“Topic One Defining Language” in “The Signifying Animal”
What is a language? You may well be wondering why I ask this question when everyone knows what a language is—it’s what you’re expressing and I’m comprehending, you say. Let’s change the question’s form a bit: How would one identify something as a language if he encountered what might be one in an obviously nonhuman species—for example, flowing kaleidoscopic color patterns on the bulbous bodies of octopus-like creatures who land in a spaceship right in one’s own backyard? And, for that matter, is the natural signing of deaf-mutes a language? The game of chess? And what about the “language” of music or art? Or suppose that pale, eyeless midgets were discovered in extended caverns far below the present floors of the Mammoth Cave-emitting very high-frequency pipings from their rounded mouths and apparently listening with their enormous, rotatable ears. How might one decide whether or not these cave midgets have an identifiably humanoid language? Only if one can say what defines a language in general and defines a human language in particular can he go on to offer possible answers to some other, very important, questions: Do certain nonhuman animals “have” a language? What is common to prelinguistic cognizing and linguistic sentencing? When does a developing child “have” a language? How may languages have developed in the human species? Answers to all these questions, of course, would have relevance to the basic issue of universals and uniqueness in human communication.
I. DEFINING CHARACTERISTICS
OF LANGUAGE GENERALLY
If anything is to be called “a language,” it must satisfy the following criteria: It must (i) involve identifiably different and nonrandomly recurrent physical forms in some communication channel, (2) these forms being producible by the same organisms that receive them, their use (3) resulting in nonrandom dependencies between the forms and the behaviors of the organisms that employ them, (4) following nonrandom rules of reference to events in other channels and (5) nonrandom rules of combination with other forms in the same channel, with (6) the users capable of producing indefinitely long and potentially infinite numbers of novel combinations that satisfy the first five criteria.1 Now, with our Octopian visitors (from the planetary system of the nearest star, Arcturus, as was later discovered) particularly in mind, I will elaborate a bit on these criteria for something being “a language” in general.
1. The Nonrandom Recurrency Criterion: Production of identifiably different and nonrandomly recurrent physical forms in some communication channel. A few years ago there was quite a flurry of excitement over apparently nonrandom, recurrent signals being received over interstellar radio receivers—was it something unusual in sunspot activity or, possibly, communication attempts from some distant form of intelligent life? (As I recall, the decision was in favor of sunspots!) As far as ordinary communication is concerned, humans have opted for the vocal-auditory channel (for reasons that will be considered a bit later), but there are many other possible channels that we are aware of, because we have also these sensory modalities (light-visual, pressure, tactile and chemical-odor, at which dogs seem particularly adept), and many others that we are not aware of (e.g., radiant wavelengths above or below the visible-to-humans range)—and what about ESP? The only critical thing is that such signals be combined in nonrandom ways to produce energy forms that are identifiably different (using special equipment if necessary) and are themselves nonrandom in distribution in time and/or space. Note that there is no requirement that the forms be discretely digital (as is generally the case for human languages); they could be continuously analogic in nature.
Back in our own backyard, having recovered from the shock, and being convinced that our octopus-like visitors intend us no harm, we note that as one Octopian pirouettes slowly, with its whole bulbous body flowing with multicolored visual forms, the other stands “silently” neutral gray—only to begin displaying and turning when the first has become “silent.” We begin to think that this may be some form of communication—not merely displays of emotional states as in the “blushing” of the chameleon.
2. The Reciprocality Criterion: These forms being producible by the same organisms that receive them. It is difficult to imagine a human society in which, say, women were the only ones who could produce language (but not comprehend it, even from other women) and men could only comprehend language (but not produce it). For one thing, knowing what we do about the intimate interactions between comprehending and expressing in the development of language in children, one would be hard put to account for the development of such a system. But what about communication between species? Yucca plants and Yucca moths certainly interact for their common survival, but one would hardly call this communication; the “language of flowers” is only a euphemism—we humans can’t smell back! But what about mule drivers and their mules? There is no question but that there is communication here, even though the mule can only “kick” back when ornery, but obviously the same organisms are not both producing and receiving the same forms—indicating clearly that communication is not the same as language. Noting that some of the flowing visual forms produced by Octopian A are reproduced in apparent “response” by Octopian B, always along with some uninterpretable contextual variations—and observing this systematically later with color movie films—we become convinced that this may, indeed, be some kind of language.
3. The Pragmatic Criterion: Use of these forms resulting in nonrandom dependencies between the forms and the behaviors of the organisms that employ them. This criterion applies most testably to the recipients of messages (behaving in appropriately differential fashion to signals received), but also to the initiators of messages (displaying a nonrandom tendency to communicate about entities and events that are proximal in other channels—see criterion 4 below). This is really the criterion that there is communication going on. Except for obviously representational art and music—and for purely affective reactions—this criterion would seem to rule out anything other than euphemistic use of phrases like “the language of art” or “the language of music.” There is no doubt that bright dogs (like my poodle, Pierre!) develop a large repertoire of appropriate behaviors dependent upon the verbal commands of their masters (e.g., fetching his bone rather than his ball when requested “go getcha bone”).
Note, however, that there is no implication here that pragmatic dependencies must be acquired through experience (learning), although this is clearly the case for human languages. In the language of the bees, when the observing bees in the hive fly the distance and direction signaled by the returned dancing bees, this satisfies the pragmatic criterion—albeit on an innate, “wired-in” basis. What, then, about communication between a human master and a completely “wired-in” computer servant? I would argue that this meets the pragmatic criterion (see Winograd, 1972, for a nice example)—but obviously not others (particularly criterion 6, combinatorial productivity). And what about our Octopian friends? Satisfaction of this criterion would be indicated by, for example, a certain color pattern in Octopian A being conditionally dependent upon the presence in front of their spaceship of some complicated scanning device and, when accompanied by various contextual color patterns, by behaviors with respect to this device (shifting its orientation, taking it back into the ship) on the part of Octopian B.
4. The Semantic Criterion: Use of these forms following nonrandom rules of reference to events in other channels. This criterion implies that for anything to be a language it must function so as to symbolize (represent for the organism) the non-necessarily-here and not-necessarily-now. Although such representing relations are clearly acquired via learning by humans, the language of the bees again tells us that this is not necessarily the case—their dance, upon returning to the hive, does symbolize the not-here (source of nectar) and the not-now (to be found at some indicated flying time in the near future) on a purely innate basis. This criterion is clearly not met by the game of chess (where the pieces, despite their names and their moves, bear no symbolizing relation to anything other than themselves) or by the “game” of mathematics (where the symbols are deliberately abstract and bear no necessary relations to anything in the real world, but by virtue of this property are potentially relatable to any set of real-world entities). Even in humans, semantic relations are not necessarily arbitrary: there is onomatopoeia (the name of a thing or event being based on its characteristic sound, e.g., cuckoo, cough, hiss, slap and wheeze in English) and there is phonetic symbolism (in my classes I like to ask the male students which blind date they would prefer, Miss Pim, Miss Bowloaf, or Miss Lavelle, and then to describe her probable appearance).
Forms in a language can also be iconic—witness much of the natural signing of the deaf as well as the gestural accompaniments by ordinary speakers (e.g., in describing my “blind dates” above)—and this, we come to infer, might well be the case, at least in part, for our Octopians: we noted that, on the first appearance of Pierre the poodle in our yard, there was a silvery blob followed by a rising line on Octopian A (possibly a question?—“what on earth [sic! ] is that?”) answered by a nondescript wobbly fuzz figuring near the bottom of Octopian B’s bulbous body (possibly meaning, “I haven’t the foggiest idea!”). Later, after we demonstrated feeding doggie biscuits to Pierre, taken from a box (for which one Octopian extended a multiple “fingered” tentacle in request—which needless to say we honored), the appearance of Pierre would produce exchanges of silvery blobs, plus some angular pattern, and one Octopian would slip into the ship and reappear with the box of doggie biscuits. And still later they would “call for” an absent Pierre with his “blob” (and that rising line pattern)—and Pierre in turn quickly learned to come to beg the Octopians for biscuits!
5. The Syntactic Criterion: Use of these forms following nonrandom rules of combination with other forms in the same channel. As will be seen, all human languages may be characterized as being hierarchical in structure (analyzable in terms of units-within-units); they are also organized temporally on a “left-to-right” basis—that is, from prior to subsequent forms at all levels—as a necessary consequence of their utilization of the vocal-auditory channel. But these by no means must be defining characteristics of language in general, and one wonders to what extent space could be substituted for time in the organization of messages by organisms using other channels. Presumably there would be some limit—for example, we might discover (although the “how” of this is not as obvious as with the preceding criteria) that the Octopians “flash” the equivalents of whole paragraphs on their N “panels” as they rotate in the process of communicating, the within-paragraph information being spatially represented—“sentences” thus being an unessential carving up of the information flow (I might note that whole paragraphs, consisting of a single sentence, multiply conjoined and embedded, are not exactly a rarity in, particularly, scholarly writing!). The diverse recursive devices (e.g., as in center-embedded sentencings like the man the girl the teacher likes married plays poker) might be entirely unnecessary in a spatially organized language like Octopian.
And yet some structuring, representing what is “natural” in sentencing based on prelinguistic cognizing experience—for Octopian “squidsters” just as for human youngsters—might be expected for all organismic languages. For humans, the two basic types of simple cognition appear to be Action and Stative Relations (and both in SVO order), the former highlighting the typically +Animate Actor as subject as against the typically “Animate (or at least relatively passive) Recipient as object and the latter highlighting the Salient Figure as subject against the “Salient Ground as object—thus Pierre chased the ball and the ball was on the grass as simple Action and Stative cognitions respectively.2
After having mastered some of the pragmatics and semantics of Octopian, we might study our video-tapes with such Naturalness Principles in mind. For examples: are the visual patterns centered on a “panel” (e.g., the silvery “Pierre” blob) characteristically representations of the more animate and/or figurai entities, and the other blobs with which they are diversely linked radially by angular lines (verb phrases?) characteristically the less animate and/or figurai? Is there some spatial ordering discernible for several sets of “conjoined” radial patterns involving same or different centered “topics,” and does this ordering typically fit our human notions of naturalness in ordering?
6. The Combinatorial Productivity Criterion: The users of the forms being capable of producing indefinitely long and potentially infinite numbers of novel combinations that satisfy the first five criteria. At any particular time, synchronically, this novelty in human language lies in the combining and not in what is combined; thus, my statements about the Octopians’ linguistic behaviors, although entirely novel as wholes (I presume), utilized thoroughly familiar lexemes (criterion 4) and constructions (criterion 5) in contemporary American English. However, it must be admitted that anyone who has done an analysis of the semantics and syntactics of telephone conversations comes away rather unimpressed with the combinatorial productivity of ordinary speakers using ordinary language. Of course over time, diachronically, human languages do display adaptive changes by both expanding the lexicon (criterion 4) and by changing the rules (criterion 5). Presumably, any organismic language would display such adaptivity—either via evolution over very long periods (for languages with innately based semantics and syntax, like that of the bees) or via learning over relatively short periods, dependent upon changing environmental conditions (for languages with individually acquired semantics and syntax, as happens with human cultures and languages in contact).
Testing for the presence of this combinatorial productivity (criterion 6) in another language—particularly a very strange one like Octopian—would undoubtedly be most difficult. Although nearly all of the communicative exchanges among the Octopians, and between them and ourselves, would seem novel in whole or in part to us, that would be no proof that they were not elaborately stereotyped patterns, analogous to most human telephone conversations. Only after we had mastered their language to the point where we could compose Octopian statements of indubitable novelty for them (and see if they comprehended them) and ask questions requiring indubitably novel answers (and see if they could produce them), could we determine if they had this crucial capability. And—horror of horrors!—we might discover that they did not, that they were entirely “programmed” like computers, and in fact were robots sent out by the real Arcturians, whatever they might be like. However, one thing this little experiment by Arcturians would have demonstrated is this: if two species each have a language by these criteria, then either directly (or mediately, via appropriate equipment) they should be able to communicate to some extent with each other; the extent and direction would depend on the amount and balance of intelligence, that organism with the lower channel capacity determining the limit on communication (the dog in effect setting limits for the master).
A. DO ANY OTHER, NONHUMAN, ANIMALS HAVE A LANGUAGE?
One can trace a continuum of levels of interorganismic communication: from proximal interactions (contacting, mating, mothering, fighting and . . . consuming!), though distal SIGNAL sending and receiving (unintentional odors such as the bitch in heat, mating calls, baboon warning and food-supply noises), through distal SIGN sending and receiving (intentional expression of affect, like growls, tail waggings, postural and facial expressions designed to influence the behavior of the receiver), to SYMBOL creating and interpreting (the food-supply dance of the bees, “play” in dogs and other higher animals, referential gesturings by chimps and humans). For any given species, we can ask, does its intraspecies communication satisfy the six criteria for something to be a language? Take the clam—which, if anything, seems to have specialized away from communication in its evolution: because, as far as I know, there is no evidence for clam-to-clam nonrandom recurrency of signals in any channel (criterion i)—and, given the limited motility and reactive capacity (“neck” retracting, shell clamping, and ...?), it seems likely that communication would be limited to chemical broadcasts at most (unless the clam has been fooling us and specializes in the development of ESP!). If criterion i is not satisfied, of course, then none of the others (reciprocality, pragmatics, semantics, syntax, or combinatorial productivity) can be met. The answer to whether this species has a language is, in a clam shell, no! So let’s move along up the evolutionary tree.
a. The Bee. Briefly (necessarily),3 the bee communicates three species-significant things: showing its pass-badge (a scent pouch that is opened on entering the hive and, if the scent is wrong—execution); locatioîi of a nectar supply (the well-known “dance,” whose angle with respect to the sun indictaes direction, whose number of turns per unit time indicates distance, and whose number of abdomen wags indicates quality of the supply); location of a new “home” (a kind of “election” in which, when local supplies have dwindled, that returning bee which gets the crowd at home to follow his dance—usually the one that has found the richest load, has had to fly the shortest distance back, and hence is the most energetic—ends up with the whole hive flying off to his new location). Because the dance forms are obviously nonrandom, because any worker bee functions equivalently as sender or receiver, because behaviors of both are nonrandomly dependent upon the messages, and because the forms have nonrandom rules of reference (to the not-here and not-now nectar locations), criteria 1-4 appear to be clearly met. As far as syntactic criterion 5 is concerned, because the messages involve three types of forms (direction, distance, and quality indicating) that must be combined in certain nonrandom ways, this would seem to be met. But what about criterion 6, the combinatorial-productivity criterion? For any given bee “speaker” or bee “listener,” a given combination of direction, distance, and quality indicators must often be novel, yet communication is successful. So we conclude that, within very narrow limits of what can be communicated, “the language of the bee” is not a euphemism—and, most remarkably, it is entirely innate (“wired in”).
b. The Bird. The varied calls of the many subspecies of birds (which must be acquired, particularly during an early “imprinting” stage) are sufficient evidence for satisfying criterion 1 (nonrandom recurrency of forms), and the back-and-forth callings that awaken us so pleasantly in the early mornings of spring are evidence for satisfaction of criterion 2 (reciprocality). In the bird-to-bird communication of crows, for example, their “alarm” and “assemble” calls are evidence for at least rudimentary pragmatics and semantics (criteria 3 and 4)—and there are even crow “dialects” (taped alarm calls of American crows have been shown to produce assembling on the part of French crows!)—but these are essentially absent in bird-to-man communication. Given criteria 1 and 2, plus use of the vocal-auditory channel, one can teach some birds (e.g., parrots, parakeets, myna birds) to talk “human” in limited ways, but these appear to be purely imitative, sensory and motor integrations—meaningless, without expressing representational significances and intentions. I well remember in the early 1950s a research assistant of O. Hobart Mowrer spending many months and thousands of trials trying to get a talkative myna bird to say something like “wanna-eat” and “wanna-drink” (both in its imitative vocabulary) differentially for seeds vs. water, when made hungry vs. thirsty and shown the appropriate reinforcers—with never a significant shift from pure chance performance. So the bird appears to be strong on criteria i and 2 and weak on 3 and 4, with both 5 (syntax) and 6 (combinatorial productivity) totally lacking—talking birds repeat phrases as wholes, never piecing together parts of different phrases to make new combinations.
c. The Dog. This species might almost be said to have developed “writing” before “talking,” because it seems to depend more on persistent (odor) than evanescent (visual or auditory) denotative signs—everytime I return to the house, my Pierre has to lift up and sniff my hand, just to make sure it’s me! However, in dog-to-dog (as well as dog-to-human) communication, connotative meanings (primitive affective Evaluation, Potency and Activity, also found to be universal in human languages—see Osgood, May, and Miron, 1975) are richly displayed and reacted to in gestural-vision and vocal-auditory channels, the dog’s tail being a particularly expressive organ: up vs. down signaling the Pleasantness, rigid vs. limp, the Forcefulness, and moving vs. stationary, the Liveliness of the animal’s affective states. Dogs can acquire extensive repertoires of differential significances for human-produced signs through training, either formal (for hunting or circus performance) or informal (as pets), and this definitely satisfies both pragmatic (3) and semantic (4) criteria. Thus, in contrast with the bird, the dog is weak on criteria 1 and 2 (dependent on what I call the Integrational Level of behavior) but strong on 3 and 4 (dependent on the representational level). But, like the bird, the dog does not satisfy 5 (syntax) and at best only minimally meets 6 (combinatorial productivity)—and that only in comprehending (e.g., behaving appropriately to the pivot phrase “pick up” plus a new object name like “your leash”).
d. The Ape. I use this term to refer to nonhuman, but close-to-human primates—rather than the chimp—because others have been shown to have similar capabilities (e.g., recently the gorilla). I will also concentrate on the Gardners’ Washoe (R. A. Gardner & B. T. Gardner, 1969; B. T. Gardner & R. A. Gardner, 1975) rather than the Premacks’ Sarah (D. Premack, 1971; A. J. Premack & D. Premack, 1972). The laboratory research with Sarah (comprehending sentencelike vertical sequences of plastic symbols and behaving appropriately; producing plastic sentence-like sequences of her own and being differentially reinforced) demonstrates the astonishingly complex cognitive capacity of a chimpanzee: ability to signal the “sameness” or “difference” of object pairs, that symbol X is or is not “the name of” various presented real objects (bananas, apples, etc.), responding appropriately to “clause” pairs conjoined by an “if-then” symbol (e.g., SARAH TAKE APPLE (“if-then”) MARY GIVE CHOCOLATE SARAH / MARY NO GIVE CHOCOLATE SARAH), and so forth. However, whether this represents comprehension and production of sentences or “simply” complex, differentially reinforced reactions remains obscure (usually only a single set of alternative responses was required on any given problem). And although there is ample evidence for semantics and pragmatics, there is little for syntax or combinatorial productivity.4 In any case, Sarah’s would be a highly arbitrary written “language” rather than the gestural-vocal one natural for the chimpanzee.
