Chapter 26Accent in Compounds and under Affixation and Cliticization
How prosodic-word formation reorganizes accent in compounds and when affixes and clitics attach.
26.1 The prosodic word and accent reapplication
In all standard-accent dialects of Hokkaido Ainu, high pitch is assigned to σ1 when σ1 is closed (has a coda), and to σ2 otherwise — with a memorized set of lexical exceptions (see Chapter 23 (The Pitch-Accent Placement Rule and the Accented/Accentless Dialect Split)). These rules evaluate the prosodic word as their domain: a prefix or compounding element that fuses with a base creates a new prosodic word, and the placement rules are reapplied from the new left edge Nakagawa (2024: 52–54); Shiraishi (2022: §5.1–5.2); Satō (2008: §2.6). Prefixation may therefore shift accent inward, and compounding suppresses all accent peaks after the first. Affixes divide accordingly: those that enter the prosodic word of their base shift its accent; those that remain outside it leave the base accent intact. The accent data thereby provide one of the clearest phonological diagnostics for the prefix/clitic boundary in the person-affix paradigm. The Yakumo irregular system, where 原則2 shifts one position further right, produces different surface outputs for the same derivations but applies the same prosodic-word logic; its characterization is given in Chapter 23 (The Pitch-Accent Placement Rule and the Accented/Accentless Dialect Split).
26.2 Compound nouns: first-element priority
When two or more free forms combine as a compound noun, the result carries a single accent peak. Nakagawa (2024: 54) states the general outcome as a first-element priority rule: only the front element's lexical accent survives; all syllables belonging to rear elements are low. For a polysyllabic front element, the accent it bears in isolation — as assigned by the placement rules to that element's own σ1 — is retained unchanged.
| front element | accent | compound | compound accent | gloss | source |
|---|---|---|---|---|---|
| kamuy | kamúy | kamuy+nomi | kamúynomi | 'pray to a deity' | Nakagawa (2024: 54) |
| horka | hórka | horka+reye+p | hórkareyep | 'crayfish' | Nakagawa (2024: 54) |
| cise | cisé | cise+kor+kur | cisékorkur | 'householder' | aynu-corpora Discord (2023–2026) (nukopoli 2024-12-09) |
| kotan | kotán | kotan+kor+kur | kotánkorkur | 'village chief' | aynu-corpora Discord (2023–2026) (nukopoli 2024-12-09) |
| ni (monosyllabic) | ní | ni+tek | nítek | 'branch' | Nakagawa (2024: 54) |
The accent positions follow from the placement rules applied to the front element in isolation. kamúy has open σ1 (ka) → 原則2 → accent on σ2 mú; that accent is the sole high syllable of kamúynomi. hórka has closed σ1 (hor) → 原則1 → accent on hór; that accent survives in hórkareyep. The monosyllabic ni carries the single available accent, giving nítek.
For cisékorkur and kotánkorkur, the community reports that both are single words with the front-element accent as their sole peak aynu-corpora Discord (2023–2026) (nukopoli 2024-12-09) ‹corpus-suggested›. The assignments follow from the principle: cise 'house' has open σ1 (ci) → 原則2 → accent on sé, which becomes the sole peak of cisékorkur 'householder'; kotan 'village' has open σ1 (ko) → 原則2 → accent on tán, retained in kotánkorkur 'village chief'.
The first-element priority is a strong default. Shiraishi (2022: §5.2) notes a set of lexicalized compound nouns that carry the iambic pattern even when the front element's isolated accent would predict otherwise: hasínaw, cepétu 'fish snout', nayétok 'stream source' all show accent on σ2 of the compound, as if the form were an underived disyllabic root. These represent lexicalized items requiring per-form verification ‹contested›. The prosodic-unit question (mora vs syllable) bearing on the weight analysis is addressed in Chapter 24 (The Prosodic Unit and the Accent-versus-Tone Analysis); attested minimal-pair contrasts in compound accent are listed in Chapter 25 (Lexical/Contrastive Accent and Minimal Pairs).
