For more information contact Luis Rocha at rocha@lanl.gov
Last Modified: October 31, 2002
This is the abstract of a talk prepared for the International interdisciplinary seminar on new robotics, evolution and embodied cognition (IISREEC).12th to 15th November 2002, Lisbon, Portugal
Abstract: Imagine that you are a stranger in a strange land who has no knowledge of the language spoken there. As a native in this strange land, I direct your attention to the right, and you see a rabbit. I then direct your attention to the left, and you see a deer. Several minutes later, I direct your attention again to the right, the rabbit is gone, but I offer (with apologies to Quine) this label: gavagai. Here is your problem: To what does gavagai refer? It seems highly likely that you would take the word to mean rabbit. At any rate 14 month old children do. These toddlers trying to learn their first language really are strangers in a strange land and yet our experiments indicate that they will consistently link the now gone object on the right to a word uttered when again they look right. And, this is so even when the direction of attention is involuntarily shifted right (e.g., when we turn their heads). Indeed, we have discovered that young children will link a word to an object only when the word and object are also linked by a common direction of attention
The problem of how we map words to objects is a form of that ubiquitous problem in cognitive science, binding problem. We know that different parts of the human brain process different aspects of sensory and other information. How are the corresponding parts appropriately linked? We also know that human learning often requires solutions that bind events and processes separated in time. How are the corresponding events linked to each other? We also know that shifts of attention – from one object (idea) to another and back are common. How do we keep track of, index, these constituents? The new results from toddlers' word learning suggests that the physical world and our bodies play an important role in solving this problem. The body's momentary disposition in space (eye gaze, pointing, posture) indexes and links cognitive processes to specific entities in the world. This is an idea first offered by Ballard, Hayoe, Pook and Rao and called “deitic reference.”
I will present evidence from a series of new experiments on toddler's mapping of words to objects. The main conclusion from these experiments is that the body's movements, and particularly orienting movements, play an in-extricable role in bringing coherence to mental life. In the discussion, I will note the similarity of the phenomena in this word learning task to the A not-B error in infant's reaching and to infants tracking of objects. Both suggest a close link between object representations and spatially directed attention. I will present examples concerning the gestures that naturally accompany spoken language and also from American Sign Language to illustrate the pervasiveness of this use of spatially directed attention to bind objects to intended referents.