Chapter 24The Prosodic Unit and the Accent-versus-Tone Analysis

24.1 Typological placement: pitch accent, not tone

Hokkaido Ainu has lexical pitch accent: the position of high pitch within a phonological word is specified in the lexicon and contrasts meaning. Pitch is not assigned to every mora or syllable independently, as in a tone language; a single accent nucleus per word marks the site of the rising boundary, and syllables before it are uniformly low Shiraishi (2022: §5); Nakagawa (2024: 48–49); Satō (2008: §2.6). The system involves no lexical tone and no vowel reduction correlating with accent position Shiraishi (2022: §5). All three primary sources for Hokkaido Ainu agree on this characterization ‹consensus›.

The surface output phonetically resembles Standard Japanese pitch accent — Tamura (1988b: 13) remarked on the similarity (reported via Shiraishi (2022: §5)) — but the two systems are organized in structurally opposite ways, as the next section sets out. Within the pitch-accent type, the further question of whether the accent nucleus is best formalized as a feature of the syllable or of the mora is taken up in § Mora or syllable?; the question of whether the system could be described as a restricted tone or register system is addressed in § Accent versus tone.

Not every description has found the system. Simeon, eliciting at dictation speed from rememberer-speakers in 1967, reported no minimal pairs for stress and tied primary prominence to the final syllable of his "contour" unit, leaving secondary stress unpredictable Simeon (1968: 11–12) — this in Hiratori, the same dialect area where Tamura was concurrently documenting lexical accent contrasts of the nísap : nisáp type. The discrepancy is best explained by method: list intonation in slow elicitation overwrites lexical pitch with phrase-level prominence, so the accent system is invisible unless connected speech or paradigmatic contrast is examined. The episode is a standing caution for reading any pre-Tamura Western description of Ainu prosody ‹speculative›.

24.2 The rising-nucleus system

The decisive structural difference from Japanese is directional. Standard Japanese is a falling-nucleus system (下り核): the nucleus marks the position after which pitch drops, and syllables up to and including the nucleus are high. Hokkaido Ainu is a rising-nucleus system (上り核アクセント): the nucleus marks the onset of the accented syllable, the pitch rises at that point, and syllables following the nucleus remain high without further contrast Nakagawa (2024: 48–49). The same asymmetry was recognized under the term 昇り核 'rising kernel' by Hattori (1961) and Tamura (1970), reported via Shiraishi (2022: §5). The consequence is that sustaining or extending the post-accent high plateau does not change a word's identity, whereas shifting the pitch rise to a different syllable does.

The community illustration of this point is apt: the first-person pronoun kuáni may be produced with a flat high plateau after the rise (ku↗áni) or with a fall after the peak (ku↗á↘ni) without any change of meaning, but a fall before the accent (ku↘áni) is perceived as anomalous or contrastive aynu-corpora Discord (2023–2026) (nukopoli, 2024-12-09; corroborated by Nakagawa (2024: 48–49)) ‹corpus-confirmed›. In the minimal pair nísap 'suddenly' : nisáp 'shin', only the syllable at which the ↗ rise occurs differs; the high plateau that follows is common to both Nakagawa (2024: 48–49, 51).

The accented vowel also tends to lengthen allophonically in polysyllables, particularly in north-eastern accent dialects — Kindaichi (1931: 4) noted forms such as ˈnuːman 'yesterday' and poˈroː 'big', describing the lengthening as "not conducted consciously" (reported via Shiraishi (2022: §5)). This allophonic lengthening is distinct from the phonemic vowel length of Sakhalin Ainu (§ Diachrony below) and carries no accent-analytical implications.

24.3 Syllable weight and the placement rule

Accent placement in the standard accent dialects (Saru, Chitose, and their neighbors) follows a two-rule system sensitive exclusively to the syllable structure of σ1 Nakagawa (2024: 50):

Only σ1's shape matters; word length beyond two syllables is irrelevant to placement. Shiraishi reformulates the same pattern as an iambic (LH) foot: accent on σ2 when σ1 is light, on σ1 when σ1 is heavy — a weight-sensitive rhythmic pattern widespread across languages Shiraishi (2022: §5.1). Kindaichi and Chiri (1936: 11) concluded from their materials that "the iambic rule is dominant" (reported via Shiraishi (2022: §5.1)).

