Chapter 22Glides, Vowel Hiatus, and the Diphthong Question

22.1 The analytical question

Hokkaido Ainu words frequently end in sequences pairing a vowel with a following y [j] or w [w]: áynu 'human', kamúy 'spirit', ohaw 'soup', kéwtum 'heart', ináw 'prayer-wand'. Two analyses have been proposed for these sequences. The first treats the final element as a coda consonant — /y/ or /w/ — parallel to the coda stops, nasals, and /r/ of the inventory. The second treats the sequences as falling diphthongs, with the final element a non-syllabic vowel ([i̯] or [u̯]), giving bimoraic nuclei [ai̯], [au̯], [ei̯], [eu̯], [oi̯], [ui̯] aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11). Phonetically the two analyses are indistinguishable: both produce the same output, and neither predicts any difference in production or perception aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11). The question is whether post-vocalic y/w behave morphologically and prosodically as consonants or as the second element of a vowel nucleus.

This chapter covers the phonemic status of coda y and w and the related question of the marginal onset sequences yi, wi, and wu. The full phoneme inventory is in Chapter 16 (The Consonant Inventory and Its Phonetic Realization) and Chapter 17 (The Vowel Inventory and Vowel Realization); the (C)V(C) syllable canon in Chapter 21 (Syllable Structure, Phonotactics, and Word-Edge Constraints). The allophonic glide that appears at morpheme-boundary hiatus — the [j] in i-y-ekarkar — is a separate process treated in Chapter 29 (Glide Epenthesis and Vowel-Hiatus Resolution).

22.2 The consonant analysis

The standard modern analysis treats post-vocalic y and w as coda consonants, assigning syllables like ay, muy, haw, and kew to the CVC slot of the (C)V(C) canon alongside cep, sin, and ker Nakagawa (2024: 29); Shiraishi (2022: §3). Three diagnostics converge on this position.

22.3 The diphthong analysis

Kindaichi & Chiri Kindaichi & Chiri (1936) (cited via Shiraishi (2022: §3)) analysed post-vocalic y and w as the second elements of falling diphthongs. Early romanization and katakana transcription followed this view: kamuy was written カムイ (four kana), treating the final element as a full vowel; aynu appeared as アイノ, reflecting the approximate quality of the rounded unstressed second element Nakagawa (2024: 26, 29). The orthographic reform that Chiri Yukie (知里幸恵, c. 1923) introduced — kamuy, ohaw, kewtum for カムイ, オハウ, ケウトゥム — was motivated by exactly the morphological and prosodic evidence set out in §2, not by any change in phonetic realization Nakagawa (2024: 29).

Phonetically the diphthong analysis is accurate. The sequences are realized as falling diphthongs — [ai̯], [au̯], [ei̯], [eu̯], [oi̯], [ui̯] — in which the second element is non-syllabic aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11). Nothing in the acoustic signal distinguishes a coda consonant from a non-syllabic vowel in this position: the two analyses predict identical phonetic outputs. The diphthong analysis is a coherent phonetic description; its weakness is at the morpho-phonological level, where it forces the statement that a "non-syllabic vowel" triggers the consonant-class allomorphs of both the nominalizer and the plural suffix.

The fullest structuralist defence of the diphthong view is Simeon's, who classes /w y/ as "nonsyllabic vowels" on four grounds: phonetic similarity to /u i/, a clustering pattern unlike that of the consonants, the narrow allophonic range they share with the vowels, and their undergoing the same elision as syllabic vowels Simeon (1968: 14–15). The cost of the analysis shows in his inventory of units: alongside rising and falling diphthongs he must admit a triphthong yay and a "geminate diphthong" ayay 'baby' Simeon (1968: 18) — sequences the coda analysis parses without any new unit types, as yay (CVC) and a.yay (V.CVC). His own morphophonemics also lists the elision and clustering behavior that the consonant analysis takes as diagnostic, so the same data sorted by modern criteria land on the opposite conclusion.

22.4 Assessment

The consonant analysis is adopted by Nakagawa Nakagawa (2024: 29), Shiraishi Shiraishi (2022: §3), and Satō Satō (2008: §2.2), and is the working assumption of this grammar ‹consensus›. The three diagnostics in §2 — nominalizer allomorphy, plural-suffix allomorphy, and pitch-accent placement — all turn on the same categorical distinction (consonant vs. vowel-final stem), and all place /y/ and /w/ in the consonant class without exception. The diphthong analysis requires an auxiliary rule reclassifying the non-syllabic vowels as consonant-class triggers in three separate morpho-phonological environments; no evidence favours this stipulation over the simpler consonant phonemicization ‹contested›.

