Chapter 151The Sakehe Refrain, Verse Meter, and the Structure of Sung Verse
The sakehe burden and the metrical/structural organization of sung genres as performance grammar: refrain types and placement, verse-line segmentation and syllable/mora-count meter, and melodic vs linguistic pitch (proto-accent reconstruction in Part XX).
151.1 The sung/recited/spoken distinction
Melody presence is the most grammar-relevant parameter dividing Hokkaido Ainu oral genres. Three verse-and-narrative traditions each carry a distinct constellation of features — melody, the sakehe ‘refrain (in sung verse)’ refrain, and rhythm-tapping — that co-condition the grammar of the text Satō (2008: 259); Endō (2022: §2.4).
| genre (Ainu) | melody | sakehe refrain | rhythm-tapping (rep / repni) | length | canonical narrator |
|---|---|---|---|---|---|
| heroic epic yúkar | yes (narrator's own tune, variable) | no (but see opening eyororpe discussed below) | yes | long (hours to multi-night) | men (traditionally) |
| divine epic kamuy yukar | yes (fixed per story) | yes | no | short (minutes to ~1 hr) | women (and men) |
| prose-style epic irupaye | no | no | no | long | women |
| prose tale uwepeker | no | no | no | medium | women (and men) |
(After Satō (2008: 259), table of Chitose oral-literature genres and features. Genre taxonomy, regional dialect-term variation, and the oyna nomenclature debate are covered in Chapter 149 (The Oral-Literature Genre System and Its Grammatical Signatures).)
The sung/prose split is not merely a performance fact. Sung verse genres deploy the poetic register (atomte itak, 'ornamented language') with its distinct morphological and lexical features, while prose (yayan itak, 'ordinary language') does not — even when the prose narrator's voice rises and falls with a narrative lilt Endō (2022: §3); Satō (2008: 263–264). The elevated register is described in Chapter 153 (The Elevated/Poetic Register: Archaic Morphology and Formulaic Diction); its TAM and evidential patterning by genre in Chapter 155 (Narrative TAM and Evidential Patterning by Genre). Folktale intonation — one-directional, low-variation melody, with upper-Saru narration more monotone than lower-Saru — sits apart from the strict melodic structure of sung verse (aynu-corpora Discord (2023–2026), nukopoli 2024, asserted, after Tamura (1996)).
The verb sakehawki ‘sing/recite a yukar’ (E-Iburi) captures the sung dimension lexically; it appears in a Abuta yukar passage meaning 'about to sing a yukar' (sakehawki anki), confirming that for at least the eastern Iburi dialects, sung performance was lexically distinguished from prose narration Satō (2009: §2.2).
151.2 The sakehe: morphology and the fixed-melody criterion
The term sakehe ‘refrain; the sake of —’ is morphologically the possessed form of sa (Japanese Wiktionary: "sa の所属形"). Tamura's Saru dictionary records a prosodic distinction: the genre-refrain sense carries initial accent (sákehe) while the common-noun sense 'sake (the beverage)' is unaccented, creating near-homophones that occasionally coincide within a single verse line (see the hancikiki example below) Tamura (1996: s.v. sákehe, sakehe). The corpus registers fifteen tokens of the form (rank 4,602 among 45,225 distinct tokens), spanning both senses; the refrain sense is attested in narrator metalinguistic commentary (see example below) ‹corpus-suggested›.
In kamuy yukar, the sakehe is a fixed refrain per story, inserted into the narration — every line or every few lines depending on the text — and held constant throughout one performance of that story Endō (2022: §2.4.2); Satō (2008: 259). The melody to which the text is sung is equally fixed per story (contrast yukar, where each narrator deploys a personal tune that may change mid-performance; see the melody section below). Together, the fixed melody and the fixed sakehe define the formal identity of a kamuy yukar text: a reciter who knows a story knows both its sakehe and its melody as an inseparable package Endō (2022: §2.4.2).
