<< Previous: Introduction Next: Morphology >>
Miluk has a fairly complex sound system, quite typical for the languages of the Pacific Northwest, characterized by a rich consonant inventory next to a simple vowel system.
1.1. Consonants
The consonants of Miluk can be divided into two basic types, viz. obstruents and sonorants. This distinction is present in all aspects of Miluk phonology, from phonotactics to morphophonology.
The system of obstruents is very rich. Maximally, four kinds of articiulation are distinguished: three types of stops/affricates (plain, aspirated, ejective) and a voiceless fricative. The constrast between plain and aspirate stops/affricates is neutralized word-finally and before another obstruent, but is still clearly phonemic in all other positions. An analysis of aspirate stops/affricates as clusters + h has been proposed by Doty (2012), but the reduplication patterns of roots beginning with an aspirate stop or affricate clearly show that these are single obstruents.
Miluk further distinguishes ten sets of obstruents according to place of articiulation, which can be grouped into front, central and back sets, plus a laryngeal set.
1.1.1. Front obstruents
Front obstruents have a three-way contrast of stops. But, like in most other North American languages, the corresponding fricatives are lacking.
bilabial: b, p, p̓
apical: d, t, t̓
1.1.2. Central obstruents (= sibilants/affricates)
This contains the three affricate/sibilant sets exhibiting all three contrasts for the affricates, and the sibilants filling the fricative slot.
alveolar: ʒ (dz), c (ts), c̓ (t’s), s
predorsal: ǯ (dj), č (tc), č̓ (t’c), š (c)
lateral: λ (dl), ƛ (tɫ), ƛ̓ (t’ɫ), ɫ
1.1.3. Back obstruents
Back obstruents include velar and uvular sets, both rounded and unrounded. Unrounded velars have a palatalized articulation unless followed by a/ā: kəmkām «kʸimkām».
velar (unrounded): g, k, k̓, x
velar (rounded): gʷ, kʷ, k̓ʷ, xʷ
uvular (unrounded): ġ, q, q̓, χ (x̣)
uvular (rounded): ġʷ, qʷ, q̓ʷ, χʷ (x̣)
1.1.4. Laryngeals
The laryngeal series is defective and has only three members: ˀ (’), h, hʷ.
1.1.5. Sonorants
The system of Miluk sonorants is less complex, with each six plain and glottalized sounds:
plain sonorants: m, n, l, γ, y, w
glottalized sonorants: m̓, n̓, l̕, γ̓, y̓, w̓
The velar or uvular sound γ is treated as a sonorant here, since it shares the phonotactic and morphphonological properties of the other sonorants. Most likely, it was not a voiced fricative but rather an approximant [ɰ] or [ʁ̞]. Evidence for a possible seventh laryngeal sonorant, viz. voiced /ɦ/, will be presented at the end of this section.
1.2. Vowels
1.2.1. Surface inventory
The vowel system of Miluk as appears in the text corpus can be described as a 4 + 1 vowel system: four vowels can appear as long or short:
a, e, i, u - ā, ē, ī, ū (spelled a·, e·, i·, u· in the corpus)
The fifth vowel is ə, which is always short. The vowels e and ē were pronounced very low, i.e. as [æ]/[æ:].
Jabobs further list several "diphthongs" of the shape au, ai, eu etc. These are better interpreted as vowel + sonorant y or w, as can be seen from morphophonological processes such infixation. E.g. the verb "tell a lie" has two forms: heusdu and hewēsēnu. This suggests a root *hews.
1.2.2. A phonemic analysis
Final vowels are always spelled as short vowels in absolute final position, but usually as long vowels if followed by an enclitic. I take the long vowel as basic, with surface shortening in absolute final position.
Based on variation of spelling in the text corpus, I propose that short i and u are non-phonemic spelling variants either of unstressed long ī and ū, or of the central vowel ə. In the latter case, u appears in labialized environments (λuw̓iyam or λəw̓iyam "eat"), and i elsewhere (t̓əm or t̓im "pack", -səm or -sim "middle suffix").
Consequently, we can posit for Miluk the following surface inventory of vowel phonemes:
short: a, e, ə
long: ā, ē, ī, ū.
Phonetically, short ə was most probably a high central vowel [ɨ], which was subject to backing and rounding next to labialized obstruents, and otherwise tended towards a fronted articulation yielding [ɪ].
