Knowledge base · Part V — Morphophonology and Sandhi · this topic as a network
28. Assimilation and Cluster Simplification: Coda /r/, Nasals, and Clusters
How does the language handle coda consonants and consonant clusters across morpheme and word boundaries?
In the Shizunai dialect, the sequence -r r- frequently assimilates to [-nr-] (Refsing 1986: 73). In addition, the sequence e-an ya becomes e-an a (Satō 2008: 385).
generated 2 claims ·
Overview
What two regressive assimilation systems operate across junctures in the language?
The sequence /t/ + /i/ undergoes regressive assimilation across a morpheme boundary to become /ci/, as in mat-ikor becoming /macikor/. (reported in Bugaeva 2012: 468–469) In Ainu phonemic alternation, regressive assimilation fundamentally predominates. (Nakagawa 2024: 36) Under resyllabification, /t_i/ undergoes palatalization and surfaces as ci [ʨi]. (Shiraishi 2022: 13.4) Both n-sandhi and r-sandhi apply word-internally as well as word-externally across morphosyntactic constituents. (Shiraishi 2022: 13.6) Sandhi is blocked when a pause intervenes between the two constituents. (reported in Shiraishi 2022: 13.6)
Coda /r/-sandhi
What happens to a coda /r/ when it precedes a consonant-initial morpheme?
In the Shizunai dialect, the morpheme sequence /-r n-/ frequently assimilates to /-nn-/. (Refsing 1986: 70) In the Shizunai dialect, the morpheme sequence /-r t-/ almost always assimilates to /-tt-/. (Refsing 1986: 70) A syllable-final /r/ regressively assimilates to a following /t/, producing the geminate sequence tt. (Nakagawa 2024: 36, Satō 2008: 12) A syllable-final /r/ regressively assimilates to a following /c/, producing the sequence tc. (Nakagawa 2024: 36, Satō 2008: 13) A syllable-final /r/ regressively assimilates to a following /n/, producing the geminate sequence nn. (Nakagawa 2024: 36, Satō 2008: 13) In Sakhalin Ainu, the flap /r/ before a consonant also debuccalized to /h/ via an intermediate stage of total assimilation ([-r-C-] > [-C-C-] > [-hC-]). (Alonso de la Fuente 2022: 5.1) Coda /r/ regressively assimilates to a following /t/, /c/, or /n/, yielding geminate /tt/, /tc/, or /nn/. (reported in Bugaeva 2012: 468–469) In Northern dialects, the compound ar-tek 'one hand' assimilates to antek. (Kindaichi & Chiri 1936) A sequence of /t/ + /i/ alternates regressively to ci across morpheme or word boundaries. (Nakagawa 2024: 36) A sequence of /r/ + /s/ alternates regressively to ss. (Nakagawa 2024: 36) In Chitose Ainu, a sequence of /r/ + /y/ alternates regressively to yy. (Nakagawa 2024: 36) The sequence /-r c-/ frequently assimilates to [-tc-]. (Refsing 1986: 73) Across form boundaries, final -r assimilates before y- as -yy-. (Satō 2008: 13) The coda r of the positional noun or assimilates to t before the postposition ta. (Satō 2008: 163) The locational noun ot 'place' is a phonologically conditioned allomorph of or before t. (Satō 2008: 184) Coda /r/ assimilates to [t] before /t/, as in rir + tursere becoming rittursere. (Satō 2008: 221) The coda r of the noun or assimilates to t before t. (Satō 2008: 289) The final -r of utar assimilates to t before t. (Satō 2008: 304) Syllable-final r turns into n before n and into t before t (e.g. akor nispa → akon nispa, akor tures → akot tures). (Shibatani 1990) A syllable-final /r/ undergoes sandhi when followed by /n/, /c/, /t/, or /r/. (Shiraishi 2022: 13.6) In northern dialects, r-sandhi triggered by /s/ occurs (e.g. tu[s]se < tur 'to fall'). (reported in Shiraishi 2022: 13.6)
Dissimilation: /r/ before /r/
How does the language treat the sequence of coda /r/ followed by onset /r/?
