Abstract: Human memory can be characterized as an elaborate network of stored representations. Researchers propose that both working memory (WM) and long-term memory (LTM) utilize same representations, and the representations they activate are in diverse patterns. So the similarity and coherence between WM and LTM are emphasized in the activation-based model, and the priming effect of LTM on WM and vice versa should be both found. By simultaneously recording event-related potentials (ERPs), the present study aimed to investigate two long-term semantic priming effects to illustrate the common representations involved in WM and LTM. Fifteen college students (mean age=20.33±1.91; 7 male) participated in the experiment. The participants were all right-hand, had normal or corrected-to-normal vision and had no neurological or psychological disorders. Each participant signed a consent form prior to experiment and was paid after experiment. We combined a WM component (short-term verbal encoding) and a LTM component (category comparison) into one task, so that WM and LTM processing could be concurrently investigated. LTM priming on WM was manipulated by asking participants to learn before short-term verbal encoding, and the priming effect of WM on related LTM should be found in category comparison. Results revealed two long-term semantic priming effects in behavior data and ERPs. The first effect was found in short-term encoding phase, which was reflected as the priming of LTM on WM. Compared with new targets in WM, the reaction time of studied targets was shorter. They elicitedearlier N2 and P3, and they also decreased N2 amplitude. These may index the so-called neural priming (or repetition suppression) of scalp potentials. However, there was no difference between new and studied distractors in WM. Furthermore, the neural priming of WM on LTM also exhibited in category comparison. Specific contents, which were formerly concerned or ignored in short-term encoding phase, led relevant representations of LTM to diverse activation patterns. N400 elicited by target category (items in which were the same category as the targets in short-term encoding phase)was more positive than that of unprimed category, which also index the neural priming of scalp potentials. But this neural priming was modulated by repetation, that is, when target category was relevant to studied targets rather than new ones, the N400 would be more negative. Results indicate that common representations are used both in WM and in LTM, supporting the activation-based model. Furthermore, these results point to the key role of attention as a modulator in the linkage between WM and LTM.

Key words: working memory, long-term memory, long-term semantic priming, Event-Related Potential (ERP), neural priming