Observer error in chest radiography is of such magnitude as to create special problems in diagnosis and treatment. A review of recent studies on the subject reveals the following: 1. 1. In interpreting a set of photofluorograms for presence or absence of a lesion a competent and experienced physician misses, in a single reading, as many as 32.2 per cent of the positive films and provides false positive diagnosis on 1.7 per cent of the negative films. 2. 2. Classification of pulmonary lesions on the basis of roentgenographic appearance is not very reliable. It is not possible to determine reliably the probable activity of a lesion from a single roentgenogram. 3. 3. In judging a pair of serial roentgenograms for evidence of progression, regression or stability of disease, two competent and experienced physicians are likely to disagree with each other in nearly one-third of the cases, and a single reader is likely to disagree with himself in about one-fifth of the pairs. 4. 4. The undesired effects of false diagnosis may be reduced considerably by the application of dual reading. In mass radiographic projects this procedure detects one-third of the lesions missed by a single reader. 5. 5. It is suggested that dual reading may be useful in clinical radiology. The main contribution results from the fact that the procedure identifies about two-thirds of the roentgenograms for which the diagnosis can be given with confidence. The remaining may be considered as roentgenologically indeterminate until further evidence is obtained.