Phonetics

All consonants are long, compared to English: with plain stops the hold is longer, with aspirated stops the aspiration is longer, and with affricates the frication is longer. The voice onset time of the aspirated and ejective stops is twice as long as that found in most non-Athabaskan languages. Young & Morgan (1987) described Navajo consonants as "doubled" between vowels, but in fact they are equally long in all positions.^[1]

Stops and affricates

All stops and affricates, except for the bilabial and glottal, have a three-way laryngeal contrast between unaspirated, aspirated, and ejective. The labial /m/ is found in only a few words. Most of the contrasts in the inventory lie within coronal territory at the alveolar and palatoalveolar places of articulation.

The aspirated stops /tʰ, kʰ/ (orthographic ⟨t⟩, ⟨k⟩) are typically aspirated with velar frication [tx, kx] (they are phonetically affricates – homorganic in the case of [kx], heterorganic in the case of [tx]).^[2] The velar aspiration is also found on a labialized velar [kxʷ] (orthographic ⟨kw⟩). There is variation within Navajo, however, in this respect: some dialects lack strong velar frication having instead a period of aspiration.^[3][4]

Similarly the unaspirated velar /k/ (orthographic ⟨g⟩) is realized as with optional voiced velar frication following the stop burst: [k] ~ [kɣ]. The unaspirated lateral /tɬ/ (orthographic ⟨dl⟩) typically has a voiced lateral release, [tˡ], of a duration comparable to the release of the /k/ and much shorter than the unaspirated fricatives /ts/, /tʃ/. However, the aspirated and ejective laterals are true affricates.

While the aspiration of stops is markedly long compared to most other languages, the aspiration of the affricates is quite short: the main feature distinguishing /ts/ and /tʃ/ from /tsʰ/ and /tʃʰ/ is that the frication is half again as long in the latter: [tsˑʰ], [tʃʰˑ]. /tɬʰ/ is similarly long, [tɬˑʰ]. The ejectives /tsʼ/, /tɬʼ/, /tʃʼ/, on the other hand, have short frication, presumably due to the lack of pulmonic airflow. There is a period of near silence before the glottalized onset of the vowel. In /tɬʼ/ there may be a double glottal release, or a creaky onset to the vowel not found in the other ejective affricates.

Continuants

Navajo voiceless continuants are realized as fricatives. They are typically noisier than the fricatives that occur in English. The palato-alveolars /ʃ, ʒ/ are not labialized unlike English and other European languages.^[5]

Navajo also does not have consistent phonetic voicing in the "voiced" continuant members. Although /z, l, ʒ, ɣ/ are described as voiced in impressionist descriptions,^[6] data from spectrograms shows that they may be partially devoiced during the constriction. In stem-initial position, /l/ tends to be fully voiced, /ʒ/ has a slight tendency to be voiceless near the offset, /z/ is often mostly voiceless with phonetic voicing only at the onset, /ɣ/ is also only partially voiced with voicing at onset. A more consistent acoustic correlate of the "voicing" is the duration of the consonant: "voiceless" consonants have longer durations than "voiced" consonants. Based on this, McDonough (2003) argues that the distinction is better captured with the notion of a fortis/lenis contrast. A further characteristic of voicing in Navajo is that it is marginally contrastive (see the voicing assimilation section).

Navajo lacks a clear distinction between phonetic fricatives and approximants. Although the pair [ɬ]~[l] has been described as a fricative and an approximant, respectively, the lack of a consistent contrast between the two phonetic categories and a similar patterning with other fricative pairs suggests that they are better described as continuants. Additionally, observations have been made about the less fricative-like nature of [ɣ, ɣʷ] and the more fricative-like nature of [j].

Sonorants

A more abstract analysis of Navajo posits two different /j/ phonemes (see the below for elaboration).

The glottalized sonorants are the result of d-effect on the non-glottalized counterparts. A strict structuralist analysis, such as that of Hoijer (1945a) and Sapir & Hoijer (1967), considers them phonemic.

