For the RM paradigm, 2-s epochs (300 ms pre-stimulus baseline) were extracted, low-pass filtered at 20 Hz (−24 dB/octave), corrected for horizontal eye movements, and rigorously screened for residual artifacts (Kayser & Tenke, 2006b) or electrode bridging (Tenke & Kayser, 2001). Artifactual or bridged surface potentials were replaced by spherical spline interpolation (Perrin, Pernier, Bertrand, & Echallier, 1989) using the available artifact-free data if possible (i.e., when less than 25% of all EEG channels contained an artifact); otherwise, a trial was rejected. Stimulus-locked ERPs were computed from correct, artifact-free trials separately for each task (i.e., words or faces) but across all other conditions (i.e., old, new, and filler items; cf. Kayser et al., 2010) to maximize the signal-to-noise ratio for the targeted N1 component. The means for the number of trials used to compute these ERP averages (M ± SD, HC vs. NH vs. AH) were 349 ±41 vs. 333 ±41 vs. 317 ±39 for faces, and 372 ±73 vs. 350 ±36 vs. 333 ±42 for words (range 207 to 447).
