Chapter 20The Laryngeals: Glottal Stop, /h/, and the Final-h Question

20.2.1 Phonetic distribution

A glottal onset [ʔ] is phonetically present before every vowel-initial syllable in Hokkaido Ainu — word-initially and at internal morpheme and word junctures. Nakagawa (2024: 31) traces the earliest systematic observation to Kindaichi (1928), a Tōhoku-Japanese speaker who characterized Ainu as inserting a glottal closure before each vowel-initial syllable. The stop's auditory salience varies with phonetic context. Tamura (1970: 586–587, cited via Shiraishi (2022: §4.3)) distinguishes a perceptible stop — arising when the following vowel carries lexical accent and/or occurs in a closed syllable — from a barely perceptible transition that amounts to no more than the syllable-division point. Attested forms in which the stop is clearly audible include [aˈʔep] 'food', [aˈʔoypep] 'tableware', [oˈʔatteke], [eˈʔunpipka], and [teˈʔeta] 'long ago' (Tamura 1970: 586, cited via Shiraishi (2022: §4.3)).

The HA syllable canon (C)V(C) permits at most one onset consonant per syllable (see Chapter 21 (Syllable Structure, Phonotactics, and Word-Edge Constraints)). When a preceding morpheme ends in a consonant, that consonant resyllabifies as the onset of the following vowel-initial syllable: cip-o → [ci.po] 'board a boat', mat-ak → [ma.tak] 'younger sister' Nakagawa (2024: 43). No [ʔ] appears in those resyllabified forms, because the onset slot is already filled. The glottal closure thus marks syllables whose onset position remains empty — a phonetically automatic default rather than a lexically specified segment.

20.2.2 Phonemic status

Whether [ʔ] constitutes an underlying phoneme /ʔ/ in Hokkaido Ainu is contested Shiraishi (2022: §4.3). Two positions have been held since the mid-twentieth century:

The accent argument applies only in pitch-accent varieties. In the accentless Shizunai dialect it does not hold, and phonemic /ʔ/ has been demonstrated there (§ below). The non-phonemic analysis stands for the accent-dialect majority and is adopted here for those varieties Chiri (1942: 461) ‹contested›.

20.2.3 Juncture gradation and prosodic integration

The phonetic realization of [ʔ] at morpheme and word boundaries varies with the prosodic integration of the flanking elements. Three degrees of boundary tightness are audible in a single compound across careful and casual speech. In itakorapiru=an 'be telling the story in full', the most separated pronunciation is [itak|orapiru…] with a full juncture stop; an intermediate version shows [itak⁀orapiru…] with partial resyllabification alongside an audible [ʔ]; and the most integrated realization is [ita.ko.rapiru…] with full resyllabification and no glottal. Acoustic measurement confirms that the presence or absence of a voiceless interval between the coda /k/ and the following /o/ distinguishes the [−C.ʔV] and [+C.V] realizations ‹corpus-suggested› (aynu-corpora Discord (2023–2026), nukopoli, 2024-12-21).

The suffix =an (fourth-person intransitive subject) provides a clear test after a stop-final stem. After sap ‘depart, leave’ the suffix =an does not trigger resyllabification of the coda /p/ onto the following vowel; a juncture [ʔ] appears instead, yielding [sapʔan] ‹corpus-suggested› (aynu-corpora Discord (2023–2026), nukopoli, 2023-12-27).

The prefixes ku=, e=, ci= are prosodically integrated with their stem, forming a single prosodic word and triggering vowel-deletion and accent-shift processes (see Chapter 31 (Personal-Affix Junctural Sandhi, =an/a= Allomorphy, and Connected-Speech Reduction) and Chapter 23 (The Pitch-Accent Placement Rule and the Accented/Accentless Dialect Split)). The prefix a=, which does not shift lexical accent, is accordingly less tightly bound Nakagawa (2024: 53). Acoustic observation of a=kor itak 'there is speech' yields [a.ko.rəʔi.tak]: the echo vowel [ə] follows coda /r/ in kor, and [ʔ] appears at the compound boundary before itak, while the a=–kor boundary shows no juncture stop ‹corpus-suggested› (aynu-corpora Discord (2023–2026), nukopoli, 2024-03-18).

