Otseqon phonology

Vowels
/e/ is somewhat low, perhaps closer to canonical [ɛ] than [e].

Short /i/ tends to become [ɪ] word-initially and before a syllabic nasal. Short /i/ becomes [ɨ] following /t tʼ/.

/o/ raises to [u] when the following syllable has an /i/. The pattern CoCoCi can go either to [CoCoci] or [CuCuCi]; this is somewhat idiosyncratic depending on the speaker/word. There are no minimal pairs between them.

Vowel devoicing
Short vowels are often devoiced between voiceless consonants or at the end of an utterance.

Vowel devoicing is stochastic and many factors influence whether a vowel is actually devoiced. Some general factors that influence vowel devoicing:


 * Vowels in more common lexical items are devoiced more often than vowels in rarer lexical items
 * H toned vowels are usually not devoiced
 * Two vowels in a row are usually not devoiced
 * Vowels before coda /h/ are frequently devoiced
 * Vowels are more frequently devoiced following /s/ than other consonants
 * The high vowel /i/ is devoiced more often than other vowels

Syllabic nasal
There is a syllabic nasal, denoted as /N/. Its place of articulation assimilates to the place of the following consonant, or tends to be [ŋ] pre-pausa.

Like other syllables, /N/ can bear tone. /N/ is generally not found at the beginning of a word, with the exception that word-initial /iN/ tends to be pronounced as [nn] / [mm] / [ŋŋ] (depending on the following consonant).

Consonants
/ʔ/ predictably occurs at the beginning of words otherwise beginning with vowels. There is an in-progress sound change causing it to be phonemic in some cases; see below. It also occurs in the common morpheme /ʔwa/ progressive auxiliary (< /iwa/ ‘stand’), as a result of reduction of the word-initial vowel.

The clicks are rare in basic vocabulary. They occur mostly in ideophones and in species names (including words for types of mythological creatures). The distinction between the two types of clicks is mainly one of [±abrupt release], i.e. the sharp release of /ᵑǃ/ versus the noisier more affricate-like release of /ᵑǀ/. Their POA is somewhat variable: the ǃ-clicks are dental-alveolar [ᵑǃ ᵑǃʼ] before non-front vowels and palatal [ᵑǂ ᵑǂʼ] before front vowels; the ǀ-clicks are generally dental, but are sometimes alveolar lateral clicks ([ᵑǁ ᵑǁʼ]).

/z/ is an affricate [ʣ] word-initially or following a long vowel. It devoices and merges with /s/ following a syllabic nasal.

Sonorants
/m n/ are realized as [w ɾ] between vowels. They are nasals everywhere else: [m n] word-initially and following a syllabic nasal (the sequences /Nm Nn/ are [mm nn]), and voiceless nasals in the clusters /hm hn/ [hm̥ hn̥].

Palatalization
Most consonants palatalize following /i/. /c cʼ s z/ go to [ʨ ʨʼ ɕ ʑ]; /p t k/ are [pʲ tʲ kʲ]; /ᵑǃ ᵑǃʼ/ are [ᵑǂ ᵑǂʼ]; /h/ becomes [ç]. /m n/ and the ǀ-clicks are largely unaffected. [tʲ] sometimes goes even farther, to [ʦ] or [ʨ].

This palatalization sometimes "bleeds through" to a following /a/, turning it into [æ]. That is, a sequence like /ika/ is phonetically [ikʲæ].

Some palatalization also takes place preceding /i/. Word-initial /si zi ci cʼi/ are [ɕi ʥi ʨi ʨʼi] and this is optionally the case word-medially (which is particularly common in the case of a word that already has another phonetic [ɕ]). /ᵑǃi ᵑǃʼi/ are also always realized with the palatal allophone.

Long /eː/ also usually triggers palatalization of the following consonant. A long of long /eː/s are the result of fusion of vowel sequences ending in /i/, and the palatalization is preserved and possibly root-internal /eː/s also now palatalize by analogy.

