AbstractThis study compares landmark location errors in cephalometric radiography (when re‐measuring radiographs) and radiographic errors (when retaking the radiograph). The samples comprised 32 remeasured and re‐digitized radiographs and a further series of 22 retaken radiographs drawn from the same overall sample of 12‐year‐old Chinese children in Hong Kong. All radiographs were recorded in natural head posture with the lips in light contact.The relative size of the errors were assessed by calculating the ‘error percentage’ for both selected dento‐skeletal and soft tissue profile measures. It is suggested that this index is clinically more meaningful than the usually used ‘method error’. It expresses the variance of the method error (me2) as a percentage of the variance of the measurement under study (standard deviation2).This error percentage was found to be doubled, on average, for measurements on the retaken radiographs. In general, measures with most landmarks in the mid‐sagittal plane showed the least increase in percentage error. The results suggest that errors arising from retaking cephalometric radiographs may effectively be greater than those usually reported. The Frankfort plane, the functional occlusal plane and the incisor long axes displayed poor reproducibility. For the soft tissues the lips served as poor angular landmarks and relatively acute angles with short ‘arms’ also