ACOUSTIC FEATURES IN ATHABASCAN U of Calgary CSU-Fresno

38
ACOUSTIC FEATURES IN ATHABASCAN DARIN HOWE SEAN FULOP U of Calgary CSU-Fresno Contact information: Darin Howe Department of Linguistics University of Calgary [email protected] 2500 University Drive NW Calgary, Alberta Canada T2N 1N4 Telephone: 403-220-6110 Fax: 403-282-3880 Sean Fulop Department of Linguistics California State University, Fresno [email protected] 5245 N. Backer Avenue M/S PB92 Fresno, CA 93740-8001

Transcript of ACOUSTIC FEATURES IN ATHABASCAN U of Calgary CSU-Fresno

ACOUSTIC FEATURES IN ATHABASCAN

DARIN HOWE SEAN FULOP

U of Calgary CSU-Fresno

Contact information:

Darin Howe Department of Linguistics University of Calgary

[email protected] 2500 University Drive NW Calgary, Alberta Canada T2N 1N4 Telephone: 403-220-6110 Fax: 403-282-3880

Sean Fulop Department of Linguistics California State University, Fresno

[email protected] 5245 N. Backer Avenue M/S PB92 Fresno, CA 93740-8001

1

ACOUSTIC FEATURES IN ATHABASCAN*

Abstract: The Jakobsonian features [grave] (redefined as the audible presence of significant low

frequency noise) and [flat] are shown to be essential in working out extravagant histori-cal shifts undergone by sibilants in the Athabascan family of languages. The import of such acoustic features to diachronic explanation on one hand, and their enduring irrele-vance to synchronic explanation on the other (SPE et seq), jointly affirm (with Ohala and Blevins) that sound change has its source in phonetics, not phonology (pace Kipar-sky), and that phonetics and phonology are distinct modules, each with its privileged principles and elements, including features (pace Steriade and Flemming). [100 WORDS]

1 Introduction.................................................................................................................................. 1 2 Audition of acoustic features. ..................................................................................................... 4 3 The grave course of Northern Athabascan thibilants. ................................................................ 9

3.1 From thibilants to laterals................................................................................................... 12 3.2 From thibilants to labials.................................................................................................... 13 3.3 From thibilants to labiovelars............................................................................................. 14 3.4 From thibilants to velars..................................................................................................... 17 3.5 From thibilants to emphatics. ............................................................................................. 19

4 The strange fate of Athabascan rounded shibilants. ................................................................. 21 5 Conclusion. ................................................................................................................................ 29 6 References.................................................................................................................................. 33

1 INTRODUCTION. In this article we reflect on sound shifts in Athabascan which have received

surprisingly little attention for their venerable documentation by the likes of Boas, Sapir, Li,

Hoijer, Krauss, Golla, Kari, Leer, and Rice. Of particular interest are the rather spectacular

* Research for this paper was supported by a grant from the Social Sciences and Research Council of

Canada (CE101-3663). For their assistance with Canadian language data, we are especially grateful to Elizabeth

Enfield (AB: Dene Tha), Mary Koyina Richardson (NT: Tłįchǫ Yatiì), Maria Myers (BC: Tsilhqot’in), Horace

Adams (AB: Dëne Sųłiné), Julius Park (SK: Dëne Sųłiné), Louise Beaulieu (NT: Dëne Sųłiné), Keith and Mary

Kunnizzie (NT: Gwich’in). Thanks also to Mike Krauss, Joyce McDonough, Sally Rice, Betty Harnum, Robert

Murray, Mike Dobrovolsky, Chris Golston and our audience at LSA 2005 (Oakland) for helpful discussions.

2

(1) *saˑxʲ

‘sand’ *-tsʰiʔ ‘head’

*-ts’ən ‘bone’

Dena’ina suj -tsʰi t s’ən

Koyukon ɬɔjx -tɬʰi t ɬən

AK-Gwich’in xih -kʰìʔ -t θ’an

Dene Tha θa -t θʰí -t θ’iəne

Tulita-Slavey fa -pʰiʔ -p’ɛneʔ

Radeli-Slavey wa -f(ʷ)iʔ -wɛnɛʔ

Tłįcho Yatiì ʍa -kʷʰì -kʷ’õ

changes undergone by sibilants in Northern members

of this family. (The languages we discuss are mapped

in (2) below.) The cognates in (1), for instance, illus-

trate that the alveolar sibilants of Proto Athabascan

(PA) —Sapir’s PA series I— have evolved into later-

als in Koyukon, into velars or dentals in Gwich’in

(AK), into dentals in Dene Tha (South Slavey), into

labials in Tulita-Slavey, into labials or labiovelars in

Radeli-Slavey, into labiovelars in Tłįchǫ Yatiì, and

into emphatics in Tsilhqot’in. Tsilhqot’in t s h isaj -t s h í -t sɛn

(2) Some Northern Athabascan languages (those discussed in this article)

3

Another class of intriguing sound changes is seen in (3): PA rounded shibilants (labialized

palatoalveolar sibilants) endure to some extent in Hupa, but have become unrounded in Navajo

(among others), retroflex in Gwich’in (among others), and labiodental in Ts’ets’aut.

(3) Hupa Navajo Gwich’in Ts’ets’aut

*-tʃʷʰəʁ ‘weep’ t ʃʷʰiw -tʃa(h) -ʈɻʰeˑ -pfa *t ʃʷ’əj ‘blow’ -tʃ’e -tʃ’i -ʈɻ’ai -pf’ɛ *ʃʷaˑ ‘sun’ ʍaː ʃá ʂɻiˑ faʔ, fʷa *ʒʷəŋʲə ‘pupil/black’ -ʍiŋ -ʒi ː ʔ -ʐɻej -fʷa

In this article we restore erstwhile suggestions (Tharp 1972, Howren 1975) that acoustic

features are necessary in understanding radical articulatory changes such as those in (1) and (3).

However, we find that such features are defined somewhat unsatisfactorily and even incorrectly

in the (phonological) literature (ibid., Jakobson et al. 1952 et seq.). We show that realigning the

definition of relevant acoustic features —[flat] and especially [grave]— to phonetic reality im-

proves their fit with Athabascan sound shifts. Specifically, we first offer a critical review of

acoustic features (§2), which we then use to explain the development of PA sibilants (§3) and

PA rounded shibilants (§4).1 We conclude that sound change has its source in phonetics, which

(unlike phonology) must include acoustic features such as [grave] and [flat] (§5).

1 Sources are as follows (in alphabetic order): Dakelh (Poser 1999); Deline-Slavey: Howren (1975), Rice

(1989), Masuzumi et al. (1994: Apx B, p. 4ff.); Dena’ina: Kari (1977a); Dene Tha (South Slavey): Rice (1989),

Howard (1990); Gwich’in: Leer (1979, 1996), Krauss and Leer (1981), Firth and Mitchell-Firth (2003); Hupa:

Golla (1964, 1977, 1996); Koyukon: Thompson (1991), Jetté and Jones (2000); Mattole: Golla (1964), Landar

(1977); Navajo: Young and Morgan (1992); Proto-Athabascan: Leer (1979, 1996), Krauss and Leer (1981), Krauss

(1982), Young and Morgan (1992); Radeli-Slavey: Howren (1975), Rice (1989); Tahltan: Nater (1989); Tanacross:

Holton (2000); Tłįchǫ Yatiì: Howren (1975), Saxon and Siemens (1996); Tolowa: Collins (1989); Ts’ets’aut: Boas

4

2 AUDITION OF ACOUSTIC FEATURES. The acoustic features [grave] and [flat] of the Jakobson-

Halle system (Jakobson et al. 1952, Jakobson and Halle 1956) were in wide use until the advent

of Chomsky and Halle’s (1968) articulatory feature system, which proved superior in character-

izing possible contrasts, natural classes, and processes in phonology (e.g., Trask 1996:188). In

spite of sporadic pleas for their retention —most notably Lass (1976:197-207) on [grave] and

Ohala (1985) on [flat]— neither feature now has any role in mainstream phonology.2 We do not

oppose the exclusion of [grave] and [flat] from current theories of synchronic phonology, but

we find these features to be important in diachronic explanation. In particular, we will argue in

§3 and §4 that both features are sine qua non in working out sound change in Athabascan. To

this end, we now revisit each feature in turn.

The feature [grave] was included in the Jakobson-Halle system to account for an appar-

ent natural class of speech sounds having no articulatory basis. For instance, [grave] is invoked

to rationalize shifts between labial and velar fricatives, notably in the histories of English and

Spanish.3 An Athabascan example is provided by the dialect of North Slavey spoken in the

Radeli district of the Northwest Territories: “There is variation among speakers between [f] and

[x]” (Rice 1977:321).

and Goddard (1924), Tharp (1972); Tsilhqot’in: Krauss (1975), Cook (1983, 1993); Tulita-Slavey: Howren (1975),

Rice (1989); Tututni: Golla (1976), Landar (1977).

2 [grave] survives (transmuted) in a few modern theories, e.g., as a putative articulatory feature [periph-

eral] in some varieties of feature geometry (e.g., Rice 1994) and as the element |u| in Dependency Phonology

(Durand 1986, Ewen 1995).

3 [grave] is typically not a sufficient condition. English x > f apparently depended on preceding labializa-

tion (laugh, cough, trough, enough, etc.). It may also have been blocked by tautosyllabic labials (Baugh, bough,

etc.) and front vowels (nigh, sigh, high, etc.), and there were many other exceptions ((al)though, th(o)rough, etc.).

5

(4) Interspeaker variation in Radeli-Slavey (Rice 1977:321)

fori ~ xori ‘quickly’ lifótõ ~ lixótõ ‘nine’ lífuʃé ~ líxuʃé ‘fork’ (fr. French la fourchette) fawéhgewe ~ xawéhgewe ‘Old Baldy’ (place name)

Jakobson et al. stated that a sound is [grave] when it contains a predominance of low frequency

spectral energy over high frequency energy. The authors suggested using a statistical skewness

metric to measure the acoustic property, so that a sound with a leftward (negative) spectral

skewness would be [grave]. Their application of this idea to their own examples, however,

seems to be mistaken. For example, they state that [f] is [grave] in view of its negative spectral

skewness, but a quick check of this proves it wrong —[f] in fact has a somewhat positive

skewness in general. On the other hand, [ʃ] is said to be nongrave in view of its positive skew-

ness, but again a quick check of this proves it wrong —[ʃ] often has a quite negative skewness.

One can only speculate about what limitations of the technology of the era led Jakobson et al.

to report erroneous acoustic facts. At any rate, there seems to be a problem with the acoustic

basis in spectral skewness for this purported “acoustic” feature.

Perhaps recognizing this, Ladefoged (1997) exhumed [grave] without ceremony, but re-

defined this feature as the presence of low frequency noise (aperiodic) energy in a sound, re-

gardless of any overall statistical metric such as skewness. As it stands this definition is hard to

apply, since every sound has some degree of low frequency noise, however small. Given that

an acoustic feature must really be an auditory feature, in adopting Ladefoged’s suggestion we

propose to define [grave] as the audible presence of significant low frequency noise in a sound.

In particular this means the low frequency noise (< 3 kHz) must not be auditorily masked by

predominant high frequency noise.

6

With this new definition, the traditional natural classes of grave and nongrave sounds

emerge correctly from the acoustic facts as measured by a simple Fourier power spectrum.