The community formulation captures the symmetry: reading /y/ and /w/ as consonants makes the consonant-like morphological behaviour fall out; reading them as vowels makes the phonetic falling-diphthong character explicit; neither reading misrepresents the phonetics aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11). Post-vocalic y and w are phonological coda consonants /y/ and /w/; their surface realization as the non-syllabic elements [i̯] and [u̯] of falling diphthongs is their expected phonetic output in that position.

22.5 Onset glides

Word-initial and post-consonantal y [j] and w [w] are onset consonants without analytical controversy: yuk ‘deer’, yam ‘potato’, wakka ‘water’, wen ‘bad’. Their onset and coda realizations are phonetically identical — onset [j] = coda [i̯], onset [w] = coda [u̯] — which is consistent with a single phoneme appearing in both positions, parallel to /m n r/ in Hokkaido Ainu Nakagawa (2024: 27–29); Shiraishi (2022: §3).

22.6 Marginal sequences: yi, wi, wu

Three onset–vowel sequences are absent from Hokkaido Ainu monomorphemic roots: yi, wi, and wu. Nakagawa states that no root has any of these as its initial syllable, and that syllables of the shape yi-C, wi-C, wu-C do not exist in the language Nakagawa (2024: 43–45); the same point is made in Shiraishi Shiraishi (2022: §4.2), with attribution to Kindaichi 1931:13 and Tamura 1988b:14. A palatal glide before a high front vowel ([ji]) and a labio-velar glide before a high back rounded vowel ([wu], [wi]) are articulatorily near-redundant transitions, and similar restrictions are typologically common. The phonotactic constraint is parallel to the absence of the onset ti from Ainu roots — its place in the table is empty, not filled by a rare allophone Nakagawa (2024: 43–44).

22.7 Glide phonemicity in Shizunai: a dialect-specific contrast

In the accent dialects — Saru, Chitose, and related varieties — the accentual contrast between fossilized and transparent compounds renders the question of /w/ phonemicity at hiatus redundant: pitch, not glide presence, carries the contrast. In Shizunai (east Hidaka), an accentless dialect, this alternative is unavailable. Okuda Okuda (2025) reports a minimal pair from a Shizunai speaker (Oda Steno) that is unique in the Hokkaido documentation:

suwat [suwat] and su'at [suʔat] are phonetically distinct. In an accent dialect this contrast would be carried by pitch — the fossilized compound would have a fixed accent pattern differing from the transparent form; in Shizunai, where pitch is not distinctive, the glide vs. glottal-stop difference bears the entire lexical distinction. Parallel contrasts for /y/ at front-vowel hiatus in Shizunai — /u'a/ ~ /uwa/, /i'o/ ~ /iyo/, /i'e/ ~ /iye/ — extend the argument beyond this single pair Okuda (2025). Okuda's conclusion: "one cannot say, without specifying the dialect, that the glides y/w are non-phonemic" Okuda (2025) ‹contested›.

The Shizunai data do not overturn the Nakagawa–Shiraishi analysis of Saru and Chitose, where accent carries the contrast the glide would otherwise mark. They demonstrate that /y/ and /w/ phonemicity at hiatus is a dialect-dependent property: phonemic in accentless Shizunai, non-phonemic in the accent dialects Okuda (2025); Nakagawa & Fukazawa (2022: §3.1). For the classification of Shizunai as accentless and its implications, see Chapter 23 (The Pitch-Accent Placement Rule and the Accented/Accentless Dialect Split).

22.8 Interaction with person-affix accent

The coda-consonant status of /y/ and /w/ has one further prosodic consequence. When ku= or e= prefixes to a stem beginning with /i/ or /u/, vowel-weakening converts the vowel to a glide, producing a CVC-closed first syllable: ku= + ipe → ku=ype /kúype/ (σ1 = /kuy/, closed, Rule 1: accent on σ1) Nakagawa (2024: 52–53). The derived coda /y/ is treated by the accent system in exactly the same way as root-internal coda /y/ in áynu. The full treatment of vowel-weakening and person-prefix accent shifts is in Chapter 31 (Personal-Affix Junctural Sandhi, =an/a= Allomorphy, and Connected-Speech Reduction) and Chapter 26 (Accent in Compounds and under Affixation and Cliticization).