Glottal(ized) consonants

Consonants involving a glottal closure – the glottal stop, ejective stops, and the glottalized sonorants – may have optional creaky voice on voiced sounds adjacent to the glottal gesture. Glottal stops may also be realized entirely as creaky voice instead of single glottal closure.^[7] Ejectives in Navajo differ from the ejectives in many other languages in that the glottal closure is not released near-simultaneously with the release of the oral closure (as is common in other languages) – it is held for a significant amount of time following oral release. The glottalized sonorants /mˀ, nˀ/ are articulated with a glottal stop preceding the oral closure with optional creaky voice during the oral closure: [ʔm – ʔm̰, ʔn – ʔn̰].

Labialized consonants

Consonants /kʰʷ, xʷ, ɣʷ, hʷ/ are predictable variants that occur before the rounded oral vowel /o/. However, these sounds also occur before the vowels /i, e, a/ where they contrast with their non-labialized counterparts /kʰ, x, ɣ, h/.

Velar /ɣ/, palatal /j/

The phonological contrast between the velar obstruent /ɣ/ and the palatal glide /j/ is neutralized in certain contexts. However, in these contexts, they may often be distinguished from each other by their different phonological patterning.

Before the rounded /o/, /ɣ/ is phonetically strongly labialized as [ɣʷ]; elsewhere, it lacks the labialization. As noted above, the lenis continuants like /ɣ/ are often very weak fricatives somewhere between a typical fricative constriction (e.g. [ɣ]) and a more open approximant constriction (e.g. [ɰ]) – this will be symbolized here as [ɰ̝]. Hoijer (1945a) describes the [ɰ̝ʷ] realization as being similar to English [w] but differing in having slight frication at the beginning of the articulation. The realization before /a/ varies between an approximant [ɰ] and a weakly fricated approximant [ɰ̝].^[8] The following verb stem^[9] has different velar allophones of the stem-initial consonant:

Word	Underlying	Phonetic	! Gloss
-ghash	/ɣàʃ/	[ɰ̝àʃ]	'make bubbling noise' (iterative, continuative)
-wosh	/ɣòʃ/	[ɰ̝ʷòʃ]	'make bubbling noise' (iterative, repetitive)

The palatal glide /j/ is also phonetically between an approximant [j] and a fricative [ʝ]. Hoijer (1945a) compares it to English [j] with a "slight but audible 'rubbiness' or frication."

The contrast between velar /ɣ/ and palatal /j/ is found before both back vowels /a, o/ as the following contrasts demonstrate:

	Word	Underlying	Phonetic	Gloss
contrast before /a/	bighaaʼ	/pìɣàːʔ/	[pɪ̀ɰ̝àːʔ]	'its fur'
	biyaaʼ	/pìjàːʔ/	[pɪ̀j˔àːʔ]	'its lice'
contrast before /o/	biwol	/pìɣòl/	[pɪ̀ɰ̝ʷòl]	'its marrow'
	biyol	/pìjòl/	[pɪ̀j˔òl]	'its breath'

Before the front vowels /i, e/, however, the contrast between /ɣ/ and /j/ is neutralized to a palatal articulation much like the weakly fricative [j˔] realization of /j/ that occurs before back vowels. However, the underlying consonant can be ascertained in verb stems and noun stems via their different realizations in a voiceless (i.e. fortis) context. The underlying velar surfaces as a voiceless palatal fricative [ç] in these environments:

Fortis context			Lenis context
Word	Phonetic	Gloss	Word	Phonetic	Gloss
hééł	[çéːɬ]	'bundle'	biyéél	[pɪ̀j˔éːl]	'her bundle'
yishhizh	[j˔ɪ̀ʃçɪ̀ʒ]	'I pick (corn)'	yiyizh	[j˔ɪ̀j˔ɪ̀ʒ]	'she picks (corn)'