Sibilants are normally not grave because they have much louder high frequency noise peaks

than their (nevertheless quite present) low frequency noise peaks, which are therefore masked

auditorily, e.g.:

(5) [s] from Dene Tha sah ‘bear’ (6) [ʃ] from Dene Tha t ʃʰaˑlʔiʔ ‘Charlie’

Inte

nsity

(dB

)

Frequency (kHz)

–10

0

10

20

30

40

2 4 6 8 10 12 14 16

Inte

nsity

(dB

)

Frequency (kHz)

0

10

20

30

40

50

2 4 6 8 10 12 14 16

Labial [f] and interdental [θ], with their relatively even spread of noise energy across a wide

frequency range, are then [grave] because their high frequency noise fails to mask the low fre-

quency noise, although it may be as loud or slightly louder, e.g.:

(7) [f] in English ‘fan’ (8) [θ] in Dene Tha t θʰah ‘carrot’

Inte

nsity

(dB

)

Frequency (kHz)

–30

–20

–10

0

10

20

2 4 6 8 10 12 14 16

Inte

nsity

(dB

)

Frequency (kHz)

–30

–20

–10

0

10

20

2 4 6 8 10 12 14 16

7

Labiovelars such as aspirated [ʍ] and labiovelarized consonants turn out to be very strongly

[grave], having a great preponderance of noise in the lower frequencies, e.g.:

(9) [ʍ] in Tłįchǫ Yatiì ʍekõ ‘dry’ (10) [w’] in Tłįchǫ Yatiì kʷ’ẽ ‘sinew’

–10

0

10

20

30

40

Inte

nsity

(dB

)

2 4 6 8 10 12 14 16Frequency (kHz)

Inte

nsity

(dB

)Frequency (kHz)

–10

0

10

20

30

40

2 4 6 8 10 12 14 16

Velar and uvular fricatives, as well as lateral fricatives, also show themselves acoustically to be

[grave] by this definition, e.g.:

(11) [ɬ]: Tłįchǫ Yatiì t ɬʰĩˑt ʃõ jatʰì (12) [ɬ]: Slav. t ɬ’áˑθɛh ‘pants’

–20

–10

0

10

20

30

Inte

nsity

(dB

)

Frequency (kHz)2 4 6 8 10 12 14 16

Inte

nsity

(dB

)

Frequency (kHz)

–10

0

10

20

30

40

2 4 6 8 10 12 14 16

Turning now to the feature [flat], it is defined —properly, it seems— as a “downward

shift of a set of formants” (Jakobson et al. 1952:31, see also Trubetzkoy 1939:127ff.). Like

[grave], this feature has diverse articulatory exponents in speech: lip-rounding, retroflexion, and

pharyngealization. American English ɚ is thus archetypal of [flat], to the extent that it involves

8

all three of these articulations (Ohala 1985). However, it is important to observe that flat articu-

lations are always secondary, in Stevens and Keyser’s (1989) sense. That is, unlike [grave],

[flat] is not a primary feature but serves only to enhance sounds (see Ohala 1985 for a lucid

discussion of this point). It is perhaps for this reason that (unlike with [grave]) there is re-

markably little evidence for the utility of [flat] in explaining substitutions, either synchronic or

(13) Arabic Argobba

sɑbijj sʷabijj ‘baby’ sɑdɑqɑ sʷadaqa ‘alms’

diachronic. Perhaps the best known case of substitution

around [flat] is Ohala’s (1985:225) in (13): Arabic pharyn-

gealized segments are borrowed as labialized segments in

the Ethiopian Semitic language Argobba (Leslau 1957). cf. risɑlɑ risɑːlɑ ‘letter’

Unfortunately, as a loan phenomenon, the Argobba pattern is strictly-speaking neither

synchronic nor diachronic. Moreover, it does not even argue compellingly for the functional

equivalency of [flat] across languages, since Ohala is careful to point out that the source seg-

ments may well have been phonetically labialized in the first place: “in some dialects of Arabic

the pharyngealized or “emphatic” consonants are accompanied by lip rounding” (ibid.).4

Shosted (2004), too, warns against exaggerating the role of [flat] in cross-language patterns like

Argobba (13). He reports that in spite of their acoustic similarity (viz., flatness) pharyngealiza-

tion and retroflexion never substitute for each other in the adaptation of numerous Arabic

loanwords into Hindi, and he goes on to demonstrate experimentally that Hindi speakers gener-

ally perceive Arabic emphatics as plain (rather than retroflex).5

4 By comparison, no significant labialization accompanies the Arabic pharyngeals [ħ, ʕ], and indeed no

labialization is involved in their adaptation into Argobba, as [h, ∅] (Leslau 1990).

5 On the other hand, Arabic listeners in Shosted’s experiment were relatively more likely to categorize

Hindi retroflexes as emphatics.

9

In the remainder of this article we consider a wide variety of consonantal sound shifts in

the Athabascan family of languages from the perspective of [grave] and [flat], as (re)defined

above. We argue most changes to be based on [grave] (like (4) above), but a few are found to

depend crucially on [flat]. The following table summarizes the relative gravity, and also the flat-

ness, of all the sounds involved in the Athabascan changes under discussion.

(14) Features: articulatory acoustic

sibilant t s, t sʰ, t s’, s, z [cor, +ant, −dist] palatoalv. t ʃ, t ʃʰ, t ʃ’, ʃ, ʒ [cor, −ant, +dist]

dent. sibil. t s, t s h , t s’, s, z [cor, +ant, +dist]

dental t θ, t θʰ, t θ’, θ, ð [cor, +ant, +dist] lateral t ɬ, t ɬʰ, t ɬ’, ɬ, ɮ/l [cor, +lat] [grave] labial pf, pfʰ, pf’, f, v [lab] velar k, kʰ, k’, x, ɣ [dor]

phar. sibil. t s, t s h , t s’, s, z [cor, −ATR] retroflex ʈɻ, ʈɻʰ, ʈɻ’, ʂɻ, ʐɻ [cor, −ant, −dist]

[flat]

lab. palat. t ʃʷ, tʃʷʰ, t ʃʷ’, ʃʷ, ʒʷ [cor, −ant, +dist, +rd]labiovelar kʷ, kʷʰ, kʷ’, ʍ, w [dor, +rd]

[grave, flat]

3 THE GRAVE COURSE OF NORTHERN ATHABASCAN THIBILANTS. Proto-Athabascan (PA) distin-

guished no less than five series of obstruents in the small area around the alveolar ridge (Leer

1979:4, Krauss and Leer 1981:190, Krauss and Golla 1981:71, Krauss 1982:73, Cook and Rice

1989:6-7): stops (15a), laterals (15b), sibilants (15c), shibilants (15d) and rounded shibilants

(15e).6 As described in §4 below, this system was simplified by historical mergers of the two

6 Many Athabascanists now consider this series to have been labialized in Pre-PA but retroflex in PA

(e.g., Leer 1996, in press:278, Rice 1998, 2004, Poser 2004). This is immaterial to our immediate discussion, but

we maintain the older view that labialization had not yet shifted to retroflexion in PA, since (as we detail in §4)

10

(15) PA coronal obstruents

a. t, tʰ, t’ b. t ɬ, t ɬʰ, t ɬ’, ɬ, ɮ c. t s, t sʰ, t s’, s, z d. t ʃ, t ʃʰ, t ʃ’, ʃ, ʒ

shibilant series (15d,e) in the Apachean and Pacific Coast

branches of Athabascan (see, e.g., Young 1983:394-6). However,

such a merger did not occur generally in Northern Athabascan,

as attested by Deg Xinag, Upper Kuskokwim, Lower Tanana,

Hän, Gwich’in, and Ts’ets’aut (see §4 for details), suggesting e. t ʃʷ, tʃʷʰ, t ʃʷ’, ʃʷ, ʒʷ

that Proto Northern Athabascan had all five series of coronal obstruents in (15). In such a con-

gested system it is reasonable to assume, as Leer (1996:197) does, that the sibilant series (15c)

was phonetically pre-alveolar, that is, dental [t s, t s h , t s’, s, z].7 Leer argues that such an articula-

tion eventually stabilized into a flattened dental series /t θ, t θʰ, t θ’, θ, ð/ in all Northern Atha-

bascan languages (henceforth NA: Krauss and Golla 1981, Rice 1998), except Denai’na (and

perhaps Ahtna), setting the stage for a chain reaction he calls the Great Northern Series Shift:8

(16) Great Northern Series Shift

“Sibilants become thibilants (interdentals) and shibilants become sibilants. …

Palatal onsets become shibilants and uvular onsets become velars.” (Leer 1996:197).

labialized shibilants endure in the Pacific Coast language Hupa (Golla 1996) and they shifted to a labial series in

the extinct Northern language Ts’ets’aut (Boas and Goddard 1924).

7 Thus in the Pacific Coast Athabascan language Tututni where reflexes of the PA sibilants survive in op-

position to four other coronal obstruent series (“apico-alveolar”, “lateral”, “flat”, and “palatal”; Golla 1976:218),

sibilants indeed “tend toward a forward (dental or interdental, lisped) articulation” (ibid.) and Landar (1977) even

(variably) transcribes them as interdental.

8 To give just one example, consider Tahltan (Nater 1989):

*sibilants > thibilants *shibilants > sibilants *velars > shibilants *uvulars > velars

*-tsʰəŋ ‘meat’ > -t θʰén’ *t ʃʰaʔ ‘beaver’ > tsʰaʔ *kʰaːn ‘rain’ > t ʃʰa *qʰuy ‘vomit’ > -kʰuː *-tsʰə ‘head’ > -t θʰíʔ *t ʃḛːχ ‘gum’ > t seːh- *kʰeʔ ‘tail’ > -t ʃʰéʔ *q’aʔ ‘arrow’ > k’aʔ *saˑxʲ ‘sand’ > θaː *-tʃʰĩːxʲ ‘nose’ > -tsʰíh *təkʰən ‘stick’ > tetʃʰən *ʁaʔ ‘hair’ > -ɣáʔ

11

On the assumption that it was encouraged by the overcrowded coronal obstruent system of

Proto NA, we interpret sibilants > thibilants as a systemically functional shift. The subsequent

shifts, viz. sibilants < shibilants < velars < uvulars, were evidently drag chains.

The fronting of NA sibilants is illustrated in (17) with Deg Xinag of Alaska, Dakelh of

British Columbia, and Dëne Sųłiné, spoken in the Northwest Territories, Alberta, Saskatchewan

and Manitoba. Note that interdentals reverted to being dental sibilants in early 19th century Da-

(17) Deg Xinag

South Dakelh

Dëne Sųłiné

*-tsʰəŋ ‘meat’ -t θʰəŋ -t s h ʌŋ -t θʰən *t sʰəŋʲəɬ ‘stone-axe’ t θʰəŋəɬ t s h et s h eɬ t θʰɛɬ *-tsʰə ‘head’ -t θʰeʔ t s h i -t θʰí *t s’ən ‘bone’ t θ’ən t s’ʌn t θ’ən *t s’iˑxʲ ‘mosquito’ t θ’ih t s’ih -- *t s’uː ‘breast’ -- t s’u t θ’ù *t ʃ’əwə ‘spruce’ t s’əvə- t s’u t s’u *səm ‘star’ -t θʰen sʌm t θʰən *-tʃʷʰəχs ‘whip’ -ʈɻʰoθ -tsʰəs t θʰaθ10 *q’a ˑ zəs ‘quiver’ k’ɔːθ k’as kʼɑkʼɑθ *jəχs ‘snow’ jəθ jʌs jaθ *-zət ‘liver’ -- -zʌs -ðəɹ

kelh, and relatively few speakers now

persevere in distinguishing them from

the nondental ones derived in the

meantime from the chain shift de-

scribed above in (16) (Story 1984:12-

3, Poser 2003:18-9).

