2 Animal Communication
from: Language Files, Sixth Edition, 1994
ed. by Stefanie Jannedy, Robert Poletto, & Tracey L. Weldon
Department of Linguistics
[scanned copy—there may be newly introduced errors]
File 2.1 True Language?
are not the only creatures that communicate. All varieties of birds make short
calls and sing
songs, cats meow to be fed or let outside, dogs bark to announce the arrival of strangers or growl and
bare their teeth to indicate their intent to attack, and so on. The fact that other animals send and re-
ceive messages is in evidence all around us. One approach to determining the nature of language is to
investigate the way other animals communicate and to explore the possibility that some species use a
system that is fundamentally the same as human language. Most people assume that only humans use
language—it is something that sets us apart from all other creatures. But is it possible that when we
examine animal communication systems we will discover our assumption was wrong?
task of comparing human language with various animal communication systems is
an easy one. First, we need a suitable working definition of "language" on which to base our compari-
sons. Unfortunately, no definition seems to adequately define language or be agreeable to everyone.
One approach to getting around this problem, suggested by the linguist Charles Hockett, is that we
identify some descriptive characteristics of language rather than attempt to define its fundamental na-
ture. Then we can determine whether a particular animal communication system exhibits these char-
acteristics as well. His list of characteristics, known as "design features," has been modified over the
years, but a standard list is provided below. From what we now know about animal communication
systems, we have found that none possesses all of these features, and thus we conclude that no nonhu-
man species uses language. Instead, they communicate with each other in systems called signal codes.
All communication systems have some features in common:
1. A Mode of Communication. This refers to
the means by which the messages are transmitted.
The mode of communication may be vocal-auditory, as in most human and most animal sys-
tems—the signals are transmitted by sound produced in the vocal tract and are received by
the auditory system. The mode may be visual (e.g., apes' gestural signals), tactile (e.g., bees),
or even chemical (e.g., moths).
2. Semanticity. The signals in any communication system have meaning.
Function. All systems of communication serve some useful purpose, from
the species to stay alive to influencing others' behavior.
Some communication systems exhibit these features as well:
This refers to the ability of individuals to both send and receive messages.
Human language exhibits this feature because each individual human can both send messag-
es (usually by speaking) and comprehend the messages of others (usually by listening). But
not all animals can both send and receive messages. For example, the Bombyx mori (silkworm)
moth uses a chemical communication system. When the female is ready to mate, she secretes a
chemical that males can trace back to her. The males themselves cannot secrete this chemical;
they can only be receivers.
--end page 20--
5. Cultural Transmission.
This is the need for some aspect of a communication system to be
learned through communicative interaction with other users of the system. Human language
exhibits this feature because humans must learn languages (even though the ability to learn
those languages is innate). Thus a child of Russian parents will learn English if that is the lan-
guage it is exposed to. In most organisms, the actual signal code itself is innate, or genetically
programmed, so an individual can no more leam a code different from the one it is pro-
grammed for than it can grow an extra eye. However, in a few systems, including certain bird
songs and chimpanzee signals, some of the signals seem to be genetically programmed or in-
stinctive, while others are learned. Therefore these systems, too, exhibit cultural transmission.
Humans, of course, must leam all the signals of their language.
This refers to the property of having signals for which the form of the signals
not logically related to its meaning (see File 1.6). The word cat, for example, does not sound
like a cat or represent a cat in any logical way. We know what the word cat refers to because we
learned the word as English speakers. If we were Spanish, the word would be gato, and if we
were Russian it would be koshka. When the relationship between a signal and its meaning is
arbitrary, then, there is nothing inherent in the form that designates its meaning. The meaning
must be learned. Most animal systems use iconic signals that in some way directly represent
their meaning, for instance, when a dog bares its teeth to indicate it is ready to attack.
7. Discreteness. This is the property of
having complex messages that are built up out of smaller
parts. Consider a sentence from a human language. It is composed of discrete units, indepen-
dent words, which are in turn composed of even smaller discrete units, individual sounds.
The messages in the animal communication systems with which we are familiar do not have
this property. Each message is an indivisible unit. Even when some animals imitate human
sounds so well, say parrots for example, these animals are merely memorizing a whole se-
quence that they reproduce, but they cannot break down the sequence into its discrete units. A
parrot trained to say Polly want a cracker and Don't go in there! will never recombine the words of
the sentence to say Polly don't want a cracker or recombine the sounds involved to say Scram, rat!
