Abstract: The late positive component (LPC) difference based on later memory performance is called the Dm effect. Many studies have shown that the LPC for items recognized in a subsequent memory test was more positive than for those that were not recognized. This effect was moderated by many factors, such as material type, encoding or orientating tasks, testing format, and study-test relationship. Most of the existing studies found that the low-frequency words showed a greater Dm effect than did the high-frequency words. However, some researchers did not find a significant Dm effect for high-frequency words. In addition, there is no consistency in whether or not access to preexisting representation is a prerequisite for the emergence of the Dm effect. In order to test the hypothesis that the Dm effect depends on accessing preexisting representation and to further explore the role of word frequency in the Dm effect, the present ERP study was designed to investigate the temporal and spatial distributions of the Dm effects for high/low-frequency characters, pseudocharacters, and noncharacters.
Eighteen healthy undergraduates participated in this study. There were seven blocks in the experiment including three real character blocks and four pseudo and noncharacter blocks. In each real character block, 118 characters (8 filler characters, 12 targets, and 98 study characters) were selected and presented on the screen during the study phase, after which the participants had to perform a test with 98 other additional characters. In each pseudo/noncharacter block, 78 stimuli (7 filler characters, 9 targets, and 62 study pseudo/noncharacter) were presented on the screen during the study phase, which was followed by a test phase with 62 additional pseudo/noncharacters. In each block, the stimuli were randomly presented on the screen for 200 ms. The duration between the onset of a stimulus and the very beginning of the next adjacent trial was randomized between 1400 ms to 1800 ms. The task was to determine whether or not each stimulus presented on the screen was a real character. An electroencephalography (EEG) was recorded continuously using an electro-cap with 119 locations (the extension of the International 10–20 System).
The results were analyzed using a five-way repeated-measures ANOVA for five time windows: 140~240 ms, 440~500 ms, 500~650 ms, 650~720 ms, and 720~800 ms. The five factors were hemisphere (left, right), location (frontal, temporal, central, parietal, and occipital), electrode, stimuli type (high- and low-frequency character, pseudocharacter, and noncharacter), and recognition (remember and forget). In addition, a three-way repeated-measures ANOVA was used for the central and occipital areas. The three factors were electrode, stimuli type (high- and low-frequency character, pseudocharacter, and noncharacter), and recognition (remember and forget). The ANOVA results indicated that the temporal and spatial distributions of the Dm effects for low-frequency characters were longer and wider than those for high-frequency characters. Moreover, the Dm effect was generally greater for noncharacters than it was for high/low-frequency characters. However, the Dm effect was not observed for pseudocharacters in any interval.
First, a significant character frequency effect was found in the Dm effect, suggesting that less commonly used characters are processed in a manner that facilitates their subsequent recognition. There might be different neural mechanisms between subsequent memory effects of high- and low-frequency characters. Second, the emergence of a subsequent memory effect may not depend on accessing preexisting representation

Key words: event-related potentials, subsequent memory effect, pre-existing representation, character frequency