The human brain possesses the ability to automatically extract statistical regularities from environmental inputs, including visual-graphic symbols and printed units. However, the specific brain regions underlying the statistical learning of these visual-graphic symbols or artificial orthography remain unclear. This study utilized functional magnetic resonance imaging (fMRI) with an artificial orthography learning paradigm to measure brain activities associated with the statistical learning of radical positional regularities embedded in pseudocharacters containing high (100%), moderate (80%), and low (60%) levels of consistency, along with a series of random abstract figures. Thirty adults passively viewed a continuous stream of these pseudocharacters. fMRI data revealed that the left occipital area and the visual word form area (VWFA) exhibited greater responses at the low consistency level than at the high and moderate levels, suggesting implicit statistical learning of positional regularities. Functional connectivity analysis further revealed significant correlations between the occipital lobe, the VWFA, and other brain regions, such as the middle temporal gyrus (MTG), the superior occipital gyrus (SOG), and the cerebellum. Moreover, neural activity showed a tendency to correlate with behavioral recognition performance. These findings demonstrate that the incidental acquisition of statistical regularities in artificial orthography arises from the coordinated activation of multiple distinct neural circuits.