True language has, in addition to the above, the following characteristics:
This refers to the ability to communicate about things that are not present in
space or time. In human language, we can talk about the color red when we are not actually
seeing it, or we can talk about a friend who lives in another state when he is not with us. We
can talk about a class we had last year, or the class we will take next year. No animal commu-
nication system appears to display this feature.
9. Productivity. This refers to the
ability to produce and understand any number of messages that
have never been expressed before and that may express novel ideas. Human language is an
"open-ended" system. However, in all animal communication systems, the number of signals
is fixed. Even if some of the signals are complex (i.e., the system exhibits feature 7, discrete-
ness), there is no mechanism for systematically combining discrete units to create new signals.
These systems are thus called closed communication systems.
the comparison of human language with animal communication systems, a debate
en about whether the two systems are qualitatively or quantitatively different. If there is merely a quan-
titative difference, then we would find an animal system that possessed all of these features, but some
would not be present to the degree that they are found in human language. If, however, the two sys-
--end page 21--
tems differ qualitatively, we would find no animal
communication system that possessed each and ev-
ery design feature. While this seems straightforward enough, there is still some disagreement on the
application of this point.
the feature of displacement, for example. It seems as if the bees' signal code
this to a limited degree, since they communicate about food that is not visible while they are transmit-
ting their message (see File 2.2). But note that we can "translate" the message of their behavior in a
number of ways. We're in the habit of interpreting the bees' message as something like "there's a food
source 40 feet from the hive at a 45° angle from the sun." In other words, our translation assumes
they're relaying a message about a distant, invisible object. But the message can be represented differ-
ently—more simply, e.g., "perform this behavior now," that is, "fly 45° for 2 minutes." This is no dif-
ferent from most messages sent in animal systems. Think, for example, of a chimpanzee who adopts a
grooming posture. This communicates the chimp's desire for another chimp to perform a particular
behavior. The bees' messages are of this type—messages sent to alter the behavior of other individu-
als; their signals may not represent objects not present. Thus, some linguists claim the bees' system ex-
hibits limited displacement, while others maintain it does not possess this feature in any degree.
any rate, we say that a communication system must have all the design
features to be con-
sidered qualitatively the same as human language, and no animal communication system has been
identified to date that meets this criterion.
--end page 22--
File 2.2 The Birds and the Bees
File 2.1 it was claimed that no animal communication system is qualitatively
the same as human
language. No animal system with which we are familiar possesses all of the design features outlined
in the previous file. In this file we will investigate two animal communication systems in detail, that of
an Italian species of honeybee and that of the European robin, in order to support the claim that, al-
though enormously complex, animal systems are quite different from human language.
animals possess some kind of "signaling" communication system. For
the spiders there is a complex system for courtship. The male spider, before he approaches his lady
love, goes through elaborate gestures to inform her he is indeed a spider and not a crumb or a fly to be
eaten. These gestures are invariant. One never finds a "creative" spider changing or adding to the par-
ticular courtship ritual of his species.
similar kind of "gesture" language is found among the fiddler crabs.
There are forty differ-
ent varieties, and each species uses its own particular "claw-waving" movement to signal to another
member of its "clan." The timing, movement, and posture of the body never change from one time to
another or from one crab to another within the particular species. Whatever the signal means, it is
fixed. Only one meaning can be conveyed. There is not an infinite set of fiddler crab "sentences." Nor
can the signal be "broken down" into smaller elements, as is possible in any utterance of human
"language" of the honeybees is far more complex than that of the
spiders or fiddler crabs.
When a forager bee returns to the hive, if it has located a source of food it does a dance that communi-
cates certain information about that source to other members of the colony.
The dancing behavior may assume one of
three possible patterns: round, sickle, and tail-wag-
ging. The determining factor in the choice of dance pattern is the distance of the food source from the
hive. The round dance indicates locations near the hive, within twenty feet or so. The sickle dance in-
dicates locations at an intermediate distance from the hive, approximately twenty to sixty feet. The
tail-wagging dance is for distances that exceed sixty feet or so.
all the dances the bee alights on a wall of the hive and literally dances on
its feet through
the appropriate pattern. For the round dance, the bee describes a circle. The only other semantic infor-
mation imparted by the round dance, besides approximate distance, is the quality of the food source.