Plosive realization
/p t k/ are aspirated word-initially and when geminated. /p t/ are voiced [b d] intervocalically (but not following a syllabic nasal). The realization of /k/ is a bit more complicated, and it can be any of [k q g x ɣ]. Generally /k/ is [q] in the vicinity of /o/, unless in contact with an /i/; intervocalic /k/ is otherwise [g], except that /k/ the sequence /aka/ is optionally lenited to [x] or [ɣ]. It is probably helpful to list the possibilities:


 * q: oqo, aqo, oqa, aqa, eqo, ɴqo, #qo, #qa, hqo, hqa, ɕqo, ɕqa
 * k: ekV́, akV́, ŋka, ŋke, #ka, #ke, hka, hke, ɕka, ɕke
 * g: egV̀, agV̀
 * kʲ: ikʲV́
 * gʲ: igʲV̀
 * x: axá
 * ɣ: aɣà

Intervocalic voicing is restricted to the onset of a low-toned syllable. Intervocalic plosive onsets of a high tone syllable are unaspirated and unvoiced.

/c/ is immune to voicing.

Marginal phonemes
pʲ, pʲʼ occur in some ideophones. They are also used to substitute for /p pʼ/ when talking to certain animals.

kʷ, kʷʼ are used to substitute for /p pʼ/ when reading sufficiently old texts, or generally when trying to make something sound archaic (they are the relatively recent diachronic source of /p pʼ/). They also occur in some ideophones.

In-progress sound changes
The following sound changes are not fully complete: they may not be used by all speakers, not used all the time, do not affect all words, etc. In general they are much more likely to affect more common words than less common ones.

ari > ai
This does not further coalesce to /eː/.

#ye > ɪ
Word-initial /ye/ is occasionally pronounced as [ɪ]. This almost merges with /i/, which is also pronounced [ɪ] initially, but word-initial /i/ has a preceding glottal stop which is missing from the vowel resulting from this sound change. This makes /ʔ/ phonemic in this one instance.

uwi > ui, iyV > iV
Glides are often absorbed following their vocalic counterparts. Note that this does not result in vowel coalescing.

Vowel coalescing
Sequences of adjacent short vowels are disallowed, and are repaired by (in most cases) fusing the vowels together into one long vowel, following the chart below:

As shown, sequences starting with /i/ are not fused but instead insert an epenthetic /y/.

In sequences of a long vowel plus a short vowel the short vowel is deleted.

Sequences of two vowels are allowed but only occur on morpheme boundaries.

Grassmann's law for ejectives
Ejectives are governed by a sort of OCP where if an ejective is followed by another ejective in the next syllable, the first turns into the corresponding tenuis plosive.

Tonal phonology
Otseqon has a register tone system opposing marked H with toneless mora. The actual number of phonetic tone levels is complicated by various factors such as downdrift. The tone-bearing unit is the mora. The system is somewhere between a full register tone system and an accentual system. Various phonological processes serve to eliminate all underlying H tones in a word except for one or two, but surface H tones can span multiple TBUs.

Lexical tone
Otseqon roots, lexical suffixes, and prefixes can be either high toned (H) or toneless. Suffixes other than lexical suffixes are all toneless with the exception of the a purely tonal H suffix in conditional clauses.

There is a third tone class possible for roots, downstepped H (!H). !H roots are identical to H roots in isolation or following an L tone, but are downstepped following an H tone. For most tonological processes, !H roots sort of behave like they have a floating L tone before the first syllable of the root, although that L tone is never actually realized. For example, while a normal H root following a high toned person prefix causes deletion of the tone from the prefix to avoid consecutive H tones, an !H root does not (and has the same pitch as the prefix), as if a plateau had formed over an intervening L tone.

H spreading
All H tones spread one TBU to the right. In imperative clauses, H spreading is unlimited, and an H tone will spread to the end of the word. It's worth noting that this can't spread into another H tone, because the only H toned suffix is the marker of conditional clauses, which obviously doesn't appear in imperatives.

H spreading is blocked if it would run up against another H and no plateau is formed.

Conditional H tone
Conditional clauses and clause-external topics acquire a purely tonal H morpheme. The position of this tone depends on the tonal class of the root. On toneless roots, it surfaces on the second TBU (recall that phonological words are minimally bimoraic) and then undergoes spreading. On H tone roots the conditional tone aligns with the right edge of the word. On bimoraic words, the conditional H tone fails to surface, but not before blocking spreading of the lexical H tone. The possible tonal patterns are given below. Note that underived Otseqon roots are generally two syllables and never longer than 3 and longer words are the result of affixation.

