Introduction

In this chapter we shall examine the main principles that govern variation in word order in SAIE, with emphasis on three striking characteristics: a high ratio of parataxis to hypotaxis; a wide range of constructions associated with a verb-final language; a high degree of topicalisation. The last of these three will be subjected to a detailed quantitative scrutiny.

Parataxis

This denotes a preference in some languages or dialects for the loose conjoining of clauses rather than the (hypotactic) use of subordination. These are to be conceived of as tendencies on a continuum, rather than absolute poles. It would be inaccurate to claim that SAIE shows a greater preference for parataxis than hypotaxis, since even the basilect shows use of a wide range of subordinating particles (for example, that, if then, etc.). Nevertheless, in comparison with most dialects of English the degree of parataxis seems significantly higher. It is not possible to give statistics here: our justification for this statement will be qualitative – showing the range of permissible paratactic clauses in the basilect.

Paratactic circumlocution

Basilectal speakers use an oral mode which favours circumlocution. In the following three examples, from three different speakers, a short pause is indicated by /, a longer pause by / / . (This notation is required since the use of commas or full stops might impose the norms of RP intonation and pause structure on the data.)

The argument about language in apes has a striking parallel in the debate among philosophers, psychologists, and computer scientists over the possibility of artificial intelligence. Much of the effort to create intelligent automata is justified by reference to an operational definition of machine intelligence proffered by Alan Turing some 40 years ago (1950). Turing asserted that the question of whether a machine can think is too imprecise and suggested that we assess the intelligence of a machine through an “imitation game.” The procedure he proposed was a bit more complex than the current conception of “the Turing test,” but the logic is unchanged. In this exercise, humans pose questions, on any subject, to either a person or a computer without knowing which. If the human interrogators are unable consistently to tell from the responses whether they are interacting with a person or the machine, then, according to Turing, we have the best evidence we could have for intelligence in a machine.

The most prominent critic of this behavioristic measure of intelligence is the philosopher John Searle (1980), who argues that simply because a machine is able to answer questions as a human would, it is not, ipso facto, engaging in thinking. Thought does not arise in a computer merely because it manipulates symbols; these symbols are devoid of meaning for the computer. Imagine, Searle bids in a now well-known thought experiment, a person ignorant of the Chinese language who is placed inside a room equipped with baskets of Chinese characters. He is also provided with a book of instructions for constructing sequences of characters (identified by shape) in response to sequences presented to him.

In case the preceding pages have not made this clear, I should state that I do not believe that any of the ape-language projects succeeded in instilling even a degenerate version of a human language in an ape. There are no persuasive data in support of syntactic patterning. At best, one or more of the animals in these studies acquired an ability, or enhanced an inchoate one, to represent things with symbols. However, the evidence for even this capacity, a prerequisite for syntactic productivity, is equivocal.

Is it absurd to consider the linguistic incompetence of apes as something requiring explanation? After all, no one would think to pose the question of, say, why Homo sapiens cannot fly, or why birds cannot swim. These questions are not asked because it is assumed that taxa separated by millions of years of evolution will differ in their adaptations, the extent of the divergence corresponding roughly to the length of that separation, barring parallel or convergent evolution. Given that our lineage diverged from the most closely related hominoid at the very least four million years ago, which separates us by eight million years of independent evolution, why is it thought likely by some that an ape species would possess a human faculty, especially an unused form of it?

The molecular geneticists King and Wilson (1975) presented an abundance of evidence that the structural proteins of humans and chimpanzees (Pan troglodytes), as well as the genes coding for those proteins, are 99 percent identical between the two species.

In Intelligence in Ape and Man, Premack (1976a) characterizes his studies as focused on intelligence rather than language per se, language being of interest only insofar as it illuminates the former. He also maintains that the language imparted to Sarah was not intended to simulate a human one. Despite these disclaimers, it is clearly language and its cognitive bases that are the topic of Premack's writings about Sarah. Furthermore, Premack implicitly and often avowedly regards Sarah's accomplishments as cognitively identical to aspects of human language. In reality, though her intellectual achievements may have exceeded those of Lana, Sarah's are probably no more languagelike than Lana's.

THE SEMANTICITY OF SARAH'S CHIPS

The major critique of Premack's work with Sarah was made by Terrace (1979a), and the following discussion relies extensively on that work. It is interesting that this most effective critique came not from a linguist anxious to beat back an interloper into the domain of human language but from a fellow behavioral psychologist, whose expertise in experiments on problem-solving by lower animals resulted in a critique “from below” rather than “from above,” that is, an argument that Sarah's performance is not different in nature from that demonstrable in rats and pigeons. In a nutshell, the Premack study, like the work with Lana and, less obviously, the signing projects, suffered from rampant overinterpretation.

As an example, consider a typical “sentence” that Sarah learned to compose in order to procure an incentive: “Mary give Sarah apple.” If one were told that Sarah reliably constructed this sequence in the presence of an apple, substituting the chip for banana if this were the incentive, one might well be impressed.

The literature on naturalistic primate communication has something for everyone – those seeking to document continuity between human language and infrahuman communication as well as those promoting a natural hiatus. Despite this lack of consensus, I think it is accurate to describe the literature as providing much more support for the discontinuity position.

