The purpose of this experiment was to develop and test a new technique for training perfect pitch. A subject was seated in a sound attenuated booth and presented with taped piano tones by means of headphones. He was required to respond by lighting a corresponding note on an enlarged metal musical staff by means of a button on a control panel in front of him. A session consisted of the presentation of 24 piano tones ranging from middle C to high B, two octaves above, and presented randomly to the subject. Five subjects were used in the experiment, three of whom were run for 36 training sessions. At the end of training, each subject was given a post-training test to measure his improvement in perfect pitch naming. All subjects made significant gains in pitch naming accuracy during the training period. The administration of a post-training test showed one subject meeting the criterion set for perfect pitch. A second subject missed it by one note. The other subjects also showed marked improvement in their pitch naming ability. A decrease in the degree of half-step errors also resulted from training for all subjects. This decrease in error (or improvement in pitch naming) was most marked for the poorer subjects but then it is obvious that they had more latitude for improvement. The complete error analysis of one subject was run. This showed that natural tones (white keys on the piano) were more accurately named than incidental tones (black keys). Further, the tones in the first octave (middle C range) were more accurately named than the notes in the octave above.