Prior to the Gardners’ work with Washoe, several psychologists (e.g. the Kelloggs’ Gua and the Hayes’ Viki) had also brought up an infant chimpanzee in their home as they would a child of their own, but the only attempts at “language” seemed to be to teach the ape to talk human—and they failed miserably (Viki ended up with about four imperfectly produced simple words like “papa” and “cup”). Given the lack of hemispheric dominance in nonhuman primates, which is critical for voluntary control over the medially located speech apparatus, this failure was not at all surprising. The decision to bring a chimp up in a human environment, but with constant exposure to the natural sign language of the deaf-mute, was long overdue, and one of the most exciting developments in decades resulted.5 Facial, manual and postural gesturings, along with strong tendencies to imitate, are characteristic of chimps in their natural state, and Washoe took readily to signing, imitating her human companions in the context of meaningful everyday events (with their motivating and reinforcing properties) and babbling away manually on her own (with her companions “shaping” her signs into the human-proper forms). And I understand that Washoe is now in a colony of chimps at Norman, Oklahoma, busily communicating with others just acquiring ... the language?
So now let’s check Washoe’s communicative performance (as reported in R. A. Gardner & B. T. Gardner, 1969) against our criteria of something being “a language.” Criterion 1 (nonrandom recurrency of forms) is obviously met by the differential use (by age 4) of some 80 gestural signs. Criterion 2 (reciprocality, both sending and receiving) is obviously met—first with humans “at the other end” but more recently with other chimps. Criterion 3 (pragmatics) is satisfied by such evidence as her making the “toothbrush” sign “in a peremptory fashion when its appearance at the end of a meal was delayed,” by her signing “open” at the door of a room she was leaving, and so forth ad infinitum. There is also no question about satisfying criterion 4 (semantics): her spontaneous “naming” of toothbrushes, with “no obvious motive other than communication”; her learning to sign “dog,” mainly to those in picture books, but then signing it spontaneously to the sound of an unseen dog barking outside; her signing “key” not only to keys being presently used to open locks but also to “not-here” keys needed to unlock locks! And there is also no question but that criterion 6 (combinatorial productivity) is satisfied: the Gardners report that as soon as Washoe had a vocabulary of a dozen or so signs (including verbs like “open” and “go,” nouns like “key” and “flower,” the pronouns “you” and “me,” and adverbials (?) like “please,” “more,” and “hurry”) she spontaneously began combining them in sequences like “open flower” (open gate to flower garden), “go sweet” (to be taken to raspberry bush), and “you me out” (you take me outdoors); she also displays the “pivot/open-class” productivity familiar in child language development (e.g., “please sweet drink,” “please key,” “hurry out,” and “please hurry sweet drink”).
But what about criterion 5 (syntax)? This has been the focus of most questioning of Washoe’s “having a language,” and in early critiques both Bronowski and Bellugi (1970, p. 672) and McNeill (1970, p. 55) stress the fact that Washoe’s “utterances” display no constraints on “word” order, her signings seemingly having free ordering (e.g., “up please” or “please up,” “open key” or “key open”). However, in an equally early commentary, Roger Brown (1970 pp. 224-230) makes several very significant points:
1. that Washoe’s linguistic performances should be compared with those of a 3- or 4-year-old deaf-mute child rather than with normal children of this age;
2. that just as normal children already control several prosodic patterns when they begin to produce combinations (e.g., the falling pitch of declaratives, the rising pitch of interrogatives), so do the deaf—and quite spontaneously (according to the Gardners) Washoe—hold for a perceptibly longer period the last sign of a sequence to indicate a question; and
3. that just as in human language development, Washoe displayed a gradual increase over time in the sign length of her “utterances”—two common before three and three common before four—and Brown asks reasonably, “why should this be so if the sign combinations are not constructions [p. 225]?”
Perhaps most significantly, Brown (1970) observes that “there is little or no communication pressure on either children or Washoe to use the right word order for the meanings they intend [p. 229]” when language is being used in contexts that are perceptually unambiguous to both producer and receiver—which is the case in much of early child language and in just about all of Washoe’s signings (and it should be noted that, although Washoe’s companions “corrected” the signings of particular lexical items, they apparently did not “correct” for sign orderings, as do most adult companions of young human children).
Relevant here is an actual experiment with Washoe (B. T. Gardner & R. A. Gardner, 1975) designed to get at evidence for recognition of “sentence constituents” in her communications. In English, answers to wb-questions require identification of the relevant NP constituent; given the sentence Roger put the key on the table, for example, the question Who did it? specifies the subject NP (Roger), the question What was put somewhere? specifies the object NP (key), and the question Where was something put? specifies the NP-head of the locative phrase. The Gardners were able to demonstrate not only that Washoe’s replies (including single-sign) to such wb-questions were significantly contingent with the correct NP constituents (at the .00001 level) but also that her 84% correct NP inclusion in replies was superior to the performance of normal children of corresponding linguistic age! It is also interesting that David McNeill, in a recent paper (1974), now concludes that the chimpanzee (Washoe data, primarily) does meet criterion 5, “. . . spontaneously adopting an apparently novel form of syntax based on social relationships (such as addressee and nonaddressee).” Generalizing, it appears that, whereas human language is designed for humans to talk about people and things, natural chimp “language” is designed for chimps to talk to other chimps (or people, in Washoe’s case).
II. DEFINING CHARACTERISTICS
OF HUMAN LANGUAGES
Human languages must, of course, satisfy all the criteria for anything to be called a language—thus having nonrandom recurrent signals in some channel, producible by the same organisms that receive them, which display nonrandom pragmatic, semantic, and syntactic dependencies that are combinatorially productive—but there are additional delimiting criteria that must be met if something is to be called a natural human language. These additional defining characteristics can be categorized (at least superficially) as either structural or junctional.
A. STRUCTURAL CHARACTERISTICS OF HUMAN LANGUAGES
For something that is a language to be called a human language, it must have the following structural characteristics: it must (7) involve use of the vocal-auditory channel, and thus (8) nondirectional transmission but directional reception and (9) evanescence in time of the forms in the channel, these characteristics requiring (10) integration over time of the information derived from the physical forms, but also (11) providing prompt feedback to the sender of his own messages. All of these structural characteristics are direct, combined functions of the physical nature of sound and the biological nature of the human organism.
7. The Vocal-Auditory Channel Criterion: All natural human languages use vocalization for production and audition for reception. This, of course, refers to the primary communication system for humans, there being many other derived systems—the most general being writing (a more lasting gestural-visual sort of channel) but also drum signals, smoke signals, and Morse code. It should be noted that the vocalic response system is both relatively “lightweight” (in terms of energy required, as compared, say, with locomotion) and minimally interfering with other ongoing activities (like toolmaking, hunting, and fighting)—properties we shall find relevant to the question of the origin of human languages.
8. Nondirectional Transmission, Directional Reception: In human languages, speaking is broadcast and hearing is selective. Broadcast transmission is simply a function of the manner of propagation of sound waves—in all directions and, conveniently, around corners (an advantage lacking in visual Octopian, by the way); selective reception is simply a function of the fact that we have a head between our ears, this interaural distance yielding phase differences for sound waves originating in all directions except along the medial line, and hence providing cues for direction of the source. This channel characteristic may well have had significant influence upon the social structures of primitive human groups—Mr. “Big-mouth” being heard by all “Little-mouths” at once, but the “Little-mouths” securing one-to-one privacy only by isolating themselves!
9. The Evanescence-in-Time Characteristic: Signals in the vocal-auditory channel jade rapidly in intensity over time. The advantage of such evanescence, of course, is that it minimizes “cluttering up” of the channel—as anyone who has tried to understand what another is saying across an echo chamber fully appreciates. The disadvantage is that it puts a heavy load on short-term as well as long-term memory—the latter undoubtedly being the major reason for the development of writing systems of various types, apparently independently in different human societies as they reached certain levels of complexity. (Parenthetically, it is sobering in this nuclear age to realize that humans have had writing systems for only about 5,000 years.)
10. The Integration-over-Time Requirement: The distribution of message forms over time on a linear “left-to-right” basis requires temporary storage and integration of information. Although there is simultancous patterning within sounds (e.g., chords in music), it is minimal in comparison with vision. A familiar example of such temporal integration in human speech is prosody—a falling intonation pattern over time signaling a statement, a rising pattern a question. This constraint also leads speakers to shift “leftward” (earlier in time) message elements that are salient to them (e.g., creating a passive, “Pierre was stung by that bee!”). Of course, our Octopian receivers would have to integrate information over space within the “panels” of their bulbous senders’ bodies, as well as store it temporarily for integration with information in succeeding “panels.”
11. Availability of Prompt Feedback: The speaker of a human language is normally capable of hearing his own messages as he produces them. This property of the vocal-auditory channel has significance for development of language in the young of the species as well as for production of language in the mature. Because it is a maxim in language development that, at all stages and in all phases, comprehension typically precedes production, it follows that the sounds of understood adult forms (like “ball”) and combinations of forms (like “that’s Mommy’s shoe”) can serve as models against which the child can correct his own productions (“bawh,” “dah Mommy show”), and even more complexly develop the syntactic niceties. In the ordinary speech of adults, one notes not only filled and unfilled pauses (time for cognizing to catch up with sentencing) but also retracings with corrections (usually of full constituent length, e.g., “Well, my dear, I did it all in one swell foop . . . /ah/ ... in one fell swoop and was on my way!”). Under abnormal conditions of delayed feedback (via the intervention of tape), one finds he cannot speak, or even read, naturally (as I found when the late Grant Fairbanks had me try to read the statement on a pack of cigarettes, “. . . the Amer-eric-c-can To-to-BACCO Comp-comp-COMPNY! . . .”); under abnormal condition of no feedback (via masking with noise), patients in psychotherapy have been found to talk more freely (but often rather incoherently) and without as much self-critical backtracking (Mahl, 1972). One wonders how the Octopian “speakers” would monitor their own communicative displays—by seeing them from within?
The signing of deaf-mutes would be ruled out, as far as being a natural human language, by criterion 7 (use of the vocal-auditory channel), but of course it would still satisfy the requirements for being a language, even in the chimp, as Washoe has demonstrated. And what about our Cave Midgets—in their domain far beneath the floors of the Mammoth Cave? While we were trying to determine if our visitors from Arcturus had something that could be called “a language,” other intrepid human explorers (linguists) were doing the same with the pale little Cave Midgets. Tape recordings of their high-frequency pipings left no doubt but that nonrandom recurrent sound forms were being reciprocally produced and received; the nonrandom dependencies of their use of artifacts in mushroom-and-worm cultivating activities upon these distinctive piping forms clearly satisfied the pragmatic and semantic criteria.
Testing for syntactic structuring and combinatorial productivity took a bit of doing, particularly because our linguists were struggling with very sore throats brought on by continuous whisperings—the big-eared Cave Midgets fly into panic at any loud, low-frequency sound. However, after many sleepless days analyzing visual displays of ultrasonic piping patterns, one linguistic genius demonstrated “noun/verb” selection rules, and, a bit later, another had a brainstorm (not surprisingly, after consuming a worm-and-mushroom pizza) and created a computer-based Cave-Midgetese synthesizer, at least for very simple utterances—and combinatorial productivity was firmly established. So there was no question but that these Cave Midgets had a language—and, given the piping sounds that went whistling around the cavern passageways plus the big ears that rotated to receive them, it seemed obvious that this language met the structural requirements for being of the human type—but what about the functional requirements?
B. FUNCTIONAL CHARACTERISTICS OF HUMAN LANGUAGES
For something to be called a human language, it must also have the following functional characteristics: (12) the semantic relations between forms and meanings must, in general, be arbitrary rather than iconic, and (13) the forms in the channel that distinguish meanings must be discretely rather than continuously variable; further, the forms in the channel must (14) be analyzable hierarchically into levels of units-within-units, with (15) large numbers of units at each higher level being exhaustively analyzable into relatively small numbers of components at each lower level; and finally, (16) extension of a language within the species, both generationally and geographically, must be via experience (learning) rather than via inheritance (maturation). The complex, visual-patterning language of the Octopians seems to fit neither the arbitrariness (12) nor the discreteness (13) criteria, and its efficiency must lie elsewhere than in hierarchical (14) and componential (15) organization. Whether the Octopian language extends itself in time and space on the basis of experience (like human language) or of inheritance (like bee language) we do not know at this point. However, because the Cave Midgets are a somewhat humanoid species of this earth, we would be most interested in seeing how their language stacks up against these functional criteria.
12. Arbitrariness of Form-Meaning Relations: In human languages the rules relating forms in the communication channel to events in other channels (cf. criterion 4) are typically arbitrary rather than iconic. We must say “typically” because (as noted earlier) human languages do display both onomatopoeia and phonetic symbolism; however, for the most part form-meaning relations are arbitrary (witness Pferd in German, cheval in French, and horse in English). Noting the prevalence of phonetic symbolism in the communication of affect—in both natural chimp (“oh-oh-oh” for joy vs. “uu-uu-uu” for sorrow, “eeee!” for fear) and natural human (“boy-oh-boy” for pleasure vs. “ugh” for disgust, “eek!” for the “shriek” of fear)—one might speculate that there has been social evolution from iconic affective signs toward arbitrary denotative signs, particularly as languages became more complex and abstract in their references.
13. Discreteness of Form Shifts Signaling Differences in Meaning: In human languages, the changes in form that convey changes in meaning are discretely rather than continuously variable. This characteristic certainly holds at the phonemic level (the abrupt shifts in distinctive features of sound that distinguish, for example, jail, gale, male, sail and tail), the morphemic level (the productive pluralizing morphemes for nouns in English are always either [-s], [z] or [-iz], conditionally dependent on the voicing or sibilance of the preceding sound, as in cats, dogs, and horses respectively), the lexical level (a graded speeding up of walk into say wok doesn’t yield the meaning of run), and the syntactic level (an NP being signaled by the singing and a VP by was singing, for a simple example). Such discreteness has certainly simplified the descriptive task for linguistic science. Whether such discrete either/or-ness holds at the level of semantic features is highly debatable—but then meaning is not usually overtly signaled in the surface forms of human languages.
14. The Hierarchical Organization Criterion: In human languages, the stream oj jorms in the channel is analyzable into levels oj units-within-units. Complex sentences are analyzable into clauses (or “sen-toids”), clauses are analyzable into immediate constituents (concatenations of NPs “hanging on” a VP), constituents into word forms (heads, modifiers, and modulators), words into morphemes (stems and affixes), morphemes into phonemes, and phonemes into distinctive phonetic features. To some extent semantic systems also display hierarchical organizations (particularly for nouns) that are describable in the form of “taxonomic tree” structures—thus the meaning of bird is entailed in the meanings of all its exemplars (sparrow, robin, eagle, etc.), the meaning of animal is entailed in those of bird, fish, mammal, and so forth. Implicit in this criterion (and in criterion 15) is the constraint that no higher-level unit can be embedded in a lower-level unit—and this raises some interesting questions about what is the proper linguistic analysis of, for example, sentences with relative clauses (I’ve met the girl who arrived late at the party, where the Wh-clause is itself a sentential elaboration of an NP), what I call “commentative” sentences (like I hope that John will be on time jor the wedding or It is a jact that John has been married bejore) or monstrous center-embedded sentences (like The boy the girl Pierre likes likes likes spaghetti!).
15. The Componential Organization Criterion: In human languages, large numbers oj units at each higher level in the hierarchy (of criterion 14) are exhaustively analyzable as near-simultaneous combinations oj relatively small numbers oj units at each next lower level. Potentially infinite numbers of sentences are analyzable into near-infinite numbers of clausal constituents, these in turn analyzable into some hundreds of thousands of word units that are themselves analyzable into some thousands of morphemes, and these being analyzable into some 40 or so phonemes that can be analyzed into an even smaller number of distinctive phonetic features for any given human language. And this componential analysis is exhaustive at all levels—no leftover pieces! Viewed from bottom to top, from combinations of the smallest units (distinctive features) to the uniquely varied patterns of the largest (sentences), this system represents a remarkably efficient way to satisfy the criterion of combinatorial productivity (6)—which anything must meet if it is to be called “a language.” But this is obviously not the only conceivable way to achieve efficiency. If, for example, a species were able to increase the number and complexity of units at some lowest, unanalyzable, level (e.g., hundreds of thousands of meaningful “morphemes”) and simultaneously increase the complexity of the simultaneous patterning of only one higher level (potentially infinite numbers and complexities of “sentences”), then perhaps even greater efficiency could be achieved, but it would take mental capacities far beyond those available to humans—perhaps those available to Arcturians, however.
In introducing this section on the defining characteristics of human languages, and proposing that they could be categorized as “structural” vs. “functional,” I added the parenthetical “hedge” (at least superficially). Now, with specific reference to the hierarchical and componential organization criteria (14 and 15), although they are in no obvious way dependent upon the peripheral sensory or motor structures of primates, they may well reflect ways in which the central nervous systems of higher organisms have evolved—and hence be just as innately determined as those reflecting the constraints imposed by using the vocal-auditory channel. Note that the same hierarchical-componential organization appears in prelinguistic behavior as well: in the comprehension of complex facial expressions of emotion, with their variable component states (upturned vs. downturned mouth, V-shaped vs. A-shaped brows, etc.); in the production of complex skilled acts, like that of “door opening,” where locomoting, arm extending, object grasping, and pulling constituents are sequenced in terms of perceptual feedbacks, are themselves composed of finer motor units, and participate in many different intentional acts (e.g., object grasping as part of the acts of “apple eating” and “hand shake greeting” as well as “door-opening”).
