Field runo plots were established in 1984 to evaluate the eects of slope length on runo, soil erosion and crop yields on newly cleared land for four consecutive years (1984±1987) on an Al®sol at Ibadan, Nigeria. The experimental treatments involved six slope lengths (60 m to 10 m at 10-m increments) and two tillage methods (plough-based conventional tillage and a herbicide-based no-till method) of seedbed preparation. A uniform crop rotation of maize (Zea mays)/cowpeas (Vigna unguiculata) was adopted for all four years. An uncropped and ploughed plot of 25 m length was used as a control. The water runo from the conventional tillage treatment was not signi®cantly aected by slope length, but runo from the no-till treatment signi®cantly increased with a decrease in slope length. The average runo from the no-till treatment was 1 . 85 per cent of rainfall for 60 m, 2 . 25 per cent for 40 m, 2 . 95 per cent for 30 m, 4 . 7 per cent for 20 m and 5 . 15 per cent for 10 m slope length. In contrast to runo, soil erosion in the conventional tillage treatment decreased signi®cantly with a decrease in slope length. For conventional tillage, the average soil erosion was 9 . 59 Mg ha À1 for 60 m, 9 . 88 Mg ha À1 for 50 m, 6 . 84 Mg ha À1 for 40 m, 5 . 69 Mg ha À1 for 30 m, 1 . 27 Mg ha À1 for 20 m and 2 . 19 Mg ha À1 for 10 m slope length. Because the no-till method was extremely eective in reducing soil erosion, there were no de®nite trends in erosion with regard to slope length. The average sediment load (erosion:runo ratio) also decreased with a decrease in slope length from 66 . 3 kg ha À1 mm À1 for 60 m to 36 . 3 kg ha À1 mm À1 for 10 m slope length. The mean C factor (ratio of soil erosion from cropped land to uncropped control) also decreased with a decrease in slope length. Similarly, the erosion:crop yield ratio decreased with a decrease in slope length, and the relative decrease was more drastic in conventional tillage than in the no-till treatment. The slope length (L) and erosion relationship ®ts a polynomial function (Y c aL bL 2 ). Formulae are proposed for computing the optimum terrace spacing in relation to slope gradient and tillage method. # 1997