Chapter 5Morphophonology

5.1 Coda alternations: h with p, t, k

Stems whose final stop has been neutralised to h (Chapter 4) show the stop whenever a vowel follows within the word. The alternation runs through both nominal and verbal morphology. With the possessive suffix (Chapter 7): teh ‘hand’ but tek-ihi ‘his hand’, otoh ‘hair’ but otop-ihi (Tangiku 2022: §2.3). With person suffixes: itah ‘speak’ but itak-an ‘we spoke’ and itak-ahci ‘they spoke’ (Tangiku 1998: §§2-2, 6). Before i the recovered t is regularly affricated (Chapter 4): mah ‘wife’ : mac-ihi ‘his wife’, kuh ‘belt’ : kuc-i ‘his belt’ (Chiri 1942: §16,6). And where the coda has shifted to s after i, the velar returns intact: sis ‘eye’ : sik-ihi ‘her eyes’, the latter frequent in Asai Take's tales (sikihi masaha ‘her eyes opened’, (Murasaki 2001: text 7)). We therefore set up underlying stem-final /p t k/ and let the neutralisation rule of Chapter 4 apply in coda position, in line with Chiri's and Tangiku's analyses and with Dal Corso's notation itak* for surface itah (Dal Corso 2024: 61).

The same alternation underlies the basic numerals: sine-h ‘one (thing)’, tu-h ‘two’, re-h ‘three’ contain the nominaliser -h, the Sakhalin outcome of *-p (Hokkaidō sinep, tup, rep); Chiri already used the suffix pair -pe (post-consonantal) ~ -h (post-vocalic) as a diagnostic of stem shape (Chiri 1942: §2(5)) (see Chapter 10).

Across word boundaries usage differs by source. Chiri describes liaison: the stop surfaces before a vowel-initial word, as in tek ani ‘with the hand’, ek-an ‘I came’, mat-ah ‘younger sister’ (Chiri 1942: §16,1), and the letters of Sentoku Tarōji show spellings like cep ukoyki ‘fish-catching’ (Tangiku & Ogihara 2001: Letter 1). In Murasaki's West-coast transcriptions, by contrast, the neutralised form stands even before a vowel — teh ani in the example above — which we attribute to the automatic glottal onset of the following word: liaison applies only where no glottal stop intervenes, hence within the word but not across it in these dialects. Between identical vowels and before stops the debuccalised segment may also surface between words in Piłsudski's texts (nax an ‘being so’, itax ki ‘make a speech’) (Dal Corso 2024: 62).

5.2 The possessive suffix and vowel copying

The possessed (“belonging”) form of the noun is built with a suffix of the shape -hV on vowel-final stems and -VhV on consonant-final stems (Chapter 7 treats its syntax). On vowel-final stems the suffix vowel copies the stem-final vowel: sapa ‘head’ → sapa-ha, cise ‘house’ → cise-he, mahpoo ‘daughter’ → mahpoo-ho (Murasaki 1979); (Tangiku 2022: §4.1). On consonant-final stems the two vowels are identical to each other but lexically determined by the stem, with strong tendencies set by the final consonant — e after y w, i or u after nasals and stops: tek-ihi, ram-uhu ‘heart’, kam-ihi ‘meat’, kotan-uhu ‘village’ ((Murasaki 1979), as tabulated in (Tangiku 2022: §4.1); (Hattori 1964)). Sakhalin has only the long form: Hokkaidō's short possessive (tek-e) has no West-coast counterpart (Tangiku 2022: §4.1).

Lengthened monosyllabic stems regularly shorten before -he, since the suffixed word is already bimoraic and the lengthening that made the free form heavy is no longer needed: kuu ‘bow’ ~ ku-he, nii ‘tree’ ~ ni-he (Chiri (1952: 463), (Chiri 1942: §72); Chapter 7). The shortening is not exceptionless, however: ree ‘name’ surfaces unshortened in the Yamada Hayo corpus (reehe tah nee manu ‘his name was this’, (Kitahara et al. 2003)). Polysyllabic stems in ii uu shorten the vowel and surface a glide before -ehe: imii ‘garment’ → imi-yehe, and ruu ‘road’ → ruwe-he (Tangiku 2022: §4.1), the coda ban on long nuclei (Chapter 4) forcing the shortening.

Two analyses of the suffix have been proposed. Chiri derives it historically from emphatic doubling of the final vowel with h inserted to break the hiatus: ona ‘father’ → onaa → onaha, siki → sikii → sikihi (Chiri 1942: §18,2). Dal Corso, for the East coast, takes the underlying form to be -(i)hi, with the first vowel of -VhV an optional epenthesis resolving the consonant + h cluster — supported by free variation of kotanuhu and kotanhu in Piłsudski's corpus — and with h deletable after nasals, giving kotanu (Dal Corso 2024: 66). The copy pattern of the West-coast data is stated most simply by Chiri's account, which we adopt for this grammar while noting that the East-coast variation favours Dal Corso's.

