9

In English there are just two series of stops, voiced (b, d, g) and unvoiced (p, t, k). The latter are generally aspirated (though it depends on phonological context).

In many common languages of Europe and Asia the unvoiced consonants are not aspirated, this is a common difficulty for English speakers learning them.

And in some other languages that English speakers do not learn as frequently there are contrasting aspirated and unaspirated unvoiced stops.

Now I've been learning Georgian for a few months and it has three series of contrasting stop consonants (and possible affricates I suppose):

  • voiced: b, d, g, j, dz
  • unvoiced aspirated: pʰ, tʰ, kʰ, chʰ, tsʰ
  • ejective: pʼ, tʼ, kʼ, chʼ, tsʼ, qʼ

Ejectives are even harder for English speakers to master than unaspirated voiceless stops. And in fact some teaching aids I've seen don't even mention the ejective quality of these stops and instead focus on the unaspirated quality, presumably because this is easier and is sufficient to distinguish the sounds for both the learners and the Georgian native speakers they will interact with.

So this has got me thinking, are there languages which contrast all three kinds of unvoiced stops: aspirated pulmonic, unaspirated pulmonic, and unaspirated ejective?

  • 3
    Close, but not exact: Korean has the three-way contrast between lightly aspirated, heavily aspirated and tensed glottis (romanised, e.g. for velar POA, as <g>, <k>, <kk> respectively). – jogloran Apr 20 '12 at 13:06
  • As a weeguk in Korea I think I got away with pronouncing these as voiced, aspirated, and unaspirated... but I didn't put as much time or effort into Korean as I have into Georgian so I know I didn't really master it. – hippietrail Apr 20 '12 at 13:48
  • I suppose what I was getting at was two contrasting kinds of unaspirated stops, but went the long way about explaining it I suppose. – hippietrail Apr 20 '12 at 19:43
  • 1
    I notice that some people evidently theorize that Proto-Indoeuropean's consonant inventory contrasted exactly what you describe: stops are voiceless (plain), voiceless (aspirated), and ejective/glottalized. Unfortunately I can only find that in Wikipedia, and it's not sourced. – Mark Beadles Apr 22 '12 at 19:11
  • 1
    @MarkBeadles: Wow that's really interesting. In fact I was wondering something along those lines yesterday or today and thought "nah it was something about gutturals". – hippietrail Apr 23 '12 at 12:05
11

Essentially all of the Na-Dene languages make such a distinction. To take Tlingit, the language I work on, the inventory of the three types of stops & affricates is as follows:

  • Unaspirated: t, ts, tʃ, tɬ, k, kʷ, q, qʷ
  • Aspirated: tʰ, tsʰ, tʃʰ, tɬʰ, kʰ, kʰʷ, qʰ, qʰʷ
  • Ejective: tʼ, tsʼ, tʃʼ, tɬʼ, kʼ, kʼʷ, qʼ, qʼʷ

The glottal stop (ʔ, also ʔʷ for some) could be either unaspirated or ejective depending on one’s analysis. The only voiced sounds in Tlingit are the sonorants /n, j, w, ɰ/, so none of the unaspirated stops & affricates are ever voiced in careful speech, and usually aren’t in fast speech either. Tlingit was described with instrumental phonetic data by Maddieson, Smith, & Bessell (2001, Anthropological Linguistics), if you care to check.

For completeness, Tlingit also has a plain/ejective fricative distinction:

  • Plain fricative: s, ʃ, ɬ, x, xʷ, χ, χʷ, h
  • Ejective fricative: sʼ, ɬʼ, xʼ, xʼʷ, χʼ, χʼʷ

The ejective fricatives are real ejectives, articulated in much the same way as ejective stops and affricates. They have a smaller aperture and greater tension with the primary articulator (the tongue) than the plain fricatives, compensating for the ‘leaky’ property of fricatives. Note the lack of *ʃʼ, which might be the result of a historical merger of *sʼ, *ʃʼ > sʼ.