These observations together point to a gradient of boundary tightness within the person-marking system: the productive object/agent prefix class (ku=, e=, ci=) is the most integrated; the a= boundary is tighter than a compound word boundary but looser than the productive-prefix class; and the =an suffix boundary is the least integrated of the person-marking elements examined here. Glide insertion at hiatus and the broader juncture phonology are treated in Chapter 29 (Glide Epenthesis and Vowel-Hiatus Resolution).

20.2.4 The Shizunai minimal pair

The accentless Shizunai dialect (静内; classified as 無アクセント by Nakagawa (2024: 50); see Chapter 160 (Hokkaido Dialect Microvariation: Phonology and Morphosyntax)) presents an environment in which [ʔ] and [w] carry lexical contrast. Okuda (2025), drawing on fieldwork by 志賀雪湖, reports a minimal pair for juncture types that in accent varieties are kept distinct by pitch placement rather than by the laryngeal itself Okuda (2025):

Because Shizunai lacks distinctive pitch accent, the contrast [suwat] : [suʔat] cannot be attributed to accent-placement differences as it would be in an accent dialect. Okuda (2025) concludes that in Shizunai both the glide /w/ and the [ʔ] at transparent compound boundaries are phonemic, and adds the explicit caution that this finding does not generalize: "one cannot say, without specifying dialect, that the transitional glide y/w is non-phonemic" Okuda (2025). Parallel contrasts are reported for /uʔa/ : /uwa/ and /iʔo/ : /iyo/ in Shizunai Okuda (2025) ‹contested›. The glide phonemicity question across Hokkaido dialects is examined in Chapter 22 (Glides, Vowel Hiatus, and the Diphthong Question).

20.4.1 Onset distribution and allophony

The phoneme /h/ occurs only in syllable-onset position in Hokkaido Ainu, categorically excluded from the coda Nakagawa (2024: 27); Satō (2008: §2.3). Three allophones are conditioned by the following vowel Nakagawa (2024: 28); Shiraishi (2022: §3.1):

The conditioning is audible but less categorical than in standard Japanese. Tamura (1998: 41, cited via Shiraishi (2022: §3.1)) reports that speakers born around 1900 showed only weak vowel-coloring of /h/, with the [ɸ] before /u/ approaching a glottal quality in particular tokens. All three realizations are predictable allophones of a single phoneme /h/; no minimal pair distinguishes them for the same following vowel (see Chapter 16 (The Consonant Inventory and Its Phonetic Realization)).

20.4.2 Coda /h/ and the final-h question

Coda /h/ is absent as a synchronic segment in Hokkaido Ainu: no native Hokkaido word carries a laryngeal in coda position Nakagawa (2024: 27); Satō (2008: §2.3). In Sakhalin Ainu, by contrast, a coda breath-release /h/ (realized as a voiceless breath, not [x]; Hattori 1964: 34, cited via Nakagawa (2024: 27)) is productive and reflects historical HA coda /r/ among other sources (see Chapter 157 (The Final-h Problem and Its Sakhalin Vowel-Length Reflex)).

Within Hokkaido Ainu, the possessed form siki-h-i 'his/her eyes' (from siki ‘eye’) contains an epenthetic /h/ inserted before the possessive suffix. This /h/ is a morphological insertion restricted to the affiliative formation and does not represent an underlying coda in the root (Kindaichi & Chiri 1936: 14, cited via Shiraishi (2022: n.8)). Whether certain citation forms of stems carry an underlying final /h/, and what diachronic relationship that /h/ bears to the Sakhalin coda and to Sakhalin vowel length, is the 'final-h question' treated in Chapter 157 (The Final-h Problem and Its Sakhalin Vowel-Length Reflex).