Comprehension, Production and Conventionalization in the

Origins of Language

Robbins Burling

University of Michigan

(May, 1998)

The Priority of Comprehension

This paper explores the implications of two observations that should be reasonably obvious, or at least familiar, but when they are considered together they lead to an unfamiliar but interesting way of thinking about the early stages of language. The first of the two observations is simply that all of us, humans and animals alike, are always able to understand more than we can say. Comprehension runs consistently ahead of production. The second observation extends the first: both humans and animals are sometimes able to interpret another’s instrumental behavior even when that other individual had no intention at all to communicate. In the first part of this paper I seek to justify these two observations. I will then consider their implications for our understanding of the origins of language.

Children, who appear to learn their first language with such magical ease, give us the most familiar example of the priority of comprehension. Parents are always convinced that their children understand far more than they can say. Linguists have occasionally been skeptical of the superior comprehension of children, partly because a vaguely behaviorist bias makes the "behavior" of speaking seem more important, than mere "passive" comprehension, but also for the much better reason that it really is very difficult to study comprehension. How do we know whether or not a child understands, and how do we know how he understands? Hold out a cookie to a child and ask "Do you want a cookie?" When he responds enthusiastically, how do we know whether he understands the words, or simply interprets the situation correctly? It is difficult to prove to the satisfaction of a linguist, let alone some kinds of hard nosed experimental psychologists, that children always understand more than they can say, but parents are rarely in doubt. At the time when one of my grandsons had a total productive vocabulary of exactly three words, one of which was a loud repeated grunt meaning "Give it to me", he could point appropriately not only in response to a request to show his eye, nose, or mouth, but also to show his elbow, knee or shoulder. He could point not only to a window or door, but to the wall, ceiling, or floor. He appeared to have a receptive vocabulary of hundreds, of words at a time when he articulated only three. Comprehension is so consistently ahead of production that we ought to recognize that much that is essential about language learning happens silently as children learn to understand. Speaking should be seen as merely the final step in a long process, the point at which language that is already under firm passive control is finally made active.

Even as adults, we understand more than we can say. We all understand dialects that we cannot produce. English speakers from the opposite sides of the Atlantic and from the southern extremities of the globe can generally understand each other with no more than an occasional hitch, but few of them would ever try to speak another’s dialect. We all understand words that we would not use. We understand some of the slang of ethnic groups or generations other than our own, even if we would not risk using it ourselves. We understand some technical terminology from fields with which we are only partially familiar. We understand, and even admire, rhetorical styles that we cannot, ourselves, duplicate. In New Guinea people have nice a way of distinguishing receptive and productive skill. They may say "I can hear that language but I cannot speak it", recognizing that it is possible to have a skilled ability to understand a language without the ability to speak.

If we had been clearer about the ability of human beings, both young and old, to understand more than they produce, we might not have waited so long to ask how much spoken human language nonhuman primates can learn to understand. Even if an ape is incapable of uttering a single spoken word, an ability to comprehend would demonstrate some genuine knowledge of a language. Anecdotal reports have suggested that captive chimps have sometimes learned to understand a good deal of spoken language even though they said nothing at all. These reports have sometimes been met with some skepticism for the same reasons that parental claims for their children’s ability to comprehend have been doubted, partly because production seems more real than "passive" comprehension, but also because it so difficult to measure skill in comprehension. Like people, apes can infer a great deal from the context in which language is used. It is always difficult to know how much any listener, even an ape, depends upon context, and how much upon the language. Hayes and Nissen suggest that Viki learned to understand a considerable amount of spoken English, but they were so eager to teach her to articulate words that they did not systematically study her comprehension (1971). As a result, Viki is remembered for her failure to speak, rather than for her success at understanding.

With the help of Savage-Rumbaugh and her colleagues, Kanzi, the famous bonobo, has now dramatically confirmed the ability of apes to learn to comprehend a significant amount of spoken language (Savage-Rumbaugh et al, 1993). At the age of eight, Kanzi was compared to a two year old human girl, and their ability to understand English was remarkably similar. Kanzi, like the girl, was able to respond correctly to a large number of different words and to a considerable variety of spoken sentences. Kanzi’s receptive skills give far better evidence of linguistic ability than has ever been shown by any nonhuman primate who has been trained to produce language or language-like signals, whether by articulating spoken words, by signing, by manipulating plastic chips, or by pressing buttons. Indeed, Kanzi’s ability to comprehend a human language seems sufficiently extensive that he should be credited with a degree of linguistic competence that linguists have most often presumed to be exclusively human. No one need fear that a bonobo or any other ape is about to give serious competition to human children in their speed or thoroughness of language learning, but I do not doubt that Kanzi has learned a good deal of English. The pattern is consistent. Not only humans of all ages, but apes as well, are always able to understand more than they can say.