This is indicated by the number of repetitions of the basic pattern that the bee executes, and the vivaci-
ty with which it performs the dance. This feature is true of all three patterns.
perform the sickle dance the bee traces out a sickle-shaped figure eight on the
wall. The an-
gle made by the direction of the open end of the sickle with the vertical is the same angle as the food
source is from the sun. Thus the sickle dance imparts the information: approximate distance, direc-
tion, and quality (see Figure 1).
--end page 23--
Figure 1. The sickle dance. In this case the food
source is twenty to sixty feet from the hive.
tail-wagging dance imparts all the information of the sickle dance with one
dition. The number of repetitions per minute of the basic pattern of the dance indicates the precise dis-
tance: the slower the repetition rate, the longer the distance (see Figure 2).
Figure 2. The tail-wagging dance. The number of times per minute the bee dances a
pattern (1-2-1-3) indicates the distance of the food source. Reprinted
bees' dance is an effective system of communication, capable, in principle, of
many different messages, and in this sense the bees' dance is infinitely variable, like human language.
But unlike human language, the communication system of the bees is confined to a single subject. It is
frozen and inflexible. For example, an experimenter forced a bee to walk to the food source. When the
bee returned to the hive, it indicated a distance twenty-five times farther away than the food source
actually was. The bee had no way of communicating the special circumstances or taking them into ac-
count in its message. This absence of creativity makes the bees' dance qualitatively different from hu-
--end page 24--
The bees' dance does give us a chance to illustrate another
very interesting property that ev-
ery natural language of the world possesses, as previously discussed. We called this property the arbi-
trariness of the linguistic sign.
we say that the linguistic sign is arbitrary, we mean that there is no
the linguistic form and its corresponding linguistic meaning (see Files 1.6 and 2.1). There is no connec-
tion between the sounds of the word tree and the concept "tree." Likewise there is no connection be-
tween a red light and the notion "stop—danger." The relationship in this case is a cultural matter. In
all human languages the relationship between the sounds and meanings of the overwhelming majori-
ty of words is an arbitrary one.
about the bees' dance? What are the forms of the sign, and to what meanings do
correspond? Are the relationships arbitrary or nonarbitrary? Consider the tail-wagging dance. One
linguistic form is the vivacity of the dance, with a corresponding meaning "quality of food source."
The relationship is clearly arbitrary, for there is nothing inherent about vivaciousness that indicates
good or bad quality. In fact, we have been careful not to say whether more vivacity indicates a greater
or lesser quality source of food. Because the relationship is arbitrary, there is no a priori way of telling.
about distance? The question here is more complicated. Remember that the slower
repetition rate, the greater the distance. On the surface this relationship may seem arbitrary, but let's
use a little physics to reword the relationship: the longer it takes to complete the basic pattern, the
longer it will take a bee to fly to the source. Thus we see that this sign is in some sense nonarbitrary.
Similarly, the direction-determining aspect of the dance is perfectly nonarbitrary.
should be remembered, however, that there are many communication systems, other
language, which contain signs that are arbitrarily related to the meanings they stand for. "Arbitrari-
ness" is not enough to make a system a language in the sense of human language.
have talked about the "language" systems of spiders, crabs, and bees.
What about the
birds? It is known that the songs of certain species of birds have definite meanings. One song may
mean "let's build a nest together," another song may mean "go get some worms for the babies," and
so on. But the bird cannot make up a new song to cope with a new situation.
French scientists have studied the songs of the European robin. They found that
songs are very complicated indeed. But, interestingly, the complications have little effect on the "mes-
sage" that is being conveyed. The song that was studied was that which signaled the robin's posses-
sion of a certain territory. The scientists found that the rival robins paid attention only to the alterna-
tion between high-pitched and low-pitched notes, and which came first didn't matter at all. The
message varies only to the extent of expressing how strongly the robin feels about his possession and
how much he is prepared to defend it and start a family in that territory. The different alternations
therefore express "intensity" and nothing more. The robin is creative in his ability to sing the same
thing in many different ways, but not creative in his ability to use the same "units" of the system to ex-
press many different "utterances," all of which have different meanings.
songs, then, seem to be no more similar to human language than are the
the spider, the claw waving of the crab, or the dancing of the bees. All these systems are "fixed" in
terms of the messages that can be conveyed. They lack the creative element of human language.
