Chapter 156Proto-Ainu Reconstruction: Segments and Accent Classes

156.1 Method and limits of Proto-Ainu reconstruction

Ainu is a language isolate: all attempts to establish a genetic relationship with any other language or family have failed Janhunen (2022: §2). For an isolate, the comparative method — which enables deep-time reconstruction by aligning cognate sets across related branches — cannot apply when only a single lineage survives. Three instruments remain available Alonso de la Fuente (2022: §1): internal reconstruction, which Alonso de la Fuente calls "by far the most powerful" tool; contact-induced-change analysis, chiefly against Japanese and, on Sakhalin, against Nivkh; and typological reasoning, which carries no independent probatory weight and must be handled with caution. The reconstruction surveyed here is accordingly shallow in time depth: internal evidence licenses proposals reaching roughly a thousand years into the past, and the one firm external anchor is the thirteenth-century Great Migration that separated the Hokkaido, Sakhalin, and Kuril varieties Alonso de la Fuente (2022: §1, §4). Deeper pre-history — the Pre-Proto-Ainu stage — is reachable only "ad nauseam by internal reconstruction" and cannot be independently validated Alonso de la Fuente (2022: §4).

The two primary sources for this chapter are Vovin (1993) — the monograph-length reconstruction of Proto-Ainu, abbreviated RPA — and the critical revision by Alonso de la Fuente (2022) (abbreviated "New"). The RPA monograph is not held in the project repository; all RPA reconstructions and associated table references are taken from Alonso de la Fuente's reproductions, and citations of Vovin 1993 therefore carry the reported-evidence status already assigned to that entry. The genetic-isolate question and the failed macro-comparisons are treated in Chapter 5 (The Genetic Position of Ainu: Isolate Status and Macro-Comparison Controversies); grammaticalization pathways for individual morphemes (the affiliative, the applicatives, the evidential nominals) are in Chapter 158 (Internal Reconstruction and Grammaticalization Pathways).

156.2 Periodization: Pre-Proto-Ainu, Proto-Ainu, Common Ainu

Alonso de la Fuente distinguishes three stages Alonso de la Fuente (2022: §4, Tables 18–19), drawing an analogy with the Slavic case (Pre-PA ≈ PIE, PA ≈ Proto-Slavic, Common Ainu ≈ Common Slavic), while noting that only Common Ainu and later can claim anything resembling a running comparative method:

The CA innovations listed by Alonso de la Fuente (Table 19) are: merger of *l and *ɾ to *ɾ; merger of palatalized and plain sibilants (*sʸ, *s > *s); shift of the sibilant before *i; aspiration of *h; and the collapse of *tʸ, *pʸ to *c. These CA-level changes account for the shared features of all attested Hokkaido lects relative to the older reconstructed stage. The Hokkaido-internal dialect archaeology built on this foundation is treated in Chapter 159 (Hokkaido Dialect Classification and Dialectometry).

156.3 The Proto-Ainu vowel system

Every documented Ainu lect — Hokkaido, Sakhalin, Kuril — has exactly five vowel qualities, a e i o u. Vovin posited a seventeen-vowel PA system (twelve short plus five long: i ü e ö E a ï ë A O o u plus long counterparts) to regularize what he called the "chaotic" distribution of the affiliative suffix -V: in Vovin's account, the suffix vowel is selected by the last vowel of the base according to a vowel-harmony-like table (RPA; reproduced in Alonso 2022, Table 2). Alonso rejects the seventeen-vowel system on phonetic and historical grounds: the proposed extra vowel qualities leave no trace in the oldest Nara-period loanword layer, and the distributional alternations attributed to them are reassignable to the five basic vowels without residue Alonso de la Fuente (2022: §2.1) ‹contested›.