Our overriding interest lies not

in the establishment of a thibilant se-

ries in NA,9 but rather in the remark-

able subsequent changes to this novel

series in individual languages. *-ɲḛˑz ‘long’ -ŋaθ -jiz -nɛð

9 The historical fronting of sibilants to thibilants is somewhat banal. In North America alone, one also

finds *s, *t s, *ts’ > θ, t θ, t θ’ in Salish (Halkomelem, Pentlatch, Straits, Comox-Sliammon) (Blake 2000); *s, *z

> θ, ð in Morley Stoney (Shaw 1980); *s > θ in Algonquian (Mahican, dialectal Munsee Delaware, Shawnee,

Kickapoo-Sauk-Fox), Siouan (Iowa-Oto, Degiha), Iroquoian (dialectal Tuscarora), Muskogean (Choctaw, Chicka-

saw), and Hokan (Karuk, various Yuman) (Mithun 2001).

10 The initial thibilant here is due to regressive sibilant assimilation.

12

As sneak-previewed in §1, thibilants have variously shifted to laterals, labials, labiovelars, ve-

lars, and emphatics. We review these changes in turn below, and in each case we implicate the

acoustic features discussed in §2, in particular [grave].

3.1 From thibilants to laterals. The first change to NA thibilants which we claim to be audito-

rily based is the merger of this series with laterals which occurred in the Alaskan language Ko-

(18)

Tanacross Koyukon

Dëne Sųłiné

*t səɬ ‘mountain’ t θɛɬ t ɬəɬ -- *t sʰəŋ ‘flesh’ -t θʰĩʔ t ɬʰən -t θʰən *t sʰiʔ ‘head’ -t θʰiʔ t ɬʰi -t θʰí *-tsʰuˑjə ‘grandchild’ -t θʰui t ɬʰuj -t θʰuɛ *-ts’ən ‘bone’ t θ’ɛn t ɬən t θ’ɛn *t s’a ˑ kʲ’ ‘dish’ t θ’áːk t ɬ’ɔk t θ’aj *saˑxʲ ‘sand’ θaːj ɬɑx t θʰaj *xʲəs ‘hill’ ʃɛθ suɬ ʃɛθ *-luˑs ‘drag’ -luːθ -luɬ -luθ *-zaʔ ‘mouth’ -ðáː -lɔ -ðá

yukon (Krauss and Leer

1981, Jetté and Jones 2000),

i.e., PA *t s, *t sʰ, *t s’, *s, *z

> NA *t θ, *t θʰ, *t θ’, *θ, *ð

> Koyukon tɬ, t ɬʰ, tɬ’, ɬ, l.

The cognates in (18) illus-

trate that laterals in Koyukon

substitute for interdentals as

found in Tanacross and Dëne

Sųłiné, for instance. *q’ʊs ‘neck’ -k’oθ q’ʊɬ -kʷ’òð

As explained above in §2 we consider thibilants to be [grave]. Crucially, so too are lat-

erals —in fact, relatively more so, since their spectral peaks are actually louder at lower fre-

quencies (see, e.g., (11) and (12) above). We accordingly opine that the merger of thibilants

with laterals in Koyukon was encouraged system-internally not only by the much-needed sim-

plification it brought to the coronal obstruent system (cf. (15)) but also by the fact that for lis-

teners (Ohala 1981, 2003) it effectively preserved (and perhaps even enhanced) the [grave] fea-

ture of thibilants. We also believe this acoustic commonality of laterals and thibilants to be

pivotal in minor shifts elsewhere, such ɬ > θ in dialectal Dëne Sųłiné, e.g., ɬekən > θekən

13

‘sweet’ (Cook 1997). The latter is a change in progress which is evidently part of a chain shift,

since the relevant dialects also show (counterfeeding) θ > h, e.g., θeda > heda ‘she/he/it is

sitting’ (ibid.). As Cook (1997:26) states, “the diachronic change indicated by these alternations

is: ɬ > θ > h.”. Apparently akin changes are reported beyond Athabascan, too, notably *l > ð

in Cree as spoken in northeastern Saskatchewan and northwestern Manitoba (Goddard 1980).

In the next few sections we argue that thibilants have expressly changed to other [grave]

sounds in other Athabascan languages, but no merger was entailed in these cases.

3.2 From thibilants to labials. NA thibilants endure in the South Slavey dialect Dene Tha,

spoken from northwest Alberta to northeast British Columbia and into southern regions of the

(19) Dene Tha Tulita

*-ts... ‘dryfish’ ʔeht θáa ehpaˑ *-ts...t ‘stay’ --11 -peʔ *t sʰiʔ ‘head’ -t θʰíʔ -pʰiʔ *t sʰəŋ ‘flesh’ -t θʰɛn -pʰɛ12 *-tθ’ən ‘bone’ -t θ’iəne -p’ɛneʔ *t s’iˑxʲ ‘mosquito’ t θ’ih p’íh, p’ĩɛ *ts’əχ ‘sinew’ -t θ’éhé p’é/p’ɛh *səŋ ‘star’ θe fẽʔ *saˑxʲ ‘sand’ θa fa *sa? ‘tent pole’ θa fa *s(ə)- PERF θɛ- fɛ- *-zeˑ ‘belt, hide’ -ðɛh -vɛh *-zaʔ ‘mouth’ -ðáʔ -va

Northwest Territories (Rice 1989, Howard

1990). However, as illustrated in (19), the

interdental series has become labial in the

most closely related dialect spoken in the

Tulita district of the Northwest Territories,

that is, PA /*t s, *t sʰ, *t s’, *s, *z/ >

NA/Early Slavey /*t θ, *t θʰ, *t θ’, *θ, *ð/ >

Tulita /p, pʰ, p’, f, v/ (Tharp 1972, Howren

1975, Rice 1989).

While Tulita-Slavey (“p dialect”,

GNWT 1990) is probably its most impres- *-zət ‘liver’ -ðeʔ -ve

11 Dëne Sųłiné -(t)θəɹ ‘stay’ (Howren 1975:583).

12 E.g., ʔɛpʰɛ ‘caribou’ —arguably the most important game animal.

14

sive examplar to date, the substitution of thibilants by labials is relatively familiar, being found

in dialectal English, notably Cockney (e.g., Sivertsen 1960). There is also consensus that th

fronting (as it is widely known) is “due to perceptual similarity” (Blevins 2004:265) and that

this similarity is in fact extreme:

Studies of infant non-linguistic speech discrimination ..., of child linguistic dis-crimination ..., and of perceptual confusions among older children and adults ..., all indicate that f/θ is always among the most difficult contrasts to discriminate, and often the most difficult. (Velleman 1988:222, references removed)

The basis of this similarity is rarely explicated, but we identify it clearly: thibilants and

labials are not only similarly quiet but similarly [grave], as discussed in §2. We propose that

this correspondence in [grave] was the precise basis for the whole-class substitution of inter-

dentals by labials in Tulita-Slavey. External factors aside, it is also tempting (if precarious) to

interpret this change as functional (see, e.g., Martinet (1955) on “marge de sécurité” (p. 47-8)

(20) Coronal obstruents in Dene Tha/Early Slavey

t θ, t θʰ, t θ’, θ, ð (<*t s, *t sʰ, *t s’, *s, *z) t, tʰ, t’ t ɬ, t ɬʰ, t ɬ’, ɬ, ɮ t s, t sʰ, t s’, s, z (<*t ʃ, *t ʃʰ, *t ʃ’, *ʃ, *ʒ/*j)

and “attraction du système” (p. 80-1)):

Early Slavey, like modern Dene Tha,

had numerous series of coronal obstru-

ents (20) but, conspicuously, it had no

labial obstruents at all (Rice 1989).13 t ʃ, t ʃʰ, t ʃ’, ʃ, ʒ (< *kʲ, *kʲʰ, *kʲ’, *xʲ, *ɣʲ)

3.3 From thibilants to labiovelars. Thibilants have become labiovelars several times in NA.

Compare for instance the above-mentioned Dene Tha with the North Slavey dialect spoken in

the Radeli district (and partly in Colville Lake) of the Northwest Territories (Hare: Howren

13 Today’s Slavey dialects have p, mostly from loanwords, as well as mb (~ m<*w/*w).

15

(21) Dene Tha Radeli Tłįchǫ Yatiì

*-ts... ‘dryfish’ ʔeht θáː ʔéhwai ɛhkʷàa *t sʰəŋʲəɬ ‘axe’ t θʰĩh kohfĩ kokʷʰĩ *t sʰiʔ ‘head’ -t θʰíʔ -fíʔ -kʷʰì *t sʰeˑ ‘rock’ t θʰe fe kʷʰe *-ts’ən ‘bone’ -t θ’iəne wɛnɛʔ -kʷ’õ *t s’əχ ‘sinew’ -t θ’ɛhɛ wiɛʔ -kʷ’é t s’a ˑ kʲ’ ‘dish’ t θ’á wáʔ kʷ’à *t s’iˑxʲ ‘mosquito’ t θ’ih wih kʷ’i(h) *t s’atɬ’ ‘moss’ t θ’áʔ wa(h) kʷ’a *səŋ ‘star’ θe we -ʍo *saˑxʲ ‘sand’ θa wa ʍa *sa? ‘pole’ θa wa ʍa *s(ə)- perf. conj. θɛ- wɛ- ʍɛ- *-zeˑ ‘belt, hide’ eðɛh ʔewé ʔewé *-zət ‘liver’ -ðé -wé -wò

1975:583, Rice 1977, 1989), as

well as with its neighboring lan-

guage Tłįchǫ Yatiì (Dogrib:

Howren 1975, Saxon and Siemens

1996). As the cognates in (21)

show, a shift to labiovelars is es-

pecially transparent in Tłįchǫ

Yatiì: /*t θ, *t θʰ, *t θ’, *θ, *ð/ >

/kʷ, kʷʰ, kʷ’, ʍ, w/. In contrast,

Radeli (“the most innovative dia-

lect”; Krauss and Golla 1981:79)

shows a reduced series: /w, w, f/.

However, we assume (with Rice

*s-it- ‘1st duopl.’ θ-ít- w-ít- w-ìd-

1977) that the Radeli dialect was earlier like Tłįchǫ Yatiì before the application of various leni-

tions (kʷ, ʍ > w; kʷ’ > w; kʷʰ > f)14 since, as shown in (22) below, the same labiovelar se-

ries as Tłįchǫ Yatiì is found in the North Slavey “kw” dialect (GNWT 1990) spoken in Deline

(and partly in Tulita), crucially located between Tłįchǫ Yatiì and Radeli.

Finally, thibilants also became labiovelars in Gwich’in, but in a more restricted context.

As shown in (23a) below, stem-final θ/ð from *s/*z weakened to -oh/-uh as part of a general

14 According to Rice (1977), weakening —especially spirantization— is very general in Radeli-Slavey

(“f” dialect, GNWT 1990): “Historically, there seems to be little question that this segment [f] is /kw[h]/” (Rice

1977:321). We suggest (tentatively) that aspiration may have caused a lifting of the noise peak in the spectrum

after the release of [kʷʰ], yielding a more level spectrum, approximating [f].

16

(22) PA > Pre-Slavey > Deline dialect

*t sʰeˑ ‘rock’ > t θʰe > kʷʰe *-tsʰəŋ ‘flesh’ > -t θʰɛ > -kʷʰɛ16 *t sʰiʔ ‘head’ > -t θʰíʔ > -kʷʰíʔ *-ts’ən ‘bone’ > -t θ’iəne > kʷ’ɛnɛ *t s’əχ ‘sinew’ > -t θ’ɛhɛ > -kʷ’é *səŋ ‘star’ > θe > ʍe ʔ *sa? ‘pole’ > θa > ʍa *s(ə)- perf. conj. > θe- > ʍe-

pattern of coda weakening (Leer 1979:9). More-

over, as shown in (23b), later stem-final θ/ð

from PA affricates15 also weakened to -oː/-uː,

though not in the conservative Gwichyah dialect

spoken in Tsiigehtchic, Northwest Territories.

