Voicing in American English stop consonants: An introduction to using speech analysis software
(Thanks to Dirk Elzinga for his help with this)
A. Introduction
(1) English stops, fricatives, and affricates come in pairs: [p/b], [t/d], [k/g], [f/v], [θ/ð], [s/z], [ʃ/ʒ], [tʃ/dʒ].
(2) The difference between the members of each pair is usually described as one of voicing.
(3) Voicing is the periodic vibration of the vocal folds.
(4) However, a closer look at voiced stops in English using instrumental analysis shows that voicing may not be involved.
(5) If voicing isn’t involved, then what is it?
B. Praat
(6) Praat is a software application designed to investigate the acoustic properties of speech. It was developed in the Netherlands by Paul Boersma and David Weenink (praat is the Dutch word for ‘speak’), who have made it freely available to the linguistics community.
(7) Open Praat.
(8) Select Read in the Praat objects dialogue box.
(9) Select Read from file from the drop-down dialogue box. Choose your file by clicking on it. The name of the file should appear in the window in the Praat objects dialogue box.
(10) Select Edit; this will open up the Edit window.
(11) Select Show spectrogram (or make sure that it’s selected). You shouldn’t need to adjust the spectrogram settings.
(12) You can measure durations in the Edit window. Highlight the part of the spectrogram (or waveform) you want to measure and read the number at the bottom of the highlighted portion. It is conventional in work on phonetics to express durations in milliseconds.
C. Using Praat to investigate stop voicing
(13) Let’s start with the file pest.wav. Open it up in the Edit window; it should look something like this:
(14) There are two ways you can see voicing using Praat. The first way involves looking at the spectrogram itself and noticing the regular alternation of light and dark bands. You can also tell Praat to show you the pulses; do this by selecting Show pulses from the Pulses menu. Do this now.
(15) For the word pest, you should see voicing only in the vowel.
(16) Notice that the picture begins with a period of silence; this is the stop [p].
(17) You should also notice that between the release of the stop and the beginning of vowel voicing, there is a period of noise. Measure the duration of this period of noise by placing the cursor at the beginning of the burst and moving it to the beginning of regular voicing.
(18) What is the number that shows up in the selected area? It should be something like 0.068960, or about 69 ms. This is the voice onset time, or the period of time required for the vocal folds to start vibrating after the release of a stop.
(19) Close this window, and open up the file best.wav It should look something like this:
(20) Look at the interval between the stop release and the onset of voicing. It is much smaller than it was for [p] (something like 15 ms).
(21) Now look at the the left side of the spectrogram; you should notice that there is no voicing at all. That is, the [b] looks just like the [p]. So what is the difference between [p] and [b]? Voice onset time.
(22) Now measure the voice onset time for the other pairs of stops ([t/d], [k/g]).
D. Conclusion
(23) Measuring voice onset time is just one of the things that Praat makes very easy. The acoustic analysis of speech is a complex, intricate, but rewarding field of study, and Praat is one of the best tools around for doing it.
Praat Exercise
For each of the stops [p, t, k, b, d, g] state whether they are traditionally classified as voiced or voiceless. Pair up the stops into voiceless/voiced pairs at the same place of articulation.
Using Praat, open the sound files for the words
and look at the spectrograms for each word in turn. Determine by visual inspection whether or not the stops are pronounced with vocal fold vibration and record your observations.
Measure and record the voice onset time for each of the stops in milliseconds (round to the nearest millisecond).
How do you explain the phonetic basis for the distinction between voiced and voiceless stops?