Chapter 157The Final-h Problem and Its Sakhalin Vowel-Length Reflex

157.1 The h phoneme and the coda question in Hokkaido

Hokkaido Ainu h is an onset segment. It occurs word-initially — hapo 'mother', huci 'grandmother' — and as the opening consonant of the affiliative suffix -hi/-he/-hu/-ha. It does not occur as a syllable coda Tamura (1996). The descriptive gap between the two scripts — Hokkaido retaining original coda consonants, Sakhalin replacing them with h — makes the phoneme appear to play an outsized role in cross-dialect description, yet for Hokkaido the question is straightforward: the h visible inside forms such as tek-ehe 'hand (of someone)' or sik-ihi 'eye (of someone)' is suffix-initial, not base-final. The base consonants — k in tek, k in sik — surface unchanged across all Hokkaido morphological environments ‹consensus›.

The analysis that makes this transparent is Janhunen's (2020, via Alonso de la Fuente (2022: §2.1)) account of the affiliative -VhV. The proto-Ainu noun base was disyllabic — *tek.e 'hand', *sik.i 'eye' — with a final vowel subsequently lost in the citation form. When the noun hi 'thing, place, time' fused with the base, the lost base-vowel was restored before the suffix-initial h: *tek.e → tek-ehe, *sik.i → sik-ihi. The h in the affiliative is therefore the onset of hi, and the suffix vowel harmonises with the base's final vowel: -ehe (base e), -ihi (base i), -uhu (base u), -aha (base a). The full morphology belongs in Chapter 44 (Concept Form vs. Affiliative (Possessed) Form) and Chapter 45 (Morphophonology of the Affiliative Suffix and Echo/Copy Vowels); the point here is that HA morphology provides no surface evidence for an underlying word-final /h/ on noun stems ‹consensus›.

A separate h-related phenomenon in Hokkaido — the prothetic h- of the Bihoro dialect (haspa 'deaf' : aspa; ham 'nail' : am) — has been analysed as an independent Bihoro innovation driven by disambiguation of initial-stressed forms, not as an archaism or an echo of any coda laryngeal (Alonso de la Fuente 2022: §2.2.5). It is discussed in the broader consonant inventory in Chapter 16 (The Consonant Inventory and Its Phonetic Realization).

157.2 Debuccalization in Sakhalin and the affiliative evidence

In Sakhalin Ainu a systematic coda-weakening rule applies: stop and fricative codas debuccalize to h (p, t, k → h), nasal m merges with n, and coda r is followed by a vowel that opens the syllable Tangiku (2022: §2.3); Alonso de la Fuente (2022: §1). A secondary rule then shifts the resulting coda h to s after a front vowel: HA sik → *sih → sis (‹SA›RA). Community members nukopoli and aldehyde8908 independently documented these alternations in the aynu-corpora Discord — kor → koh (nukopoli, 2024) and the sik / sis split (aldehyde8908, 2025) aynu-corpora Discord (2023–2026) — both fully corroborated by Tangiku (2022: §2.3) ‹corpus-confirmed›.

The mat / mah / mac-ihi triplet is the strongest case for internal reconstruction via the affiliative. The ‹SA›RA surface form mah would predict an affiliative *mah-ihi, but the attested form is mac-ihi, with an affricate c absent from the citation form of both dialects Alonso de la Fuente (2022: §1); Tangiku (2022: §2.3, §5). The affricate arises from the regular t → c alternation before front i that applies at the morpheme boundary when the ‹SA› uniform affiliative suffix -ihi is appended: the underlying *t surfaces as c before the connecting -i-. The affiliative therefore functions as a transparent window onto the base consonant that debuccalization hides on the citation form. In HA, where the affiliative connecting vowel matches the base's own final vowel (a for mat), the same alternation does not arise: HA mat-aha keeps t.

The practical consequence — noted by community member nukopoli (2023, aynu-corpora Discord (2023–2026)) ‹corpus-confirmed› — is that ‹SA› concept-form etymology is often unrecoverable from the citation form alone: a surface -h coda could descend from any of p, t, k, or (via the secondary rule) any front-vowel-final stem whose coda lenited to h then shifted to s. The HA cognate or the ‹SA› affiliative — often both — supply the disambiguation.