A canonical instance of sakehe-every-line placement comes from a marlin-tuna kamuy yukar narrated by Nabezawa Nepuki (鍋沢ネブキ) of Biratori. The sakehe tusunapanu precedes every narrative word-group, producing a strict alternation of refrain and narrative that generates the rhythmic frame of the performance (Endō (2022: §2.4.2), reporting Kayano (1996: s.v. sakehe); narrator 鍋沢ネブキ, Biratori):
| refrain | narrative segment | refrain | narrative segment |
|---|---|---|---|
| TUSUNAPANU | Okikurmi | TUSUNAPANU | Samayunkur |
| TUSUNAPANU | u-tura hine | TUSUNAPANU | repa kusu arki |
| TUSUNAPANU | ki hi kusu | TUSUNAPANU | etok-oho ta |
('Okikurmi and Samayunkur went offshore to fish, so… before that… ahead of [them]…'; the sakehe in capitals precedes every narrative phrase.)
Attested sakehe from the aligned Hokkaido corpus and oral-literature archives show the range from onomatopoeic to opaque. The following is the thunder-god's sakehe from a Biratori kamuy yukar, with the narrator's own gloss preserved:
‘hum-pakpak (the sakehe of the thunder-god)’
Biratori Ainu oral literature 1969; Saru, Kaizawa Turushino 貝澤とぅるしの, Biratori; kamuy yukar 'フㇺ パㇰパㇰ（雷の神が自叙する神謡）'
The narrator's own metalinguistic comment (biratori/001/004#170): humihiruy だわ。sakehe. 'it's a loud sound — that's the sakehe.' Onomatopoeia of thunder's cracking/pattering.
The next example, from a sparrow-feast kamuy yukar, illustrates a structural property of the Saru verse line: the refrain hancikiki immediately precedes a narrative line whose first word is the common noun sakehe 'sake (millet wine)', creating a momentary homophony between the genre-term and the beverage:
‘[HANCIKIKI] — we made sake’
Biratori Ainu oral literature 1969; Saru, Nabezawa Nepuki 鍋澤ねぷき, Biratori; kamuy yukar 'ハンチキキ（スズメの酒盛り）'
The sakehe hancikiki precedes a narrative line that opens with the homophonous common noun sakehe 'sake (beverage)'. The refrain and the narrative vocabulary coincide accidentally.
The story-type and sakehe are demonstrably not in fixed correspondence: the same narrative situation (marlin-tuna confrontation) carries sakehe tusunapanu in Saru (Biratori), tussay turi tussay matu in Horobetsu, and nissō in Chitose (where the protagonist is swapped for the fish okina) Endō (2022: §2.4.2). Kayano's Saru dictionary preserves a long multi-word sakehe (tunoyake renoyake kutunke kamuyke kamuy cikappo hum hum hum) as one illustration of the form's capacity for extended, rhythmically organized strings Kayano (1996: s.v. sakehe).
151.3 Melody: fixed per story in kamuy yukar, personal in yukar
The melodic contrast between kamuy yukar and yukar is the clearest formal divide in the genre system. A kamuy yukar text comes with a melody that is bound to that story and is transmitted alongside the words and the sakehe; a performer who learns the story learns all three as an integrated unit Endō (2022: §2.4.2). Prayer (kamuy nomi) also has a melodic delivery, but its melody is improvised and personal — different each time and not transmitted with the words — a property Nakagawa (via Endō (2022: §2.3)) uses to distinguish prayer from the fixed-word, fixed-melody incantation.
Yukar stands apart: each narrator constructs or inherits his own tune for a borrowed story, and the melody is liable to change mid-performance Endō (2022: §2.4.3). The absence of a sakehe in yukar corresponds to this melodic instability: where the kamuy yukar uses the repeating sakehe to bind melody, narrative line, and rhythmic unit together, the yukar uses rhythm-tapping (rep ‘keep time’ / repni ‘rhythm-stick’ — 'the act of keeping time', confirmed in the Mukawa-dialect dictionary as 拍子棒 Satō (2008: 259)) to provide rhythmic scaffolding without a textual refrain. Listeners add a sung chorus (hetce ‘chorus response’; corpus count 13) during yukar performance, a further rhythmic layer absent from kamuy yukar Satō (2008: 259).
Satō notes that an opening formula — the eyororpe ‘opening formula (yukar)’ — may recur at the start of a yukar and could represent a vestige of a refrain that yukar once possessed Satō (2008: 259) ‹speculative›. The inference is Satō's own and rests on structural analogy with kamuy yukar rather than on textual attestation.