We suggest that (as part of final weakening) θ/ð

were replaced by ʍ/w, which coalesced with the

preceding vowel, making it round and back. *-zət ‘liver’ > -ðəʔ > -wéʔ

(23) Gwichyah Gwich’in

a. *səs θoh θoh ‘belt’ *jəχs ʒoh ʒoh ‘snow’ *-ɲḛˑz nd ʒuh nd ʒuh ‘long’ *χəz xaoh xoh ‘pus’

b. *-ʔḛˑt s -ʔìeθ -ʔjùˑ ‘step’ (perf.) *-t’aˑt s’ -t’ieθ -t’juˑ ‘cut’ (perf.) *-tʰeˑt ʃ’ -tʃiθ -tʃuˑ ‘lie’ (plur. perf.) *-ʔaˑt ʃ’ -ʔeθ -ʔoˑ ‘go’ (dual, perf.) *-wət ʃ’ -vað -voh ‘my cheek’

15 Leer (1996) argues that the shift to thibilants applied twice in the history of Gwich’in: first to PA sibi-

lants, e.g., (i), and later to sibilants derived from PA shibilants (via the above-mentioned drag-chain), e.g., (ii).

Early Pre-Gwich’in

Late Pre-Gwich’in

Gwichyah Gwich’in

i. *-t’aˑt s’ *-t’et θ *-t’eθ -t’ieθ ‘cut’ (perf.) *-ʔḛˑt s *-ʔìt θ *-ʔìθ -ʔìeθ ‘step’ (perf.)

ii. *-tʰeˑt ʃ’ *-tʲʰit s *-tʲʰiθ -t ʃiθ ‘lie’ (plur. perf.) *-ʔaˑtʃ’ *-ʔet s *-ʔeθ -ʔeθ ‘go’ (dual, perf.)

16 E.g., ʔɛkʷʰɛ ‘caribou’; cf. Tłįchǫ Yatiì ʔɛkʷʰo .

17

Returning to acoustic features, our recognition of thibilants as [grave] rationalizes their

development into labiovelars in the histories of Tłįchǫ Yatiì, Radeli-Slavey, Deline-Slavey, and

Gwich’in. It is difficult to imagine that listeners directly misapprehended thibilants as labiove-

lars, so we suggest that speakers expressly enhanced the [grave] feature of thibilants by adding

rounding, i.e. [flat] (*θ > θʷ > ʍ, etc.; compare esp. (8) with (9)/(10) in §2).17 Moreover, it is

again tempting to view this shift as having been encouraged systemically by the clogged cor-

onal obstruent system on one hand, and the absence of labiovelars on the other. Gwich’in alone

had labiovelar obstruents, but crucially it lacked [w], *w having been lost to adjacent vowels,

except syllable-initially where it strengthened to v, e.g., *tʃ’əwə ‘spruce’ > t s’iːviː.18

3.4 From thibilants to velars. We now describe two restricted cases in which velars substitute

for thibilants. The first comes from Dene Tha. As discussed in the preceding two sections, the

Northern thibilant series generally survives intact in this South Slavey dialect (Rice 1989,

Howard 1990),19 but Howren (1975) recounts that fricatives are replaced by velars on occasion.

(24) Dene Tha: Standard Wrigley

*s(ə)- PERF. θe- xe- *səŋ ‘star’ θe xẽʔ

*-zaʔ ‘mouth’ -ðáʔ -ɣáʔ

For instance, Howren ascribed the forms in (24) to

the chief of the Slavey Band at Fort Wrigley, NT, in

1970. His transcriptions were apparently borne out

by spectral analyses (not provided): *-zeˑ ‘skin’ -ðéh -ɣéʔ

17 A reverse change, labiovelar > dental, famously occurred before front vowels in Greek, e.g., *kʷis

‘who’ > tis, *penkʷe ‘five’ > pente, *gʷhe/ormos ‘warm’ > thermós, etc. We suggest that front vowels, being

acute, “disenhanced” the [grave] feature of labiovelars, rendering them acoustically comparable to dentals.

18 *w > mb/p in the other languages, for example, *weːʃʷ ‘knife’ > Tłįchǫ Yatiì mbɛh, Dene Tha mbɛh,

Deline pɛh, Radeli piɛ(h) (Krauss and Leer 1981:6).

19 Thibilants, like all nonnasal consonants, reduce to [h] syllable-finally (Saxon 1979, Rice 1989:93).

18

I … subjected several of the questionable utterances to acoustic analysis, using a Kay

Electric 6061A Sona-Graph. In general, the spectrograms —especially those of the voice-

less fricatives— confirmed my auditory impressions; the acoustic spectra of sounds in

question bear a marked resemblance to those of obviously velar articulation. Either they

are altogether velar, or they are dentals with a velar coarticulation. (Howren 1975:596)

While limited, this reported substitution of [x, ɣ] for [θ, ð] is expected on our view that velars

are perceptually similar to thibilants because they share a critical acoustic property: [grave].

The second case of velar substitution for thibilants is more systematic but less obvious.

It occurred in Gwich’in as spoken in Alaska (Leer 1979, 1996, Krauss and Leer 1981). The

forms below show that the historical series /*t θ, *t θʰ, *t θ’, *θ, *ð/ survives as such before back

vowels (25a), but has transformed in two ways before front vowels (25b): it became palatoal-

veolar /tʃ, t ʃʰ, t ʃ’, ʃ, ʒ/ in Canada, but velar /k, kʰ, k’, x, ɣ/ in Alaska.

(25) Gwich’in: Early Canada Alaska

a. *t səɬ ‘mountain’ *t θah? t θah t θah *-tsʰəŋ ‘meat’ -*t θʰàjʔ -t θʰa ĩʔ -t θʰa ĩʔ *t s’ən ‘bone’ *t θ’an t θ’an t θ’an *səs ‘belt’ *θaθ θoh θah *-zəkʲʰ ‘time elapses’ *-ðákʰ -ðákʰ -ðákʰ

b. *t sʰeˑ ‘rock’ *t θʰiˑ t ʃʰiˑ kʰiˑ *-tsʰiʔ ‘head’ *-t θʰìʔ t ʃʰìʔ kʰìʔ *-tsʰuˑjə ‘grandchild’ *-t θʰiuiː -tʃʰiː -kʰioiː

*t s’a ˑ kʲ’ ‘dish’ *t θ’èkʰ t ʃ’ìkʰ k’ìkʰ *t s’iˑ- ‘straight’ *t θ’iˑ- t θ’iˑ- k’iˑ- *saˑxʲ ‘sand’ *θejh ʃeih xih *-zaːt ‘be long’ *ðètʰ -ʒitʰ ɣìtʰ

19

The impetus behind both changes, we suggest, is that front vowels —as acute20— spoil the

gravity of thibilants preceding them due to consonant-vowel coarticulation (Hardcastle 1999).

The feature [grave] was simply conceded in Canadian Gwich’in, as thibilants became palatoal-

veolar under this coarticulation; see Bhat (1978) for numerous other such instances of

[+anterior] → [−anterior] / ____ [−back]. By contrast, the feature [grave] was preserved in the

Alaskan dialect, as speakers began replacing thibilants with velars —grave consonants which

are less prone than coronals to consonant-vowel coarticulation. While exceptional, such a “pre-

emptive measure” is consistent with our impression that NA speakers are sensitive to the fea-

ture [grave] and that they adjust their speech around this feature —syntagmatically in this in-

stance.

3.5 From thibilants to emphatics. In Tsilhqot’in, spoken in the Central Interior of British Co-

lumbia (Krauss 1975, Cook 1983, 1993), the thibilants of NA have become a pharyngealized

(26) Dëne Sųłiné Tsilhqot’in

*t sʰəŋʲəɬ ‘axe’ t θʰɛɬ t s h ĩɬ *t sʰiʔ ‘head’ -t θʰí -t s h í *t sʰəŋ ‘meat’ -t θʰən -t s h ə *t s’… ‘stay (pl)’ -t θ’i -t s’i *saˑxʲ ‘sand’ t θʰaj saj *s(ə)- perf. conj. θɛ- sɛ- *jəχs ‘snow’ jaθ jəs *-zeˑ ‘belt, hide’ -ðe si? *-zaʔ ‘mouth’ -ðá -zí

(“emphatic”) sibilant series, as demon-

strated by the cognates in (26). To explain

the formation of this [flat] series, we hy-

pothesize an intermediate stage of labiali-

zation in the history of Tsilhqot’in, i.e., PA

/*ts, *t sʰ, *t s’, *s, *z/ > NA /*t θ, *t θʰ,

*t θ’, *θ, *ð/ > Pre-Tsilhqot’in /t θʷ, t θʷʰ,

t θʷ’, θʷ, ðʷ/ > Tsilhqot’in /t s, t s h , t s’, s,

z/. On one hand, such labialization would

*ŋʲḛˑz ‘long’ -néð -nez

20 The Jakobson-Halle feature system posits [acute] as the opposite of [grave].

20

have preserved (and indeed strengthened) the [grave] feature of thibilants even as they reverted

to sibilants in modern Tsilhqot’in. On the other hand, labialization would critically bring in the

feature [flat]. Our suggestion that labialization may have arisen from thibilants is thus useful in

avoiding the arbitrary substitution of [grave] by [flat] in Tsilhqot’in.

We believe our hypothesis to be reasonable on the strength of our discussion of the in-

dependent development of labiovelars (*t θ, *t θʰ, *t θ’, *θ, *ð > *t θʷ, *t θʷʰ, *t θʷ’, *θʷ, *ðʷ >

kʷ, kʷʰ, kʷ’, ʍ, w) described in §3.3 and also given that labialization was contrastively present

elsewhere in Tsilhqot’in, e.g., kʷʰiʃkʷʰiʃ ‘stellajay’, kʷ’εlh ‘penis’, qʷ’εs/qʷ’εs ‘cloud’, naxʷ

‘twice’, -ʔʷεɬ ‘ swim’ (Cook 1983:125-7). Of course we also recognize that labialization is rela-

tively marked on coronals (cf. PA rounded shibilants in §4). Presumably, then, if lip-rounding

developed from thibilants in Tsilhqot’in (as we suggest) there would have been pressure to

shake off either the coronal articulation of those segments involved, or else their labialization.

The first scenario —loss of coronality— was played out in Tłįchǫ Yatiì, Radeli-Slavey, Deline-

Slavey, and Gwich’in as discussed in §3.3. The second set-up is the one we propose for

Tsilhqot’in: labialization was replaced by equally [flat] pharyngealization in coronals.

The latter move to pharyngealization was facilitated by factors both internal and exter-

nal to Tsilhqot’in. Internally, Tsilhqot’in has uvulars (e.g., bəq’uni ‘his throat’; Cook

1983:125), which crucially involve a pharyngeal gesture according to many theorists (Shahin

2002 and references therein). Indeed across languages “there is a common sound change of

uvulars to pharyngeals” (Blevins 2004:198), as seen, for instance, “in every branch of Semitic”

(ibid.), in Wakashan (Jacobsen 1969), and in Haida (Eastman and Aoki 1978). More pointedly,

uvulars pattern with emphatic coronals in triggering flattening consonant harmony in

Tsilhqot’in, e.g., t s’iqi [t s’əiqəi] ‘woman’ (cf. t s’isi [t s h əisaj] ‘sand’) (Cook 1983, 1993). Accord-

21

ingly Cook and Rice (1989:38) specifically identify the focus of this process as a “pharyngeal

feature associated with the uvular series and a dental series in Chilcotin” (Cook and Rice

1989:38).21

Externally, Tsilhqot’in is adjacent to Interior Salish languages which abound not only in

pharyngeals /ʕ, ʕ, ʕʷ, ʕ / but indeed in emphatics such as /ʃ, t ʃ, l, l , ɹ, ɹ / (Mattina 1979, Kuipers

1981, Bessell 1998). For instance, Bessell and Czaykowska-Higgins (1991:5-7) find 56 roots

with emphatics in the Southern Interior Salish language Nxa’amxcin, while Shahin (2002:183)

reports 172 such roots in the Northern Interior Salish language St’át’imcets. Thus Cook and

Rice (1989) describe pharyngealization in Tsilhqot’in as “an areal feature shared by Interior

Salish” (p. 39).