The phonemic analysis does not end here though: the high long vowels ī and ū can be analyzed as /əy/ and /əw/, since they go exactly parallel with combinations of /ə/ with the other for sonorants. E.g. the reduplicated third person object forms with infixed a/ā of t̓əm "pack" and ƛū "marry" are:
t̓əm > dəmt̓ām
ƛū > ƛūƛau
The latter becomes fully regular if analyzed as /ƛəw/ → /ƛəwƛaw/. Compare also the following pair:
p̓əls > p̓als-t "tear it"
pīx > payx-c "paint it"
which again points to /pəyx/ as underlying form of pīx. We thus arrive at the final underlying phonemic vowel system for Miluk:
short: a, e, ə
long: ā, ē
In the following sections, I will spell the vowels from the sample texts as written down by Jacobs. For the citation form of morphemes and roots however, a surface phonemic re-spelling is used. The underlying form will be resorted to only for the discussion of morphophonological processes.
1.3. Phonotactics
1.3.1. Syllable strucure
The syllable structure of Miluk roots and words is moderately complex when compared to other languages of the Pacific Northwest. For the following description, I will use these abbreviations:
C: any consonant
K: obstruent
R: resonant
V: vowel
The most complex form of a syllable is KCVRKK, although full clusters to the left and right of the vowel are extremely rare, as e.g. in the word ƛxanxt "encounter it", which is morphologically complex, composed of the root √ƛxənx plus infix -a- and suffix -t.
In final KK clusters, the second obstruent is restricted to t, c, and s. The clusters -Kt and -Kc are very frequent, since -t and -c are common verb suffixes. Roots with a final KK-cluster are very rare, e.g. ˀelqs "afraid".
Final clusters of the shape RR are only allowed at the word level if the first element is a semivowel (overt or underlying), e.g. ilaχq̓ain (/yəlaχq̓ayn/) "shaman", ƛīn (/ƛəyn/) "straight", cūm (/cəwm/) "lie down". On the root level, there is no constraint on final RR-clusters. If however the first element is a nasal or liquid, such roots always appear in words which either carry a suffix, or the cluster is broken up by an infix, e.g. the root √λəml "spear (fish)" which appears as intrasitive λəm:le / λəml̓iyam or transitive λəm:al.
For the initial KC-clusters, there appear to be little constrains. One rigid constraint is that the glottal stop ˀ never appears as first part of an initial cluster. Clusters of the shape KR- are rare but do occur, e.g. c̓mīχʷən "trickster".
1.3.2. Vowel harmony
Miluk has a constraint against the cooccurrence of e/ē and a/ā in a single word. The front vowels e/ē change to a/ā whenever they occur together due to affixation. This harmonization can affect root vowels as well as vowels in affixes.
Vowel harmony in root vowels:
Vowel harmony in affixes:
At the underlying phonemic level, this results in a very limited number of vowels which can appear in a Miluk word, viz. either e/ē/ə or a/ā/ə.
1.3.3. Vowel lowering
Following the uvular obstruents and the glottal stop ˀ, only low vowels can appear. This constraint affects basic roots as well affixes or reduplicated syllables. By default, ə, and also ī and ū (= underlying əy and əw) are changed to e, ey and ew, which can become a, ay and aw, respectively, as a result of vowel harmony.
Lowering of vowels in suffixes:
In the last example, we can also find a low vowel inserted under stative reduplication: naqaq-. Other examples include taiχaχ- "go to river", ɫāqaq- "wait", helqeq- "climb". With consonants other than uvular obstruents, the inserted vowel is ə (often spelled i or u): umīdəd- "follow", qaλəλ- "fall".
1.3.4. De-glottalization of ejectives
Within a Miluk word, only one ejective stop/affricate is allowed to occur. If suffixation or reduplication would lead to the occurrence of two ot three ejectives, then only the rightmost ejective is retained, whereas all other ejectives are changed to the corresponding plain stop.
Intensive reduplication:
Stative reduplication:
Suffixation:
Stative reduplication + suffixation:
De-glottalization of ejectives is further found in intransitive verb forms that have the suffix -a or the pluralizing suffix -īda:
In fact, the suffix -īda itself also displays de-glottalization of the pluralizing element -t̓- which is found in the suffixes -īt̓a/-āt̓a (direct plural object) or -ūt̓- (middle plural).
