Variation in the qualities of epenthetic vowels in Tashlhiyt Berber

Many of the phonetic transcriptions given below are linked to sound clips (.wav files). Except in transcriptions marked with an initial D, the speaker is Mohamed Elmedlaoui. (Starred transcriptions were spoken by Lahcen Damiri.) The examples follow the numbering of examples given in J. Coleman (2000), The Phonetics and Phonology of Tashlhiyt Berber Syllabic Consonants, in Transactions of the Philological Society, for which this page was specifically composed.

A postscript version of an earlier (1999) paper of mine, The Nature of Vocoids Associated with Syllabic Consonants in Tashlhiyt Berber. In J. J. Ohala, Y. Hasegawa, M. Ohala, D. Granville and A. C. Bailey, eds. Proceedings of The XIVth International Congress of Phonetic Sciences. Vol. 1. 735-8,  can be downloaded from here.  But please note that the analysis given in the Transactions paper is different from the ICPhS paper, as I have abandoned the idea that different qualities of epenthetic vowels might be contrastive.

Examples of Tashlhiyt syllabic consonants

Some Tashlhiyt words which, according to Dell and Elmedlaoui, consist entirely of consonants:
(7)  /tf̩tk̩̩ts̩tt/ `you sprained it'

/tl̩di/ `she pulls'

/ts̩kr̩t/ `you did'

/tn̩ʃf̩t/ `you grazed the skin'

/tχ̩zn̩t/ `you stored'

Labialization of velar and uvular obstruents is usually contrastive, e.g. D/ikti/ `hot' vs. D/ikwti/ `recall'. Consonants may be contrastively geminate, even in word-initial position:
(9a) /t̩tsχ̩χan/ `dip (in sauce)'
b) /t̩tbd̩dal/ `exchange'
c) /t̩ˁtˁaʒin/ `plate'
d) /s̩ʃ/ `eat'
e) /t̩ˁtˁaf/ `hold'
f) /t̩ˁtˁasˁ/ `sleep'
g) D/t̩tg̩wgwa/ `be washing clothes'
h) D/t̩tg̩ga/ `be' [progressive]

As in many other Afroasiatic languages, many words are formed by intercalating independent vowel and consonant melodies, e.g. singular D/aʃakwʃ/ 'pile of stones', plural D/iʃukaʃ/.

Phonetics of Tashlhiyt syllabics

The tap(s) of the alveolar sonorant /r/ are almost always accompanied by short, central vocoids of an [ə]-like quality:
(14a) /afruχ/ [afəruχ] `boy' (optional [ə])

/tafruχt/ [tɛfəɾʊχt] `girl'
b) /aʁwrrˁaf/ [aʁwrˁ:af] `bowl with handle'
c) /aʁwrˁmi/ [aʁwŏɾʊmɪ̆] `date-stone'

/aggwrn/ [ag:wʊɾə̆n] `flour'

/aqwrˁnˁif/ [aqwŏɾə̆nif] `palm bough'

Dell and Elmedlaoui (1985) analyzed CC-initial words such as /kdˁu/ `smell', /gbu/ `pierce' etc. as having complex onsets. However, I analyse them as CəC-, so that e.g. [kdˁu] is phonological disyllabic /kədˁu/.

(16a) /kwdˁiχ/ `he smelled' is realised as [kʊdˁiχ] (with 2 voicing intervals, [ʊ] and [i]) in 9/10 tokens and as [kwdˁiχ] (with only [i] voiced) in 1 token out of 10.

b) /ikwbar/ `squirrels' is realised as D[ikʊbar] (with 3 voicing intervals, [i], [ʊ] and [a]) in 8/10 tokens and [ikwbar] (with 2 voicing intervals, [i] and [a]) in 2/10 tokens.

c) /kdˁu/ `smell!' is realised as [kədˁu] in 8/10 tokens and as [kdˁu] in 2/10 tokens.

