Discrimination of the musical octave was measured by the method of constant stimuli with roving standards and four response categories. For each of three musically trained listners, and two of three untrained listeners, the estimated value of the subjective octave was ’’stretched’’ slightly sharper than the physical octave of 1200 cents. The magnitude of this stretch, about 20 cents on the average, replicated earlier findings obtained by the method of adjustment. A Thurstonian decision-theory model, providing iterative parametric solutions, generated an excellent fit to listeners’ psychophysical functions. The present experimental task is best described by a noncategorical model of interval perception.