AbstractAcontinuousrecord reveals that the incidence of bedload in a coarse‐grained river channel changes from flood to flood. Long periods of inactivity encourage the channel bed to consolidate sufficiently so that bedload is largely confined to the recession limb of the next flood‐wave. But when floods follow each other closely, the bed material is comparatively loose and offers less resistance to entrainment. In this case, substantial amounts of bedload are generated on the rising limb. This is confirmed by values of bed shear stress or stream power at the threshold of initial motion which can be up to five times the overall mean in the case of isolated floods or those which are the first of the season. This produces a complicated relationship between flow parameters and bedload and explains some of the difficulties in establishing bedload rating curves for coarse‐grained channels. Besides this, the threshold ofinitialmotion is shown to occur at levels of bed shear stress three times those at the thresholds offinalmotion. This adds further confusion to attempts at developing predictive bedload equations and clearly indicates at least one reason why equations currently in use are unsatisfactory. Bedload is shown to be characterized by a series of pulses with a mean periodicity of 1.7 hours. In the absence of migrating bedforms, it is speculated that this well‐documented pattern reflects the passage of kinematic waves of particles in a slow‐moving traction carpet. The general pattern of bedload, including pulsations, is shown to occur more or less synchronously at different points across the strea