However, de-glottalization is not found with transitive direct suffix -a, which indicates that the intransitive suffix -a must be phonemically distinct. The most straightforward way to account for this is to assume some kind of glottalized articulation of the intransitive suffix -a, which latter therefore I tentatively re-spell as -aˀ.
1.3.5. Two h-phonemes?
Evidence from phonotactic constraints indirectly points at two different h-phonemes in Miluk. One of these behaves as a laryngeal obstruent: with the root √ǯənh, the final h can undergo stative reduplication in the intransitive activity verb ǯinhehēnu, which is only possible for obstruents (cf. Section 2). Further, the reduplicated syllable shows the lowered vowel e, which is otherwise only found with uvular obstruents and the glottal stop.
The other putative h-phoneme behaves like a sonorant, which can be seen with the root √dəwh. The simple direct form dūha follows the pattern of sūla (√səwl) by taking the simple suffix -a. With final obstruents, the suffix would take the shape -iya (e.g. pūkʷiya from √pəwkʷ). The applicative direct form is dūhay̓a, with a glottalized y̓ in the suffix -ay̓a. Glottalized y̓ only appears following plain sonorants (e.g. č̓ūlay̓a), whereas it is deglottalized following an obstruent (e.g. wusāya). Finally, there is no constraint for this sonorant h-phoneme to be followed by a high/mid vowel, as can be seen from the local, inverse and passive forms for this root, all of which have the shape dūhi- + characteristic ending (the vowel i most likely represents the long high vowel /ī/ (= /əy/).
In Jacobs' notes on the sounds of Miluk, no mention is made of any phonetic variation associated with h, so we completely depend on internal evidence here. Most likely, the first h-phoneme was a voiceless fricative [h], whereas the second must have been a voiced approximant [ɦ]. The distribution in the lexicon can be partly reconstructed: every instance of h followed by a high or mid vowel must be identified as representing the sonorant, e.g. hīt "sell" (/ɦəyt/), hūd "mock" (/ɦəwd/). In other cases where h is followed by a low vowel, the phonemic interpretation is ambiguous. Yet the frequency of e/ē/a/ā following h is much higher than found with front, central or velar obstruents; this clearly must be the result of the constraint for high/mid vowels following obstruent /h/. The ambiguity is however sometimes resolved by certain morphophonological processes such as reduplication of CVRK-roots. With intensive reduplication, the reduplicated syllable takes a default vowel, i.e. /ə/ unless C is a uvular obstruent or /ˀ/, in which case the defaul vowel is /e/ (or /a/ through vowel harmony). Forms like hekhelk̓u, haqhal:aq̓, hachāyac or hačhāyač̓ clearly point towards the obstruent /h/.
In this sketch however, I will not distinguish between /h/ and /ɦ/, in the first place due to the still preliminary nature of my hypothesis, and due to lack of evidence for distiguishing them systematically.
<< Previous: Introduction Next: Morphology >>
1.1. Consonants
The consonants of Miluk can be divided into two basic types, viz. obstruents and sonorants. This distinction is present in all aspects of Miluk phonology, from phonotactics to morphophonology.
The system of obstruents is very rich. Maximally, four kinds of articiulation are distinguished: three types of stops/affricates (plain, aspirated, ejective) and a voiceless fricative. The constrast between plain and aspirate stops/affricates is neutralized word-finally and before another obstruent, but is still clearly phonemic in all other positions. An analysis of aspirate stops/affricates as clusters + h has been proposed by Doty (2012), but the reduplication patterns of roots beginning with an aspirate stop or affricate clearly show that these are single obstruents.
Miluk further distinguishes ten sets of obstruents according to place of articiulation, which can be grouped into front, central and back sets, plus a laryngeal set.