An alternative account of the affiliative -(V)hV shape comes from Janhunen's vowel-dropping hypothesis (Janhunen 2020, reported via Alonso 2022 §2.1): the marker was originally an undefined final consonant *-x; the synchronic shape reflects the lost second vowel of a once-disyllabic base — *CVC.Vx > CVC, but *CVC.Vx > CVC-Vx when the marker follows — and the sequence later fused with the noun hi 'time, place, thing' to yield the current possessive shape. Alonso calls this "a triumph of internal reconstruction" Alonso de la Fuente (2022: §2.1). Community discussion on the aynu-corpora Discord has independently noted this as the more parsimonious account, opposing the introduction of a phonemically unmotivated *H aynu-corpora Discord (2023–2026) (antitwilight, 2024-12-23 ‹contested›). The morphophonology of the affiliative suffix — its echo vowels, allomorphic conditioning, and distributional constraints — is treated in Chapter 45 (Morphophonology of the Affiliative Suffix and Echo/Copy Vowels).

The reconstructed PA vowel system is the same five-quality inventory that all modern lects share Alonso de la Fuente (2022: §2.1) ‹consensus›. Whether PA also had distinctive vowel length is the central suprasegmental dispute, addressed in §6 below.

156.4 The Proto-Ainu consonant system: Vovin's reconstruction

For the modern Hokkaido consonant system as a baseline, see Chapter 16 (The Consonant Inventory and Its Phonetic Realization). Vovin proposed the following PA inventory (RPA; reproduced in Alonso de la Fuente (2022: Table 3)):

The half-voicing of *p *t *k is motivated by alternations in pre-scientific wordlists — e.g. Dixon (1883) ⟨jabush⟩ for capus 'lip' — which suggest that the stops were phonetically intermediate between voiced and voiceless at an earlier period (via Alonso de la Fuente (2022: §2.2)). Contemporary reviews of RPA raised two structural objections: the absence of *w, a phoneme present in every attested Ainu lect (Sidwell 1996, via Alonso 2022 §2.2.3); and the background assumption of an Austroasiatic affiliation that distorted the consonant correspondences (Sidwell 1996, Abondolo 1997, Witczak 1999; all via Alonso de la Fuente (2022: §1)).

156.5 Revisions to the consonant correspondences: Alonso 2022

Alonso de la Fuente's revision addresses eleven correspondence classes Alonso de la Fuente (2022: §2.2, Table 5). The table below lists each class in RPA notation, its attested reflexes across the lect sample, and the revised reconstruction (all second-hand via Alonso 2022, citing Vovin 1993):

The most consequential revisions are summarised in the subsections below.

The two revised PA consonant inventories proposed by Alonso de la Fuente are (Table 14; Alonso de la Fuente (2022: Table 14)):

Items in {curly braces} are motivated by typological reasoning only; items in [square brackets] are plausibly allophonic. The choice between the two scenarios depends on resolving the *g/*hd/*s tangle in §4.3 and remains open ‹contested›.

156.6 Dialect correspondence sets

The cognate sets below illustrate the principal PA reconstructions against the lect sample (Alonso 2022, Tables 1–6, 14, 18; via Alonso de la Fuente (2022)). Lect abbreviations follow Alonso 2022 §0: Y = Yakumo, HO = Horobetsu, SA = Saru (Alonso's convention), B = Bihoro; ‹SA› RA = Sakhalin Raiciska; ‹KU› KD/KK/KV = Kuril sources (Dybowski/Klaproth/Voznesenskij), NA = ‹SA› Nairo:

The ‹SA› final-consonant reflexes — sik → sis, cup → cuh — are instances of debuccalization: stop codas weaken to h, then shift further to s after front vowels Tangiku (2022: §2.3). The underlying stop resurfaces in the ‹SA› affiliative/possessive form (sik-ihi, not *sis-ihi), confirming the reconstructed coda Tangiku (2022: §2.3, §5). The full diachrony of this process is treated in Chapter 157 (The Final-h Problem and Its Sakhalin Vowel-Length Reflex) and Chapter 162 (Sakhalin and Kuril Ainu: The External Comparison).

