Cephalometric landmark validity (the difference between the estimated landmark and the true landmark) has surprisingly not previously been comprehensively evaluated, and no previous study has examined the validity of cephalometric angles and distances. The aim of this study was to investigate the validity of 15 commonly used skeletal and dental cephalometric landmarks, and the subsequent effects on 17 angles and distances. Small steel balls were glued on to 30 Chinese dry skulls to represent the true anatomical landmarks. The skulls were mounted in a purpose-designed skull holder and two cephalograms recorded of each skull, one with and one without the steel balls on the landmarks. Validity was expressed as the difference in the measurements between the assessments made with and without the steel ball markers. Measurements were made relative to X and Y co-ordinates which were constructed from reference points (steel balls) glued intracranially to the skulls. Seven out of the 10 skeletal landmarks and all five dental landmarks, were found to be non-valid along the X or the Y axes (P < 0.05). The standard deviations of the validity errors were large, being 1.0-2.5 mm, along at least one axis, for eight of the skeletal landmarks and three of the dental landmarks. Four of the cephalometric angles (SNA, SN/MnP, MxP/MnP, and LI/MnP) and three of the distances (N-Me, MxP-Me, and lower incisor edge to APg) were also found to be invalid (P < 0.05). The validity errors were greater for angles involving dental landmarks and for angles dependent on four landmarks compared to those dependent on three. The standard deviations of the validity errors for the skeletal angles ranged from 0.9 to 1.8 degrees, except for ANB (0.4 degrees), and for the dental angles from 3.2 to 5.8 degrees.