14
‘The Evolution of Language: From Hand to Mouth
Michael C, Corballis
In 1866, seven years alter the publication of Darwin's Origin of Species,
the Linguistic Society of Paris famously banned all discussion of the evo-
lution of language. The main difficulty, it seems, was the widespread
belief that language was uniquely human, so that there was no evidence
to be gained from the study of nonhuman animals. This meant that lan-
guage must have evolved some time after humans split off from the great
apes. Because there was litte evidence to be gained from the fossil evi-
dence, any theory as to how language evolved was largely a matter of
speculation—and no doubt argument. Of course, evolution was itself a
contentious issue, and was attacked vigorously by the Church. In the
case of language, the conflict between science and religion would have
been exacerbated by the long-standing view that language was gifted by
God.
In more recent if not more enlightened times, the ban seems to have
lifted, but the contention remains. Although there are probably few
who would argue that language is a gif from God, there are still those
who maintain that language evolved in a single step, in all-ornone
fashion. This is sometimes called che “big bang” theory of language
evolution and has been most clearly articulated by the linguist Derek
Bickerton (1995), but itis also implicie in much of the writing of Noam
Chomsky. It smacks a litle of the miraculous, although the appeal i,
‘ous genetic mutation (eg, Crows 2002) or
to some emergent physical property rather than to God. Against this is
the view that language evolved in incremental fashion, through natural
selection, as maintained by Pinker and Bloom (1990) and elaborated by
Jackendof
of human evolution.
A related issue is whether the origins of language are to be
found in the behavior or communication systems of animals of more
mote likely to be t0 a
(2002). This view is clearly aligned with the Darwinian view404
pres 1
specifically, of our primate forebears. In 1966, Chomsky wrote as
follows:
The unboundedness of human speech, as an expression of limitless though is
an entirely different matter {from animal communication], because of the
freedom from stimulus control; and the appropriateness 10 new situations
‘Modern studies of animal communication s0 far offer no counterevidence 0 the
Cartesian assumprion thac human language i based on an entirely diferent prin
ciple. Each known animal communication system either consists of a fixed
number of signals, each associated with a specific range of eliciting systems oF
Snternal states ora fixed number of “linguistic dimensions," each associated with
1 non-lingustic dimension. (pp. 77~
Although Chomsky has also argued that language did not evolve
through natural selection, the idea that language is uniquely haman has
not precluded the notion that it is a product of natural selection. For
example, although Pinker and Bloom (1990s see also Pinker, 1994) argue
that language evolved in the hominid lineage through natural selection,
they agree with Chomsky that nothing resembling language has been
demonstrated in any nonhuman species,
Chomsky appears to have moderated his conclusion in a recent co-
authored article (Hauser, Fitch, &¢ Chomsky, 2002}, in which ie is argued
that there is a distinction berween what the authors call the faculty of
language in the broad sense (FLB) and the faculty of language in the
narrow sense {FLN}. It is clear that most animals, including primates,
are able to make communicative sounds and perform intentional acts, It
is also clear that primates can use sounds or actions in symbolic fashion.
For example, vervet monkeys give different warning cries ¢0 distinguish
between a mumber of different threats, such as snakes, hawks, eagles, or
leopards. When a monkey makes one of these cries, the troop members
act appropriately, clambering, up trees in response 10 a leopard call or
running into the bushes in response to an eagle call (Cheney & Seyfarth,
1990). These cries bear no obvious relation to the sounds emitted by the
predators they stand for, and in that sense they are symbolic. More com-
pellingly, perhaps, the bonobo Kanzi is able to use visual symbols on a
keyboard to refer to objects and actions; these symbols are again ak
in the sense that they were deliberately chosen by the human keepers so
as not to resemble what they stand for (Savage-Rumbaugh, Shankes, 8
Taylor, 1998).
Ree405
Recursion
“The Evolution of Language
Based on such arguments, Hauser et al. (2002) argue that FLB is shared
by other species, including birds and other mammals, although they also
point out that the use of symbols in these examples does not mean that
the symbols have all of che properties of words. The critical ingredient
that is missing from FLB, and that characterizes FLN, is recursion.
