ABSTRACT
Premack (1983a, 1983b, 1984) has argued that there is a profound disparity in the abstract problem-solving ability of language-trained chimpanzees and those that have not been exposed to such training. Chimpanzees trained on a language in which plastic pieces symbolically serve as words (Premack, 1976) can make correct judgments not only about the resemblance of physical objects, but also about the resemblance of abstract relationships. Non-language-trained chimpanzees, however, fail problems of the latter type. For example, in a matching-to-sample task both language-trained and non-language-trained chimpanzees will place apple with apple or bottle with bottle, but not apple with bottle. The disparity is seen when the animals are asked to judge the abstract similarity of identity and nonidentity relations by matching, for example, a pair of apples to to a pair of bottles rather than to a pair of nonidentical items like a cork and a spoon, or alternatively, matching a pair consisting of a padlock and an eraser to a paired can and paperweight rather than to a pair of shoes. Although Premack's language-trained chimpanzees correctly matched the
identity and nonidentity relations independently of the objects used to instantiate them, his non-language-trained chimpanzees failed the task. Interestingly, a similar, but age-related, disparity in humans is reflected in demonstrations that young children typically cannot solve relational oddity problems until they are about 5 years of age (see review by House, Brown, & Scott, 1974).
