Abstract：In everyday life, we encounter familiar and unfamiliar faces in many different environments. Many studies have demonstrated that face recognition is in fact a complex cognitive achievement. Although prosopagnosic individuals are severely impaired with explicit recognition of familiar faces, there have had substantial physiological and behavioral evidences for “covert face recognition”. Many evidences from event-related potential (ERP) investigations of brain activity with healthy participants have confirmed that covert face recognition and overt recognition have different neural bases at retrieval. But at encoding, the neural distinction between covert and overt face-related encoding processes is unknown. A powerful method of examining the neural basis of encoding is to measure neural activity during the study phase of an experiment and then to sort out these measurements according to subsequent memory test performance. ERP experiments employing intentional tests have typically shown positive deflection of waveforms recorded during study of faces that are subsequently recognized, comparing with waveforms recorded during study of faces that are subsequently unrecognized. The deflection is called “difference due to later memory” (DM) effect. However, DM effects for covert recognition in incidental tests have not been consistently observed, owing to methodological ambiguities in prior studies which often compared incidental tests with intentional tests. In fact, brain activity in one test can reflect not only covert but also overt recognition. So it is important to distinguish covert and overt recognition for specific episodes in one test. An experiment was conducted with subsequent memory paradigm and novel faces as stimuli. ERPs were recorded while participants were studying visually presented faces. The ERPs were sorted by whether or not participants would later recognized the studied face in a forced-choice recognition where two faces (one studied and one new) were presented concurrently, and by whether or not they would indicated that the choice was remembered from the study list. If they could not choose a studied face, subjects were encouraged to guess. Study trials from which the studied faces were later selected and remembered (remembered trials) were compared with those from which the studied faces were later selected and not remembered (primed trials), in order to measure the ERP difference (Dm effect) associated with later overt recognition. Primed trials were compared with study trials from which the studied faces were later not selected (forgotten trials), in order to measure Dm effect associated with later covert recognition. The results showed that comparing with forgotten trials, primed faces were associated with more negative waveforms at study over the frontal-central electrodes during 400~500 ms. This negative-going Dm effect was distinct from the positive-going Dm effect associated with later overt recognition by virtue of its polarity and topography, that is, comparing with the primed trials, remembered trials were associated with more positive waveforms at study over the parietal electrodes starting 400ms after face onset. We suggested that the negative-going Dm effect over the frontal-central electrodes could index the elaborated processing of encoding information into the perceptual representation system, and the positive-going Dm effect over the parietal electrodes could index the processing of encoding information into the episodic memory system. So only the stimuli which were encoded into the episodic memory system could be retrieved consciously during a later test. If the stimuli were not encoded into the episodic memory system, rather, they were encoded into the perceptual representation system by the elaborated processing, then minimal activation of corresponding memory traces would be triggered during tests, which could give rise to covert recognition.