초록 close

In this article, I address the learnability problem of linguistic rules with a large number of exceptions, using a computational framework. I investigate how children in Philadelphia learn short-a tensing, a rule subject to complex conditioning and many exceptions. Based on corpus statistics from the CHILDES database, I show that the tensing rule can be productively applied by children, confirming the Tolerance Principle (Yang 2005, Yang 2016), a computational model that makes quantitative predictions about the productivity of rules based on the distribution of input data. Also, I simulate a child’s acquisition of the short-a tensing rule with heterogeneous inputs and demonstrate that the Tolerance Principle correctly predicts that children receiving limited input may fail to master the rule.