The problem of the timbre of the speaking voice has been studied by various investigators. However, little extended work has been done with the singing voice, which it appears is considerably different from the normal speaking voice. The key to a much‐needed revision and “codification” of many methods of teaching voice lies in a precise knowledge of what physiologic and acoustic phenomena are taking place in “good” and in “poor” tone‐production. This paper summarizes the results of a three year investigation to determine the ideal frequency characteristics of good voices, as a first step in laying the foundation for a scientific pedagogy of voice. Nearly a thousand oscillograms have been taken, each long enough to include at least one complete cycle of the voice‐vibrato. They comprise various qualities from best to poorest from over 40 male and female voices of all degrees of training, at frequent points throughout the pitch series, on various vowels, and at various intensities. A study of these records, including harmonic analysis to 30 components of a selected group of 46 wave forms, brings to light four definite characteristics of good voice‐quality, which are discussed together with their physiologic implications.