A sequence of two consecutive /r/ sounds across a boundary dissimilates to nr. (Nakagawa 2024: 37, Satō 2008: 13, Shibatani 1990) The sequence /r+r/ undergoes regressive dissimilation to become /nr/, as in ku=kor rusuy becoming /ku=kon rusuy/. (reported in Bugaeva 2012: 468–469) A stem-final -rV sequence dissimilates to -nr- before r- or to -nn- before n-. (Murasaki 2025: 88) The verb kar 'to pick, gather' is normally pronounced kan before words beginning with r, such as the cohortative particle ro. (Satō 2008: 108) The consonant r can alternate to n before n. (Satō 2008: 127) When a word ending in r is followed by a word beginning with r, the first r alternates to n (e.g. kar ruwe becomes kan ruwe). (Satō 2008: 210)
Derivation walkthroughs
How are underlying representations of /r/-sandhi mapped to surface phonetic forms?
Except for t+i → ci, sandhi alternations do not always apply obligatorily, varying by dialect, speaker, and even within the same speaker's utterances. (Nakagawa 2024: 38) Older speakers tend to exhibit a higher frequency of sandhi alternations, and the tendency can carry over into their Japanese speech. (Nakagawa 2024: 38) The interrogative particle ya fuses with a preceding stem-final consonant r, as in amkir + ya > amkir_a and nankor + ya > nankora. (Shiraishi 2022: 13.7)
Coda-nasal place assimilation
How does a coda nasal /n/ assimilate in place to a following labial or velar consonant?
In Hokkaido Ainu, final /-n/ frequently assimilates in place of articulation, becoming [m] before labials and [ŋ] before velars. (Refsing 1986: 68) Voicing of stops occurs through progressive assimilation when k, t, or p follows the nasals m or n. (Kindaichi & Chiri 1936) In Raichishka, /m/ is distinctly pronounced as [m] sentence-finally without becoming [n], but assimilates to [n] when followed by k, t, or n. (Murasaki 2025: 2) Coda -m assimilates in place of articulation to a following s- (becoming -ns-) or k- (becoming -nk-). (Murasaki 2025: 89) Before the bilabial consonants p and m, both n and m are phonetically realized as [m]. (Nakagawa 2024: 58) The alveolar nasal [n] optionally velarizes to become [ŋ] before [k]. (Shibatani 1990) Final n assimilates to following bilabial sounds (m and p) and becomes m (e.g. pon-pe → pompe, pon menoko → pom menoko). (Shibatani 1990)
Coda /n/-sandhi
What happens to a coda /n/ before the continuant consonants /s/, /y/, and /w/?
In the Shizunai dialect, the morpheme sequence /-n w-/ almost always assimilates to /-nm-/ or /-mm-/. (Refsing 1986: 70) In the Shizunai dialect, the morpheme sequence /-n s-/ almost always assimilates to /-ys-/. (Refsing 1986: 70) In the Shizunai dialect, the morpheme sequence /-n r-/ sometimes assimilates to /-yr-/ or /-rr-/. (Refsing 1986: 70) In the Shizunai dialect, the morpheme sequence /-n y-/ frequently assimilates to /-yy-/. (Refsing 1986: 70) In Sakhalin Ainu, yayne can assimilate to nayne when preceded by coda -n. (Murasaki 2025: 31, 96) A syllable-final /n/ regressively assimilates to a following /y/, producing the geminate sequence yy. (Nakagawa 2024: 36, Satō 2008: 13) A syllable-final /n/ regressively assimilates to a following /s/, producing the sequence ys. (Nakagawa 2024: 36, Satō 2008: 13) The conjunctive particle wa alternates to ma when preceded by coda n. (Satō 2008: 34) Coda /n/ assimilates to a following /s/ by becoming /y/, as in pon sisam becoming /poy sisam/. (reported in Bugaeva 2012: 468–469) A vowel can be epenthesized between final -n and initial w- (-nuw-) or 'ike (-ni ike). (Murasaki 2025: 88) The progressive assimilations m+w → mm, n+w → nm, and p+w → pp are lexically restricted