Automatic and continuously recording samplers are deployed in a Hertfordshire gravel-bed stream to show that bedload transport is related to stream power. The pattern is similar to that already established for North American channels but, because the record is so detailed, it is possible to identify the cause of the considerable scatter that is normal in such relationships. A major factor is the Occurrence of rhythmic pulses in bedload discharge that are not matched by similar fluctuations in hydraulic variables. It is suggested that these pulses reflect downstream differences in the concentration of mobile particles in a slow-moving traction carpet, and that they may be likened to kinematic waves. The record also reveals that the threshold of sediment transport-always presumed hithero to be associated with incipient motion-is related to thecessation of bedload transport in a river flood. Indeed, the mean value of stream power at the finish of bedload transport is only 20 per cent of that prevailing at the moment of incipient sediment motion. Because of this, there is an inevitably poor correlation between actual bedload transport rates and those predicted by bedload equations which rely upon a single traction threshold. These new data show that the general inverse relationship between bedload discharge and waterdepth: grain-size ratio proposed by Bagnold (1977Bagnold ( ,1980) ) is not universal. Transport efficiency for this gravel-bed stream is typically 0.05 per cent of available stream power, which compares with 1.6 per cent for a river moving both gravel and sand, and 5 per cent for another channel where bedload is composed predominantly of sand-sized particles. It is argued that coarse and fine-grained alluvial channels may need to be considered separately. By allowing for differences in traction threshold at the beginning and end of bedload events, and by averaging bedload discharge flood by flood in order to smooth out the effect of pulses, it is possible to.achieve a reasonably good prediction of average bedload transport rate in terms of stream power.