A New Pitch Scale

Abstract

The mel scale derived by Stevens and associates has proved extremely useful as an explanatory concept in communications, frequency filtering, and critical band theory. Undoubtedly there exists a need for some such scale, but attempts to reproduce the mel scale have generally foundered on individual differences. Fractionation of frequency does not lend itself readily to standardization. Several years experience has convinced the writer that fractionation of frequency is possible for persons without absolute pitch, but that pitch scales among individuals are so unlike that an “average” scale is not very convincing. Large and variable constant errors are introduced when grossly disparate frequency regions are presented for comparison judgments. A method was used presenting only relatively narrow frequency regions. Given A-B (musical intervals avoided) subject is asked to set a tone C such that B-C = A-B. Subject is then given B-C and asked to set D; and so on up and down the frequency scale. It can be shown that if A-B is wide enough, a mel scale similar to the usual fractionation scale is obtained for any individual; but a critical interval exists furnishing a relatively invariant mel scale which is not changed if the initial interval is made even narrower. Thus the error in the fractionation-mel scale is measured and a new scale furnished freed of this error and much more like among individuals.