151.4 Function of the sakehe: the protagonist-index debate
The functional motivation of the sakehe has generated competing accounts. Nakagawa proposes that it originally indicated the protagonist: the refrain was an identifying tag that signalled whose voice or experience was being narrated Endō (2022: §2.4.2), reporting Nakagawa 2001b: 7 ‹contested›. The counter-evidence is substantial: the same story-type does not consistently carry the same sakehe across dialects (the marlin-tuna case above), and many sakehe have no recoverable semantic or deictic link to their protagonist. The thunder-god's hum pakpak can be read as onomatopoeia of thunder's sound (the narrator's own gloss supports this), but that is iconic, not indexical, and it is not generalizable to the full sakehe inventory Endō (2022: §2.4.2).
A performer's own metalinguistic statement, recorded in the AA研 archive, puts the sakehe at the centre of the sung/prose boundary: the narrator Kawakami Matsuko explains that a menoko yukar is performed sakekoye ('told with the refrain'), and that when the refrain is omitted the same material is told as uwepeker:
menokoyukar って sakekoye するんだけども sakehe 言わないで uwepeker に言うの。
‘A menoko yukar is told with the refrain (sakekoye); but when one does not say the sakehe, one tells it as a uwepeker (prose tale).’
ILCAA Ainu materials 1976; Saru, 川上まつ子, AA研「menokoyukar 神謡6（節なし）」[aa-irc/019#0]
A metalinguistic (code-switched) statement in which the narrator names the criterion that separates sung verse from prose: presence versus absence of the sakehe refrain. The same narrator adds (aa-irc/019#66) that 'the sakehe is long, the uwepeker is short' — the refrain is also a length-diagnostic. The m/g lines gloss only the Ainu words embedded in the Japanese-metalanguage frame shown in the ain surface line.
The community discussion archives one live genre-boundary case that bears on the sakehe's defining role. A kamuy yukar narrated by Oda Suteno (織田ステノ, 'Lazybones Child') carries a long sakehe (manayta sanke tar panke tar kocupu) but an unusually short narrative body, and its closing move — the protagonist self-identifying as a human boy (sine wen toranne hekaci, 'a single weak miserable child') — formally resembles a yayeisoytak (self-narration) rather than a divine-epic conclusion (aynu-corpora Discord (2023–2026), gengojiro 2025-12-12, proposed). This text, with melody and a long sakehe but a human-protagonist resolution, resists placement in either kamuy yukar or yukar, illustrating that the sakehe alone does not uniquely define genre membership ‹contested›.
One additional philological datum: the title-line AEKIRUSHI in Chiri Yukie's Chiri Yukie (1923) has been proposed to parse as a=e-kir-usi 'the place one has memory of' (aynu-corpora Discord (2023–2026), 0penke 2024-03-19, proposed, after 片山龍峯 2003) rather than deriving from ekiru 'turn back'. If correct, the title-line encodes the hero's logophoric self-location, consistent with the divine-epic pattern of first-person self-presentation ‹contested›.
151.5 Verse metre: two analytical systems
The metrical organization of HA verse has been described under two competing frameworks. The long-standing analysis treats verse as syllable-count oriented, with five syllables per line as the basic unit. Okuda's more recent work identifies a second system, accent oriented, in which the first pitch-accent falls in a fixed rhythmic position regardless of line length. The two frameworks are not simply rival accounts of the same data: Okuda argues they are genre-differentiated and co-active within a single performer's repertoire ‹contested› Okuda (2022: §2, §3).
151.5.1 Syllable-count orientation
The syllable-count description goes back to Kindaichi (1931: 165, via Okuda (2022: §2.1)) and was adopted by Tamura (1973: 56, via Okuda (2022: §2.1)): the basic line unit is five syllables, with four to seven as the attested range. The five-syllable target is met through a set of compositional techniques catalogued by Tamura and Satō Satō (2008: 264); Okuda (2022: §2.1):
- use of formulaic verse phrases already calibrated to syllable count;
- choice between synonyms for their syllable count;
- insertion or omission of final particles;
- insertion of the dummy verb ki ‘do (light verb)’ between a verb and its following element;
- use of the uncontracted ku- person prefix rather than its reduced form;
- prefixing an initial u to a line when reciting;
- splitting a long phrase across two five-syllable lines;
- not eliding vowels that would contract in everyday speech;
- not pronouncing the line-final syllable (a performance strategy, not a morphological rule).
