Chapter 19The Rhotic /r/ and Coda-r

19.1 The phoneme /r/ in the consonant inventory

Hokkaido Ainu has one rhotic phoneme, /r/, the sole coronal sonorant that is neither a nasal (/m/, /n/) nor a glide (/y/, /w/) in the eleven-consonant system described in Chapter 16 (The Consonant Inventory and Its Phonetic Realization) Nakagawa (2024: 27); Shiraishi (2022: §3); Satō (2008: §2.2). The phoneme occurs in both onset and coda positions within the canonical (C)V(C) syllable, and its coda realization — carrying a distinctive vocalic release and triggering a dense assimilation system — makes it the most morphophonologically active segment in the language. The full assimilation rules are given in Chapter 28 (Assimilation and Cluster Simplification: Coda /r/, Nasals, and Clusters); the present chapter covers the phoneme's phonetic realization, the echo-vowel phenomenon, and the diagnostics that separate /r/-final syllables from genuine /rV/ syllables.

19.2 Onset realization and positional variants

In syllable-onset position /r/ is realized as an alveolar flap [ɾ] Nakagawa (2024: 28); Shiraishi (2022: §3.1). Speaker variation and historical change produce a broader allophonic range; the table below collects documented realizations.

The allophonic range — flap, trill, dental-approximant — places Ainu /r/ in a different perceptual space from Japanese ラ行, whose typical word-initial and post-nasal realization leans lateral. Tamura's Saru dictionary grammar is reported by community members to emphasize this perceptual distance explicitly aynu-corpora Discord (2023–2026) (aomidori, 2023-12-11, citing Tamura 1996) ‹speculative›; the source most directly available, Shiraishi (2022: §3.1), records [l] and [d] as occasional allophones attributed to Tamura (1988b: 13) rather than as the default, which remains the flap [ɾ]. In Sakhalin Ainu the cognate of Hokkaido /r/ is described as retroflex [ʈ], acoustically close to [t] and distinct from the Hokkaido flap Shiraishi (2022: §3.1).

19.3 Coda /r/ and the echo-vowel

The coda realization of /r/ carries a vocalic release — known variously in the literature as the re-sounding, the echo-vowel, or the echoing vowel — that has attracted sustained analysis since the earliest systematic descriptions Shiraishi (2022: §3.2). In careful speech the release is audible as a short vowel copying the nucleus of the syllable: kar [kaɾ(a)], kor [koɾ(o)], sir [siɾ(i)]. In recitation or emphatic delivery the release extends to a full vowel: [piriˑka], [mokoroˑ] Shiraishi (2022: §3.2).

Batchelor (1905) and the early Kindaichi transcriptions wrote post-/r/ vowels as full phonemic segments — hence spellings such as koro, pirika, Ainurakkuru — without distinguishing a genuine /rV/ syllable from the echo-vowel release of a /r/-final syllable Batchelor (1905); Shiraishi (2022: §3.2). Yukie Chiri (知里幸恵), in correspondence with Kindaichi around 1923, established that /r/-final words form a coherent phonological category and that the post-/r/ releases in Batchelor's transcriptions are phonologically spurious Shiraishi (2022: §3.2). She identified three diagnostics that separate /r/-final from /rV/-final words Nakagawa (2024: 30); Shiraishi (2022: §3.2):

1. The echo-vowel weakens before a vowel-initial suffix; a genuine final vowel does not. In mokor=an 'sleep.4S' the echo is reduced before the suffix, whereas the final /o/ of poro=an 'be.big.4S' remains unreduced.
2. Coda /r/ undergoes morphophonological sandhi; /rV/ does not. The /r/-final verb a=kor 'have' assimilates its coda before the onset of a following word (a=kor nispa → [akonˈnispa] 'a wealthy man'), whereas a word ending in /rV/ such as poro 'big' shows no corresponding change. The sandhi system is treated in Chapter 28 (Assimilation and Cluster Simplification: Coda /r/, Nasals, and Clusters).
3. The nominalizer selects its consonant-allomorph after coda /r/. The nominalizer -p/-pe ‘nominalizer’ surfaces as -pe after a consonant and -p after a vowel. A /r/-final stem takes -pe:

The minimal pair etor [etoro] 'snot' versus etoro [etoːro] 'to snore' is now well established in the literature: in the /rV/-final word the vowel preceding /r/ is measurably longer, whereas the echo-vowel in the /r/-final word is audibly shorter Nakagawa (2024: 30); Shiraishi (2022: §3.2). Asai (1976) confirmed the length difference acoustically Shiraishi (2022: §3.2). An additional independent diagnostic, noted by Satō (2000), is that in pre-pausal position the final vowel is repeated only for /r/-final words Shiraishi (2022: §3.2).

The direction of vowel-copying is predominantly anticipatory: pirka surfaces as [pirĭka] and arki as [ariki], with the echo anticipating the following vowel rather than perseverating the preceding one Nakagawa (2024: 30). A perseveratory case, citarpe [citarube] (Asai 1976), shows that the copying direction is not categorically fixed Shiraishi (2022: §3.2). Northern dialect speakers vacillate between -r and -rV in certain lexical items (kikir ~ kikiri), suggesting ongoing redistribution between the two categories in those varieties Shiraishi (2022: §3.2).

One articulatory account of the echo-vowel, proposed within the community, treats it as a timing offset: the tongue-tip flap gesture begins before the mouth has finished the preceding vowel, so the flap releases into the vowel-gesture and produces an audible copy of the nucleus aynu-corpora Discord (2023–2026) (nukopoli, 2023-12-11) ‹speculative›. Published treatments describe the release as a phonological property of coda /r/ without committing to a specific articulatory model Shiraishi (2022: §3.2).

