The purpose of the current study was to examine the effect of different reference frames on lower limb joint moments during gait with a view to identifying a recommended convention for clinical purposes. Data were collected from 10 subjects (nine able-bodied adults and one child with diplegic cerebral palsy) whilst walking at a self-selected speed. Calculations utilised a three-dimensional inverse dynamics model. For each joint, moments were expressed in four different reference frames. Three of the frames were orthogonal: laboratory frame; anatomical frame of proximal segment; anatomical frame of distal segment. The fourth reference frame was a non-orthogonal joint coordinate system (JCS). Significant differences in the joint moment profiles during gait were found with alternative references frames. This was apparent primarily for the transverse plane joint moments for able-bodied adult gait. For the pathological gait pattern presented, there were also marked differences in the frontal plane joint moments and more subtle ones in the sagittal plane. Whilst it is recognised that all possible reference frames for the expression of the net moment vector are mathematically valid, a decision needs to be made as to which one is more clinically useful. It is proposed that the non-orthogonal JCS is most logical on the basis of what, biomechanically, the joint moment actually represents.