As far as I’m aware, nearly all the Athabaskan languages have the same three-way distinction between unaspirated, aspirated, and ejective stops and affricates. Navajo is particularly well documented phonetically, as is Witsuwitʼen. For the latter see Sharon Hargus’s immense grammar (Hargus 2007). I think Joyce McDonough has some solid instrumental phonetic data for Dëne Sųłiné and Slave as well, but I don’t have a reference offhand. Eyak also has the same patterns, but I don’t know of any phonetic investigation of it. The Proto-Na-Dene language is thus logically reconstructed with the same three-way distinction.

| improve this answer | |
8

The Penutian language Wintu exhibits the following contrasts in the stop inventory:

╔═══════════╦═════════╦══════════╦══════════════╦═══════╦════════╗  
║   Stops   ║ Labial  ║ Alveolar ║ Postalveolar ║ Velar ║ Uvular ║  
╠═══════════╬═════════╬══════════╬══════════════╬═══════╬════════╣  
║ Voiced    ║ b       ║ d        ║ -            ║ -     ║ -      ║    
║ Aspirated ║ pʰ      ║ tʰ       ║ -            ║ -     ║        ║  
║ Ejective  ║ pʼ      ║ tʼ       ║ tʃʼ          ║ kʼ    ║ qχʼ    ║  
║ Voiceless ║ p       ║ t        ║ tʃ           ║ k     ║ q      ║  
╚═══════════╩═════════╩══════════╩══════════════╩═══════╩════════╝  

For the front stops, they exhibit the contrasts you mention (in addition to a plain voiced manner).

EDIT: The inventory above is not atypical of North American languages, for example Kiowa also contrasts voiced/voiceless/aspirated/ejective. And many of Africa's !Kung languages exhibit extensive contrasting manners of articulation among the stops; Ekoka contrasts up to 5 manners at each point of articulation, and that's not including clicks. It's not clear if this is the kind of thing you're asking about, though.

| improve this answer | |
  • Yes it's what I was asking about. Basically it seemed that several features go into a sound though the name might reflect only one feature. I was wondering whether each individual feature was salient or whether certain features tend to go together and add up to a contrasting sound only in combination... well I'm not expressing this well but you certainly answered my question. – hippietrail Apr 23 '12 at 11:35
  • 2
    Just to add another example, while ejectives are very rare in Austronesian languages they are found in Yapese (in Micronesia) and Waima'a (spoken in East Timor). Waima'a has a contrast between the following stop types: voiceless unaspirated, voiceless aspirated, voiceless ejective and (plain) voiced. This paper discusses the details. – Gaston Ümlaut Apr 23 '12 at 12:03
3

Voiced, unaspirated, aspirated, and ejective stops are distinguished in Dakota (a Souian Native American language) as well.

For example, ta with an unaspirated t is the postposition "to, towards," whereas t'a with the ejective is "to die," tʰa with the aspirated tʰ is a possessive marker (derived from the longer tʰawa) and de with the voiced d translates as "this".

The same distinctions appear for the sequence b, p, pʰ, and p' , and all but the voiced segment also appear in the sequences k, kʰ, k' and tʃ, tʃʰ, and tʃ'.

Hope this helps!

| improve this answer | |
3

This is quite common. I would argue that that Georgian pattern is almost the same thing as the aspirated-unvoiced-ejective pattern. This variant where the plain stop is voiced occurs frequently in other Caucasian languages as well as Georgian, and also shows up pretty frequently in North America.

To start looking for answers to questions like this, I would recommend this UPSID search tool, mixed with a little bit of programming and/or manual sorting.

http://web.phonetik.uni-frankfurt.de/upsid_find.html

This is a way of searching the UCLA Phonological Segment Inventory Database. This is a database of the phonetic inventories of 451 languages. These languages were chosen to be representative of human languages as a whole. It is not an attempt to record the phonetic inventories of every language, as can be seen by the fact that it doesn't even have English. If you're a little more adventurous, you could try downloading PHOIBLE 2.0 and writing software to search through that. That's a database with over 3000 phonological inventories.

A simple way to start would be to look for every language where unvoiced stops of the aspirated pulmonic, unaspirated pulmonic, and unaspirated ejective types occur.

