Cosmogenic burial dating using ²⁶Al and ¹⁰Be nuclides is a promising technique for establishing chronologies of fluvial deposits ranging in age from ~0.5 – 5 Ma. Accurate age determinations, however, require an assumption that sediment is buried with a known initial ²⁶Al/¹⁰Be ratio. Here, we evaluate the reliability of burial dating of fluvial terraces along the Yellow River in Lanzhou, China. Rapid and continuous burial of terrace deposits by thick accumulations of windblown loess following abandonment provides deep burial and independent age control. Samples of fluvial gravel from 6 terraces yield burial ages ranging from ~ 0.90 Ma to 3.02 Ma. Comparison of these results to independent ages determined from magnetostratigraphic and loess-paleosol studies demonstrates a high degree of correspondence between ages in the lower three terraces. In the higher terraces, however, burial ages are systematically older than magnetostratigraphic determinations, a result that requires a previous burial history. Our results are consistent with the incision history of the Yellow River, which initiated upstream at ~1.7 Ma following breaching of upstream lake margins; subsequent excavation of basin sediment provided an abundant source of material with low initial ²⁶Al/¹⁰Be ratios. Overall, our results suggest that long-term storage and burial of sediment in continental-scale river systems may be a principal source of error in the application of burial dating.