A study was designed to compare use of an numerical rating scale (NRS) and a visual analogue scale (VAS) for subjective assessment of lameness, using sheep as a model. The NRS consisted of 5 divisions, 0, 1, 2, 3, and 4; 4 of these divisions (1-4) described lameness. The VAS used a 100-mm horizontal line with vertical bars at either end; one end was labeled 'sound' and the other was labeled 'could not be more lame.' Two independent observers graded lameness in 62 sheep, and between- and within-observer differences were assessed for each scoring system to compare the NRS with the VAS. Results indicated no significant differences between the 2 observers scoring lameness, using either the VAS or the NRS. The scores obtained, using the VAS, were not normally distributed, although differences between scores for the 2 observers were. The NRS scores followed a normal distribution pattern. Investigation of repeated measurement for the same sheep, using both scales, revealed no significant difference between either. A comparison of the NRS and VAS scores made by each observer indicated that although correlation was good (observer 1; r = 0.94; observer 2; r = 0.95), there was not perfect agreement. The maximal NRS score of 4 was associated with VAS values > 68 mm, indicating that the NRS divisions did not reflect equal increases in lameness. The VAS and NRS scores for each observer were highly reproducible, although they were more variable for sheep that were regarded as moderately lame. Results indicate that although the NRS and VAS compared favorably with respect to repeatability, reproducibility, and use by 2 observers, the VAS is inherently more sensitive. In addition, the NRS and VAS should not be used interchangeably.