So, I on three tabs of the search engine. I search for:

  • voiceless, no aspiration etc., plosive
  • voiceless, aspirated, plosive
  • voiceless, ejective-stop

I then copy the the languages list sections of the output pages, including all the messy numbers and phoneme lists and stuff, and assign them to 3 variables in Python 3.7.5 (which I have on my computer) based on which search they came from.

t = '''<insert text from voiceless, no aspiration etc., plosive>'''.split()
th = '''<insert text from voiceless, aspirated, plosive>'''.split()
tt = '''<insert text from voiceless, ejective-stop>'''.split()

The .split() function here converts things strings of text into lists strings using any spaces as dividers.

I then run a fairly simple little program to find what words are in all three lists but aren't just numbers, and without repeating everything 3 times.

for i in t:
    if (i in th) and (i in tt):
        mayBeAnExample = True
        for j in i:
            if j not in "ABCDEFGHIJKLMNOPQRSTUV-!?'~":
                mayBeAnExample = False
        if mayBeAnExample:
            print(i)

Note that "ABCDEFGHIJKLMNOPQRSTUV-!?'~" is every character used in UPSID language names, as can be seen on this page: http://web.phonetik.uni-frankfurt.de/upsid_languages.html

I get the following input back:

ACOMA
AHTNA
ARCHI
ARMENIAN
BATS
DAKOTA
GEORGIAN
HAIDA
HUPA
JACALTEC
JAQARU
KABARDIAN
KLAMATH
LAK
NAVAJO
PICURIS
QUECHUA
RUTUL
SOUTHERN
SOUTHERN
NAMBIQUARA

Now, the method I used involves splitting the text up into different words using spaces as dividers, but some of the UPSID languages have multi-word names, so I have to check against the names against http://web.phonetik.uni-frankfurt.de/upsid_languages.html to see if there are any that have been messed up by my program. (Actually, I don't in this case because I recognize these names from similar searches I've done very recently, but, in general, you might have to.)

In this output we can see that the last three words need to be fixed. The first "SOUTHERN" comes from "SOUTHERN KIWAI" (a Trans-New-Guinea language), which has plain voiceless stops, but not the other two. Southern Nambiquara, however, (a language from Matto Grosso Brazil) has all three types of stop (as well as two implosives), so both instances of the word "SOUTHERN" got taken along for the ride and accepted, while "KIWAI" was correctly omitted.

Thus our fixed list is:

  1. ACOMA
  2. AHTNA
  3. ARCHI
  4. ARMENIAN
  5. BATS
  6. DAKOTA
  7. GEORGIAN
  8. HAIDA
  9. HUPA
  10. JACALTEC
  11. JAQARU
  12. KABARDIAN
  13. KLAMATH
  14. LAK
  15. NAVAJO
  16. PICURIS
  17. QUECHUA
  18. RUTUL
  19. SOUTHERN NAMBIQUARA

19 languages out of 451 = 4.21%, so now you know what it actually means when I say that something is "quite common". It's interesting to note that all of these languages are from either the Americas or the Caucasus. It's also worth noting that some of these languages have more than just these three series of stops, like how I already said Southern Nambiquara has implosives. You can look each of them up to find out. Also, for each language on this list, you can probably find several more in the same language family or part of the world.

You may have noticed that the UPSID describes what you and Wikipedia say are voiced stops in Georgian as actually voiceless stops. Perhaps this is dialectical variation. UPSID really just records one dialect of each language, and the language names it uses are sometimes too vague to even look up the precise language easily. Alternatively, this could just be a different interpretation or just error on someones part.

It is worth noting that plain unvoiced stops sound fairly similar to plain voiced stops. Complicating this even further, Germanic languages like English often pronounce their "voiced" stops as plain voiceless stops in positions where our "voiceless" stops would always be aspirated. This is part of why plain voiceless stops sound voiced to English speakers in certain parts of words. This sort of thing is, no doubt, why they are often written as such in the orthographies of languages. This all makes sense when you consider that there is actually a continuum from voiced to unvoiced to aspirated, when thought of in terms of voicing onset time. Different languages place the allophones of their phonemes on this spectrum differently.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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