1. Stop voicing alternations
In the next few posts, I'll be looking at cases where cognate sets show a discrepancy in stop voicing between different languages, for instance Fayu
doû ~ Waritai
toû, both meaning "short" and Sikaritai
a-tá ~ Biritai
a-ɾa "muscle, meat". I'd like to explore whether there are any regular patterns to this (spoiler: there's at least one), how frequent this actually is, and which direction it goes in. Please feel free to criticise as much as you like. Language abbreviations are listed at the bottom.
1.1 Definition of a stop in this context
Firstly, I'll clear up what I mean by "stops", given that voiced/voiceless pairs aren't always stops in LP languages, Here's a summary of what I consider to be prototypical voicing pairs:
| Voiceless | Voiced |
| Labial | P = [p, ɸ, h] | B = [b, β, m] |
| Coronal | T = [t, s] | D = [d, ɾ, l, n] |
| Dorsal | K = [k] | G = [ɡ, ɣ] |
These all rely on well-established sound changes or allophony already known to exist in LP, so I'll just give a short explanation of these:
- PLP *p fricates in the majority of the languages, at least allophonically, and no LP languages distinguish any two of /p ɸ h/; these can all be safely treated as equivalent to P.
- Almost all nasal consonants appear only as allophones of the voiced stops, so any instances of /m n/ can be treated as B D - the notable exceptions being Faia, which has /n̥ m/ arising from clusters *pd bd (shown in Clouse and Clouse 1993, p.8), and Rasawa which has an adjectival suffix -nu/-no of uncertain origin.
- /β ɾ l ɣ/ are lenided allophones of voiced stops in most cases, although [β] seems to arise sometimes from fortition of *u in non-syllabic positions e.g. Sa keⁱawo vs. Ra kaɾiaβo "drink". Also, in some cases D may be reflected as ∅, e.g. Do bou-wa vs. Ka bouɾu-wa "bite, chew".
- /s/ always reflects PLP *t, although the exact mechanisms behind its development are unclear; Clouse says "The most puzzling change that occurred was the emergence of *s in PFW and PT. Although the phone [s] occurs in all of the languages, in no instance are there cognates where the reflex *s is chosen for all of the languages; *t is always chosen as the reflex in one or more of the languages, but not consistently in the same languages" (Clouse 1997, p. 140). I think it's safe to assume that [s] ⇒ T in the vast majority of cases, especially before /i/.
- It's known that PLP did not have a voiced equivalent to *k, so all instances of /ɡ/ are likely secondary developments, either allophonic or phonemic. Out of the languages with described phonologies, /ɡ/ is only phonemic in Doutai and Abawiri.
- I'm treating all instances of /kʷ ɡʷ/ as equivalent to K G; their phonemic status is posited for Sikaritai and Abawiri (it's worth noting that the work on Abawiri is a lot more in-depth than for Sikaritai and more reliable).
1.2 Consistent D-devoicing in Waritai?
There's only one instance of voicing alternation which Clouse mentions as a specific change, which is
*d >
d~t in Waritai, an East Tariku language. Exploring this further, I believe that this is indeed a predominantly consistent change word-initially, which is somewhat obscured by an entirely separate process of aphaerisis.
1.2.1 Examples
This devoicing can be seen in numerous examples, some of which are shown below:
PLP *du "bird" > tu
"comb" > îtɛ (cf. Ki dida, Do îdɛ, but also Ka titî)
"drink" > kwito-wa (Ka kwida-wa)
"pull" > tokidi-wa (Do dokudɛ)
"red" > tia (Si dí)
"snake" > təbi (Si débí)
Although there are also plentiful exceptions to this:
"cheek" > ɾaku (cf. Do dakû, but also Ka a-takuw)
"elbow" > ɾikû (Ka a-ɾiki)
"grass" > ɾû (Do duk)
"spider" > ɾoba (Do dobá)
"wide" > daɾisa (Do daɾisa)
And even a number in the other direction:
"heel" > ɾɛkû (cf. Er otoku, Ka a-tou)
"knee" > ɾûda (Ka a-sûa)
*to "leg" > ɾo (Ka a-to)
"vein" > ɾi (Bi atei)
?*auCo "wing" > doɾû (Si a-sóod, Pa aˈtɔɾu)
What's striking is that these last five are all body part terms which commonly take a prefix
a- (alternating with
e-,
o- etc.) in LP languages (cf. all the Kaiy forms), so these may be a case of intervocalic voicing. Of all the body parts in the Waritai list, only one instance has retained this prefix (
oro "knee") and this comes from the least reliable source (Voorhoeve
1975); in contrast it's plain
ɾo in the generally trustworthy Clouse. This is supported by other examples of D- where cognates have
at- or
ad-:
"root" > ɾi (cf. Si atgid̚, Ka atifɛɾi)
"sand" > do (cf. Pa ˈarɔ, possibly Si oɾɛta)
And "cheek" shown above could also be interpreted as such. A comparison can even be drawn to the retention of a voiced phoneme in "spider" mentioned above cognate to Papasena
aropa (at least in one wordlist; it's given as
oˈɸa in a different one). A more tenuous example is
ɾoutɛ "tongue" vs. Sikaritai
a-siɡ̚ja (/a.sík.ia/) and Eritai
e-tida; the
i~
u alternation is not unprecedented, and there are a few other instances of
ita :
îi̯a after (extra-)high vowels, e.g. "crocodile"
fɛita vs. Ob /héik.ia/. Waritai also stands out in that many examples where D's voicing is retained reflect it as
ɾ, which may imply that all cases where it was not devoiced are (or were historically) in lenition environments, where the realisation [ɾ] is areally almost universal. These examples are abundant enough that I can suggest a shift
*at, *ad > d /#_. As I may get to later, this is mirrored in the velars in at least one case (
gaî "hand" vs. Ka
a-kai), and possibly also in the labials (
boɾû "nose" vs. Pa
a-ˈpoɾuta,
baɾiso "skin" vs. Ob
a-hadid).
