Catalytic Factors in Speech Perception

Abstract

Speech from which all formant information except the second formant frequency has been removed by filtering and infinite clipping is still highly intelligible. Spectrograms of such speech exhibit a normal formant band and higher “harmonic” bands representing clipping distortion. This spectral configuration was approximated by use of a multivibrator whose period was controlled by the output of a second formant tracker. A voice operated gate was interposed between the multivibrator and the output terminals of the device. The resulting signal, although possessing a spectrogram closely comparable to that of the filtered-clipped speech already mentioned, is not at all speechlike and has zero intelligibility. Addition of random noise or a signal repetitive at typical glottal rates to the multivibrator control signal renders the signal speechlike and somewhat intelligible. It appears that the perturbations in individual multivibrator periods caused by the noise or repetitive signal are catalytic factors which enable this synethetic speech to be perceived as such. [Work supported by NASA.]