A class of lexical exceptions breaks the default: when σ1 is open yet accent falls on σ1, the word must be memorized. Nakagawa lists around twenty such items, including cápe 'cat', húci 'grandmother', húre 'red', kéra 'taste', míci 'father', mína 'laugh', nísap 'suddenly', páse 'heavy', sísam 'Japanese person', síno 'really', túki 'cup', túnas 'fast', and yúpo 'elder brother' Nakagawa (2024: 51). These exceptions produce minimal accent pairs:

These pairs establish that accent position is phonemically contrastive and that the weight rule, though dominant, does not determine accent in the exception class. The full inventory of such pairs is the subject of Chapter 25 (Lexical/Contrastive Accent and Minimal Pairs); accent behavior under compounding and affixation is treated in Chapter 26 (Accent in Compounds and under Affixation and Cliticization).

24.4 Mora or syllable as the accent domain?

In Japanese pitch accent the mora is the accent-bearing unit: a CVC syllable may, in principle, bear accent on either of its two moras. For Hokkaido Ainu the evidence points to the syllable as the relevant domain.

Three observations converge on this conclusion. First, the placement rule (§ above) is stated entirely in terms of whether σ1 is open or closed — syllable weight rather than mora count. No sub-syllabic accent position within a CVC syllable is ever distinctive; all coda consonants (/p t k m n r s y w/) are equivalent in their weight-conferring effect. Second, the phonetic realization — the pitch rise — is timed to the onset of the accented syllable, not to an internal mora boundary. Third, the analysis of coda /-y/ and /-w/ as consonants rather than as the second element of a diphthong (argued independently in Chapter 22 (Glides, Vowel Hiatus, and the Diphthong Question) on morphological and phonotactic grounds) is consistent with syllable-based accent: ka.muy 'deity' (two syllables, σ2 = CVC) receives accent on σ2 as an undivided unit; there is no evidence that the internal /-y/ constitutes a separate accent-bearing mora Nakagawa (2024: 29); Shiraishi (2022: §3, §5.1).

Person-prefix accent shift provides additional evidence. The prefixes ku=, e=, and ci= integrate with the stem as a single phonological word; the weight of σ1 in the new word is recalculated with the prefix included. A stem whose open σ1 would otherwise take σ2 accent by 原則2 shifts its accent one syllable toward the word boundary Nakagawa (2024: 52); Shiraishi (2022: §5.1):

The contrasting prefixes a= and eci= do not trigger the shift: omáre → a=omáre retains the stem's accent Nakagawa (2024: 53). These two prefixes are prosodically less integrated with the stem, a gradient that parallels their position in the person-affix paradigm (see Chapter 62 (Architecture of the Personal-Affix System: The Four Persons and the S/A/O Paradigms) and Chapter 26 (Accent in Compounds and under Affixation and Cliticization)).

A moraic restatement of 原則1/2 is formally possible: a closed σ1 has two moras (CV + C), so accent could be described as associating with the second mora of the word — belonging to σ1 when σ1 is heavy, to σ2 when σ1 is light. This restatement covers the same data but posits mora-counting that is not independently evidenced elsewhere in Ainu phonology. The sources consulted do not argue for it: Shiraishi characterizes the pattern as iambic and syllable-based Shiraishi (2022: §5.1), and Nakagawa states the rule in syllable-structure terms Nakagawa (2024: 50) ‹contested›.

24.5 Accent versus tone: the analytical question

The pitch-accent label is used in two overlapping senses in the Japanological tradition: as a cover term for any system in which pitch is phonemic, and in the narrower sense of a system with one positionally specified nucleus per word. All three primary sources use it in the narrow sense as applied to Hokkaido Ainu. Pre-nuclear syllables are uniformly low; post-nuclear syllables are uniformly high; no syllable-by-syllable tonal specification is posited. Ainu is not a tone language in the cross-linguistic typological sense ‹consensus›.

Two analytical positions on the structure of the exception class merit explicit grading.

Position A — gradient weight-sensitive accent with a lexical exception class ‹consensus›. The system is weight-sensitive iambic accent; the exception class is a bounded set of forms with lexically specified initial accent. The default rule accounts for the vast majority of the lexicon; exceptions must be memorized individually. This is the position of Nakagawa Nakagawa (2024: 50–51), Shiraishi Shiraishi (2022: §5.1), and Satō Satō (2008: §2.6), as well as Kindaichi and Chiri (1936: 11) (reported via Shiraishi (2022: §5.1)).

