The graphical representation of the waveform from a section of speech can show amplitude (volume), and, to a lesser extent, periodic aspects of the sound. But speech sounds are made up of many frequencies, and so it is typically useful to perform spectral analysis to produce a representation of which frequencies are present. Roach shows several examples where the waveform is displayed above the spectogram.

Using wave and spectogram, it is possible to see that (from Roach p43-46):

• vowel sounds are typically periodic (repeating pattern) and have a few formant frequencies (like notes in a chord) which differ depending on the vowel. It turns out that "the formant with the lowest frequency (formant 1) corresponds roughly to the traditional open/close dimension of vowels...Formant 2...corresponds roughly to the front/back dimension of vowels" (Roach p43).
• voiceless fricatives (/s/, /f/) are aperiodic, don't have formants. They are made up of a large number of frequencies (bit like white or pink noise)
• voiced fricatives (/z/) have periodic and aperiodic sounds
• plosives are characterised by a period of silence (shown as zero amplitude wave) followed by a transient aperiodic sound and then (if they are voiced) by a periodic sound
• nasals are similar to vowels but have less energy and harder to identify formants (the signal is compromised as a result of having to come through the nose)
• approximants are similar to vowels

Useful book readings for this lecture:

• Roach: "Phonetics", ch. 6
• Fromkin: "An Introduction to Language" pp 399-401
• Crystal: "Encyclopedia of Language" ch. 23,24