The deep dullness of heart and superior mediastinal vessels was percussed on 110 male university students (erect and recumbent) by one physician, wire markers were fastened to their chests, and teleoradiographic films were taken. All results were analyzed statistically. In these students the mean errors (difference between percussion and x-ray borders) were less than 1 cm. in most intercostal spaces; but much greater allowance for possible error had to be made in percussing any one individual—for example, a range of over 3.5 cm. in the fourth left intercostal space in erect students. The differences in error between erect and recumbent positions could be largely attributed to dispersion of x-rays. For practical purposes the variation between two percussions on the same student was as great as between different students because the correlation between the errors in repeated percussion of the same students, although significant, was low. From one-third to one-half of the variation in the cardiac (but not the superior mediastinal) region was attributable to diaphragm movement, owing to the impossibility of securing the same midphase of respiration on any two occasions. The risk of wide error in any individual percussion could be greatly lessened by taking the average of several independent readings, thereby reducing the effect of diaphragm movement. Stature, weight, and chest size or shape had no important relationship to the percussion error. Ten students, taken at random, sufficed to show the difference in error of two experienced physicians. Since each observer should know his own percussion error, the amount of information obtainable from a small series of films is illustrated.