N170 Amplitude to Rare Neutral Faces in an Oddball Condition Reflects Prediction Error

Visual mismatch negativity (vMMN) of event-related potentials (ERPs) reflects automatic change detection in serially presented stimuli under the predictive coding framework. Previous studies have shown modulations of P1 and N170 components in response to changes in facial expressions, though these studies have not fully controlled for stimulus probability and low-level visual features. We recorded P1 and N170 to facial expressions and investigated their associations to vMMN in 36 participants of wide age range (19–65 years, M = 46.19, SD = ±13.07) and varying levels of depressive symptoms (Beck's Depression Index-II, M = 15.06, SD = ±12.43). Neutral, happy, and sad faces were assigned as deviant (p_deviant = 0.14) and standard stimuli in separate oddball conditions allowing comparison of responses to physically identical stimuli in the deviant and standard positions. A control condition with seven basic facial expressions (p = 0.14 for each) and no repetitive stimuli served as a control for predictive processing. We found evidence of predictive deviance detection in face processing, indicated in the N170 amplitude, but only for neutral faces (deviant > standard and deviant > control). Depressive symptoms did not correlate with ERP amplitudes or latencies, while the N170 amplitude increased with age. These findings suggest that, under conditions controlling for low-level features and stimulus probability, prediction error is robustly indexed by N170 amplitude in response to neutral faces. No such effects were observed for emotional (happy or sad) faces, or in earlier ERPs (P1) or latency measures.