26.3 somo and somósomo as a prosodic diagnostic
The negative adverb somo ‘not’ illustrates the prosodic-word boundary transparently. As an independent word, somo has open σ1 (so) and thus accent on σ2 by 原則2: somó. This is its accent in phrases such as ka somó ki 'does not do even', where somo is a separate prosodic word. When doubled to form the intensified negation somosomo 'not at all, definitely not', the pair becomes a single prosodic word: only the front element's accent survives, giving somósomo with a single high syllable on mó aynu-corpora Discord (2023–2026) (nukopoli 2024-12-09) ‹corpus-suggested›. The contrast ka somó ki (two prosodic words) versus somósomo (one prosodic word) instantiates the same mechanism as the compound table above. The negation system and somo are treated in Chapter 126 (Standard Clausal Negation with somo).
26.4 Derivational prefixes
26.4.1 The reflexive prefix yay-
The reflexive prefix yay- ‘self, reflexive’ always opens the prosodic word as σ1. yay is a closed syllable (CVC), so 原則1 places the accent on σ1 — on yáy — regardless of the shape of the following stem Nakagawa (2024: 53). The published example is yáykoytak 'talk to oneself' (< yay-ko-itak). The community additionally reports yáyesiru 'concentrate (on oneself)' as a further illustration that yáy holds the accent whether the following syllable is consonant-initial or vowel-initial aynu-corpora Discord (2023–2026) (nukopoli 2024-11-10, attributed to Nakagawa (2024)). Lexicalized yay- derivatives may deviate and require per-form verification. The distribution of yay- and the co-reflexive si- in Chitose is examined in Satō (2007); the morphosyntax of both reflexive prefixes is set out in Chapter 78 (The Applicative System: Overview and the Preverb-vs-Applicative Debate).
26.4.2 Applicative prefixes as prosodic-word heads
The applicative prefixes e- ‘APPL (body-part/location)’, ko- ‘APPL (goal/recipient)’, and o- ‘APPL (path)’ open the prosodic word as a light (CV) σ1 when they attach to a base Nakagawa (2024: 53); (Bugaeva 2006). Because σ1 is open, 原則2 shifts the accent to σ2 of the new prosodic word, which is the σ1 of the original base. The result is that prefixation moves the high syllable one position earlier in the string compared with the free base.
| free / phrasal form | accent | e- form | accent | gloss (e- form) | source |
|---|---|---|---|---|---|
| puni | púni | e=puni | epúni | 'lift up (something)' | Nakagawa (2024: 53) |
| kotan ne (phrase) | kotán ne | e=kotan-ne | ekótanne | 'belong to a village, be a villager' | aynu-corpora Discord (2023–2026) (nukopoli 2024-11-10) |
In ekótanne, the prefix e- creates a new light σ1 → 原則2 → accent on σ2 (kó). The free word kotán accents σ2 (tán) by the same rule; in the applicative form the high syllable has moved one position to the left. The ekótanne example is a community extrapolation from the general one-word analysis of applicative-prefixed forms aynu-corpora Discord (2023–2026) (nukopoli 2024-11-10), consistent with the published epúni Nakagawa (2024: 53) ‹corpus-suggested›. The morphology and syntax of the applicative prefixes are treated in Chapter 78 (The Applicative System: Overview and the Preverb-vs-Applicative Debate).
26.4.3 Causative suffixes and the lexicalized koré
Derivational suffixes generally do not displace accent in Hokkaido Ainu Nakagawa (2024: 53). The causative suffixes -re, -te, -e attached to monosyllabic bases retain the base accent: e → ére; ku → kúre; kar → káre.
