic, mechanical, photocopying, recording or otherwise, 
without the prior permission of Oxford University Press. 
Library of Congress Cataloging-in-Publication Data 
Loritz, Donald, 1947– 
How the brain evolved language / Donald Loritz. 
p.  cm. 
Includes bibliographical references and index. 
ISBN 0-19-511874-X; 0-19-515124-0 (pbk.) 
1.  Language and languages—Origin.  2.  Biolinguistics.  3.  Grammar, 
Comparative and general.  4.  Human evolution.  I.  Title. 
P116.L67  1999 
401—dc21 
98-29414 
1 3  5 7  9 8 6  4 2  
Printed in the United States of America 
on acid-free paper 

This book is dedicated to my family 
and 
to the memory of 
Walter A. Cook, S.J. 

This page intentionally left blank 

ACKNOWLEDGMENTS
This book has been so long in the writing that I can hardly begin to recount 
and thank the many people who have helped it on its way. Foremost among 
them are certainly Paula Menyuk and Bruce Fraser, who introduced me to the 
many, many subtleties of generative psycholinguistics. Then I must thank Steve 
Grossberg for teaching me to avoid “homunculi” as cheap solutions to hard 
psychological problems. 
In the late 1970s and 1980s, there was little appreciation of neural networks. 
During these years I must thank in particular Jan Robbins, Walter Cook, and 
Sam Sara for supporting me on little more than simple faith. By the late 1980s, 
however, neural networks had become more widely respected, and with the help 
of Bernard Comrie, Charles Ferguson, Winfred Lehmann, and Betty Wallace 
Robinett, and the support of my colleagues, I succeeded in securing tenure 
and setting aside some time to think heretically about how language could 
possibly work without a computational homunculus. 
For such sense as the following pages might make, I am indebted also to 
Allen Alderman, Kathy Broussard, Craig Chaudron, Dick Chen, Brad Cupp, 
Lisa Harper, Beatriz Juacaba, Jee Eun Kim, Elena Koutsomitopoulou, Bernard 
Kripkee, Mark Lewellen, Lise Menn, Afsar Parhizgar, Lisette Ramirez, Bill Rose, 
and Elaine Shea for their insights and support. I fear I have not yet achieved 
the rigorous standards my teachers have set, but the time has come to collect 
my attempts at explaining language so that better minds might bring them to 
fruition. 
December 1998 
D. L. 

This page intentionally left blank 

ONE 
· 
TWO 
· 
THREE 
· 
FOUR 
· 
FIVE 
· 
SIX 
· 
SEVEN 
· 
EIGHT 
· 
NINE 
· 
TEN 
· 
ELEVEN 
· 
TWELVE 
· 
CONTENTS 
Lought and Thanguage  3
Jones’s Theory of Evolution  21
The Communicating Cell  36
The Society of Brain
Adaptive Resonance
Speech and Hearing
52 
74 
90 
Speech Perception  109
One, Two, Three  123
Romiet and Juleo  133
Null Movement  143
Truth and Consequences  161
What If Language Is Learned by Brain Cells?  171
Notes  195
References  203
Index  219

This page intentionally left blank 

HOW  THE  BRAIN
EVOLVED  LANGUAGE

This page intentionally left blank 

• 
O 
N 
E
• 
Lought and Thanguage
A baby wildebeest, born on the Serengeti, learns to walk and run in a matter 
of minutes. The precocious wildebeest is marvelously adapted. But once a wilde­
beest can outrun a lion, its adaptation is largely done. Humans must adapt, 
too, but they must continue to adapt, first to mother, then to brother, then to 
teacher, then to work, then to him, then to her, then to babies, and so it goes. 
The world to which humans must adapt is a world of words: his word, her word, 
my word, your word, word of mouth, word of honor, word of scripture, word of 
law, the adaptive word. 
A baby wildebeest, born on the Serengeti, learns to walk and run in a matter 
of minutes, but it takes two full years before a baby human makes a word, learns 
to talk. And it is another ten years before the human child learns to talk back. 
Even then, it is an immature criterion that would claim the teenager has mas­
tered language. Human beings’ fascination with language persists as long as 
their fascination with life. How does a baby learn language? How does one talk 
to a teenager—or, for that matter, to a parent? What do words mean? Or, since 
I know what I mean, what do your words mean? And what do you think my words 
mean? Does she think I’m a nerd? Does he think I’m a bimbo? Will my boss 
think I’m disloyal if I don’t say yes? How can I adapt? 
Where Is Language? 
One doesn’t have to be a philosopher to ask such questions, and in fact, it may 
be better if one is not. The celebrated learned men of history never concerned 
themselves much about the learning of language. (After all, teaching children 
language is women’s work!) Thus absolved of responsibility for explaining lan­
guage in its most human terms, the wise men of history were freed to conflate 
and confuse thought and language as they pleased. 
3 

4
•
 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
This is not to deny that down through the ages many amusing books have 
been written on the topic of language. For example, in 1962 the philosopher 
J. L. Austin published a book exploring the question Is the King of France bald? 
(Since there is no longer a King of France, Austin concluded the question is 
neither true nor false, but simply void.) By and large however, previous phi­
losophers of thought and language have simply been ignorant of one impor­
tant modern fact: thought happens in the brain! 
Aristotle, to pick one particularly influential example, thought that the 
main function of the brain was to cool the blood.
1
 In hindsight, the ancients’ 
ignorance of the brain and its function was quite understandable. Locked up 
in its bony carapace, the brain, which resisted exposure to the warrior’s sword, 
resisted as well the anatomist’s scalpel. And even when the ancients noticed 
there was a brain beneath all that bone, they couldn’t see it do anything. It didn’t 
beat or breathe or bend. What ancient could have imagined that the brain 
created ideas with the same electrical forces as Zeus’s thunderbolts? Real knowl­
edge, Aristotle thought, was lodged in the heart, and even today, when we know 
something well, we say we “know it by heart.” So we can understand as well the 
ancients’ belief that knowledge was mysteriously dissolved in the blood. 
Finally, 2,000 years after Aristotle, Harvey showed that the heart pumped 
blood through the body, circulating nourishment to the organs of the body. 
Knowledge had to be somewhere else. But the microscope had not yet been 
invented, and when the seventeenth-century eye looked at the brain, the first 
feature it noted was that the brain, like the heart, had several connected, fluid-
filled chambers, called ventricles. (In Figure 1.1, a horizontal section of the brain 
exposes the main, lateral ventricles.) To seventeenth-century philosophers, the 
meaning was obvious: the brain was just another pump. Following Galen and 
Harvey, Descartes thought it pumped an animating fluid (animus) through the 
nerves, thereby causing muscles to move. He specifically thought that the 
pineal gland at the base of the ventricles was a kind of master valve, which con­
trolled hydraulic pressure in the system. To Descartes, this brain-pump was just 
so much more plumbing. Hydraulically moving the muscles was important, but 
it was just machinery; it could have nothing very much to do with thought. For 
Descartes, thought happened somewhere else. Thought happened in the mind, 
not in the brain. 
But where was the mind? For Descartes, language, mind, and thought were 
all essentially the same thing. Descartes would have asserted that it makes no 
more sense to ask Where is the mind? than it does to ask Where is language? or 
Where is algebra? Such questions, to use Austin’s term, were simply void. Lan­
guage, thought, and mind were abstract sets of formal relations. They could 
relate things in places to other things in other places, but they were not them­
selves in some place. For Descartes, thought and language, mind and mean­
ing, algebra and geometry, were all essentially the same sort of thing, which is 
to say they weren’t really things at all. 
In the seventeenth century this dualism of mechanics and mind, of things-
in-the-world and things-not-in-the-world, had a confirming parallel in the 
Church’s natural-supernatural dualism of life and afterlife. In a sense, Descartes 