study of higher animals also reveals no "language" systems that are
creative in the way hu-
man language is. Wolves use many facial expressions, movements of their tails, and growls to express
different degrees of threats, anxiety, depression, and submission. But that's all they can do. And the
sounds and gestures produced by nonhuman primates, the monkeys and apes, show that their signals
are highly stereotyped and limited in terms of the messages they convey. Most important, studies of
such animal communication systems reveal that the basic "vocabularies" produced by either sounds
or facial expressions occur primarily as emotional responses to particular situations. These animals
have no way of expressing the anger they felt "yesterday."
--end page 25--
The philosopher and mathematician Rene Descartes pointed out more than three hundred
years ago that the communication systems of animals are qualitatively different from the language used by humans:
It is a very
remarkable fact that there are none so depraved and stupid, without even
excepting idiots, that they cannot arrange different words together, forming of them a
statement by which they make known their thoughts; while, on the other hand, there
is no other animal, however perfect and fortunately circumstanced it may be, which
can do the same. ("Discourse on Method")
goes on to state that one of the major differences between humans and animals
that human use of language is not just a response to external, or even internal, emotional stimuli, as
are the grunts and gestures of animals. He warns against confusing human use of language with "nat-
ural movements which betray passions and may be... manifested by animals."
the studies of animal communication systems provide evidence for Descartes'
between the fixed stimulus-bound messages of animals and the linguistic creative ability possessed by
the human animal.
--end page 26--
File 2.3 Primate Studies
species of animals have communication systems that are much more complex than
imagine but that still appear to be very different from human language. The great apes (gorillas, chim-
panzees, and orangutans), for example, communicate with facial expressions, gestures, and calls to
express anger, dominance, fear, danger, acceptance in a group, and the like. But the precise meanings
of the various expressions of a particular band of apes are by no means easy for the human outsider to
decode. As complex as these systems are, they nevertheless lack displacement and productivity. Apes
apparently do not communicate about things that are not physically present, nor can they combine
their independent gestures or calls in novel ways to create new meanings.
great apes are, however, very intelligent creatures and Homo sapiens'
nearest relatives in
the animal kingdom. Chimpanzees, for example, are said to share 99% of their genetic material with
human beings. This biological similarity of ape and human, as well as the apes' intelligence, has
prompted some scientists to wonder if language could be taught to apes, even though it does not oc-
cur naturally. Many such projects have been conducted, most in the past twenty years or so. The ape
used most often has been the chimpanzee because it is considered by many to be the most intelligent
of the great apes and the most social, and the easiest to procure and handle. An orangutan and a goril-
la have also been used.
experiments have generated both exuberance and disappointment and, since their
ception, a debate about their results that continues to the present day. On the one hand, there are still
some scientists who maintain that they have indeed taught an ape human language. On the other,
there are many scientists who dispute this claim and have proposed alternative explanations for the
behaviors some researchers assumed could only have been language use. We will return to this debate
first such experiment conducted in the
wanted to raise a baby chimpanzee in a human environment to determine if the chimp would acquire
language on its own, just as a human child does, by virtue merely of being exposed to it. They there-
fore decided not to give training or "forcible teaching" to the chimp they acquired at 71/2 months and
whom they named Gua, other than that which would be given a human infant. Gua was raised along-
side the Kelloggs' newborn son, Donald, and the development of the chimp was compared to the
boy's. W. Kellogg stated that his intent was to determine how much of human language ability de-
rived from heredity and how much from education. He reasoned, a bit naively in retrospect, that what
the chimp could not learn would be those aspects of language that a human inherently knows.
Kellogg admitted one violation of this program when at one point he attempted to mold Gua's lips in
an effort to teach her to say papa. This effort proved unsuccessful. (They tried for several months.) The
duration of the Kelloggs' experiment was rather short in comparison to those that were to follow: only
the 1950s Keith and Cathy Hayes decided to raise Viki,
a female chimp, also as much like a
human child as possible, believing that with the proper upbringing a chimp could learn language. The
Hayeses believed that they could teach Viki to speak, even though doubt was emerging among scien-
tists at the time about whether the chimpanzee's vocal anatomy could even produce human speech
sounds. The Hayeses, however, believed that the vocal tract of the chimp was similar enough to a
--end page 27--
for it to be able to articulate human sounds. They had no aversion to
"training," and their
program included first teaching Viki to vocalize on demand (this alone took five weeks), and then
shaping her lips with their hands into various configurations that yielded consonant sounds. After
three years, Viki could "speak" three words: cup, mama, and papa, although they were accompanied by
a "heavy chimp accent"; it sounded as if Viki were only whispering. The Hayeses reported that Viki
could, however, "understand" many words, but they offered no experimental proof of this. The
Kelloggs' and Hayeses' experiments were not viewed by scientists as successful attempts to teach lan-
guage to apes.