to words ending in m, n, or p followed by the particle wa (conjunctive, case, or final particle). (Nakagawa 2024: 37) Across form boundaries, final -n before w- can alternate as -nm- [mm] or as -ww-. (Satō 2008: 13) The particle wa becomes ma after n, so ahun wa becomes ahun ma. (Satō 2008: 191) The sequence k-an wa becomes k-an ma. (Satō 2008: 385) Final n becomes y before a following y or s (e.g. pon yuk → poy yuk, pon seta → poy seta). (Shibatani 1990) A syllable-final /n/ undergoes sandhi when followed by /s/, /y/, or /w/. (Shiraishi 2022: 13.6) n-sandhi can create sequences such as [uw] that are otherwise illicit in the language. (Shiraishi 2022: 13.6) The form 'éy-sere is phonologically composed of the first-person singular objective prefix 'en- plus sére. (Tamura 1972: 21)
Domain asymmetry: /r/-sandhi and /n/-sandhi differ in boundary sensitivity
How do the domains of /r/-sandhi and /n/-sandhi differ in boundary sensitivity?
In recordings of the Saru dialect, n-sandhi never applies across a sentence boundary. (Shiraishi 2022: 13.6) In the Saru dialect, r-sandhi can apply across a sentence boundary. (Shiraishi 2022: 13.6) n-sandhi is a P-structure rule sensitive to syntactic boundaries, whereas r-sandhi is a Pure phonological rule that is not sensitive to syntactic boundaries. (Shiraishi 2022: 13.6)
Lexicalized sandhi outputs
Which lexical items in the language preserve sandhi outputs as their current form?
The alternation m <-> n occurs between word variants. (Kindaichi & Chiri 1936) The alternation n <-> r occurs between word variants. (Kindaichi & Chiri 1936) The progressive alternation t+s → tc is limited to specific historical changes such as pet + sam → petcam and sinot + sa → sinotca. (Nakagawa 2024: 37) When the locative particle ta is combined with the base form of the locative noun or, it is pronounced otta via phonemic alternation and can be written as a single word. (Nakagawa 2024: 297) The combination of nankor followed by the sentence-final particle na is frequently used and often pronounced as nankonna. (Nakagawa 2024: 310) The conjunction form pa is an allomorph of wa 'and, -ing' whose initial w became p after a preceding p. (Satō 2008: 175) The verb kor exceptionally alternates to koy before the noun yupi 'elder brother', an alternation that does not occur before other words beginning with y. (Satō 2008: 229) The conjunctive particle wa surfaces as pa after the coda p of yap. (Satō 2008: 289) kan is a reduced or assimilated variant of kar 'to make'. (Satō 2008: 369) kon is a reduced or assimilated variant of kor 'to have'. (Satō 2008: 369)
Bihoro and eastern-dialect cluster simplification
What cluster simplifications occur in the Bihoro dialect beyond productive sandhi rules?
The form komo in Shiranuka is considered to be a contracted form of ka omo. (proposed in Nakagawa 2024: 329) Regressive consonant assimilation (such as tk > kk, pt > tt, rs > ss) is characteristic of eastern Hokkaido dialects, especially Bihoro. (Nakagawa & Fukazawa 2022: 8.3) The form rokan is derived from rok yan. (Satō 2008: 386)
Rule inventory and domain summary
What are the productive rules for assimilation and cluster simplification in the language?
The sequence /m+w/ undergoes progressive assimilation to become /mm/, as in isam wa becoming /isam ma/. (reported in Bugaeva 2012: 468–469) The sequence /n+w/ undergoes mutual assimilation to become /mm/, as in an wa becoming /am ma/. (reported in Bugaeva 2012: 468–469) Across a morpheme boundary, nasal or liquid consonants assimilate to following glides, nasals, or /n/ according to consonant assimilation rules. (Murasaki 2025: 88)