Tonogenesis in Dene languages

To understand the reconstruction of PA that will concern us in the remainder of the blog post, one needs to have a basic understanding of the syllable structure in Leer's reconstruction of Pre-PA, the common ancestor of Eyak, Tlingit, and the Dene languages. Most of the properties listed below have analogues for other languages in the family that are also described in the paper, but we will gloss over these details, focusing only on Pre-PA:

1. Pre-PA syllables have an onset, which consists of a single consonant. Stops in this position may be plain, aspirated, or glottalised.
2. A syllable may have a coda, which consists of several consonants (usually no more than two). At the end of a syllable, plain and aspirated consonants are not contrasted.
3. There are four vowels (i,e,a,u) and vowels without a stigma are said to be reduced, and denoted with a lowercase v. Leer reconstructs this reduction as a simple shortening of the vowel. A rime cannot consist of a reduced vowel alone, but full vowels or reduced vowels followed by coda consonants are both valid.
4. The shape of the rime in Pre-PA is V (·/ˀ) (C) (C) (C)*. The first coda consonant may be a sonorant (R) or an obstruent (X), while the second is always an obstruent.
5. If the vowel of the rime has glottal stigma, the first coda consonant cannot be a sonorant.

Constriction

In the transition from Pre-PA to PA, a general phenomenon occurred which Leer calls suprasegmentalisation, in which the vowel takes on suprasegmental properties of a neighbouring coda consonant. For example, nasal sonorant consonants such as *m or *n in Pre-PA turned a preceding V or V· into a nasalised long vowel. Myriad other examples of suprasegmentalisation and spirantisation (an obstruent transforming into a fricative) are presented in the paper, and in particular, the author is able to reconstruct stem variation in PA verbs and nouns in terms of the stigmatic theory. The only phenomenon we will focus on in this post is the one that Leer calls constriction. It is the suprasegmentalisation of glottalisation, either in the stigma or in a coda consonant. Constriction in the vowel will be denoted by V' (the apostrophe is actually placed over the V in the text). The rule that describes how Pre-PA glottalisation became constriction in PA is given by

Pre-PA *VˀX, *vRˀX, *V·RˀX → PA *V'·X}.

A number of other rules are listed that add constriction, such as Pre-PA *vXˀ → PA *v'Xˀ, with Xˀ undergoing spirantisation (transforming into a fricative) if it is followed by another obstruent. This rule is somewhat interesting because it shows that a feature may spread leftward from a coda consonant to the vowel while the coda consonant remains unchanged. Notably, constriction does not occur with Pre-PA *V·Xˀ, which is the same in PA.

Finally, tone appears

We saw that in the transition from Pre-PA to PA, vowel glottalisation disappeared and was suprasegmentalised into vowel constriction. No Dene language spoken today actually has vowel constriction as Leer describes it. Instead, we find that constricted vowels turned into tone: low tone in some languages, and high tone in others. Interestingly, the (historically) low-marked language Tsuut'ina (Sarcee), spoken in central-southern Alberta, developed three tones:

Proto-Athabaskan                        Tsuut'ina
-----------------------------------------------------------
lengthened                          ->  marked high tone
short                               ->  unmarked mid tone
constriction (both short and long)  ->  marked low tone

in the Southern Dene Languages (Navajo and Apache), we have the following development:

Proto-Athabaskan                        Southern Dene
-----------------------------------------------------------
lengthened                          ->  long high tone
constricted long                    ->  long low tone
short constricted/unconstricted     ->  short low tone

Both of these peoples are thought to have migrated south from around the same geographical area, which would make it unsurprising if these tonal developments are related somehow. Leer also includes a couple of examples of Chipewyan alongside Tsuut’ina and Navajo, illustrating that Chipewyan has high tone wherever vowel constriction in the proto-form is present (as indicated by the tone in the other two languages).

As mentioned before, some neighbouring languages have opposite tone but are otherwise very similar, such as Northern/Southern Tutchone. The tone-flipping does not impede communication between these two groups, and a footnote goes so far as to say that the speakers of these languages do not find the fact that the tones are reversed in the other language interesting at all—John Ritter is quoted as calling it a “near universal ‘yawn factor’”. It seems that speakers of these languages are less aware or concerned with the high/low contrast than they are with the marked/unmarked contrast, which is much more uniform across the family.

I find this fascinating, as I am familiar with a partial analogue of this phenomenon all the way across the world in Singapore. I have family members who speak Hokkien and some who speak Teochew, two very similar mutually intelligible languages. One difference between Hokkien and Teochew is that Teochew has a high-level tone where Hokkien has a rising tone, and emphasising the perceived “musicality” of Teochew by exaggerating these high tones is a mild source of amusement among some family members (both Hokkien and Teochew) that I have asked about this.

What exactly was constriction?

The “glottalisation becomes constriction becomes tone” theory explains, with minimal exceptions, the source of tone in Dene languages. However, given that no extant Dene language has vowel constriction, we are left wondering what exactly this was, and how it morphed into two different “mirror-image” tone systems. This question was considered in 1985 by John Kingston. He notes that in some languages, a glottal stop after a vowel can induce a creaky voice (popularly known as “vocal fry”) on the vowel, whereas in other languages glottal stops induce a “tense” voice, which is characterised by higher frequencies.

Creaky voice involves low-frequency vibration of the vocal folds, and could therefore have transformed into low tone, while tense voice could become high tone. Thus one could hypothesise that vowel constriction was realised as creaky voice, which soon became accompanied by low tone. In some places, both of these properties disappear, leaving an atonal language. In others, the creaky voice could have disappeared, leaving behind a low-marked tonal language. In the final case, the low and lax creaky voice became a higher “tense creaky voice”, which leaves behind a high-marked tone.

This hypothesis, though somewhat complex, is somewhat more palatable than the naïve hypothesis that the tones simply switched places out of nowhere somewhere along the line. Leer discredits this line of thinking, writing that if spontaneous tone reversal was a real thing, we would have more examples of it in tone languages around the world. But the creaky-voice theory isn’t perfect either, and Leer closes the paper by indicating possible further research that could be done in the area.

References

• John Kingston, “The phonetics and phonology of Athabaskan tonogenesis,” Unpublished manuscript (1985), University of Texas at Austin.
• Michael E. Krauss and Victor Golla, “Northern Athapaskan Languages,” Handbook of North American Indians, Vol. 6. Subarctic, (1991), Smithsonian Institute, Washington.
• Fang-Kuei Li, “Chipewyan Consonants,” Bulletin of the Institute of History and Phonology, Academica Sinica, Ts’ai Yüan P’ei Anniversary Volume, Supplementary Volume 1 (1933), 429-467.
• Jeff Leer, “Tonogenesis in Athabaskan,” Cross-linguistic Studies of tonal phenomena: Tonogenesis, typology, and related topics (1999), Shigeki Kaji (ed.), 37-66.