The internal reconstruction of cep 'fish' is Alonso de la Fuente's illustration of what internal reconstruction can recover for an isolate Alonso de la Fuente (2022: Table 18):

156.7 Proto-Ainu suprasegmentals: the length–pitch dispute

The most contested area of PA reconstruction concerns the suprasegmental system. Sakhalin and Kuril lects have distinctive vowel length (a aa e ee i ii o oo u uu), confined to open stressed syllables; Hokkaido Ainu has pitch accent and no distinctive length. Hattori (1967, via Alonso 2022 §2.3) first stated the systematic correspondence: ‹SA›/‹KU› long vowel ↔ HA high pitch — e.g. HA húre 'red' ↔ ‹SA› huure; HA túki 'cup' ↔ ‹SA› tuuki; HA sísam 'Japanese person' ↔ ‹SA› siisam Alonso de la Fuente (2022: §2.3); Tangiku (2022: §2.1). Three reconstructions of this correspondence have been proposed:

Alonso de la Fuente's chapter is internally inconsistent on this point: §2.3 sides with Satō's pitch-first view (c), while §2.4 endorses Hattori's length-first view (a), and the chapter identifies Vovin's mora-counting assumption in RPA as "the source of all problems" Alonso de la Fuente (2022: §2.3, §2.4). This inconsistency cannot be resolved from the sources consulted here; the reconstruction of PA prosody is genuinely unsettled ‹contested›.

A further complication is the trochaic→iambic shift hypothesis. Satō (2015/2016) and Shiraishi (2017) — both via Alonso de la Fuente (2022: §2.3) — propose that PA/Common Ainu was trochaic and that accent spread rightward to produce the historical iambic pattern visible in most Hokkaido dialects. ‹SA› archaisms supply the key diagnostic forms: ‹SA› Nairo noociw : HA nocíw 'star', and niisew : HA niséw — where ‹SA› length corresponds to HA low pitch rather than high, reversing the expected correspondence and motivating the rightward-shift analysis. The aynu-corpora Discord has noted this debate as an unresolved live question, recognising the Hattori/Satō opposition accurately aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11 ‹contested›). Itabashi's dedicated study of the ‹SA› vowel-length ↔ HA pitch-accent historical relation Itabashi (2001) has not yet been incorporated in detail here; the diachrony is captured via Hattori (1967) and Itabashi (2004) as reported in Alonso 2022 and Tangiku 2022. The relationship between the accent classes and the synchronic accent-placement rules is treated in Chapter 23 (The Pitch-Accent Placement Rule and the Accented/Accentless Dialect Split) and the prosodic-unit debate in Chapter 24 (The Prosodic Unit and the Accent-versus-Tone Analysis).

156.8 Open questions and honest gaps

Several questions bearing on PA segments cannot be settled from the sources consulted. The choice between Alonso's two consonant scenarios (§4.3) turns on evidence that neither phonological typology nor the attested reflex distribution yet settles decisively. The palatalized-series proposal (§4.5) rests entirely on typological reasoning and is not required by the philological data; Shiratori 2026 adds some support, but the hypothesis remains speculative. The RPA monograph (Vovin 1993) is not in the project repository, so every page reference to RPA is second-hand via Alonso 2022, and load-bearing RPA citations should be verified against the original before finalization Vovin (1993). The Hattori (1967) length-pitch correspondence paper, Hattori and Chiri (1960), Asai (1974), and the Satō (2015/2016) accent papers are all cited second-hand via the Handbook chapters read.

The three-stage model (§2) is heuristic. Pre-Proto-Ainu is a theoretical construct reached only by sustained internal inference, without external validation; the thirteenth-century Great Migration provides the one firm absolute anchor. The comparison with PIE/Slavic stages, while structurally illuminating, cannot import the evidential standards that the comparative method sustains in well-attested families — a limitation intrinsic to isolate reconstruction Alonso de la Fuente (2022: §1, §4); Janhunen (2022: §2).