Re-examination of artificial grammar paradigm (in Chinese)

Abstract

This paper reviewed many researches on implicit learning with the finite-state grammar. This grammar is a inclusive concept, a concept without any irrelevant attributes, a concept with limited variance. In the previous researches, it was treated only as a tool to generate strings. It’s biases during string generation was ignored. These biases include: The strings were ill-proportioned according to path, what the learning strings expressed was not what the testing strings tested, and the learning strings in each path couldn’t express the rules of this path. So the string set of previous researches didn’t express the grammars designed by experimenters completely, accurately and exclusively. Subjects who didn’t learn grammar sufficiently, and what he learned may not be the grammar which experimenter initialized. Therefore, the evidence of implicit learning and its characteristics was not sufficient and strong.