words are not very many when one is trying to prove human language capability.
and Beatrice Gardner believed, contrary to the Hayeses, that chimps were not capable of producing
human speech sounds, so trying to teach a chimp to speak was fruitless. But since chimps are manual-
ly dexterous and use gestures to communicate naturally, the
Language (ASL) to a chimp they named Washoe. Washoe was not raised as a human infant but was
brought up with minimal confinement in a stimulating atmosphere. Spoken English was not allowed
in her presence; the
motivated to learn ASL. She also received deliberate training. Objects were presented to Washoe and
the trainers molded Washoe's hands into the shapes for their signs. Eventually, in order to be reward-
ed, she had to produce the signs herself and with greater and greater accuracy. The experiment was
considered at the time to be a great success. By the time Washoe was five years old, she had acquired
132 signs. More important, she had supposedly invented her own novel combinations, such as dirty
Roger, where dirty was used as an expletive, and water bird, upon seeing a swan on a lake.
searcher since them. Subsequent endeavors have all involved either ASL or invented languages that
used visual signs such as plastic chips or "lexigrams," symbols composed of geometric shapes.
1972 Francine Patterson began to teach ASL to a gorilla named Koko. This project has been
one of the longest of its kind, and Patterson has made some of the most dramatic claims for such a
project's success. According to Patterson, Koko knows several hundred signs and has invented many
of her own combinations, such as finger bracelet for 'ring'. In addition, she supposedly understands
spoken English—so well, in fact, that Koko occasionally rhymes, putting together such signs as bear
and hair even though the signs themselves have no visual similarity to each other. Koko also substi-
tutes homonyms for words when she cannot think of the sign, such as eye for I or know for no. Koko
also supposedly uses her signs to insult people and things she doesn't like. After being reprimanded
one day, for example, Koko is said to have called Patterson a dirty toilet devil. Patterson's most as-
tounding claim, however, is that "Koko is the first of her species to have acquired human language."
and David Premack began in 1966 to work with a
chimpanzee named Sarah. Their
methods were quite a bit different from those discussed above. Rather than treat the chimp like a hu-
man child, David Premack decided to try to find and use the best possible training procedure. The
"language" used was also atypical. Instead of ASL, Premack used differently shaped and colored plas-
tic chips. With each chip he arbitrarily associated an English word. Communication between the train-
ers and Sarah involved placing these chips on the "language board." Sarah was taught how to do one
type of "sentence" at a time. Typically, her task was to choose an appropriate chip from a choice of two
or to carry out a task indicated on the language board. Premack intended to teach Sarah the names of
objects as well as the names of categories of objects. He originally claimed to have taught her 130
signs, including category names such as color and concepts such as same and different.
Rumbaugh wanted to design an ape language experiment
with as much of the training
taken out of the hands of human trainers as possible. He reasoned that if the training were automated,
one could avoid cueing the animal and the training could be more efficient and constant and require
fewer humans, thus leaving more time for experimental verification of claims. He and his associates
--end page 28--
a computer that could execute certain commands, such as dispensing food or
slides, in response to an operator lighting up the proper symbols. The symbols of this invented "lan-
guage" were lexigrams—various combinations of nine different geometric figures, such as a big circle,
a little circle, and a large X. The operator of the machine was, of course, a chimp; her name was Lana.
Lana did learn to use her keyboard quite well, and Rumbaugh initially thought that he had succeeded
in teaching a chimp some human language.