Ritualization

Comprehension plays a crucial role in the origin of animal signals, for signals become communicative not when they are first produced, but only when they are first understood. The gestures and vocalizations by which animals communicate with one another develop from acts that were originally purely instrumental (Tinbergen 1952). Instrumental acts are the movements or noises that form a part of the ordinary business of living–moving around, eating, scratching, yawning. Although instrumental behavior is produced with no communicative intent whatsoever, conspecifics may still be able to interpret it. Only after such behavior has come to convey some sort of meaning to another animal, can it develop into a specifically communicative signal. A classic example is a dog's snarl.

Snarls began as simple instrumental gestures, nothing more than a part of getting ready to bite. The lip had to be moved out of the way of the teeth, but at first, the gesture had no communicative intent and probably no communicative result. Eventually, however, potential victims came to recognize the retracted lip as a signal that a bite was immanent. Those clever enough to read the signs would then be encouraged to flee, and so they could avoid the bite and live to reproduce. Comprehension, in other words, came before any communication was intended by the snarler. Comprehension was the first step but, once the victims were able to understand, the aggressor was presented with a new opportunity. By retracting his lip as if to bite, he might manage to frighten off his enemy but avoid the much riskier activity of really biting. It might even help to move the lip in a stereotyped or exaggerated manner and so reduce the sign’s ambiguity. As production and comprehension of the signal evolved together, the sign can be said to have become "ritualized", modified from a purely instrumental act into a stereotypic communicative signal.

The instrumental lip movement evolved into a communicative snarl, transmitting information that was useful both to the aggressor and to his potential victim. All this happened, of course, under the slow but relentless pressures of natural selection, and it required no individual learning. The term "phylogenetic ritualization" is sometimes used for this process so as to emphasize that signals like the snarl develop by slow evolution, not by rapid learning, but the point that I want to stress here is that the process has to start with comprehension. The ritualization of the lip movement could not even begin until it was understood. Other animal signs probably began much as did the snarl. Some sort of instrumental gesture or noise that was already being made for purposes other than communication was understood by other animals. Only then could it be ritualized into a specifically communicative signal.

By recognizing that comprehension has priority over production, both in our own language and in the origin of animal signals, we can start to solve a puzzle that has hovered over the first appearance of language: What could the first speaker have hoped to accomplish with her first words if no one else was around with the skills to understand her? The puzzle disappears as soon as we recognize that communication does not begin when someone makes a sign, but when someone interprets another's behavior as a sign. Comprehension must have been ahead from the very beginning. The original behavior that was understood in a language-like way could not have been intended as a sign at all. A lonely producer who tries out a new kind of sign will almost certainly fail to communicate. A lonely comprehender, on the other hand, may gain considerable advantage by being able to interpret another's actions even when no communication at all had been intended. At every stage of evolution, the selective pressures favoring skill at comprehension are likely to have been considerably more insistent than the selective pressures favoring skill at production. Producers often benefit by not giving themselves away. Comprehenders have little to lose and much to gain by understanding more.

The precocity of comprehension implies that at every point along the evolutionary path toward language, understanders needed to be ready before another complexity could be added to production. More accurately: The only innovations in production that can be successful, and so consolidated by natural selection, are those that conform to the already available receptive competence of conspecifics. At every point, production would have been limited by and directed by the ability to comprehend that was already found in the population. Only when others were able to understand, would a speaker be able to use new linguistic tricks This disposes of any mystery about the communicative usefulness of the first word-like signs. They would not even have been produced with communicative intent. Their communicative value came from the skill of the receiver, not from the intent of the producer.