The phonemic status of ‹SA› coda h is itself debated. Alonso de la Fuente Alonso de la Fuente (2022: §1) transcribes it as a velar fricative /x/, reflecting its realization as [ax], [iɕ], [uf] conditioned by the preceding vowel. Tangiku Tangiku (2022: §2.3) and community member thegodofneet (2025, aynu-corpora Discord (2023–2026)) treat it as a glottal or pharyngeal breath [h], citing the Hattori (1964) dialect dictionary's description that only breath, not modal voicing, exits, and raising the further question of whether the coda segment and the word-initial onset h are the same phoneme ‹contested›. The two transcriptions have no consequence for the diachronic correspondences laid out here, since both agree on the underlying source consonants; but they diverge in the phonological analysis of contemporary ‹SA› Ainu and its segmental inventory. The broader consonant inventory discussion is in Chapter 16 (The Consonant Inventory and Its Phonetic Realization).

‹SA›-internal variation further limits how far back the debuccalization can be projected. The Taraika (east coast) and Tarantomari dialects preserve p / t / k in coda position where the main ‹SA› dialects have h, and Esutoru distinguishes -is from -ih Tangiku (2022: §2.3). Debuccalization thus developed within the Sakhalin branch rather than in the common ancestor, and the HA forms represent the conservative retention of whatever coda consonants Proto-Ainu had.

157.3 The Sakhalin vowel-length reflex and proto-Ainu prosody

Alongside the consonant correspondences, Hokkaido and Sakhalin diverge in their prosodic systems. Sakhalin Ainu carries phonemic vowel length in open stem syllables (a aa e ee i ii o oo u uu); Hokkaido Ainu has a pitch accent and no contrastive length Tangiku (2022: §2.1); Alonso de la Fuente (2022: §2.3). Hattori (1967, via Alonso de la Fuente (2022: §2.3); Tangiku (2022: §2.1)) first stated the systematic correspondence: a ‹SA› long vowel in an open stem syllable aligns with a HA high-pitch syllable in the same position. A closed-syllable constraint qualifies the generalization: where the stem syllable carries a coda consonant (retained in HA, debuccalized to h in ‹SA›), the ‹SA› vowel is short while HA retains the high pitch on that syllable Alonso de la Fuente (2022: §2.3).

The súpki / suhki pair illustrates the closed-syllable constraint directly. The compound *sup+ki 'reed' keeps a closed first syllable regardless of dialect: in HA, p closes the syllable and the pitch falls on súp-; in ‹SA›, p → h closes the syllable differently but the syllable remains closed, so the vowel is short. The resulting ‹SA› form suhki shows a short vowel in the first syllable and a coda h, while HA súpki shows high pitch and the original coda p.

The diachronic interpretation of the length–pitch correspondence is the most contested point in proto-Ainu suprasegmental reconstruction. Three positions are current in the literature, surveyed comprehensively in Alonso de la Fuente (2022: §2.3) and addressed in detail in Itabashi (2001):

• (a) Length-first (Hattori 1967; De Boer 2010, both via Alonso de la Fuente (2022: §2.3)): Proto-Ainu had contrastive vowel length and no pitch accent; Hokkaido acquired its pitch system through prolonged contact with Japanese, while Sakhalin preserved the older length contrast.
• (b) Both length and pitch (Vovin 1993, via Alonso de la Fuente (2022: §2.3); Itabashi (2001)): Proto-Ainu carried both phonemic length and pitch simultaneously; the Sakhalin branch lost pitch under Amur–Sakhalin contact pressure, retaining length; Hokkaido retained pitch and neutralised length.
• (c) Pitch-first (Satō 2015–2016, via Alonso de la Fuente (2022: §2.3)): Proto-Ainu had pitch accent only; ‹SA› length is a phonetically conditioned phenomenon restricted to open syllables and represents a Sakhalin-internal development — possibly compensatory lengthening following coda weakening — rather than an inherited contrast.

