Data collection and reduction procedures, coherently structured in protocols, are necessary in gait analysis to make kinematic and kinetic measurements clinically comprehensible. The current protocols differ considerably for the marker-set and for the biomechanical model implemented. Nevertheless, conventional gait variables are compared without full awareness of these differences. A comparison was made of five worldwide representative protocols by analysing kinematics and kinetics of the trunk, pelvis and lower limbs exactly over the same gait cycles. A single comprehensive arrangement of markers was defined by merging the corresponding five marker-sets. This resulted in 60 markers to be positioned either on the skin or on wands, and in 16 anatomical landmark calibrations to be performed with an instrumented pointer. Two healthy subjects and one patient who had a special two degrees of freedom knee prosthesis implanted were analysed. Data from up-right posture and at least three gait repetitions were collected. Five corresponding experts participated in the data collection and analysed independently the data according to their own procedures. All five protocols showed good intra-protocol repeatability. Joint flexion/extension showed good correlations and a small bias among protocols. Out-of-sagittal plane rotations revealed worse correlations, and in particular knee abduction/adduction had opposite trends. Joint moments compared well, despite the very different methods implemented. The abduction/adduction at the prosthetic knee, which was fully restrained, revealed an erroneous rotation as large as 30 degrees in one protocol. Higher correlations were observed between the protocols with similar biomechanical models, whereas little influence seems to be ascribed to the marker-set.