16. The Transferrai-via-Learning Criterion: Human languages are transferred to other members of the species, both generationally over time and geographically over space, via experience (learning) rather than via inheritance (maturation). There is no evidence whatsoever that the offspring of speakers of some particular human language find it easier to acquire that language than any other; in other words, children come into the world cognitively equipped to speak any human-type language—a Japanese infant can learn to speak English or Papago just as (asily as Japanese. Of course, this Japanese infant couldn’t learn Cave-Midgetese very well (without special equipment), and the same would hold for Cave Midget youngsters learning any existing human language, but this could be due simply to the special adaptation of the Cave Midgets to their cavernous environment over the millennia.
How might our intrepid linguists determine whether or not Cave-Midgetese meets the functional criteria for a language being humanoid in type? Form-meaning arbitrariness (12) and discreteness rather than continuousness in signaling meaning shifts (13) would both require pretty thorough familiarization with the language; given that, one might then determine if the components of piping “word” units (analogous to our phonemes) have essentially random relations to their meanings (12) and also whether the shifts in pipings are discretely or continuously variable with respect to shifts in meaning (13) (e.g., if a “vowel” frequency shifted continuously from low to high in order to modulate the meaning of an otherwise constant form to convey “doing something very slowly” to “doing it very rapidly”).
But both of these “tests” already presuppose satisfaction of both hierarchical (14) and componential (15) organization, and in some ways “tests” of these characteristics would be easier: One would merely need to demonstrate (using the ultrasonic visual displays) that distinguishable piping forms enter arbitrarily into diverse larger forms that have distinctive meanings, and that these larger forms enter similarly into more diverse, still larger piping sequences, probably separated by identifiable “pauses”—the differences in these larger forms, of course, displaying predictable contingencies with behavioral (pragmatic) and referential (semantic) phenomena, as indicated by the use of a vastly improved computerized Cave-Midgetese synthesizer. So, given such demonstrations, it would appear that “the language of Cave Midgets” is indeed a human-type language—developed by an early branch of hu-manoids that, in the search for bigger and juicier mushrooms and worms, happened to end up in caverns deep in the earth.
III. NONDEFINING CHARACTERISTICS
OF HUMAN LANGUAGES
All of the defining characteristics considered so far are, ipso facto, universals. There are many other characteristics of human languages-some of them apparently absolute but many of them only statistical universals—which, I would argue, are not defining characteristics. That is, if one encountered something that was a language by criteria 1-6 and, further, was a human-type language by criteria 7-16, then, if it failed to display any one or more of the characteristics to be discussed in this section, it would still be considered a human language—albeit a rather strange one. We can categorize nondefining characteristics fairly reasonably into two types: (i) those that reflect certain intellectual and cultural traits that are common to the human species rather than linguistic regularities per se; (2) those which reflect dynamic interactions among principles of psycholinguistic performance and typically yield statistical rather than absolute universals of language. Only very small samples of such nondefining characteristics can be offered here by way of illustration.
A. BASED ON INTELLECTUAL AND
CULTURAL CHARACTERISTICS OF HUMANS
1. Propositionalizing
All known human languages can be used to create propositional sentences that, in principle, are testable as to their truth or falsity. All humans seem capable of cognizing certain regularities and relations in their physical and social environments and expressing these in linguistic assertions—the sun always rises in the east, a robin is a bird, bears hibernate in the winter, and so on ad libitum. Without such capacities, there could be no science, to be sure, but would we wish to claim that the language of a society of humanoids sans science was therefore not a human-type language? The language of our Cave Midgets might well be so characterized. Interesting in this connection is the complete absence of propositional sentences (apparently) in Washoe’s signings—and, in fact, McNeill (1974) points out that chimpanzee “syntax” seems to be based on sociality more than objectivity, with the addressee-non-addressee distinction being crucial for word ordering (witness Wash-oean “utterances” like you me out, Roger Washoe out, you Naomi peekaboo and you tickle me-Washoe).
2. Prevarication
In all known human languages, messages can be intentionally false, deceptive, or meaningless. This, in a way, is a consequence of being able to propositionalize. All humans seem to be able to produce sentences analogous to the moon is made of green cheese, I was not involved in planning the Watergate Caper, and even colorless green ideas sleep furiously. If speakers of an otherwise humanoid language—let’s say Cave-Midgetese—couldn’t prevaricate, then, of course, they could have neither fiction nor poetry, would be unable to “grease” the social grind with convenient “little white lies,” would be incapable of sarcasm (“Thanks a lot!”) or irony (saying “Isn’t that just great!” while looking at a flat tire on one’s car) or any of the little clevernesses with language, like saying of a certain woman “she’ll make someone a nice husband.” But would such lackluster communication rule out a language as humanoid? I doubt it. One might note in passing that “play” in higher animals (e.g., Pierre making a ferocious growling charge at me that ends in kissing my hand) is a kind of prevarication.
3. Reflexiveness
In all known human languages, messages can be used to talk about other messages and/or their components. In all human languages one can say things like “the word bachelor can refer to ‘an unmarried adult human male’ but also to ‘one who has received the lowest college degree’ or even to ‘a young male fur seal kept off the breeding grounds by the older males.’ ” In writing, we usually make use of tricks like italicizing, capitalizing, and quoting to signal the fact that such forms are words about words, rather than ordinary communications. And witness performative sentences like I christen thee “the Jimmy Carter” for some aircraft carrier of the future, or illocutionary speech acts more generally (see Searle, 1969) like I state that S, I promise that S, I fear (am amazed, am sorry) that S, and I know (believe, doubt, etc.) that S—what I have called “commentative” sentences where the so-called matrix sentence (/verb that . . .) is some kind of comment on the embedded sentence. If our Cave Midget friends lacked this ability to use language to comment on language, then of course they could have no philosophy, no linguistics, no psycholinguistics . . . and no puns. But they could still be said to have a humanoid language, I think.
4. Learnability
Any natural human language can be acquired by any normal human being. If the acquirer is a child, we speak of first-language learning; if it is an adult, already fluent in his native language, then we speak of second-language learning. We might sometime come across a language used by humanoids so primitive or unintelligent that they just simply were not capable of learning the complex structures of any ordinary human language—and this might well apply to the hypothetical Cave Midgets—but I still think we would have to classify it as a human language if it met all of the defining criteria. And the same, of course, would apply in reverse: If it were some humanoid species whose spaceship landed in our backyard, and if they used the vocal-auditory channel and obviously displayed all of our defining characteristics—but their language was so complex in lexicon and in both length and embeddings of sentences that no Earthly human could learn it—then I think we would still humbly have to admit it as a human language.
5. Translatability
Any natural human language can be translated into any other human language. Both the preceding nondefining characteristic (learnability) and this one (translatability) assume, of course, that the lexicon of the acquirer or of the destination language of the translation can be expanded as necessary or be handled by circumlocutions (e.g., the “iron horse” used by American Indians as a translation of “locomotive”). And the same complexity constraints apply as above—either the ordinary human source language could be beyond the intellectual capabilities of the destination language users (translating English into Cave-Midgetese) or the extraordinary humanoid source language could be too complex for the users of ordinary human destination languages (translating the language of those godlike humanoids from outer space into ordinary English). But again, I would argue that these limitations would not rule out the essential “humanness” of the languages involved.
B. BASED ON LANGUAGE PERFORMANCE PRINCIPLES
AND THEIR INTERACTIONS
1. Selection and Combination Rules
Across all languages and levels of units, rules of selection and combination of alternative forms are statistical rather than absolute univer-sals (cf. criteria 14 and 15, hierarchical and componential organization of human languages, above). At the phonological level, each language selects from those differences in sound which the human vocalic system makes possible a small subset of differences that will make a difference in meaning (i.e., are phonemic); certain differences (voiced vs. non-voiced) are much more probable statistically than others (lip flattening vs. lip rounding) across languages. At the syntactic level, each language selects as its grammar a limited subset of “rewrite rules” (for expansion in expression and for contraction in comprehension) from an indefinitely large number of possibilities; again, certain types of grammatical rules are statistically more probable than others (e.g., NP ↔ N + A about twice as likely across languages as NP ↔ A + N and for good psycholinguistic reasons). At the semantic level, each language selects from a potentially infinite number of features some subset for differentiating among the items in its lexicon; although many semantic features are universal, many are not—and, in any case, the distributions of feature weights in languages are statistically variable. This differential rule selection and combination at all levels is the reason why any particular human language must be learned—yet is learnable.
2. A Progressive Differentiation Principle
Across all languages and levels of units, a principle of progressive differentiation of meaning-signaling forms operates, but the extent of differentiation varies statistically. In phonology, for example, only if a high-low vowel distinction is phonemic will a lips flattened-rounded distinction also be phonemic—never the reverse order of differentiation. In syntax, one example would be the Keenan-Comrie (1977) “Accessibility Hierarchy” for relative clause formation: Only if a language permits relativization of the subject NP will it also have relativization of the object NP, and only if these are both developed in the language will relativization of the indirect object NP occur—and so forth, but never with the reverse order of differentiation. Greenberg (1966) documents evidence for progressive differentiation across languages for units at all levels; at the semantic level, for example, only if a language already makes a distinction between singular and plural will it also have a further distinction of plural into dual vs. indefinitely plural. Berlin and Kay (1969) offer evidence for progressive differentiation of color terms cross-linguistically—from the most primitive bright vs. dark, to differentiation of the bright into red vs. nonred, further in the red region into red vs. yellow, and later in the blue region into blue vs. green. Our own cross-cultural studies of affective meaning (Osgood, May, & Miron, 1975) suggest that generalized affective Positiveness vs. Negativeness differentiates into ±Evaluation + ±Dynamism and thence differentiation of Dynamism into ±Potency + ± Activity (i.e., from a one-into a three-feature system). Although the principle of progressive differentiation seems to be a universal of human languages, its interaction with other principles results in a statistical distribution of degrees of differentiation.
3. A Least Effort Principle
Across all languages and levels of units, a principle of least effort operates statistically, such that the higher the frequency-of-usage level (1) the shorter the length of forms, (2) the smaller the number of forms, and (3) the larger the number of different meanings (senses) of the forms used. This principle comes from G. Kingsley Zipf (1949), and he offers the entirely delightful analogy of a skilled artisan working at a long bench, with his production (sentence composition) space at one end and his various tools (here, lexical forms) spread out along it; obviously, it would be most efficient to have the tools most often used closest at hand along the bench, and these tools themselves lightweight, few in number and multipurpose in function. Using language data from English mainly, but also some from Chinese and Latin, Zipf was able to report functions that very neatly supported these hypotheses, and there seems to be no reason to doubt that the same would appear for languages generally. Does the Least Effort Principle hold for linguistic levels other than the lexicon? Greenberg (1966), noting Zipf’s pioneering studies (p. 64), reports relevant data at the phonological level. Although I know of no explicit evidence, one would expect the same to hold for NP and VP constituents—e.g., that unmodified head nouns would be more frequent than elaborated ones (indeed, that diachronically there would be pressures toward simplification, e.g., from knob of the door to door’s knob to doorknob) and that unmodulated verbs (simple presents and pasts, walks and walked) would be more frequent in ordinary language than modulated ones (like t vas walking or had been walking or certainly would have still been walking).
4. Affective Polarity
Across all languages and levels of units, it is statistically universal that affectively positive forms are distinguished from affectively negative forms (1) by marking (either overt or covert) of the negative members of pairs and (2) by priority of the positive members of pairs in both development (in the language and in the individual) and form sequencing in messages. Again Greenberg (1966) provides massive evidence at all levels: thus in phonology, marked nasal vowels are never more frequent in a language than unmarked nonnasal vowels, and Jakobson’s general theory of phonemic development (Jakobson & Halle, 1956) includes the principle of priority for unmarked poles of features; thus in syntax it appears to be a universal (Greenberg, 1963) that affirmation is unmarked (X is Y) and negation is marked (X is NOT Y), and the unmarked active construction (X verbed Y) is universally the basic, natural form and the marked passive (X WAS verbed BY Y) is always viewed as a transformation; thus, in semantics, Animate is unmarked but INanimate is marked overtly, and affectively Positive tall is unmarked but Negative short is marked covertly (note that we normally say how tall is John, not how short is John, unless we are already assuming he is to some degree short). As far as priority of Positive members in the development of language in the species is concerned, the mere fact that it is characteristically Positives that are marked to produce Negatives (happy/UNhappy but not sad/unsad, although this, too, is only statistically universal, as witness untroubled/troubled) clearly implies that the Positives already exist to be marked; as far as priority in individual language development is concerned, DiVesta (as reported in Boucher & Osgood, 1969) has shown that in qualifier elicitation from children of various ages the Positives of familiar opposites (good-bad, big-little, etc.) typically appear earlier and hold tighter frequencies than the Negatives. As to sequencing of such pairs in language production, note first that in stating opposites one usually goes from Positive to Negative (strong-weak rather than weak-strong, fast-slow rather than slow-fast, and so on), and then that familiar idiomatic phrases tend to follow the same rule (see Cooper & Ross, 1975) — no more ands or buts, they hunted fore and aft, are you for or against me, the pros and cons of it, but definitely not the reverse orders.
5. The Pollyanna Principle
Across all languages and levels of units, it is statistically universal that affectively Positive forms and constructions are more diversified, more frequently used, and more easily processed cognitively than affectively Negative forms and constructions. The greater diversity of Positives shows up nicely in our cross-linguistic semantic differential data on (now) some 30 language-culture communities around the world—in the eight-octant space defined by ±Evaluation, ±Potency and ±Activity factors (which system is itself a human universal), the + + + octant (Good, Strong, and Active) is much more densely populated with concepts than the octant - - - (Bad, Weak, and Passive), and the same holds for the Positive vs. Negative directions of each factor taken separately. Greenberg (1966), Boucher and Osgood (1969), and Hamilton and Deese (1971) all present evidence that the Positive members of pairs of word forms are significantly more frequent in usage than their Negative counterparts. As to the ease of cognitive processing, both H. Clark and his associates (Clark, 1971; Clark & Chase, 1972) and Hoosain (1973) in my own laboratory have shown that simple sentences with overt (nots) or with covert negatives (short, ugly, iveak, etc.) take significantly longer to comprehend, and both also find that the same holds for incongruent complex sentences conjoined with but as compared with congruent ones conjoined with and—incongruence being itself a form of negativity. Perhaps the most striking evidence for the Pollyanna Principle will be offered in a paper in preparation by Osgood and Hoosain (1979): Measuring the times required for simply saying appropriately “positive” or “negative” to single words from all sorts of pairs presented randomly, there was a highly significant difference of about 50 msec favoring Positives. In other words, it is easier to “simply get the meaning” of affectively Positive words than the meaning of affectively Negative words, and this even when usage frequency was biased in favor of the Negatives.
IV. HOW MAY HUMAN LANGUAGES
HAVE ORIGINATED?
This is a question that has intrigued speculative philosophers of all periods—and it is as purely speculative as anyone could hope, because there is little likelihood that any of the hypotheses will ever be tested empirically. There is no question but humans have the propensity for vocal-auditory language: All known normal human groups have such a language; all normal children develop competence in a native language of great complexity in what has often been called “a remarkably short time” (although there are many hours in the busy days of childhood); and, as we have seen (criteria 1-16), all human languages are fundamentally of the same type. Yet, given the essential arbitrariness of the phonological, syntactic, and semantic rules of each, particular human languages must be learned.
A. SPECULATIONS ON THE ORIGIN OF LANGUAGE
How did some genius man-ape “get the idea” of communicating with others of his kind by means of vocalizations having distinctive referential properties, thereby enabling him to influence their behaviors? All of the speculations I briefly characterize below—except the first, “mystical” theory—will be seen to contain a grain or two of probable truth, but they are all obviously insufficient. Furthermore, they are limited to “wording” (usually emoting or labeling) and have nothing to say about “sentencing”—but, like the holophrastic stage in child language, it seems certain that the expression of ideas (whole cognitions) would be the fundamental communicative unit.
1. The “Ding-Dong” (Mystical) Theory
Assuming it to be given (i) that meanings are somehow inherent in words and (2) that objects have the power to evoke the words that refer to them, the process of language origin is simply that man-ape sees object DING, DING causes him to say “dong,” and “dong” contains the meaning ding. Nonsensical as this is, it nevertheless has understandable roots in primitive (?) human behavior—the strong tendency toward reification of words: The infant in its random babbling happens to produce the noise “ma-ma” and the fond parent exclaims “Why, the little darling knows me!”; Malinowski (1938) has aptly dubbed this the “bucket theory” of meaning—words, like little buckets, are assumed to pick up their loads of meaning in one person’s mind, carry them across the intervening space, and dump them into the mind of another—and he notes that in some societies a man never reveals his real name (rather, inscribing it on a piece of wood or stone and burying it in some secret place) lest some ill-wisher practice magic on it. Other obvious criticisms are: (1) why don’t all languages have the same names for things? (2) what about names for object-less “things” like function words, verbs, and abstract concepts generally?
2. The “Bow-Wow” (Imitative) Theory
Assuming it to be given (1) that animals and many other things make or have characteristic sounds and (2) that our man-apes had a spontaneous tendency to imitate noises heard, the process of language origin is simply that dog-produced “bow-wow” already has dog meaning as a perceptual sign and man-aped imitative “bow-wow” elicits the same meaning via stimulus generalization. Suggestive evidence would be the use of imitative sounds in most tribal ceremonies and the commonness of onomatopoeia across human languages. Critique: (1) most words in languages are not onomatopoeic; (2) because in different languages there are quite different “natural” imitations of the sounds that dog, cow, cat, rooster, etc., make, and they all follow their own phonological rules, we would have to assume that our man-apes already had a language. However, this theory has a grain of probable truth as a “starter” on the path toward vocal-auditory language.
3. The “Pooh-Pooh” (Interjectional) Theory
Given (1) that, like ape, man-ape has a repertoire of unlearned vocal expressions of affect (grunts, groans, screams), (2) that these are nonrandomly occasioned by the situations he is in, and (3) that these situations (as complexes of perceptual signs) have meanings for him, the “idea” of language originates in the transfer of the meaning of a situation to the vocal interjection as a sign of that situation. Thus Man-ape A breaks the thong he is using to tie his stone axe-head to its handle-stick and mutters a disgusted “pooh-pooh”; and later he says “pooh-pooh, pooh-pooh, pooh-pooh” to ridicule Man-ape B for tripping clumsily over a log—and soon the whole tribe is going around “pooh-pooh”-ing each other! Not only does this theory seem quite reasonable as a “starter” for vocal-auditory language, but it has a pretty solid grain of probable truth in it—prelinguistic cognizing (the meanings of perceived situations) is prior to, and necessary for, linguistic cognizing, a speculation on which I will elaborate momentarily. However, by way of critique: (1) how does the man-ape get beyond the ape—in the naming of entities (NPs) and relations (VPs)? and (2) how does he get from emoting into describing, explaining, warning and the like—that is, into sentencing?