Alonso de la Fuente's own treatment is internally divided: §2.3 endorses view (c) as "most economical", while §2.4 states that "Hattori was right when he suggested Proto-Ainu had only vowel length, and no pitch accent; pitch accent may be the result of language contact after Japanese influence" — a restatement of view (a) Alonso de la Fuente (2022: §2.3, §2.4) ‹contested›. The two sections advocate opposite positions without reconciliation; neither can be taken as Alonso's settled conclusion. Itabashi Itabashi (2001) provides the dedicated historical treatment of the SA-length ↔ HA-pitch diachronic relation; its arguments should be weighed before any definitive account of PA prosody is committed to, and the fuller treatment belongs in Chapter 24 (The Prosodic Unit and the Accent-versus-Tone Analysis).

A trochaic-to-iambic shift hypothesis (Satō 2015–2016; Shiraishi 2017, both via Alonso de la Fuente (2022: §2.3)) adds a further layer: if Proto-Ainu or Common Ainu was trochaic and the pitch accent spread rightward, the resulting iambic system would leave certain archaisms — ‹SA› noociw : HA nocíw, ‹SA› niisew : HA niséw (length ↔ low pitch in HA) — as the recoverable fingerprints of the older system. Community member nukopoli (2023, aynu-corpora Discord (2023–2026)) noted the Hattori/Satō dispute directly, identifying the competing claims from within the community's reading of the literature ‹contested›. Antitwilight (2024, aynu-corpora Discord (2023–2026)) takes a position against Satō's posited `*H` segment and prefers Janhunen's vowel-dropping account (see also §1 on the affiliative), arguing that introducing an unattested consonant for the purpose of accounting for exceptional accent is less economical than assuming base-vowel loss ‹contested›.

The pitch-accent system of Hokkaido, including the placement rules and the dialect-internal variation among standard-accent, irregular, and unaccented dialects, is treated in full in Chapter 23 (The Pitch-Accent Placement Rule and the Accented/Accentless Dialect Split) and Chapter 24 (The Prosodic Unit and the Accent-versus-Tone Analysis). The proto-Ainu segmental reconstruction on which the coda correspondences above rest is in Chapter 156 (Proto-Ainu Reconstruction: Segments and Accent Classes), and the broader Hokkaido–Sakhalin dialect picture in Chapter 159 (Hokkaido Dialect Classification and Dialectometry) and Chapter 162 (Sakhalin and Kuril Ainu: The External Comparison).

157.4 Orthographic and typological implications

For Hokkaido, the orthographic consequences of the debuccalization asymmetry are minimal: HA retains original coda consonants on the surface, and the citation forms write exactly what is heard — tek, sik, cup, mat. The apparent h that appears medially in affiliative forms is always the onset of the suffix -hi and is morphologically transparent.

For Sakhalin orthography the picture is more consequential. Writing the citation form as mah, cuh, teh follows the phonetic surface but conceals the base consonant that the affiliative restores and the Hokkaido cognate corroborates. Thegodofneet (2025, aynu-corpora Discord (2023–2026)) proposed writing the underlying p/t/k values directly and reading them as h by convention — a strategy with analogues in French and Korean orthographic traditions, which preserve historically silent consonants as relics of earlier forms. Aldehyde8908 (2025, aynu-corpora Discord (2023–2026)) countered that some ‹SA› subdialects genuinely surface with p/t/k (Taraika, Tarantomari), so writing the surface form is the accurate representation of those dialects, while the uniform h convention would misrepresent them ‹contested›. The dispute reflects the genuine ‹SA›-internal variation noted above and has no consensus resolution in the sources consulted. Ijäs Ijäs (2023) addresses the SA–HA affiliative alternation in pedagogical terms, making the base-consonant restoration accessible to learners encountering ‹SA› citation forms for the first time.

Typologically, the ‹SA› debuccalization belongs to a widespread class of coda-weakening processes in which place distinctions are neutralised toward a laryngeal default. The resulting system — a single coda h for historically distinct p, t, k — would be irrecoverable from ‹SA› alone; the Hokkaido cognates and the ‹SA› affiliative together supply the two independent verification routes that Alonso de la Fuente Alonso de la Fuente (2022: §1) calls an illustration of internal reconstruction for an isolate. The genetic position of Ainu — an isolate for which no comparative method in the strict sense applies — makes this type of internal, morphologically mediated evidence especially load-bearing; for the macro-comparative context see Chapter 5 (The Genetic Position of Ainu: Isolate Status and Macro-Comparison Controversies).