Symbols

Premack’s (1976) alternative method, mentioned previously, was based on plastic tokens that his chimpanzees could manipulate and place on a magnetic board. This procedure had two advantages over the sign language method. First, a sentence of tokens could be easily recorded and analyzed objectively. Second, since the tokens were under the animal’s continuous view, sentence formation placed a light burden on short-term memory.

Using this token procedure, chimpanzees were trained in progressively more complex tasks. With a repertoire of about 130 tokens, the vocabulary of a chimpanzee named Sarah allowed for the design of a variety of problems including naming, interrogation, and object classification. Chimpanzees were trained to comprehend commands involving two-, three-, and four-word sentences. For example, when shown a piece of apple, they were required to place the tokens for “Give apple” in the board; subsequently, the chimpanzee had to correctly choose the token corresponding to its name from two different names and place it at the end of the string (“Give apple Sarah”); and finally, a token corresponding to a trainer’s

Higher Cognitive Processes 307 name was added so as to form a request of the form “Debbie give apple Sarah.” Test performance was above 90% correct when sentences involved two words. However, correct responses dropped below 75% correct immediately after a shift to four-wo rd strings. Errors included choosing the wrong token and placing the tokens in an incorrect sequence (e.g., “Give Debbie apple Sarah”). With sufficient training, performance in four-word sentences exhibited some improvement but, curiously, the chimpanzees refused to continue with the training.

A somewhat more elaborated procedure involved lexigrams-—arbitrary geometric patterns symbolizing objects and actions. Lexigrams were presented as keys in a keyboard; chimpanzees were trained to press the keys to form strings that could be roughly translated into English phrases such as “Please machine give piece of apple.” In one study combining language comprehension and production, chimpanzees Sherman and Austin were first trained to share food (e.g., apple) and instruments (e.g., wrench) and subsequently were exposed to situations that required their cooperation to obtain a particular reward. For example, one chimpanzee learned to request a tool from the other to obtain food by using the correct tool and food lexigrams. Once obtained, the food was shared (Savage-Rumbaugh, Rumbaugh, & Boysen, 1978).

Notice that the procedure used by Premack (1976) aimed mainly at the resolution of problems based on tokens, but symbols were not use for communication, thus lacking an important feature of language. It is not surprising that the performance of these chimpanzees, impressive as it is, can still be understood in nonlinguistic terms. Sign-language training procedures had communicative properties, but their results were difficult to evaluate and the possibility of experimenter cueing difficult to eliminate. The lexigram procedure, which had none of these drawbacks, has also been applied to dogs with some success (Rossi & Ades, 2008). A dog named Sofia received training from an early age to use symbols displayed in a keyboard to request objects. Initially, the lexigrams were placed closed to the object they represented, but, eventually, the distance was increased until the lexigrams appeared only as part of a keyboard. Sofia was required to paw a particular symbol when requesting specific objects. Sofia showed some evidence of categorization by using a given symbol to request access to a variety of related objects (e.g., toys). Sofia did not use the keyboard when alone, but only when in the presence of humans, a fact suggesting a communicative function of the lexigrams.