These techniques confirm that the syllable target is a compositional constraint on the verse line rather than a property of the grammar at large. They are the mechanisms through which the sung verse actively shapes morphosyntax: word choice, prefix realization, and particle use are all subordinated to the five-syllable frame. This is the most direct route by which genre conditions grammatical form in Hokkaido Ainu.
151.5.2 Accent-orientation metre
Okuda's comparative analysis of divine epics and lyric songs identifies a second system that he terms accent-orientation (アクセント指向): the first pitch-accent of the line falls in the first disyllabic foot, regardless of how many syllables the line contains Okuda (2022: §3). Under this system, a four-syllable line is metrically well-formed as long as the accent is initial; the syllable-count system, by contrast, disfavours four-syllable lines.
The scansion evidence comes from divine epics recorded by narrators in Monbetsu and Chitose. In the divine epic Amamecikappo narrated by Hatozawa Wateke (幌別/Monbetsu; text in Tamura 2001, via Okuda (2022: §3.3)), of 90 non-refrain lines, 14 are four-syllable lines; of these 14, twelve carry initial accent on the first syllable, matching the accent-orientation prediction. Representative lines (● = first accented syllable; ○ = unaccented):
| line | syllables | scansion | gloss |
|---|---|---|---|
| si-né-a-mam-pus | 5 | ○●○○○ | 'one ear of grain' |
| ci-tá-ta-ta-ta | 5 | ○●○○○ | 'I crushed and crushed' (iterative) |
| kí-wa-tap-ne | 4 | ●○○○ | 'and then' |
| c-é-a-hun-ke | 4 | ●○○○ | 'I invited (it) into the house' |
(After Okuda (2022: §3.3), Hatozawa Wateke, Biratori/Monbetsu; primary text Tamura 2001.)
In Chitose divine epics, the pattern holds with similar consistency. The text Wawo kamuy (Shirasawa Nabe; 226 lines) has 62 four-syllable lines, of which 56 carry initial accent; Apehuci kamuy (same narrator; 220 lines) has 79 four-syllable lines, of which 71 do Okuda (2022: §4). The same reciters' lyric songs (yaysama) are overwhelmingly five-syllable (e.g., 33 of 34 lines in Hatozawa Wateke's lyric sample), confirming that the same performer switches between systems by genre Okuda (2022: §3.2).
151.5.3 Genre distribution and the debate
Okuda maps the two systems onto genres and musical rhythms Okuda (2022: §3, §5.1):
| metre | primary genres | four-syllable lines | theoretical type |
|---|---|---|---|
| syllable-count orientation | lyric song (yaysama), prayer (inonno itak) | disfavoured (<5% attested) | pure-syllabic (Lotz type A) |
| accent orientation | divine epic (kamuy yukar), heroic epic (yukar; both systems) | admitted (15–30%); accent obligatorily initial | accentual-patterning (Lotz type B2) |
Okuda's claim that a single performance tradition deploys two such metrically distinct systems is, on his own account, without parallel in the reported typological literature Okuda (2022: §5.1) ‹contested›. The traditional syllable-count description (Kindaichi, Tamura) remains well evidenced for lyric and prayer; the accent-orientation analysis adds explanatory coverage for the divine and heroic epics, where four-syllable lines with initial accent resist treatment under a pure syllable-count rule. The two analyses are therefore complementary across genres rather than mutually exclusive ‹contested›.
A further hypothesis, explicit in Okuda and flagged as conjectural, relates the syllable-count system to Japanese 7-5 mora metre: the contact with Japanese literary tradition could have reinforced or introduced the strict syllable target Okuda (2022: §5.2) ‹speculative›. No positive evidence is offered and the hypothesis rests on structural analogy alone.