The longest tokens of epenthetic [ʊ] in my recordings of Elmedlaoui's speech occur in /ʃʃʁw[ʊ]rˁʃ/ `beat!' (duration of [ʊ] = 66 ms), /akw[ʊ]r/ `to steal' (54 ms), and /agw[ʊ]l/ `to hang' (58 ms). For comparison, the shortest tokens of the lexical vowel /u/ are /jukrr/ `he drags', /jugl/ `he hung' ([u] duration = 44 ms), and /ugmn/ `they drew (water)' (39 ms). In Elmedlaoui's pronunciation of [managu], [u] duration was 53 ms, whereas in [aggwrn], the vocoid before [r] has a duration of 57 ms.
(17a) [managw] ~ [managu] `when'
b) D[aggwrn] ~ D[aggurn] `flour'
c) [aχwlil] ~ [aχlul] `snot'

Some results of the distribution analysis

Epenthesis was never observed between two voiceless consonants, except for one instance of /qs/ in the word /lˁqsˁtˁtˁ/ `poem', but this phonemic representation may be wrong. It is a nativized loan from the Arabic al qasˁiida. Damiri considered the correct phonemic transcription to be /lˁqsˁitˁtˁ/, which he pronounced with the vowel [i]. But Elmedlaoui amended this transcription to /lˁqsˁtˁtˁ/, pronounced with an epenthetic [ɛ], [lˁqɛsˁtˁtˁ].

Pronunciations by Elmedlaoui of two words in particular are highly problematic for Dell and Elmedlaoui's account of epenthesis. First, /nʃfat/ `skin! (v. pl.)' was pronounced [ənʃfat], with an initial epenthetic [ə]. This is unsurprising if the initial /nʃ/ is a syllable coda preceded by an empty nucleus (i.e. /(ə)nʃ.fat/). Since the preceding word, /za/, ends in a vowel, it is possible that the initial epenthetic [ə] of [ənʃfat] reflects the articulatory transition from [ə] to [n]. Second, both tokens of /tmzħ/ (token 1, token 2) `she jested' were spoken with a final epenthetic [].

Minimally distinct diphones

Each pair of words in (20) has a common sequence of two consonants, listed in the second column. According to Dell and Elmedlaoui's proposal, therefore, they should have the same epenthetic vowel or secondary articulation qualities. However, they do not. Those on the left have markedly fronter qualities than those on the right.
Fronter Backer
a) gn /gn/[gɪnj] `to sleep' gnnu[gən] `sew' [v.impf.]
b) ql /iqql/[qə̟l] `he waited' /ismuqql/[qʊl] `he waited'
c) gl /trglt/[gɪlj] `you locked' /jugl/[gʊl] `he hung'

In the word pairs listed in (21), those on the left have rounded syllabic consonants, whereas those on the right have unrounded consonants.
Rounded Unrounded
a) zn /uzˁnˁkwdˁ/[zʊ̟ŋ]`gazelle' /azn/[zə̠n] `send'
b) drt /tudrt/[dɵɾət]`life' [n.f.] /tajdrt/[dəɾət]`ear of wheat'
c) gl /jugl/[gʊl]`he hung' /trglt/[gɪlj] `you locked'

In (21), rounded epenthetic vowels are found in words containing /u/, and unrounded epenthetic vowels in words lacking /u/. (22a) has a syllabic with a closer secondary articulation than the corresponding syllabic in (22b). In this example, it is unclear how the phonological environment might affect the quality of the epenthetic vowel.

Closer More open
22) a) D/abʁjus/[bɤʁ] `stupid person' b) D/tsbʁt/[bʌʁ] `you painted'

In addition to these binary contrasts, I have also found some ternary contrasts of secondary articulation, such as:
23a) km Front, close: D/kmmi/[kɪm] `you' [f.2sg.]
Central, open: D/uglχkm/[kəm]`I hung you' [f.2sg.]
Close, back, rounded: D/akmmus/[kʊmw] `bundle' []

In (23a) the secondary articulation of the syllabic is clear, [ɪ]-like; in (23b), it is central, and neither especially clear nor dark; and in (23c) it is labiovelarized. This set of distinctions parallels the system of full vowels in Tashlhiyt. The epenthetic vowel of (23a) may anticipate the subsequent vowel /i/, and that of (23c) the subsequent /u/. However, the range of phonetic distinctions in the secondary articulation of syllabic consonants is not restricted to [ɪ]-like, [ʊ]-like and [ɐ]-like: a few groups of words containing a syllabic consonant with a wide variety of secondary articulations from the interior of the vowel space is presented in (24).


