I'm wondering for how long the voicing persists into closure here, if at all. At about 50% into the closure I would say it's noise only from there on, nevermind F1. But what about the oscillations before that point?

waveform and spectrogram of /p/ in "type of"

2 Answers 2


Your spectrogram is not clear. You should use a narrow-band spectrogram to see formants efficiently. So change your settings. To obtain this visualisation, you should use a small window length as it is indicated here: http://www.fon.hum.uva.nl/praat/manual/Intro_3_2__Configuring_the_spectrogram.html

After that, if it appears harmonics with a high energy, so this segment is voiced. Like in these examples: https://www.krisyu.org/blog/supp-material-cantonese-creak-perception.html


This is probably background noise. To check this, compare a section with no speech. For example, during closure of /p/ you can get very strong apparent "coarticulation" from preceding /a/ where the formants of /a/ are identifiable -- if there is an echo. You can't analyse noise with just a tiny snippet of a spectrogram, you need an actual acoustic wave sample of a full utterance.

  • Thank you. I am afraid that a section with no speech looks different; there is much more irregular noise. Also, there isn't much echo in the recording. And, most importantly, other word-final stop closures usually show some very obvious phonation but for about 20 m, not as in this example. For the reasons mentioned, I think it's some kind of phonation rather than just noise.
    – Grzegorz
    Jul 29, 2019 at 21:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.