4 THE STRANGE FATE OF ATHABASCAN ROUNDED SHIBILANTS. This section shows that the

acoustic-auditory features [flat] and [grave] described in §2 shed vital light on the various

changes undergone by the rounded shibilants of PA. As a point of departure, the obstruent sys-

tems of Pre-PA —Krauss’ (1964) Proto-Athabaskan-Eyak— and of PA are presented in (27)

and (28), respectively, after Leer (1979:4), Krauss and Leer (1981:190), Krauss and Golla

(1981:71), Krauss (1982:73), and Cook and Rice (1989:6-7), among others. Focusing first on

21 As back dorsal consonants, uvulars are [grave], and to the extent that they involve a pharyngeal gesture,

they are also [flat]. It is perhaps unsurprising, then, that in the Pacific Coast Athabascan language Hupa (Golla 1996)

uvular obstruents have become [grave, flat] /w/, for example:

*ʁeːŋʲ ‘kill (tr.)’ > -weːn *t ʃḛːχ ‘pitch’ > -tʃeːw *-ʁuʔ ‘tooth’ > -wuʔ *t’aːχ ‘fly’ > -t’ɑw *-ʁaːtʰ ‘shake (out)’ > -wat *-neːχ ‘do (so)’ > -nəw *ʁa ‘fur’ > -waʔ *t ʃḛʁ ‘ear’ > -t ʃiwʔ

22

dorsals, it is standardly assumed that the velar series were phonetically fronted, apparently to

enhance their contrast with corresponding uvulars.22 As the arrows between (27) and (28) de-

pict, this phonetic fronting eventually resulted in the labialized velar series becoming palatoal-

veolar whereas the nonlabialized series remained velar, apparently to avoid a merger with the

preexisting shibilants (cf. Martinet 1955).

(27) Pre-Proto-Athabascan obstruents (28) Proto-Athabascan obstruents

t, tʰ, t’ t, tʰ, t’

t ɬ, t ɬʰ, t ɬ’, ɬ t ɬ, t ɬʰ, t ɬ’, ɬ, ɮ

t s, t sʰ, t s’, s t s, t sʰ, t s’, s, z

coro

nals

t ʃ, t ʃʰ, t ʃ’, ʃ

X t ʃ, t ʃʰ, t ʃ’, ʃ, ʒ

k, kʰ, k’, x [k, k h , k’, x]

t ʃʷ, tʃʷʰ, t ʃʷ’, ʃʷ, ʒʷ

coronals

kʷ, kʷʰ, kʷ’, xʷ [k , k ʰ, k ’, xʷ] kʲ, kʲʰ, kʲ’, xʲ, ɣʲ

q, qʰ, q’, χ q, qʰ, q’, χ, ʁ

dorsals

dors

als

qʷ, qʷʰ, qʷ’, χʷ

(q(ʊ), qʰ(ʊ), q’(ʊ), χ(ʊ))23 In time the fronted velars of PA, too, underwent coronalization in most daughter languages, but

only recently in some (e.g., Dëne Sųłiné (Krauss 1982), Dakelh (Poser 2004)) and only restrict-

22 Cf. Keating (1984) on “polarization”; also Ladefoged and Maddieson (1996:46). The phonetic use of

palatalization to distance velars from uvulars survives robustly in eastern Ahtna (Kari 1977b:284-5) as well as in

Hupa (Woodward 1964:200, see also Gordon 1996), and is commonplace elsewhere. For instance, it is an areal

feature of the Northwest Coast, affecting Wakashan (e.g., Lincoln and Rath 1980), Salishan (e.g., Nater 1984,

Blake 2000) and Tsimshian (e.g., Tarpent 1987, Dunn 1995) among others.

23 Krauss and Golla’s (1981) imply that PA retained a distinct labialized uvular series /qʷ, qʷʰ, qʷ’, χʷ/,

but all other sources assume that these Pre-PA segments had by then fissioned as /q(ʊ), qʰ(ʊ), q’(ʊ), χ(ʊ)/, with ʊ

replacing adjacent Pre-PA *ə.

23

edly in others (e.g., Ahtna (Kari 1978b), Deg Xinag (Kari 1978a)). Moreover, the historical ve-

lar series survives as such (along with uvulars) in more than a few modern languages, from

Alaska’s Denai’na (Kari 1977a) to California’s Hupa (Golla 1996).24

The latter (Hupa) is “a remarkably conservative … and … only surviving language of

the California Athapaskan subgroup” (Golla 1977:355) and it is of particular interest to us in

(29) Proto-Athabascan Hupa

*-tʃʷʰəʁ ‘weep’ -tʃʷʰeh *t ʃʷʰaˑn ‘excrement’ t ʃʷʰaŋʔ *t ʃʷʰətʃʷʰ ‘firewood’ t ʃʷʰitʃ

cf. *ʒʷəŋʲə ‘black’ -ʍiŋ *t’eˑʃʷ ‘charcoal’ t’ehʍ

(30) [t ʃʷʰ]: Hupa t ʃʷʰola ‘five’25

that it alone preserves the rounded shibilants

of PA in some measure (as we describe be-

low), e.g., (29). Consider the spectrum for

Hupa [tʃʷʰ] in (30). Crucially, peaks are

stronger at lower frequencies than at higher

frequencies in this spectrum, suggesting that

the rounded shibilant is not only [flat] (via

rounding) but also [grave] (see §2). We sug-

gest that rounded shibilants in PA were simi-

larly [flat] as well as [grave] since both

acoustic features played key roles in the de-

velopment of these segments in daughter lan-

guages. Frequency (Hz)

0 11025

-20

0

20

Let us consider first those languages in which the historical series was retained as a

separate yet altered series. On one hand, the feature [flat] but not [grave] was preserved as

rounded shibilants grew to be retroflex (i.e., *tʃʷ, *tʃʷʰ, *t ʃʷ’, *ʃʷ, *ʒʷ > ʈɻ, ʈɻʰ, ʈɻ’, ʂɻ, ʐɻ) in a

24 E.g., *-nəq’ ‘swallow (perf.)’ > -nəq’ vs. *-nḭkʲ ‘move hand (perf.)’ > -nikʲ.

25 We are grateful to Danny Ammon for a recording of this form.

24

handful of contiguous Northern languages (Krauss 1973, Leer 1979:4, Krauss and Golla 1981):

Deg Xinag, Upper Kuskokwim, Lower Tanana, Hän, and Gwich’in. This change is illustrated in

(31) with Gwich’in, which is spoken in the Northwest Territories, the Yukon, and Alaska.26 On

(31) Gwich’in Ts’ets’aut

*t ʃʷeːjəʔ ‘heart’ -ʈɻiˑʔ -pfɛˑ *-tʃʷʰəʁ ‘weep’ -ʈɻʰej -p fʰɛ *t ʃʷʰaˑn ‘excrement’ ʈɻʰɪnʔ pfˣa *t ʃʷʰətʃʷʰ ‘firewood’ ʈɻ ʰoˑ pfʷʰó *t ʃʷ’əj ‘blow’ -ʈɻ’ej -p f’ɛ *t ʃʷ’ḛN ‘calf (leg)’ -ʈɻ’oˑʔ -pf’u *t ʃʷ’əʒʷ ‘boil’ ʈɻ’oh apf’o *t ʃʷ’ətʃʷ’əʔ ‘kidney’ -ʈɻ’oˑ -pf’u *ʒʷəŋʲə ‘pupil/black’ -ʐɻej -fʷa *ʃʷaˑ ‘sun’ ʂɻiˑ f(ʷ)aʔ *-ʃʷḛʔ? ‘to call’ -ʂɻìʔ -fɛ

the other hand, the feature [grave] but

not [flat] was preserved as the historical

series became labiodental (i.e., *tʃʷ,

*t ʃʷʰ, *t ʃʷ’, *ʃʷ, *ʒʷ > pf, pfʰ, pf’, f, v)

in Ts’ets’aut, an extinct language of Brit-

ish Columbia (Boas and Goddard 1924,

Tharp 1972, Krauss 1973:944-7, 1977),

as also shown in (31).27 In comparison to

the Gwich’in change, that of Ts’ets’aut

seems immoderate, and it is tempting *xʲəʃʷ ‘bear’28 ʃoh fu

to ascribe it a functional teleology (cf. §3.2 and §3.3): Pre-Ts’ets’aut, like PA (28), had a sur-

plus of coronal obstruents but, conspicuously, it had no labial obstruents, aside from some p’s

derived from *w (e.g., -pɛ < -weːŋʲ ‘swim SG’).29

26 We interpret this change as categorical phonologically but only gradient phonetically, since PA rounded

shibilants may well have been enhanced by phonetic retroflexion (Michael Krauss, p.c.) and conversely, retroflex-

ion in Deg Xinag et al. may be enhanced by lip rounding.

27 A few forms such as ‘firewood’, ‘pupil’ and ‘sun’ display vestigial rounding, i.e., [flat].

28 Stem-initial *xʲ > ʃ is regular in Gwich’in, as is stem-final *əʃʷ> oh (Leer 1996). The development of

Ts’ets’aut is obscure in this form; possibilities include *xʲəʃʷ > (xʲə)ʃʷə > fu, and *xʲəʃʷ > ʃəʃʷ (coronalization)

> ʃʷə(ʃʷ) (regressive sibilant assimilation and coda attrition) > fu. Huld (1983) argues for PA *ʃʷəʃ ‘bear’, which

seems a better fit for the Ts’ets’aut reflex.

