A three-part study was designed in order to test the applicability of the fluid-level method for registration of natural head position. In the first part the fluid-level method was utilized to make two repeated cephalometric radiographs of 33 young adults. The reproducibility of the craniovertical, craniocervical and cervicohorizontal relationships was comparable with previous results with the mirror method. In the second part, the fluid-level method was compared with the mirror method when used by two radiographers for repeated radiographs in 40 subjects. The reproducibility of the craniocervical and cervicohorizontal angles was less accurate with both methods and the only difference between them was a slightly better reproducibility of the craniovertical angle with the fluid-level method for one of the radiographers. In the third part, it was shown that using the fluid level, a patient can be transferred from a standing to a sitting position in the cephalostat without any systematic change in craniovertical, craniocervical or cervicohorizontal relationships.