Chapter 158Internal Reconstruction and Grammaticalization Pathways

158.1 The scope and tools of internal reconstruction

Ainu is a language isolate (Chapter 5 (The Genetic Position of Ainu: Isolate Status and Macro-Comparison Controversies)): the comparative method, which requires a set of related languages with regular correspondences, is unavailable. Three tools remain for recovering Ainu prehistory: internal reconstruction, which exploits alternations, paradigm irregularities, and dialect differences within Ainu itself; contact-induced-change analysis, drawing on the documented Japanese, Nivkh, and Tungusic contact layers; and typological inference, useful as a heuristic but carrying no independent probatory value Alonso de la Fuente (2022: §1). Internal reconstruction is by far the most powerful of the three, accounting for essentially all dialectal differences that can be connected to recoverable earlier forms. The time depth it can reach is shallow — roughly 1,000 years — bounded by the 13th-century Great Migration that separated Common Ainu into its Hokkaido and Sakhalin branches Alonso de la Fuente (2022: §1, §4).

The reconstructible ancestor of the modern Hokkaido varieties is termed Proto-Ainu (PA); the immediate pre-Migration common ancestor is Common Ainu (CA); what internal reconstruction reaches beyond PA through recursive application is termed Pre-Proto-Ainu (Pre-PA). The segmental inventory and the contested suprasegmental reconstruction (length versus pitch) are treated in Chapter 156 (Proto-Ainu Reconstruction: Segments and Accent Classes); the SA/KU contrast arising from the 13th-century split is treated in Chapter 162 (Sakhalin and Kuril Ainu: The External Comparison). This chapter focuses on grammatical morphology: what the alternation patterns within Hokkaido Ainu, together with the SA contrast, allow us to infer about the diachronic sources of bound morphemes and the grammaticalization clines that produced them.

158.2 Grammaticalization pathways — an inventory

The etymology of Hokkaido Ainu functional morphology is, by comparative standards, unusually transparent. As Alonso observes, "when it comes to functional words and TAM formatives, the etymology is still transparent" Alonso de la Fuente (2022: §3.1). The major pathways fall into four structural types. The theoretical framework for grammaticalization clines, layering, and renewal drawn on here is (Hopper & Traugott 2003).

158.3 The affiliative suffix -(V)hV: two accounts

The affiliative (possessed) suffix surfaces in modern Hokkaido dialects as -(V)hV, where both vowels vary with the base in complex ways. Two analyses of this distribution have been proposed, and the choice between them has direct consequences for the PA vowel inventory.

Vovin (1993, via Alonso de la Fuente (2022: §2.1)) attributed the distributional complexity to a 17-vowel PA system: the suffix vowel was selected through vowel harmony with the final vowel of the base. The machinery was extensive — twelve short and six long vowels — but all three major reviewers of the 1993 monograph found it unmotivated by the available evidence (Sidwell 1996; Abondolo 1997; Witczak 1999, all via Alonso de la Fuente (2022: §1)).

Janhunen (2020, via Alonso de la Fuente (2022: §2.1)) offers a more economical account via internal reconstruction. Pre-PA nouns were disyllabic: stem *CVC.V#. When the second syllable reduced and was eventually lost, the lost vowel was reanalysed as an incipient affiliative marker. Subsequently, the common noun hi ‘thing, time, place’ fused with this stranded vowel: base vowel + hi → -(V)hV. The word tek ‘hand’ illustrates the pathway: PA *tek.e (disyllabic) → tek with loss of second syllable; lost e + hi → affiliative -ehe. Alonso calls this "a triumph of internal reconstruction" and adopts it as the preferred account, concluding that the PA vowel system contained only the modern five vowels Alonso de la Fuente (2022: §2.1).

Convergent evidence from the ‹SA› varieties supports the underlying hi analysis: in ‹SA›, the base consonant resurfaces in the affiliative form even when it is otherwise debuccalized — sik-ihi 'eye-POSS' rather than the expected *sis-ihi, which would follow from ‹SA›'s regular k → s/_i rule (see Chapter 157 (The Final-h Problem and Its Sakhalin Vowel-Length Reflex)). The affiliative suffix must therefore preserve access to the underlying base consonant that is otherwise hidden by final debuccalization Tangiku (2022: §2.3, §5). Community discussion records the same pattern: ‹SA› san + hi = sanihi, makan + hi = makanihi (nukopoli, aynu-corpora Discord, 2024-12-16) aynu-corpora Discord (2023–2026), consistent with the hi-fusion account ‹corpus-confirmed›. The morphophonological details of the modern affiliative paradigm are treated in Chapter 45 (Morphophonology of the Affiliative Suffix and Echo/Copy Vowels).

