Learning a Language from Inconsistent Input: Regularization in Child and Adult Learners