Recursion lies at the heat of grammat, and enables us to create a poten
tial infinity of sentences that convey an infinity of meanings. Recursive
language is well understood even by quite young children, as illustrated
by the well-known children’s stor
isis the house chat Jack built
This isthe male shat lay inthe house that Jack built,
This isthe rat that ate the malt that lay in the house that Jack built
This isthe eat that wosried the rat that ate the malt tha layin the house that
Jack buile
Young children quickly understand that the sentence can be extended ad
infinitum, The recursive rules of grammae also allow phrases co be moved
around instead of simply being tacked on to the beginning. For example,
if one wanted to highlight mait in the story, one could embed phrases as
follows:
Tie male chat the rat that the cat killed ate lay in the house that Jack built
Tr seems clear that this highly flexible, recursive property is absent
from communication among nonhuman species. Although many birds
emit long sequences of sounds, these ate essentially repetitive, born of
insistence, perhaps, rather than an attempt t0 convey new information,
The same is true of primates, and it is evident even at the acoustic level
that there is a variation and novelty in human vocal output that simply
does not exist in the vocalizations of primates, even great apes (Arcadi,
2000}. A visitor from Mars would soon discern that there is something
special about human vocal output, even if she had no understanding of
swhac was being said,
Protolanguage
Compared to the recursive sophistication of human language, animal
communication systems are at best only weakly combinatorial, For the406
Chapter 14
language to the great apes, and especially to our closest relatives, the
chimpanzee and bonobo. It soon emerged that chimpanzees are essen
tially unable to speaks in one famous example, a baby chimpanzee reared
in a human family proved able to articulate only three o four words,
and was soon outstripped by the human children in the family (Hayes.
1952). It was then realized that the failure to speak may have resulted
from deficiencies of the vocal apparatus, and perhaps of cortical control
cf vocal output, and subsequent attempts have
action and visual representations. For example, the chimpanzee Washoe
‘was taught over 100 manual signs, based loosely on American Sign Lan
guage, and was able to combine signs into two- or three-“word™
sequences to make simple requests (Gardner & Gartner, 1969). The
bonobo Kanzi has mastered the use of a keyboard containing 256
symbols representing objects and actions, and can construct meaningful
sequences by pointing to the symbols. He supplements this vocabulary
with gestures of his own invention. Although he makes fall use of this
vocabulary, his manual utterances appear ¢0 be limited to only two oF
three “words.” Surprisingly, chough, he has shown an impressive ability
to follow instructions conveyed in spoken sentences with as many as
seven or eight words (Savage-Rumbaugh et al, 1998).
“There scems to be a general consensus, though, that these exploits
ace not language. As Pinker (1994, p. 340) pur it, the great apes “just
don’t ‘get it."" Kanai’s ability so understand spoken sentences, although
seemingly impressive, was shown to be roughly equivalent to that of a
2'J-year-old gil (Savage-Rumbaugh et al. 1998) and is probably based
fon the extraction of two or three key words rather than on a fall decod-
ing of the syntax of the sentences. His ability 10 produce symbol
sequences is also at about the level of the average 2-year-old human, fn
bbuman children, grammar typically emerges between the ages of 2 and
4 years, so thac the linguistic capabilities of Kanzi and other great apes
is generally eaken as equivalent to that of children in whom grammar
has not yet emerged. Bickerton (1995, p, 339), who wrote that “[eJhe
chimps’ abilities at anything one would wane to call grammar were next
to nil,” has labeled this pregrammatical level of linguistic performance
“protolanguage.”