The question that we should ask, therefore, is not "Why did the first speaker try to communicate if no one was around who shared his talents?" The answer to this is very simple: "He didn’t. It would have been useless". A much better question is "Why would anyone make word-like signs in the absence of any intention to communicate?" A plausible answer to this question is that the first interpretable language-like signs were instrumental acts. Once these instrumetnal acts could be interpreted by conspecifics, it became possible to conventionalize them as deliberate communicative signals. This implies that word-like signs could have had an origin that is quite similar to that of animal signals like the snarl, but there is one crucial difference. Almost all animal signals have been ritualized by the long process of natural selection. Early word-like signs, on the other hand, could have been conventionalized within the lifetime of a single individual.

The process that I am calling "conventionalization" is sometimes referred to as "ontogenetic ritualization" (Tomasello and Call 1997, 299-302). By using the word "ritualization", this phrase acknowledges the parallels between the origin of animal signals such as the dog’s snarl (phylogenetic ritualization) and the origin of signals that depend upon individual learning (ontogentic ritualization). I prefer to keep the jargon under at least partial control by calling the latter process "conventionalization" (or, when I want to be very explicit "ontogentic conventionalization") but, whatever it is called, it must be distinguished from the "ritualization" that is phylogenetic. If we are to find examples of conventionalization today, we should look for instrumental acts that can be interpreted by conspecifics, but that then become conventionalized as communicative signals. Such instrumental acts can be found among both humans and apes. Indeed, conventionalization can take place so easily that we hardly realize that it is happening.

Conventionalization

Consider, for example, the simple and familiar "arms-up" gesture by which toddlers ask to be picked up. This begins instrumentally. It is simply one part of a baby's adaptation to the impinging world, in this case a part of his interaction with bigger people. After being lifted often enough by adult hands that have been placed under his arms, a baby learns to spread and then raise his arms in anticipation. Adults, in turn, learn to recognize the gesture, and it quickly becomes conventionalized into a stylized request. The arm-up gesture is so common that we might almost suppose it to be an innate and species-wide signal, but it is more dependent upon learning than, for example, our facial expressions of anger or joy. Unlike the words of a language and unlike our "quotable gestures" such as the bye-bye wave, on the other hand, the arms-up gesture is not ordinarily learned either by imitation or by direct instruction, but rather through mutual adjustment to the actions of other people. It is conventionalized from an instrumental gesture, but it comes to act as a deliberate communicative signal.

The begging gesture–hand extended, palm upward with the fingers together–is learned in much the same way. Humans share this gesture with chimpanzees so it has deep roots, but it requires more learning by each infant than do the calls and gestures that form the inherited communicative repertory of each species.

A parallel example, this one audible rather than visible, is provided by the humble grunt. Lorraine McCune and her colleagues have studied grunting in human and nonhuman primate infants and they have followed the development of human grunting from a purely instrumental noise to a communicative signal (McCune et al. 1996, McCune in press). They were able to distinguish three stages of grunting in the children they observed. First came effort grunts that occurred when babies exerted themselves, as when reaching for an object, when changing position, or when crawling. Effort grunts occurred in the first month of life and, of course, we all still make them. Those observed by McCune were purely instrumental, a by-product of a baby’s exertion. Attention grunts appeared a bit later and occurred when children were paying attention to something by looking at it or by touching it, but they were made without any indication of special effort or any sign of an intention to communicate. These attention grunts could still have been noticed and responded to by caretakers, however, and the children could have discovered that they could attract attention with a grunt. Finally, the children made communicative grunts. These occurred while the child looked at his mother, reached toward her or tugged at her when trying to attract, or be certain of, her attention. Communicative grunts appeared during the second year, close to the time when words began to be used. Like words, the communicative grunts were deliberate communicative signals.

Examples of conventionalized instrumental acts that I find even more interesting than those of human children come from the observations of Michael Tomasello and his co-workers who have studied the communication of young chimpanzees who were growing up in a semi-naturalistic situation at the Yerkes Primate Center Field Station in Georgia USA (1985, 1989, 1994). These young chimpanzees use a wide variety of gestures to communicate with each other and with adults. For example, infants develop idiosyncratic ways to let their mothers know that they want to nurse. These gestures begin when a baby simply pushes his mother's arm aside so that he can reach the nipple. Mother's learn to recognize this instrumental act and this, in turn, permits the gesture to be conventionalized until the infant needs only to touch his mother in a characteristic way, and she will understand that he wants to nurse. The interesting point is that the gestures are quite idiosyncratic. Each infant uses them only with its own mother, never with another individual, So each pair is free to develop its own convention.

