A detailed knowledge of Himalayan development is important for our wider understanding of several global processes, ranging from models of plateau uplift to changes in oceanic chemistry and climate1, 2, 3, 4. Continental sediments 55 Myr old found in a foreland basin in Pakistan5 are, by more than 20 Myr, the oldest deposits thought to have been eroded from the Himalayan metamorphic mountain belt. This constraint on when erosion began has influenced models of the timing and diachrony of the India–Eurasia collision6, 7, 8, timing and mechanisms of exhumation9, 10 and uplift11, as well as our general understanding of foreland basin dynamics12. But the depositional age of these basin sediments was based on biostratigraphy from four intercalated marl units5. Here we present dates of 257 detrital grains of white mica from this succession, using the 40Ar–39Ar method, and find that the largest concentration of ages are at 36–40 Myr. These dates are incompatible with the biostratigraphy unless the mineral ages have been reset, a possibility that we reject on the basis of a number of lines of evidence. A more detailed mapping of this formation suggests that the marl units are structurally intercalated with the continental sediments and accordingly that biostratigraphy cannot be used to date the clastic succession. The oldest continental foreland basin sediments containing metamorphic detritus eroded from the Himalaya orogeny therefore seem to be at least 15–20 Myr younger than previously believed, and models based on the older age must be re-evaluated.