The accuracy of the body condition scoring for ewes using lumbar palpation was assessed by mean of an experiment conducted under conditions close to those occurring everyday on the farms. Thirteen operators evaluated the body condition score twice in a sample of 48 Lacaune breed dairy ewes, using a semi-quantitative grid (from 0 to 5 points, with a 0.25 point precision). Under the experimental conditions, the repeatability of the method, i.e. the comparison between the scores given by the same operator to the same animals for the two successive scoring notations, was good and even very good (Kappa test), if a 0.25 point precision is tolerated. On the other hand, the reproducibility, i.e. the comparison between the scores given to the same animals by differents operators was poor. Therefore, this method is easy to master for the management of a given flock by a single person. However in order to make comparisons between flocks, in reference to body condition standards set up for different farming systems, this method needs greater precision. Some of the application conditions required for this are discussed.