Young chimps also learn to use a considerable number of other idiosyncratic gestures. Some slap the ground, stamp their feet, or throw things as an invitation to play. They direct an adult’s hand or point to their side when they want to be tickled. They present their back when they would like to be groomed. They beg with an extended hand. Many of these gestures vary from one individual to another, and many are never made to a young chimp by an older animal. This makes it impossible to learn them by imitation. Nevertheless, these communicative gestures of young chimps are under far less tight genetic control than a dog’s snarl. They have to be learned by each individual, conventionalized in the course of ontogeny

Signal types

The chart places conventionalized instrumental acts in the context of other forms of animal and human communication. Examples of human and animal signals are listed on the left, and their most relevant properties are shown at the right. The two sets of rows at the top are all examples of signals that I like to call "gesture-calls". This term is simply a way to recognize the unity of the auditory and visible aspects of mammalian signaling and, at the same time, to acknowledge the similarity of one component of human communication to the communication of other mammals. We do not usually think of human beings as having "calls" but our laughter, screams, and sobs join with our bodily postures and with our facial expressions to form a thoroughly primate system of communication. This is the gesture-call system of the human primate, unique to our species in its details, just as the details of each gesture-call system are unique to its species, but consisting of signals that are very much like those of other primates both in the way they are produced and in the kinds of messages they convey (Burling 1993). Our gesture-calls have been built into each of us by the long process of phylogenetic ritualization. Like the gesture-calls of other primates, they need, at most, to be triggered off by the experiences that come to each individual in the normal course of maturation. They are narrowly determined by our genetic inheritance.

Toward the bottom of the chart are the most language-like parts of human communication. These include language itself, both the spoken languages of hearing people and the signed languages of the deaf, and also the gestures that Kendon has aptly called "quotable " (1993). These include hand signals such as the V-for victory sign, the thumbs-up gesture, the head screw that suggests that someone is crazy, and a great many more. Like the words of a language, these quotable gestures have to be learned. They differ much more from one community to another than do our gesture-calls, and they form a part of the community’s cultural tradition. In addition to these word-like gestures, we also use a number word-like noises, such as oh-oh and tsk-tsk. These noises are not quite words because they do not conform to the phonological system of a language (which is why they are difficult to spell) nor fit into its syntax. Like true words and like quotable gestures, however, these not-quite-words have to be learned, they are passed down by tradition, and they vary from one community to another. By analogy with "quotable gestures" they might be called "quotable noises".

Between the gesture-calls and the language-like signs on the chart are the conventionalized gestures and noises that have already been discussed. They share some properties with gesture-calls and other properties with language.

These three types of communicative signals differ most sharply in the way they are acquired. Both gesture-calls, and conventionalized gestures and noises begin as instrumental acts, but snarls and other gesture-calls became communicative by being ritualized through the long process of natural selection. The arms-up gesture, communicative grunts and nursing pokes have to be learned, or in a sense, invented by each individual while interacting with others. Like language and like quotable gestures and noises, the conventionalized signals have to be learned by each individual, but only the language-like signals are learned by imitation. Only they can be perpetuated as a part of the cultural tradition of a community.

In addition to being learned, conventionalized gestures share one other important characteristic with language: Instead of grading into one another they are in contrast. There are no half-way signals between two conventionalized gestures any more than there are half-way signals between two contrasting words. A different way of making this point is to say that the conventionalized gestures belong to a digital system, while our gesture-calls form an analog system. Giggles, laughs, and guffaws are connected by a continuum of signals that are intermediate both in the way they are formed and in the meaning they convey. Many of our facial expressions, such as those that show our anger, joy, and fear, also grade into each other. So, apparently, do many or most of the gesture-calls of the great apes (Marler 1976). The conventionalized gesture by which a chimpanzee infant shows that it wants to nurse, on the other hand, does not grade into anything else. Unlike a laugh, it does not occur in a range of slightly varied forms with related but slightly varied meanings. An arms-up gesture is unambiguously a request to be picked up. No intermediate gestures connect it to the equally unambiguous begging gesture. These signals are as discrete as human quotable gestures, and this makes them considerably more language like than are gesture-calls. It is true, of course, that some animal signals are discrete. The famous vervet alarm calls, for example, appear to be safely distinct from one another, but if ape and human gesture-calls are predominantly graded, the discreteness of vervet alarms is hardly relevant to the phylogeny of language

Words and Conventionalized Gestures

Ontogenetically conventionalized acts are more word-like than are gesture-calls but they are by no means words, and once we have isolated them as a special type of signal we can see both the ways in which they resemble words and the ways in which they differ.