1.2.2 Interaction of voicing and tone
Tonal information is only available consistently for Iau and Abawiri, often for Obokuitai, Doutai and Sikaritai as well, and sometimes for Edopi; also stress is marked on the Papasena and Kwerisa entries. I list below the instances of the Waritai
*d > T and
*d > D correspondences where tonal information is available in cognate forms; I'll exclude intervocalic instances of D = D and the previously established VtV = dV to keep this somewhat manageable
1)
*d > T
| English | Waritai | Other languages |
| "bird" | tu | Si, Ob dù, Pa du, ˈdu, Ia du˨˩si˦˥, Ab dúke |
| "drink" | kwito-wa | Pa ˈkʷig̚ˈdawa |
| "fish" | tɛɾi | Ob tê, Si dèd, Pa ˈdɛrig̚, Ia dʊ˥˩, Ab tīe |
| "pull" | tokidi-wa | Pa ˈdawıkig̚ˈda |
| "red" | tia | Si, Ob dí, Pa diˈtɛ |
| "short" | toû | Pa daᵘg̚ˈtɛ |
| "snake" | təbi | Si débí, Pa daˈwi, ˈdauᵘi |
| "tomorrow" | tia | Pa ˈdia |
2)
*d > D
| English | Waritai | Other languages |
| "cold' | dɛkɛi | Pa aɾɛˈkaig̚ |
| "heavy" | dɛɸɛî | Pa aˈkudɛˈɸɛig̚, Kw auˈɛɾɛpi |
Honestly, I don't think much can be gained from this. If I treat PLP as having had *H, *L and *HL tones (like Foley does; if LH existed it was probably rare), then D-devoicing is seen before likely H ("red"), L ("bird") and HL ("fish") tones as well as in stressed ("drink", "pull") and unstressed syllables ("short", "tomorrow"). The two counter examples are both intervocalic in cognates so can be ignored. While tone is elusive here, as is usually the case with LP languages, it seems to have zero correlation with stop voicing.
1.2.3 Summary
Two shifts must to be at work here; namely 1) devoicing of initial D, and 2) voicing and/or flapping of T,D in
#a_V position with subsequent loss of the initial vowel. I think these are the only contradictory examples in my database:
i) D → d /#_ fails:
- ɾû "grass"
- diapoɾi "many"
- ɾita-ɾi "small"
- daɾisa "wide"
(1) is reflected as
buɾu in Kaiy, which is probably Waritai's closest relative (likewise Si
bid, Er
buɾu, Ob
sɛɾuɾub̚, Bi
saɾudu, Pa
aɸiˈɾi, but also Do
dû, Kw
du), which suggests that the PLP form – or at least the PET form – was something like
*budu (the Ob and Bi forms can't be derived from the same form as all the others; they're part of their own dialect chain anyway). I think it's fair to say that it's not a strong counter-example. (2) is a more confusing case; the only likely (partial) cognates are Edopi
odi.aᵘ.abe and Biritai
dîaɣa; Edopi at least suggests a similar intervocalic origin, although admittedly Biritai is more closely related to Waritai. (3) has no obvious cognates whatsoever except the very tenuous Papasena
aˈɾiku, which has a tempting initial vowel but an unexplained /t/ = /k/ correspondence. The last case is also annoying; the only clear cognate is Doutai
daɾisa; this could reasonably be put down to borrowing as the form is entirely identical and Doutai is Waritai's closest neighbour.
ii) a[T,D] → d /#_ fails:
- tɛiɸo "thick"
- səbɛɾɛ "thin"
(1) and (2) are cognate to Kaiy
ataipo and
atibiso respectively; however, no other cognates seem to have initial
a- (Do
téipó "thick", Do
sabaɾɛ, Si
sɛbɛkio-kwɛ, Er
sɛbaki, Ob
sɛbɛso "thin"). I'm tempted to put this down to an innovation in Kaiy, since there's no evidence for an initial vowel elsewhere.
1.2.4 Conclusion
My final verdict is that there was regular devoicing of PLP word-initial
*d in Waritai, which has been replaced by loss of initial
*a with flapping of following
*t and
*d; these changes must have happened following the breakup of the East Tariku languages as they are restricted to Waritai. These changes can be codified as follows:
- *d → t /#_
- *t,d → d~ɾ /#a_V
- *a → ∅ /#_ɾ
Tentatively I might be able to expand the second and third ones:
- *p t k → b d ɡ /#a_V
- *a → ∅ /#_[b,d,ɡ]
Although this extension is based off of only three examples.
Abbreviations:
Ab: Abawiri
Aw: Awera
Bi: Biritai
Da: Dabra
De: Deirate
Do: Doutai
Du: Duvle
Ed: Edopi
Er: Eritai
Fai: Faia
Fay: Fayu
Ia: Iau
Ka: Kaiy
Ki: Kirikiri
Kw: Kwerisa
Ob: Obokuitai
Pa: Papasena
PCT: Proto-Central Tariku
PELP: Proto-East Lakes Plains
PET: Proto-East Tariku
PFW: Proto-Far West
PLP: Proto-Lakes Plain
PWT: Proto-West Tariku
PT: Proto-Tariku
Ra: Rasawa
Sa: Saponi
Se: Sehudate
Si: Sikaritai
Tab: Taburta
Tau: Tause
Wa: Waritai
We: Weirate