LOUGHT  AND  THANGUAGE 
•
5
 
Figure 1.1. 
Horizontal section of the cerebrum. The (lateral) ventricles are 
exposed. Occipital lobe (O), parietal lobe (P), frontal lobe (F), lateral ventricle 
(L), corpus callosum (C), fissure of Rolando (R). (Kirkwood 1995, 15. Reprinted by 
permission of Churchill Livingstone.) 
extended the conception of the supernatural to include not only angels but 
also algebra, algorithms, and language. These otherworldly entities had a truth 
that, to Descartes, was obviously true, a priori, before and independent of any 
empirical experience. One could only find this truth by doubting empirical, 
in-the-world experience and by believing in a priori, not-in-the-world truths, 
truths like the existence of God. 
But how could you or I, mere mortals both, know in the mind that even 
we ourselves exist, let alone so sublime a being as God? Was there a less pre­
sumptuous a priori truth from which we could deduce these larger Truths? 
Perhaps the most famous “rationalist” deduction of this sort was Descartes’s 
proof of his own existence: 
cogito ergo sum 
(1.1) 
think-I therefore exist-I 
“I think therefore I am.” 
Unfortunately, as Nietzsche later observed, Descartes’s “proof” turns out 
to be uselessly circular: in Latin, the -o on cogito and the form sum itself both 
indicate first-person I. Consequently, as the literal gloss of 1.1 emphasizes, the 
premise I think presupposes the conclusion I am. To illustrate this point, con­

6
•
 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
sider 1.2 and 1.3 (here and elsewhere a * means “something seems wrong with 
this sentence”): 
*You think therefore I am. 
(1.2) 
or 
*Thought exists therefore I am. 
(1.3) 
Without its presuppositions, Descartes’s proof fails utterly. In Descartes’s 
defense, we should perhaps consider the context of his times. The Reforma­
tion had put reason at odds with God, and Descartes had a larger agenda than 
to vainly prove his own existence. But the proof is still false. Even a genius can­
not deduce truth from faulty premises. 
Tabula Rasa 
Well before Nietzsche, many philosophers objected to Descartes’s dualistic 
method. Descartes’s contemporary Francis Bacon strenuously objected to 
Descartes’s introspective method. Francis Bacon (and, coincidentally, 400 years 
earlier, Roger Bacon) espoused a rather distinctively English empiricism. Un­
like Descartes’s dualism, this empiricism was a triadism that divided the uni­
verse into Soul, Mind, and Matter. Leaving the supernatural aspects of Soul to 
God, empiricism proceeded to focus on the material aspects of Matter. But 
neither Bacon was a rigorous scientist by modern standards. (In what was ap­
parently his first and only scientific experiment, Francis Bacon stuffed a chicken 
with snow to see if the snow would inhibit decay. The only reported result was 
that Bacon caught cold and died.) The relationship of Mind to Soul and Mat­
ter was little advanced by their methods. It wasn’t until a hundred years after 
Descartes that empiricism found a clear voice in the philosophy of John Locke. 
For Locke, Mind was just a blank slate, an erased tablet of Matter, a tabula rasa. 
Experience wrote upon the tablet, thus creating Mind. Of course rationalists 
objected that this explained no more than cogito ergo sum. If, as the empiricists 
would have it, there was such a tablet, then where was it? Where was Mind? 
And if this tablet were writ upon in language, then where was language? Void 
questions all! So rationalism survived until 1849, when Claude Minié invented 
the conical bullet. 
Before 1849, bullets were musket balls. Musket balls had a frustrating habit 
of curving unpredictably in flight, so prior to 1849, opposing armies would line 
themselves up, shoulder to shoulder, in order to give the opposing team a rea­
sonable chance. Even then, when a musket ball did happen to score, it tended 
to shatter the skull, causing massive damage to the brain beneath. Minié’s 
conical bullet, on the other hand, flew true. Even better, it was frequently able 
to create a surgically clean hole in the skull and a nice, focused wound (a focal 
lesion) in the underlying brain tissue. 

LOUGHT  AND  THANGUAGE 
•
7
 
As a result of this technological advance, a young doctor in France, Pierre 
Paul Broca, obtained a sizable cohort of war casualties whose brain lesions dis­
turbed their ability to speak but otherwise left the patients’ minds and be­
haviors remarkably intact. In 1861, Broca presented the discovery that such 
aphasia occurred almost exclusively when injury was sustained to a relatively 
limited area of the left half of the brain. Several years later a Viennese doctor, 
Karl Wernicke, discovered that injuries to another region on the left side of 
the brain caused a second kind of aphasia. Whereas “Broca’s aphasics” had 
difficulty speaking but relatively little difficulty comprehending language, 
“Wernicke’s aphasics” had no difficulty speaking but great difficulty compre­
hending. Where is language? had seemed a void question, but suddenly—and 
quite unexpectedly—it had an answer. 
Where Language Is 
Language was in the brain! This finding, utterly implausible to the ancients, 
was supported by copious and irrefutable evidence: spoken output was gener­
ated in Broca’s area, and heard input was processed in Wernicke’s area. The 
scientific community instantly and earnestly undertook the study of the brain. 
It was no longer the seventeenth century. Leeuwenhoek had long since 
invented the microscope, and within a generation of Broca, scientists had 
trained it on the brain. In 1873, Camillo Golgi discovered that chromium-
silver salts would selectively stain brain cells, thus making them clearly visible 
under the microscope. Using Golgi’s staining method, Santiago Ramón y Cajal 
charted the microstructure of the brain in encyclopedic detail, and by the dawn 
of the twentieth century, it had become an established scientific fact that mind 
was brain. And since brain was made up of white matter and gray matter, mind 
was matter. Rationalism was dead. 
For their discovery of the brain’s previously invisible structure, Golgi and 
Ramón y Cajal were awarded the 1906 Nobel Prize.
2
 Their work also engaged 
them in a famous debate. Ramón y Cajal believed each cell was a separate cell, 
wholly bounded by its cell membrane and unconnected to its neighbors, but 
his microscopes weren’t powerful enough to prove it. On the other hand, 
Galvani had long before shown that electricity made a dissected frog’s leg twitch. 
It could therefore be readily inferred that there was electrical communication 
among nerve cells. But how could electrical impulses be transmitted if the wires 
weren’t connected? Golgi maintained that the myriad cells of the nervous sys­
tem must form a continuous network. 
In the early 1900s many more researchers joined in this debate. Using ever-
more-powerful microscopes, they took ever-closer looks at nerve cells. In the 
end, Sherrington, Adrian, Dale, Loewi, and others proved that Ramón y Cajal 
was right, earning in the process Nobel Prizes for their efforts. Neurons were 
discrete cells separated by a synaptic gap. This gap was small, but it was big 
enough to electrically insulate each cell from the next. So how did neurons 
pass their messages across the synapse? They passed their electric messages 

