Abstract
The proportion of the words that were common to the study phase and the test phase was manipulated in two experiments. Subjects previously studied either 20% or 80% of the words appearing in the test. In Experiment 1, the subjects performed a word fragment completion test. In the 20% condition, reading words produced more priming than generating words. In the 80% condition, however, generating words produced priming as much as that of the 20% condition. Experiment 2 examined this proportion overlap effect on fragment cued recall test. The generation effect was observed in the 80% condition, but not in the 20% condition. These findings indicate that performance on memory tests is determined by the manipulations of both encoding and retrieval conditions.
