Is there a real distinction in say, a spectrogram, between unaspirated voiceless stops and their voiced counterparts before a (voiced) vowel? For example, /ka/ and /ga/. Are they actually different phonetically? And if so, how different.... picoseconds of voicelessness? Or is it significant?

My question comes from noticing that in Zulu, an aspirated /k/ often becomes non-aspirated when there's a nasal preceding it I believe (e.g., inkukhu (chicken) instead of inkhukhu), and it sounds very close to a /g/.... particularly when followed by a vowel. So, my hypothesis is that the two sounds become so similar, that they probably lose their ability to be distinct in that environment. Anyone know any specifics?

3 Answers 3


Phonetically, the main theory I've heard is that voiced/voiceless/aspirated consonants are distinguished by voice onset time. VOT is the time delta between when the consonant stops and when the vocal folds start vibrating.

If the VOT is positive, then there's a gap between the consonant ending and the vowel beginning. This is aspiration.

If the VOT is close to zero, then the consonant ends and the vowel begins right after. This is a "tenuis" (= "normal") consonant.

If the VOT is negative, then the vocal folds are vibrating during the consonant itself. This is a voiced consonant.

In other words, aspirated — tenuis — voiced are all different points on a continuum. Some languages have all three as distinct, like Ancient Greek, while others have only two. And the ones with two categories can choose different places to draw the line: French has strongly voiced [d] and barely unvoiced [t], while English has strongly aspirated [t] and barely voiced (or even not voiced at all!) [d].

It's entirely possible for an aspirated consonant to become unaspirated or voiced after a nasal—nasals are generally voiced, so this is straightforward assimilation. However, it's also possible for the opposite to happen! In Swahili, for example, tenuis stops became aspirated after nasals.

  • It is also worth noting that in many languages including English and German, phonemically voiced plosives /b d g/ tend to be unaspirated voiceless when in onset and not preceded by a voiced sound, or when in coda and not followed by a voiced sound.
    – Nardog
    Oct 8, 2018 at 4:21
  • Also in English, the voiced–voiceless contrast in stops is neutralized when preceded by /s/ in onsets, making /p/ in spy more like /b/ in by than /p/ in pie, so the choice we transcribe e.g. spy with /p/ is rooted in phonology than in phonetics.
    – Nardog
    Oct 8, 2018 at 4:21
  • @Nardog: I don’t know enough phonetics to describe the details, but my understanding is that there are some ways in which the plosives in sC clusters are phonetically more similar to aspirated voiceless plosives than to “voiced” plosives. Maybe tone? Oct 8, 2018 at 4:40
  • @sumelic I've heard that too, though I'm also not a skilled enough phonetician to say exactly what's going on. It doesn't help that my dialect still aspirates the /p/ in "spy".
    – Draconis
    Oct 8, 2018 at 4:47
  • @sumelic Hmm, I hadn't heard that, but apparently you're right. From Cruttenden (2014: 47): "vowels following /p, t, k/ generally start from a higher pitch and vowels following /sp, st, sk/ have this higher pitch, which argues for /p, t, k/." This paper goes into more detail. Bottom line is, the evidence of course supports the analysis with /sp, st, sk/, but when only looking at the consonants, that in spy is closer to by than pie.
    – Nardog
    Oct 8, 2018 at 5:06

In Zulu is there is a contrast between aspirated voiceless, unaspirated (weak) ejective, and ostensively voiced stops. Aspirated consonants deaspirate after a nasal, so /izim-phaphe/ → [izim-paphe] 'feathers', cf. /izim-pete/ → [izim-pete] 'knock-kneed person'; see also your example "chicken". While /k, kh/ neutralize, /g/ remains distinct. However, distinguishing k and g (especially) but also t vs d is difficult. The voiced obstruents cause breathy voicing on the following vowel, and in fact they may be entirely devoiced during the closure if not preceded by a nasal. The main phonetic correlate of voicing of obstruents in Zulu (Xhosa, Swati) is causing breathiness on the vowel. There are a few contexts in Swati and Zulu where breathiness is not conditioned by a consonant (e.g. the first syllable of yebo "yes").

Long voice onset time is sufficient to identify aspirated stops. Breathy voicing and related pitch-lowering should be enough to distinguish b,d,g from unaspirated p,t,k, in case ejection is weak for a speaker / token.


The main difference between [k] and [g] is situated at the level of vocal folds. When [g] is pronounced the vocal folds vibrate, but not in the case of [k]. This situation is easily distinguishable on a spectrogram. [g] has what it is called a voice bar which does not appear on the spectrographic representation of [k]. Here a course which evokes the voice bar: http://ec-concord.ied.edu.hk/phonetics_and_phonology/wordpress/learning_website/chapter_3_consonants_new.htm

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.