The stem-initial velar of the noun stem /xéːɬ/ has a voiceless fortis realization of [ç] (as [çéːɬ]) when word-initial. When intervocalic, it is realized as lenis [j˔] (as [-j˔éːl]). Likewise, the underlying velar of the verb stem /xɪ̀ʒ/ is a voiceless [ç] after the preceding voiceless [ʃ] and lenis [j˔] when intervocalic. Thus, the alternation of [ç ~ j˔] in the two contexts is indicative of an underlying velar consonant. Similarly before the back vowels, the velar continuant has the alternations [x ~ ɰ̝] and [xʷ ~ ɰ̝ʷ] as shown in the examples below:

	Fortis context			Lenis context
	Word	Phonetic	Gloss	Word	Phonetic	Gloss
before /a/	haníłháásh	[hànɪ́ɬxáːʃ]	'you make it boil'	hanílgháásh	[hànɪ́lɰ̝áːʃ]	'it comes to a boil'
before /o/	ałhosh	[ʔàɬxʷòʃ]	'he's sleeping'	áhodilwosh	[ʔáhòtɪ̀lɰ̝ʷòʃ]	'he's pretending to be asleep'

An underlying palatal /j/ can be determined by alternations which differ from the velar alternations. However, /j/ has two different alternation patterns which have led to the positing of two distinct phonemes. Incidentally, the two different phonemes are also connected to two different reconstructed consonants in Proto-Athabascan. One of these /j/ phonemes is considered an obstruent as it has a fricative realization of [s] in fortis contexts. It is often symbolized as a palatalized (or front velar) fricative /ɣ̑/ (in Americanist phonetic notation) and is a reflex of Proto-Athabascan *x̯. It may be considered coronal because of its coronal voiceless allophone.

	Fortis context			Lenis context
	Word	Phonetic	Gloss	Word	Phonetic	Gloss
before /i/	sin	[sɪ̀n]	'song'	biyiin	[pɪ̀j˔ìːn]	'her song'
before /a/	honissą́	[hònɪ̀sːã́]	'I'm wise'	hóyą́	[hój˔ã́]	'she's wise'
before /o/	hanisóód	[hànɪ̀sːóːt]	'I drive them out'	hainiyóód	[hàɪ̀nɪ̀j˔óːt]	'she drives them out'

In the above examples, the fortis realization is [s] in the stems [sɪ̀n], [-sã́], [-sóːt] while the lenis realization is the glide [j˔] in the corresponding [-j˔ɪ̀n], [-j˔ã́], [-j˔óːt]. Since the fortis reflex of this phoneme is [s] there is also a neutralization between this /j/ phoneme and the alveolar /s/ phoneme. The alveolar phoneme has a [s ~ z] alternation in fortis-lenis contexts:

Fortis context			Lenis context
Word	Phonetic	Gloss	Word	Phonetic	Gloss
séí	[séɪ́]	'sand'	bizéí	[pɪ̀zéɪ́]	'her sand'

Thus, the different alternations also distinguish between underlying /j/ and underlying /s/.

The other underlying (or morphophonemic) palatal /j/ is considered a sonorant and has an invariant [j˔] realization in both fortis (voiceless) and lenis (voiced) contexts. This phoneme is relatively rare, occurring in only a few morphemes. It is a reflex of Proto-Athabascan *y (as symbolized in Americanist notation). Two examples are below:

	Fortis context			Lenis context
	Word	Phonetic	Gloss	Word	Phonetic	Gloss
before /a/	yaaʼ	[j˔àːʔ]	'louse'	shiyaʼ	[ʃɪ̀j˔àʔ]	'my louse'
before /o/	honishyóí	[hònɪ̀ʃj˔óɪ́]	'I'm energetic'	honíyóí	[hònɪ́j˔óɪ́]	'you're energetic'

A further distinction between the different phonemes are found in the context of d-effect.

