In this study. short-term headcut retreat was monitored from 46 active bank gullies. selected in the Guadalentin and the Guadix basin in Southeast Spain, The measurements were carried out manually using an orthogonal reference system fixed by erosion pins around the gully heads. between April 1997 and April 1999 with a 1-year interval. The average volumetric retreat rate for all gullies was 4.0 m(3) year(-1), corresponding with an average linear retreat rate of 0.1 m year(-1), but more erosion took place during the first monitored year (1997-1998) compared to the second (1998-1999). An interplay of spatial variations in rainfall distribution and tension crack activity is assumed to be responsible for the important difference in annual headcut retreat. compared to the small difference in annual rainfall amounts, Statistical analysis showed that the present drainage-basin area (A(p)) was the most important topographical factor explaining average gully headcut retreat rate. both in terms of annual eroded volume (V-e) and annual linear retreat (R,), and expressed by the power relationships V-e = 0.04 A(p)(0.38) (R-2 = 0.39) and R-1 = 0.01 A(p)(0.23) (R-2 = 0.39). The V-e-A(p) relationship was compared with the relationship between original drainage-basin area (A(o)) and total eroded bank gully volume (Vol), i.e. Vol = 1.71 A(o)(0.60) (R-2 = 0.65). The importance of runoff generation from a drainage basin is shown by the positive correlation of linear headcut retreat and the runoff curve number (CN), representative for the conditions in the drainage basin, High CN values tend to coincide with higher annual eroded volumes in the relationship between present drainage-basin area (A(p)) and annual eroded volume (V-e). but this affect was not observed in the relationship between original drainage-basin area (A(o)) and total eroded bank gully volume (Vol). Stepwise multiple regression selected the relevant environmental parameters explaining annual eroded volume and linear retreat. In both equations. the present drainage-basin area explained the largest part of the variation. The CN was selected as another common parameter. Height of the headcut was the second most important variable explaining annual eroded volume, indicating the role of energy transfers and undercutting at the headcut. Linear retreat was further explained by the average slope of the present drainage-basin area, representing the effect of decreasing transmission losses and increasing flow velocity with steeper catchment slopes, and by the sand content. decreasing the cohesion of the soil material, promoting soil fall and headcut retreat. Spatial extrapolation of the measured volumetric retreat rate of 4.0 m(3) year(-1) revealed that active bank gully heads contribute up to 6% of the sediment yearly filling up the Puentes reservoir. Estimated gully ages (i) based on the ergodic principle, and (ii) by linear extrapolation of actual gully retreat rates in the past, range between 63 years and 1539 and between 64 and 1720 years, respectively. The high correlation between the gully ages estimated by the two methods is attributed to the fact that most gullies have not reached the evolutionary stage of significantly declining retreat rates.