Repetition suppression (RS) refers to the reduction of neuronal responses to repeated stimuli as compared to non-repeated stimuli. The predictive coding account of RS proposes that its magnitude is modulated by repetition probability (P(rep)) and that this modulation increases with prior experience with the stimulus category. To test these hypotheses, we examined the RS and it’s modulation by P(rep) for three stimulus categories for which participants had different expertise, using EEG methodology. Cantonese speakers watched paired stimuli (S1-S2) of Asian faces, Chinese written words, and animal pictures with the S2 being the same or different from S1. Attributes of S1 (e.g., the sex of the first face) served as a cue for repetition probability of S2. Time-point by time-point Topographic Analyses of Variance (TANOVA) for words showed significant repetition effects across several intervals (92-140, 150-248, 260-488, and 502-560ms), and expectation effects during 789-844ms. Significant repetition effects were also identified for faces (207-358ms), and animals (324-486ms). Timing and topographies suggest N250r effects for all three stimulus categories, but TANOVA comparisons indicate earlier and distinct topographic distributions of repetition effects for words versus faces (151-263, 277-445m) and animals (148-242, 266-437ms), and for faces versus animals (209-316ms). These results suggest that repetition effects differ between stimulus categories, presumably depending on prior experience and stimulus properties, such as spatial frequencies. Importantly, we did not find any EEG evidence for effects of P(rep) potentially manipulating expectancy. Such null findings of P(rep) effects do not support the general predictive coding account of repetition suppression.