Position B — two prosodic word-classes ‹speculative›. The exception class (open σ1, initial accent) and the default class (weight-conditioned accent) could be reframed as two lexical register classes: Class I words carry a lexical feature placing accent on σ1 regardless of weight; Class II words are unspecified and realize weight-determined accent. This account is structurally parallel to two-register systems described for some East Japanese and Ryukyuan pitch-accent varieties. The Sakhalin vowel-length correspondence (§ Diachrony below) supplies a diachronic motive: the HA exception class corresponds systematically to SA long-vowel forms, suggesting the two HA classes may reflect a former segmental length distinction rather than a discrete tonal specification. This reanalysis has not been proposed in the sources consulted; the synchronic data do not require it; it is noted here as a possible alternative formalization ‹speculative›.

Community discussion noted a phonetic resemblance between Hachijō-language intonation and Ainu pitch contours aynu-corpora Discord (2023–2026) (nukopoli, 2024-12-08). No systematic published comparison of the two systems has been consulted here; the significance of any resemblance — intonational, accentual, or areal — cannot be assessed from the sources consulted ‹speculative›.

24.6 Accent dialects, irregular dialects, and accentless varieties

Not all Hokkaido varieties have contrastive pitch. Nakagawa and Fukazawa classify Hokkaido dialects into three prosodic types Nakagawa & Fukazawa (2022: §3.1); Nakagawa (2024: 49–50):

The irregular accent of Yakumo shifts the iambic peak one position right when both σ1 and σ2 are open: compare Yakumo notakám vs Saru notákam, Yakumo sinuyé vs Saru sinúye, Yakumo hoyupú vs Saru hoyúpu; but when σ2 is closed both varieties agree: parúnpe in both Nakagawa (2024: 50–51); Nakagawa & Fukazawa (2022: §3.1).

The accentless status of Shizunai is contested. Refsing (1986: 73) treated it as an accent dialect with rules parallel to Saru Refsing (1986: 73). Nakagawa argues from museum-archive recordings that the same word appears in both accent patterns across tokens from one speaker — nisáppone five times vs nísappone twice — and reads this as accent-neutralization, not a stable system Nakagawa (2024: 50) ‹contested›. One phonological consequence of Shizunai's accentless status: the contrast between glide [w] and glottal stop [ʔ] at a hiatus position must carry the functional load that pitch contrast carries elsewhere, as argued for the pair suwat 'pothook' vs suʔat 'pot bail' in Okuda (2025); see Chapter 22 (Glides, Vowel Hiatus, and the Diphthong Question).

The accentless dialects differ in their accentless origin. Bihoro, Kushiro, and Shiranuka lost the pitch contrast without a compensatory change. Sakhalin Ainu is also accentless but preserves a vowel-length opposition where Hokkaido developed pitch; neither opposition has neutralized in Sakhalin Nakagawa & Fukazawa (2022: §3.1). Dialect-level consonant features that intersect with accent — including eastern h-elision when accent falls on σ2 — are treated in Chapter 160 (Hokkaido Dialect Microvariation: Phonology and Morphosyntax).

24.7 The Sakhalin length–Hokkaido accent diachrony

The correspondence between HA accent class and SA vowel length runs through the lexicon systematically: HA exception-class words (open σ1, initial accent despite the weight rule) correspond to SA long-vowel forms; HA regular-class words (weight-conditioned accent) correspond to SA short-vowel forms Nakagawa (2024: 27, 51); Nakagawa & Fukazawa (2022: §3.1):

Hattori (1967) reconstructed vowel length as the earlier contrast: Proto-Ainu had a long-vowel/short-vowel distinction, Hokkaido replaced it with accent while Sakhalin retained length (reported via Nakagawa (2024: 27); Nakagawa & Fukazawa (2022: §3.1)). On this account the HA exception class reflects historically long syllables; the weight-conditioned default (原則2) is a Hokkaido innovation. Satō dissents on the directionality in his contribution to the Handbook (reported via Nakagawa & Fukazawa (2022: §3.1)) ‹contested›. Itabashi (2001) argues the length–pitch diachrony in detail Itabashi (2001); that study was not consulted directly for this grammar and the claims attributed to it here rest on its treatment in Nakagawa (2024) and Nakagawa and Fukazawa (2022).

The correspondence also supplies diachronic motivation for Position B (§ above): if the HA exception class reflects historically long vowels rather than a distinct tonal category, then the two HA accent classes may be the synchronic residue of a former segmental opposition rather than evidence of a register system in Proto-Ainu. Whether this warrants a formal two-class reconstruction is a question for the comparative work in Chapter 156 (Proto-Ainu Reconstruction: Segments and Accent Classes) and Chapter 157 (The Final-h Problem and Its Sakhalin Vowel-Length Reflex) ‹speculative›.