The varying contextual realizations of these three underlying segments are summarized in the following table:

Underlying segment	Lenis			Fortis	D-effect
	before /a/	before /o/	before /i, e/
/ɣ/	ɰ̝	ɰ̝ʷ	j˔	x	k
/j/ < Proto-Ath. *x̯	j˔	j˔	j˔	s	ts
/j/ < Proto-Ath. *y	j˔	j˔	j˔	j˔	jˀ˔

Voicing assimilation

The voiced continuants /z, l, ʒ, ɰ̝/ at the beginning of stems vary with their voiceless counterparts /s, ɬ, ʃ, x/, respectively. The voiceless variants occur when preceded by voiceless consonants, such as /s, ɬ, ʃ, h/ while the voiced variants occur between voiced sounds (typically intervocalically). For example, the verb stems meaning 'spit it out', 'be burning', 'spit', and 'be ticklish' have the following forms with alternating voiced and voiceless stem-initial consonants:

Phonetic forms	Orthographic forms	Gloss
[zóːh ~ sóːh]	-zóóh ~ -sóóh	'spit it out'
[lɪ̀t ~ ɬɪ̀t]	-lid ~ -łid	'be burning'
[ʒàh ~ ʃàh]	-zhah ~ -shah	'spit'
[ɰ̝ʷòʒ ~ xʷòʒ]	-wozh ~ -hozh	'be ticklish'

Since the voicing is predictable, it can be represented more abstractly as an underlying consonant that is underspecified with respect to voicing. These archiphonemes can be indicated with the capital letters /Z, L, Ʒ, Ɣ/.^[10] The variant voicing of the stem-initial consonant can be found in the context of subject person prefixes which are added to the verb stem:

Phonetic form	Orthographic form	Underlying segments[11]	Gloss
[hàɪ̀tɪ̀zóːh]	haidizóóh	hàìtì-∅-Zóːh	'he spits it out'
[hàtɪ̀sóːh]	hadisóóh	hàtì-ʃ-Zóːh	'I spit it out'
[hàtòhsóːh]	hadohsóóh	hàtì-oh-Zóːh	'you two spit it out'
[tɪ̀lɪ̀t]	dilid	tì-∅-Lìt	'he's burning'
[tɪ̀ʃɬɪ̀t]	dishłid	tì-ʃ-Lìt	'I'm burning'
[tòhɬɪ̀t]	dohłid	tì-oh-Lìt	'you two are burning'
[tɪ̀ʒàh]	dizhah	tì-∅-Ʒàh	'he spits'
[tɪ̀ʃàh]	dishah	tì-ʃ-Ʒàh	'I spit'
[tòhʃàh]	dohshah	tì-oh-Ʒàh	'you two spit'
[jɪ̀ɰ̝ʷòʒ]	yiwozh	∅-Ɣòʒ	'he's ticklish'
[jɪ̀ʃxʷòʒ]	yishhozh	∅-ʃ-Ɣòʒ	'I'm ticklish'
[wòhxʷòʒ]	wohhozh	∅-oh-Ɣòʒ	'you two are ticklish'

As the above examples show, the stem-initial consonant is voiced when intervocalic^[12] and voiceless when it is preceded by a voiceless /ʃ-/ sh- first person singular subject prefix^[13] or a voiceless [h] in the /oh-/ oh- two person dual subject prefix.

Another example of contextual voicing of verb-stem-initial consonants occurs when a voiceless /-ɬ-/ -ł- classifier prefix occurs before the stem as in the following:

Phonetic form	Orthographic form	Underlying segments	Gloss
[tìːlzáːs]	diilzáás	tì-Vt-ɬ-Záːs	'we two dribble it along'
[jɪ̀tɪ̀sáːs]	yidisáás	jìtì-ɬ-Záːs	'he dribbles it along'
[tɪ̀sáːs]	disáás	tì-ʃ-ɬ-Záːs	'I dribble it along'
[tòhsáːs]	dohsáás	tì-oh-ɬ-Záːs	'you two dribble it along'

In the verb-form [tìːlzáːs] diilzáás ('we two dribble it along'), the /Z/ occurs between a voiced [l] and the voiced stem vowel [áː]. Thus it is realized as a voiced [z]. Here the /-ɬ-/ classifier is voiced due to the d-effect of the preceding /Vt-/ first person dual subject prefix. In the other verb-forms, the stem-initial /Z/ is preceded by voiceless /-ɬ-/ classifier which results in a voiceless realization of [s]. In the surface verb-forms, the underlying /-ɬ-/ classifier is not pronounced due to a phonotactic restriction on consonant clusters.