The verb koré 'give' is a well-attested departure. A regular causative derivation from kor 'have' + -e would yield *kóre (σ1 kor = CVC → 原則1 → σ1 accented); the attested form is koré, with accent on σ2 Nakagawa (2024: 53); Shiraishi (2022: §5.2). Nakagawa's account is that the morpheme boundary between kor and -e has eroded sufficiently for the form to be treated as a monomorphemic disyllabic root, whereupon σ1 ko is open → 原則2 → accent on ré. The community proposes alternatively that koré is a phonological contraction of korpare 'give to someone', which independently yields σ2 accent aynu-corpora Discord (2023–2026) (nukopoli 2024-12-26) ‹contested›. Both accounts predict the same surface accentuation; the accent koré is attested in the Saru corpus:
tópenpe atay póka ka somo a=koré
‘She couldn't even be given the price of one sweet.’
ILCAA Ainu materials 1976; Saru, 川上まつ子, isoytak 独話2, AA研アイヌ語資料 [aa-irc/004#105]
The σ2 accent of koré reflects the iambic pattern of a lexicalized open-initial disyllable (σ1 ko = CV, open → 原則2). The co-existent ditransitive korpare 'give to someone' is attested alongside it in the Saru corpus.
26.5 Person-affix accent
26.5.1 The shifting class: ku=, e=, ci=, en=
The person-marking prefixes ku= (1SG agent/subject), e= (2SG), and ci= (inclusive 1PL) form a single prosodic word with their base Nakagawa (2024: 52); Shiraishi (2022: §5.1); Satō (2008: §2.6). Each prefix contributes a light (CV) σ1; 原則2 applies and the accent falls on σ2 of the new word — one position earlier than the base accent when the base's own σ1 was open:
| base | accent | prefixed form | accent | gloss | source |
|---|---|---|---|---|---|
| hopuni | hopúni | ku=hopuni | ku=hópuni | 'I rise' | Nakagawa (2024: 52) |
| nanu | nanú | e=nanu | e=nánu | 'your face' | Nakagawa (2024: 52) |
| yupihi | yupíhi | ku=yupihi | ku=yúpihi | 'my elder brother' | Shiraishi (2022: §5.1) |
In each case the base has open σ1 and accents σ2; after the CV prefix is added, σ1 of the new prosodic word is the prefix itself (open) and σ2 — which is the original σ1 of the base — receives the high pitch.
In Saru and Chitose, ku= and ci= additionally lose their vowel before a base-initial vowel other than i. The conditions for this vowel deletion are described in Chapter 31 (Personal-Affix Junctural Sandhi, =an/a= Allomorphy, and Connected-Speech Reduction) and Chapter 22 (Glides, Vowel Hiatus, and the Diphthong Question). The accent shift precedes deletion, so the shifted accent survives in the contracted surface form: ku= + omáre → accent shift → ku=ómáre → vowel deletion → k=ómare; ci= + etú → c=étu Nakagawa (2024: 52).
Before a base-initial i, vowel deletion is blocked and glide formation (i → y) substitutes; the accent shift still applies: ku=ipe → ku=ýpe / kúype. When the base-initial i is in a closed syllable, glide formation is itself blocked by the phonotactic constraint against CCC clusters: ku= + inkar → ku=ínkar with no glide and accent on σ2 Nakagawa (2024: 52–53).
The 1SG object prefix en= behaves analogously: it opens the prosodic word as a closed σ1 (en, CVC), so 原則1 places the accent on σ1. In Nakagawa's textbook the phrase wa én=kore shows the accent on én, consistent with this rule aynu-corpora Discord (2023–2026) (nukopoli 2024-11-10, attributed to Nakagawa (2024)) ‹corpus-suggested›.