8
•
 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
using chemicals, called neurotransmitters. Doubt of the world and belief in truth 
were now clearly behaviors of the brain: 
Thus, both doubt and belief have positive effects upon us, though very differ­
ent ones. Belief does not make us act at once, but puts us into such a condition 
that we shall behave in a certain way, when the occasion arises. Doubt has not 
the least effect of this sort, but stimulates us to action until it is destroyed. This 
reminds us of the irritation of a nerve and the reflex action produced thereby; 
while for the analogue of belief, in the nervous system, we must look to what are 
called nervous associations—for example, to that habit of the nerves in conse­
quence of which the smell of a peach will make the mouth water. (Peirce 1877:9) 
Never Mind the Mind 
But Peirce was ahead of his time. Twenty years later, in America, Peirce’s “prag­
matic” perspective developed into behaviorism. Behaviorism came in many 
flavors, but one lineage descended from Peirce to Dewey to Thorndike to 
Watson to Lashley. In the formulation of John B. Watson, behavior could be 
observed and scientifically reduced to a series of stimulus-response events, 
“habits of the nerves,” occurring along a chain of neurons. Mind was just an 
unobservable and useless abstraction. All of creation, from the lowliest animal 
to the highest form of social organization (then widely believed to be either 
the assembly line or the Prussian army), could be pragmatically analyzed solely 
in terms of stimulus-response chains of command. Behaviorism, in the social 
form of totalitarianism, promised a well-regulated society in which every ani­
mal want could be provided by eliciting strict, learned, obedient responses to 
the stimuli of an all-powerful, all-loving dictator. 
Predictably, this utopian vision was especially popular among the ruling 
and managerial classes, who had never worked on an assembly line or directly 
experienced the new, improved, conical bullet. Many, following Herbert Spen­
cer (1862) and later “social Darwinists,” envisioned themselves to be “super­
men,” a new species which had evolved through natural selection to a point 
“beyond good and evil” (Nietzsche 1883). However, after World War II and the 
likes of Hitler and Stalin, this utopian vision began to lose some of its appeal, 
even among the controlling classes. In his 1948 utopian novel Walden Two, the 
celebrated Harvard behaviorist B. F. Skinner attempted to dissociate behavior­
ism from these infamous European practitioners. As Skinner spun the story, 
everyone—more or less regardless of race, creed, color, or, for that matter, 
genetics—could be educated to perfection through the application of “pro­
grammed learning.” In programmed learning, students were methodically 
rewarded for correct answers and punished for incorrect answers. In this way, 
it was believed that good habits would be efficiently “learned” and bad habits 
would be efficiently “extinguished.” 
In the United States, however, there was a new class of university students: 
World War II veterans whose college tuition was paid as a war benefit. These 

LOUGHT  AND  THANGUAGE 
•
9
 
students and vocal, war-hero labor union leaders let it be known that they did 
not consider any chain of command to be utopian. Whether on the front line, 
the assembly line, or the school registration line, they did not want to be pro­
grammed! By the mid-1950s, opposition to Skinner had become widespread, 
but it was inchoate. Behaviorism had been politically refuted by the European 
experiment with totalitarianism, but Skinner’s scientific authority as a Harvard 
professor was still unassailable, and there were no viable alternatives to his 
psychological theories. 
In 1957, amid mounting popular disdain for behaviorism, Skinner pub­
lished a scholarly book, Verbal Behavior. In it, he sought to show that behavior­
ism had developed far enough beyond the study of lab rat behavior to undertake 
the explanation of human language. In 1959, two years after the publication 
of Verbal Behavior, Noam Chomsky, a young linguist at the Massachusetts Insti­
tute of Technology, published a disdainful review of it in the journal Language. 
Not only did Chomsky find Skinner’s analyses of language naïve, but he found 
them to be proof of the vacuity of behaviorism in general. 
Skinner didn’t reply directly to Chomsky’s review, but he did write another 
book, Beyond Freedom and Dignity, to which Chomsky also gave a bad review. These 
reviews of Skinner and behaviorism made Chomsky an instant, popular cham­
pion of freedom and dignity, opening a new chapter in the confusion of thought 
and language. 
Finite Mind, Infinite Language 
Reaching back to rationalism for support, the thrust of Chomsky’s argument was 
that language was not a “thing” like a stimulus or a response, a punishment or a 
reward. Language was a unique—and uniquely human—module of mind. Thus, 
twentieth-century generative grammar became grafted onto a Cartesian dualism. 
The resulting generative philosophy has depended heavily on what I call the 
“generative deduction,” the basic form of which may be given as follows: 
(1a)  The human brain is finite, but
(1b)  an infinity of sentences exists,
(1c)  which can be generated by rule,
proving language is infinite. Nevertheless, 
(2a)  normal human children acquire language quickly and effortlessly, 
(2b)  even though no one teaches language to young children, 
(2c)  and only human children so acquire language. 
Therefore, 
(3)  language is innate. It is not so much learned as it is “acquired.” 

10  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
The premises of the generative deduction have come under attack from 
many quarters, but it has not yet been refuted. Consider, for example, Jacken-
doff’s 1994 witty defense of premise 1. First, Jackendoff opens the dictionary 
at random and generates a large number of sentences by a simple rule: 
A numeral is not a numbskull. 
A numbskull is not a nun. 
A nun is not a nunnery. 
. .  .  
These are all completely absurd, but they are sentences of English nevertheless. 
There will be something like 10
4
 × 10
4
 of them = 10
8
. Now let’s put pairs of these 
sentences together with since, like this: 
. .  . 
Since a numeral is not a numbskull, a numbskull is not a nun.
. .  . 
Since a numeral is not a numbskull, a numbskull is not a nunnery.
. .  . 
And so on it goes, giving us 10
8
 × 10
8
 = 10
16
 absolutely ridiculous sentences. Given 
that there are on the order of ten billion (10
10
) neurons in the entire human 
brain, this divides out to 10
6
 sentences per neuron. Thus it would be impossible 
for us to store them all in our brains. (Jackendoff, 1994:21) 
Although 10
16
 does not quite qualify as mathematical infinity, it certainly 
seems infinite for human purposes. This infinity of language was at the nub 
of Chomsky’s arguments against Skinner in 1959, and premise 1 of the gen­
erative deduction has stood unrefuted and irrefutable until the present day. 
For the past forty years, a variety of biologists, psychologists, teachers, and 
child-language researchers have contested premise 2, arguing that children are 
taught language and do in fact learn in the process. But premise 1 forms the 
basis for a strong logical defense of premise 2. Chomsky has introduced that 
defense with a different quotation from Peirce: 
You cannot seriously think that every little chicken that is hatched has to rum­
mage through all possible theories until it lights upon the good idea of picking 
up something and eating it. On the contrary, you think that the chicken has an 
innate idea of doing this; that is to say, that it can think of this, but has no fac­
ulty of thinking anything else. . . . But if you are going to think every poor chicken 
endowed with an innate tendency towards a positive truth, why should you think 
to man alone this gift is denied? (Peirce, quoted in Chomsky 1972, 92) 
Peirce called the ability to come up with new theories abduction, a logico­
cognitive process which he believed was more important than either of the 
logical processes of induction or deduction. Chomsky asked essentially the same 
question of children and language: one cannot seriously think every little child 
that is born has to rummage through all possible grammatical theories until it 
lights upon the one right way of making words into sentences. Language could 

