Kanzi: The Ape at the Brink of the Human Mind (30 page)

Patricia Greenfield disappeared for a few days, working nonstop to take in all the data that were available regarding the contexts surrounding Kanzi’s utterances. Unlike previous studies of apes who employed symbols, we had recorded the events surrounding Kanzi’s every utterance, as well as the utterance itself. It is not especially helpful simply to know that Kanzi said
Austin hamburger
—is he calling Austin a hamburger? It is helpful to know that he said this in response to being asked what he wanted to do after he pointed to the door. It is also helpful to know that after he said this, I asked, “Do you want to take some hamburger to Austin?” to which Kanzi responded with a chorus of positive
Waa
vocalizations. I told him he could get some hamburger out of the refrigerator, which he did. Then he gave it to me to carry as we went to visit Austin. When he arrived, I asked him if he still wanted to give the hamburger to Austin and he vocalized a positive
Whuh
, took the hamburger from me, and handed it to Austin. With details such as these recorded for all of Kanzi’s combinations, Patricia was able to make sense of what Kanzi had been trying to say.

When she returned, she said that she felt there was evidence of syntactical structure in Kanzi’s multiword utterances. I was surprised, as such structure was not apparent to me from talking with Kanzi. Patricia pointed out that you would not expect it to be, since the patterns appeared only when you
looked across many utterances that were dissimilar in meaning. She was anxious to write a scientific paper with me on this point. I hesitated, as I knew that we would have a difficult time convincing people that Kanzi used syntactical structures, given the reception the work had already received.

Most linguists were not prepared to grant language to another species, for such a view would undermine the very foundation of their field. Linguists assume that the structure of the human mind will eventually be laid bare by linguistics. They believe this will happen once linguistics reaches its goal of describing the underlying structures that unify all languages and drive language acquisition in all children, regardless of their culture. If apes, too, can be said to learn language, then language cannot be uniquely human and the goal of using it to characterize the basic structure of the human mind can never be attained.

Patricia had been challenging the assumptions of linguistics with child data, and now she was ready to do so with ape data. From where she stood, linguistics as a discipline was obsessed both with syntax as the mark of language and with its Chomsky-inspired certainty that humans are unique in possessing language abilities. At the very least, this sort of close-minded thought ignored evolutionary reality: We are sibling species with chimpanzees, with whom we share 99 percent of our genetic blueprint. In principle, therefore, the likelihood that chimpanzees and humans share at least some elements of language competence is rather high.

By its nature, the major body of linguistic theory deals with modern language, specifically its acquisition and structure. Few linguists appear to have realized that this perspective may distort the way questions are asked about the
origin
of language and the nature of its precursors in apes. It may be inappropriate to use the model of modern language structure as a guide for finding evidence of language in its earliest stages. For instance, it should not be a surprise to state that at one point in human history, language (and its syntax) was invented, not learned. In
which case, when looking for the precursors of language abilities in apes, one would be more realistic to look for the capacity for the invention of rules rather than the knowledge of existing rules. Furthermore, an evolutionary perspective of the origin of language implies that the rules would reflect the adaptation and behavior of the early humans who invented it. Similarly, if apes do have the fundamental ability to invent rules of syntax, those rules are likely to match an ape’s rather than a human’s mode of adaptation in some way.

This shift away from the anthropocentric, or human-centered, view pursued by strict linguists was an important conceptual guidance that Patricia brought to her analysis of Kanzi’s utterances. Her experience in cultural anthropology in Africa and Central America was vital in developing this perspective. She learned at first hand how the threads of a different culture influence its entire fabric, in material and linguistic realms. Insightfully, she applied this to the assessment of Kanzi’s language, insisting that we should take a bonobo’s perspective, not a human’s.

A second conceptual guidance we brought to our study enabled us to eschew the strict linguist’s question: Do apes have language? The answer to this question, of course, is that they don’t. But what would we say of a one-year-old human infant? Does she have language? A two-year-old? A five-year-old? Language emerges as the infant matures, and so it seemed reasonable for us to take a developmental approach in our study. In a chapter we contributed to a book called
“Language” and Intelligence in Monkeys and Apes: Comparative Developmental Perspectives
, we explained our perspective in this way: “We need to look for parallels between ape language and human language in the
earliest stage of development
and, having established these, see how far the apes can travel down the path toward human language.”
¹
In other words, we would not be looking to see if apes have language; instead, our search would be for what elements of language are present and what elements are absent at a particular stage of development.

Nevertheless, there are some fundamental criteria that a body of rules must meet if it is to be recognized as grammar, no matter
what the behavioral perspective or stage of development. In the previous decade, linguists and psychologists have offered various criteria, and Patricia and I assembled five of these as central.