Bickerton has further suggested, howeves, that protolanguage may
be the precursor of true language, not only in development buat also in
evolution, an idea adopted by Jackendoff (2002) in a recent inSuential
book. Yer protolanguage has been taught to such diverse creatures asthe
great apes, dolphins, a sea lion, and an African Gray parrot, implying
based om manual407 The Evolution of Language
parallel evolution. Further, it has never been observed in the wild, An
alternative view, then, is that itis not a precursor to language but rather
is indicative of a general problem-solving ability. For example, chim
ppanzees have been observed to solve mechanical problems by combining
implements, such as joining ewo sticks together to rake in food that
would not he reachable using either stick alone (Kohler, 1925; Tomasello,
1996). The combining of symbols to achieve some end, such as food,
may in principle be no different
Nevertheless, prorolanguage may also be said to characterize the
skills of the 2-year-old child, as well as persons with
a following damage to Broca’s area—or perhaps, as
earch suggests, damage to the left precentral area of the
nsula, a cortical structure underlying the frontal and temporal lobes
(Dronkers, 1996}. Given that protolanguage seems to underlie language
both in development and in terms of neural representation, itis perhaps
reasonable co suppose that language did in fact grow out of protolan-
guage, whether conceived as a linguistic ability or simply as a capacity
for problem solving, in human evolution, The question then is, when did
this happen?
expt
.
&
The
2s
ee
=
ee
From Protolanguage to Language
Because great apes have not acquited any communicative skill beyond
prorolanguage, despite strenuous efforts to teach them, we can be rea
ly suse thar the language capacity evolved some time after the split
sen the hominid line and the line leading to modern chimpanzees
and bonobos, The carliest fossil skull tentatively identified as a bipedal
hominid is Sahelantbropus tehadensis, discovered in Chad, and dated
berween 6 and 7 million years ago (mya) (Brunet et al., 2002). This date
is probably very close to the time of the chimpanzee-hominid split, esti-
mated at between 6,3 and 7.7mya by a DNA-DNA hybridization tech:
nique (Sibley & Alquist, 1984). Another early fossil, Orrorin
tugenensis, is perhaps more securely identified as bipedal, and is dated
from between 5.2 and 5.8 mya (Galik et al., 2004). The early hominids
‘were distinguished from che great apes by facultative bipedalism, but
een: with respect to brain size and what litte is known of their cognitive
capacities, they probably were litde different from present-day chim.
panzees. There is ieee reason 10 suppose that grammatical
language emerged within the 4 or $ million years of early hominid
nplyics evolution,408
(Chapter 14
Dramatic changes began 10 occur beginning some 2mya, with the
emergence of the genus Horio. Stone tool industries have been dated
from about 2.5mya in Ethiopia (Semaw et al., 1997) and have been ten
tatively identified with H. rudolfensis. However, these tools, which
belong to the Oldowan industry, are primitive, and some have suggested
that H. rudolfensis and H. abilis, the hominid traditionally associated
with the Oldowan, should really be considered australopithecines (e.f
Wood, 2002). The true climb to humanity and to language probably
began with the emergence of the larger-brained H. erectus around
L.8mya, and the somewhat more sophisticated Acheulian tool industry
dating fom around 1.5mya (Ambrose, 2001). But even tool manufac-
ture may not be an especially good guide co the advance of cognition,
since the Acheulian industry remained fairly static for over a million
years and even persisted into the culture of early H. sapiens some
125,000 years ago (Walter et al., 2000).
Other changes associated with H. erectus may give a better guide
to the emergence of more sophisticated cognition, Erectus marked the
progression from facultative to obligate bipedalism and the full striding
sait characteristic of modem humans. From about 1.6mya, some
members of this species strode our of Africa and into Asia, and erectus
fossils in Java have been dated to as recently as 30,000 years ago (Swisher
etal, 1994), But perhaps the surest signs of intellectual advance have to
do with the size and development of the brain.
Bigger Brains
According to estimates based on fossil skulls, brain size increased from
457 ce in Australopithecus africanus, to 552cc in H. habs, to 854¢c in
carly H. erectus (also known as H. ergaster}, to 1,016ce in later H.
erectus, to 1,552 in H. neanderthalensis, and back to 1,355ec in H.
sapiens (Wood & Collard, 1999). These values depend partly on body
size, which probably explains why H. neanderthalensis, being slightly
larger than modern humans, also had a slightly lager brain, but the
picture is clearly one of a progressive increase, first clearly evident in
early Homo.