The debate between the two accounts is ultimately ‹contested› rather than resolved: the five-vowel reconstruction has the stronger case given the reviewers' response to Vovin's system, but the internal reconstruction of a lost disyllabic base rests on inferences that cannot be confirmed without a comparative method.

158.4 The evidential nominals: sensory nouns to clause-final markers

The four formal nouns that build the Hokkaido evidential system — ruwe, hawe, siri, humi — are the possessed forms of four sensory nouns, each of which has grammaticalized into a clause-final evidential predication marker. The grammaticalization path in each case runs: sensory noun → possessed formal noun → clause-final evidential marker Alonso de la Fuente (2022: §3.1); Kindaichi & Chiri (1936); Hirosawa (2026: 65).

The grammaticalization is not complete: the lexical source meaning remains recoverable, and the formal nouns retain nominal properties (they take the full affiliative suffix: ruwehe, humihi) while losing others (they do not accept demonstrative or attributive modification in their evidential function) Dal Corso (2018: 24–25). This partial grammaticalization is reflected in synchronic dialect layering: Tokachi and Ishikari use the bare concept forms ru / sir / haw / hum where Saru and Chitose use the possessed forms, and Horobetsu has extended the possessed forms into conjunctive and adverbial-clause roles Takahashi (2018: 107–112); Takahashi (2013: 129–131). Chiri (1936, via Hirosawa (2026: 65)) already noted the cline from 'there is a trace of X' to 'there is a fact of X' for ruwe.

The dialect distribution encodes the grammaticalization timeline as a spatial stratigraphy: bare concept forms (Tokachi) represent the least grammaticalized stage; possessed forms (Saru) the next; conjunction/adverbial function (Horobetsu) the most advanced. All three stages coexist synchronically across the Hokkaido dialect space — a textbook case of layering (Hopper & Traugott 2003) ‹corpus-confirmed› (corpus token counts above from string search on the Hokkaido 167,890-sentence corpus). The full semantics and synchronic distribution of each form belong to the evidential chapters: Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential), Chapter 122 (siri ne — the Direct-Perception / Situational Evidential), Chapter 123 (hawe ne — the Reportative / Hearsay Evidential), Chapter 124 (humi ne — the Non-Visual Sensory Evidential); the overall schema is in Chapter 120 (The Nominalization-plus-Copula Evidential Schema); the formal defective noun class is defined in Chapter 35 (The Formal / Defective Noun Set (ruwe, hawe, siri, humi, hi, pe, kur)).

The copula ne ‘be, copula’ that closes the evidential predication also has a grammaticalization history: Alonso identifies it as an anaphoric pronoun that developed a copular function Alonso de la Fuente (2022: §3.1). This shift is treated in Chapter 96 (The Copula ne and Predicate-Nominal Clauses) ‹speculative›.

158.5 Free verbs grammaticalized as bound suffixes

The most transparent set of grammaticalization events in Ainu involves free verbs reanalysed as derivational suffixes on nouns and verbs. Alonso documents six such suffixes with unambiguous lexical sources (Alonso 2022 §3.1, Table 16; drawing on Dettmer 1989): Alonso de la Fuente (2022)

The same free verbs continue in full predication and in compound verb constructions, maintaining a synchronic continuum between free and bound use rather than a categorical split. The morpheme database (mdb.aynu.org) corroborates the overlap: honkor = hon 'stomach' + kor 'have' (conf. 0.85), and the compound verb kor an (progressive) shows the same root in its free form Alonso de la Fuente (2022: §3.1). Community discussion adds evidence for ongoing productivity of the -kar suffix in action-verbalizer derivations (nimawekar, osorokar, sikayekar, yaykara; nukopoli, aynu-corpora Discord, 2024-11-12) aynu-corpora Discord (2023–2026) ‹corpus-confirmed›.

Related pathways in the functional vocabulary follow the same template: ani ‘hold’ > instrumental/comitative particle 'with, by'; kusu ‘reason, cause’ > causal conjunction 'because'; hon + tom 'belly + middle' > locative 'in the middle of' Alonso de la Fuente (2022: §3.1). These are treated in the postposition and clause-linkage chapters; they confirm that the free-to-bound pathway is an ongoing mechanism, not a closed Pre-PA residue.