24) kr a) /tskrt/[kə̟r]`do' [v.perf.2sg.]

c) /jukr/[kʊr] `he stole'

b) /ikrzawn/[kər]`he ploughed for you'


d)  /akrkur/[kʌr] `blackbird'


Variation in quality of epenthetic vowels

Although the existence of variation can be demonstrated, it is not the case that epenthetic vowels are always of a particular category for a given word. For example, the epenthetic vowels [ɤ] and [ɐ] are both found after [k] in different tokens of /jukrr/ `he drags', [ʊ] and [ʏ] after [r] in /turmt/ `you tasted', [ʏ]and [ɐ] (again) after [dˁ] in /tsdˁrˁtˁ/ `you let fall', [ɪ], [] or [ɐ] after [k] in /trkst/ `you hid', and [ɤ] or [ə] after [t] in /trksas/ `she hid him'. Many other such examples could be given. Therefore, though the occurrence of epenthetic vowels is best explained by the presence of a phonologically empty nuclei in the syllabification of a word, their qualities are not given in the lexicon, contrary to my proposal in Coleman (1996, 1999).

Vowel-vowel coarticulation?

In many of the examples presented above, local context appears to play a role in determining these fine-grained details. However, the direction and domain of coarticulation is not clear from the data currently available. Disharmonious examples, such as D/lw[ə]r/ `run away' (not [lwʊr]), are problematic.

Phonology of Tashlhiyt syllabics

In (30) a Tashlhiyt lexical vowel corresponds to a Tamazight lexical schwa; and in (31), a syllabic consonant (optionally preceded by an epenthetic vowel) in Tashlhiyt corresponds to a lexical schwa or another lexical vowel in Tamazight.
(30) Tashlhiyt Tamazight, Kabyle
a) /aʒʒ/ /əddʒ/ `leave, abandon'
b) /izg[ə]r/ /əzgər/ `to cross'
c) /inχarˁ/ /inzər/ `nose'

(31) Tashlhiyt Tamazight
a) /im[ə]q̩qurˁ/ /aməqqwrˁan/ `big'
b) /im[ə]z̩ˁzˁi/ /aməʒtˁuħ/  `small'
c) D/[ə]t̩ˁtˁaʒin/ /adˁaʒin/  `plate'
d) D/[ə]s̩ʃ/ /əʃʃ/ `eat'
e) D/[ə]t̩ˁtˁaf/ /ətˁtˁəf/  `hold'
f) D/n[ə]k̩̩ki/ /nəkki/  `me'

Epenthetic vowel prediction

(38) Dell and Elmedlaoui's transcriptions Parser input Parser output
a) /̩m/ [isləm] islm is.ləm 
b) /tχz̩.nakkw/ [tχ(ə)z̩nakkw] tχznakkw təχ.zə.nakkw
c) /tbχ̩.lakkw/ [təbχ̩lakkw] tbχlakkw təb.χə.lakkw
d) /̩t/ [maratəgt] maratgt ma.rat.gət 
e) /̩.rin/ [it.əb(ə)d̩.rin] itbdrin it.bəd.rin
f) /tngdt/ [tn̩(ə)gətt] tngtt təng.tət
g) D/ts.bʁ`t/ [ts(ə)bʁ`t] tsbʁt təsb.ʁət
h) /r̩.gl̩χ/ [r̩gl̩əχ] rglχ rəg.ləχ
i) /iχn̩g/ [iχn̩əg] iχng iχ.nəg
j) /iχn̩gt/ [iχn̩əgət] iχngt iχ.nəgt
k) /iʃkd/ [iʃkəd] iʃkd iʃ.kəd
l) /i.ʃnn.qas/ [iʃn̩nəqas] iʃnnqas  iʃ.nən.qas
m) /aristajl̩qqajd/ [-jɪl-] aristajlqqajd a.ris.taj.ləq.qajd
n) /inawl̩mas/ [-wʊl-] inawlmas i.nawl.mas

In (38c) and (h), my parser added epenthetic vowels in places where none were found in Dell and Elmedlaoui's transcriptions. But in these two cases, close examination of my recordings of Elmedlaoui's pronunciation (click on each syllable of the following transcriptions separately) shows that the second schwa in [təb.χə.lakkw] and the first schwa in [rəg.ləχ] are empirically attested epenthetic vowels.

A few final examples:

lʁwrˁd 'hill'
xzr 'to give a look of anger or blame'
tzgr 'she crossed (e.g. the road)'
zʁwrˁ 'to grease'
zrˁ 'to watch'
gwmr 'to hunt or fish'
tˁnˁgdˁtˁ 'you drowned'
rgl 'lock!'
kʃm 'to enter'
nnqqwrˁtˁ 'silver'
gwdm 'to turn upside down'
ssrksxt 'I hid him'