1.1.1. Front obstruents
Front obstruents have a three-way contrast of stops. But, like in most other North American languages, the corresponding fricatives are lacking.
bilabial: b, p, p̓
apical: d, t, t̓
1.1.2. Central obstruents (= sibilants/affricates)
This contains the three affricate/sibilant sets exhibiting all three contrasts for the affricates, and the sibilants filling the fricative slot.
alveolar: ʒ (dz), c (ts), c̓ (t’s), s
predorsal: ǯ (dj), č (tc), č̓ (t’c), š (c)
lateral: λ (dl), ƛ (tɫ), ƛ̓ (t’ɫ), ɫ
1.1.3. Back obstruents
Back obstruents include velar and uvular sets, both rounded and unrounded. Unrounded velars have a palatalized articulation unless followed by a/ā: kəmkām «kʸimkām».
velar (unrounded): g, k, k̓, x
velar (rounded): gʷ, kʷ, k̓ʷ, xʷ
uvular (unrounded): ġ, q, q̓, χ (x̣)
uvular (rounded): ġʷ, qʷ, q̓ʷ, χʷ (x̣)
1.1.4. Laryngeals
The laryngeal series is defective and has only three members: ˀ (’), h, hʷ.
1.1.5. Sonorants
The system of Miluk sonorants is less complex, with each six plain and glottalized sounds:
plain sonorants: m, n, l, γ, y, w
glottalized sonorants: m̓, n̓, l̕, γ̓, y̓, w̓
The velar or uvular sound γ is treated as a sonorant here, since it shares the phonotactic and morphphonological properties of the other sonorants. Most likely, it was not a voiced fricative but rather an approximant [ɰ] or [ʁ̞]. Evidence for a possible seventh laryngeal sonorant, viz. voiced /ɦ/, will be presented at the end of this section.
1.2. Vowels
1.2.1. Surface inventory
The vowel system of Miluk as appears in the text corpus can be described as a 4 + 1 vowel system: four vowels can appear as long or short:
a, e, i, u - ā, ē, ī, ū (spelled a·, e·, i·, u· in the corpus)
The fifth vowel is ə, which is always short. The vowels e and ē were pronounced very low, i.e. as [æ]/[æ:].
Jabobs further list several "diphthongs" of the shape au, ai, eu etc. These are better interpreted as vowel + sonorant y or w, as can be seen from morphophonological processes such infixation. E.g. the verb "tell a lie" has two forms: heusdu and hewēsēnu. This suggests a root *hews.
1.2.2. A phonemic analysis
Final vowels are always spelled as short vowels in absolute final position, but usually as long vowels if followed by an enclitic. I take the long vowel as basic, with surface shortening in absolute final position.
Based on variation of spelling in the text corpus, I propose that short i and u are non-phonemic spelling variants either of unstressed long ī and ū, or of the central vowel ə. In the latter case, u appears in labialized environments (λuw̓iyam or λəw̓iyam "eat"), and i elsewhere (t̓əm or t̓im "pack", -səm or -sim "middle suffix").
Consequently, we can posit for Miluk the following surface inventory of vowel phonemes:
short: a, e, ə
long: ā, ē, ī, ū.
Phonetically, short ə was most probably a high central vowel [ɨ], which was subject to backing and rounding next to labialized obstruents, and otherwise tended towards a fronted articulation yielding [ɪ].
The phonemic analysis does not end here though: the high long vowels ī and ū can be analyzed as /əy/ and /əw/, since they go exactly parallel with combinations of /ə/ with the other for sonorants. E.g. the reduplicated third person object forms with infixed a/ā of t̓əm "pack" and ƛū "marry" are:
t̓əm > dəmt̓ām
ƛū > ƛūƛau
The latter becomes fully regular if analyzed as /ƛəw/ → /ƛəwƛaw/. Compare also the following pair:
p̓əls > p̓als-t "tear it"
pīx > payx-c "paint it"
which again points to /pəyx/ as underlying form of pīx. We thus arrive at the final underlying phonemic vowel system for Miluk:
short: a, e, ə
long: ā, ē
In the following sections, I will spell the vowels from the sample texts as written down by Jacobs. For the citation form of morphemes and roots however, a surface phonemic re-spelling is used. The underlying form will be resorted to only for the discussion of morphophonological processes.
1.3. Phonotactics
1.3.1. Syllable strucure
The syllable structure of Miluk roots and words is moderately complex when compared to other languages of the Pacific Northwest. For the following description, I will use these abbreviations:
C: any consonant
K: obstruent
R: resonant
V: vowel
The most complex form of a syllable is KCVRKK, although full clusters to the left and right of the vowel are extremely rare, as e.g. in the word ƛxanxt "encounter it", which is morphologically complex, composed of the root √ƛxənx plus infix -a- and suffix -t.