LOUGHT  AND  THANGUAGE 
•
11
 
not be learned unless every child was endowed with an innate tendency toward 
a correct, universal grammar. 
Following Chomsky’s suggestions, researchers undertook a series of math­
ematical analyses, collectively referred to as “learnability theory,” to investigate 
the conditions under which language could be learnable (Gold 1965, 1967; 
Hamburger and Wexler 1975; Wexler and Culicover 1980; see Pinker 1984, 
1989, for approachable reviews). The gist of their argument was the following. 
If you say potayto and I say potahto, how is a child to learn which one to say? This 
argument becomes more convincing as one considers, not just the 5,000 or 
10,000 words that a child might memorize, but also the fact that the child knows 
how to transform these words à la Jackendoff into an infinite number of sen­
tences (premise 1 again). Chomsky’s seminal example was the “passive trans­
formation,” as of 1.4 into 1.5: 
John saw her. 
(1.4) 
She was seen by John. 
(1.5) 
Instead of 1.5, why doesn’t a child ever say 1.6*, 
*Saw by John was she. 
(1.6) 
or any of the other 118+ possible permutations of 1.5? “Because the child 
never hears those other 118+ permutations,” you may say. But the child has 
likely never heard the exact permutation which is 1.5, either. Nevertheless, 
every child has learned to produce passive sentences like 1.5 by the age of 
six or so (premise 2a). 
“Well, the child doesn’t memorize rote sentences,” you reply. “He remem­
bers patterns.” But exactly how does he remember patterns? No one in his right 
mind sits down and teaches a child of four that “to transform an active sen­
tence pattern into a passive sentence pattern, one positionally exchanges the 
subject and direct object, prefaces the subject with the word by, appropriately 
changes the grammatical case of the moved subject and direct object, precedes 
the main verb with the tensed auxiliary of be, agreeing in number and person 
with the new subject, and replaces the main verb by its past participle.” 
You might instead argue that the child learns language patterns by imitat-
ing adult speech, and this was in fact the explanation proposed by behavior­
ists. Unfortunately, child-language researchers quickly found that children don’t 
imitate adult speech. Consider the following, oft-quoted transcript from McNeill 
1966: 
Child  Nobody don’t like me.
Mother  No, say “Nobody likes me.”
Child  Nobody don’t like me.
Mother  No, say “Nobody likes me.”

12  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
Child  Nobody don’t like me.
Mother  No, say “Nobody likes me.”
Child  Nobody don’t like me.
Mother  No, say “Nobody likes me.”
Child  Nobody don’t like me.
Mother  No, say “Nobody likes me.”
Child  Nobody don’t like me.
Mother  No, say “Nobody likes me.”
Child  Nobody don’t like me.
Mother  No, say “Nobody likes me.”
Child  Nobody don’t like me.
Mother  Now listen carefully. Say “Nobody likes me.”
Child  Oh, nobody don’t likes me.
To maintain that language is “learned,” it appears one needs a better theory 
of learning than imitation. 
Although generative philosophy has demonstrated the failure of behav­
iorism to most observers, it has not been without its critics. For example, the 
claim that language is rule-based (premise 1c) extends back to the foundations 
of modern linguistics in the eighteenth century, but for forty years, nonlinguists 
have objected that language cannot be rule-governed, because semantics, the 
meaning system of language, is not rule-governed. After all, what rule could 
definitively tell you what I mean when I say I love you? But semantics has little 
to do with the generative deduction. Chomsky has argued that “such under­
standing as we have of [language] does not seem to be enhanced by invoking 
thoughts that we grasp, public pronunciations or meanings, common languages 
that we partially know, or a relation of reference between words and things” 
(1993, 25), and as Jackendoff’s A nun is not a nunnery illustrates, sentences can 
be grammatical even if they are meaningless. That is, leaving meaning aside, 
how is one even to explain syntax, if not as acquired through the agency of an 
innate, rule-governed system? 
Recently, many cognitive psychologists have attacked premise 1c by dem­
onstrating that pattern-based neural networks can exhibit linguistic behaviors 
similar to that of rule-based systems (Rumelhart and McClelland 1986a). But 
to date these demonstrations have been more semantic than syntactic. Also, 
the fact that rulelike behavior can be elicited from an artificial neural network 
does not preclude the possibility that the brain functions at some other, more 
interesting level like a rule-based digital computer. 
My discomfort with the generative deduction originated with premise 2a, 
that children learn language “effortlessly.” To be sure, childhood in middle-
class America in the latter half of the twentieth century has been mostly child’s 
play, but even privileged children display the temper tantrums of the “terrible 
twos,” and these are nothing so much as results of the child’s frequently frus­
trated efforts at communication. Nor do mommies and au pairs find the ter­
rible twos “effortless.” Nevertheless, the claim that toddlers learn language 
effortlessly seems never to have been challenged directly, and I am unaware 

LOUGHT  AND  THANGUAGE 
•
13
 
that generative philosophers have ever independently proposed an objective 
measure of child effort. The problem, no doubt, is that effort is an intrinsically 
subjective, “nonscientific” concept. Society devalues child labor because no one 
pays children a salary, and no one hears children complain—no one except 
mommies and au pairs, that is, but “scientific” society doesn’t pay them salaries, 
either. 
“Hard science” often tries to distance itself from such social issues, but when 
the object of scientific inquiry is language, it is hard to maintain distance. As a 
kind of compromise, sociolinguists (Ferguson and Slobin 1973) and “function­
alists” (Bates and MacWhinney 1982; MacWhinney 1987a) have attacked premise 
2b by redefining learning away from the narrow terms of behaviorism into more 
general terms of interaction in the social environment. We learn that the sky 
is blue, that birds fly, and that ice is slippery from the physical environment 
without a teacher, but no one claims this knowledge is innate. Sociolinguistic 
functionalism argues that we learn language from the social environment in 
much the same way. But how do we learn that birds fly and ice is slippery? Gen­
erative philosophers have justifiably objected that this sort of learning (a) is 
not itself well understood and so (b) barely begins to address deeper problems 
like how we understand the sentence I don’t think penguins can fly. 
Finally, biologists have often attacked premise 2c, the human uniqueness 
of language, citing dancing bees and signing apes as evidence of the evolution 
and learning of language in other species. Nevertheless, not even the proud­
est trainer invites his animals to cocktail parties. Whatever their language, 
animals’ language is still a far cry from human language. 
Although locally convincing, none of these attacks has proved generally 
fatal to the generative deduction, much less added up to a viable alternative 
theory of thought or language. Taken together, though, they indicate that 
something is amiss with the generative deduction. Forty years after first postu­
lating that children have an innate “language acquisition device,” generative 
philosophers have as yet been unable to find its place in human biology, and 
generative theory has found itself increasingly at odds with the rest of science 
and society. Chomsky himself has become defensive, asserting that “no one 
knows anything about the brain” (Chomsky 1988, 755), and asking, 
how can a system such as human language arise in the mind/brain, or for that 
matter, in the organic world, in which one seems not to find systems with any­
thing like the basic properties of human language? That problem has sometimes 
been posed as a crisis for the cognitive sciences. The concerns are appropriate, 
but their locus is misplaced; they are a problem for biology and the brain sci­
ences, which, as currently understood, do not provide any basis for what appear 
to be fairly well-established conclusions about language. (Chomsky 1994, 1) 
The preceding is neither a crisis for biology nor a crisis for linguistics; it is 
a crisis for Science. The assertion that no one knows anything about the brain 
may have been defensible in 1936, when Turing initiated “the study of cogni­
tive activity from an abstract point of view, divorced in principle from both 
biological and phenomenological foundations” (Pylyshyn 1979). It may also 

