Estimation of body composition in live cattle is difficult and usually performed using subjective approaches (visual and handling evaluation). Today, there is no reference methodological framework for the standardization of the determination of body fatness in cattle by ultrasonography. The objective of this study was to compare ultrasonographic measurements of fat and muscle subcutaneous tissue thickness at the buttock, lumbar, back and rib level with the conventional body condition scoring method. We performed ultrasonography on 160 cows from 5 different breeds (Normande, Holstein, Montbeliarde, Charolaise, Salers) by selecting populations of animals with the widest range of body condition scores possible (from 0.5 to 4.5 on a 0-5 scale). The study was carried out with four experimenters working with two different ultrasound devices and repeating the measurements two times. Low variation between assessors was observed at the lumbar and the rib sites. Measurements performed on dairy cows allow good correlation between body scores and ultrasound indicators at the buttock (fat, r=0.709), lumbar (fat and muscle, r=0.829) and rib (fat and muscle, r=0.707)) sites. The best result was obtained with the measurement of fat thickness + muscle at the lumbar site on the Normandy cows (R=0.856, P<0.001). A combination of these ultrasound indicators provides an accurate and objective assessment of body composition on dairy cows. However, no relationship was found between body scores and ultrasound indicators on beef cows. Ultrasonography is an easy, objective and non-invasive way to determine body composition in dairy cows. This work is offering an opportunity to standardize this technique and use it routinely on a large number of animals in the field.