Indeed, Chomsky (1975, p. 59) has suggested that language may
have arisen simply as a consequence of possessing an enlarged brain,
without the assistance of natural selection:
We know very little about what happens when 10" neurons are crammed into
something the size of a basketball, with forther conditions imposed by the
409408 The Evolution of Language
specific manner in which this system developed overtime. It would be a &.
erxor co suppose that all properties, or the interesting seuctures that eval
can be “explained” in terms of aatueal selection.
Nevertheless, the increase in brain size itself may have depended on
natural selection, and recent research has brought to light two genetic
mutations that may have a bearing on this. It is of some interest that
both mutations resulted in the inactivation of genes, suggesting that we
may owe our humanity at least in part to the Joss of genes rather than
to the incorporation of new ones.
‘One of these mutations has to do with a gene on chromosome 7
that encodes the enzyme CMP-N-acetylneuraminic acid (CMP-NewSAc)
hydroxylase (CMA). An inactivating mutation of this gene has resulted
in a deficiency in humans of the mammalian sialic acid N-glycolyineu-
raminic acid (NeuSGe). This acid appears to be absent in Neanderthal
fossils as well as in humans, bat itis present in present-day primates. It
also seems to have been downregulated in the chimpanzee brain, and
through mammalian evolution, leading to speculation that inactivation
of the CMAH gene may have removed a constraint on brain growth in
‘human ancestry (Chou et al., 2002). Chow et al. applied molecular clock
analysis to the CMAH genes in chimpanzees and other great apes, as
as to the pseudogene in humans, which indicated that the mutation
occurred some 2.1 mya, leading up to the expansion in brain size.
The other inactivating mutation that also may have contributed to
the increase in brain size has to do with a gene on chromosome 7 that
encodes for the myosene heavy chain MYHI6, responsible for the heavy
‘masticatory muscles in most primates, including chimpanzees and goril
feo Jas, as well as the early hominids. Molecular clock analysis suggests that
hes = this gene was inactivated around 2.4 mya, leading to speculation that the
diminution of jaw muscles and their supporting bone structure removed
a further constraint on brain growth (Stedman et al. 2004). Ie is a matter
of further speculation as to why this seemingly defeterious mutation
became fixed in the ancestral human population. It may have had to do
with the change from predominantly vegetable diet to a meat-eating
‘one, or it may have had to do with the increasing use of the hands rather
than the jaws to prepare food (Curtie, 2004),
R= More generally, however, these mutations, and the resulting
bra. increase in brain size, may have been selected because of a change in
envionment. With the global shift to a cooler climate after 2.Smya,
much of southern and eastern Africa probably became more open and
sparsely wooded (Foley, 1987). This left the hominids noc only more
RUAPDE OUR EU Reman ae410
Chapter 14
exposed to attack from dangerous predators, such as saber-tooth cats,
Tions, and hyenas, but also obliged to compete with them as carnivores.
The solution was not to compete on the same terms bat to establish whar
Whiten (1999, p. 175) has termed a cognitive niche, relying on social
cooperation and intelligent planning for survival. It scems reasonable to
suppose that language evolved in this context. Language is expensive in
terms of neural circuitry as it takes up a good deal of the left hemisphere
{and some of the right) in modern humans (Dick et al, 2001), so selec-
tion for mote sophisticated language may itself have been one of the
drivers of inereasing brain size.
Although Bickerton (1995) and others bave argued that language
evolved in a single step from protolanguage, it is more likely that ie devel:
oped gradually, through progressive refinements based on natural selec-
tion. One scenario has been proposed by Jackendoff (2002) and is
summarized in figuce 14.1. Although recursion is not explicitly men-
tioned in Jackendoff’s scheme, it seems likely that phrase structure may
have evolved before the flexible, recursive properties of grammar
emerged, allowing phrases to be sequenced and embedded according to
flexible rales.