14  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
have been defensible in the late 1950s, when the foundations of generative 
philosophy were being laid. But since then, some two dozen Nobel Prizes have 
been awarded for discoveries in brain science. To date, at the end of the twen­
tieth century, some thirty-three Nobel Prizes have been awarded for discover­
ies about the human brain and nervous system (Ramón y Cajal in 1906, Golgi 
in 1906, Sherrington in 1932, Adrian in 1932, Dale in 1936, Loewi in 1936, 
Erlanger in 1944, Gasser in 1944, Hess in 1949, Békésy in 1961, Hodgkin in 
1963, Huxley in 1963, Eccles in 1963, Hartline in 1967, Wald in 1967, Granit 
in 1967, Axelrod in 1970, von Euler in 1970, Katz in 1970, Guillemin in 1977, 
Schally in 1977, Yalow in 1977, Sperry in 1981, Hubel in 1981, Wiesel in 1981, 
Levi-Montalcini in 1986, Cohen in 1986, Sakmann in 1991, Neher in 1991, 
Fischer in 1992, Krebs in 1992, Gilman in 1994, Rodbell in 1994). The prob­
lem today is not that “no one knows anything about the brain.” The problem 
is that we know so much about the brain and its abnormalities in so much detail 
that it becomes difficult to step back and see how the brain might do some­
thing so normal and so large as language. 
The Von Neumann Limit (
ν) 
The great mathematician John Von Neumann is credited with having invented 
the modern serial computer’s organization into “procedural memory” (pro­
gram) and “declarative memory” (data). Although Ramón y Cajal had won the 
1906 Nobel Prize for showing that the brain is a massively parallel processor, 
Von Neumann declared that “the nervous system [is] a computing machine in 
the proper sense, and that a discussion of the brain in terms of the concepts 
familiar in computing machinery is in order” (Von Neumann 1958, 75). 
Von Neumann went on to claim that since serial computers could do 
everything parallel computers could do and then some, they were in principle 
of design superior to parallel computers. It follows logically from this premise 
that the digital computer is, in principle of cognitive design, superior to the 
human brain, and that the computer scientist could be in this respect supe­
rior to God Almighty. Of course, most computer scientists have been too modest 
to make these deductions in public, but in the privacy of classified documents, 
these obvious implications sold a lot of computers to a world military/security 
establishment bent on being almighty (Roszak 1986). Flush with money and 
power from these contracts, Von Neumann and his followers overlooked one 
small factor in their calculations, however, death. 
It turns out that the fatal flaw in the generative deduction is in its least-
examined premise, premise 1a. While a serial computer might well be able to 
do everything a parallel computer can do, it can’t always do those things in 
the 10
9
-odd seconds of a human’s allotted lifetime. The reason every little child 
doesn’t have to rummage through all of the words and sentences his little head 
can hold is not so much that every little child is born with a language acquisi­
tion device endowing him with an innate tendency toward universal grammar. 

LOUGHT  AND  THANGUAGE 
•
15
 
It is that the human mind is infinite. To see how this is so, let us recall Jackendoff’s 
conclusion: 
And so on it goes, giving us 10
8
 × 10
8
 =10
16
 absolutely ridiculous sentences. Given 
that there are on the order of ten billion (10
10
) neurons in the entire human 
brain, this divides out to 10
6
 sentences per neuron. Thus it would be impossible 
for us to store them all in our brains. 
Recall how Jackendoff got to these numbers. He combinatorially paired 10
4
 words 
into 10
8
 simple sentences and then combinatorially paired those 10
8 
simple sen­
tences to create 10
16 
compound sentences. But if words and sentences can com­
bine, why can’t neurons combine? In fact, Ramón y Cajal showed that neurons do 
combine: each brain cell makes synaptic connections to thousands and thousands 
of other brain cells. Brain cells function in subnetwork combinations with other 
brain cells. And just how many subnetwork combinations can brain cells make, 
you ask? Well, it so happens there is a formula for combinations: 
n! 
———— 
(1.7)
(n – k)!k! 
If we assume that each of Jackendoff’s 10
10
 neurons is used in language 
and that each makes some 10
3
 synaptic connections, and if we further assume 
that every word is represented by 10
6 
connections combining in simultaneous 
activation, then with n = 10
13
 synapses taken in combinations of k = 10
6
, 1.7 
becomes 1.8: 
10
13
!
ν = ——————— ≈ 10
7,111,111 
(1.8)
(10
13
 – 10
6
)!10
6
! 
10
There is no computer large enough to compute the value of 
ν exactly, but 
7,111,111
 is a reasonable approximation.
3
 And how big is 10
7,111,111
? Well, 10
300 
is a generous estimate of the number of atomic particles in the known universe, 
so not only can your brain store Jackendoff’s 10
16
+ sentences, it can also store 
a name for every particle in the universe! And it will still have room for nam­
ing all the particles in 10
7,110,811
 more universes!! Compared to Jackendoff’s 10
16 
sentences—or the brief candle of a human life—your mind’s capacity is, for 
all human purposes, infinite. 
10
But are we to seriously think that every little child that is born has to rum­
mage through all possible 10
7,111,111
 combinations until it lights upon the one 
good grammatical idea for a sentence? Only if we process them serially. Be­
cause letter follows letter, word follows word, and sentence follows sentence, 
the wise men of history from Aristotle to Von Neumann supposed that serial 
language must be the product of a serial process: because language is serial, 
they supposed thought must also be serial.
4
 But unlike Von Neumann’s serial 
computer, it takes no more time for a parallel-processing brain to abduce 
7,111,111
 theories than it does to abduce 1 theory! Think about it this way. When 
you go to the zoo and see a zebra, do you start going through all the names in 

16  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
your head—aardvark, ant, anteater, antelope . . . —until you come to zebra? Of 
course not. Zebra comes to mind as quickly when you see a zebra as aardvark 
does when you see an aardvark. Neither is this trick of fast, “content-address-
able” memory unique to humans. When a lion sees an aardvark or a zebra, I’m 
sure it knows immediately what’s for dinner, even though it doesn’t have a 
printed menu. When a pigeon sees a hawk or a fox, it immediately decides 
whether to duck or to fly. In the meantime, the little serial computer, which 
had to rummage through all its possible plans of action, would have been lunch. 
If we are prepared to think that every poor pigeon is innately endowed with a 
fast, content-addressable, parallel-processing brain, why should we think to 
Homo loquens alone this gift is denied? 
Adaptive Grammar and the Plan of the Book 
It is important not to confuse thought and language. Just because language 
is manifestly serial, it does not necessarily follow that language must be com­
puted by a serial processor. Although language is serial, thought is parallel. 
Although Turing machines are serial processors, the human brain is a paral­
lel processor. 
While generative philosophy’s fundamental confusion of thought and lan­
guage cannot be ignored, the primary objective of this book is not an attack 
upon generative linguistics or artificial intelligence. Within their just premises, 
generative linguistics has identified numerous previously unnoticed cogni­
tive phenomena which demand explanation, and serial computers have often 
proved themselves to be valuable, if not very adaptive, devices. On the other 
hand, children are not Turing machines, and so long as children are needed, 
they need a theory of their own. What is needed is a theory of how human 
language, which functions to serve our minute-by-minute social adaptation, 
arose as part of the same adaptive, evolutionary process which led to Homo 
loquens. This book attempts to develop such a theory. Chapters 2–5 outline the 
theory’s foundations, from elementary evolutionary and biological principles 
of life through Stephen Grossberg’s adaptive resonance theory (ART). Chap­
ters 6–12 apply ART to language, creating in the process that specific applica­
tion of ART which, to give my critics a convenient target, I will call adaptive 
grammar. 
Adaptation occurs on many timescales. Over the ages, each phylum evolved 
as an adaptation to a changing Earth. Over the millennia each species evolved 
as an adaptation to changing habitats. These are the timescales on which ad­
aptation is usually discussed. But each individual human also must adapt to 
society over the course of a lifetime, and the neurons which encode our daily 
thought and language must adapt even faster, on a scale of seconds. English 
speakers, for example, must adapt every a or the to their listeners’ knowledge. 
Thus, we say a dog until our listener knows which dog we’re talking about, and 
thereafter, we say the dog. On these latter timescales, it is common usage to call 
adaptation learning. 

