The relative importance of frequency distortion and changes in time constants in the intelligibility of speech

A recording technique developed by Burnett permitted changes to be made in the relative proportions of intermodulation and harmonic distortion introduced by squaring the input signal. This process was achieved by sending equal parts of the signal through a squaring device, and through a 90-degree phase dividing network followed by a squaring device, and combining the squared outputs. By coupling this procedure with the time-changing features inherent in the sound recording process, we were able to set up a sequence of intellgibility tests whose scores provided solutions to sets of simultaneous equations involving the variables of frequency doubling, frequency halving, and doubling and halving transition times. It was possible to set up an overspecified series of tests, so that more than one solution could be derived from the experiments. Lowering of the characteristic frequency proved to have the most serious effect upon intellgibility. The next most serious was stretching the transition times. Doubling the frequency by second-harmonic distortion was considerably less damaging, though still significant. Reducing the transition times to half their initial values proved relatively trivial.