Fig. 1.

Experimental design of Experiment 1 (A) and schema of the cross-repetition pattern analysis (B). A total of 120 novel faces were studied over three scanning runs. (A) Each face was repeated four times. They were categorized post hoc, as remembered faces and forgotten faces, according to performance on the recognition memory test administered after a 1-hour delay. Each presentation of the remembered faces (R1 to R4) and forgotten faces (F1 to F4) was separately modeled. (B) Pattern analysis was based on independent structural ROIs (top row) (10). Activation pattern in a given ROI was extracted for each presentation (middle row), and then subjected to Pearson correlation analysis. The encoding variability hypothesis predicts that the degree of pattern similarity for subsequently remembered faces is lower than that for subsequently forgotten faces (bottom row).