151.6 Alliteration and rhyme
Philippi's influential judgment (1979, via Okuda (2022: §2.2)) was that rhyme in Ainu poetry is 'quite accidental' and alliteration 'sporadic' — not prosodically functional as in Mongolic or Tungusic traditions. Okuda formalizes the reasoning: given five vowels and roughly twelve onset consonants, chance repetition in a corpus of five-syllable lines is high, and a sequence counts as technique only when it demonstrably exceeds chance and is not forced by morphological reduplication, which is pervasive in Ainu Okuda (2022: §2.2).
A counter-position has been raised in the community literature. Tangiku (2020, 北大言語 アーカイヴ報告書) argues that an 18th-century Ainu yukar — ルウェサニウンクㇽ — uses deliberate head-rhyme, end-rhyme, and imperfect rhyme as compositional devices (aynu-corpora Discord (2023–2026), nukopoli 2023-12-13, asserted, after Tangiku 2020). If the Tangiku analysis survives scrutiny, it would demonstrate at least one historical period or textual tradition in which rhyme was a conscious structural choice rather than an artefact of phonological probability ‹contested›. The Tangiku source has not been consulted directly for this grammar; the claim is flagged as needing verification against the primary text.
151.7 Fourth-person narration in sung verse
Both yukar and kamuy yukar are narrated in the first person from the protagonist's viewpoint. The grammatical marker of this 'first person' is, in Saru and Chitose, the fourth-person prefix a= and suffix =an, not the everyday first person ku= Satō (2008: 259–261); Endō (2022: §2.4). The choice of the fourth-person forms is grammatically productive: in elegant language, =an combines with singular verbs (e.g., ek-an 'I come.SG'), a combination that everyday Saru disallows (*ek-an is ill-formed there but arki-an 'we come.PL' is fine) Satō (2008: 263). This number-agreement override is a diagnostic of the elegant register in sung verse — see Chapter 153 (The Elevated/Poetic Register: Archaic Morphology and Formulaic Diction).
The Iburi dialect tradition, canonically represented by Chiri Yukie's Chiri Yukie (1923), uses instead the exclusive first-person plural ci= / =as as a singular hero-narrator marker, consistently throughout the collection Satō (2008: 261); Satō (2004). The regional split is systematic: a= / =an in Saru and Chitose; ci= / =as in Iburi (Horobetsu and related traditions) — with Chitose showing an intermediate pattern where ci= / =as may appear in the opening lines before reverting to a= / =an Satō (2008: 261).
The motivation for the non-everyday person-marking in oral narrative is discussed under two competing accounts: the shamaness-oracle origin theory (standard), in which the verse is framed as spirit-speech, and Satō's 'indirection principle' (間接性の原理), in which the god's first-person speech inside a dream required indirectness to soften the taboo on direct divine address Satō (2008: 259–261) ‹contested›. The logophoric and direct-speech dimensions of the fourth person as used in narration are treated in Chapter 150 (First-Person Narration, the Logophoric Fourth Person, and Reported Discourse) and Chapter 68 (Honorific and Logophoric Uses of the Fourth Person); reference tracking across verse episodes in Chapter 139 (Reference Tracking and Argument Continuity); the full paradigm in Chapter 62 (Architecture of the Personal-Affix System: The Four Persons and the S/A/O Paradigms).
One grammatically significant consequence of sung-verse performance is quotative density. Because narration is cast as direct speech and closes with a self-identification formula (…sekor … isoytak 'so X told the story'), the quotative particle sekor ‘quotative’ with speech verbs (ye ‘say’, hawki ‘say/recite’, isoytak ‘tell a story’) clusters heavily in verse text-closings. The evidential tail hawe ne / hawe as (< haw ‘voice’ + possessed form) carries much of the reportative marking in these contexts Endō (2022: §2.4); the broader evidential system is treated in Chapter 120 (The Nominalization-plus-Copula Evidential Schema) and Chapter 121 (ruwe ne — the Inferential / Visual-Trace Evidential).
References cited in this chapter
aynu-corpora Discord (2023–2026) ·Biratori Ainu oral literature (1969) ·Chiri Yukie (1923) ·Endō (2022) ·ILCAA Ainu materials (1976) ·Kayano (1996) ·Okuda (2022) ·Satō (2004) ·Satō (2008) ·Satō (2009) ·Tamura (1996)