Chapter 16The Consonant Inventory and Its Phonetic Realization

16.1 The phoneme inventory

Hokkaido Ainu has eleven consonant phonemes: /p t k c s h m n r y w/ Nakagawa (2024: 27); Satō (2008: §2.2); Shiraishi (2022: §2). The inventory is organized by a straightforward place-and-manner grid. A twelfth segment, the glottal stop [ʔ], is posited as a phoneme in some analyses and treated as a syllable-boundary phonetic element in others; that debate and its implications for the syllable canon are addressed in Chapter 21 (Syllable Structure, Phonotactics, and Word-Edge Constraints).

Two consonants are excluded from syllable-final position: /c/ and /h/ never occur as codas in Hokkaido Ainu Nakagawa (2024: 27); Satō (2008: §2.3); Shiraishi (2022: §3). All other consonants appear in both onset and coda, subject to the phonotactic constraints set out in Chapter 21 (Syllable Structure, Phonotactics, and Word-Edge Constraints). The vowel inventory that these consonants pair with is described in Chapter 17 (The Vowel Inventory and Vowel Realization).

16.2 Onset realizations

In syllable-initial position the consonants realize as follows Nakagawa (2024: 28); Shiraishi (2022: §3).

The affricate /c/ has a secondary realization [ts] before /u/. Native Ainu writers consistently avoided the palatalized kana (チャ・チュ・チェ) for the syllables ca cu ce; Kirikae attributes this to the absence of a genuine palatalization contrast at the /c/ position Kirikae (1997: 102). Recordings from Chitose illustrate the conditioned [ts] allophone: the form eci=kay 'you (pl.) do' surfaces as [et͡sɨ̥kaj] with concurrent vowel devoicing aynu-corpora Discord (2023–2026) (nukopoli, 2024-12-20; corroborated by Shiraishi (2022: §3)).

The fricative /s/ is laminal, with its precise place of articulation conditioned by the following segment. Before /i/ — whether as onset or coda — /s/ palatalizes to [ɕ] Nakagawa (2024: 28); Shiraishi (2022: §3). Community recordings suggest the palato-alveolar realization before /i/ is phonetically distinct from Japanese [ɕ], with less hissing and a blade rather than lamino-palatal contact aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11) ‹corpus-suggested›; this granularity is not addressed in the academic literature consulted.

For /h/, colouring by the following vowel is well established Nakagawa (2024: 28); Shiraishi (2022: §3), but the degree is weaker in Ainu than in Japanese. Speakers born around 1900 showed particularly reduced colouring, with /h/ before /u/ approaching a pure glottal articulation rather than the bilabial [ɸ] (Tamura 1998, reported via Shiraishi (2022: §3)). The suggestion that [ɸ]~/u/ reflects Sprachbund contact with Japanese phonology has been raised in community discussion but not established in published instrumental work aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11) ‹speculative›.

16.3 Coda realizations

The stops /p t k/ are unreleased in coda position — [p̚ t̚ k̚] — a property resembling Korean final stops and distinct from English released codas Nakagawa (2024: 29); Shiraishi (2022: §3). Community recordings note that some older archival speakers produced fully released coda stops, with the implosive realization more consistent for /p/ than for /k/ aynu-corpora Discord (2023–2026) (aomidori, 2023-12-05) ‹corpus-suggested›.

Coda /s/ shows allophony conditioned by the preceding vowel. After /i/, coda /s/ surfaces as [ɕ]; after other vowels it is [sʲ]~[s] Nakagawa (2024: 29–30); Shiraishi (2022: §3). The contrast appears in cis 'cry' (→ [tɕiɕ]) vs as 'stand' (→ [asʲ]~[as]). Before the particle wa, coda /s/ retains plain [s] regardless of the preceding vowel: as wa 'standing, then' → [as wa], with no palatalization aynu-corpora Discord (2023–2026) (aomidori, 2023-12-11; attributed to Tamura 1996 Saru dictionary front matter, not directly verified here ‹speculative›).

Coda /m/ is invariant [m]. Coda /n/ undergoes place assimilation addressed below. Coda /r/ carries a vocalic release treated in § The rhotic /r/. The glides /y/ and /w/ appear as coda off-glides [i̯] and [u̯], which contributes to the argument for their consonant status (see § Glides).

