The ability to categorize objects--such as dogs, cats, cars, furniture and people--is one of the foundations of early perceptual and cognitive development. Many of the objects in these categories tend to have features that co-occur together, such that things that have wings and beaks tend also to have claws and lay eggs. It is now well established that infants in the first year of life are adept at learning such co-occurrences among static non-moving features, and by the second year of life they can do so for dynamic features involved in motion (e.g., things with legs walk and things with wheels roll). It remains to be seen if infants are able to learn about such co-occurring features if they are rarely, if ever, observed together. This ability is important because it is impossible for infants to observe every feature of an object at the same time, and consequently they must generalize from their experience to develop expectations about unseen features of objects. The research supported by the award will use the infant-controlled habituation method to address this issue by examining whether infants and young children between 18 and 26 months of age are able to learn that objects that move in different ways tend to have a co-occurring object part even though they never experience the motion and the part together. In other words, if infants and young children learn that objects with legs walk and that things with legs have eyes, will they infer that things with eyes walk?
The findings will be of direct relevance to developmental psychology. The results will help researchers to understand how infants and young children learn about the complex features of object categories in the world, and they will demonstrate that children possess a potentially powerful--but as yet unexplored--way to generalize their knowledge in the first years of life. The project also has the potential for practical application for early educators because it could help to elucidate a completely novel way to teach young children how to make broader generalizations from limited data. This is important because classroom activities aim to promote the transfer of knowledge from the original instructional context and the kind of learning examined in this work may help to enable this transfer automatically. In a similar vein, it is also feasible to use the limits of this kind of learning to predict which math and science problems are likely to pose difficulty to children when entering school. In addition to the practical benefits for society, the proposed work will provide an outstanding training opportunity for undergraduate students to gain extensive one-on-one experience in conducting scientific research that will prepare them for graduate programs, and potentially, a career in science or education.