LOUGHT  AND  THANGUAGE 
•
17
 
Because adaptation is such a universal property of life, adaptive grammar 
is rooted in many disciplines, and this book must range from Cambrian to 
Cenozoic, from molecules to minds to language, across evolutionary biology, 
neurobiology, psychology, linguistics, mathematics, and computer science. Even 
if I could master all of these fields and their subfields, I would need to sim­
plify, and even still, every reader would find some chapter difficult and another 
chapter oversimplified. For this, and for the many lacunae in my own knowl­
edge, I beg the reader’s indulgence. In the end, however, a viable adaptive 
grammar will be the product of many minds, and to attract a broad resonance, 
I have sacrificed much detail to readability. 
Chapter 2 begins by taking a long, long view of adaptation and adaptive 
communication, starting with the evolution of the brain cell out of the “pri­
mordial soup.” That soup has left few fossils, so this is an admittedly specula­
tive view, but it is justified on three accounts. First, the story of evolution 
establishes the biological principle of self-similarity, which is needed to extend 
adaptive grammar beyond the simplest biological and linguistic examples. 
Second, imagining how the simplest, two- and four-celled brains evolved is a 
preliminary thought experiment in Grossberg’s method of minimal anatomies, 
the reasonable method of explaining complex neural processes in terms of 
simpler neural processes. This preliminary thought experiment shows how basic 
neural mechanisms that would eventually be needed for language had already 
evolved as early as 600 million years ago. Finally, the story of evolution coinci­
dentally gives the uninitiated reader a narrative thread that can tie together 
the numerous biological facts upon which subsequent chapters will build. 
Chapter 3 studies the single-celled organisms called neurons. But whereas 
chapter 2 was necessarily speculative, there is relatively little that is speculative 
about chapter 3. Twentieth-century science has laid neurons out plainly before 
us, photographed by electron microscopes and dissected with biochemical 
scalpels. With the rapid progress of biological science, every speculation in this 
domain quickly becomes a testable—and tested—hypothesis. 
The neuron is ultimately a social creature, so chapter 4 brings the neuron’s 
billion-year evolution to its culmination in the complex society of the human 
brain. Chapter 4 first looks at the large-scale organization of the brain as it 
appeared to Broca and Wernicke: hindbrain, midbrain, and forebrain, right 
hemisphere and left hemisphere, front and back. These views are now fairly 
familiar to the educated reader, but chapter 4 at its end takes some more un­
usual, less traditional views: it also looks at the brain top to bottom, inside out, 
and splayed flat. 
Once upon a time, there were four myopic neuroscientists. Peering through 
the microscope, they happened upon a brain cell. The first, observing a cell 
body shaped like a pyramid, said, “This is a pyramidal cell.” The second, ob­
serving the spines on the cell’s dendrites, said, “This is a spiny cell.” The third, 
observing its spherical neurotransmitter vesicles, said, “This is a spherical cell.” 
The fourth, noting the concentration of glutamate in those vesicles, said, “This 
is an excitatory cell.” Of course, they all were right, but by using ever-thicker 
lenses to study ever-smaller objects, our myopic neuroscientists eventually 

18  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
“could no longer see the forest for the trees.” Only a handful of neural net­
work researchers persisted in studying the neural forest that Ramón y Cajal 
discovered, explored, and charted. 
Following these researchers, chapter 5 turns off the electron microscope 
and uses Grossberg’s adaptive resonance theory to describe the midscale orga­
nization of the brain. Earlier theories, like behaviorism, were built on the de­
tails of how individual neurons behaved (as described in chapter 3), while later 
theories, like generative philosophy, were built on the organization of gross 
modules of the brain like Broca’s area and Wernicke’s area (described in chap­
ter 4). But these theories could not integrate their levels of description to ex­
plain how an entire brain full of neurons is capable of writing War and Peace— 
or, for that matter, guiding me to my office every morning—even though no 
individual neuron supervises the process. ART, by contrast, describes such a 
brain in terms of minimal anatomies that are organized on a scale of two to a 
few thousand neurons. Only when we understand how and why these minimal 
anatomies work do we begin to understand how and why larger systems like 
thought and language must work as they do. 
Adaptive resonance theory has been constructed principally by the analy­
sis of mathematical models. Insofar as adaptive grammar is more concerned 
with its linguistic validity than its mathematical validity, chapter 5 will only touch 
upon the mathematical foundations of ART. To help the general reader visu­
alize many of the most important features of ART—contrast enhancement, 
noise suppression, resonance, self-similarity, and neural rebounds—without 
close mathematical study, the main points of chapter 5 are presented by means 
of a graphical computer simulation. This will be a relief to many readers, but 
it would be wrong to suppress the mathematics entirely. Unlike first-year 
calculus’s thin gruel of missile trajectories, the mathematics of ART is humane 
and exciting, so I have tried to keep just enough of it to encourage the intrepid 
reader to venture into ART’s primary literature. 
Linguistics often views language as atomic sounds (phones) built up into 
the progressively larger structures of phonemes, morphemes, words, phrases, 
sentences, and discourse. It happens that this approach also suits the method 
of minimal anatomies, so chapters 6–12 follow the same general plan, apply­
ing ART to these successive levels of language, in the process constructing that 
corpus of explanations I call adaptive grammar. 
Chapter 6 is transitional between neuroanatomy and linguistics. It describes 
how the mouth produces phones and how they are received by the ear and 
passed along auditory nerve pathways to the brain. 
After the physical, physiological, and phonetic description of speech and 
speech sounds in chapter 6, the problem of phonemes is addressed in chapter 7. 
This is the basic problem of how you can say potayto and I can say potahto, yet 
we can both still mean the same tuber. Similar questions have been investigated 
by ART, but these have mostly been about vision—for example, how we can 
stably identify grass as “green” even though daylight itself changes in color from 
dawn to noon to dusk. At this phonemic level, where neurons in the minimal 
anatomies between the ear and Wernicke’s area process the sound spectrum 