In final KK clusters, the second obstruent is restricted to t, c, and s. The clusters -Kt and -Kc are very frequent, since -t and -c are common verb suffixes. Roots with a final KK-cluster are very rare, e.g. ˀelqs "afraid".
Final clusters of the shape RR are only allowed at the word level if the first element is a semivowel (overt or underlying), e.g. ilaχq̓ain (/yəlaχq̓ayn/) "shaman", ƛīn (/ƛəyn/) "straight", cūm (/cəwm/) "lie down". On the root level, there is no constraint on final RR-clusters. If however the first element is a nasal or liquid, such roots always appear in words which either carry a suffix, or the cluster is broken up by an infix, e.g. the root √λəml "spear (fish)" which appears as intrasitive λəm:le / λəml̓iyam or transitive λəm:al.
For the initial KC-clusters, there appear to be little constrains. One rigid constraint is that the glottal stop ˀ never appears as first part of an initial cluster. Clusters of the shape KR- are rare but do occur, e.g. c̓mīχʷən "trickster".
1.3.2. Vowel harmony
Miluk has a constraint against the cooccurrence of e/ē and a/ā in a single word. The front vowels e/ē change to a/ā whenever they occur together due to affixation. This harmonization can affect root vowels as well as vowels in affixes.
Vowel harmony in root vowels:
hāwiya "raise s.o."
wāsdidiyam "return (pl.)"
mikman:ak "beat it (int.)"
wāsdidiyam "return (pl.)"
mikman:ak "beat it (int.)"
< √hēw + -iya (hēwi "grow")
< √wēst ("return") + stative reduplication + -iyam
< √meng + intensive reduplication + -a2-
< √wēst ("return") + stative reduplication + -iyam
< √meng + intensive reduplication + -a2-
Vowel harmony in affixes:
γālānu "talk"
χaɫtām "make (passive)"
ƛ̓āis-əja "net (allative)"
χaɫtām "make (passive)"
ƛ̓āis-əja "net (allative)"
< √γal + -ēnu (γala "word")
< √χaɫ + -d + -ēm
< ƛ̓āis- + -əje
< √χaɫ + -d + -ēm
< ƛ̓āis- + -əje
At the underlying phonemic level, this results in a very limited number of vowels which can appear in a Miluk word, viz. either e/ē/ə or a/ā/ə.
1.3.3. Vowel lowering
Following the uvular obstruents and the glottal stop ˀ, only low vowels can appear. This constraint affects basic roots as well affixes or reduplicated syllables. By default, ə, and also ī and ū (= underlying əy and əw) are changed to e, ey and ew, which can become a, ay and aw, respectively, as a result of vowel harmony.
Lowering of vowels in suffixes:
tēyχeu "go to river"
sλaq̓ai "bathe"
naqaqayam "run (pl)"
sλaq̓ai "bathe"
naqaqayam "run (pl)"
< √tēyχ + -ū
< √sλeq̓ + -ī
< √nəq + -əyam
< √sλeq̓ + -ī
< √nəq + -əyam
In the last example, we can also find a low vowel inserted under stative reduplication: naqaq-. Other examples include taiχaχ- "go to river", ɫāqaq- "wait", helqeq- "climb". With consonants other than uvular obstruents, the inserted vowel is ə (often spelled i or u): umīdəd- "follow", qaλəλ- "fall".
1.3.4. De-glottalization of ejectives
Within a Miluk word, only one ejective stop/affricate is allowed to occur. If suffixation or reduplication would lead to the occurrence of two ot three ejectives, then only the rightmost ejective is retained, whereas all other ejectives are changed to the corresponding plain stop.