The initial consonant of noun stems also display contextual voicing:

Phonetic form	Orthographic form	Underlying segments	Gloss
[sàːt]	saad	sàːt	'language'
[pɪ̀zàːt]	bizaad	pì-sàːt	'his language'
[ɬɪ̀t]	łid	ɬìt	'smoke'
[pɪ̀lɪ̀t]	bilid	pì-ɬìt	'his smoke'
[ʃàːʒ]	shaazh	ʃàːʒ	'callous'
[pɪ̀ʒàːʒ]	bizhaazh	pì-ʃàːʒ	"his callous"
[xʷòʃ]	hosh	xòʃ	"cactus"
[pɪ̀ɰ̝ʷòʃ]	biwosh	pì-xòʃ	"his cactus"

Here an intervocalic context is created by inflecting the nouns saad, łid, shaazh, hosh with a [pɪ̀-] bi- third person prefix which ends in a vowel. In this context, the stem-initial consonant is voiced. When these nouns lack a prefix (in which case the stem-initial consonant is word-initial), the realization is voiceless.

However, in some noun stems, the stem-initial continuant does not voice when intervocalic: [ʔàʃĩ̀ːh] ashįįh ('salt').

Dorsal place assimilation

The dorsal consonants /k, kʰ, kʼ, x, ɣ/ (written ⟨g⟩, ⟨k⟩, ⟨kʼ⟩, ⟨h⟩, ⟨gh⟩) have contextual phonetic variants (i.e. allophones) varying along place of articulation that depend on the following vowel environment. They are realized as palatals before the front vowels ⟨i⟩ and ⟨e⟩ and as velars before the back vowels ⟨a⟩ and ⟨o⟩. Additionally, they are labialized before the rounded back vowel ⟨o⟩. This likewise happens with the velar frication of the aspirated /tʰ/.

Phoneme	Allophones
	Palatal	Velar	Labial
k	[c(ʝ)]	[k(ɣ)]	[k(ɣ)ʷ]
kʰ	[cç]	[kx]	[kxʷ]
kʼ	[cʼ]	[kʼ]	[kʼʷ]
x	[ç]	[x]	[xʷ]
ɣ	[j˔]	[ɰ̝]	[ɰ̝ʷ]
tʰ	[tç]	[tx]	[txʷ]

Coronal harmony

Navajo has coronal sibilant consonant harmony. Alveolar sibilants in prefixes assimilate to post-alveolar sibilants in stems, and post-alveolar prefixal sibilants assimilate to alveolar stem sibilants. For example, the si- stative perfective is realized as si- or shi- depending upon whether the stem contains a post-alveolar sibilant. For example, while sido ('it is hot' perfective) has the first form, shibeezh ('it is boiled' perfective), the stem-final /ʒ/ triggers the change to /ʃ/.

D-effect

A particular type of morphophonemic alternation (or mutation) occurring in Athabascan languages called d-effect is found in Navajo. The alternation in most cases is a fortition (or strengthening) process. The initial consonant of a verb stem alternates with a strengthened consonant when it is preceded by a /-t-/ (orthographic ⟨-d-⟩) "classifier" prefix or the /-Vt-/ first person dual subject prefix.^[14] The underlying /t/ of these prefixes is absorbed into the following stem. D-effect can be viewed prosodically as the result of a phonotactic constraint on consonant clusters that would otherwise result from the concatenation of underlying segments.^[15] There is thus an interaction between a requirement for the grammatical information to be expressed in the surface form and an avoidance of having sequences of consonants (see the syllable section for more on phonotactics).

