Abstract: The automatic detection of facial emotion changes is crucial for survival. Numerous studies using event-related potential (ERP) technique have found that the amplitude of emotion-related visual mismatch negativity (EMMN) could be used to test the automatic processing of facial emotion. Previous studies suggested that deviant - standard stimulus (D-S) pair (different/same) and emotion type (negative/non- negative) might modulate the EMMN effect, however, the evidence so far was mixed. Therefore, we conducted a meta-analysis to analyze the findings of 35 studies (involving 721 healthy participants) on EMMN. Results showed that: (1) EMMN effects emerged at both the early- (0~200 ms) and late- (200~400 ms) stages, demonstrating that infrequently presented deviant stimulus elicited more negative ERPs at both the early- and late-stages. This suggests that EMMN reflects the probability effect of early- and late- stages emotion-related ERP components; (2) the type of D-S pair moderated the EMMN effect at the early- but not the late-stages. Specifically, the EMMN effect of different D-S pairs was significantly larger than that of the same D-S pairs at the early-stage; (3) in the studies of same D-S pairs, the evidence between equiprobable and non-equiprobable paradigm showed no significant differences in EMMN at both stages; (4) a negative bias was found in both early and late EMMN, i.e., the EMMN elicited by the angry, fearful, angry faces was significantly larger than that of happy faces. These results indicate that the EMMN effect is affected by experimental manipulations such as D-S pair type and emotion type.

Key words: emotion-related visual mismatch negativity (EMMN), meta-analysis, negative bias, deviant- standard stimulus pair