The methods currently used to evaluate cultivator weed control are imprecise and often biased. A new method was tested. An experiment was carried out at two sites in 1993 and again in 1994, using a split-plot design with four repetitions. Cultivation was carried out on weed populations simulated with rye grass (Lolium spp L.) and mustard (Sinapis alba L.) in 1993 and on both simulated and natural weed population in 1994 with Rabe Werk, Hiniker and Kongskilde cultivators. Weed control was assessed by the effective weed kill and the percentage of the area of the simulated weed population which was affected, and by counts and visual observations on the natural weed population. The new method provided an objective assessment of the cultivation treatments, especially by the effective weed kill index. The area affected approach was not suitable for all cultivation treatments. There was agreement between the proposed method and the conventional methods. This validated the proposed new method. Overall, the Hiniker with ridgers performed the best. The Hiniker without ridgers, and the Kongskilde with or without ridgers performed similarly. The Rabe Werk without goosefoot attachment resulted in the poorest weed control. The degree of weed displacement varied among the cultivators and was associated with weed density. The degree of weed displacement onto the crop row from the inter-row area was low ($<$1% in 1993 and up to 3.4 in 1994). The weed displacement distance was related to weed density and the degree of weed uprooting produced by cultivation.

No description available.