ABSTRACT
In the artificial grammar learning (AGL) transfer paradigm Ss are instructed to memorize a set of strings, which (unbeknownst to them) were generated by a finite state grammar. They are then informed that the strings followed by a set of rules, and asked to categorise novel strings as following or violating those rules. Unlike the standard paradigm (Reber, 1967), the novel strings are composed from a different "alphabet" from the training strings, although the underlying rules are unchanged. Ss are nevertheless able to make the required distinction at above chance (and control) levels.