The fortition is typically a change from continuant to affricate or continuant to stop (i.e. adding a period of closure to the articulation). However, other changes involve glottalization of the initial consonant:

Prefix consonant + Stem-initial consonant[16]		Surface consonant	Example
t- + -Z	→	-ts	/tʃʼéná-t-Zìt/ → [tʃʼéná-tsìt] chʼénádzid ('he woke up')
/t- + -L	→	/-tl	/ʔánéìnì-t-Laː/ → [ʔánéìnì-tlaː] ánéinidlaa ('you repaired it')
/t- + -Ʒ	→	/-tʃ	/ʔákʼídíní-t-Ʒéːʔ/ → [ʔákʼítíní-tʃéːʔ] ákʼídíníjééʼ ('you spit on yourself')
/t- + -j	→	/-ts	/nìː-Vt-jòɬ/ → [nìː-tsòɬ] niidzoł ('we two are driving them along') (cf. /jìnòː-jòɬ/ yinooyoł 'he's driving them along')
/t- + -Ɣ	→	/-k	/jì-Vt-Ɣòʒ/ → [jìː-kòʒ] yiigozh ('we two are ticklish') (cf. /jì-ɣòʒ/ yiwozh) 'he's ticklish')
/t- + -ʔ	→	/-tʼ	/nànìʃ-t-ʔìn/ → [nànìʃ-tʼìn] nanishtʼin ('I'm hidden')
/t- + -m	→	/-mʼ	/jì-Vt-màs/ → [jìː-mʼàs] yiiʼmas ('we two are rolling along') (cf. /jì-màs/ yimas 'he's rolling along')
/t- + -n	→	/-nʼ	/náːtòː-t-nìːt/ → [náːtòː-nʼìːt] náádooʼniid ('she said again')
/t- + -j	→	/-jʼ	/xònì-Vt-jóí/ → [xònìː-jʼóí] honiiʼyóí ('we two are energetic') (cf. /xònìʃ-jʼóí/ honishyóí 'I'm energetic')

The two occurrences of t- + -j in the chart above reflect two different patterns of d-effect involving stem-initial /j/. Often different underlying consonants are posited to explain the different alternation. The first alternation is posited as a result of underlying t- + -ɣ leading to a d-effect mutation of /dz/. The other is t- + -j resulting in /jˀ/. (See the velar /ɣ/, palatal /j/ section for further explanation.)

Another example of d-effect influences not the stem-initial consonant but the classifier prefix. When the /-Vt-/ first person dual subject prefix precedes the /-ɬ-/ (orthographic ⟨-ł-⟩) classifier prefix, the /-ɬ-/ classifier is realized as voiced [l]:

Prefix consonant + Classifier consonant		Surface consonant	Example
t- + -ɬ-	→	-l-	/jì-Vt-ɬ-Ʒõ̀ːh/ → [jìː-l-ʒõ̀ːh] yiilzhǫǫh ('we two tame it')

Other

• n > high tone
• expressive x clusters

Acoustic phonetics

Navajo F1 vs F2-F1 in Hz with mel scaling based on the median formant values in McDonough (2003). Here the vowels are labeled in orthography: ⟨ii⟩ = /iː/, ⟨i⟩ = /i/, ⟨ee⟩ = /eː/, ⟨e⟩ = /e/, ⟨aa⟩ = /aː/, ⟨a⟩ = /a/, ⟨oo⟩ = /oː/, ⟨o⟩ = /o/.

F1 vs F2-F1 in Hz with mel scaling.^[17]).

McDonough (2003) has acoustic measurements of the formants of Navajo long and short vowel pairs as pronounced by 10 female and 4 male native speakers. Below are the median values of the first (F1) and second (F2) formants for this study.

