- The fortis plosives are spelled p t k and the lenis plosives b d g.
- The fortis and lenis affricates are spelled c and z.
- The rhotic and palatal approximants are spelled r and y.
- A circumflex marks /j w/ as syllabic: ŷ ŵ.
Kennan has the monophthongs /a ɛ ɔ ɪ ʊ/, spelled a e o i u, and the rising diphthongs /ɛj ɔw ɛo aj oa/, spelled ei ou eo ai oa.
Additionally, /wɛo/ may perhaps be considered a phonemic triphthong, as this sequence not only occurs with much greater frequency than distributional averages would suggest, but shows morphological alternation with other vowels as well. However, this picture is complicated by the fact it becomes [lɛo] after labial consonants, and that both weo and leo often are clearly composed of separate w-l plus eo.
- The fortis/lenis contrast in Kennan is primarily a matter of late and early VOT (voice onset time) respectively. In brief descriptions, late VOT may be described as aspiration and early VOT as voicing, especially in prevocalic position. A lenis consonant gains voicing at any point in the range between immediately before its articulation and midway though it, whereas voicing does not begin until markedly after a fortis consonant's articulation is finished. Lenis consonants are always completely voiced when following another voiced or lenis segment, even across syntactic boundaries; the midway VOT occurs after voiceless consonants and in utterance-initial position.
- When followed by another consonant, the VOT distinction is realized primarily as a voicing contrast on that instead: approximants are quite prototypical glides and liquids after lenis consonants, but undergo devoicing and frication following fortis consonants. For example, dw may be phonetically transcribed as [dw] or [tw], while tw is phonetically [tʍ].
- Although listed among the fricatives, /h/ may more accurately be described as the [late VOT] feature attached to a null consonant, as it has the same affect on following segments that fortis consonants do, while normally lacking any independent frication. The exception is that a true glottal fricative appears in intervocalic environments that do not cross a word boundary.
- The other fricatives, /f s/, are universally voiceless, but do not show any of the delayed-VOT effects that fortis consonants do.
- While the affrication difference between s and c is rigorously maintained, the fully voiced realization of the lenis affricate z is deaffricated to [z] by many speakers.
- Nasals assimilate to the point of articulation of any following consonant, even across a syntactic boundary. Only before vowels and in utterance-final position do they contrast POA; by late Kennan (ca. 600 YP) even the utterance-final distinction was neutralized.
- Each approximant has a basic and a devoiced allophone. The devoiced realizations occur in two environments: after fortis consonants, and in word-final position. The devoiced allophones of r l y w are [ɹ̥ ɬ ç ʍ].
- Word-final nasals are usually devoiced as well.
- Approximant devoicing also tends to affect the diphthongs ai ei ou in word-final position, where they are most often pronounced [aç ɛç ɔʍ].
- All eight approximant allophones (basic and devoiced) can additionally be syllabic; these versions can be transcribed [ɹ̩ l̩ i u] for the basic approximants, and [ɹ̥ ɬ̩ ç̩ ʍ̩] for their devoiced counterparts, although [i̥ u̥] may be preferable for the latter two.
- The vowels i u are quite distinct from the consonants y w. The former pattern closely with the other vowels in that they are never fully voiceless, do not alternate with glides, have a "lax" vocal tract setting, and vary with bimoraic counterparts. They are pronounced [ɪ ʊ], and tend to be quite central - in fast speech, they sometimes appear as simply unrounded and rounded schwas respectively. Meanwhile, y w pattern closely with the other approximants r l in that they exhibit voicing and syllabicity alternations in all the same environments. Their syllabic realizations [i u] are pronounced markedly closed, and have a "tense" vocal tract setting, as do the basic glides [j w].
- The mid vowels e o are most often mid-open [ɛ ɔ], but mid-closed [e o] occur under prosodic stress.
- Stressed a is simply [a]. Unstressed a is underspecified for frontness and rounding, and varies among [æ a ɑ ɒ], often depending on the following consonant. Generally, [ɑ] appears before velar consonants, rounded [ɒ] before all Cw clusters, [æ] before Cj clusters, and in other environments a may vary freely among any of these and [a].
- The underlying sequences ja and wa, when unstressed, merge into /ɛ ɔ/. For morphological reasons it is simpler to treat and spell these as instances of e o.
Kennan syllabic structure is rather complex, so we will approach the topic from multiple angles.
For a soundbite description, it may be simplest just to say that all consonant clusters consist of the pattern (D)C(A), where A is any approximant (r l y w) and D is any of the permitted codas (s c z r l m n), and that the full (D)C(A) can occur in any position - initial, medial, or final. But this misses a lot of important detail.
On closer examination, Kennan syllables can be divided into full syllables, which have a full vowel (a e i o u) and minor syllables, which do not. Stressed syllables of content words are always full; other syllables can be full or minor. The underlying structure of Kennan contains only full syllables, while minor syllables are surface realizations of underlying codas or approximants that are no longer adjacent to a vowel, and which cease to be independently syllabic when placed next to one. Save for recent loanwords, Kennan morphemes are overwhelmingly monosyllabic at the underlying level.
