Erosion plot studies at the Hilton Experimental Site, Shropshire, UK, evaluated the effects of set-aside on runoff, erosion and soil organic matter (SOM) dynamics. Ten runoff plots (slope angles 7-15°) were put to a grass ley in April 1991. Runoff and erosion rates during nine years were low, despite the occurrence of potentially erosive rains. Mean runoff was 0·24 per cent of precipitation (SD = 0·20,n= 89 plot years), compared with a 15-year mean value of 0·13 per cent (SD = 0·04) on permanent (control) grassland. The mean plot erosion rate was 0·21 t ha-1 a-1 (SD = 0·14,n= 69 plot years). Mean SOM content increased consistently and significantly on the set-aside plots, by a mean value of 1·07 per cent by weight (i.e. 1·07 g per 100 g of soil) in ten years, from 2·04 per cent (1991) to 3·11 per cent (2001). Soil erodibility after six years of set-aside (sampling date 24 April 1997) was determined using a drip-screen rainfall simulator. Soil aggregate stability was higher on the grassed soils, compared with set-aside and bare arable soils. Despite no significant (P> 0·05) differences between grassland and set-aside soils, both these treatments were significantly (P< 0·001) different from bare soils. Contrary to the usually strong relationship between SOM and clay content, the percentage silt (r= 0·52,P< 0·001,n= 50) exhibited a stronger correlation with SOM than percentage clay content (r= 0·37,P< 0·01,n= 50). Furthermore, there were equally strong correlations between SOM and cumulative particle size fractions of both clay and silt contents (r= 0·51,P< = 0·001,n= 50). In the absence of significant quantities of clay in these sandy soils, silts assist binding of SOM, which has significance for assessing soil carbon sequestration potential. Finally, future SOM and soil organic carbon benchmark work at the Hilton Site will improve harmonization of global SOM databases and enhance international estimates of rates and potential for soil carbon sequestration.