In casual connected speech, word-final /r/ is susceptible to deletion, producing reduced forms such as ku for kur 'person' and ko for kor 'have', while the verb koro tends to stabilize its full form in many contexts aynu-corpora Discord (2023–2026) (nukopoli, 2024-12-11) ‹corpus-suggested›. This casual reduction is distinct from the diachronic Sakhalin coda-/r/ → /h/ change treated in Chapter 157 (The Final-h Problem and Its Sakhalin Vowel-Length Reflex).

19.4 Devoicing of coda /r/ after voiceless-stop onsets

When the onset of a CVC syllable is a voiceless stop /k/, /p/, or /t/, the coda /r/ is devoiced to [ɾ̥]. The verb nukar 'look at' illustrates the pattern: the final syllable kar has onset /k/ and coda /r/, and the rhotic surfaces as [ɾ̥], giving [nukaɾ̥] aynu-corpora Discord (2023–2026) (aomidori, 2023-12-11) ‹corpus-suggested›. The pattern is phonetically parallel to the partial devoicing of sonorant codas reported for other languages after voiceless-onset syllables; Shiraishi's description of /r/ as a flap with voicing-conditioned variants Shiraishi (2022: §3.1) is consistent with the conditioning, though no instrumental study targeting this specific alternation has been found in the sources read for this grammar.

The devoiced coda in turn affects the echo-vowel: when [ɾ̥] releases, the following vocalic copy is likewise voiceless, yielding a realization such as [nukaɾ̥ɨ̥h] in citation-form speech aynu-corpora Discord (2023–2026) (aomidori, 2023-12-11). Vowel devoicing in comparable environments — high vowels between voiceless consonants — is independently attested in Hokkaido Ainu and is described in Chapter 17 (The Vowel Inventory and Vowel Realization); the rhotic devoicing here extends the same voicelessness-spreading dynamic to the coda rhotic of CVC syllables with voiceless onsets.

19.5 Morphophonological activity of coda /r/

Coda /r/ is the dominant trigger of the regressive assimilation rules described in Chapter 28 (Assimilation and Cluster Simplification: Coda /r/, Nasals, and Clusters). When a /r/-final morpheme or word precedes a consonant-initial morpheme or word, the /r/ assimilates to, or is absorbed into a geminate with, the following consonant. The principal alternations, drawn from Nakagawa (2024: 35–37) and Shiraishi (2022: §6.1), are: /r/ + /t/ → [tt] (a=kor tar → [aˈkottar]); /r/ + /n/ → [nn] (a=kor nispa → [akonˈnispa]); /r/ + /c/ → [tc] (retar cikap → [retatˈcikap] 'swan'); /r/ + /s/ → [ss] (ar suy → [asˈsuy] 'once'); and, uniquely, a dissimilatory /r/ + /r/ → [nn] (ku=kor rusuy → [kukonˈrusuy] 'I want it') Nakagawa (2024: 35–37); Shiraishi (2022: §6.1).

Unlike nasal-place assimilation (which is bounded by the phrase in Shiraishi's 2001 analysis), /r/-sandhi operates across phrasal and sentence boundaries as a postlexical rule Shiraishi (2022: §6.2); aynu-corpora Discord (2023–2026) (antitwilight, 2024-11-23) ‹corpus-confirmed›. The rules are not written in standard orthography, which notates underlying forms; they are fully productive in speech Nakagawa (2024: 37–38). Personal-affix junctures that feed the same /r/-sandhi rules are treated in Chapter 31 (Personal-Affix Junctural Sandhi, =an/a= Allomorphy, and Connected-Speech Reduction), and the echo-vowel's interaction with possessive-form derivation is treated in Chapter 45 (Morphophonology of the Affiliative Suffix and Echo/Copy Vowels).

19.6 Dialectal variation and the Sakhalin coda-/r/ reflex

Across Hokkaido varieties, coda /r/ shows at least one documented lexically specific neutralization: in some dialects mokor 'sleep' surfaces as mokon, with /r/ replaced by [n] aynu-corpora Discord (2023–2026) (keycja, 2024-07-20) ‹speculative›. Nakagawa and Fukazawa's dialect classification documents extensive morphophonological isoglosses across the Hokkaido area Nakagawa & Fukazawa (2022: §3), though the specific /r/ → /n/ change in this lexical item has not been mapped to a named dialect zone in the sources read for this grammar. The broader picture of dialectal consonant correspondences is taken up in Chapter 160 (Hokkaido Dialect Microvariation: Phonology and Morphosyntax).

In Sakhalin Ainu, coda /r/ has merged with /h/ in syllable-final and word-final position, so that Hokkaido kor 'have' corresponds to Sakhalin koh. The change is attested in the Sakhalin corpus, and the distribution of Sakhalin -h vis-à-vis the retained /rV/ type has been the subject of dedicated distributional analysis (cited in aynu-corpora Discord (2023–2026) nukopoli, 2024-12-19). Because the Sakhalin /r/ → /h/ shift belongs primarily to the diachronic and comparative treatment, details are routed to Chapter 157 (The Final-h Problem and Its Sakhalin Vowel-Length Reflex). The Sakhalin onset cognate of Hokkaido /r/ is the retroflex [ʈ] mentioned in §2 above, distinct both from the Hokkaido flap and from the Sakhalin /h/ that the coda reflex produces.