Growth of a permanent, valley‐bottom gully from 1964 to 2000 was determined annually from survey and sediment‐discharge data and compared with runoff and baseflow discharges. Data were analysed to test the hypothesis that rates of gully growth decay exponentially with time in response to shrinking catchment area caused by gully enlargement. Also, monthly values of growth rates and runoff, averaged over the 36‐year record, were analysed with mass‐wasting data to determine the extent to which colluvium availability affected growth rates seasonally.

From 1964 to 2000, the gully volume increased by 9200 m3, accounting for 34 per cent of sediment yield from the watershed. There were tight power‐law relationships between annual growth rates and annual runoff, with runoff exponents of 1·57 and 1·30 for headward and volumetric growth, respectively. Increases in gully length, area, and volume were fitted successfully assuming an exponential decay in growth rate with time. Rather than being due to a decrease in catchment area, however, the decline in growth rate was caused by a 77 per cent decrease in the ratio of runoff to baseflow, which also widened the gully and reduced the mean slope of its banks. Order‐of‐magnitude seasonal changes in erosion efficiency, defined as the fraction of stream power used to evacuate sediment from the gully, were roughly correlated with colluvium availability, as indicated by seasonal changes in the number of bank mass‐wasting events. No more than 2·2 per cent of stream power was used to evacuate sediment during any month. This study demonstrates the danger of attributing declining rates of gully growth to a shrinking catchment area if corroborative runoff and baseflow data are not available. Moreover, it illustrates that stream power alone provides only a rough and physically indirect measure of erosion potential.