The existence of fmax has been perennially observed in various earthquakes. However, whether fmax is caused by source processes (Papageorgiou and Aki, 1983) or site effects (Hanks, 1982) remains a controversial issue. With abundant near‐fault strong ground‐motion records, the Wenchuan earthquake provides an opportunity to validate the issue. By fitting the acceleration spectra of 10 seismic stations in the proximity of the fault on which the mainshock occurred, a set of fmax values is obtained along the fault. The most notable feature of this set is its V‐shaped distribution. The minimal fmax value is recorded at the 051PXZ station near the epicenter, and fmax increases as the distance deviates farther from the station. The maximal fmax is recorded at the most northern station with respect to the epicenter. We estimate the cohesive zone size of the ruptured fault from the observed fmax based on the source–cause model by Papageorgiou and Aki (1983). The sizes are remarkably consistent with the widths of the fault zone obtained from a totally independent study of the fault zone trapped waves (Li et al., 2009; Li et al., 2010). Therefore, the present study provides strong support to Papageorgiou and Aki’s source–cause model for fmax. Moreover, the V‐shaped nonuniform distribution of fmax indicates a complicated rupture process of the Wenchuan earthquake, wherein the most energetic thrust faulting produces a thicker fault zone (approximately 400 m) around the epicenter, and an exhausted strike‐slip faulting produces a thinner fault zone at the end portion of the fault (170–200 m).