This study used a directed-lying paradigm to explore the possible differentiation between identification and classification processes involved in facial recognition. The directed-lying paradigm made it possible to measure honest and deceptive classification responses for familiar and unfamiliar faces. In general, the participants were more accurate at identifying familiar faces than unfamiliar faces. Results based on the two a priori markers for facial familiarity showed that the N400f occurred with both identification and classification, while P600f only appeared with identification. Regardless of whether they had to lie or tell the truth, the participants had more positive-going N400f and P600f in the middle and right parietal regions when they were identifying familiar faces compared to when they were identifying unfamiliar faces. Classifying faces as “familiar” (under both truth and lie conditions) elicited more negative-going N400f in the central and right temporal regions. Both conventional and SPM analysis gave convergent evidence of these observations.
The results showed that the N400f was associated with both the identification and classification processes in facial recognition. The difference is that in the identification process, familiar faces were found to elicit less negative-going N400f than unfamiliar faces. In contrast, in the classification process, faces classified as familiar elicited more negative-going N400f than those classified as unfamiliar. This suggests that it is likely that, despite sharing the same N400f, the two steps in facial recognition involve two discrepant cognitive processes. Previous studies have consistently found that the N400f is involved in discriminating familiar faces from unfamiliar ones 
. Previous researchers have also related the N400f to high-level attentional processing 
. They found that familiar faces elicited more negativity than unfamiliar faces at Cz and Pz; however, such effects were diminished when attention was directed to another demanding task. It is plausible that the N400f also reflects top-down control in the processing of familiar faces.
In this study, regardless of whether the participants were telling the truth or lying, faces classified as “familiar” elicited more negative-going N400f than those classified as “unfamiliar.” These results were interesting because lying did not seem to affect the N400f. In other words, classification could be determined by the intention of categorizing faces as familiar or unfamiliar rather than by the possible regulatory processes of lying.
Moreover, there were different polarity effects for identification (i.e., familiar>unfamiliar) and classification (i.e., “familiar”<“unfamiliar”). Compare this to the findings of previous studies that found N400f amplitude differences: familiar faces classified as familiar elicited more negative-going amplitudes than unfamiliar faces classified as unfamiliar 
. This suggests that the N400f can also be attributed to an intended classification of the facial stimuli rather than to the mere identification of familiarity based on facial features.
The P600f was associated only with the identification of faces, not with lying. Familiar faces elicited a more positive-going component than unfamiliar faces in the middle parietal and occipital regions. This identification process seems to be independent from the classification process since the interaction was not significant. Our findings are consistent with a previous study that reported that the P600f component was related with more positive amplitudes for familiar faces than for unfamiliar faces at the Pz channel, regardless of whether attention was directed to another demanding task or not 
. The association of face identification with the P600f is somewhat counter-intuitive. One would expect the perception of facial features to be a prerequisite for identification, which precedes classification, but the results of this study suggest that some sub-process of identification (reflected by the P600f) takes place even after classification (reflected by the N400f). Rugg and Curran offered a plausible explanation for the temporally split phenomena underlying face identification 
. According to their findings, the recognition of an object involves a dual process: a judgment on familiarity followed by a recollection of the information associated with the object. The identification sub-process, as reflected by the P600f in this study, suggests that the participants probably accessed the information related to the familiar faces. This proposition is supported by previous studies which related the P600f to the processing of the demographic characteristics (e.g., name and occupation) of known persons or the visual characteristics of known faces 
. The recollection of information about familiar faces is likely to be automatic as the participants were not instructed recollect such information in the task used in this study.
In the directed-lying task used in the study, the intention to lie or tell the truth was prompted before the presentation of a face. The appearance of the face would have enabled the participants to extract the facial features embedded in the stimuli (associated with N170). Setting the intention to lie as part of the classification process appears to occur at around 280 ms (denoted by N400f). Our results also suggest that attention would be allocated as part of the identification process at round 300 ms. The extraction of semantic information about faces, which is independent of the decision to lie (classification), is likely to occur at around 500 ms.
An interesting finding in this study is the late onset of the classification process, elicited around 760 ms in the bilateral temporal to occipital regions. Temporally, this signal is distinct from the N400f and P600f components. Other studies have reported a face-specific negative component (N700) elicited about 700 ms post stimulus from intracranial electrodes placed on the cortical surface of the ventral and lateral brain regions 
. The N700 was found to be related to semantic priming in a task involving learning and identifying face names 
, perhaps reflecting a top-down process. However, the contribution of this component in the processing of facial familiarity is beyond the scope of this study and should be investigated in further studies.
Our findings support the differential roles of identification and classification processes in the recognition of familiar faces. The N400f appears to be linked with both identification and classification, while the P600f appears to be primarily linked with identification. The dual identification processes revealed in this study are likely to involve an earlier judgment of familiarity and a later recollection of information related to familiar faces. Using electrophysiological measures together with brain imaging could further differentiate the role of these two components.
This study has a few limitations. The faces used were of personal acquaintances, but previous studies have shown that learned familiarity 
and celebrity faces 
evoke different processes. Therefore, caution should be taken when generalizing the current findings to female adults and those with different demographic characteristics to the participants in this study. The results may also not be generalizable to other types of familiarity. More studies on the neural processing of face identification and classification are still needed.
This study also has implications for lie detection. However, there are at least two things that need to be done first. First, this paradigm needs to be tested on more samples to obtain a consistent standard for lie detection. Second, the measures should be adjusted to the special neural characteristics of each person so that the method is able to detect lies better in individual cases.