Postnatal Growth
‘There is reason to suppose that recursive grammar depends not on brain
size per se but rather on postnatal growth of the brain. Human devel-
lopment is characterized by “secondary altriciality”; that is, the human
brain undergoes most of its growth alter bieth. In macaque newborns,
the brain at birth has a volume of about 70% of adult sizes in chimpan-
ees the figure is about 40%s in humans, itis only 25% (Coqueugniot,
Hublin, Veillon, Houet, & Jacob, 2004). The brain is at its most plastic
during growth, which may explain the so-called critical period for the
development of language. The optimal period is probably between 2 and
4 years, although there is evidence that grammar can be acquired later
in childhood (Vargha-Khadem et al., 1997}, but probably not after
puberty
The idea that growth may be critical to the emergence of recursive
grammar also gains some support from work with artificial networks.
Elman (1993) tried to determine whether a network with recurrent loops
could acquire the rules underlying sequences of symbols by testing
whether the network could learn t0 predict the nest symbol in the
sequence. The network was able to learn simple grammars but was at
4aut The Evolution of Language
Pre-existing poate conoeptulstuctre
Use of moos in a nonshuatonspec ation
/ \
Use olan open uninted elas of eyo CConcatnation of mooi
+ ‘
Development fs phonclelal combinatorial Use of symbol postion
system i eclargs ope, united Gass of to convey base semantic
symbos (posi silabian fst then ponomos) reiatone
N 4
(Powanainge extno)
$
Hierarchical phrase st
Ve
“
‘Symbols that expiety encode
‘Sbatract mantle relations
Y |
ee ee
\ Meaney Enctontoconroy
ae
‘| a
bertnd
Figure 14.1
Hypothesized evolutionary steps in the evolution of language. Sequential steps
sce ordered top to bottom; parallel, independent steps are shown side by sce
Steps unique ro humans are shown in bold type (after Jackendoff, 2002)
first unable vo deal with recursive grammars in which phrases were
phrases. This problem was at least partially sur-
‘mounted when Elman introduced a “growth” factos, which he simulated
by degrading the system carly on so that only global aspects of the input
‘were processed, and then gracually decreasing the “noise” in che system
so that it was able to process more and more derail. When this was done,
embedded in othe
the system was able to pick up some of the recursive quality of grammar,
and $0 begin to approximate the processing of true language.
‘One might expect secondary alteiciality to have evolved along with
the increasing brain size, since restrictions on the size of the birth canal
RN aChapter 14
‘would presumably have forced earlier birth; indeed, it has been estimated
thae if it were to conform to the general primate pattern, human bieth
should occur after 18 months of gestation, not 9 months (Krogman,
1972). Yer there is recent evidence that early H. erectus showed an
apelike pattern rather than a humanlike pattern of postnatal brain
growth, suggesting tha secondary altreiality may not have emerged until
fairly late in the genus Homo (Coqucugniot et al, 2004; Dean et al
2001). Evidence based on dental enamel suggests that this fearure may
not have been present even in H. neanderthalensis, bat had at least begun
to emerge in H. antecessor and H. heidelbergensis, the immediate fore-
runners of H. sapiens (Ramirez-Rossi & Bermudez de Castro, 2004). The
difference between H. sapiens and earlier Homo may lie not in the rele
ative size of the brain at birth but in the speed of brain growth.
Komarova and Nowak (2001) suggest that language is more accurate
the longer the period of acquisition, but this must be balanced against
the high cost of learning, The pressure may have been toward longer
periods of growth as language skills became more critical to biological
fieness
a summary, the expansion of brain size from some 2mya may well
have signaled the beginnings of more sophisticated language and the
emergence of phrase stsucture and grammatical categories. Bur the emer
gence of fully recursive language may not have come about until later in
the evolution of Homo, when postnatal growth was characterized not
only by secondary altriciality but also by a slowing of postnatal growth,
This scenario may well provide an evolutionary framework for the
sequential processes of language evolution outlined by Jackendoff
(2002)
Did Language Evolve from Manual Gesture?