Plateauing
Otseqon tends to avoid "troughs", HLH sequences. In most cases where there would be an HLH the H tones fuse and form an HHH plateau. This is true whether the left H tone is a specified H tone or results from spreading.

The only exception to plateauing is that plateaus do not form with H introduced from conditional marking. This is true for both a root H tone + conditional tone and for high tone person prefix + toneless root + conditional tone.

H deletion
In sequences where an H tone is immediately adjacent to another H tone one of them is deleted. Root tones are prioritized, so when an H toned prefix precedes an H root the first tone is deleted, and when an H toned lexical suffix follows an underlyingly monomoraic H root the second tone is deleted (were the root not monomoraic a plateau would form instead). When two H toned prefixes are adjacent the second H is deleted. This happens before plateaus are formed, which can make it look like nothing is deleted at all, since the second prefix can "recover" its tone by participating in a plateau.

Downstepped H roots (!H) behave as if they had a preceding L tone, so H on prefixes is not deleted before a !H root but rather forms a plateau.

Note that in the case of H + H + LL one can tell the second H tone has been deleted because it doesn't spread onto the root, although the prefix has recovered its own H tone from spreading from the first H.

Downdrift
Otseqon exhibits automatic downdrift, the lowering of the pitch range of H tones following L tones.

L tones also downdrift slightly in Otseqon, although not as dramatically as H tones.

The absence of downdrift is a prosodic marker of various clause types, such as questions. Downdrift is mostly associated with declarative clauses.

Accentual tone
It would be easy to assume that toneless morae come out with low tone, but this is actually not the case. Rather, sequences of toneless morae take on a tonal melody beginning with a low tone and rising to a high tone, although the high tone of the accentual tone is not as high as that of lexical H tones. This is referred to as "accentual tone", as the tone is determined entirely by the placement of the H accent.

This means the actual tonal pattern of toneless roots is ˩˧˧˧˧, while the tone pattern of H roots is ˦˥˩˧˧ (see §§ HL upstep below).

Accentual tones are not entirely inert in the phonology. In an H toned verb marked with the conditional H tone they form a plateau (or more of a slope) with the conditional H tone (provided there are enough intervening toneless morae), so the tonal contour of a word like HHLLLH is a high-rise followed by a sharp fall and then a gradual rise to the end of the word ˦˥˩˨˧˦.

Accentual tones plateau with the accentual tone of the following word in verb - object sequences if the object is toneless.

Some syntactic constructions prevent the insertion of accentual tone. Accentual tone does not appear on vocatives, on proper nouns in subject position (it does apply to objects), right-dislocated "afterthought" constituents, or adverbs in immediately preverbal position. In imperative clauses there is unlimited H spreading, which prevents insertion of accentual tone on H-toned words (however, it is still inserted on toneless words).

Final lowering
The entire pitch range is lowered at the end of an utterance in declarative and related clauses. On words ending with an L, this manifests as a fall from the accentual tone to an extra-low tone, while on words ending with H the H tone rises barely if at all and is essentially the same level as an L tone.

If a final H tone is multiply-linked as a result of spreading or a plateau, all surface tones are lowered.

HL upstep
H tone before L is upstepped slightly. The main significance of this is that the second H tone in a spread H tone is a bit higher than the first one, that is, the spread H tone is really sort of a high-rising tone, and the tonal contour of an underlyingly HHLLL word is basically high-rise followed by sharp fall before levelling out to the accentual tone level.

The way this works with downdrift is that it does not raise the downdrifted H level but does go slightly above it.

Declaratives
Declaratives are marked by downdrift and final lowering.

Polar questions
In polar questions the pitch range is widened and shifted upward somewhat compared to declaratives and downdrift is absent. Rather than final lowering like is found in declaratives there is a slight rise toward the end of the utterance.

na questions
This type of polar question is marked by a final particle na and expresses that the speaker expects an affirmative answer (akin to ‘…, right?’). Intonationally, these are more similar to declaratives than to other polar questions, i.e. they have downdrift and a similar pitch range. However like other questions they lack final lowering, although the rising intonation is not as apparent.

Imperatives
Imperatives also lack downdrift and final lowering, but their base pitch level is comparable to declarative clauses. The most distinctive feature of imperative clauses is that instead of spreading one TBU to the right, H tones spread all the way to the end of the word.