4. The “Yum-Yum” (Gestural) Theory
This was what E. L. Thorndike (see below) dubbed the speculations of Sir Richard Paget (most recently, 1944) about the origin of human languages. Given (1) that our man-ape was a fluent and total gesture maker (with noisy mouth gestures accompanying other bodily gesturings and posturings) and (2) that his gesturing behavior followed the same principle of least effort as that proposed by G. K. Zipf (earliest in The Psychobiology of Language, 1935—see Section IIT,B,3), language originates as the least effortful and least interfering vocal parts of the gesturing are substituted for the total. Thus anticipatory behavior toward a juicy food-object includes tummy rubbing and lip smacking; when the latter is accompanied by voiced exhaling, “yum-yum-yum” is produced. Now imagine two hungry man-apes squatting beside a grubby rotten log: if they keep rubbing their tummies they can’t keep on digging for juicy grubs, but they can keep on “yum-yum”-ing (except when swallowing!)—so the vocalizations become the signs. This also seems reasonable as a “starter” on the path to a human language, and it is consistent with evidence for short-circuiting and amplitude reduction in the acquisition of representation mediators in sign learning (cf. Osgood, 1956, 1979). Critique: (1) we are still left with the problem of how the meaning of “yummy” is abstracted from the situational context (so that Man-ape A can point in some direction, say “yum-yum” to Man-ape B, and get B to go to the grubby log); and again, (2) how do we get beyond the ape and satisfy functional criteria 12-16?
5. The “Babble-Lucky” (Associational) Theory
This is E. L. Thorndike’s “dubbing” of his own theory (1943) of the origins of language. Given (1) that man-apes already had strong tendencies for vocalic play (“babbling,” like human infants), (2) that they already existed in social groups and were surrounded with various natural and artifactual objects, and (3) that they had the cognitive capacities for symbolizing the meanings (Thorndike simply called these X’s) of such objects, language evolves as a result of chance associations of certain random babbles with certain objects and events, these happening to be observed and imitated by the group, and thus becoming socially standardized (hence the “lucky”). Imagine that a bright man-ape spies some clams along the lake shore and happens to babble “uk-uk”; since he already has a meaning for the perceptual sign of CLAM, he associates “uk” with that meaning; gathering up some clams, he brings them back to the cave, calling “uk-uk-uk!” as he comes; the other, less gifted, man-apes imitate him, while observing and feasting on the clams—and “uk” thus comes to refer to CLAM object and all that it signifies. This theory includes the referential properties of others, but it has the advantage of allowing arbitrariness of form-meaning relations while permitting “bow-wow”s, “pooh-pooh”s and “yum-yum”s as starters. It also introduces the critical notion of social standardization. Critique: (1) maybe this mechanism is too chancy and (2) too susceptible to social confusion (what with some man-apes in the group “uk-uk”ing, others “yum-yum”ing, and still others “whiss-whiss”ing about the same clams); most critically, (3) it remains simply a theory of “wording,” with nothing to say about “sentencing.”
B. SPECULATIONS ON THE EVOLUTION OF LANGUAGES
Is it possible that some prehuman simian species like Pithecanthropus had a human-type language? If so, we would expect it to have been more primitive than existing human languages, which presumably have evolved from simpler origins. It seems most unlikely that humanoids suddenly started talking as they dropped from the trees. Many writers have suggested what at least some of the lines of language evolution must have been. For example, in a paper submitted to a volume To Honor Roman Jakobson, J. Bronowski (1967) suggests the following: (i) increasing capacity to delay outgoing vocalization to an incoming auditory message; (2) increasing capacity to separate the affective reactions to, and the denotative significances of, messages; (3) the “prolongation of reference,” thus increased ability to refer backward and forward in time; (4) increasing internalization of language, from being primarily a means of social communication to becoming a means of reflecting and reasoning; (5) the “structural activity of reconstitution,” increasing analysis and synthesis of messages into rearrangeable components. One might also consider my defining junctional characteristics—(12) arbitrariness of semantics, (13) discreteness of signals, (14) hierarchical organization, (15) componential organization, and (16) trans-ferral via learning—as variables that increase in complexity as languages evolve. Unfortunately, unlike skulls and tools, languages leave no traces in or on the earth, and so there is no direct evidence available on language evolution. But what about indirect evidence?
1. The Question of “Primitive” Human Languages
Among the many thousands of extant human languages, are there some less and some more “evolved”? The Old Look answered “yes”: It was suggested that there were stages in development—from isolating languages like Chinese, through agglutinating languages like Turkish, to highly inflecting languages like ... of course! . . . Latin; this notion was dropped when it was realized, with some embarrassment, that most modern Indo-European languages, along with English, were less inflecting, more isolating, than Latin. The New Look answers with a resounding “NO!”: Ethnolinguists have found the languages of culturally primitive (near Stone Age level) peoples fully as complex as those of highly civilized (??) peoples “like us”; complexity of language-based conceptual systems (for kinship relations, mythology, etc.) appears to vary quite independently of levels of technological development; so the very question of “relative primitiveness” of extant human languages is meaningless.
But perhaps we need to take a Fresh Look—and ask ourselves just what the criteria of “primitiveness” might be.
a. Structural Simplicity Vs. Complexity? Just how this would be indexed—numbers of phonemes, of semantic features, of syntactic rules, or maybe ease of descriptive linguistic analysis?—is entirely unclear; if anything, it would appear that there probably was prehistoric increase in overall complexity but decrease during recent historic times.
b. Size of Vocabulary? This clearly covaries with cultural development (but one should use lay speakers in such comparisons), and, in principle (given criterion 6, combinatorial productivity) any language can be used to talk about anything.
c. Efficiency in Communication? In the information-theoretic sense, most languages at most levels (e.g., N-features/N-phonemes) run at about a 50% redundancy level in order to maximize the probabilities of messages “getting through”; in my foreign research travels, however, I have noticed rather marked differences in the length of signs used to communicate the same messages in airlines (e.g., no smoking, fasten seat belts) — but then, cultures where we might expect language primitive-ness usually don’t have airlines! Maclay and Newman (1960) devised an interesting measure here: the number of morphemes needed to communicate the same information when Person A tells Person B on the other side of a screen which forms to select from a set to match his own; there were marked effects of both negative feedback and homogeneity of forms upon increasing the N of morphemes, but unfortunately no comparative studies across languages were made—only American English subjects being used.
d. Degrees of Concreteness? Roger Brown (1970, pp. 19-22) reports a comparison of the nouns and verbs used by adults vs. children in terms of their concreteness (vs. abstractness)—a kind of “picturability” (a point-at-able/non-point-at-able ratio). For nouns, 67% of the children’s but only 16% of the adults’ 1000 most frequent were “picturable”; for verbs, 67% of children’s but only 33% of adults’ were “picturable.” Might this be used as an index of “primitiveness”? Perhaps—but this too, would seem to be more cultural than linguistic, when adult speakers are being compared.
e. Ease of Learning? It seems reasonable that more primitive languages should be easier to learn, but the “how” of going about this is obscure. Could one show that isolating, uninflected Chinese is learned more quickly as a first language than is, say, highly agglutinative, inflected, and left-branching Turkish—and just what criteria of “learning” would be used? Could it be shown that Chinese can be learned more quickly by Turks as a second language than Turkish can be learned by Chinese? As Uriel Weinreich has amply documented in his Languages in Contact (1953), the interactions at all levels of units in second-language learning are incredibly complex and bi-language specific. A related index would be the degree of difference (linguistically described) between adult and child language—comparatively across languages, of course—but at just what level do we define the “child” language?
So, having taken a Fresh Look at the question of relative “primitiveness” of extant languages in terms of criteria that might be used, we seem to be about where we started—with a set of possible criteria but little or no available evidence.
2. The Notion of Recapitulation
The notion that ontogeny (development of the individual organism) recapitulates phylogeny (evolution of its species) is a familiar one. Could we get some idea of how language might have evolved in human species from the regularities of language development in its contemporary offspring? In an early tracing of stages in language development, Ervin and Miller (1963) report that in the course of language development there are increases in the number of phonemic distinctions, the number of grammatical classes, and the average length of utterances; holophrastic words (with contextually redundant remainders of full cognitions unexpressed) appear before word combinations (constructions), lexical word forms (distinctively semantic) before function word forms (primarily syntactic), and ordering rules (e.g., Agent-Action-Recipient) before inflections (which permit permutations in ordering).
More recently, Brown (1973), taking a more sentential approach, specifies five stages in development: (1) expression of relations or roles within simple sentences (the case roles of various noun phrases in relation to the verb phrase); (2) modulations of meaning within the simple sentence (modulating the N with number tags, with qualifiers, and the like, and the V with tense indicators and adverbials); (3) diversifying the modalities of simple declarative sentences (into yes-no questions, wb-constituent questions, negations, imperatives and the like); (4) embedding of sentences, expressing simple cognitions, within others (relative clauses, object noun-phrase complements, and the like); and (5) coordination of simple sentences and propositional relations (simple and/but complexes, adverbial main/subordinate clauses, and so forth). It should be stressed that Brown believes this 1-5 ordering to be that of development in children.
If we have some faith that ontogeny does tend to recapitulate phylogeny, then we might speculate as follows about the evolution of human languages:
1. They probably began with holophrastic expressions of complete perception-based cognitions, with purely semantic (often affective) and pragmatic functions.
2. As the sheer numbers of such expressions increased, the number of phonemic distinctions and grammatical classes had to increase, along with socially agreed upon rules of sequencing the expression of such classes.
3. Starting with the simplest sentential expressions of full cognitions (single-word nominals related in action and stative relations by single-word verbs, like dog took leg-bone and woman (is) in cave), the pressures to express finer semantic distinctions led to modulations into NPs and VPs (your mangy mutt sneakily stole away with my juicy leg-bone and my woman must have been in your warm cave) and the needs for more pragmatically effective social communication led to diverse modalities of sentencings (like where is my woman?, she is not in our cave, and tell her to come home).
4. Relatively late in the evolution of languages, complexes of cognitions came to be conjoined in coordinate and embedded sentential forms, with the rather aesthetic need to avoid repetitive orderings within sentoids (clauses) leading to diverse inflections and transformations—all of which eventuated in sentence lengths getting longer and longer (like this one!).
C. A “NATURALNESS” (REPRESENTATIONAL) THEORY OF THE ORIGIN
OF LANGUAGES
I’ve been unable to come up with a catchy, rhyming dubbing for my own speculations—but perhaps some reader will succeed where I’ve failed and let me have his suggestion. The broadest notion underlying the general theory of cognizing and sentencing I’ve been working on (Osgood, 1979) is this: that, both in the evolution of the species and in the development of the individual human, the cognitive structures that interpret sentences received and initiate sentences produced are established in prelinguistic experience, via the acquisition of adaptive behaviors to entities perceived in diverse action and stative relations. I suppose one might also call this an “article of faith.” However, it follows from two assumptions that would rather obviously seem to be true: (1) that humanoids, before they had language, must have had capacities (a) for cognizing the significances of events going on around them, and (b) for learning to behave appropriately in terms of such significances—if the species were to survive; and (2) that children of contemporary humans, before they have language, display exactly the same capacities for acquiring the significances of perceived events and reacting with appropriate intentional behaviors.
This notion implies what I have called a Naturalness Principle for sentencing in language behavior—namely, that the more sentences produced or received correspond in their surface structures to the cognitive structures developed in adaptive prelinguistic perceptuomotor experience, the greater will be the ease of processing them. This functional principle in turn has potentially testable implications for what should be universal (1) in the development of the language by the young of our species, (2) in the processing of language by the adults, and (3) in the evolution of language in the species itself. As to children (1), the greater the correspondence of alternative structures to the prelinguistically established ones, the earlier should be the acquisition of their processing—in imitating, in comprehending (Simply Acting out), and in expressing (Simply Describing)—both within any given language and cross-linguistically for bilingual children (cf. Slobin, 1973); as to adults (2), the greater the correspondence, the greater will be the speed of comprehending and producing sentences and the more frequent will be the use of such sentences in various communication tasks (e.g., in Simply Describing perceived events); as to language evolution (3), the greater the correspondence, the earlier such structures should appear in, and the more universal they should be across, human languages.6
1. A Bit of Behavior Theory
The specification of what is “natural” involves hypotheses deriving both from Representational Neobehaviorism (see Osgood, 1979) and from psychological intuitions about what should be natural in the prelinguistic cognizing of young children7—in contrast to linguistic intuitions about what is grammatical in the sentencing of adults. So a bit of behavior theory is in order. Our concern here will be limited, first, to presenting a basic sign-learning principle and its extension to an equally basic feature-learning principle—both being part of my own generalization of classical Hullian (Hull, 1943) mediation theory to a representational and componential mediation theory of meaning—and, second, to emphasizing the intimate parallelism between nonlinguistic and linguistic cognizing (or between “Things and Words,” reversing appropriately the title of Roger Brown’s justly famous book, 1958) by offering an “emic” principle and an “ambiguity” principle of neobehaviorism.
a. A Sign-Learning Principle. Just like apes (and even rodents) before them, it seems likely that humanoids developed the capacity for “getting the meanings of” wholistic perceptual signs of things prior to the capacity for analyzing out the distinctive features that make the differences in meanings. This behavioral principle may be stated as follows: When a percept that elicits no predictable patter?i of behavior has repeated, contiguous, and reinforced pairings with another percept that does elicit predictable behavior, the former will become a sign of the latter as its significance, by virtue of becoming associated with a mediation process, this process (1) being some distinctive representation of the total behavior produced by the significant and (2) serving to mediate overt behaviors to the sign appropriate to (“taking account of”) the signifie ate. Such a principle has been implicit in all of the speculations above (except the “ding-dong” theory) about the origins of language (e.g., the X’s in Thorndike’s “babble-lucky” theory). Note that, both in evolution of the species and in development of the individual, nonlinguistic percepts of familiar entities in diverse action and stative relations (DADDY SPANKING FIDO; KITTY BEING ON PILLOW) will acquire meanings prior to the linguistic percepts (word forms) that will later represent (“stand for,” “refer to”) them—and, indeed, the representational processes formed in such prelinguistic experience will typically provide “prefabricated” mediators in later language acquisition. The representational mediation process that comes to be associated with a sign (perceptual or linguistic) as a dependent event in the nervous system is its significance (in comprehending); the same process as an antecedent event is the intention behind the overt behaviors mediated by a sign (in expressing).
b. A Feature-Learning Principle. Just as the overt behaviors made to significates are typically a set of overt responses which together constitute and “act,” so also in theory are the mediation processes derived from this total behavior, and now elicited by a sign, a set of mediator components. To the extent that pairs of signs elicit reciprocally antagonistic mediator components, these componential antagonisms will become the “differences that make a difference” in meaning.
A simple, but paradigmatic, demonstration (Lawrence, 1949) will illustrate: In a simple T-maze, with the upper “arms” at the choice point having both BLACK vs. WHITE walls and CHAINS vs. NOCHAINS (soft curtain) distinguishing right from left sides (but with random right-left locations across trials), members of one of four possible groups of rat subjects will be rewarded with food pellets if they go toward the BLACK and punished by sudden loss of support if they go toward the WHITE, with the CHAINS/NO CHAINS location being random with respect to differential reinforcement; being reasonably bright little fellows, these rats rapidly learn to get food and avoid falls. The crucial thing here is that, after this experience, they will learn to go to the WHITE and avoid the BLACK much more rapidly than to go to the CHAINS and avoid the NO-CHAINS; in other words, they have learned to “pay attention to” the differences that make a difference in meaning—here, anticipated reinforcement (“hope”) vs. anticipated punishment (“fear”)—in an otherwise constant behavioral situation and to “disregard” differences that do not make a difference.
I call this little experiment “paradigmatic” because it suggests a general model for the development of both phonemic and semantic distinctive features in human languages. In this connection, it is interesting to me that, although many linguists and most psycholinguists assume that phonemic and semantic feature distinctions must be acquired via experience (given the obvious fact that languages differ in what features come to “make a difference”), I have searched in vain for any explication of the crucial how of this learning. Note that this feature-learning principle is a logical extension of the sign-learning principle, once a componential conception of the mediation process is substituted for an undifferentiated global conception. This is the entrée, I think, of Neobehaviorism to a theory of meaning and reference—and, with structuring of the semantic system, ultimately to a theory of sentencing.
The “emic” principle of Neobehaviorism. By virtue of the fact that both things and organisms are mobile with respect to each other, along with the fact that environmental contexts are changeable, it follows that the distal signs of things will be variable through many stimulus dimensions. Thus, for the human infant, MOTHER’S FACE will vary in retinal-image size as she approaches him and in brightness and hue as time-of-day shifts from dawn through midday into twilight; similarly, for man-ape on the hunt, retinal size of ANTELOPE percept must have varied in size and hue as he moved stealthily in on his prey. But because these varying percepts are associated with the same significate and the behavior it produces (cf. the sign-learning principle above)—mother eventually coddles and comforts baby and antelope eventually gets killed and eaten—there will be extension of the common mediator (meaning) across such sets of percepts. Therefore the differences within such sets will be differences that do not make a difference in meaning.
This is the “that-ness” or “thing-ness” in perception, and it is the basis for the constancy phenomenon—long familiar to psychologists. Note that what we have here is a class of variable signs having a common significance: if we substitute sounds for signs, we have the definition of the “phoneme” (a class of physically different sounds having the same significance in the phonemic code, e.g., the /k/ in key, cope and coo) ; similarly, at the lexical level, word forms like “mother” and “linguist” retain their denotative significance regardless of variations in intonation, stress, or voice quality of the speaker. Equivalently on the output side of the equation, classes of nonlinguistic and/or linguistic behaviors come to be associated with the common mediation process and therefore can be said to be expressions of the same intention—the child (older now) or the man-ape locomoting toward, reaching for, grasping, and biting the APPLE object . . . but not for a percept-class having a different significance, e.g., a RED-HOT COAL!