Criticisms of the Early Projects
results and conclusions of these projects have been critically questioned on
two fronts. In the late
1970s, Herbert Terrace began a project similar to that of the
ously named Nim Chimpsky (hoping that when Nim learned language, the joke would be on Noam
Chomsky, the noted linguist who claimed such a thing was impossible). Terrace's concern was to
prove that a chimp could acquire and display some use of grammar. Terrace believed, as did most re-
searchers at this time, that evidence of human language capability was the use of grammar and not
just the use of signs. (Current researchers are concentrating on the way signs are used by animals.) By
the time Nim was four years old, he had acquired 125 signs, and Terrace felt Nim had indeed acquired
human language abilities as well. This project was the first to videotape all interactions between
chimp and trainer, however, and it was by reviewing these tapes that Terrace decided he must reverse
his initial claim and instead acknowledge that the ape's use of signs was very different from human
language. He noted that there were many dissimilarities between Nim's and a human child's acquisi-
tion of "language." Nim, for example, almost never initiated signing. Upon reviewing the tapes. Ter-
race found that only 12% of Nim's signs were spontaneous and a full 40% were mere repetitions of
what the trainer had just signed. This subtle interaction was never noticed by the trainer at the time. In
addition, Nim's signing was invariably a request for food or social reward; he never made unsolicited
statements or asked questions. Quite unlike a human child, he never took turns and was more likely to
interrupt his trainer's signing than not. There was also no evidence that Nim knew any grammar. His
combinations had variable word order, and more importantly, Nim rarely went beyond two-word
combinations; even when he did, the additional signs added no new information. For example, Nim's
longest utterance was give orange me give eat orange me eat orange give me eat orange give me you.
called into question the results of all previous experiments. He reviewed tapes
Washoe and Koko that had been made for a PBS special and concluded that they too had been cued by
their trainers. He and others leveled even more serious criticisms of the Premack project, arguing that
the training procedure taught problem solving and not language, and that Premack's conclusions
were not well founded, given his experimental design and his results. Consider Premack's claim that
Sarah learned the word insert. As proof of this Premack offered that in one task, when Sarah saw Sarah
banana pail insert on her language board, she correctly executed the task. When the word insert was
tested against the word give, however, Sarah could not distinguish the two. Premack likewise claimed
Sarah knew the prepositions on and in but never administered a test where Sarah would have to dis-
tinguish one from the other. Following instructions did not have to involve Sarah understanding a
sentence on the language board, but rather recognizing, for example, a banana chip and a pail chip
and imitating what she had been trained to do in the first stage of the test—in this case, insert the ba-
nana in the pail (a banana couldn't go on an upright pail.)
revelations had a great effect on the field of animal language studies. Funding
was thereafter hard to come by, and many scientists responded with new cynicism to any and all
claims of animal language researchers. Sue Savage-Rumbaugh maintains that both the initial easy
--end page 29--
of claims in this field and the post-Terrace cynicism are too extreme. She has
project with several chimpanzees and believes she has made the first real progress ever in teaching
some human language skills to an ape. She has, however, leveled a different criticism against previous
ape language studies. She believes that looking for evidence of grammatical capabilities in apes was
far too premature. She has considered a more fundamental and critical question: when apes use a sign,
do they know what it means ?
question is by no means easy to answer and it is ironic that the early
researchers in this
field took it for granted that when an ape produced a sign, it was using it in the same way humans do,
as an arbitrary symbol to represent something. It is precisely this use of the symbol that Savage-
Rumbaugh has considered and researched. Note that this approach represents a departure from the
attempt to assess animal language capabilities in terms of descriptive "design features," such as pro-
ductivity and displacement.
language does have these features that distinguish it from other communication
tems, but a more fundamental difference is perhaps that of symbol use. Understanding the concept of
symbol is difficult, partly because symbol use is innate to us. The use of a symbol involves a special re-
lationship with at least three components. First, there is the physical, external substance of a word
such as tree—either ink on a page in the shape of the letters t-r-e-e or a spoken word with a particular
acoustic pattern. Second, it has a relationship to a real tree somewhere, which is sometimes known as
the referent because the word tree refers to it. But, as you will see in File 7.1, not every word has a refer-
ent (e.g., love, Martian, Godzilla). The important relationship involved in a symbol is the mental repre-
sentation we have for the word tree, an idea of tree that is called up when we hear, say, or see the word
tree. Note that "mental representation" does not mean "image" or "picture"—not every word can
have one of these either. Whereas one might be hard pressed to say what the referent or mental "im-
age" of words such as "silly" or "love" is, we certainly possess some mental representation associated
with them—an idea of what each means. Mental representations have an existence separate from their
referents and can be manipulated independently of them. Thus we can think and talk about things
that are not present; in fact, we can talk about things that don't even exist (e.g., unicorns).
