The purpose of this study was to compare the shoulder and elbow joint loads during the tennis serve. Two synchronised 200 Hz video cameras were used to record the service action of 20 male and female players at the Sydney 2000 Olympics. The displacement histories of 20 selected landmarks, were calculated using the direct linear transformation approach. Ball speed was recorded from the stadium radar gun. The Peak Motus system was used to smooth displacements, while a customised inverse-dynamics program was used to calculate 3D shoulder and elbow joint kinematics and kinetics. Male players, who recorded significantly higher service speeds (male=183 kin hr(-1): female=149 kin hr(-1)) recorded significantly higher normalised and absolute internal rotation shoulder torque at the position when the arm was maximally externally rotated (MER) (male=4.6% and 64.9 Nm: female=3.5% and 37.5 Nm). A higher absolute elbow varus torque (67.6 Nm) was also recorded at MER, when compared with the female players (41.3 Nm). Peak normalised horizontal adduction torque (male=7.6%: female=6.5%), normalised shoulder compressive force (male=79.6%: female=59.1%) and absolute compressive force (male=608.3 N: female=363.7 N), were higher for the male players. Players who flexed at the front knee by 7.6degrees, in the backswing phase of the serve, recorded a similar speed (162 kin hr-1), and an increased normalised internal rotation torque at MER (5.0%), when compared with those who flexed by 14.7degrees They also recorded a larger normalised varus torque at MER (5.3% v 3.9%) and peak value (6.3% v 5.2%). Players who recorded a larger knee flexion also recorded less normalised and absolute (4.3%, 55.6 Nm) peak internal rotation torque compared with those with less flexion (5.6%, 63.9 Nm). Those players who used an abbreviated backswing were able to serve with a similar speed and recorded similar kinetic values. Loading on the shoulder and elbow joints is higher for the male than female players, which is a reason for the significantly higher service speed by the males. The higher kinetic measures for the group with the lower knee flexion means that all players should be encouraged to flex their knees during the backswing phase of the service action. The type of backswing was shown to have minimal influence on service velocity or loading of the shoulder and elbow joints.