But there is also a “how-ness” or “where-ness” in perception. Neither human child nor man-ape emits the various responses expressing the same intention at random. Neither would emit biting, then grasping, then reaching movement—in thin air—in that order or before approaching the APPLE, anymore than Caesar would have announced “Vici, vidi, veni”! Rather, the distinctive percepts as stimuli converge with the stimulus effects of the common mediator to modulate the probabilities of alternative movements (expressions)—thus, the small visual angle of APPLE (ON-TABLE-OVER-THERE) plus its meaning as an edible object converge on locomoting toward, a larger visual angle plus the same meaning converge on reaching for, and so on. This fusion of convergent and divergent hierarchies is what is called control and decision in representational neobehaviorism. Note that there is thus a syntax of behaving just as there is a syntax of talking—and again, the former is clearly prior to the latter and therefore can serve as a cognitive model for the latter.
c. The “Ambiguity” Principle of Neobehaviorism. When percepts are constant, but the mediators associated with them are variable, we have the conditions for perceptual ambiguity—the same sign having more than one meaning. Familiar examples from the psychological laboratory would be the Necker Cube (an outline cube which flips between “from the side” to “from above” perspectives) and the Miles Kinephantoscope (where the shadow of a rotating bar on a rod can be seen as “flapping toward me,” “flapping away from me,” “shrinking and expanding,” “rotating left or right,” etc.), and there are many others. The child may be ambiguated by a fuzzy black something on the floor (BALL OF THREAD or SPIDER?)-until it moves; our man-ape may be ambiguated by a sudden movement of some bushes (WOMAN-APE or SABER-TOOTHED TIGER?)-until it growls!
This is strictly analogous to linguistic ambiguity—homonomy, and more generally (and finely) polysemy of forms, at lexical (he went to the BANK), syntactic (the SHOOTING OF THE HUNTERS was terrible), and pragmatic (CAN you open the window? as an indirect request or an inquiry as to capability) levels. Disambiguation via contextual signs can occur within channels: perceptual-by-perceptual, as when that pretty, long-haired person says “where’s the toilet?” in a DEEP BASS VOICE; linguistic-by-linguistic, as in he ROWED to the bajik. Or it can be across channels: perceptual-by-linguistic, as when the experimenter induces irresistible shifts in the Kinephantoscope perceptions of the subject from “pirouetting to the left” to “flapping behind” just by saying those words; linguistic-by-perceptual, as when one co-ed says to her companion as another co-ed in a ramz-miniskirt just passes them on a campus path, “she also dyes her hair”! The crosschannel disambiguations have a most significant implication—namely, that at some deep cognitive level, nonlinguistic (perceptual) and linguistic “wordings” and “sentencings” must share the same representational system—otherwise, events in the two channels would pass each other like ships in the night.
2. Some Speculations About Naturalness in Sentencing
In discussing the syntactic criterion (5) for anything to be called a language, in connection with Octopian, I introduced this Naturalness notion. It is assumed that the prelinguistic structures of the simple cognitions underlying sentoid (clause) comprehending and expressing are tripartite in nature (cf. Greenberg, 1963), and further that these cognitions are “SVO” in their ordering of components (ENTITYi-RELATION-ENTITY2). This makes the very strong prediction that human languages that are SOV in adult type will display evidence of originating from SVO: There is diachronic evidence for many shifts in type from SOV to SVO in historic times, but not a single case of an SVO-to-SOV shift, without external pressures (invasion, cultural dominance, etc.); in the “diachronics” of language in children, one would expect evidence for SVO ordering in the two-to three-word stages of development in SOV languages before the adult language ordering takes over.8
It is also assumed that there are two basic types of SVO cognitions: (1) expressing action relations (where ACTOR-ACTION-RECIPIENT is the natural ordering in cognizing, because of the characteristically +animate—and often +human—semantic coding of actors as against the -animate—or at least relatively passive—coding of Recipients); and (2) Stative relations (where FIGURE-STATE-GROUND is the natural ordering, because of the Gestalt-like characteristic +Salience of Figures and “Salience of Grounds). Thus, both for all human children and for our hypothetical man-apes en route to language, one would expect prelinguistic cognizings like MAN KICK DOG (action) and DOG ON FLOOR (stative) to be more natural than DOG KICKED BY MAN or FLOOR UNDER DOG. If the underlying Naturalness Principle is valid, then one would expect the earliest sentencings—by contemporary child and by now-extinct man-ape—to be active (rather than passive) and “figure-ative” (rather than “ground-ative”). The fact that we don’t even have any names for the latter suggests that the cognizing of stative relations may well be more basic and primitive than the cognizing of action relations (again, cf. Osgood, 1979, for elaboration).
What about evidence for these naturalness predictions? Only a bit of the most relevant evidence can be given here. Osgood and Bock (1977) present a reanalysis of data on the simply describing of a variety of little demonstrations with balls, blocks, tubes, poker chips, and the like by adult American English speakers (previously reported in a paper by the first author, titled “Where Do Sentences Come From?”). Evidence for naturalness in describing both action and stative relations was overwhelming: For stative relations, despite the grammatical availability of GROUND-STATE-FIGURE sentencings (e.g., a plate is holding a ball, a spoon and a poker chip or a ball is being held by the man), these rarely occurred (we got equivalents of a ball, a spoon and a poker chip are on a plate [20/0/6]9 and of the man is holding a ball [24/1/1]); for action relations, despite the availability of ordinary passives (e.g., affirmatives like the tube was hit by the orange ball and negatives like the tube was not hit by the orange ball), again they practically never occurred (we got equivalents of the orange ball hit the upright tube [21/0/5] the orange ball did not bit[missed, passed, etc.]the tube [21/5/0—the “flubs” not expressing negation in any way]). Our Center for Comparative Psycholinguistics is right now in the process of collecting such simply describing data from native speakers of some 20 languages around the world, using a color-film containing 70 similar demonstrations designed with “psycholinguistic malice aforethought”; we want to see how diverse languages equivalently take account of various cognitive distinctions (presumably universal) in the surface structures of their sentences.
At this point one might reasonably ask why do “unnatural,” yet grammatical, sentencings ever occur? The title of the Osgood and Bock (1977) paper, “Salience and Sentencing: Some Production Principles,” implies the answer: All languages provide ways for their speakers to move constituents of messages that are salient to them forward (earlier) in time of production; this, by the way, is consistent with the Hullian motivation principle (Hull, 1943)—performance equaling habit strength times drive. The Naturalness Principle itself is based on the notion of the relative salience of Actors over Recipients and Figures over Ground, other things being equal; the principle can be overridden by (i) the vividness (inherent salience, affective intensity, of semantic features) and/or (2) the motivation of speaker (attributed salience, due to topicality, interest, identification, etc.) with respect to nonactor or nonfigure constituents.
Kay Bock’s thesis (summarized in Osgood and Bock, 1977), investigated the effects of vividness, using a variety of optional transformations in English. In the following examples, I capitalize the noun-phrase alternatives that were assumed to be + Vivid in each case: the dative (the boy tossed the FRISBEE/ball to the dog/ST. BERNARD / the boy tossed the ST. BERNARD/dog the ball/FRISBEE), the passive (the BULLDOZER/truck crushed the flowers/DAFFODILS / the DAFFODILS/flowers were crushed by the truck/BULLDOZER), the genitive (the powwow was held in the WIGWAM/tent of the chief/SITTING BULL / the powwow was held in SITTING BULL’S/the chief’s tent/WIGWAM), and a number of others. The prediction was that where the +Vivid alternative was already in the temporally prior position (the boy tossed the FRISBEE to the dog or the boy tossed the ST. BERNARD the ball) there would be few shifts in reproduction, but where the +Vivid was given in the temporally subsequent position (the boy tossed the ball to the ST. BERNARD or the boy tossed the dog the FRISBEE) there would be significant transformational shift tendencies. These vividness predictions had to be tempered by the Naturalness Principle, of course: If the given structures were already +Natural, they would tend to be reproduced that way; if they were “Natural, they would tend to be shifted back into the natural form in reproduction, regardless of the Vividness relations. In other words, 50% of the time Vividness and Naturalness Principles would reinforce each other and 50% of the time they would compete with each other. The results were generally consistent with these expectations.
An hypothesis derived from “putting on the booties of childhood” received interesting support from both cross-language linguistic data and from experimental psycholinguistic data, as reported by Osgood and Tanz (1977). The intuition was about the prelinguistic child’s cognizing of bitransitive situations (prototypically, Person A transfers inanimate object X to Person B), where the X is perceptually a part of the transfer relation and Person B perceptually the real object (little Suzie perceiving BIG-BROTHER GIVES-BOOK-TO BIG-SISTER)—and this led to the title of our paper, “Will the Real Direct Object in Bitransitive Sentences Please Stand Up?”
For the cross-language data, a number of predictions were verified such as: (1) that the so-called indirect object (D for dative) should be more frequently marked than the so-called direct object (O for objective) and that this should be more often the case when D is closest to the verb (i.e., the unnatural transform); (2) the strong prediction—that in languages where O is marked in unitransitive clauses but not in bitransitive, the same marker will be applied to the D—was upheld in a surprising number of cases; (3) that in languages where nouns are incorporated in verbs, O-incorporation should be greater than D-incorporation (clearly upheld); and (4) that the natural SVOD structure should be less constrained transformationwise than the unnatural SVDO structure. Three psycholinguistic experiments supported the predictions (5) that natural SVOD forms would be more faithfully recalled as given than SVDO forms; (6) that unnatural SVDO forms would be more frequently shifted to SVOD in recall; and (7) that as probes for single-word associations, V should produce O more often than D, and O should produce V more often than D does. So it would appear that, despite what grammarians have been telling us for centuries, the real direct object in bitransitive sentences is the so-called indirect object!
Finally, the Naturalness Principle makes certain predictions about the ordering of clauses in complex, conjoined sentences—namely, that where there is a natural order in prelinguistic experience with complex events, this order in sentencing will be either required or at least preferred across languages. In the simple junction mode of conjoining, either order is natural (Mary sang and John played the guitar or the reverse). But for the temporal mode there is a natural order in experience (and thus Mary got dressed and went to the party, but not! Mary went to the party and got dressed);10 similarly for the causal mode (it was raining and John got wet, but definitely not! John got wet and it was raining) and the combined causal-temporal mode (Dan ate some poisonous mushrooms and got sick but definitely not! Dan got sick and ate some poisonous mushrooms).
Opačić (1973) measured processing times (for judging whether a given pair of clauses, then a given conjoiner, was “possible” or “impossible”) and then the complex sentence the subject produced if he had said “possible”—for these and other modes of conjoining. For all modes in which two orders are possible, one natural (N) and the other unnatural (U), natural-given and natural-retained (NN) had the shortest latencies, UN were next in ease of processing (i.e., where the subject restores the natural order), UU came next (literal memorizing), and NU (shifting from the natural to the unnatural ordering) was the most time-consuming.
Results of this sort are most encouraging and support the general notion that the cognitive structures developed in prelinguistic perceptual experience do in fact provide the “most natural” cognitive bases for comprehending and producing sentences—both in the evolution of language in the human species and in the development of its contemporary children. But, of course, neither this nor the demonstrably intimate parallelism between nonlinguistic and linguistic cognizing means that such prelinguistic capacity in itself constitutes “a language.” It lacks three of the criteria for anything being a language—nonrandomly recurrent signals in some channel, reciprocality in sending and receiving such signals, and combinatorial productivity—precisely because such prelinguistic cognizing cannot be abstracted from the perceptual and motor chains that bind it to reality.
__________
Reprinted from Psycholinguistic Research: Implications and Applications, edited by Doris Aaronson and Robert W. Rieber, © 1979 by Lawrence Erlbaum Associates, Inc.
NOTES
1. In formulating my own notions about “what is a language” here and in the sections on human languages that follow, I am indebted to the participants in a conference I attended on Universals of Language, sponsored by the Social Science Research Council in 1961 and later published in a volume under the same title edited by Joseph H. Greenberg (1963)—particularly papers by Greenberg, Charles S. Hockett, and Uriel Weinreich.
2. See Osgood (1979) for detailing of these Naturalness Principles—including, of course, the fact that human languages also utilize adverbial substitutes for simple junction (and) and disjunction (but) which permit clausal permutations (which might well be unnecessary in Octopian).
3. See K. von Frisch (1974) and James L. Gould (1975) for updating references on communication in the bee.
4. With regard to semantics, particularly impressive to me was the fact that Sarah would make the same “semantic-feature” choices for the plastic sign of, say, an APPLE (Round rather than Square, Stemmed rather than Nonstemmed) as for the apple object itself—when the sign itself happens to be a flat purple triangle!
5. I can’t resist noting (with no little remorse) that during my tenure on the University of Illinois Research Board in the early 1960s a request for support of precisely this type of study was voted on favorably, but the young couple who were to raise the baby chimp backed out—otherwise this breakthrough might have occurred at the University of Illinois!
6. And, I would add, the greater the correspondence of competence-based grammars to prelinguistically based cognitive structures, the higher should be their evaluation in competition with alternative grammars.
7. In preliminary mimeographed versions of my evolving APG (Abstract Performance Grammar) I have used as a parenthetical subtitle “Picking Oneself Up by One’s Booties”!
8. During the summer and fall of 1977, two of my students tested this radical hypothesis—S. N. Sridhar and Annette Zehler in Bangalore and Hiroshima, respectively, Kannada and Japanese both being SOV languages, but otherwise rather different (cf. Osgood, 1979).
9. With an N of 26 subjects, the first value is the number fitting the naturalness prediction, the second the number of “flubs” (e.g., a subject opened his eyes too late!), and the third is the number contrary to the prediction.
10. I use the! symbol to emphasize the fact that although such sentences as these may be completely unacceptable (indeed, mind-boggling), they are not ungrammatical—which raises some interesting questions about the notion of grammaticality, which we cannot go into here.
REFERENCES
Berlin, E., and Kay, P. Basic color terms: Their universality and evolution. Berkeley: University of California Press, 1969.
Boucher, J., and Osgood, C. E. The Pollyanna hypothesis. Journal of Verbal Learning and Verbal Behavior, 1969, 8, 1-8.
Bronowski, J. In To honor Roman Jakobson. The Hague: Mouton, 1967.
Bronowski, J., and Bellugi, U. Language, name and concept. Science, 1970, 168, 669-673.
Brown, R. Words and things. Glencoe, 111.: Free Press, 1958.
Brown, R. Psycholinguistics: Selected papers. New York: Free Press, 1970.
Brown, R. A first language: The early stages. Cambridge, Mass.: Harvard University Press, 1973.
Clark, H. H. The chronometric study oj meaning components. Paper presented at the CRNS Colloque Internationale sur les Problèmes Actuels de Psycholinguistique, Paris, December 1971.
Clark, H. H., and Chase, W. G. On the process of comparing sentences against pictures. Cognitive Psychology, 1972, 3, 472-517.
Cooper, W. E., and Ross, J. R. World order. In Papers from the parasession on functionalism. Chicago: Chicago Linguistics Society, 1975.
Ervin, S. M., and Miller, W. R. Language development. In Child psychology, 62nd Yearbook, National Society for the Study of Education. Chicago: University of Chicago Press, 1963.
Gardner, B. T., and Gardner, R. A. Evidence for sentence constituents in the early utterances of child and chimpanzee. Journal of Experimental Psychology: General, 1975, 104, 244-267.
Gardner, R. A., and Gardner, B. T. Teaching sign language to a chimpanzee. Science, 1969, 165, 664-672.
Gould, J. L. Honey bee recruitment: The dance-language controversy. Science, 1975, 189, 685-692.
Greenberg, J. H. Some universals of grammar with particular reference to the order of meaningful elements. In J. H. Greenberg (Ed.), Universals of language. Cambridge, Mass.: MIT Press, 1963.
Greenberg, J. H. Language universals. In T. A. Sebeok (Ed.), Current trends in linguistics: III. Theoretical foundations. The Hague: Mouton, 1966.
Hamilton, H. W., and Deese, J. Does linguistic marking have a psychological correlate? Journal of Verbal Learning and Verbal Behavior, 1971, 10, 707-714.
Hoosain, R. The processing of negation. Journal of Verbal Learning and Verbal Behavior, 1973, 12 y 618-626.
Hull, C. L. Principles of behavior: An introduction to behavior theory. New York: Appleton-Century-Crofts, 1943.
Jakobson, R. C., and Halle, M. Fundamentals of language. The Hague: Mouton, 1956.
Keenan, E., and Comrie, B. Noun phrase acceptability and universal grammar. Linguistic Inquiry, 1977, 8, 63-100.
Lawrence, D. H. Acquired distinctiveness of cues: I. Transfer between discriminations on the basis of familiarity with the stimulus. Journal of Experimental Psychology, 1949, 39, 770-784.
Maclay, H., and Newman, S. Two variables affecting the message in communication. In D. Wilner (Ed.), Decisions, values and groups. New York: Pergamon Press, 1960.
Mahl, G. F. People talking when they can’t hear their voices. In A. Siegman and B. Pope (Eds.), Studies in dyadic communication. New York: Pergamon Press, 1972.
Malinowski, B. The problem of learning in primitive languages. Supplement in C. K. Ogden and I. A. Richards (Eds.), The meaning of meaning. New York: Harcourt Brace, 1938.
McNeill, D. The acquisition of language. New York: Harper, 1970.
McNeill, D. Sentence structure in chimpanzee communication. In K. J. Connolly & J. Bruner (Eds.), The growth of competence. New York: Academic Press, 1974.
Opačić, G. Natural order in cognizing and clause order in the sentencing of conjoined expressions. Doctoral dissertation, University of Illinois, 1973.
Osgood, C. E. Behavior theory and the social sciences. Behavioral Science, 1956, 1, 167-185.
Osgood, C. E. Lectures on language performance. Berlin, Heidelberg, New York: Springer, 1979.
Osgood, C. E., and Bock, J. K. Salience and sentencing: Some production principles. In S. Rosenberg (Ed.), Sentence production: Developments in research and theory. Hillsdale, N.J.: Lawrence Erlbaum Associates, 1977.
