The classic artificial grammar paradigm (Reber, 1967), in which participants judge whether unpronounceable letter strings obey or disobey an underlying rule structure (i.e., a grammar), was tested in two separate 2-session experiments over a 1-week interval. Whereas all but one previous experiment (Reber & Allen, 1978) with this task has focused on performance in one session, the present experiments are focused on the effects of retention, as well as transfer to new letters sets. Both halves of the experiments featured an acquisition phase, in which participants retyped exemplar strings quickly, and a test phase, in which participants judged whether entirely new strings obeyed the rule structure and rated their confidence. Half of the participants were re-exposed to the same grammar in the second session with new letters, and other half of the participants learned a new grammar with a new letter set. Additionally, in the second experiment, participants trained with either all positive exemplars or a mix of positive and negative exemplars. In the first experiment, both groups were faster, more confident, and more accurate in their grammatical judgments in the second session. However, the retention group, who were re-exposed to the same grammar, showed an advantage during the acquisition phase (with implicit measures), but not during the test phase (with explicit measures). In the second experiment, test performance was best without negative exemplars, but all groups improved on their implicit measures, although not on test accuracy. Taken together, these experiments demonstrated how improvement on performance with a complex task is dependent on the type of training and the degree to which knowledge of task rules is critical. For implicit components of the task, in which rule knowledge is less emphasized, performance improves with practice. For explicit components of the task in which rule knowledge is more emphasized, performance remains stable across practice.
展开▼