Next, let us examine the core of a full syllable and work outward from there. As a general template, this core can be described with the pattern C(A)V(D). But it is not sufficient to declare this the "syllable pattern" and move on. Most notably, most medial clusters that can be formed by this pattern can also be found in word-initial or word-final position. That is to say, an underlyingly initial syllable can begin with a "coda" provided it is restricted to s c z, and a underlyingly final syllable can end in an "onset" C(A). In either case, initial "codas" and final approximants then become minor syllables on the surface level. Yet, it is once more not sufficient to say that minor syllables occur only peripherally to full ones and move on; for one thing, medial minor syllables exist as well, across morpheme boundaries (and especially in compound words). For another, a great many unstressed grammatical words consist only of minor syllables.
Another angle is that it may be appropriate to reconsider the usual representation of sonority peaks as the most central element of phonological structure. Instead, Kennan's structure may be considered to revolve around sonority valleys - that is to say, the usual syllable notation is in terms of required nuclei plus whatever material is permitted to surround them, whereas Kennan may be more appropriate to describe as requiring a cluster-central C, around which other material is permitted, in the pattern (V)(D)C(A)(V). It is true that stressed content words all have at least one full vowel as a syllable nucleus. but affixes and free grammatical words do not require them.
Finally, let us look at some examples.
- hon is a simple CVD monosyllable with no minor syllables.
- sdroyr is a fairly typical shape for a Kennan root, DCAV.CA. It has a single stressed vowel, and is thus an underlying monosyllable. On the surface, it has two additional minor syllables. There is a full DCA cluster, sdr, on the left, and a CA cluster, yr, on the right; both the s and the yr form minor syllables. The CA cluster cannot be interpreted as DC, because y is not a permissible coda. The minor syllable becomes a full syllable when the nominative suffix -as is added: sdroyras, DCAV.CA-VD, [sdɹɔ.jɹas].
- kadrten is a compound of kadr and ten; the minor syllable in the first root is simply sandwiched in the middle of the word.
- dwisklckoz is another compound with a minor syllable sandwiched in the middle. This one, however, is formed from two original minor syllables - a final l and an initial c, which combine into a single minor syllable. The whole word is pronounced [dwɪs.kɬ̩ʦ.kʰɔʣ]. This breaks down to CAVD.CA-D.CVD.
- yr, an unstressed deictic particle, has no full vowel at all. It is a minor syllable. Like before, y is not a permissible coda, and thus the r fills an A slot and gains syllabicity. So properly, it should only be pronounced [jɹ̩], not [iɹ]; however, free grammatical words are somewhat laxer about this than content words, and in fast speech the pronunciation could easily come out either way, or simply be indeterminate between the two. The closely related particle hŷr can similarly end up [çɹ̩] or [hiɹ], although the latter would be the more proper form.
- Gemination is possible for any consonant in intervocalic position, but such a geminate cannot cluster with anything else unless it would be a valid DC. Thus nny is a valid geminate cluster, but tty would not be, as t is not a possible coda.
- The sibilant s cannot follow a coda affricate.
- f can (and frequently does) appear in the approximant slot after s c t. Resulting clusters can still appear with preceding codas under the usual rules; sf and cf are also possible DC clusters that can take an A, but tf isn't because t is not a possible coda. Thus sfr is a permitted DCA cluster, and rtf, but not tfr.
- Similarly, s can appear in the approximant slot after k p, although this is much less common. h cannot appear in this slot at all.
- Coda nasals assimilate to the point of articulation of a following nasal, plosive or affricate, but not a fricative.
- w cannot be found in the approximant slot after p b f m.
- l cannot be found there after t d.
When morphology causes the collision of two consonants that do not make a legal (D)C(A) cluster, one or more is lenited or deleted according to the following rules, which are applied in order until a legal cluster is reached:
- In any cluster of four or more consonants that are not resolved by a medial minor syllable, the most medial consonant that is lowest on the sonority hierarchy is identified as the C, and any consonants immediately adjacent to it are deleted until there is only one consonant on each side of the C. For example: ansktra → ankra.
- After a vowel, lenis plosives before fortis ones reduce to nasals (which assimilate for POA): abta → amta → anta
- Following any plosive, a fortis plosive is reduced to a fricative: p t k → f s h. Exception: intervocalic geminates are allowed.
- A lenis plosive followed by a fricative becomes fortis: adfa → atfa, adsa → atsa, and adha → atha
- Following a fortis plosive, h is deleted as an independent consonant: atha → ata.
- The combination ts reduces to c: atsa → aca.
- After any plosive, affricate, or fricative in the C slot, voiced plosives and nasals reduce to approximants in the A slot: b m → w, d z n → r, and g → y. Exception 1: intervocalic geminates are allowed. Exception 2: no such reduction occurs when the sequence is becoming a DC instead of a CA; ansma → answa, but asmya → asmya, while intervocalic tm will always reduce to tw because t can't be a coda.
- Whenever fortis plosives still find themselves in a D slot despite the best efforts of all the preceding rules, t reduces to s, but k p will take over the C slot by force, re-triggering the first deletion rule - unless this would cause an important grammatical morpheme to be lost, in which case the k or p itself is deleted instead.
- w becomes l after a labial: pw → pl. The reverse happens after t d: dl → dw.
Note that the above rules do not cover all possible consonant combinations, just the most common ones. In general, other clusters will act in parallel to these to the extent reasonably possible, but can have rather idiosyncratic outcomes.