Osgood, C. E., and Hoosain, R. Polly anna II: It is easier to ‘simply get the meanings’ of affectively positive than affectively negative words. Manuscript in preparation, 1979.
Osgood, C. E., May, W. H., and Miron, M. S. Cross-cultural universals of affective meaning. Urbana, 111.: University of Illinois Press, 1975.
Osgood, C. E., and Tanz, C. Will the real direct object in bitransitive sentences please stand up? In Linguistic studies presented to Joseph Greenberg. Saratoga, Calif.: Anma Libri, 1977.
Paget, R. A. The origin of language. Science, 1944, 99, 14-15.
Premack, A. J., and Premack, D. Teaching language to an ape. Scientific American, 1972, 227, 92-99.
Premack, D. Language in chimpanzee? Science, 1971, 172, 808-822.
Searle, J. R. Speech acts: An essay in the philosophy of language. London: Cambridge University Press, 1969.
Slobin, D. I. Cognitive prerequisites for the development of grammar. In C. A. Ferguson & D. I. Slobin (Eds.), Studies in child language development. New York: Holt, Rinehart and Winston, 1973.
Thorndike, F. L. The origin of language. Science, 1943, 98, 1-6.
von Frisch, K. Decoding the language of the bees. Science, 1974, 185, 663-668.
Weinreich, U. Languages in contact. New York: Linguistic Circle, 1953.
Winograd, T. Understanding natural language. Cognitive Psychology, 1972, 3, 1-191
Zipf, G. K. The psychobiology of language. Boston: Houghton Mifflin, 1935.
Zipf, G. K. Human behavior and the principle of least effort. Cambridge, Mass.: Addison-Wesley, 1949.
Human Communicative
Behavior: A Biological Model
A. LINGUISTICS AND LANGUAGE
THE DEFINITION OF LANGUAGE
The vast majority of texts dealing with the subject of linguistics define it as “the science of language” (Robins 1964) or, less cryptically, as “the scientific study of language” (Lyons 1968). The prominent inclusion of the term science in this definition is probably premature, since the manner in which scientific method can apply to an essentially descriptive discipline such as linguistics is far from clear (cf. Botha 1968; Derwing 1973; Dingwall 1978). Our concern here, however, is not directly involved with questions of methodology, but rather with the object of study, language. Definitions of this phenomenon run the gamut from “[the] set (finite or infinite) of sentences, each finite in length and constructed out of a finite set of elements” (Chomsky 1957: 13) to “a communication system that is capable of transmitting new information” (Lieberman 1975:6). Either of these extreme definitions would appear quite applicable to various zoosemiotic systems (see Thorpe 1972:28-29) as well as to the output of various mechanical devices such as a digital clock. This is clearly not what the majority of linguists mean by language.
For them, language is equivalent to human communication only. Cairns and Cairns make this abundantly clear when they state in a recent introductory text:
The most complicated communication system in the animal kingdom is, of course, that of human beings, Language. One purpose of linguistics is to describe those properties of Language that are common to all human beings and that differentiate it from the communication systems of other animals. [1976:5]
Since the aim is to define language as a species-specific communication system, the output modality of this system is often stressed, since it clearly distinguishes it from systems employed by other animals (see Dingwall 1975). Thus, the standard definition of language adds this feature for good measure, as the following definitions demonstrate:
Language . . . refer [s] to that verbal communication system developed and used by humans. [Glucksberg and Danks 1975:23]
. . . [t]he subject matter of linguistics is confined to verbal messages only. [Sebeok 1977:1056]
The problem with using an output modality as the defining characteristic of language is that it is inconsistent with the goal set forth by Cairns and Cairns of describing “those properties of Language that are common to all human beings” (emphasis added). While it is true that only humans use articulate speech to communicate, it is not true that all humans who communicate use articulate speech. Either one defines language in such a way as to exclude humans that communicate non-verbally or one defines it in such a way as to include all humans who communicate. If the latter option is chosen, one cannot claim that it is a communication system confined to humans alone (see fig. 1).
Despite all protestations to the contrary (see Chomsky 1972), linguistics has remained essentially a descriptive discipline devoted to the study of corpora that consist of sets of sentences. Language is a noun with no verbal form. The goal of linguistics is the structural analysis of an entity, not a behavior. Until very recently, this entity has been studed in virtual isolation from the organism in which it resides and from the contexts in which it is produced. The isolation is reinforced by the distinction between competence and performance that separates knowledge of the structure of language from its use and asserts that competence alone is the appropriate domain of theoretical linguistics (Chomsky 1965).
Recent suggestions for broadening the scope of linguistics to include pragmatics have certainly been steps in the right direction. It is unfortunate, however, that many of these suggestions involve simply adding various formal devices to a grammar of competence in the above sense (see Lakoff 1971; Ross 1970) and that they are restricted to one output modality—speech.
It is not my intention here to argue whether the definitions of language outlined above are appropriate for linguistics. They are certainly not inconsistent with the traditional descriptive approach of the discipline. I will argue, however, that such definitions are inappropriate for interdisciplinary fields such as psycholinguistics and neurolinguistics which study a behavior rather than an entity. In these interdisciplinary fields, all behavior bearing on communication is subject to study; study is not limited to behaviors such as judgments of grammaticality, synonymy, etc., that are needed to establish a corpus, or to speech and its derivatives. The communicative behavior of populations that would be considered distinct by linguists intersects, as suggested in figure i. Human communicative behavior is not an entity such as language which one either possesses in its entirety or not at all. Rather, it is a mosaic of structures, skills, and knowledge that does not develop in children as a whole, does not disappear in dysphasics as a whole, and undoubtedly did not evolve as a whole. Finally, such behavior cannot profitably be studied in isolation from the environment in which it takes place or from the peripheral and central structures of the organism that subserve it.
Fig. l. The continuum of communicative behaviors evinced by various populations of primates. The interesting areas of overlap are labeled. A: Various reflexive and affective aspects of communicative behavior. B: Similarities in developmental sequences of communicative behavior. C: Similarities between communication systems in development and in dissolution. D: Continuities between dysphasic and normal communicative behavior, e.g., slips of the tongue, the tip-of-the-tongue phenomenon, etc.1
What is being argued, then, is that the definition of language proposed by linguists is both too narrow and too monolithic to characterize adequately the types of communicative behavior addressed by fields such as psycholinguistics and neurolinguistics, which study performance (as opposed to competence). Furthermore, it is suggested that an adequate model for these interdisciplinary fields must be tied to what is known about the biological bases of behavior in general and about communicative behavior in particular. Before attempting to elaborate such a model, let us first consider how we might define communicative behavior in primates. (See Chomsky 1980 for an exposition of his latest views on the issues raised in this section together with extensive invited comment on these views.)
B. BIOLOGY AND
COMMUNICATIVE BEHAVIOR
THE DEFINITION OF COMMUNICATIVE BEHAVIOR
While, as we have seen, the linguists’ definition of language is too narrow for our purposes, definitions of communicative behavior formulated by those working within a biological framework often appear too broad. Many definitions simply require some correlation of behaviors between two organisms. Thus, for Altmann, it is “a process by which the behavior of an individual affects the behavior of others” (1967:326), a statement which is so general that the event of two individuals bumping into each other accidentally would constitute an instance of communication. Certainly, the factor of an external stimulus affecting behavior of an organism in some manner is a necessary but not sufficient condition for communication to take place. If one is interested solely in the transfer of information, one can concentrate on the receiver organism alone. For information transfer to occur, it is only necessary that signals fall within the acuity range of the receptor organs of the organism and that the signals are interpreted unambiguously by its nervous system. The nature of processing that occurs in the nervous system will depend on the type of nervous system and type of stimulus. In general, what one observes is a progression both phylogenetically and ontogenetically in primates from reflexive to affective to cognitive processing systems. While even so broad a term as “transfer of information” may seem inappropriate for reflexive behaviors such as jumping at a loud noise, surely this term does apply to behaviors involving a degree of cognition, such as the interpretation of a cloud formation to signify rain. But is this really what is meant by communication?
MacKay (1972), in a most perceptive analysis of communicative processes, suggests that the kind of behavior we have been discussing does indeed involve transfer of information, but that it should not be termed communication. Communication in his view involves an interaction between organisms such that the sender organism’s behavior is goal directed towards the receiver organism. The essentials of MacKay’s proposal are schematized in figure 2.
Fig. 2. The essentials of MacKay’s model of the sender organism. The receptor system R of the organism monitors the action of its effector system E in field F. Goal direction involves essentially the comparison of the current state of field F (IF) with the goal criterion IG in evaluator C, so that a mismatch may be communicated to the organizing system O.
In this figure, the receptor system R of the organism monitors the action of the effector system E in field F. Goal direction involves essentially the comparison of the current state of field F (IF) with the goal criterion IG in evaluator C, so that a mismatch may be transmitted to the organizing system O. Many organisms, including man, appear to possess neural feature detectors which would determine to some degree the information included in IF. Thus a bank of feature filters might be added to the model between the receptor system and IF. One might also complicate O by including in it the capacity to produce a variety of responses as well as a mechanism for determining priorities in the production of these responses. Once again, one can imagine a progression here from a relatively small repertoire of stereotyped and automatic responses in communication systems of the reflexive and affective type to an extremely large and potentially infinite repertoire of novel and propositional responses in communication systems of the cognitive type.
Note that goal direction as defined in this model resembles somewhat Austin’s concept of “speech act” when defined in a modality-free manner (see Bates 1976). The locutionary act is the joint result of the organizing and effector systems. The illocutionary act might be equated with IG, the goal criterion, while the perlocutionary act is analogous to the comparison of IF (the state of field F) with IG (the goal criterion in the evaluator).
There is one basic difference between my interpretation of Mac-Kay’s system presented above and his own. In order for his system to provide for the continuities in communicative behavior schematized in figure I, it is essential that goal-directed behavior not necessarily imply volitional control. In large measure, what I have termed reflexive and affective communication systems do not involve such control. If we do not make this distinction, then a vast proportion of what has been termed animal communication cannot be considered communication at all. Such a distinction allows us to differentiate between non-human primate communication in the wild, which is goal directed but largely nonvolitional, and the acquired communicative behaviors of great apes in the laboratory, which is both goal directed and volitional. But what about the person who blushes in embarrassment or the poker player who cannot keep a poker face? Are they communicating in terms of this model? There can be no doubt that there may be a transfer of information in these situations and that it is nonvolitional. That such behaviors are goal directed in MacKay’s sense appears doubtful. On the other hand, a dog who, after whining and nudging his master in order to gain attention, scratches at the door to be let out is certainly engaging in goal-directed behavior.
Goal direction on the part of the sender organism is not enough to ensure communication, as MacKay seems to imply. The receptor organism must, of course, be equipped to process such goal-directed behavior. This requires, as noted before, that the behavior can be perceived by the organism and that its central nervous system, operating in terms of a relatively closed or open genetic program (see Mayr 1974), is capable of interpreting the behavior of the sender organism unambiguously. Thus, if we should approach the cage of a rhesus monkey who in turn approaches us with teeth bared, we might interpret its expression anthropocentrically as a smile and even act on this interpretation by reaching out to pet it. Criteria for the communication process would have been met, save for the last stage, interpretation; the receiver must have knowledge of the communication system being employed, and in this case the receiver did not. It is not necessary, however, that both communicators be of the same species.
A GENERAL BIOLOGICAL MODEL OF COMMUNICATIVE BEHAVIOR
Our attempt to clarify the nature of communicative behavior in the previous section reveals that we have already begun to operate within a biological framework—one that involves monitoring the environment as well as considering peripheral structures (such as receptor and effector systems) and central processing structures (such as the evaluator and the organizing system). We are now ready to investigate a biological model of communication in more detail. We shall confine ourselves to primate communication, particularly that of the great apes, since it is reasonable to expect that if homologies in communicative behavior (i.e., behaviors that could in principle be traced back to a common ancestor) are to be discovered, they will be discovered among our closest relatives (see Hodos 1976 for a discussion of the concept of behavioral homology and Dingwall 1979 for a detailed application of this concept to the evolution of human communication). Unfortunately, much of the research on the evolution of human communicative behavior has been based on a classification of animals termed the scala naturae or Great Chain of Being, in which animals are arranged according to their imagined complexity, with man (or God) at the top. Comparisons of animals arranged in such a manner are meaningless if one is interested in the phylogeny of traits (see Hodos and Campbell 1969). It is not obvious to me what can be learned about the evolution of human communication from comparisons in terms of a set of features, such as those proposed by Hockett (1960), which are derived from the logical analysis of communicative behaviors and are applied helter-skelter throughout the animal kingdom, without regard to the structures that subserve the behavior and without regard to phylogenetic proximity to man. It is interesting that when such features are applied to primates alone, they fail to distinguish man from other members of the order (Altmann 1967).
Hierarchical Input/Output System
In biology, models of behaviors such as communication are typically presented in the form of hierarchical input/output systems. Such systems generally include specification of peripheral receptor and effector structures as well as some details of the routes taken by the tracts relaying neural impulses to and from a central processor that encompasses much of the central nervous system. A model of this type is depicted in figure 3 and, with further simplification, in figure 4, panel B.
In the latter figure, it is proposed that three basic transductions are involved in communicative behavior. The first (Ti) can be labeled cognition and involves the capacity of the brain of an organism to produce and to comprehend, if they impinge upon it, a variety of concepts which may be either simple or complex. Cognitive psychologists such as Piaget have to some extent clarified the nature of this transduction as it develops in humans. Ethologists and comparative psychologists have investigated its nature in other species.
Fig. 3. Schematic diagram of neural mechanisms for human communication.
Fig. 4. Aspects of a general biological model of the communicative behavior of primates. (T1 = transductionr, LTS = long-term store; STS = short-term store; SIS = sensory information store; NS = neurosign.)
One cannot at present judge whether an animal is in command of a particular concept unless the animal signals it in some manner. This involves linking the concept with a neurological state, the neurosign (NS), that mediates either its production or its recognition. T2, or what might be termed coding, involves the linkage between an abstract conceptual structure and an equally abstract sensorimotor representation, which is only imperfectly reflected in the various states of its production and recognition. When speech alone is considered, it is this transduction that theoretical linguistics regards as its major concern. (Thus, meaning is generally taken to involve an abstract conceptual structure, while speech sounds are taken to involve an equally abstract phonetic representation.) Recent work by Sachs (1967), Bransford and Franks (1972), Kintsch (1974), and others has provided the first clear evidence of the nature of conceptual structure, while at the same time neuroscientists have demonstrated in event-related potential studies that what is being termed the neurosign is more than a theoretical construct (see Dingwall and Whitaker 1974). On the other hand, a decade of work in experimental psycholinguistics has made it abundantly clear that current linguistic theory fails to provide a viable model of either the processes underlying T2 or its endpoints (cf. Fodor et al. 1974; Watt 1970, 1974; Dingwall and Shields 1973; Wanner 1977; Wanner et al. 1977).
The third transduction (T3), which we shall term transmission, involves the processes of signal production and detection. It is this transduction, because of its relative accessibility to study, about which most is known, thanks to the efforts of physiologists, physiological psychologists, hearing and speech scientists, experimental phoneticians, and others interested in this aspect of communication (see Lass 1976).
It should be noted that in a biological model of communication, effector and receptor segments involve quite different neurological substrates, as even the simplified account in figure 3 clearly shows. There is considerable psychological as well as neurological evidence that production and comprehension (of which these segments represent the late and early stages, respectively) involve quite different processes. This is true whether one examines these aspects of communication in normal adults, in children acquiring communicative abilities, or in persons whose communicative abilities are impaired in various ways (see Straight 1976 for a review of some of the evidence). Similarly, while it is clear that man is the only species that is capable of producing articulate speech, it is not at all clear that he is the only species that can process it (see Burdick and Miller 1975; Fouts et al. 1976; Kuhl 1978).
To what extent is the processing represented by the three transductions independent of each other? Let us first consider T3. We know that trained phoneticians can abstract from a language phonetic elements which they cannot further process. In clinical cases of pure word deafness and pure alexia, words are heard or seen, yet are not further processed, even though they represent elements of the native language of the patient. Cases exhibiting neologistic jargon or glossolalia may represent the random production of sound sequences without further encoding. There is evidence that both transmission and coding can take place without cognition. We have all had occasion to process materials we could not understand; Lewis Carroll’s “Jabberwocky” is an extreme example. There are cases of aphasia known as “isolation of speech area,” where the patient shows little evidence of comprehension or volitional production of speech, but does repeat verbal material that has been presented. Despite the fact that such material does not appear to be processed cognitively (i.e., via T1), it nevertheless appears to be processed by T2. Thus, if the patient is presented with a grammatically incorrect sentence such as (I):
(1) She buy a dress yesterday,
it is corrected to form (2):
(2) She bought a dress yesterday.
Presented with items in a foreign language, a patient will pronounce them, but with an accent characteristic of his or her own language (see Whitaker 1976 for a detailed discussion of such a patient). A similar type of automatic processing is observed in speech errors. Thus, a German speaker intending to say (3):
(3) Er hat in Berlin drei Tage in der Woche gearbeitet,
exchanged the italicized words to produce (4):
(4) Er hat in Berlin drei Wochen im Tag gearbeitet.
Note that all appropriate inflectional changes have been made, even though this was not the intended sentence (see Bierwisch 1970). There is an automatic character to coding and transmission which is also typical of other overlearned outputs, such as driving a car and dancing. This factor of automatization appears to be tied to a critical period in individual human development, at least in the case of speech output (Oyama 1976), and has a plausible neurological explanation (see Dingwall 1975).
Can the processes represented by T1 take place independently of T2 and T3? This question involves the essential problem of the relationship of language and thought. Most investigators agree that thought is possible without language (see Slobin 1975a for a recent discussion). If one narrows the definition of language to include only T2 and T3, then the evidence we have discussed above would indicate that language can occur without thought, whereas communicative behavior as defined in this paper clearly must involve T1.
It is conceivable that communicative behavior can occur without coding (T2). Halliday (1975:13) has noted that the earliest stages in language development do not involve T2, but rather involve a direct mapping of concepts onto expression. Thus, each complex concept is tied to one unanalyzed unit of expression. This system lacks duality of patterning in Hockett’s sense (1960). Similar by-passing of T2 may characterize much of the reflexive and affective vocalizations of primates.2
Finally, it should be pointed out that two elements which are essential for communication have not been specified in either figure 3 or figure 4B: a series of feature filter systems to provide feedforward (defined by MacKay 1972 as “information in advance of any external action”), as well as a feedback mechanism providing visual, tactile-kinesthetic, and auditory information of ongoing activity. (See Borden 1979 for a detailed review of evidence relating to the role of feedback in human communication.)