16.4 Nasals and nasal assimilation

Hokkaido Ainu has two nasal phonemes: /m/ (bilabial) and /n/ (dental). In onset position both are phonetically invariant. As codas, /m/ remains [m], while /n/ undergoes regressive place assimilation Nakagawa (2024: 30); Shiraishi (2022: §3):

• /n/ → [ŋ] / __ /k/ (dorsal assimilation): honkor 'be pregnant' → [hoŋkor]; kimunkamuy 'mountain deity' → [ki.muŋ.ka.muy] aynu-corpora Discord (2023–2026) (nukopoli, 2024-11-16; corroborated by Nakagawa (2024: 30)).
• /n/ → [m] / __ /p m/ (labial assimilation): anpe 'true thing' → [ampe] Nakagawa (2024: 30).

These rules apply to /n/ only. The bilabial nasal /m/ does not shift to [ŋ] before /k/: a form such as okamkir, containing underlying /m/ before /k/, retains [m] aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11) ‹contested›. The orthographic implication is that /m/ and /n/ remain distinctly spelled before dorsals — a practice consistent across the standard Romanization — though the systematic phonetic documentation of this non-assimilation across dialects and lexical items is not available in the sources read for this grammar.

The sandhi extension of these nasal assimilation rules across morpheme and word boundaries — including /n/ + /s/ → [y] + /s/ (e.g. pon seta → [poy seta] 'puppy') — is treated in Chapter 28 (Assimilation and Cluster Simplification: Coda /r/, Nasals, and Clusters).

16.5 Voicing

Hokkaido Ainu has no phonemic voicing contrast among obstruents. The pairs /p/~[b], /t/~[d], /k/~[g], /s/~[z] are non-contrastive, and the phonemic transcription uses the voiceless symbol throughout Nakagawa (2024: 25, 27); Satō (2008: §2.2); Shiraishi (2022: §3.1). The form hapo 'mother' and its allophonic variant habo are the same phonemic form; similarly, pet 'river' surfaces as [pet] or [bet] depending on context and speaker.

Voiced allophones arise in three phonetic environments Nakagawa (2024: 25, 27); Shiraishi (2022: §3.1): (i) intervocalic position (hapo ~ [habo]); (ii) post-nasal position (tampe ~ [tambe]); (iii) post-/r/ position (unarpe ~ [unarbe]). Of these, post-nasal voicing is the most frequent: Yamashita (2000) measured approximately 40% application in that environment (reported via Shiraishi (2022: §3.1)). Voicing likelihood also varies by place of articulation: labial /p/ voices most readily, velar /k/ least. Asai (1969) observed that ape 'fire' frequently voices to [abe] while cikap 'bird' seldom voices to [tɕigap] (reported via Nakagawa (2024: 28)). Voicing additionally correlates with speaker sex; male speakers voice more frequently than female speakers (Kindaichi 1928, reported via Nakagawa (2024: 27)).

Voice onset time measurements from a Chitose recording session give approximate values of +8 to +15 ms for /p/ and +20 ms for /k/ in onset position; an aspirated [kʰ] variant reached +30 to +43 ms aynu-corpora Discord (2023–2026) (nukopoli, 2024-12-20) ‹corpus-suggested›. These figures are consistent with an absence of contrastive aspiration and substantial speaker latitude; a published acoustic VOT study of Hokkaido Ainu has not been consulted for this grammar.

16.6 The rhotic /r/

The phoneme /r/ realizes primarily as the alveolar flap [ɾ]. In some speakers trill [r] and [d]-like articulations occur as well; Sakhalin varieties have a retroflex [ʈ] that may be hard to distinguish from [t] (reported via Shiraishi (2022: §3)). Archival recordings reportedly show /r/ with an approximant quality, and /s/ palatalizing more broadly than before /i/ alone aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-05). Community discussion, citing the front-matter phonology of Tamura's Saru dictionary, stresses that Ainu /r/ 'does not resemble [l] at all' (aomidori, aynu-corpora Discord (2023–2026), 2023-12-11); Shiraishi, by contrast, lists [l] alongside [d] as possible realizations Shiraishi (2022: §3). The community characterization could not be page-verified in sources read for this grammar and is reported here as a community assertion ‹speculative›.