Intensive reduplication:
dəm-t̓ām "pack it"
bəs-p̓əl:as "tear it"
haq-hal:aq̓ "name it"
bəs-p̓əl:as "tear it"
haq-hal:aq̓ "name it"
< √t̓əm + intenstive reduplication + -ā2-
< √p̓əls + intenstive reduplication + -a2-
< √helq̓ + intenstive reduplication + -a2-
< √p̓əls + intenstive reduplication + -a2-
< √helq̓ + intenstive reduplication + -a2-
Stative reduplication:
mīgək̓ānīda "play ball"
ɫuġʷāq̓ʷ "flow"
ɫuġʷāq̓ʷ "flow"
< √mīk̓ + stative reduplication + -ēn + -īda
< √ɫq̓ʷ + stative reduplication + -ā2-
< √ɫq̓ʷ + stative reduplication + -ā2-
Suffixation:
čǯiʒāt̓a "pull them out"
hūʒīt̓a "do to them"
hūʒīt̓a "do to them"
< √č̓ǯ + -ʒ + -āt̓a
< √hūc̓(ū) + īta
< √hūc̓(ū) + īta
Stative reduplication + suffixation:
laġawidədīt̓a "tell them"
< √laġawīt̓ + stative reduplication + -īt̓a
De-glottalization of ejectives is further found in intransitive verb forms that have the suffix -a or the pluralizing suffix -īda:
gʷubəba "lined up"
sginλiya "go ahead"
hādida "sing (pl.)"
sginλiya "go ahead"
hādida "sing (pl.)"
< √k̓ʷb + stative reduplication + -a
< √sgənƛ̓ + -əy + -a
< √hāt̓ + -īda
< √sgənƛ̓ + -əy + -a
< √hāt̓ + -īda
In fact, the suffix -īda itself also displays de-glottalization of the pluralizing element -t̓- which is found in the suffixes -īt̓a/-āt̓a (direct plural object) or -ūt̓- (middle plural).
However, de-glottalization is not found with transitive direct suffix -a, which indicates that the intransitive suffix -a must be phonemically distinct. The most straightforward way to account for this is to assume some kind of glottalized articulation of the intransitive suffix -a, which latter therefore I tentatively re-spell as -aˀ.
1.3.5. Two h-phonemes?
Evidence from phonotactic constraints indirectly points at two different h-phonemes in Miluk. One of these behaves as a laryngeal obstruent: with the root √ǯənh, the final h can undergo stative reduplication in the intransitive activity verb ǯinhehēnu, which is only possible for obstruents (cf. Section 2). Further, the reduplicated syllable shows the lowered vowel e, which is otherwise only found with uvular obstruents and the glottal stop.
The other putative h-phoneme behaves like a sonorant, which can be seen with the root √dəwh. The simple direct form dūha follows the pattern of sūla (√səwl) by taking the simple suffix -a. With final obstruents, the suffix would take the shape -iya (e.g. pūkʷiya from √pəwkʷ). The applicative direct form is dūhay̓a, with a glottalized y̓ in the suffix -ay̓a. Glottalized y̓ only appears following plain sonorants (e.g. č̓ūlay̓a), whereas it is deglottalized following an obstruent (e.g. wusāya). Finally, there is no constraint for this sonorant h-phoneme to be followed by a high/mid vowel, as can be seen from the local, inverse and passive forms for this root, all of which have the shape dūhi- + characteristic ending (the vowel i most likely represents the long high vowel /ī/ (= /əy/).
In Jacobs' notes on the sounds of Miluk, no mention is made of any phonetic variation associated with h, so we completely depend on internal evidence here. Most likely, the first h-phoneme was a voiceless fricative [h], whereas the second must have been a voiced approximant [ɦ]. The distribution in the lexicon can be partly reconstructed: every instance of h followed by a high or mid vowel must be identified as representing the sonorant, e.g. hīt "sell" (/ɦəyt/), hūd "mock" (/ɦəwd/). In other cases where h is followed by a low vowel, the phonemic interpretation is ambiguous. Yet the frequency of e/ē/a/ā following h is much higher than found with front, central or velar obstruents; this clearly must be the result of the constraint for high/mid vowels following obstruent /h/. The ambiguity is however sometimes resolved by certain morphophonological processes such as reduplication of CVRK-roots. With intensive reduplication, the reduplicated syllable takes a default vowel, i.e. /ə/ unless C is a uvular obstruent or /ˀ/, in which case the defaul vowel is /e/ (or /a/ through vowel harmony). Forms like hekhelk̓u, haqhal:aq̓, hachāyac or hačhāyač̓ clearly point towards the obstruent /h/.
In this sketch however, I will not distinguish between /h/ and /ɦ/, in the first place due to the still preliminary nature of my hypothesis, and due to lack of evidence for distiguishing them systematically.
<< Previous: Introduction Next: Morphology >>
No comments:
Post a Comment