We aimed to devise a simple and reproducible method of shoulder strength measurement with reference to the Constant and Murley shoulder score. Fifteen male and 15 female patients were tested. The average age was 32 years. None had previous shoulder pathology. We used a simple spring balance and an electronic load cell to measure force. The system was attached to the floor and connected to the patients arm with an adjustable strap. Strength was measured at 90 degrees of flexion and abduction (in the plane of the scapula), with both dominant and non-dominant limbs. The strap was placed on the upper arm a fixed distance from the tip of the acromion. In addition measurements were taken with the strap attached at the patients wrist. Measurements were repeated after a rest period. There was no significant difference between strength measured in abduction or flexion or between dominant and non-dominant limbs. The mean measurement for females (13.38kg) is 56 of the male mean (23.92kg). The mean measurement with the strap at the wrist was 47.6 of the mean measurement with the strap on the arm for females and 56.25 for males. In their original description Constant and Murley use the method described by Moseley to measure shoulder strength. This involved the examiner pulling down on a spring balance held at arms length by the patient. The value assigned to the scale (0.5 kg scores one point to a maximum of 25) is arbitrary. By placing the strap on the upper arm the lever arm is shortened and a higher strength measurement is obtained, the position is standardised by measuring a fixed distance from the tip of the acromion. By adjusting the scoring scale accordingly a greater margin for error and thus greater accuracy and reproducibility can be obtained. In addition there is a large difference in mean strength between males and females that is not taken into account in the Constant and Murley score. A simple multiplier could be used to correct this factor. We suggest minor modifications to strength measurement for the Constant and Murley score could improve accuracy and reproducibility and correct scoring differences between the sexes.