25

We now turn to those languages in which the

rounded and unrounded shibilants of PA have merged

into a single series. The Northern language Dena’ina

and the Apachean language Navajo, both exemplified

in (32) on the next page, are conventional in that all

shibilants —whether historically rounded or not—

have collapsed into a single unmarked series: /tʃ, t ʃʰ,

t ʃ’, ʃ, ʒ/. PA shibilants were also amalgamated in the

Pacific Coast branch, but in more interesting ways. On

one hand, the historical merger resulted in a uniformly

retroflex series in the Oregon subgroup (Golla

1976:219), as Tolowa illustrates in (32).30 On the other

hand, this merger has yielded a mixed series in the

Californian language Hupa: /*t ʃʰ, *ʃ, *ʒ/ and /*t ʃʷʰ,

*ʃʷ, *ʒʷ/ have merged as labialized /tʃʷʰ, ʍ/ (32a,b),31

whereas /*t ʃ, *tʃ’/ and /*t ʃʷ, *t ʃʷ’/ apparently merged as non-labialized /tʃ, tʃ’/ (32c,d). In fact,

however, there is evidence that ejective shibilants, too, had earlier been obligatorily rounded

29 Some *w’s apparently went to m instead, e.g., *wən ‘lake’ > mɛː.

30 Nonejective affricates spirantize in Tolowa —another “Oregonian hallmark” (ibid.).

31 The obligatoriness of rounding in earlier Hupa shibilants is confirmed by *xʲ > ʃ → ʃʷ → ʍ, e.g.:

*nətʃʰĩːxʲ ‘nose’ > -nt ʃʷʰɪʍ Compare: Outer Inlet Dena’ina: -ntʃʰiʃ *q’ɑjʔ-lixʲ ‘willow’ > q’ajlɪʍ Outer Inlet Dena’ina: q’əjliʃ *tənəxʲ ‘bearberry’ > tənəʍ Upper Inlet Dena’ina: tənəs (<tənəʃ) *xʲən ‘song’ > ʍɪŋ Dëne Sųłiné: ʃən

cf. *q’əʃ ‘alder’ > q’ɪʍ

26

(Golla 1970:29),32 suggesting that the historical merger of PA shibilants had uniformly favored

the rounded series in archaic Hupa.33

(32) Dena’ina Navajo Tolowa Hupa

a. *t ʃʷʰətʃʷʰ ‘firewood’ t ʃʰətʃʰ t ʃʰiʒ ʂɻəʂɻ t ʃʷʰit ʃ *-tʃʷʰəʁ ‘weep’ -tʃʰəʁ -tʃa(h) -ʂɻaʔ t ʃʷʰɪw *t ʃʷʰaˑn ‘excrement’ t ʃʰun t ʃʰãːʔ ʂɻan- t ʃʷʰaŋʔ *-tʃʷʰət ‘push, shove’ -- -tʃʰit -ʂɻət -tʃʷʰit *nət ʃʰĩːxʲ ‘nose’ -nt ʃʰix t ʃʰi (ː)ʃ- -mĩːʂɻ -nt ʃʷʰiʍ *-tʃʰən ‘smell’ -tʃʰan -tʃĩh -ʂɻən -tʃʷʰin *-tʃʰixʲ ‘ochre/red’ t ʃʰix -tʃiːh -ʂɻik -- *-tʃʰiː ‘make (1,IMPF)’ -- -tʃʰí -ʂɻiː -tʃʷʰeː *t ʃʰaʔ ‘beaver’ t ʃʰu t ʃʰaːʔ -- t ʃʷʰaʔ *t ʃ(ʷ)ʰ ‘grandmother’ t ʃʰu- -tʃʰó ʂɻúː (xo)t ʃʷʰoː

b. *ʃʷaː ‘sun’ -- ʃá -ʂɻiː34 ʍaː *-ʒʷəŋʲə ‘dark/black’ -ʒəni -ʒi ː ʔ -ʂɻən -ʍiŋ *-ʃḛːq’ ‘spit (perf.)’ -ʃaq’ ʃéːʔ ʂɻéːk’iʔ -ʍeːq’

c. *t ʃʷeˑn ‘day’ t ʃan(i) t ʃi ʂɻíːniʃ t ʃeːnis *-weː-t ʃʷ’e- ‘m.-in-law’ -vatʃ’a ´-bíʒí -- -meːt ʃ’eː- *t ʃʷeːjəʔ ‘heart’ -- -tʃéí -ʂɻéːyeh -tʃeːʔ *weːt ʃʷ ‘boil (cook)’ -latʃ -peːʒ -- -met ʃ *t ʃḛʁ ‘ear’ -tʃəʁa -tʃaːʔ -ʂɻeʔ -tʃiwʔ *-tʃ(ʷ) ‘shin’ -tʃata -tʃát35 ʂɻáːtʰeʔ -- *t ʃḛːχ ‘gum’ t ʃaχ t ʃeːh ʂɻéː t ʃeh *t ʃ’əχt ‘hat’ t ʃ’əχ36 t ʃ’ah -ʈɻ’at --

32 E.g., *tʃ’əxʲtɬ’ ‘frog’ (Krauss and Leer 1981) > Earlier Hupa tʃʷ’al (Boas and Goddard 1924) > Mod-

ern Hupa tʃ’ahl (Golla 1996) (cf. Navajo tʃ’aɬ).

33 Rounding (acoustically [flat]/[grave]) may have persevered only in [tʃʷʰ, ʃʷ/ʍ] because it was more

easily preserved in the long and heavily aspirated noise of these segments (cf. Golla 1964:108, n. 3).

34 Movement of the sun.

35 ‘leg’; ‘shin’ = t ʃástʰis < -tʃáːt ‘leg’ + -tʰis ‘over’.

27

d. *t ʃʷ’əj ‘air, blow’ t ʃ’əj -tʃ’i -ʈɻ’íː -tʃ’e *jiːt ʃʷ’ ‘breath’ -jitʃ’ -jih -jeʔʂɻ -jeʔtʃ’ *t ʃ’… ‘guts’ -tʃ’ik’a -tʃ’íːʔ ʈɻ’íːk’eʔ -tʃ’eːkʲ’eʔ *t ʃ’əwə ‘spruce’ t ʃ’əva- t ʃ’ó ʈɻ’ameʔ t ʃ’əme- *t ʃ’ənʔ ‘towards’ t ʃ’(ən) -tʃ’ĩʔ --37 t ʃ’ənʔ

The preference for flat shibilants in Pacific Coast Athabascan —retroflex in Tolowa and

rounded in (earlier) Hupa— is surprising from the point of view of articulatory markedness, but

is predicted by an acoustic-auditory view of Athabascan sound change. The rounding contrast

among shibilants was, we suggest, especially disadvantageous within the PA sytem (28) be-

cause this secondary articulation has a valuable noncontrastive function in differentiating shibi-

lants from sibilants. The key auditory distinction of shibilants is their relatively lower resonant

frequency, which results from their slightly larger front cavity. In well-studied languages which

oppose shibilants to sibilants, such as English or French, it is found that lip rounding regularly

enhances this opposition, because such an articulation serves to enlarge the front cavity of

shibilants and lower their resonant frequency further (Ladefoged and Maddieson 1996:148,

Flemming 2002:55, Johnson 2003:118).38 Thus, the development of the PA sibilant system (28)

36 Archaic.

37 Cf. closely-related Tututni -ʈʂ’ənʔ ‘towards’. According to Golla (1976) retroflexes in Tututni are pho-

netically more fronted than in Tolowa (i.e., presumably apicoalveolar rather than fully sublaminoprepalatal), and

they are therefore exceedingly difficult to hear and transcribe consistently (e.g., Landar 1977). Yet the diachronic

pattern is the same: retroflexes are the reflexes of PA shibilants, whether originally flat (*-t ʃʷʰəʁ ‘weep’ > -ʂeh;

*jiːtʃʷ’ ‘breath’ > -jiʔʂ, etc.) or not (*-tʃʰən ‘smell’ > ʂən; -t ʃʰixʲ ‘ochre/red’ > -ʂək; *tʃḛʁ ‘ear’ > ʂəɻɛː, etc.).

38 Such phonetic enhancement is language-specific, and it is not reported for other Athabascan languages

which have merged the rounded and unrounded shibilants of PA into an unmarked series, such as Dena’ina or Na-

28

into a two-way opposition between /t s, tsʰ, t s’, s/ and /tʃ(ʷ), t ʃʷʰ, t ʃʷ’, ʃʷ/ in earlier Hupa is sen-

sible from an acoustic-auditory perspective, albeit marked in terms of articulatory typology.

This explanation extends to the development of the PA system (28) into a two-way op-

position between /ts’, s/ and /ʈɻ’, ʂɻ/ in Oregon Athabascan. Rounded and unrounded shibilants

merged into a [flat] series, not an unmarked one, because the flatness served to enhance the

auditory opposition of shibilants to corresponding sibilants. It is also expected that retroflexion

should have replaced rounding as the dominant enhancement of shibilants in Oregon Athabas-

can (as in Gwich’in and other Northern languages; see above): our auditory account crucially

invokes [flat], and rounding and retroflexion are both articulatory exponents of this feature. In

fact, these articulations are not at all mutually exclusive; according to Golla (1976:219) “the

retroflex reflexes of the PA *dž/g -series [i]n Chetco-Tolowa, and apparently in Umpqua …

are realized as strongly r-colored, lip-rounded, palatal sibilants [š r ] and [tš r ’].”

Finally, we raise a detail about PA that has been side-stepped in our discussion so far:

the two series of PA shibilants were fully distinguished only stem-initially (Krauss and Golla

1981:72). The handful of Northern languages which provide evidence of the historical distinc-

tion between two series of shibilants (viz. Deg Xinag, Upper Kuskokwim, Lower Tanana, Hän,

Gwich’in, and Ts’ets’aut) suggest that /*ʃ, *ʒ/ were in fact merged with rounded /*ʃʷ, *ʒʷ/

stem-finally. The fact that Gwich’in, for instance, shows the round reflex o(h) for all stem-final

shibilant fricatives (33) has led researchers to conclude that all stem-final shibilants were flat in

PA (see especially Leer 1996:212, also Krauss 1977, Leer 1979, Rice 1998:77). In other words,

vajo in (32). McDonough (2003) remarks explicitly that “rounding is not a salient property of the articulation of

Navajo sh” (p. 130).

29

(33) Gwich’in

*q’əʃʷ ‘alder’ > k’oh *xʲəʃʷ ‘bear’ > ʃoh *t’eˑʃʷ ‘charcoal’ > t ʃ’ìoh *-tḛːʃʷ ‘lie PL.MOM.IMPF’ > -t ʃìoː

*-ʔa ː ʃʷ ‘go 2.MOM.IMPF’ > -ʔòː *jəʃʷ ‘wolf’ > ʒòh

stem-finally /*s, *z/ were opposed only to /*ʃʷ, *ʒʷ/39 (not

/*ʃ, *ʒ/), in much the same way that alveolar sibilants

were eventually opposed only to flat shibilants in Pacific

Coast Athabascan (e.g., early Hupa, Tolowa), and we can

assume the same auditory/acoustic explanation: while

marked, obligatory lip-rounding served to emphasize the

opposition of /*ʃʷ, *ʒʷ/ to /*s, *z/ stem-finally in PA.

*ʁḛːʒʷə ‘egg’ > ɣòːʔ

5 CONCLUSION. Phonetic awareness proffers many rewards to the student of sound change.

Perhaps most satisfying is how seemingly strange shifts become sensible upon consideration of

the articulation and perception of the sounds involved. Hoijer’s (1942) demystification of *t > k

in Eastern Apachean is a case in point. Hoijer pointed out that Athabascan t and k were phone-

mically /tʰ, kʰ/, and that phonological aspiration had acquired a velar realization in Apachean

phonetics: [tˣ, kˣ]. This articulation was adopted by speakers apparently to enhance the distinct-

iveness of the aspirated stop series within the rich obstruent system of Athabascan. Indeed, the

same articulation has been reported for phonologically aspirated stops in other unrelated lan-

guages which (like Athabascan) make a three-way laryngeal distinction among stops, from

Lakhota (Rood and Taylor 1996) to Thai (Haas 1968:166, n. 6).40 Crucially, sound change oc-

curred (in Ohala’s (2003) “mini” sense) precisely when some (Eastern) Apachean (quite rea-

sonably) misheard each other’s [tˣ] as [kˣ], and therefore as /kʰ/.

39 As mentioned in fn. 6, Leer (1996) and Rice (1998) assume that rounded shibilants became retroflex al-

ready in PA. Hence Leer’s (1996:212) wording: “the reflexes of Pre-PA stem-final *š and *šʷ had already merged

in PA as *šʳ or its voiced counterpart *žʳ.”

30

Another obvious benefit of phonetic explanation is that its principles are physical, and

therefore constant, thereby accounting for the neogrammarian impression that sound change is

regular. For instance, Hoijer remarked that outside of Apachean, “it would appear that only in

Mattole have t and k taken on a heavy spirantal [x]-glide. Similarly, the data available on the

northern Athapaskan languages indicate that the [x]-glide is found only in Chipewyan” (p.