158.6 Aspectual an and the posture-verb pathway

The existential verb an ‘be, exist (SG)’ derives from a posture verb meaning 'sit down (SG)'; the suppletive plural rok ‘sit down (PL)’ corroborates the source, since the SG/PL suppletion retains the posture semantics Alonso de la Fuente (2022: §3.1). From existential predication, an extends to durative and progressive aspect in the constructions kor an and wa an (see Chapter 110 (The Existential-Aspect System: kor an Progressive and wa an Resultative-Perfect)).

A contested question is whether the personal suffix =an (4th-person intransitive subject; also the inchoative -an in the table above) derives from the same a=~an root. Nakagawa and Itabashi treat an= as the historically older form from which the reduced a= prefix descends Nakagawa (2024: §person); Itabashi (2001). This derivation is disputed: nukopoli (aynu-corpora Discord, 2026-04-29) notes that the high productivity of a= as an independent prefix weighs against deriving it from the longer an= form, and proposes an independent origin from the existential/aspectual root aynu-corpora Discord (2023–2026) ‹contested›. The synchronic person paradigm, where the question bears on its distribution, is treated in Chapter 67 (The Indefinite/Fourth Person: Forms, Reference, and Agent-Defocusing).

158.7 The nominalizers -p/-pe and the cep case

The nominalizer suffix -p/-pe derives participant nouns and thing-nouns from predicates. Its lexical source is the free noun pe ‘thing, object’, which shifted from independent nominal to clause-final derivational suffix by the same pathway as the evidential formal nouns above Alonso de la Fuente (2022: §3.2). The synchronic distribution of -p/-pe and the co-occurring event/place nominalizer -i/-hi (< hi 'thing, place') is treated in Chapter 101 (The Nominalizers -p/-pe and -i/-hi (Participant, Event, Place, Fact)).

The word cep ‘fish’ provides the most completely reconstructible example of this pathway and of internal reconstruction in Ainu generally. Its morphological analysis is:

The three-way comparison — Saru cep, Horobetsu ciʔep, SA ceh — is Alonso's flagship illustration of the CA→dialect split Alonso de la Fuente (2022: §4, Table 18). Horobetsu retains the glottal stop *ʔ from CA */ciʔep/, Saru has lost it, and the SA debuccalization (final p > h) is the regular ‹SA› reflex. That three independent phonological processes left distinct residues corroborates the reconstruction in a way that a single-dialect analysis could not.

A community refinement (nukopoli, aynu-corpora Discord, 2023-12-21) proposes that the ci= prefix in this reconstruction reflects a middle/anticausative rather than resultative function — the same ci= that produces intransitive/middle stems in modern Ainu (see Chapter 87 (Middle, Anticausative, and the (Non-)Passive Question)) — with the semantic reading 'the thing that gets-eaten (by itself)' rather than 'the thing that all eat' aynu-corpora Discord (2023–2026) ‹speculative›. Both readings yield 'fish'; the middle-voice alternative is of interest because ci= as a middle prefix is otherwise well-attested in the modern grammar, whereas the resultative/inclusive reading presupposes a PA function of ci= that is less clearly distinct in the modern system.

158.8 The applicative prefixes — a disputed diachrony

The applicative prefixes e-, o-, and ko- introduce oblique arguments as core verbal arguments. They are treated functionally in Chapter 78 (The Applicative System: Overview and the Preverb-vs-Applicative Debate) and individually in Chapter 79 (The e- Applicative), Chapter 80 (The ko- Applicative), Chapter 81 (The o- Applicative); Bugaeva (2006) and Bugaeva (2010) remain the primary analytical references. Their diachronic origin Alonso assigns to the grammaticalization of locative/relational nouns into verb prefixes, a widespread pathway typologically Alonso de la Fuente (2022: §3.1); (Peterson 2007). The correspondence is transparent in type — relational nouns shifting into applicative prefixes — but the specific source nouns for e- and o- have not been individually identified with confidence, making the claim ‹speculative› for those two prefixes.

