Efficient decoding of even brief and slight intensity facial expression changes is important for social interactions. However, robust evidence for the human brain ability to automatically detect brief and subtle changes of facial expression remains limited. Here we built on a recently developed paradigm in human electrophysiology with full-blown expressions (Dzhelyova et al., 2017), to isolate and quantify a neural marker for the detection of brief and subtle changes of facial expression. Scalp electroencephalogram (EEG) was recorded from 18 participants during stimulation of a neutral face changing randomly in size at a rapid rate of 6 Hz. Brief changes of expression appeared every five stimulation cycle (i.e., at 1.2 Hz) and expression intensity increased parametrically every 20 s in 20% steps during sweep sequences of 100 s. A significant 1.2 Hz response emerged in the EEG spectrum already at 40% of facial expression-change intensity for most of the 5 emotions tested (anger, disgust, fear, happiness, or sadness in different sequences), and increased with intensity steps, predominantly over right occipito-temporal regions. Given the high signal-to-noise ratio of the approach, thresholds for automatic detection of brief changes of facial expression could be determined for every single individual brain. A time-domain analysis revealed three components, the two first increasing linearly with increasing intensity as early as 100 ms after a change of expression, suggesting gradual low-level image-change detection prior to visual coding of facial movements. In contrast, the third component showed abrupt sensitivity to increasing expression intensity beyond 300 ms post expression-change, suggesting categorical emotion perception. Overall, this characterization of the detection of subtle changes of facial expression and its temporal dynamics open promising tracks for precise assessment of social perception ability during development and in clinical populations.