Perceiving emotional expressions from faces is essential for social interactions. However, while the intensity of facial expressions is highly variable in everyday-life, perceptual tuning to facial expression of various intensities remains largely unknown. Here we recorded scalp electroencephalogram (EEG) from 18 adult participants during fast periodic visual stimulation (FPVS): 100-sec sequences presenting an individual neutral face at a rapid rate of 6 Hz. Expressive faces displaying anger, disgust, fear, happiness, or sadness (one emotion per sequence) were introduced every five faces (i.e., at an oddball rate of 1.2 Hz). During each sequence, the intensity of the 1.2 Hz expression-change increased every 20 secs in 20% steps. A frequency-domain analysis of the whole EEG epoch showed a medial-occipital response at the 6 Hz base rate and its harmonics. The 6 Hz amplitude decreased with increasing intensity steps (i.e., along the sequence), indicating adaptation to low- (e.g., luminance) and/or high-level (e.g., individual face) visual cues that repeated at 6 Hz. In contrast, and most importantly, the 1.2 Hz oddball frequency and its harmonics revealed an expression-change specific activity already at 40% of intensity for all facial expressions except sadness, and increasing in amplitude with increasing intensity steps mainly over right occipito-temporal sites. Interestingly, a time-domain analysis revealed that early expression-change specific activities (peaking at 140 and 250 ms after stimulus-onset) linearly increase with intensity steps whereas a later response (peaking at 370 ms) shows a more abrupt sensitivity to expression-change intensity, suggesting early gradual expression-change detection followed by categorical facial expression perception. Overall, thanks to FPVS-EEG, we isolate objective markers of sensitivity to changes of facial expression and provide new insights into the temporal dynamics and tuning functions of this critical visual categorization ability for efficient social functioning.