219). Hoijer was surely gratified to learn that the shift *t > k was eventually replicated in dia-

lectal Dëne Sųłiné (Chipewyan), as first reported by Haas (1968). According to Cook (2004),

who provides the examples in (34), this shift was exhaustive in dialects such as Fond du Lac,

(34) Standard vs. shifted Dëne Sułiné tʰu

tʰɛn setʰá jaɬtʰi

hustʰón

kʰu kʰʌn sekʰá jaɬkʰi

huskʰón

‘water’ ‘ice’ ‘my father’ ‘s/he speaks’ ‘I hold it’

(35) PA > Pre-Mattole > Mattole *-tʃʰĩːxʲ ‘nose’ > -tʃʰiʃ > -tʃxix *ʒʷəŋʲə ‘black’ > -ʃin > -xin *ʃʷaˑ ‘sun’ > ʃaː > xaː ‘light’ *q’əʁəʃʷ ‘alder’ > k’iɣiʃ > k’iɣix

Saskatchewan, which now completely lacks t. In the

other language with velarized aspiration mentioned by

Hoijer, the extinct Californian language Mattole, the

affected stops were evidently reanalyzed as stop+[x]

sequences: /tʰ, kʰ, t ʃʰ/ > /tʰx, kʰx, tʃʰx/ (Li 1930).

Apparently, too, the velar realization of aspiration had

extended to /ʃ/ in Mattole, yielding perhaps [ɧ], which

was eventually reanalyzed as /x/, as shown in (35).41

Happily, many other unobvious (and therefore

potentially informative) sound shifts have occurred in *t’eˑʃʷ ‘charcoal’ > t’eːʃ > t’eːx

the history of Athabascan. In this article we focused on those which affected PA sibilants in the

Northern branch (§3) and PA rounded shibilants in the Northern and Pacific Coast branches

40 Haas’ trusted report of velarization in aspiration does not apply to Standard Thai (Abramson 1989).

41 See Vaux (1998) for various arguments that voiceless fricatives and aspirated stops form a natural class

based on [spread glottis].

31

(§4). We argued that the acoustic features [grave] (redefined as the audible presence of signifi-

cant low frequency noise in a sound) and [flat] (§2) are essential in understanding these shifts.

As such, our study is obviously contrary to strong claims that effort minimization is the only

phonetic motivation for regular historical change (e.g., Pagliuca and Mowrey 1987, Mowrey

and Pagliuca 1995, Bybee 2001:77-81). Indeed, Blevins (2004) remarks that sound changes like

θ > f (cf. our §3.2) are unintelligible in terms of articulatory minimization:

I know of no studies suggesting that labiodental fricatives involve less articulatory effort than dental fricatives; in fact, since dental fricatives are frequently the lenited counterparts of dental stops, while bilabial fricatives result from lenition of bilabial stops, there is the suggestion in sound-change typology that labio-dental fricatives involve more articulatory effort than dental fricatives. (Blevins 2004:292)

Rather, our demonstration that sound change in Athabascan is largely guided by [grave]

and [flat] supports theories such as Blevins’ (2004) and Ohala’s (1981, 2003) which blame lis-

teners for sound change: “listeners’ errors constitute the main and the essential factor in sound

change (assuming sound change is taken as “new pronunciation norm”)” (Ohala 2003:686, n.

5). For instance, we saw that Athabascan history is replete with listeners who have misheard

and misanalyzed one [grave] sound as another (e.g., thibilant → lateral/labial/velar/labio-velar),

or one [flat] sound as another (e.g., lip-rounding → retroflexion/pharyngealization).

In most cases, however, it is also clear that speakers held partial responsibility, not for

initiating the sound changes, but for introducing greater variation in the speech signal, thereby

unwittingly misguiding their listeners. Of particular interest are instances in which speakers

have introduced special phonetics to highlight a distinction. As Ohala (2003:683) states: “There

is, to be sure, much cognitive activity —teleology, in fact— in producing and perceiving

speech, but all the evidence we have suggests that this is directed toward perserving, not re-

32

placing, pronunciation norms.” Ironically, our study revealed many cases in which pronuncia-

tions aimed at “preserving” led instead to sound changes, that is, to listener misapprehensions:

(36) phonetics … used to enhance … led to sound change

a. dentalization [+anterior] in sibilants (vs. [−anterior] shibilants)

/tsʰ/ [t s h ] → “t θʰ” /t θʰ/ (e.g., Northern Athabascan)

b. lip-rounding [grave] in thibilants (vs. nongrave sibilants)

/t θʰ/ [t θʷʰ] → “kʷʰ” /kʷʰ/ (e.g., Tłįchǫ Yatiì)

c. palatalization [+high] in velars (vs. [−high] uvulars)

/kʷʰ/ [k ʰ] → “t ʃʷʰ” /t ʃʷʰ/ (e.g., Proto-Athabascan)

d. retroflexion [+round] in labialized shibilants (vs. plain shibilants)

/tʃʷʰ/ [ʈɻʷʰ] → “ʈɻʰ” /ʈɻʰ/ (e.g., Gwich’in)

e. lip-rounding [−anterior] in shibilants (vs. [+anterior] sibilants)

/tʃʰ/ [t ʃʷʰ] → “t ʃʷʰ” /t ʃʷʰ/ (e.g., early Hupa)

f. retroflexion [−anterior] in shibilants (vs. [+anterior] sibilants)

/tʃ’/ [ʈɻ’] → “ʈɻ’” /ʈɻ’/ (e.g., Tolowa)

g. velarization [spread glottis] in stops (vs. unas-pirated plain stops and ejectives)

/tʰ/ [tˣ] → “kˣ” /kʰ/ (e.g., dial. Dëne Sųłiné)

Cases such as (36b) suggest that speakers and listeners alike show sensitivity to [grave]

(see also AK-Gwich’in thibilant > velar in §3.4). Moreover, the relative abundance of sound

shifts in our study which make crucial reference to [grave] and/or [flat] suggests generally that

these features are markedly salient in the Athabascan family. However, it does not follow that

these features are present in individual phonologies. As mentioned in §2, it appears that syn-

chronic phonologists have not missed [grave] and [flat] at all since abandoning them in the

1970’s. For instance, one might imagine that there are consonantal distinctions or synchronic

phonological processes affecting consonants which make crucial reference to these features, but

no compelling examples have been produced to date (see Flemming 2002 for a list of possibili-

33

ties). Indeed, the Jakobson-Halle system was forsaken precisely because articulatory features

consistently provided simpler analyses of synchronic (consonantal) distinctions and processes

—even those for which acoustic features seemed equally descriptively adequate (Halle 2002).

This all leaves us with the impression that acoustic features like [grave] and [flat] are

linguistic elements to be sure, but they are not part of the phonology proper. Rather, the impor-

tance of such features to diachronic explanation on one hand, and their enduring irrelevance to

synchronic explanation on the other (Chomsky and Halle 1968 et seq.), jointly affirm that pho-

netics and phonology are distinct modules, each with its privileged principles and elements, in-

cluding features (e.g., Gerfen 2001, Gussenhoven and Kager 2001, pace Steriade 1997, Flem-

ming 2001), and that sound change has its source in phonetics, not phonology (e.g., Ohala

2003, Blevins 2004, pace Kiparsky 1995).

6 REFERENCES

Abramson, A. S. 1989. Laryngeal control in the plosives of Standard Thai. Pasaa 19:85-93. Bessell, Nicola, and Czaykowska-Higgins, Ewa. 1991. The phonetics and phonology of postvelar

sounds in Moses-Columbia Salish (Nxa'amxcin). Paper presented at the Annual Meeting of the Canadian Linguistics Association, Queen's University, Kingston, Ontario, May 1991.

Bessell, Nicola Jane. 1998. Local and Non-Local Consonant-Vowel Interaction in Interior Salish. Phonology 15:1-40.

Bhat, D. N. S. 1978. A general study of palatalisation. In Universals of human language, eds. Joseph H. Greenberg, Charles Ferguson and Edith Moravcsik, 47-92. Stanford: Stanford University Press.

Blake, Susan. 2000. On the distribution and representation of schwa in Sliammon Salish: descriptive and theoretical perspectives, University of British Columbia: Doctoral dissertation.

Blevins, Juliette. 2004. Evolutionary phonology: the emergence of sound patterns. Cambridge: Cam-bridge University Press.

Boas, Franz, and Goddard, Pliny Earle. 1924. Ts'ets'aut, an Athapascan language from Portland Ca-nal, British Columbia. International Journal of American Linguistics 3:1-35.

Bybee, Joan L. 2001. Phonology and language use. New York: Cambridge University Press. Chomsky, Noam, and Halle, Morris. 1968. The sound pattern of English. New York: Harper & Row. Collins, James. 1989. Nasalization, lengthening, and phonological rhyme in Tolowa. International

Journal of American Linguistics 55:326-340. Cook, Eung-Do. 1983. Chilcotin flattening. Canadian Journal of Linguistics 28:123-132. Cook, Eung-Do, and Rice, Keren. 1989. Introduction. In Athapaskan linguistics, eds. Eung-Do Cook

and Keren Rice, 1-62. Berlin: Mouton.

34

Cook, Eung-Do. 1993. Chilcotin flattening and autosegmental phonology. Lingua 91:149-174. Cook, Eung-Do. 1997. The acquisition and deacquisition of Chipewyan consonants: University of

Calgary. Cook, Eung-Do. 2004. A grammar of Dëne Sųłiné (Chipewyan). Winnipeg, MB: Algonquian and

Iroquoian Linguistics. Dunn, John A. 1995. Sm'algyaẋ: A reference dictionary and grammar for the Coast Tsimshian lan-

guage. Juneau: Sealaska Heritage Foundation. Durand, Jacques. 1986. Dependency and non-linear phonology: Croom Helm linguistics series. Lon-

don; Dover, N.H.: Croom Helm. Eastman, Carol M., and Aoki, Paul K. 1978. Phonetic segments in Haida (Hydaburg dialect). In Lin-

guistic and literary studies in honor of Archibald A. Hill, 237-249. Berlin: Mouton. Ewen, Colin. 1995. Dependency relations in phonology. In The Handbook of Phonological Theory,

ed. John A. Goldsmith, 570-585. Cambridge, MA, and Oxford, UK: Blackwell. Firth, William G., and Mitchell-Firth, Eleanor. 2003. Gwichyah ts'at Teetl'it Gwich'in Ginjik Gwi'di-

nehtl'ee'. Fort McPherson, NT: Gwich'in Social and Cultural Institute. Flemming, Edward. 2001. Scalar and categorical phenomena in a unified model of phonetics and

phonology. Phonology 18:7-44. Flemming, Edward S. 2002. Auditory representations in phonology. New York: Routledge. Gerfen, Chip. 2001. A critical view of licensing by cue: The case of Andalusian Spanish. In Segmen-

tal phonology in Optimality Theory, ed. Linda Lombardi, 183-205. Cambridge: Cambridge University Press.

GNWT. 1990. Dene Standardization Project. Yellowknife: Department of Culture and Communica-tions, Government of the Northwest Territories.

Goddard, Ives. 1980. Eastern Algonquian as a genetic subgrouping. In Papers from the Eleventh Al-gonquian Conference, ed. William Cowan, 143-158. Ottawa: Carleton University.

Golla, Victor. 1964. An etymological study of Hupa noun stems. International Journal of American Linguistics 30:108-117.

Golla, Victor. 1976. Tututni (Oregon Athapaskan). International Journal of American Linguistics 42:217-227.

Golla, Victor. 1977. A note on Hupa verb stems. International Journal of American Linguistics 43:355-358.

Golla, Victor. 1996. Sketch of Hupa, an Athapaskan language. In Handbook of North Americans In-dians, ed. Ives Goddard, 364-389. Washington: Smithsonian Institute.

