Many roentgenographic signs have been described (2,4,8, 9, 11, 13, 18) which are considered characteristic of specific congenital heart abnormalities, but no single study has examined the frequencies with which each of these signs is distributed over a broad spectrum of congenital defects. This study evaluates the occurrence of 20 roentgenographic parameters as found in 9 uncomplicated congenital heart lesions. The 9 lesions include secundum type of interatrial septal defect (IASD), primum type of interatrial septal defect, interventricular septal defect (IVSD), patent ductus arteriosus (PDA), tertralogy of Fallot, coarctation of the aorta, pulmonary valve stenosis, aortic valve stenosis, and complete transposition of the great vessels. These defects comprise about 87 per cent of all congenital heart deformities (10). An accurate evaluation of the 20 roentgenographic parameters and their occurrence in different congenital heart lesions required a rapid and uniformly consistent method of handling data. The digital computer was utilized to compile and analyze over 3,000 bits of information collected during this study. The computer analysis has included absolute and relative frequency distributions for the various radiographic signs. From this data a frequency distribution matrix (Fig. 3) has been constructed from which the computer, using Bayes' theorem of inverse probability (14, 22), or a radiologist by visual inspection, can diagnose one of the 9 different cardiac lesions. A discussion of the Bayesian method as applied to computer diagnosis is presented. Computers are finding increased and varied applications in radiology (5, 15–17, 20, 23). But before a computer can be intelligently programed to perform a task or solve a problem, the radiologist must objectively identify and methodically analyze all the roentgenographic information used in his decision-making processes. Then the computer can evaluate this information statistically and determine its validity. The results obtained may cause the radiologist to reassess and reorganize his pattern recognition. Thus, not only is the computer fast and consistent, but the exacting, systematic manner of film analysis imposed is an excellent tool for teaching and research. Once instructed, the computer will go completely through the programed sequence and consider each roentgenographic parameter in relation to every diagnosis for which it has information; it is tireless and never has mental lapses. Material and Methods A total of 231 cases have been reviewed. Figure 2, A shows the number of cases evaluated for each of the 9 different “pure” congenital heart defects. The lack of sufficient case material necessitated omission of the rarer cardiac lesions from the study at this time. Every case has been proved at autopsy, surgery, or by cardiac catheterization and angiography. The different roentgenographic parameters recorded for each patient are illustrated by the worksheet in Figure 1.