26.5.2 The non-shifting class: a=, eci=
The indefinite-person prefix a= and the 2PL prefix eci= leave the accent of the base unchanged Nakagawa (2024: 53); Shiraishi (2022: §5.1); Satō (2008: §2.6). They attach outside the prosodic word of the stem:
| base | accent | a= / eci= form | accent | gloss |
|---|---|---|---|---|
| omare | omáre | a=omare | a=omáre | 'was put (somewhere)' (4th/indef.) |
| etu | etú | eci=etu | eci=etú | 'your (pl.) nose' |
| onaha | onáha | a=onaha | a=onáha | 'his/her mother' (4th/indef.) |
a= and eci= also do not trigger the vowel contraction that the shifting prefixes undergo in Saru and Chitose after the accent shift. The combined phonological evidence — no shift, no contraction — aligns with the morphosyntactic assessment in Chapter 62 (Architecture of the Personal-Affix System: The Four Persons and the S/A/O Paradigms) that a= and eci= are prosodic clitics, while ku=, e=, ci=, and en= are prefixes proper. The position-class arrangement of person affixes in the verb word is set out in Chapter 56 (The Verb Word and Its Position-Class Template).
26.5.3 Secondary accent on personal-number suffixes
When a personal-number suffix (=an, =as) attaches to a vowel-final verb stem, the suffix carries a secondary (lower-level) accent alongside the primary stem accent Shiraishi (2022: §5.1). The verb iruska 'be angry' with the plural suffix gives irúska=àn (primary accent á on rú, secondary accent à on àn); ipe 'eat' with =an gives ipé=àn. When the stem ends in a consonant, the secondary accent on the suffix is lost: cis=an 'we cry' and sap=an 'we go out' carry only the primary accent on σ1 (both stems have closed σ1 → 原則1) with no secondary peak on =an Shiraishi (2022: §5.1). The number-suffix paradigm is described in Chapter 56 (The Verb Word and Its Position-Class Template) and Chapter 62 (Architecture of the Personal-Affix System: The Four Persons and the S/A/O Paradigms).
26.6 Accent-neutral suffixes: -pa and the affiliative -hV
The plural suffix -pa and the affiliative (possessed) suffix -hV — realizing as -hi, -he, -ha etc. according to the vowel-copy rule — are both accent-neutral: the stem's placement-rule accent survives into the suffixed form without adjustment Nakagawa (2024: 53); Tamura (1996). They attach outside the prosodic word of the stem and do not produce the reapplication that the prefixal and applicative morphology triggers. The morphophonology of the affiliative suffix and the echo-vowel alternation are treated in Chapter 45 (Morphophonology of the Affiliative Suffix and Echo/Copy Vowels).
26.7 Prosodic-word status as a morphological diagnostic
The accent patterns reviewed in this chapter offer a phonological diagnostic for grammatical category that is independent of morphological segmentation. A morpheme that enters the prosodic word of its host and shifts the accent is a prefix in the prosodic sense; a morpheme that leaves the host accent untouched is a clitic or a non-incorporating suffix. Applied to the Hokkaido Ainu person-affix paradigm, the test produces: prefixes — ku=, e=, ci=, en=; clitics — a=, eci= Nakagawa (2024: 52–53); Shiraishi (2022: §5.1). The same prosodic-word criterion distinguishes compound-internal structure (a single prosodic word, one accent peak) from phrasal juxtaposition (two prosodic words with independent accent peaks).
The compound formation discussed here and the sandhi processes of Part V share the same domain boundary. Coda assimilation (r+t→tt, n+s→ys, etc.) applies most consistently within the prosodic word; the accent test provides an independent means of determining which phonological domain a given morpheme occupies before examining the assimilation data. These interactions are taken up in Chapter 28 (Assimilation and Cluster Simplification: Coda /r/, Nasals, and Clusters) and Chapter 31 (Personal-Affix Junctural Sandhi, =an/a= Allomorphy, and Connected-Speech Reduction). The underlying prosodic unit and the weight analysis are addressed in Chapter 24 (The Prosodic Unit and the Accent-versus-Tone Analysis).
References cited in this chapter
aynu-corpora Discord (2023–2026) ·Bugaeva (2006) ·ILCAA Ainu materials (1976) ·Nakagawa (2024) ·Satō (2007) ·Satō (2008) ·Shiraishi (2022) ·Tamura (1996)