The effectiveness of tropical grass species in strips of different length in trapping sediment from cropland was assessed, and the influence of filter length was determined. The assessment was made under natural rainfall which induced sheet and rill erosion in run‐off plots and then using simulated run‐off which caused concentrated erosion. The evaluated grasses were elephant grass, lemon grass, paspalum and sugarcane. Run‐off plots were on a 10% slope in a randomized complete block design replicated three times. Filter lengths were 2.5, 5 and 10 m against a 10‐m‐long sediment source area planted with maize on a clay loam soil. The results show that sediment trapping effectiveness (TE) increases nonlinearly with increasing filter length for all grasses. Under natural rainfall, more than 70% of sediment was trapped in the first 5 m, and lengthening the strip to 10 m only resulted in a marginal increase in TE. With concentrated run‐off, more than 70% of sediment was trapped in the first 5 m and lengthening the strip to 10 m resulted in a significant increase in TE. Paspalum and lemon grass performed significantly better than other grasses (P < 0.05), owing to their spreading growth pattern over the soil surface. Paspalum also has the highest root density in the upper 0.3‐m layer of the soil followed by lemon grass, hence offering the greatest resistance to erosion from concentrated flow. The results demonstrate that tropical grass filter strips provide a viable means for reducing the sediment flux from cropland.