Figure 1.

The left panel displays a 110-ms steady-state portion of the vowel /ae/ produced by a 21-year-old male speaker saying the word bat, with a 35-ms analysis segment of the vowel marked by two vertical cursors on the time domain (as displayed in TF32). The top portion (A) is an amplitude by time waveform, and the bottom portion (B) its frequency by time spectrogram. The right panel (C) displays the amplitude by frequency spectral slice (overlay of linear predictive coding spectrum on fast Fourier transform spectrum) of the marked analysis segment of the vowel, with vertical gridline markings in 1000-Hz intervals, and horizontal lines in 10-dB increments. The frequency of the white vertical cursor on the spectra (C), correspond to the frequency of the white horizontal cursor on the spectrogram (B).