5.3 Person prefixes and clitics before vowel-initial stems

The person markers (Chapter 13) end in a vowel (ku= e= eci= i=) or in n (an=), and their combination with vowel-initial stems is the main laboratory of hiatus resolution. Vowel-final markers simply stand before the stem vowel, which keeps its glottal onset: Chiri transcribes an-e ‘we eat it’ as [aneː] and e-e ‘you eat it’ as [eeː], with the clitic vowel short and the accented stem vowel lengthened (Chiri 1942: §1(2)); Murasaki's apostrophe spellings of the type 'e'ani ‘you’ make the same point for the 1960s data. In the nineteenth-century West-coast sentences recorded by Dobrotvorsky and in Sentoku Tarōji's letters the pattern is directly attested with an= before vowel-initial stems.

Identical vowels at a marker boundary may coalesce: in Asai Take's tales eci=itak-ahci ‘they call you’ is realised ecitakahci ((Murasaki 2001: text 14)). We analyse this as the same-vowel contraction that Chiri documents inside compounds (keran ‘tasty’ from kera-an ‘flavour-exist’, sinan from sine-an ‘a certain’; (Chiri 1942: §14,1)). Unlike vowels in derivation are instead separated by a glide that copies the frontness or rounding of the first vowel: i-uta → iyuta ‘pound (grain)’, u-osurpa → uwosurpa ‘throw one another away, divorce’ (Chiri 1942: §15,1) — the same glide insertion seen in imi-yehe and ruwe-he above.

The n-final clitic an= assimilates to a following sonorant: an=ye ‘we say’ is pronounced ayye, an=wante ‘we know’ awwante (Chiri 1942: §16,3); before s the nasal dissimilates to y (ay-sike from an=sike ‘my load’), and before the desiderative the sequence an=kor rusuy surfaces as an=kon rusuy (§5.4). The corpus spelling an=yee ((Kitahara et al. 2003)) is morphophonemic; we take the phonetic value to be [ayyee], following Chiri.

5.4 r-alternations and the two kinds of -rV stems

Since r is banned from codas, a stem-final *r surfaces with an echo of the preceding vowel: koro ‘have’, nukara ‘see’, mukara ‘axe’, kisaru ‘ear’ for Hokkaidō kor, nukar, mukar, kisar; less commonly the r joins the general coda neutralisation and yields h, as in utah beside utara ‘people’ and mahnekuh ‘woman’ ((Chiri 1942: §16,4); (Dal Corso 2024: 64)). Word-internally the old clusters survive as h + stop: ohta ‘at’ from or ta, unahpe ‘aunt’ (Tangiku 1998: §3-2).

Hattori took the echo vowel to be a real, historically added vowel, so that the Sakhalin verbs end in vowels and Hokkaidō preserves the older shape; Chiri treated it as an automatic repetition of the preceding vowel. Tangiku (1998) shows that for the West coast Chiri was right at the level of the underlying form: verbs in -rV fall into two classes that behave differently under suffixation. Class I verbs are those whose Hokkaidō cognates end in coda r — koro ‘have’ ~ kor, nukara ‘see’ ~ nukar, mokoro ~ mokor: they are consonant-final at the deeper level, and their surface vowel is the epenthetic echo, inserted only to open the syllable. Class II verbs are vowel-final in Hokkaidō as well — tura ‘accompany’, poro ‘be big’, kira ‘flee’ — and their final vowel is lexical, part of the stem itself. Nothing in the citation forms tells the two apart; only their behaviour under suffixation does.

The diagnostic environments are two. First, the causative: -te attaches to consonant-final stems and -re to vowel-final ones, and Class I koro, nukara, kara, mokoro select -te with the regular shift of r to n before it (kor-te → konte), while Class II poro, kira, tuuri select -re like any vowel-final stem (Tangiku 1998: §4). Lexicalised causatives preserve the same treatment: cinupunte ‘ennoble’ from ci-nupuru-te (Tangiku 1998: §4). Second, the desiderative rusuy and rayki ‘long to’: after the same four verbs the two rs dissimilate (kon rusuy, nukan rusuy, mokon rayki), whereas tura rusuy, koore rusuy keep their vowels (Tangiku 1998: §3-4).

The shape of the alternation is itself a dialect diagnostic, perhaps the sharpest single one the verbal morphology offers. Hokkaidō causativises its r-final stems with a third allomorph -e — kor-e ‘give’, nukar-e ‘show’ — so the cognate verbs there never develop the nasal (Chiri 1942: §55). Sakhalin lacks -e entirely: a Class I stem takes -te with r → n on both coasts, from Dobrotvorsky's 1860s sentences and Sentoku Tarōji's East-coast letters ((Tangiku & Ogihara 2001: Letter 1)) to Asai Take's tales a century later. A -rV verb that yields a causative in -nte or a desiderative in -n rusuy is thereby identifiable as Sakhalin at a glance, and the pair konte : kore draws the isogloss between the two branches in a single word.