1.
Each component must have independent symbol status
. For example, if Kanzi uses a combination such as
Chase dog
, it must be shown that both of these words can occur in other utterances, where they each have a different meaning, such as
Chase banana
and
Dog play-yard
. If all symbols are not free to combine with all other symbols in any manner, then it cannot be assumed that their union really is a legitimate instance of combining on the part of the ape.

2.
The relationship between the symbols must be reliable and semantic
. Some combinations such as
Ice TV
, which Kanzi made in order to ask that the television be turned on and ice be brought to him while he watched it, are combinations of things that Kanzi wants, but the two symbols lack an intrinsic grammatical relationship between them. It just happened that Kanzi put them together. However, an utterance like
Matata bite
, used to comment upon the first time his mother ever disciplined him by actually biting him, illustrates Kanzi’s ability to use symbols that have a meaningful relationship to each other. In such a case, unlike the
Ice TV
example, use of either
Matata
or
bite
alone would not convey half of the message. This is because the message is in the combining and the consequent relationship that is structured by the combination of one symbol with the other to convey a special meaning in a particular context.

3.
A rule must specify relations between categories of symbols across combinations, not merely a relation between individual symbols
. This means that a grammatical relationship such as action-object must be represented in the body of data by many different action terms and many different object terms (for example,
Grab Austin, Hide Austin, Bite ball, Chase ball, Hug ball
, and so on). When this occurs with a stable order between the categories of action and object, it can be legitimately inferred that Kanzi has some understanding of these categories themselves. Without such an understanding, he would be unable to order the combinations consistently by putting the action first and the object second across many different individual
symbols, all combined in different ways. In looking at Nim’s data, Terrace found ordering rules, but these rules were tied to specific symbols, some that Nim always put in the first position and others that he always placed in the second position. Kanzi, by contrast, formed combinations like
Grab Austin
and
Austin go
, illustrating that a symbol like
Austin
could be used in either position depending on how it functioned communicatively in the sentence.

4.
Some formal device must be used to relate symbol categories across combinations
. This means that given that Kanzi understands such broad categorical distinctions as agent and object, he must be able to demonstrate this understanding by ordering them in some reliable rule-based manner. In this case, the rule Kanzi uses is the same as that used by all other English speakers, the rule of order; place the action term before the object term. This does not mean that Kanzi is aware that he knows this rule, or even that none of his utterances ever violates it. It means only that at some primitive level Kanzi must have some sort of cognizance of the rule because he tends to use it.

5.
The rule must be productive
. This simply means that the rule must be able to be applied to new situations and to function communicatively in those situations from the first use. Thus, once Kanzi has learned the action-object rule, he should be able to apply it to produce appropriate combinations in new situations. An example of this is Kanzi asking someone to play
Tickle ball
, meaning to tickle him by rubbing the ball all over his body. This was a game that Kanzi invented, not us, as he loved the feel of the ball over his body. He followed the correct ordering rule as he made this new combination, suggesting that for him, the rule had become productive.

In a scientific paper on our work, we explained that “different subsets of criteria have been emphasized by different investigators.”
²
We had deliberately set ourselves a tough goal to reach.

Throughout the language program with Kanzi we kept a continuous record of his use of the keyboard, his gestures, and his
accompanying behavior. For the formal analysis of structure, we chose a five-month period, beginning in April 1986, when Kanzi was five and a half years of age. During that time, we recorded Kanzi’s 13,691 utterances, just a little more than 10 percent of which comprised two or more elements. About half of these multiword utterances met the criterion of spontaneity (that is, not produced in response to or in partial imitation of a caregiver), and so we finished up with a corpus of 723 combinations. This is a considerably larger body of data than any child studies have used to investigate grammatical development in humans.

We had already satisfied criterion 5, that of productivity. And we knew that Kanzi’s use of lexigrams met criterion 1, that is, independent symbolic status. From a very early age, Kanzi demonstrated an understanding of a one-to-one relationship between a symbol and an object or action. And we were satisfied that his use of the “go” gesture fulfilled the criterion of an independent symbol, just as such gestures do in deaf children who invent a sign language. In our study, therefore, we focused on criteria 2, 3, and 4, the reliability of the relationships between symbols used and the presence of a rule-based ordering to the symbols. We set out to examine whether there was a structure in Kanzi’s utterances.

One very clear rule that emerged during the five-month period concerned combinations that involved doing something to an object, such as Kanzi hiding a peanut or biting a tomato. During the first month, Kanzi showed no particular ordering of such symbols; sometimes he put the action first, sometimes the object.
Hide peanut
occurred just as often as
peanut hide
, for instance. But thereafter he began to follow a rather strict order, that of putting the action first and the object second:
hide peanut, bite tomato
, and so on. Kanzi apparently learned this rule from us, as we would touch lexigrams simultaneously as we spoke, and English generally employs an action-object ordering rule. Human children learn the rule at the two-word stage, somewhere between the first and second year.