one disputes that humans use their words in this way. But how are we to know if
when it uses a sign in the same way we might, really has a mental representation for it? Savage-
Rumbaugh has suggested that in all previous experiments apes were not using their signs symbolical-
ly. Apes had merely learned to associate certain behaviors (making or seeing a particular sign) with
certain consequences (e.g., getting something to eat)—similar to a dog, for example, which, upon
hearing the word walk, knows it's going to get to go for a walk. This is an extremely subtle distinction
for humans to perceive, since the use of symbols comes naturally to us. We interpret other creatures'
signals to us in the same way we interpret those from each other, but that doesn't necessarily mean
they're intended in the same way. For this reason Savage-Rumbaugh has pointed out the necessity of
proper experiments that prove the ape has truly acquired a word in the same way a human has. She
has criticized the claims of previous projects either because they were not based on testing with prop-
er controls or because they had not been tested at all.
is one to find evidence of a mental phenomenon? One must still find it in the
the animal, or in the "processes of the exchange" with the trainer, but one must be more discriminat-
ing about what counts as evidence. In addition, Savage-Rumbaugh reasoned, apes had not learned to
use symbols given the training techniques used previously, which had assumed the symbol aspect of
sign use would come naturally to the ape; therefore, apes must specifically and intentionally be taught
Savage-Rumbaugh and her colleagues have worked extensively with
two male chimpanzees,
Sherman and Austin, attempting to teach them language skills with the computer and the language of
lexigrams used with Lana. They have found that the use of symbols by humans is not a single holistic
--end page 30--
but is a complex of independent abilities and behaviors. For example, the
ability to pro-
duce a symbol was found to be composed of at least three separate abilities. Using the association of a
lexigram and an object to request the object is one of these (and a display of it does not prove the user
has a mental representation for the symbol). Naming is another behavior that involves giving the lexi-
gram associated with an object without expecting to consume or receive it. The third ability involved
in production is called comprehension of the symbol. One might find it difficult to separate these three,
but to the chimpanzee they all had to be taught separately, and the presence of one ability could not be
assumed because of the presence of another. Savage-Rumbaugh points out the extreme importance of
another aspect of symbol use and human communication that had previously been overlooked: that of
the role of the receiver or listener. This in itself was also found to be a complex of skills and behaviors,
each of which had to be taught separately to Sherman and Austin.
Savage-Rumbaugh claims to have been successful at teaching the
chimps these skills as well
as the links between them (the coordination of these occurs naturally in humans). She has conducted
numerous experiments that indicate that the skills, as she has described them, have indeed been ac-
quired. She has acknowledged Terrace's criticisms of other projects but maintains that Sherman and
Austin do not evidence Nim's shortcomings. They take turns, their utterances are not imitations of
their trainers, and supposedly they produce messages not only because they must, but because they
wish to make statements.
project certainly has made real progress both in clarifying what human language
are and in teaching them to apes. Criticisms have been leveled, of course. Some suggest that, again,
the apes have been skillfully trained and still do not comprehend what they are saying or use their
signs symbolically. After all, it is perhaps impossible to know if another creature has a mental repre-
sentation for a word; we couldn't see it even if we opened up its brain. Savage-Rumbaugh might re-
spond that this criticism is a reflection of a cynical attitude rather than scientific considerations. How-
ever, given past experience and the tendency to overinterpret, there is a need to scrutinize the newest
claims in this field.
Savage-Rumbaugh's most recently begun project must be mentioned.
She has started to work
with another species of chimpanzee. Pan paniscus, which she claims is more intelligent than Pan troglo-
dytes, which has been used in all other projects. She claims that the new chimp she has been working
with, Kanzi, has learned to comprehend spoken English just by being exposed to it and has spontane-
ously begun to use the keyboard with lexigrams to make requests and comment on his environment.
Savage-Rumbaugh reports both anecdotal observations and the results of tests that might substantiate
these astonishing claims. Again, these newest claims are difficult to accept without further confirma-
tion and the demonstration of the kind of objective scrutiny and testing that was advocated at the in-
ception of the Sherman and Austin project.
--end page 31--