LOUGHT  AND  THANGUAGE 
•
19
 
of speech, previous ART analyses map quite directly onto issues of linguistic 
perception. 
But speech spectra change more quickly than Apollo drives his team across 
the sky: language happens quickly in time, and serial order becomes a funda­
mental problem. For example, when children first learn to count, they tend to 
count “one, two, three, eight, nine, ten.” What happens to the middle of this 
series? Behaviorist accounts in terms of stimulus-response chains (one stimu­
lates two stimulates three, etc.) obviously had some missing links, but the issue 
is also critical for parallel models of cognition: how can a parallel brain encode 
serial order? Among major “connectionist” theories, only ART offers a suffi­
ciently detailed analysis of neural architecture to solve this fundamental lin­
guistic problem. Chapter 8 outlines the general solution. 
Intending to honor the queen, Rev. William Archibald Spooner, Fellow and 
Warden of New College, Oxford, is claimed to have offered a toast to “our queer, 
old dean.” In his memory, such quaintly twisted phrases now bear the slightly 
derisive eponym of spoonerisms. Psychologists, however, are not at all derisive. 
In 1951, it was the lowly spoonerism that spelled the beginning of the end for 
behaviorism. In that year, Karl Lashley (a student of J. B. Watson, the “Father 
of Behaviorism” himself) first noted the impossible problem spoonerisms posed 
for behaviorism: in spoonerisms, stimulus-response chains not only lose some 
links but they must also split apart and rejoin in totally nonhabitual recombi­
nations. Worse, Lashley noted that these recombinative reversals occurred fre­
quently and ubiquitously not only in disorders like dyslexia but also in normal 
speech and common behaviors like dance and tpying [sic]. Chapter 9 explains 
the spoonerism as a natural interaction of ART rebounds (chapter 5) and ART 
serial-learning anatomies (chapter 8). Rather unexpectedly, this leads to a 
deeply rhythmic analysis of word structure (morphology), one uncannily remi­
niscent of recent linguistic theories of “metrical phonology.” 
Lashley died in 1958, so it was left to Chomsky to administer the coup de 
grace to behaviorism. Chomsky saw that spoonerisms were not only isolated 
error phenomena but actually instances of a more general linguistic process 
called metathesis. In particular, Chomsky saw that sentences like 1.9 and 1.10 
could also be related by metathesis: 
Spoonerisms slew behaviorism. 
(1.9) 
Behaviorism was slain by spoonerisms. 
(1.10) 
Chomsky’s theory of linguistic metathesis was built on the mathematical 
principles of Alonzo Church’s lambda calculus (1941). The lambda calculus is a 
recursive grammar of algebra and the design specification for a kind of com­
puter called a pushdown-store automaton. Pushdown-store automata are especially 
suited to recursive operations upon data that are structured in binary trees, so 
Chomsky explained that sentence 1.10 was derived from 1.9 by a “passive trans­
formation” on syntactic trees, which moved behaviorism to the tree position of 
spoonerism and vice versa. Chomsky’s theories generated widespread and well­

20  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
funded enthusiasm: if recursion and pushdown-store automata could explain 
language, then certainly they could explain human intelligence in general! 
Although Chomsky cautiously distanced himself from such glib enthusiasm, 
his work became very much a cornerstone of a generation’s research in “artifi­
cial intelligence.” 
But if language is produced by brain cells, what could it mean to “move 
behaviorism to the tree position of spoonerism”? Could it mean that in transforming 
1.9 into 1.10, some behaviorism-cell and some spoonerism-cell actually exchange 
places in the brain? Of course not. So chapter 10, “Null Movement,” states the 
obvious and rejects Chomsky’s basic explanation of metathesis: nothing moves. 
It then extends the analyses of chapters 7–9 to develop an alternative explana­
tion of how words can “move” if neurons can’t. In place of movement, chapter 
10 borrows the serial organization principles of chapter 8 and organizes syntax 
around the neural representation of topicality. 
Chapter 11 is a kind of cadenza. As we progress from phoneme to mor­
pheme to phrase to sentence, we require larger and larger minimal anatomies 
to describe phenomena. When we finally reach the stage of social discourse 
and meaning, our anatomies are no longer minimal, and as we approach the 
Von Neumann limit, they take on the behavior of free will. Just as Einstein 
rejected Heisenberg’s uncertainty principle,
5
 many “hard” scientists today re­
ject the notion of free will, but when, as sometimes happens, the activity of 
1 subnetwork in 10
7,111,111
 causes an entire brain to change its “mind-set,” sci­
entific prediction and explanation can no longer be 100% accurate. In chap­
ter 11 the multiple themes of the earlier chapters become intertwined in ways 
that science is reluctant to entertain. Truth and meaning are explored as logi­
cal and social constructs, respectively, and self-similarity is reinvoked with the 
epigram “A human being is a neuron’s way of making another neuron.” 
Chapter 12, “What If Language Is Learned by Brain Cells?” is like ontog­
eny recapitulating phylogeny. The themes of the preceding chapters are reca­
pitulated, but this time from the perspective of the individual language learner. 
At the outset, the unborn fetus is not only affected by its genetic inheritance 
but also exposed to a raft of environmental hazards, so phenomena of disor­
dered language learning are treated first. When I began this book after many 
years away from the language disorders literature, I confess to having been 
skeptical of the many and varied complaints of “learning disability.” But as it 
evolved, adaptive grammar itself began to convince me not only that learning 
disabilities were real but also that nearly everyone is learning disabled. But just 
as the chapter appears ready to end with a prescription of learning pills for 
everyone, it returns to a reconsideration of learning—normal language learn­
ing, nature’s way of enabling a less-than-perfect assemblage of neurons in the 
human brain to adapt and survive. Our computers are wonderful and our 
medicines are wonderful, but it is most wonderful of all that nearly every 
human child, despite genetic defects and a hostile environment, learns a 
human language and survives. 

EVOLUTION
•
21
• 
T 
W 
O
• 
Jones’s Theory of Evolution
The Sanskrit language, whatever be its antiquity, is of a 
wonderful structure; more perfect than the Greek, more 
copious than the Latin, and more exquisitely refined than 
either, yet bearing to both of them a strong affinity, both in the 
roots of verbs and in the forms of grammar, than could possibly 
have been produced by accident; so strong, indeed, that no 
philologer could examine them all three, without believing 
them to have sprung from some common source, which, 
perhaps, no longer exists. 
Sir William Jones (1786) 
We identify transformational grammar with Chomskyan grammar, but ideas 
about language “transformations” were in the air well before Chomsky’s name 
was attached to them. We call the heliocentric solar system the Copernican 
system, but Aristarchus first proposed it nearly 2,000 years before Copernicus. 
We call the theory of evolution “Darwin’s theory of evolution,” but the central 
notion of evolution was proposed nearly 100 years before Darwin by the emi­
nent English philologist Sir William Jones. Following Jones, European philolo­
gists embarked upon the reconstruction of what they called the “Aryan language,” 
the ancestor of all modern Indo-European languages. (Because Hitler mis­
appropriated the term “Aryan,” this ancestral language is now called “Proto-
Indo-European.”) Within a generation, it had been conclusively proved that 
languages as diverse as English and Sanskrit had, in fact, descended from this 
now-extinct language. Since these philological reconstructions did not extend 
back more than a few thousand years, they did not challenge the biblical 
account of creation, and Jones’s theory of evolution quickly became established 
scientific fact. Within a generation, the philologists’ ideas and methods were 
adopted by Lamarck, Chambers, Wells, and many other predecessors of Darwin 
who proposed the evolution of nonhuman species within biblical time. So 
Darwin did not really so much invent the theory of evolution as apply it cor­
rectly to biology. 
But even after the philologists explained the descent of modern languages 
and Darwin explained the descent of man, there remained something about 
21 