The rules are ordered: suffixation of -te and rusuy applies to the consonant-final stem, then the echo vowel is inserted, and only then do the person-number suffixes attach (Tangiku 1998: §7). Hence the plural of koro is koro-hci, with the post-vocalic allomorph -hci rather than post-consonantal -ahci (§5.5), and the desiderative likewise takes -ahci outside it:

The deletion also has a sharp outer boundary. The causative and the desiderative belong to the verbal complex and reach the underlying stem, but the homophonous onset of the converb teh ‘and, having …-ed’ does not trigger the rule: koro teh is never reduced to the expected sandhi outcome *koh teh, although r + t yields ht word-internally (§5.5) (Tangiku 1998: 225). In the ordering just given the echo vowel is in place before clause-level material attaches, and is then protected. The corpus bears the prediction out without exception: koro teh, mokoro teh and yaycisekoro teh recur throughout the West-coast texts, and the reduced form never occurs — the affix/converb boundary is exactly where the rule stops.

In Asai Take's Odasu tales the same morphology recurs (nukan rusuyahci ‘they wanted to see’, (Murasaki 2001: text 52)), so the system holds across the West-coast dialects. A solitary counterexample is e-hecinte ‘let play there’ from hecire ‘dance, play’, the only verb in -V₁rV₂ with unlike vowels that takes -te; Tangiku connects it with the East-coast and Hokkaidō variant heciri (Tangiku 1998: §5-2). The choice of echo vowel is occasionally lexical rather than copying — kisaru ‘ear’ matches the Hokkaidō possessed stem kisar-u (Tangiku 2022: §2.4), and the possessed form kisaru-hu appears in the Yamada Hayo corpus ((Kitahara et al. 2003)).

5.5 Plural allomorphy and connected speech

The verbal plural suffix has the shape -ahci after consonants and -hci after vowels, including inserted echo vowels: cis-ahci ‘they wept’, itak-ahci ‘they spoke’, tura-hci, koro-hci, ariki-hci (Tangiku 1998: §6) (see Chapter 8). Furukawa reports a Rayciska tendency for stems in u to take -ahci with vowel replacement (konupuru → konupurahci ‘they like it’), which the Odasu speaker does not share (yaykonopuru-hci) — dialect or idiolect variation that the small corpus cannot resolve ((Furukawa 1967); (Tangiku 1998: note)).

On the East coast the suffix shows further erosion: after monosyllabic open-syllable verbs the h may drop (nu-ci ‘they heard’ beside expected nu-hci), a process never found in the West-coast sources (Dal Corso 2024: 65–66); Chiri had already noticed the conditioning, assigning -si -ci to long-vowel stems and -hsi -hci to short ones (Chiri 1942: §41). The parallel deletion in the possessive (kotanu for kotan-[h]u) was noted in §5.2. Sentoku Tarōji writes itaku=hci ‘they say’ with an East-coast support vowel where the West coast has itak-ahci ((Tangiku & Ogihara 2001: Letter 1)) — we read his -u- as the same cluster-resolving epenthesis that Dal Corso posits for -VhV.

Connected speech is governed by a small set of assimilations and dissimilations, most of them already catalogued by Chiri (1942: §§14–16): r assimilates to a following n (cise-kon-nispa ‘master of the house’ from koro), r + t gives tt in Hokkaidō but ht in Sakhalin (kapah tuki ~ kapara tuki ‘thin cup’), n becomes m before labials (pom mahpoo ‘little girl’), and n m dissimilate to y before s (amay-su ‘rice pot’ from amam-su; East-coast tuy śata from tum sa ta ‘in the middle’, (Dal Corso 2024: 65)). Vowel contraction trims frequent collocations (setoko from si-etoko ‘ahead of oneself’), and h flanked by a consonant and a vowel deletes (sinotoripi from sinot-horipi ‘play-dance’) (Chiri 1942: §§14–15).

Reduplication, finally, is lexical rather than inflectional in the recorded material. Body-part and place nouns show fossilised doubling of a final root (etu-puy ‘nose’ : etu-puy-puy ‘nostril’, supuy-puy ‘anus’, (Chiri 1942: §19,1)), verb-stem reduplication marks plurality or iteration of the event ((Dal Corso 2024: 55–56)), and sound-symbolic doublets such as purupuru carry the refrains of Asai Take's sung tales ((Murasaki 2001: texts 34, 54)). We find no productive grammatical reduplication in the corpus: plurality is carried by -hci/-ahci and suppletive stems (Chapter 8).