Memory Components
Associated with each transduction are various kinds of knowledge stored in either long- or short-term memory which are necessary for the processing that goes on at the levels of these transductions (see fig. 4C).
On the input side of the hierarchical model in figure 4, there is evidence that after initial auditory or visual analysis, which undoubtedly involves feature detectors of some sort, information is stored briefly in a relatively gross form in a sensory information store (SIS). This preliminary processing may be sufficient to determine whether a signal engages the dominant hemisphere of the brain in further processing. There is considerable empirical evidence that what we term coding is carried on in this hemisphere. If the coding mechanism is engaged, then the signal is further processed, utilizing knowledge of the communication system held in long-term storage (LTS). The recognition process involves periods of short-term storage (STS) at various levels of analysis. As processing continues, perhaps in parallel with the above stages, employing not only shared knowledge of the communicators but also general knowledge of the world, the form of the signal is stored in short-term memory only long enough to abstract the gist of the message (see Pisoni and Sawusch 1975).
Thus, very shortly after hearing a sentence, one’s memory for lexical items or syntax is near chance levels, but memory for semantic content is extremely good (Sachs 1967). If materials containing identical semantic content are presented in a simple versus complex linguistic form, then the complex materials will take longer to process; however, questions about content are answered equally rapidly once the material is processed (Kintsch and Monk 1972).
Bransford et al. (1972) have clearly shown that sentence processing is constructive (i.e., involves inference based on knowledge of the world). Thus, when sentences are inferentially related, as in the case of (5) and (6) below, subjects in recognition tasks are invariably confused about which one they were presented during the acquisition stage.
(5) The racoons raced up the tree and the dogs circled around them.
(6) The racoons raced up the tree and the dogs circled around it.
On the output side, knowledge of the world also plays a role in almost any conversational exchange. If, for example, I mention to a detective that I was held by my abductors in a car, he might ask me if I were put into the trunk. He asks this because his general knowledge of cars and abductions makes it a reasonable question. I might continue by telling him that I was driven to the mall where I was finally released. This statement assumes shared knowledge of the identity and perhaps location of the mall in question. What Grice (1968) has termed conversational postulates clearly involve both kinds of knowledge associated with Ti.
Note that on both the input and the output sides of the system model knowledge of the communication system is directly involved in performance, rather than abstracted away from it, as in Chomsky’s model involving the competence/performance distinction. I agree with other investigators who question the validity of this distinction for interdisciplinary fields such as psycholinguistics and neurolinguistics (see Watt 1974; Whitaker 1970). If such a distinction is to be employed at all, it must be redefined so that competence applies to cognition (T1), which appears not to be lateralized in the brain, while performance relates to coding (T2) and transmission (T3), which are.
Finally, in the last stage of output, short-term memory once again plays a role, providing a buffer for strings, perhaps of syntagma length, being readied for transmission (Laver 1970). It is at this stage that errors in ordering can occur, which happens in slips of the tongue and in dysphasia. Whether storage at later levels of the output process exists is not yet known.
Fig. 5. Elaborated models of the transductions involved in coding and transmission. These particular models, which are based on Pisoni and Sawusch (1975) on the input side and on Fromkin (1971) (for T2) and on Cooper (1972) (for T1) on the output side, are confined to the verbal input/output modality of humans.
It is clear that at each stage in processing a different code is employed in memory. The code of the sensory information store involved in transmission differs from that of the short-term store involved in coding. The form of storage differs in turn from the conceptual memory for semantic content, which we might term mentalese following Fodor (1975).3
Elaborated Models of Specific Stages of Communication
Investigators from many fields have provided us with elaborated models, based on a variety of empirical data, of the processes and structures involved in the transductions we have been discussing (see fig. 4A for a listing of models). While psychologists have generally concentrated on the representation of conceptual structures in memory (see Kintsch 1974; Frederiksen 1975), scholars working in artificial intelligence (AI) and in linguistics have been more interested in the coding processes that link the neurosign to such structures (see Chomsky 1965; Winograd 1972). Researchers in the area of experimental phonetics have, as one might expect, concentrated on aspects of the transmission stage (T3) of communication. Figure 5 presents representative models of coding and transmission on both the input and output sides. Unfortunately, such models have been almost exclusively confined to speech processing.
We have already discussed in some detail the types of processes represented on the input side in figure 5. The model schematized there is based on that of Pisoni and Sawusch (1975). Note the parallel nature of the processing involved, as indicated by the double-headed arrows interconnecting various components. Exactly how such processing proceeds is far from understood. There is little evidence, for example, to support either the motor theory of speech perception or the related analysis-by-synthesis models of such processing (cf. Lieberman 1977; Fodor et al. 1974).
The output side presents a model of coding based on Fromkin’s study of speech errors (1971) and on a model of transmission based on the work of the Haskins Laboratories group (Cooper 1972). While both models are based on empirical evidence, it is doubtful that either Fromkin or Cooper would claim that their proposals represent more than a gross approximation of the highly complex psychological and neurological processes that must be involved.
The most elaborate general model of the neurological bases of communicative behavior is that developed by Lamendella (1976, 1977). Its main features are summarized in figure 6. This model can be applied to the development of communicative behavior in the child, given the general rule of caudal to rostral maturation of the brain. The application is illustrated in table I, where I have arranged communicative functions with behavioral examples under the rubrics of the reflexive, the affective, and the cognitive processing systems referred to earlier. In light of Lamendella’s views on recapitulation (1976), the developmental sequence outlined in table I gives insight into the phylogeny of communicative behavior in the primate order.
Fig. 6. Overview of human communication. (I) Lateralized systems of right hemisphere; (2) primary neocortical sensory systems; (3) primary neocortical motor systems; (4) volitional subsystem; (5) anterior affective subsystem; (6) posterior affective subsystem; (7) appetitive subsystem; (8) lateralized system of left hemisphere.
Table 1. The development of communicative behavior in children based on Lamendella’s neurological model. These behaviors develop in the sequence from left to right and from top to bottom and are grouped under the neurological substrate that is held to be principally responsible for them.
In general, I find this grand scheme quite plausible and am in complete agreement with Lamendella that there is no reason to think that the brain divides behaviors into the neat packages that our taxonomies indicate. Although the model may be generally plausible, it should be noted that its representation of the localization of functions is based on scant empirical evidence. An example which may prove very important on this point is American Sign Language (ASL). It is by no means clear that ASL is lateralized to the dominant hemisphere in the deaf (cf. McKeever et al. 1976; Manning et al. 1977), as Lamendella’s diagram indicates. This may be the reason that ASL can be processed by the great apes, who, as far as we now know, do not evince lateralization of communicative functions. Furthermore, the limbic system may not play as great a role in human and nonhuman communicative behavior as Lamendella and others have imagined (see Jürgens and Ploog [1976] and Dingwall [forthcoming] for further discussion).
Lamendella’s view of human communication, however, differs significantly from my own in several ways. First, he embraces the linguists’ définition of language, stressing that it is a communication system of a definite type, “characterized by quite specific syntactic, morphological, and phonological organization” (Lamendella 1976:401). Such a communication system is qualitatively different from that of nonhuman primates, and this narrow view of language leads to many of the problems discussed above, particularly the arbitrary exclusion of many humans who communicate from the class of language-users. Schnitzer, in a review of evidence from various types of aphasia that clearly demonstrated that phonological processing could be bypassed in human communicative behavior, came to the conclusion that “language without phonology is language” (1976:159). I concur with Schnitzer’s conclusion, but not with his nomenclature. I would prefer to reformulate his finding by stating that human communicative behavior cannot be defined in terms of a particular input/output modality.
Second, I disagree with Lamendella’s view that the stages of communicative development listed in table I do not represent a gradual accretion of the elements that make up a human communication system. I have stated earlier that human communication is a rich mosaic of structures, skills, and knowledge, and that it develops gradually in the child. If the child develops normally, then a particular input/output modality using the auditory medium will gain primacy over other modes, although the other communicative skills displayed in figure 6 will be present in the background. If the child does not develop normally or if speech becomes no longer available in adulthood, then other communicative skills, typically involving visual or tactile sensation, may be used predominantly. In fact, this multimodal backup system is one characteristic I would use to differentiate primate communication systems from those of other animals.
Now let us consider behavioral models of communication. Such models generally consist of taxonomies of observable communicative behaviors relating to the inputs and outputs of the organism and to the contexts in which such behaviors take place. As Savage and Rum-baugh (1977) have pointed out, such taxonomies may concentrate on (I) the physical aspects of the communicative act (the input/output modalities, the media used, etc.), or (2) the functions served by such acts (identification, contact, expression, etc.), or (3) abstract properties (design features) of such acts (openness, duality of patterning, etc.). Sebeok (1968) provides detailed examples of these approaches. Many of the best known taxonomies, such as those of Hockett (i960) and Hymes (1962), combine all three characteristics listed above. Sets of design features selected to characterize properties of animal communication in a general sense can provide insight into similarities and differences between communication systems, but such comparisons provide little insight into phylogeny. Thus, if one compares the signing behavior of pongids and the songs of birds with human communication following Hockett, then one finds seven and nine differences respectively out of a possible sixteen (see Dingwall 1979). The similarities found between birdsong and human communication are due to the bias in Hockett’s system towards a particular input/output modality. At the end of this paper, I shall propose a set of features based on characteristics of the biological model. These features, in contrast to those presented by Hockett, can be used to measure the complexity of primate communication systems, and, at the same time, they are phylogenetically relevant.
At least three aspects of the observable communicative behavior of primates are of importance in the development of a general biological model. First, we need to survey the totality of communicative behaviors in terms of the input/output modalities that man and other primates are capable of using. This will give us some insight into the nature and the complexity of primate communication systems. Second, we need to discuss the role of context in communication. Biology has always been concerned with the interaction of the organism and its environment. Although we have recognized for quite some time the sensitivity of the communicative behavior of nonhuman primates to various aspects of the environment, we are only beginning to appreciate the existence of that quality in human communication. Third, we need to investigate the aspects of the biological model that are engaged by varied tasks involving the same communicative behavior. We shall see that communicative behaviors to which we assign a single designation, such as writing, are indeed not monolithic capacities but rather aggregates of skills.
Table 2 provides an overview of the diversity of factors involved in primate communication systems. Communication by primates typically makes use of a variety of media, which are usually employed simultaneously. We may isolate articulate speech as the central element in face-to-face communication in normal humans, but we must realize that it is nevertheless embedded in a complex of acoustic, visual, tactile, and even olfactory signals, which may on occasion contradict the message the words transmit. Similarly, if we look at signing behavior in deaf persons, we find that even though hand gestures are the predominant means of communication, various aspects of facial expression and body movement interact with those gestures to help convey information. It is interesting to note that studies of nonhuman primate communication in the wild are unclear whether the acoustic or the visual medium is more important (see Menzel 1971).
Table 2. Primate communication systems: input/output modalities, media, and signal types. (Asterisks mark signal types presumed unique to human communication.)
In humans, and perhaps in some degree in nonhuman primates, almost any imaginable type and combination of input/output modalities can be recruited for purposes of communication. One can learn to write with a pencil grasped between one’s teeth or to gesture with one’s feet. One can learn to “read” using a series of electrodes attached to one’s stomach or through letters traced on the palm of one’s hand. The recent success in teaching great apes to communicate using the visual medium has sparked attempts to use similar modes of communication with humans who lack the ability to speak. As one might expect, these efforts have met with considerable success.
Communicative behavior is highly complex not only because of its varied and simultaneously occurring signal types, but also because it occurs in multifaceted environments. While it may be true that the ability to function in isolation in a physical context is a unique characteristic of human communication,4 it can hardly be termed the norm. Where physical context is lacking or is not part of the shared knowledge of the communicators, it is usually supplied by descriptions, illustrations, or (on radio, for example) sound effects. Communication can be degraded by lack of knowledge of the communication system (as is often the case when dealing with foreigners, and is possibly the case when dealing with children) or by pathology. Contextual cues then become even more important for proper interpretation. Bloom (1970) was the first to point out the need for “rich interpretation” of early childhood utterances, taking into account aspects of the physical context. In cases of aphasia, the physical context becomes extremely important for cuing the meaning of communicative acts.
As sociolinguists have documented in detail, the social context is highly relevant to communicative behavior. This is as true of nonhuman primate societies as it is of human societies. For example, while the use of various lexical and grammatical devices in a human communication system such as Japanese to signal differences in sex, age, and status may be more subtle, it parallels nonverbal behavior among primates that signals the same types of sociological distinctions (see Lancaster 1975).
Finally, the linguistic context of communication must be considered. If we regard normal verbal behavior simply as a set of sentences unrelated to one another, we neglect to explain many phenomena such as anaphora, stress placement, article choice, and use of connectives. (See Isenberg 1968 for an early formal treatment of discourse phenomena.) Both monologue and dialogue must take account of linguistic context in order to function adequately. The linguistic context clearly has effects on performance. Thus, Wright (1972) has demonstrated that subjects make fewer errors when responding to questions if the questions are cast in the same grammatical form as the linguistic material upon which they are based. Thus, given a sentence such as (7):
(7) The crowd was held back by the police.
a question in the passive, such as (8), will probably be answered correctly, as well as more rapidly, than one in the active such as (9):
(8) Was the crowd held back by the police?
(9) Did the police hold back the crowd?
There may well be parallels to the role played by linguistic context in human communication systems in the communicative behavior of nonhuman primates. For example, Savage and Rumbaugh (1977) have recorded a definite ordering in the nonverbal exchanges that take place between chimpanzees. While physical context naturally plays a role, it is difficult to determine from the available evidence if either social or linguistic contexts are involved in the communicative behavior of great apes who have mastered humanlike communication systems.
The study of presupposition in pragmatics deals with the conditions under which a sentence is appropriate to the context in which it is used. There is no reason why this concept of presupposition cannot be broadened to cover the appropriateness of communicative behavior to the contexts in which it takes place.
Communicative Behavior as an Aggregate of Skills
Finally, one must not forget that the complex behavioral phenomenon we have been describing is the product of an equally complex organism that is ultimately controlled by the brain. There is no a priori requirement for the ultimate control mechanism to compartmentalize behavior in a manner isomorphic with human systems of nomenclature. It should therefore be obvious that behaviors instinctively felt to be monolithic because they have a single designation in some language are actually a complex aggregate of skills.
This is perhaps most graphically demonstrated in various aphasie syndromes. The array of deficits displayed by patients suffering from one of the so-called pure aphasias termed “alexia without agraphia” is a typical example (see Benson and Geschwind 1969). In such patients, verbal comprehension and production are typically uninvolved. However, these patients cannot read letters, words, or musical notation, although they may be able to read numbers (even roman numerals) as well as familiar logos. They can write, either voluntarily or to dictation, but cannot read what they have written. Objects, but not colors, can be recognized and appropriately named. Patients can verbally spell words, can understand spelled words, and can copy letters, but they cannot transcribe print into script.
Table 3. An aggregate of skills involved in communicative behavior. (“Haptic-sounded” refers to sounds made with the hands, such as finger snapping; “haptic-artifactual” refers to the use of mechanical means to make sounds with the hands, e.g., drumming.)
The example describes an aphasia that is a breakdown in communicative behavior due to brain injury, and it reveals an aggregate of skills that is characteristic of the entire continuum of communicative behavior presented in figure 1. Some of these skills are listed in table 3. As already noted, input and output, comprehension and production, are different processes neurologically, and it is clear that they are differentially involved in this type of aphasia. At a finer level of resolution, it is clear that particular modalities and types of input and output are implicated. The patient cannot read, but inputs using other modalities or types can be processed by the system. Various transformations of inputs are affected in this syndrome. While letters and pictures can be copied slavishly, it appears that they are not being decoded, because they cannot be transcribed. Just as the phonetician imitates sounds or words of an unfamiliar verbal communication system, this patient can copy letters; transduction (T3) alone is operating. There is evidence that some graphic symbols can bypass various levels of processing and be comprehended directly, that numerals and logos appear to be processed as gestalts (see the discussion of parallel phenomena in Schnitzer 1976). Even more detailed aspects of particular transductions may be involved. Thus, particular semantic categories of items such as colors cannot be named. Various aspects of the memory components listed in figure 4C are involved in aphasias such as anomia, as well as in the tip-of-the-tongue phenomenon characteristic of normal speech. Finally, there is clearly a continuum of communicative behavior ranging from the reflexive to the affective to the cognitive (propositional) that may be quite differentially affected in aphasia. Note, however, that this same continuum is evinced in the development of communicative behavior in children (see table 1). Some of the types of speech behaviors that make up this continuum, following research by Van Lancker (1975), are listed in figure 7.
C. ASPECTS OF CHANGE IN
COMMUNICATIVE BEHAVIOR
Behavior implies flux. Whereas language may be regarded as a static entity, a corpus of sentences, to be analyzed at a particular point in time, communicative behavior must be regarded as ever changing. The aspects of this behavior which we have focused on in this paper are dynamic, whether they involve evolution, development, or dissolution (see fig. 1).
Slobin (1975b) has discussed with numerous examples four competing forces which he believes contribute to change in human communication systems. He expresses these as imperatives: (1) to be clear, that is, to express what we have termed simple and complex concepts with as little distortion as possible, (2) to be humanly processible in ongoing time, (3) to be quick and easy, that is, to relay as much information as possible within a given time span, and (4) to be expressive, both in the sense of relaying all that one has in mind (being se-mantic) and in the sense of relaying what one wishes in an effective manner (being rhetorical). It is clear that these forces are basically incompatible with stasis. The first two may be regarded as constraints on change, while the second two are stimuli for change.
Fig. 7. A hypothetical continuum of propositional and automatic speech modes and their properties. (Van Lancker 1975:173)
What is the goal towards which such unstabilizing forces strive? If one were to give it a name, it might be efficiency. An efficient communication system may be defined as one that is capable of transmitting the greatest amount of information in the least amount of time with the least ambiguity and the greatest intelligibility. The complex of changing environmental pressures that would call for increased efficiency are easy to imagine. They involve a number of factors in a complex feedback relationship, such as more efficient toolmaking, more control over the environment, more learning, and more efficient social organization (see Dingwall 1979).
