It is a great advantage and satisfaction to work from curves taken by the phonograph itself, because it is then possible to reproduce and verify by actual hearing the sound supposed to be represented by given curves. This ought always to be done, even with the phonograph, and the slightest change of timbre should be carefully noted; for the given curve represents not really the original sound, but the sound given back by the phonograph. Every plate or membrane vibrates more readily and strongly to some pitches than to others, and in a composite sound it is almost inevitable that some of its elements are recorded with more, and others with less, than their original relative force. Different vibrators hence yield quite different-looking phonograms for the same sound. In vowels it seems fortunately to happen that considerable changes in the relative force of the component vibrations may occur without altering the vowel. It is the pitch of the components which matters; and that is not altered by an efficient vibrator. Assuming a knowledge of the methods by which the record dug vertically into the phonographic cylinder is displayed horizontally for examination and analysis, we note that a satisfactory Fourierian analysis demands a sufficient number of measurements.