The issues in the primate-communication literature bearing on the origin of language lend themselves to a number of contrasts. Some of these contrasts are dichotomous, while others define a continuum of possibilities, with language at one end and primate systems, if not at the other, than at least elsewhere. It will be apparent that several of the oppositions presented in the following pages are conceptually linked; in a more meticulous treatment, some might prove to be essentially synonymous oppositions.

VOCAL–AUDITORY VERSUS OTHER CHANNELS

Human linguistic exchanges typically employ a visual channel, parallel to the Vocal–Auditory one, which contains information conveyed by such “paralinguistic” means as hand gestures, body attitude, and facial expression. In addition, the Vocal–Auditory channel, through such aspects as pitch, stress, and volume, carries information supplementary to the “words” themselves, information about the speaker's current emotional disposition in general, toward the listener, and regarding the objects of reference in the discourse, as well as clues about the speaker's personal identity, age, gender, geographic origin, and so forth. It is the verbal content of the Vocal–Auditory channel, nonetheless, that is properly considered the central information in language; far more of what is distinctive in human communication would be lost by deleting the words of a conversation than by blocking out the gestures and facial expressions.

THE NATURE OF LANA'S SYMBOL STRINGS

The assessments of Lana's performance by the Lana researchers take a strong form of the tendency among ape-language researchers to overinterpret their animals' behavior in the direction of human language, elevating to linguistic status behavior more cogently explained in simpler terms. Lana's “sentences” provide a good example.

It is clear that most of Lana's sentences were learned as rote sequences of key presses, yet Rumbaugh et al. gloss each lexigram within them with an English word. The inappropriateness of this practice is most clearly seen in the rendering of the request indicator as “please” (for example “please machine give apple”). Not only is there no evidence to suggest that Lana had any notion of the meaning of “please” or even a child's rudimentary understanding of the sociolinguistic rules governing its usage, but there is also little reason to believe that the other, more concrete terms were meaningful for her either. On the contrary, there is evidence that they were not: long after Lana had learned to “request” these incentives with her sentences, she required 1,600 trials to learn to push the corresponding lexigram when shown an M & M or a banana and asked for its name (Rumbaugh and Gill 1977). The Lana researchers have denied ever suggesting that every element in Lana's strings had linguistic significance (Savage-Rumbaugh and Rumbaugh 1980c), but one must wonder why these elements were nonetheless given English terms. Presumably, this was done to indicate to readers of the publications the concepts the experimenters intended Lana eventually to attain and to convey the pragmatic significance of each term, that is, what would result from its use.

EARLY STUDIES

Before the Gardners' innovative attempt, described below, to inculcate a visual human language in a nonhuman primate, at least half a dozen projects had been undertaken with the aim of either actively conferring a spoken language on an ape or observing the possible “natural” acquisition of one within a human home environment. The linguistic results in each case were dismal.

Witmer (1909) reported on a chimpanzee that had been trained to approximate crudely the word “mama.” Furness (1916) succeeded, after much labor, in getting an orangutan to utter a discernible “papa” and “cup.” Hayes and Hayes (1951) attained the greatest success among these early projects; at the end of six and a half years of home-raising a chimpanzee, Viki, the Hayeses had managed to teach her to utter “papa,” “mama,” “cup,” and, less successfully, “up.” But Viki's articulation was poor and there was little evidence that these words served a referential, that is, symbolic, function for her.

On the other hand, the apes' “comprehension” of spoken words and phrases substantially exceeded their productive abilities. For example, Gua, a chimpanzee home-reared by the Kelloggs (1933), outpaced the Kelloggs' young son in the number of phrases to which she could respond correctly. This was true up to the end of the fourth month of the study, after which the boy surpassed Gua. Rather than positing understanding, Kellogg (1968) speaks of the chimpanzee “reacting correctly,” but he does not address the question of how much of this reacting could be attributed to linguistic decoding as distinguished from nonlinguistic cues and contextual information.

This book is about the experiments carried out over the past two decades in which it was attempted to impart a language, either natural or invented, to an ape. The debate engendered by these projects has been of interest – consuming for some, passing for others – to all of those whose concerns include the enduring questions of human nature, among them anthropologists, psychologists, linguists, biologists, and philosophers.

An adequate treatment of the linguistic capabilities of apes entails consideration of a number of related issues, each of which is an interesting problem in its own right. Continuities in primate mentality, the relationship between language and thought in the individual and in the species, and the origin of language in, again, both the ontogenetic and the phylogenetic senses, are themes that will recur throughout this work. Development of parts of the argument will require moderately technical excursions into American Sign Language grammar, recursive rules in language, the neuropsychology of language, naturalistic primate communication, and language acquisition in children. A grounding in the last topic, in particular, is crucial to the argument, for the apelanguage dispute is essentially a quarrel about how similar the performance of the linguistic apes is to that of young children acquiring language.

The method followed in this book is one of detailed (though, hopefully, not tedious) critical analysis of the experimental methods and conclusions of the ape-language projects. The analysis is based on data summaries, published anecdotes, and experimenters' conclusions – in short, on published material rather than on primary data.