BACKGROUND AND PURPOSE: Although the assessment of cervical motion is routinely performed in clinical practice, no standard protocol for this procedure has ever been established formally. The specific aim of the present study was to select from four different measurement protocols the one which was most stable in terms of reproducibility and was appropriate for clinical and/or medicolegal applications. METHOD: A repeated measurement, test-retest of cervical motion study design using an ultrasound-based system for three-dimensional (3D) motion analysis; cervical range of motion was measured along the six primary directions: flexion; extension; right and left rotation; and right and left lateral flexion, in 20 healthy subjects who were tested twice over a period of lasting from one to four weeks. 'Protocol A' (reciprocal--intermittent testing) consisted of moving the head along a given primary direction, return to the neutral position, a pause and then motion to the opposite primary direction and return to neutral position. These movements were repeated three times. 'Protocol B' (reciprocal--continuous testing) was identical to Protocol A, but without the pause between the primary directions. 'Protocol C' consisted of three repetitions of the same primary direction with a break between two consecutive primary directions. Three sets of six randomly ordered primary directions constituted 'Protocol D'. RESULTS: Protocol D was associated with a significantly smaller range of motion and with the least intra-test reproducibility, as indicated by the coefficient of variation. The differences between the other protocols were largely negligible. CONCLUSION: In routine clinical practice, either of protocols A, B or C may be applied.