A more structurally ambitious proposal appears in the community discussion. Antitwilight (aynu-corpora Discord, 2026-06-04) argues that the applicative prefixes were originally pre-nominal, not pre-verbal: certain compound nouns and fixed adverbs of the form *e-ka-nok, *ka-o-suy would then be nominal collocations later reanalysed as verb-prefix constructions when the prefix migrated leftward to the verbal host aynu-corpora Discord (2023–2026) ‹speculative›. The hypothesis draws on the well-attested presence of applicative-prefix forms inside nominal compounds in the modern language (see Chapter 90 (Oblique Incorporation and the Syntactic-vs-Lexicalist Debate)) but has no direct support in the Handbook sources read for this chapter. It is noted here as a community-generated lead that warrants investigation of the old compound-noun record before further evaluation.

158.9 Layering, renewal, and what the dialect record shows

Internal reconstruction in Ainu proceeds primarily through the synchronic dialect record, where different dialects have advanced different distances along the same grammaticalization clines, leaving stages of a single pathway distributed across the Hokkaido dialect space. Two contrasting patterns emerge: layering (coexistence of older and newer forms with overlapping function) and renewal (the newer form replaces, rather than supplements, the older).

The evidential nominals are the clearest layering case. The Tokachi bare concept forms ru / sir / haw / hum, the Saru possessed forms ruwe / siri / hawe / humi, and the Horobetsu conjunctive extensions of the same possessed forms coexist synchronically as three stages of one cline. The forms at each stage are not synonymous across all contexts, and the distributional differences between stages reveal the grammaticalization constraints: the possessed form (ruwe) has wider contextual distribution than the bare concept form (ru), and the bare form retains combinatorial restrictions that the possessed form has shed Takahashi (2013: 129–131); Takahashi (2018: 107–112).

The plural marker in the Saru/Ishikari region illustrates renewal rather than layering. The older plural form ci (surviving in Horobetsu, Shizunai, Shiranuka) was replaced in the Saru/Ishikari core by an innovated pa, which spread and displaced the older form in those dialects rather than accumulating beside it Nakagawa & Fukazawa (2022: §4.2). The corresponding ‹SA› form (a)hci is likely the archaic stage, placing ci as intermediate Tangiku (2022: §7.5).

The free-verb-to-bound-suffix pathways (§ above) offer the purest layering: the root kor ‘have, hold’ functions simultaneously as a free transitive verb, as a compound element in the progressive kor an, and as the bound possessive-verbalizer suffix -kor. No Hokkaido dialect has lost any of these uses; the layers accumulated without replacement. Reconstruction of a relative chronology for these layers — which function is oldest, and when the bound use separated from the free — remains out of reach given the approximately 1,000-year horizon of reliable internal reconstruction and the absence of any direct chronological evidence internal to Hokkaido Ainu Alonso de la Fuente (2022: §1, §4).

158.10 Classifier relics — a cautionary assessment

Alonso (2022 §3.2), following Austerlitz, Kindaichi, and Chiri (1936), considers whether certain final consonants of body-part and nature nouns are relics of an earlier classifier or noun-class system in Pre-PA. The proposed class markers include: *k (body/round: tek 'hand', sik 'eye', nok 'egg'); *t (long/rolling: pet 'river', not 'chin', at 'cord'); *m (long/small: kem 'needle', yam 'chestnut'); and *r (elongated: ker 'footwear', sir 'rock') Alonso de la Fuente (2022: §3.2, Table 17–18).

Alonso himself rejects the hypothesis on diagnostic grounds: classifier systems typically require that the class marker occur independently as a free morpheme or in some derivational use beyond noun-final position. Ainu's proposed class-final consonants fail this test — they are invariably noun-final and never function as independent words or derivational triggers. He concludes that the pattern "illustrates the dangers of wishing to see too much in the data" (Alonso 2022 §3.2). The word-final consonant inventory of Hokkaido Ainu is phonologically constrained (see Chapter 16 (The Consonant Inventory and Its Phonetic Realization) and Chapter 160 (Hokkaido Dialect Microvariation: Phonology and Morphosyntax)) but does not cross-classify nouns in any grammatically active synchronic sense. The productive numeral classifiers of modern Hokkaido Ainu — -p~-pe (things) and -n~-iw (people, from aynu ‘human being’) — derive from the common nouns pe and aynu through the noun-to-nominalizer pathway already attested in the -p/-pe nominalizer above; they are treated in Chapter 55 (Numeral Classifiers, Ordinals, and Quantifier Syntax) ‹speculative›.