One difference is the greater degree of iconicity of the conventionalized gestures and noises. It is true that the sign languages of the deaf have considerably greater iconicity than does spoken language, but as linguists like to insist, most spoken words have no resemblance at all to the things they stand for. Linguists illustrate this by such obvious examples as the words for head. The French say tête and the Germans say kopf, and any other form would do equally well so long as it is accepted by the community. Many, though not all, of the gesture-calls of both humans and animals might also be regarded as arbitrary. A dog’s wagging tail tells us that he is happy, while the wagging tail of a cat conveys a very different emotion. The relationship between the form and the meaning of a tail wag may seem to be every bit as arbitrary as the relation between the form and meaning of a spoken word. Unlike words, but like other animal signals, however, tail wags, are firmly set by the genetic inheritance of the species. This gives the ritualized arbitrariness of gesture-calls an utterly different basis than the conventional arbitrariness of words.

Conventionalized instrumental gestures are far from arbitrary, for they reflect the instrumental origin of the gestures or noises from which they were derived. In spite of their conventionalization, for example, the arms-high and begging gestures, retain a good deal of the iconicity of their instrumental origins. Comparing the iconicity and arbitrariness of various types of signs is difficult because most varieties of signs, even words, show a mixture of arbitrariness and iconicity so the differences are far from sharp. Nevertheless, the relatively high degree of iconicity of conventionalized signs seems clear.

Another difference between the conventionalized signs and language is that none of the conventional signs are used symmetrically between two individuals. A parent and child can use the same words with one another, but parents do not use the arms-up gesture to their children as a request to be picked up. If a parent uses the arms-up gesture it is in playful imitation of the child, not a serious request. Chimpanzee infants make nursing gestures to their mothers, but mothers do not make the same gesture to their children. Moreover, unlike most words, none of the conventionalized gestures are names for things. They are, instead, imperatives–requests or commands for an action by another individual. Even the attention grunt is a request for attention.

Finally, it must be emphasized once more that the conventionalized signs are not learned by imitation. Without imitation, it is impossible for an entire community to share the same sign, or for signs to be passed down from one generation to the next. There is no indication that chimpanzee nursing gestures, infant grunts, or the arms-up gesture are learned by imitation. In the absence of imitation, individuals, or pairs of communicating individuals are free to differ in the forms of their signals. Ontogenetically conventionalized gestures do not lead a species across the boundaries of culture.

In spite of their imperative and asymmetrical use, the absence of imitation, and their relatively high degree of iconicity, conventionalized signs resemble words in important ways. Like language, but unlike gesture calls, the conventionalized signs are learned, conventional, and discrete These characteristics make them a much more promising source from which to imagine early language growing than is any part of a gesture-call system.

Conclusions

As soon as we recognize that comprehension had to come first in the phylogeny of language, just as it came first in the history of animal calls and gestures and just as it comes first for each individual child, we are led to ask some new questions about the first stages of human language. We should ask about the kinds of selective pressures that might have driven our prehuman and early human ancestors toward an increasingly skilled ability to interpret the instrumental acts of others. We should also ask how, at later stages, they could have begun to understand the acts of others in increasingly word-like and then sentence-like ways. The origins of comprehension should, after all, be less mysterious than the origins of production. Producers may have excellent reasons not to give themselves away so it is often highly advantageous not to communicate. On the other hand, any animal, including a human animal, has little to lose and potentially a great deal to gain by understanding as much as possible from the behavior of conspecifics: What is that fellow likely to do? What does she want? Why is she moving off in that direction? What does that grunt mean? The more one animal can infer from the actions of others, the more skillfully he can plan his own behavior. As mutual comprehension improves, of course, a time will come when it will be advantageous for individuals to exploit the comprehension of others. Then they can adapt their own production to the comprehension skills of their conspecifics. They can act in deliberately informative ways. Then, and only then, does the coevolutionary development of productive and receptive skills begin.

