I (think that I) understand that sonorants are sonorous because airflow is continuous, non-turbulent, and so resonate at such a frequency that the vocal tract acoustic filter amplifies them. And I understand the voiceless obstruents are not sonorous, because the occlusion in the vocal tract impedes airflow/makes it turbulence, with noise being made by air pressure fluctuations rather than resonance.

What I don't understand is how voicing works with obstruents. They involve trilling action at the vocal folds, yes? But the occlusion in the vocal tract means they aren't sonorous. So would the waveform of a voiced obstruent be (quasi)periodic, or aperiodic? Do they have a frequency, and do they resonate? Where does the majority of the noise come from: the build up of air pressure, or the vibration of the vocal folds?

I'm probably wrong on much of the above, so if someone could set me straight on the whole thing, it would be much appreciated. Thanks!

  • 1
    In the SPE account, voiced obstruents do not "involve trilling action at the vocal folds". Voiceless sounds are those with the vocal folds spread sufficiently to prevent spontaneous voicing, and sounds that are not voiceless are voiced. This, for example, makes glottal stops voiced. There is no vibration required for voicing (in that account). See, the problem here is that you're thinking about acoustics, but SPE is almost totally about articulation.
    – Greg Lee
    Commented Apr 25, 2016 at 23:55

1 Answer 1


There are three linguistic terms which have a common Latin origin, sonor: "sonorant", "sonority" and "sonorous". A sonorant is a sound produced with a vocal tract cavity configuration where spontaneous voicing is possible: see Chomsky & Halle 1968 The sound pattern of English p. 300-301 for discussion of the latter concept. Thus vowels, glides (including [h,ʔ], liquids and nasals are [+sonorant], anything else is [-sonorant]. "Sonority" is a less-well defined concept, but is approximately equal to the inverse of resistance to airflow – as such, it is not a binary property but sub-classifies sounds along a scale of at least 5 points (probably not a continuous scale). A [+sonorant] sound need not have vocal fold vibrations, because that is controlled by separate features governing the approximation of the vocal folds. Sounds that have high sonority likewise are not necessarily voiced, for the same reason. So, one can spread the vocal folds when producing a bilabial nasal, and [m̥] satisfies the criteria for being [+sonorant]. Without such active spreading, there would be vocal fold vibration.

Taking [z] as one example of a voiced obstruent, because of the lingual construction, spontaneous voicing is not possible, but voicing is (see SPE p. 300-1). There are a number of accommodations possible with a voiced fricative, such as lowering the larynx, raising the velum, slightly opening the velum, moving the constriction forward. This does become a zero sum game with stop, which can sustain voicing only for a short time until supraglottal pressure rises to the point that voicing stops.

What allows voicing is trans-glottal pressure drop; a trans-glottal pressure drop requires a sufficiently open vocal tract that oral air pressure does not become too high. Turbulence is kind of orthogonal, though it's hard to have turbulence with a really open vocal tract. Now: you use the term "sonorous", but unfortunately that is only marginally a term in linguistics. It might be used to refer to RMS amplitude, or it might be used to refer to "having higher sonority" but that requires a standard of comparison to be meaningful. Because sonority is inversely related to resistance to airflow, resistance to airflow is what leads to oral pressure buildup, oral pressure buildup reduces trans-glottal pressure drop, and a trans-glottal pressure drop is a requirement for voicing... there is some relationship between sonority and voicing.

There is quasi-periodic vocal fold vibration during production of voiced obstruents (though for some dialects in some positions in English, phonologically voiced stops are actually not produced with vocal fold vibration); you can verify this with an electroglottograph. Anything that vibrates quasi-periodically has a computable frequency, and it's not hard to compute in Praat, if you set the analysis parameters right.

W.r.t. "the noise", I assume you're speaking of the acoustic energy during the production of a voiced stop. Whether you mean "per msc." or "summed up", the laryngeal component contributes almost all of the acoustic energy. Unlike voiceless stops, there is very little pressure buildup during a voiced stop, so very little if any release burst.

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.