Learning Words from Context

Word learning is a notoriously difficult induction problem because word meaning is severely underdetermined by positive examples. One way of solving this problem early in development is to use prior knowledge in a word learning situation. Some have argued that young word learners have a number of “mapping” assumptions that reduce the overall hypothesis space, whereas others have argued that young children have a combination of “sampling” and “mapping” assumptions. In this research, we demonstrate that children may rely on a different mechanism -word learning is directed by associations among the context and a new word. These are novel findings pointing to a powerful associative mechanism underlying word learning.