A focus on improved comprehension might give us a different picture of the sequence by which new features enter language than does a focus on production. It is not at all obvious that we would expect the same sequence if we ask how people might have built up their understanding as we would expect if we ask how they would come to produce an increasingly complex language. If, for example, we assume that our forebears used single words before joining them together into orderly sequences, we will want to ask how the joining could have begun. A focus on comprehension should lead us to ask how understanders could start to make inferences from the sequence of the words they hear, even when the producer had made no effort to arrange them in orderly ways. The first step, quite plausibly, could have been nothing more than a gradual increase in the frequency of individual words. As more and more words were used, they would begin to bump up against each other. They would emerge in more rapid succession, but without any deliberate structure having been imposed upon them. Even without intending to do so, however, a speaker might use words in consistent ways. If he thought chronologically, for example, he might utter his words in a sequence that iconically reflected the chronology of events. Once comprehenders began to perceive the chronological significance of the word order that they heard from others, producers might find it advantageous to exploit that understanding in order to communicate more precisely. A rudimentary iconic syntax would then become possible. As soon as we recognize that it becomes useful for speakers to use a new form of communication only after his interlocutor has the ability to understand it, we should ask, for every feature that must enter a language, why producers would have begun to use it without any intention of communicating. How could understanding develop even before speakers began to exploit it?

The questions that I have raised in this concluding section of the paper, are different from those that have most often been asked by those of us who are interested in the origin of language, but they arise naturally as soon as we recognize the central role of comprehension. I believe these questions deserve careful thought and debate.

References Cited

Burling, R. 1993. Primate calls, human language, and nonverbal communication. Current Anthropology, 34, 25-53.

Hayes, K. J. & Nissen, C. H. 1971. Higher mental functions of a home-raised chimpanzee. In Behavior of Nonhuman Primates (Ed. by A. M. Schrier & F. Stollnitz), pp. 59-115. New York: Academic Press.

Kendon, A. 1993. Human gesture. In Tools, Language and Cognition in Human Evolution (Ed. by K. R. Gibson & T. Ingold) pp. 43-62. Cambridge: Cambridge University Press.

Marler, Peter. 1976. Social organization, communication and graded signals: The chimpanzee and the gorilla. In Growing points in ethology. (Ed. by P. P. Bateson and R. A. Hinde) pp. 239-80. Cambridge: Cambridge University Press.

McCune, Lorraine, Marilyn M. Vihman, Liselotte Roug-Hellichius, Diane Bordenave Delery, and Lakshmi Gogate. 1996. Grunt communication in human infants (Homo sapiens). Journal of Comparative Psychology, 110(1), 27-37.

McCune, Lorraine. In press. Children’s Transition to Language: A Human Model for Development of Vocal Repertoire in Extant and Ancestral Primate Species? In The Origins of Language: What Nonhuman Primates Can tell Us. (Ed. by Barbara J. King.) Santa Fe: School of American Research Press.

Savage-Rumbaugh, E. S., Murphey, J., Sevik, R. A., Brakke, K. E., Williams, S. L., & Rumbaugh, D. M. 1993. Language Comprehension in Ape and Child. Monographs of the Society for Research in Child Development. Serial No. 233, Vol. 58, Nos. 3-4, 1993.

Tinbergen, N. 1952. Derived activities: Their causation, biological significance, origin and emancipation during evolution. Quarterly Review of Biology, 27, 1-32.

Tomasello, M., George, B. L., Kruger, A. C., Farrar, M. J., & Evans, A. 1985. The development of gestural communication in young chimpanzees. Journal of Human Evolution, 14, 175-186

Tomasello, M., Gust, D., & Frost, G. T. 1989. A longitudinal investigation of gestural communication in young chimpanzees. Primates, 30, 35-50

Tomasello, M., Call, J., Nagell, K., Olguin, R., & Carpenter, M. 1994. The learning and use of gestural signals by young chimpanzees: A trans-generational study. Primates, 35, 137-154.

Tomasello, M. and Call, J. 1997. Primate Cognition. New York and Oxford: Oxford University Press.