METHODS & DESIGNS A computer-controlled language training system for investigating the language skills of young apes* DUANE M. RUMBAUGHt, TIMOTHY V. GILL, for linguistic "productivity" (Hockett, 1960) as we and JOSEPHINE V. BROWN know it in man, and, if the answer is in the affirmative, Georgia State University. Atlanta. Georgia 30303 (b) What is the limit of that language capacity? Will it E. C. von GLASERSFELD and PIER PISANI allow for conversation with man or between apes? Willit . University of Georgia, A thens, Georgia 30601 provide for creative expressions, as descriptions or questions? and The authors of this paper have undertaken a long-term HAROLD WARNER and C. L. BELL study to help answer these questions, which are viewed Yerkes Primate Research Center ofEmory University as reasonable extensions of the work of others. The Atlanta, Georgia 30322 approach and methods have been designed (a) to enhance the objectivity and efficiency of inquiry into A computer-controlled language training system was the language-relevant behaviors of ape Ss, and (b) to designed and constructed to enhance the objectivity and efficiency of inquiry into the language-relevant behaviors develop a technology that will allow for systematic of apes. The system allows the S to gain control over the investigation of the parameters influencing the events of the 24-h day in direct correspondence with its acquisition of these behaviors. competence in using a keyboard on which each key To se rve th ese ends, a computer-controlled represents a word. Various incentives can be obtained through the selection and depression of appropriate keys environment was designed and constructed within which in accordance with rules of sentence structure monitored the ape Ss can come to control increasingly the events of by a computer. The system is flexible and allows for the 24-h day. As language-relevant skills are perfected, eventual conversation between man and ape, with the the Ss will be able to choose between incentives-Toods, computer as the intermediary. A Teletype records all drinks, play objects, playmates, and human that transpires. Achievements of the chimpanzee S over the course of the first 8 months of the system's associates and experiences-e.g., music, films, play, and operation attest to the worth of the system and training other kinds of social interactions. The system provides methods. for possible conversation between man and ape, with the computer serving as the reliable intermediary, monitor, The capacity of true, productive language has long and recorder of all linguistic events that occur. The been held by many professionals and laymen as uniquely computer record will allow detailed examination of human, but the recent studies by Gardner and Gardner language acquisition. Although the system is currently (1971) and by Premack (1971) give reason to question used with ape Ss, there is no reason why, in modified this view. These investigators have demonstrated that the form, it might not be used also with human infants, both chimpanzee (Pan) can acquire behaviors that suggest the normal and retarded. mastery of a vocabulary and at least a circumscribed A language, and methods of monitoring the use ability to use these words in novel chains apparently thereof, had to be devised to launch the project. The appropriate to newly encountered test and living skills of psychologists, a linguist, a computer specialist, a situations. The capacity of the chimpanzee for mastery biomedical engineer, an electronics technician, and a of rules of sentence structure, however, remains in skilled behavioral research technician collectively question, for this was not the primary line of inquiry in contributed to the development of the study program, either of these studies. which included the design of the language system All studies have their limitations, and the ones (Yerkish), the methods for displaying and recording the referred to above are not without theirs. Their linguistic messages, the training program for the apes, pioneering and creative qualities are commendable, but and the fabrication of the entire system. The system is their methods and scope leave unanswered two questions now operational, and the experiences of the first 8 of primary importance: (a) Do apes have the capacity months have served to validate the team's deliberations and decisions. It is premature to conclude that our ape Ss are clearly using language, but it is clear, even now, that the approach and the system promise success in answering, *This research was supported by NICHD Grant HD 06016-01,02 and by NIH Grant RR-00165. at least in part, the questions posed above. The following tPlease address reprint requests to Duane M. Rumbaugh, Department of Psychology, Georgia State University. 33 Gilmer report provides details regarding the system and methods Street, S.E., Atlanta, Georgia 30303. of inquiry. Bebav. Res. Meth. It lnatru., 1973, Vol. 5 (5) 385 Table 1 (1960) called "semanticity" and is unquestionably a Design Elements prerequisite oflinguistic communication. Display System 0 X0 Consequently, it was decided that a visual language 2 3 .