Language is often equated with speech. Yer it has become inceeasingly
evident that the signed languages invented by deaf communities down
the ages have all of she grammatical and semantic sophistication of
speech (Armstrong, Stokoe, 8 Wilcox, 1995; Emmorey, 2002; Neidle,
MeLaugblin, Bahaw, & Lee, 2000). Chomsky (2000, pp. 121-122) puss
it succinctly
‘Though highly specialized, the language faculty is not tied to specific sense
modalities, contrary to what was assumed not long ago. Thus, the sign language
of the deaf is structurally very similar to spoken language, and the cour
acquisition is very similar. La
413ee — TT
ie
oR
413
The Evolution of Language
fon language acquisition... The analytic mechanisms of the language faculty
sccm to be triggered in much the same ways whether the input i auditory, visual,
even tactual, and seems t0 be localized in the same brain areas, somewhat
sutprisingly
Further, language is seldom wholly one or the other. Both manual
and facial gestures normally accompany specch and are closely syn-
c’chronized with it, implying a common source, and the gestures carry part
of the meaning (Goldin-Meadow & McNeill, 1999). The visible move-
_ments of the face can also influence the perception of speech, as in the
McGurk effect, in which dubbing sounds onto a mouth that is saying
something different alters what the hearer actually hears (McGurk &
MacDonald, 1976). Although we can communicate without having to
see the person we are talking to, as on radio or cell phone, speech in the
al world is rendered more eloquent and meaningful with the addi-
tion of bodily movements.
These considerations raise the possibility thar language itself might
have evalved from manual gestures rather than from animal calls,
although it may always have been a mixture of both—perhaps with ges-
tures punctuated by grunts gradually giving way to vocalizations embel-
lished by gestures. The idea that language may have its roots in manual
gesture goes back at least to the eighteenth-century philosopher Condil
Jac (1971/1746), but has been advocated many times since, often inde-
pendently (e.g., Armstrong, 1999; Armstrong et al., 1995; Corballis,
2002; Givin, 1995; Hewes, 1973; Rizzolatti 8 Arbib, 1998), From an
evolutionary point of view, the idea makes some sense, because nonhu:
man primates have little if any cortical control over vocalization but
excellent cortical control over the hands and arms. Human speech
required extensive anatomical modifications, including changes to the
vocal tract and co innervation of the tongue, and the development of
cortical control over voicing via the pyramidal tract, The vocalizations
of other primates are probably largely emotional, controlled by the
limbic system rather than the cortex (Ploog, 2002). The human equiva-
lents aze laughing, crying, shrieking, and the like. The modifications nec.
essary for articulate speech arrived late in hominid evolution, and may
not have been complete until the emergence of H. sapiens—or even later,
We have also seen that efforts to reach great apes anything resemabling,
speech have proven futile, but there has been reasonable success in teach-
ing them to communicate through gestures or by pointing to visual
symbols,
ae414
Chapter 14
‘The Late Emergence of Speech
Fossil evidence suggests that articulate speech emerged late in hominid
evolution, which gives further grounds for supposing that it may have
been preceded by a gestural system. One piece of evidence has to do with
the hypoglossal canal at the base of the tongue. The hypoglossal necve,
which passes through this canal and innervates the tongue, is much larger
in humans than in great apes, probably because of the important role of
the tongue in speech. Fossil evidence suggests that the size of the
hypoglossal canal in early australopithecines, and perhaps in H. babilis,
was within the range of that in modern great apes, while that of the
Neanderthal and early H. sapiens skulls was well within the modern
‘human range (Kay, Cartmill, &¢ Barlow, 1998), although this has been
disputed (DeGusta, Gilbert, & Tamer, 1999). A further luc comes from
the finding chat the thoracic region of the spinal cord is relatively larger
in humans than in nonhuman primates, probably because breathing
during speech involves extra muscles of the thorax and abdomen. Fossil
evidence indicates that this enlargement was not present in the early
hominids or even in H. ergaster, dating from about 1.6mya, but was
present in several Neanderthal fossils (MacLarnon & Hewitt, 1999).
‘The production of articulate speech in humans depends on the low-
cring of the larynx. According to P, Lieberman (1998; P. Lieberman,
Crelin, 8 Klatt, 1972) this adaptation was incomplete even in the Nean-