22  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
humankind that evolutionary theories seemed still unable to explain. That 
something was descent of language—not just English or French or Chinese, 
not even Proto-Indo-European, but language itself. This great web of mean­
ing that we humans call language—can it not be called our soul, or at least our 
mind? And however might evolution account for that? For centuries, fabulous 
fakirs and sober scientists had attempted to produce horses that count and dogs 
that talk. In no case, however, did any succeed in producing or finding animal 
language that seemed anything like human language. Absent Darwinian evi­
dence that language has evolved gradually, generative philosophy set forth the 
bold assertion that language is a specific and uniquely human development, 
undescended from Darwinian nature. Generative philosophy allowed this de­
velopment to be physiologically associated with brain regions like Broca’s area 
and Wernicke’s area, but it implied that, in Chomsky’s words, any attempt to 
study language as having evolved from general animal intelligence was “adapta­
tionist hogwash” (Chomsky 1988). This further implied not only that behav­
iorism had been defeated by language but that its essential procedure, the 
extrapolation of research findings about animal behavior to human behavior, 
was sterile. 
Beginning in the 1960s, generative psycholinguists therefore began to 
explore language anew, as an autonomous module of mind. But by the mid­
1970s, there were some disturbing reports from these frontiers. Researchers 
into sign languages of the congenitally deaf (especially ASL, American Sign 
Language; Klima and Bellugi 1979) had established that these visual languages 
were capable of expressing human thought as completely as spoken language. 
We will revisit this topic in chapter 9, but since sign language does not involve 
primary speech cortex or hearing cortex, language began to look like an ad­
aptation of a more general intelligence.
1 
Then Philip Lieberman showed that apes and other quadrupeds had been 
incapable of speech, not because of some profound cognitive deficit, but sim­
ply because of the shape of their vocal tracts. As we shall see in chapter 6, speech 
sounds depend upon the shape of the vocal tract. When the human species 
assumed an erect posture, the vocal tract bent 90°, its acoustic properties 
changed dramatically, and speech became possible. But apes’ vocal tracts re­
mained unbent and physically incapable of producing most speech sounds— 
despite the best efforts of fabulous fakirs and sober scientists. 
Armed with these two pieces of evidence, ethologists stopped trying to 
teach apes to speak and started trying to teach them ASL. A series of chimps 
and gorillas (Gardner and Gardner 1969; Patterson 1978; see Premack 1985 
for a summary) rapidly learned vocabularies of hundreds of signs, and what 
is more, they learned to combine them into new signs and sentences. For 
example, upon seeing a swan, a chimp called Washoe combined the signs for 
water and bird. 
Thus, an evolutionary link was established between animal language and 
human language—at least in the minds of many biologists and ethologists. But 
generative philosophers remained unimpressed. To them, Washoe had signed 
nothing more remarkable than “there’s some water; there’s a bird.” Apes still 

JONES
’
S  THEORY  OF  EVOLUTION 
•
23
 
had not demonstrated the ability to generate novel and infinite sentences. 
Syntax remained a summit which only humans had scaled. 
Nevertheless, by the turn of the twenty-first century, even some of his MIT 
colleagues began to abandon Chomsky’s hard line (Pinker and Bloom 1990, 
Pinker 1994). It became acceptable, even in some linguistic circles, to say that 
language had evolved. The naked ape stood upright, bending his vocal tract. 
This gave him a significantly enriched inventory of “calls,” and the rest is his­
tory. Unfortunately, this account still does not explain some of the more re­
markable aspects of language, let alone of mind. The generative objection is 
that this account does not explain why, for example, we can say sentences 2.1 
and 2.2 but not 2.3: 
The man who is
1
 dancing is
2 
singing a song. 
(2.1) 
Is
2
 the man who is
1
 dancing singing a song? 
(2.2) 
*Is
1
 the man who dancing is
2
 singing a song? 
(2.3) 
For a deeper explanation of language, we must look deeper into evolution. 
We must go back in time, long before the hominids, long before the prehomi­
nids. In order to understand how human language and communication make 
survival possible, we must understand how intercellular communication among 
the first one-celled life forms made multicelled life forms possible. And to 
understand this, we must go back long before even the dinosaurs, back to when 
the only living things were rocks. 
Consider that a rock lives and dies. Take a common crystal of baking soda 
(sodium bicarbonate, NaHCO
3
), and drop it into a supersaturated solution of 
baking soda. Behold, the crystal grows. And while it is true that there is more 
to life than growth, note how this growth is like life. The growth of a crystal is 
self-similar: a small crystal grows into a large crystal, and just as a small boy grows 
into a big boy, its essential structure remains unchanged. 
Moreover, crystals reproduce. If we split our crystal in two, each half grows 
into a crystal which is also self-similar to its parent. And while it is true that 
there is more to life than growth and reproduction, consider also that our poor 
crystal can die. 
Baking soda is hardly a diamond among crystals, but it is spectacular in its 
death. Complete this elementary-school thought experiment by dropping a 
teaspoon of vinegar (flavored acetic acid) on our crystal. Immediately, the crys­
tal undergoes a fiery death, leaving behind only a pile of soda ash, a puddle of 
water, and a cloud of carbon dioxide. Thus, even the lowliest rocks have the 
essentials for a tale of life, death, and, sometimes, transfiguration. What is 
wanting is a little more personality, a salt of the earth with a little more spice. 
In 1951, looking for that spice, Stanley Miller concocted a primordial soup 
of water, methane, ammonia, and hydrogen. In his laboratory, he subjected 
this mixture to artificial lightning, seeking to simulate early conditions on the 
planet Earth. After a week, Miller began to find amino acids in his soup. At the 

24  • 
HOW  THE  BRAIN  EVOLVED  LANGUAGE 
time, only a handful of scientists appreciated what Miller had found, but in 
1953, Watson and Crick suggested “a structure for the salt of deoxyribonucleic 
acid” (DNA). This salt of DNA organized the assembly of amino acids into 
proteins, the stuff of life.
2 
One molecule of baking soda is just like any other molecule of baking soda. 
What gives proteins and DNA personality is their complexity. DNA chains four 
nucleic acid bases—guanine (G), cytosine (C), adenine (A), and thymine (T)— 
along a backbone of phosphate and deoxyribose-sugar molecules. Each sub-
chain of three bases defines a codon, which in turn defines one of twenty amino 
acids. Chains of codons form genes, and chains of amino acids form proteins. 
DNA is also sexy. One chain of nucleotides does not make a complete DNA 
molecule; it must be embraced by a second chain, its complementary image. 
When DNA reproduces, the two chains unwrap, and each builds a new mirror-
image partner (figure 2.1). 
It is still a long way from an amino acid to a sexy, self-replicating strand of 
DNA, and although Miller’s experiment has been superseded by much more 
sophisticated studies (e.g., Orgel 1979; Cairns-Smith 1985; see Dawkins 1986 
and Dennett 1995 for highly readable introductions to evolutionary biology), 
his experiment is still superb allegory, for Stanley Miller’s test tube did not only 
bring forth amino acids; it also brought forth that old villain, acetic acid. Life, 
death, and simple salts. The life of the first nucleic acids cannot have been much 
less treacherous than the life of a baking-soda crystal in a third-grade classroom. 
Even the earliest molecules of life had to be constantly on the lookout for killer 
chemicals. 
Despairing of the odds of bringing forth life in a hostile primordial soup 
(even given two billion years), I did not know how to continue this chapter. 
Then I gazed out my window at the rain, and I saw a test tube in every drop 
and a laboratory in every puddle, and when I multiplied billions of drops 
by billions of puddles by billions of years, the emergence of complex self-
replicating molecules no longer seemed so miraculous. Like a Polynesian ex­
plorer happening upon an island without snakes and without disease, all that 
life needed was one puddle of Eden, protected against predators by an imper­
meable barrier of rock. 
Nearly impermeable, that is, because at some point, on some day, some one 
self-replicating, complex molecule had to bravely go where no such molecule 
had gone before, and it had to survive. To survive outside paradise, it had to 
clothe itself; it had to bring a barrier with it. In billions of puddles over billions 
of years, probably billions of colonies of replicating molecules tried on many 
different clothes, but in a watery world, it was the cell that survived. 
The Cell 
In the cell, a bilayer of phospholipid hydrocarbon molecules created a puddle 
within the primordial puddle. A phospholipid hydrocarbon is a fatty molecule, 
one end of which is electrochemically repelled by water while the other is at­

JONES
’
S  THEORY  OF  EVOLUTION 
•
25