( 5 whose lexical items were graphic designs of an "abstract" nonrepresentational kind would be used in the present project. For various technical reasons (e.g., ~ JVVV\. automatic coding and projection of the Yerkish words), • 0 it was desirable to design these abstract symbols in such 6 7 8 9 a way that they could be combined by superimposing a number of basic design elements. Nine single design Elements 2,.(,7 Orange Elements 1, 7, 8 Orange elements, plus combinations of two, three, or four of them, yielded 255 symbols which were then further differentiated by three basic colors and three easily ji ~ discriminable color combinations (see Table 1). BLANKET BOX In the present implementation of the system, 125 of these combinations are used as word symbols, or Elements 1,5,6 Red Elements 5,9 Red lexigrams. The colors of the lexigrams constitute a simple semantic code. Red lexigrams, for instance, are used to designate concepts of ingestible items, i.e., food ~ B and drink; blue lexigrams are used to designate activity NUT . WATER concepts, etc. (see Table 2). This semantic code is somewhat extended by the fact that lexigrams Elements 1,3,8,9 Blue Elements 1,2,3,5 Blue corresponding to certain conceptual classes are characterized by the presence or absence of specific ~ ~ design elements. TO BITE TO GROOM Table 2 Generic Conceptual Categories and Corresponding Design Elements DESIGN OF THE YERKISH LANGUAGE AND INITIAL TRAINING Design Design Attempts at establishing linguistic two-way Color of Elements Elements Generic Semantic Lexigram Present Absent Category communication with nonhuman primates, using a spoken natural language such as English, Russian, or 6 Animate being (items that Japanese, have been unsuccessful (ef. Kellogg, 1968; act spontaneously) Ploog & Melnechuk, 1971). In the early days of modern Violet 3,6 Human beings 3,6,7 Members of the team primatology, Yerkes (1925) had already observed that (persons of S's normal great apes are physiologically handicapped with regard environment) to the production of speech-like sounds. More recently, 7 6 Physical objects of environ­ it has been shown (Lieberman, 1968; Lieberman et al, ment 1969) that the vocal apparatus of apes lacks the Orange 6 9 "Categorial concepts" (e.g., anatomical and neurological features considered beginning, top, piece, etc.) 5 3,4,7 Edible items indispensable for the emission and modulation of many Red of the sounds that constitute phonemes in natural 5 3,4,6,7 Drinkable items human languages. Green 6 7 Parts, of body The studies of the Gardners (1969, 1971) and of 7 Ambiental conditions and Premack (1970, 1971), the first using American Sign events (e.g., heat, light, Language (ASL), the second, an artificial language of music) Blue-Gray 6, 7 States of physical objects colored plastic shapes, have demonstrated, however, that (adjectives such as clean, a chimpanzee can acquire at least some of the skills that cold, dry, open, etc.) are usually subsumed under the rather vague generic 3 Activities (except percep­ term "language." Their success was made possible by a Blue tion) switch from a vocal to a visual means of communication. 6 3,7,8,9 Perceptual activities The data collected from Washoe and Sarah show that a "Prepositions" and other chimpanzee can both acquire and retain the symbolic Black particles designating use of visual items. This is equivalent to what Hockett relational concepts 386 Behav. Res. Meth. & Instru., 1973, Vol. 5 (5) The Yerkish Language always unequivocal as to the syntactic structure they are intended to express. Thus, for instance, the sentence The concepts designated by lexigrams are, on the "visiting professors may be boring" (an example cited by whole, much the same as the meanings of English words. Schank, 1972, and Lindsay & Norman, 1972) is The one important difference is that, while nearly all ambiguous, not because the receiver cannot decide on English words are ambiguous (semantically, the meaning of the individual words, but because he syntactically, or both), lexigrams are always does not know whether it is the "professors" who do the unequivocal. The meaning of many lexigrams, therefore, visiting or whether someone else is visiting them. That is is somewhat restricted in comparison to the meaning of to say, the ambiguity concerns the assignation of certain the "corresponding" English word. The lexigram for roles that arc important in the designated situation. "stick," for instance, means "thin, longitudinally The grammar of Yerkish has been designed to avoid extended, solid object" only, and it has none of the this second kind of ambiguity. It was derived from the other meanings of the English noun "stick" or those of correlational grammar developed for computer the verb. applications in the area of automatic language analysis One of the major aims in designing Yerkish was to and machine translation (von Glasersfeld, 1964, 1970).