How does the pressure for more efficient communication affect the biological model we have been discussing? At least three basic changes can be imagined: (1) changes in the character and the amount of storage required, (2) changes in the character and the amount of processing required in each transduction, and (3) changes in the character of the predominant input/output modality.
As the amount of knowledge required for survival increases, there is obviously pressure for the memory components of cognition to expand. Table 4 presents the changes involved in increasing the efficiency of a communication system within the primate order. As the shifts in communication shown here take place, both the amounts of storage and of processing required for coding (T2) and transmission (T3) increase. Whereas the child, as noted earlier, initially links concepts directly to output, he or she later develops the type of complex intermediary processing suggested in figure 5. Finally, the character of the predominant input/output modality may change from that which is based on some combination of body movement plus affective vocalization to that which involves articulate speech, possibly with an intermediate stage where hand gestures rather than speech are predominant. There is good evidence to believe that speech is one of the most efficient means of relaying information between organisms that has evolved (see Lieberman 1975). Note that the changes outlined above apply to ontogeny and quite plausibly to phylogeny and are accompanied in both instances by increase in brain size.
D. DETERMINANTS OF COMPLEXITY
IN PRIMATE COMMUNICATION SYSTEMS
It makes little sense to ask whether any of the populations depicted in figure 1 (to which one might add computers, for good measure) are in possession of some uniform capacity called language or human communication. It should be evident by now that the determination whether one communicates in a human manner cannot be made in absolute terms; rather, some gradient measure must be employed. In conclusion, I should like to suggest the following set of six gradient features based on our biological model—each of which could be associated with a scale from 1 to 7 as a measure of comparative degree of applicability. I would also like to submit that, unlike Hockett’s design features, these appear capable of producing a taxonomy within the primate order that is in accord with phylogenetic proximity.
Table 4. Some of the changes involved in moving towards a more efficient communication system within the primate order. All the transitions indicated are gradient rather than absolute. (*It is possible that there was an intermediate stage where gesture rather than speech was foregrounded.)
1. Length of Immaturity. It is well known that there is a trend toward prolonged growth and maturation periods in primates. The longer the period of immaturity, the greater the time available for learning.
2. Multimodality. The extent to which the communication system makes use of a complex of media simultaneously and to which alternative input/output modalities can sustain communication.
3. Degree of Volitional Control. The extent to which the communication system is of the cognitive type as opposed to the reflexive or the affective.
4. Efficiency. The extent to which the communication system is capable of transmitting a great amount of information in a small amount of time with little ambiguity and great intelligibility.
5. Degree of Coding. A measure of the complexity of the coding processes on both the input and the output sides.
6. Degree of Context-Freeness. The extent to which communication can take place in isolation from contextual cues.
Suppose we could quantify these features and then apply them to random samples of the populations we have been discussing throughout this paper, namely:
A. Nonhuman primates in the wild.
B. Nonhuman primates who have acquired humanlike communication systems.
C. Normal children acquiring language.
D. Humans with verbal disorders.
E. Normal human adults.
What would we find? My speculation is that nonhuman primates in the wild and normal adult humans would be significantly different on these measures. But what of the remaining populations? While length of immaturity would place children and the verbally impaired in a group with normal adults, the remaining features would most probably differentiate these two populations from the latter. Of the groupings that might occur, the following three appear most plausible on present evidence:
Other possibilities are not ruled out entirely, however.
SUMMARY
We have seen that neither linguistics nor biology offers at present a viable framework for the study of human communicative behavior. In this chapter, I have attempted to develop a biological model of such behavior which combines the positive aspects of both linguistics and biology while avoiding their shortcomings. In elaborating this model, I have made use of insights, both theoretical and empirical, from such allied fields as neurolinguistics, psycholinguistics, artificial intelligence, experimental phonetics, and ethology.
Human communicative behavior is viewed as a hierarchical input/output system which is mediated by the central nervous system and peripheral structures of the human organism, and is carried out within a complex context involving physical, social and linguistic aspects. Language, the subject matter of linguistics, forms but one element of this complex behavior. This element, which I have termed coding, involves the transduction between an abstract conceptual structure and an equally abstract sensorimotor representation. Coding may be mediated by various input/output modalities, not just the ear and vocal tract, as is often assumed by linguists.
Neither language nor human communicative behavior, of which it is a part, is a unitary phenomenon. Rather, as has been demonstrated, both involve a mosaic of structures, skills, and knowledge. Numerous examples have been provided, showing: that comprehension and production involve different processes; that the three transductions we have discussed (cognition, coding, and transmission) can occur independently of one another; and further, that each transduction involves an aggregate of dissociable skills. These examples characterize not only the dissolution of human communicative behavior in instances of pathology, but its normal development in children as well as its normal usage by adults. That dissociations of transductions can be demonstrated provides some support for the view that language (coding) is task-specific (see Chomsky 1972, 1980).
On the other hand, it cannot be assumed a priori that human communicative behavior viewed in its totality constitutes a species-specific behavior without parallel in the animal kingdom. It seems likely that a behavior as complex as human communication did not evolve by some sudden genetic saltation, but rather in a mosaic fashion. Using the concept of behavioral homology developed by Hodos (1976) and elaborated in Dingwall (1979 and forthcoming), we may even be able to trace to some extent the probable evolutionary course of this behavior. Some aspects of this question are addressed in the final two sections of the chapter.
It should be clear that after hundreds of years of writing grammars and studying their properties, we have only begun to investigate in an empirical manner the broader question: what is the nature of the communicative behavior of our species, and how did it evolve?
__________
Tables and figures are reprinted with the permission of the publisher from Die neueren Sprachen 77:3/4 (1978) :271-99. Verlag Moritz Diesterweg, Frankfurt am Main, West Germany.
NOTES
This chapter represents a modified and somewhat expanded version of an article with the same title, which appeared in a special issue of the journal Die neueren Sprachen (77:3/4, 1978), devoted to language and behavior.
1. The extent to which great apes have been able to acquire languagelike behaviors is at present a hotly debated question. For an evenhanded discussion, from an evolutionary point of view, of the issues involved in this debate see Desmond (1979).
2. If information can be conveyed without coding, then what pressures could conceivably result in the emergence of such a transduction and its elaboration? As depicted in figure 8, Liberman and Studdert-Kennedy (1977) have recently suggested that the elaboration of this transduction allows for the communication of a large number of concepts employing relatively short strings of lexical items that are constructed from a limited inventory of elementary signals. Thus, as the child grows in cognitive capacity and, perhaps, as early hominids came under selection pressures for such growth, the system of communication had to change in order to remain efficient, that is, capable of transmitting the greatest amount of information in the least amount of time with the least ambiguity and the greatest intelligibility. Other aspects of the communication system may also be affected by this drive for increasingly efficient transfer of information. Some of these are outlined from a phylogenetic perspective in table 4 of this paper.
Fig. 8. A rationale for complex coding in human vocal communication (derived from Liberman and Studdert-Kennedy 1977).
3. There appear to be memory stores specific to language, since neither anterograde or retrograde amnesias affect language per se.
4. It should be noted, however, that chimpanzees who have mastered sign language have been observed to retrieve reliably objects which are not physically present. This is true even if the object in question is one among many different objects stored in a chest at some distant location. Thus, it is possible that the characteristic mentioned above is not unique to human communication.
REFERENCES
Altmann, S. A. 1967. The structure of primate social communication. In Social communication among primates, S. A. Altmann, ed., 325-62. Chicago: University of Chicago Press.
Bates, E. 1976. Language and context. New York: Academic Press.
Benson, D. F.. and Geschwind, N. 1969. The alexias. In Handbook of clinical neurology, vol. 4, P. Vinken and G. Bruyn, eds., 112-40. Amsterdam: North-Holland.
Bierwisch, M. 1970. Fehler-Linguistik. Linguistic Inquiry 1:397-414.
Bloom, L. 1970. Language development. Cambridge: MIT Press.
Borden, G. J. 1979. An interpretation of research on feedback interruption in speech. Brain and Language 7:307-19.
Botha, R. P. 1968. The function of the lexicon in transformational generative grammar. The Hague: Mouton.
Bransford, J.; Barclay, J.; and Franks, J. 1972. Sentence memory: A constructive versus interpretive approach. Cognitive Psychology 3:193-209.
Bransford, J., and Franks, J. 1972. The abstraction of linguistic ideas: A review. Cognition 1:211-49.
Burdick, C., and Miller, J. 1975. Speech perception by the chinchilla: Discrimination of sustained /a/ and /i/. Journal of the Acoustical Society of America 58:415-27.
Cairns, H., and Cairns, C. 1976. Psycholinguistics: A cognitive view of language. New York: Holt, Rinehart and Winston.
Chomsky, N. 1957. Syntactic structures. The Hague: Mouton.
——. 1965. Aspects of the theory of syntax. Cambridge: MIT Press.
——. 1972. Language and mind. New York: Harcourt Brace Jovanovich.
——. 1980. Rules and representations. New York: Columbia University Press, 1980.
Cooper, F. S. 1972. How is language conveyed by speech? In Language by ear and by eye, J. Kavanagh and I. Mattingly, eds., 25-45. Cambridge: MIT Press.
Derwing, B. 1973. Transformational grammar as a theory of language acquisition. New York: Cambridge University Press.
Desmond, A. J. 1979. The ape’s reflexion. New York: The Dial Press/James Wade.
Dingwall, W. O. 1975. The species-specificity of speech. In Develop?nental psycholinguistics: Theory and applications, D. Dato, ed., 17-62. Washington, D.C.: Georgetown University Press.
——. 1978. Towards a reconstruction of the methodology of experimental
linguistics. In A survey of linguistic science, W. O. Dingwall, ed., 135-57. Stamford, Conn.: Greylock Publishers.
——. 1979. The evolution of human communication systems. In Studies
in neuro linguistics, vol. 4, H. Whitaker and H. A. Whitaker, eds., 1-95. New York: Academic Press.
——. Forthcoming. The evolution of human communicative behavior. New York: Academic Press.
Dingwall, W. O., and Shields, J. 1973. From utterance to gist. Four experimental studies of what’s in between. Unpublished manuscript.
Dingwall, W. O., and Whitaker, H. A. 1974. Neurolinguistics. Annual Review of Anthropology 3:323-56.
Fodor, J. 1975. The language of thought. New York: T. Crowell.
Fodor, J.; Bever, T.; and Garrett, M. 1974. The psychology of language. New York: McGraw-Hill.
Fouts, R.; Chown, W.; and Goodin, L. 1976. Transfer of signed responses in American Sign Language from vocal English to physical object stimuli by a chimpanzee (Pan). Learning and Motivation 7:458-75.
Frederiksen, C. 1975. Representing logical and semantic structure of knowledge acquired from discourse. Cognitive Psychology 7:371-458.
Fromkin, V. 1971. The non-anomalous nature of anomalous utterances. Language 47:27-52.
Glucksberg, S., and Danks, J. 1975. Experimental psycholinguistics. New York: John Wiley and Sons.
Grice, H. P. 1968. The logic of conversation. Unpublished manuscript.
Halliday, M. A. K. 1975. Learning how to mean: Explorations in the development of language. London: Edward Arnold.
Hockett, C. F. i960. The origin of speech. Scientific American 203(9):89-96.
Hodos, W. 1976. The concept of homology and the evolution of behavior. In Evolution, brain and behavior: Persistent proble?ns, R. B. Masterton, W. Hodos, and H. Jerison, eds., 153-67. New York: John Wiley and Sons.
Hodos, W., and Campbell, C. B. G. 1969. Scala naturae: Why there is no theory in comparative psychology. Psychological Review 76:337-50.
Hymes, D. 1962. The ethnography of speaking. In Anthropology and human behavior, T. Gladwin and W. Sturtevant, eds., 13-53. Washington, D.C.: Anthropological Society of Washington.
Isenberg, H. 1968. Überlegungen zur Texttheorie. ERIC/PEGS Paper #57. Washington, D.C.: Clearinghouse for Linguistics, Center for Applied Linguistics.
Jürgens, U., aud Ploog, D., 1976. Zur Evolution der Stimme. Archiv für Psychiatrie und Nervenkrankheiten 222:117-37.
Kintsch, W. 1974. The representation of meaning in me?nory. New York: John Wiley and Sons.
Kintsch, W., and Monk, D. 1972. Storage of complex information in memory: Some implications of the speed with which inferences can be made. Journal of Experimental Psychology 94:25-32.
Kuhl, P. 1978. Predispositions for the perception of speech-sound categories: A species-specific phenomenon? In Communicative and cognitive abilities—Early behavioral assessment, F. Minifie and L. Lloyd, eds., 229-55. Baltimore: University Park Press.
Lakoff, G. 1971. On generative semantics. In Semantics, D. Steinberg and L. Jakobovits, eds., 232-96. New York: Cambridge University Press.
Lamendella, J. 1976. Relations between the ontogeny and phylogeny of language: A neorecapitulist view. Annals of the New York Academy of Sciences 280:396-412.
——. 1977. The limbic system in human communication. In Studies in neurolinguistics, vol. 3, H. Whitaker and H. A. Whitaker, eds., 157-222. New York: Academic Press.
Lancaster, J. 1975. Primate behavior and the emergence of human culture. New York: Holt, Rinehart and Winston.
Lass, N., ed. 1976. Contemporary issues in experimental phonetics. New York: Academic Press.
Laver, J. 1970. The production of speech. In New horizons in linguistics, J. Lyons, ed., 53-75. Baltimore: Penguin.
Liberman, A. M., and Studdert-Kennedy, M. 1977. Phonetic perception. Haskins Laboratories Status Report on Speech Research 50:21-60.
Lieberman, P. 1975. On the origins of language: An introduction to the evolution of human speech. New York: Macmillan.
——. 1977. Speech physiology and acoustic phonetics. New York: Macmillan.
Lyons, J. 1968. Introduction to theoretical linguistics. New York: Cambridge University Press.
MacKay, D. M. 1972. Formal analysis of communicative processes. In Nonverbal communication, R. A. Hinde, ed., 3-25. New York: Cambridge University Press.
Manning, A.; Goble, W.; Markman, R.; and LaBreche, T. 1977. Lateral cerebral differences in the deaf in response to linguistic and nonlinguistic stimuli. Brain and Language 4:309-21.
Mayr, E. 1974. Behavior programs and evolutionary strategies. American Scientist 62:650-59.
McKeever, W.; Hoeman, H.; Florian, V.; and Vandeventer, A. 1976. Evidence of minimal cerebral asymmetries for the processing of English words and American Sign Language in the congenitally deaf. Neuropsych ologia 14:413-23.
Menzel, E. W. 1971. Communication about the environment in a group of young chimpanzees. Folia Primatologia 15:220-32.
Oyama, S. 1976. A sensitive period for the acquisition of a nonnative phonological system. Journal of Psycholinguistic Research 5:261-83.
Pisoni, D., and Sawusch, J. 1975. Some stages of processing in speech perception. In Structure and process in speech perception, A. Cohen and S. Nooteboom, eds., 16-35. New York: Springer-Verlag.
Robins, R. H. 1964. General linguistics. London: Longmans, Green.
Ross, J. 1970. On declarative sentences. In Readings in English transformational grammar, R. Jakobs and P. Rosenbaum, eds., 222-72. Waltham, Mass.: Ginn.
Sachs, J. 1967. Recognition memory for syntactic and semantic aspects of connected discourse. Perception and Psychophysics 2:437-42.
Savage, S., and Rumbaugh, D. 1977. Language and communication: A perspective. In Language learning by a chimpanzee: The Lana project, D. Rumbaugh, ed., 287-309. New York: Academic Press.
Schnitzer, M. 1976. The role of phonology in linguistic communication: Some neurolinguistic considerations. In Studies in neuro linguistics, vol. 1, H. Whitaker and H. A. Whitaker, eds., 139-60. New York: Academic Press.
Sebeok, T. A., ed., 1968. Animal communication. Bloomington, Ind.: Indiana University Press.
Sebeok, T. A. 1977. Zoosemiotic components of human communication. In How animals communicate, T. A. Sebeok, ed., 1055-77. Bloomington, Ind.: Indiana University Press.
Slobin, D. 1975a. Language and thought. (Audiocassette). Berkeley: University of California, Extension Media Center.
——. 1975b. Language change in childhood and in history. Working paper #41. Berkeley: University of California, Language Behavior Research Laboratory.
Straight, H. S. 1976. Comprehension versus production in linguistic theory. Foundations of Language 14:525-40.
Thorpe, W. H. 1972. The comparison of vocal communication in animals and man. In Non-verbal communication, R. A. Hinde, ed., 27-47. New York: Cambridge University Press.
Van Lancker, D. 1975. Heterogeneity in language and speech: Neurolinguistic studies. Working Papers in Phonetics #29. Los Angeles: University of California.
Wanner, E. 1977. Review of The psychology of language by J. Fodor, et al. Journal of Psycholinguistic Research 6:261-70.
Wanner, E.; Teyler, T.; and Thompson, R. F. 1977. The psychobiology of speech and language—An overview. In Language and hemispheric specialization in man: Cerebral ERPs. Progress in clinical neurophysiology, vol. 3, J. E. Desmedt, ed., 1-27. Basel: S. Karger.
Watt, W. C. 1970. On two competing hypotheses concerning psycholinguistics. In Cognition and the development of language, J. R. Hayes, ed., 137—220. New York: John Wiley and Sons.
——. 1974. Mentalism in linguistics, II. Glos sa 8:3-40.
Whitaker, H. 1976. A case of the isolation of the language function. In Studies in neurolinguistics, vol. 2, H. Whitaker and H. A. Whitaker, eds., 1-58. New York: Academic Press.
Whitaker, H. A. 1970. A model for neurolinguistics. Occasional paper #10. Colchester: University of Essex.
Winograd, T. 1972. Understanding natural language. Cognitive Psychology 3:1-191.
Wright, P. 1972. Some observations on how people answer questions about sentences. Journal of Verbal Learning and Verbal Behavior 11:188-95.
We use cookies to analyze our traffic. Please decide if you are willing to accept cookies from our website. You can change this setting anytime in Privacy Settings.