Oral Vowels from McDonough (2003)

Vowel	F1	F2		Vowel	F1	F2
iː	372	2532		oː	513	957
i	463	2057		o	537	1154
eː	487	2195		ɑː	752	1309
e	633	1882		ɑ	696	1454

An earlier study (McDonough, Ladefoged & George (1993)) has measurements from seven female speakers:

Oral Vowels from McDonough, Ladefoged & George (1993)

Vowel	F1	F2		Vowel	F1	F2
iː	315	2528		oː	488	943
i	391	2069		o	558	1176
eː	498	2200		ɑː	802	1279
e	619	2017		ɑ	808	1299

Tonological processes

Navajo nouns are simple: /kʰṍ/ [kʰṍ] kǫ́ ('fire'), /pi-tiɬ/ [pìtìɬ] bidił ('his blood'). Most long nouns are actually deverbal.

In verbs, with few exceptions, only stems may carry a high tone: C V(ː)(C)(T). Prefixes are mostly single consonants, C-, and do not carry tone. The one exception is the high-tone vocalic prefix /ʌ́n/-. Most other tone-bearing units in the Navajo verb are second stems or clitics.

All Navajo verbs can be analyzed as compounds, and this greatly simplifies the description of tone. There are two obligatory components, the "I" stem (for "inflection") and the "V" stem (for "verb"), each potentially bearing a high tone, and each preceded by its own prefixes.^[19] In addition, the compound as a whole takes 'agreement' prefixes like the prefixes found on nouns. This entire word may then take proclitics, which may also carry tone:

clitics=	agreement–	(prefixes–	I-stem)	+	(prefixes–	V-stem)
tone			tone			tone

(Hyphens "–" mark prefixes, double hyphens "=" mark clitics, and plus signs "+" join compounds.)

Any high tones on clitics spread to the next syllable of the word only if it is short and located immediately before the verbal stem. This can be seen with the iterative clitic /ná/꞊. Compare

ha=ni-sh+ł-chaad

hanishchaad

'I card it (wool)'

and

ha=ná=ni-sh+ł-chaʼ

hanáníshchaʼ

'I usually card it (wool)'

where the clitic ná= creates a high tone on the following short pre-stem syllable in bold, but,

ha=ná=ni-iid+ł-chaʼ

hanániichaʼ

'we usually card it (wool)'

and

ha=ná=da=ni-oh+ł-chaʼ

hanídanołchaʼ

'you (pl) usually card it (wool)'

where it does not.

• conjunct prefixes in verb stems are unmarked for tone (with a few exceptions) – they assimilate to the tones of neighboring prefixes
• tones in disjunct prefixes and stems are underlying specified
• certain enclitics (like the subordinator ⟨=go⟩) affect the tones of preceding verb stems

Peg elements, segment insertion

All verbs must be at least disyllabic. Some verbs may only have a single overt nonsyllabic consonantal prefix or a prefix lacking an onset, or no prefix at all before the verb stem. Since all verbs are required to have two syllables, a meaningless prefix must be added to the verb to fulfill the disyllabic requirement. This prosodic prefix is known as a peg element in Athabascan terminology (Edward Sapir used the term pepet vowel). For example, the verb meaning "she/he/they is/are crying" has the following morphological composition: Ø-Ø-cha where both the imperfective modal prefix and the third person subject prefix are phonologically null morphemes and the verb stem is -cha. In order for this verb to be complete a yi- peg element must be prefixed to the verb stem, resulting in the verb form yicha. Another examples are verb yishcha ('I'm crying') which is morphologically Ø-sh-cha (Ø- null imperfective modal, -sh- first person singular subject, -cha verb stem) and wohcha ('you [2+] are crying') which is Ø-oh-cha (Ø- null imperfective modal, -oh- second person dual-plural subject, -cha verb stem). The glide consonant of the peg element is ⟨y⟩ before ⟨i⟩, ⟨w⟩ before ⟨o⟩, and ⟨gh⟩ before ⟨a⟩.