In coda position /r/ carries a vocalic release — described as 'resounding' by Kindaichi (1931, reported via Shiraishi (2022: §3.2)), 'parasitic vowel' by Peng (1970, reported via Shiraishi (2022: §3.2)), and 'echo vowel' by Tamura (1988b, reported via Shiraishi (2022: §3.2)) — whose quality roughly copies the preceding vowel: pirka 'good' → [pirĭka], arki 'come (pl.)' → [arĭki] Nakagawa (2024: 30); Shiraishi (2022: §3.2). This release distinguishes /-r/ from /-rV/ sequences. The minimal pair etor [etoro] 'snot' vs etoro [etoːro] 'to snore' — where the /-rV/ word has a measurably longer pre-rhotic vowel (Asai 1976, reported via Shiraishi (2022: §3.2)) — is now well established Nakagawa (2024: 30); Shiraishi (2022: §3.2).

After voiceless stops /k p t/, coda /r/ devoices: nukar 'look at' may surface as [nukaɾ̥ɨ̥h] aynu-corpora Discord (2023–2026) (aomidori, 2023-12-11) ‹corpus-suggested›. The sandhi diagnostics for /-r/ vs /-rV/ — resyllabification behavior, nominalizer allomorphy, and the cross-word assimilation rules — are treated in Chapter 28 (Assimilation and Cluster Simplification: Coda /r/, Nasals, and Clusters).

16.7 Glides /y/ and /w/ as consonants

The glides /y/ [j] and /w/ [w] function as phonemic consonants in both onset and coda Nakagawa (2024: 29); Shiraishi (2022: §3). An earlier analysis by Kindaichi and Chiri Kindaichi & Chiri (1936) treated these as the second elements of diphthongs; the consonant analysis was established by Nakagawa and Shiraishi on the basis of two independent diagnostics.

The first diagnostic is nominalizer allomorphy. The nominalizer suffix selects -p after a vowel and -pe after a consonant; morphemes ending in /y/ or /w/ take the consonant allomorph — okay-pe, siw-pe — establishing that /-y/ and /-w/ pattern with coda consonants Shiraishi (2022: §3); Nakagawa (2024: 29). The second diagnostic is accent: syllables closed by /y/ or /w/ do not bear independent accent, unlike syllables closed by /i/ or /u/. Thus kamuy 'deity' parses as ka.muy (two syllables, accent on the second), and aynu 'person' as ay.nu Nakagawa (2024: 29). Community members corroborate this analysis through parallel allomorphy: the plural suffix (ante → okayte) and causative allomorphs (-re/-te) both treat /-y/ as a consonant coda aynu-corpora Discord (2023–2026) (nukopoli, 2025-12-03) ‹consensus›.

In onset position, /y/ and /w/ appear before vowels in native morphemes, but the sequences *yi, *wi, *wu are absent from monomorphemic roots, arising only across morpheme boundaries Nakagawa (2024: 43–45). These phonotactic gaps are part of the syllable inventory treated in Chapter 21 (Syllable Structure, Phonotactics, and Word-Edge Constraints).

The phonemic status of onset /y/ and /w/ is dialect-dependent and remains unsettled ‹contested›. For Saru and Chitose, Nakagawa treats these as surface glides derived at phoneme junctures, not underlying phonemes Nakagawa (2024: 34–35). In Shizunai, the absence of contrastive pitch means that the distinction between [w] and [ʔ] at a hiatus position — illustrated by the minimal pair suwat 'pothook' vs suʔat 'pot bail' — must be carried by the glide itself, making /w/ phonemically contrastive there Okuda (2025). Satō (2003) further posits underlying /y w/ in Chitose for non-predictable hiatus positions while treating glide appearances after ku- and e- as surface-derived (reported via Shiraishi (2022: §4.3)). Okuda concludes that no dialect-neutral statement about glide phonemicity is possible Okuda (2025).

The accent behavior of syllables closed by /y/ or /w/ is treated in Chapter 23 (The Pitch-Accent Placement Rule and the Accented/Accentless Dialect Split). Glide insertion, hiatus resolution, and person-affix sandhi involving glides are in Chapter 31 (Personal-Affix Junctural Sandhi, =an/a= Allomorphy, and Connected-Speech Reduction). For dialect-level consonant variation — including Bihoro cluster assimilation (tk→kk, rs→ss) and h-elision and insertion in eastern varieties — see Chapter 160 (Hokkaido Dialect Microvariation: Phonology and Morphosyntax).