Golla, Victor Karl. 1970. Hupa grammar, University of California, Berkeley: Doctoral dissertation. Gordon, Matthew. 1996. The phonetic structures of Hupa. University of California Working Papers

in Phonetics 93:164-187. Gussenhoven, Carlos, and Kager, René. 2001. Introduction: phonetics in phonology. Phonology

18:1-6. Haas, Mary R. 1968. Notes on a Chipewyan Dialect. International Journal of American Linguistics

34:165-175. Halle, Morris. 2002. Introduction. In From memory to speech and back, ed. Morris Halle, 1-17. Ber-

lin: Mouton de Gruyter. Hardcastle, William J. ed. 1999. Coarticulation: Theory, data and techniques. Cambridge, UK:

Cambridge University Press. Hoijer, Harry. 1942. Phonetic and phonemic change in the Athapaskan languages. Language 18:218-

220. Holton, Gary Michael. 2000. The phonology and morphology of the Tanacross Athabaskan language,

University of California, Santa Barbara: Doctoral dissertation.

35

Howard, Philip G. 1990. A dictionary of verbs of South Slavey. Yellowknife, Northwest Territories: Department of Culture and Communications, Government of the NWT.

Howren, Robert A. 1975. Some isoglosses in Mackenzie-drainage Athapaskan: First steps toward a subgrouping. In Northern Athapaskan Conference 1971, ed. Annette McFadyen Clark, 577-618. Ottawa: Ethnology division, National Museum of Man.

Huld, Martin E. 1983. Athapaskan bears. International Journal of American Linguistics:186-195. Jacobsen, William H., Jr. 1969. Origin of the Nootka pharyngeals. International Journal of American

Linguistics 35:125-153. Jakobson, Roman, Fant, Gunnar, and Halle, Morris. 1952. Preliminaries to speech analysis. Cam-

bridge, MA: MIT Press. Jakobson, Roman, and Halle, Morris. 1956. Fundamentals of language. The Hague: Mouton. Jetté, Jules, and Jones, Eliza. 2000. Koyukon Athabaskan dictionary. Fairbanks: Alaska Native Lan-

guage Center. Johnson, Keith. 2003. Acoustic and auditory phonetics, 2nd ed. Oxford: Blackwell. Kari, James. 1977a. Dena'ina noun dictionary. Fairbanks: Alaska Native Language Center. Kari, James. 1977b. Linguistic diffusion between Tanaina and Ahtna. International Journal of

American Linguistics 43:274-288. Kari, James. 1978a. Ingalik stem list. Fairbanks: Alaska Native Language Center. Kari, James. 1978b. Ahtna stem list. Fairbanks: Alaska Native Language Center. Keating, Patricia A. 1984. Phonetic and phonological representation of stop consonant voicing. Lan-

guage 60:286-319. Kiparsky, Paul. 1995. The phonological basis of sound change. In The Handbook of Phonological

Theory, ed. John A. Goldsmith, 640-670. Cambridge, MA, and Oxford, UK: Blackwell. Krauss, Michael. 1964. Proto-Athapaskan-Eyak and the problem of Na-Dene: I phonology. Interna-

tional Journal of American Linguistics 30:118-136. Krauss, Michael. 1973. Na-Dene. In Linguistics in North America, ed. Thomas Sebeok, 903-978. The

Hague: Mouton. Krauss, Michael. 1975. Chilcotin phonology: A descriptive and historical report, with recommenda-

tions for a Chilcotin orthography. Ms. University of Alaska, Fairbanks. Krauss, Michael. 1977. Proto-Athapaskan-Eyak fricatives and the first person singular. Alaska Native

Language Center Research Papers 1:1-60. Krauss, Michael, and Leer, Jeff. 1981. Athabaskan, Eyak, and Tlingit sonorants. Alaska Native Lan-

guage Center Research Papers 5. Krauss, Michael. 1982. Proto-Athapaskan *k in Chipewyan, 1742-1800: Philological evidence. In-

ternational Journal of American Linguistics 48:73-82. Krauss, Michael E., and Golla, Victor Karl. 1981. Northern Athapaskan languages. In Subarctic, ed.

June Helm, 67-85. Washington: Smithsonian Institution. Kuipers, Aert. 1981. On reconstructing the Proto-Salish sound system. International Journal of

American Linguistics 47:323-335. Ladefoged, Peter, and Maddieson, Ian. 1996. The sounds of the world's languages. Oxford, UK:

Blackwell. Ladefoged, Peter. 1997. Linguistic phonetic descriptions. In The handbook of phonetic sciences, eds.

William J. Hardcastle and John Laver, 589-618. Oxford, UK: Blackwell. Landar, Herbert. 1977. Three Rogue River Athapaskan vocabularies. International Journal of Ameri-

can Linguistics 43:289-301. Lass, Roger. 1976. English phonology and phonological theory: synchronic and diachronic studies.

Cambridge, UK: Cambridge University Press. Leer, Jeff. 1979. Proto-Athabaskan verb stem variation, part one: Phonology. Alaska Native Lan-

guage Center Research Papers 1:1-100.

36

Leer, Jeff. 1996. The historical evolution of the stem syllable in Gwich'in (Kutchin /Loucheux) Ath-abaskan. In Athabaskan language studies: Essays in honor of Robert W. Young, eds. Eloise Jelinek, Sally Midgette, Keren Rice and Leslie Saxon, 193-234. Albuquerque: University of New Mexico Press.

Leer, Jeff. in press. How stress shapes the stem-suffix complex in Athabaskan. In Athabaskan pros-ody, eds. Sharon Hargus and Keren Rice. Amsterdam: John Benjamins.

Leslau, Wolf. 1957. The phonetic treatment of the Arabic loanwords in Ethiopic. Word 13:100-123. Leslau, Wolf. 1990. Arabic loanwords in Ethiopian Semitic. Wiesbaden: Harrassowitz. Li, Fang Kuei. 1930. Mattole - An Athabaskan language. Chicago: University of Chicago Press. Lincoln, Neville J., and Rath, John C. 1980. North Wakashan comparative root list. Ottawa: National

Museums of Canada. Martinet, André. 1955. Économie des changements phonétiques; traité de phonologie diachronique:

Bibliotheca romanica. Ser. 1: Manualia et commentationes, 10. Berne: A. Francke. Masuzumi, Barney, Grandjambe, Dora, and Cizek, Petr. 1994. North Slavey terminology and con-

cepts related to renewable resources: An interim report for the Department of Renewable Re-sources, Government of the Northwest Territories. Yellowknife, NWT: Dene Cultural Insti-tute.

Mattina, Anthony. 1979. Pharyngeal movement in Colville and related phenomena in the Interior Salishan languages. International Journal of American Linguistics 45:17-24.

McDonough, Joyce. 2003. The Navajo sound system. Dordrecht: Kluwer Academic Publishers. Mithun, Marianne. 2001. The languages of Native North America. Cambridge, UK: Cambridge Uni-

versity Press. Mowrey, Richard, and Pagliuca, William. 1995. The reductive character of articulatory evolution.

Rivista di Linguistica 7:37-124. Nater, Hank F. 1984. The Bella Coola Language. Ottawa: National Museums of Canada. Nater, Hank F. 1989. Some Comments on the Phonology of Tahltan. International Journal of Ameri-

can Linguistics 55:25-42. Ohala, John J. 1981. The listener as a source of sound change. Chicago Linguistic Society 17:178-

203. Ohala, John J. 1985. Around flat. In Phonetic linguistics: Essays in honor of Peter Ladefoged, ed.

Victoria Fromkin, 223-241. Florida: Academic Press. Ohala, John J. 2003. Phonetics and historical phonology. In The handbook of historical linguistics,

eds. Brian D. Joseph and Richard D. Janda, 669-686. Oxford, UK: Blackwell. Pagliuca, William, and Mowrey, Richard. 1987. Articulatory evolution. In Papers from the Seventh

International Conference on Historical Linguistics, eds. A. G. Ramat, O. Carruba and G. Bernini, 459-472. Amsterdam: John Benjamins.

Poser, William. 1999. Carrier monosyllabic noun stems. Ms. University of Pennsylvania, Philadel-phia.

Poser, William. 2003. Dʌlk’ʷahke: The first Carrier writing system. Ms. University of Pennsylvania, Philadelphia.

Poser, William. 2004. Dating velar palatalization in Carrier. Ms. University of Pennsylvania, Phila-delphia.

Rice, Keren. 1977. Continuants in Hare. International Journal of American Linguistics 43:315-326. Rice, Keren. 1989. A grammar of Slave. Berlin, New York: Mouton de Gruyter. Rice, Keren. 1994. Peripheral in consonants. Canadian Journal of Linguistics 39:191-216. Rice, Keren. 1998. Les langue athapascanes du nord: Une introduction. Recherches amérindiennes

au québec 28:75-92.

37

Rice, Keren. 2004. Language contact, phonemic inventories, and the Athapaskan language family. Linguistic Typology 8:321-383.

Rood, David S., and Taylor, Allan R. 1996. Sketch of Lakhota, a Siouan language. In Handbook of North American Indians, ed. Ives Goddard, 440-482. Washington, D.C.: Smithsonian Institu-tion.

Sapir, Edward. 1931. The concept of phonetic law as tested in primitive languages by Leonard Bloomfield. In Methods in social science: A case book, ed. Stuart A. Rice, 297-306. Chicago: University of Chicago Press.

Saxon, Leslie. 1979. Proto-Northeastern Athapaskan stem-final consonants. Ms. University of To-ronto.

Saxon, Leslie, and Siemens, Mary. 1996. A Dogrib dictionary. Rae-Edzo, Northwest Territories: Do-grib Divisional Board of Education.

Shahin, Kimary. 2002. Postvelar harmony. Amsterdam: John Benjamins. Shaw, Patricia A. 1980. Theoretical issues in Dakota phonology and morphology: Outstanding dis-

sertations in linguistics. New York: Garland Pub. Shosted, Ryan. 2004. Flatly contradicted: The non-equivalency of a distinctive feature in Hindi and

Levantine Arabic. Paper presented at the Ninth Conference on Laboratory Phonology, June 24-26, 2004, University of California-Berkeley.

Sivertsen, Eva. 1960. Cockney phonology. Oslo: University Press. Steriade, Donca. 1997. Phonetics in phonology: The case of laryngeal neutralization. Ms. University

of California, Los Angeles. Stevens, Kenneth N., and Keyser, Samuel Jay. 1989. Primary features and their enhancement in con-

sonants. Language 65:81-106. Story, Gillian. 1984. Babine and Carrier phonology: a historically oriented study. Dallas: Summer

Institute of Linguistics. Tarpent, Marie-Lucie. 1987. A grammar of Nisgha, University of Victoria: Doctoral dissertation. Tharp, George W. 1972. The position of the Tsetsaut among Northern Athapaskans. International

Journal of American Linguistics 39:14-25. Thompson, Chad. 1991. The low topicality prefix k'i- in Koyukon. Studies in Language 15. Trask, R. L. 1996. A dictionary of phonetics and phonology. London, UK: Routledge. Trubetzkoy, Nikolai Sergeevich. 1939. Grundzüge der Phonologie. Prague: Travaux du cercle lin-

guistique de Prague 7. Vaux, Bert. 1998. The laryngeal specifications of fricatives. Linguistic Inquiry 29:497-511. Velleman, Shelley L. 1988. The role of linguistic perception in later phonological development. Ap-

plied Psycholinguistics 9:221-236. Woodward, Mary F. 1964. Hupa phonemics. In Studies in California linguistics, ed. William Bright,

199-216. Berkeley: University of California Press. Young, Robert W. 1983. Apachean languages. In Southwest, ed. Alonso Ortiz, 393-400. Washington,

D.C.: Smithsonian Institution. Young, Robert W., and Morgan, William, Sr. 1992. Analytical lexicon of Navajo. Albuquerque: Uni-

versity of New Mexico Press.